Introduction

Hematopoietic stem cells (HSCs) endowed with full regenerative potential in adult vertebrates are generated during a narrow-time window of few days during embryonic development. These cells, at the root of all blood and immune cells in the body, emerge from intra-embryonic aortic vessels and, more specifically, from a specialized type of vascular tissue called the hemogenic endothelium (HE, (Wu and Hirschi 2021)). A series of seminal studies have evidenced the autonomous production of repopulating HSCs within an intraembryonic region called the Aorta-Gonad-Mesonephros (AGM), before their appearance in other hematopoietic organs (A. L. Medvinsky et al. 1993; Müller et al. 1994; Garcia-Porrero, Godin, and Dieterlen-Lièvre 1995; A. Medvinsky and Dzierzak 1996; Cumano, Dieterlen-Lievre, and Godin 1996). Thereafter, the endothelial origin of HSCs was evidenced (Jaffredo et al. 1998; de Bruijn et al. 2002; T. E. North et al. 2002). The direct visualization, in real-time, of the emergence of precursors of hematopoietic stem and progenitor cells (HSPCs) from the dorsal aorta, termed the Endothelial-to-Hematopoietic Transition (EHT), was finally achieved in vitro (Eilken, Nishikawa, and Schroeder 2009), ex-vivo from mouse sections (Boisset et al. 2010), and in vivo in the zebrafish embryo (Bertrand et al. 2010; Kissa and Herbomel 2010), using live microscopy. This opened the way to a more detailed analysis of the characteristics of the HE at the transcriptional level, revealing its transient nature and its early hematopoietic commitment (Swiers et al. 2013).

The HE is characterized by inherent heterogeneity and is not only contributing to the formation of precursors to long-term HSCs (pre-HSCs) but also to more restricted progenitors (Hadland and Yoshimoto 2018). These progenitors can be born from HE sub-types that are found in extra-embryonic source such as the yolk-sac – in which case it gives rise to erythro-myeloid progenitors that will sustain erythro-myelopoiesis until birth (Frame et al. 2016) -, or both from extra and intra-embryonic sources (the yolk sac and the dorsal aorta) in which cases progenitors are biased, such as for example in the mouse, toward T- and B-innate-type lymphoid progenitors (Yoshimoto et al. 2011; 2012) or toward less restricted, multipotent progenitors (Hadland and Yoshimoto 2018; Dignum et al. 2021). Importantly, while some of the non-HSC derived progenitors born during embryonic life only support the functions of the immune system during embryonic/ fetal life, others can persist in the adult to exert tissue-resident functions, as has been mainly described in the mouse (Ghosn et al. 2019). However, studies in human embryos support the idea that developmental hematopoiesis is highly similar in mice and humans (Ivanovs et al. 2011; 2017).

A key question regarding the issue of the capacity of the HE to give birth to cells endowed with different hematopoietic potential is whether this comes from HE cells of distinct origin (in which case the HE would be a mosaic of precursors with distinct potential) or if this results from extrinsic environmental cues that impose variability in the constantly evolving developmental context (for a discussion, see (Barone et al. 2022)).

To address this type of question, high-resolution experimental settings need to be developed, if possible at the single-cell resolution. Along this line, recent developments of transcriptomics, including single cell RNAseq, spatially resolved in situ transcriptomic approaches, and in situ barcoding, have been invaluable (Weijts, Yvernogeau, and Robin 2021). Among those high-resolution approaches, cell fate mapping and lineage tracing approaches that include imaging technologies have been very instrumental, particularly when performed with the zebrafish model that reproduces many aspects of developmental hematopoiesis in higher vertebrates (Orkin and Zon 2008). Recently, work performed in the zebrafish embryo has brought evidence for heterogeneity of hematopoietic stem cell precursors being born from the HE, in the ventral floor of the dorsal aorta and independently from HSCs, including a wave of transient T-lymphocytes (using temporally-spatially resolved fate-mapping, (Tian et al. 2017)) and lympho-myeloid biased progenitors born from a myeloid-lymphoid biased Spi2+ HE that appears to co-exist with an erythroid-biased HE (using single-cell RNA-sequencing, (Xia et al. 2023)).

While light starts to be shed on the molecular and signaling cues that appear to regulate HE sub-specification and a continuum from arterial endothelium, HE maturation, and subsequent heterogeneity in HSPCs (Zhu et al. 2020), the essential molecular and cell biological properties that support HE functional plasticity remain to be determined.

Here, using the zebrafish embryo as a model, we complement our previous work describing essential molecular and mechanistic features of EHT cell emergence (Lancino et al. 2018). With the support of high-resolution live imaging and image analysis, the generation of new transgenic fish lines that express functional markers of cell polarity (podocalyxin) and of inter-cellular adhesion (Junctional Adhesion Molecules - JAMs), as well as functional and genetic interference, we bring insight into additional key features of the EHT process that take their root in the differentiation and maturation of the HE.

Results

Apicobasal polarity determines emergence types

Our previous work describing the morphodynamic evolution of cells emerging from the aortic floor through the EHT (hereafter designated as “EHT cells”), in the zebrafish embryo, revealed the unusual feature of a cell extruding from the plane of a tissue while maintaining its luminal/apical membrane until the very end of the release process, thus contributing to its peculiar crescent-shaped morphology (Kissa and Herbomel 2010; Lancino et al. 2018, and see the cartoons in Figure 1A). However, to our knowledge, the polarity status of EHT cells has not been investigated so far and the maintenance of a bona fide apical domain has never been proven (with the luminal membrane enriched in apically targeted proteins and physically delimited by apical polarity factors and tight junction complexes (Rodriguez-Boulan, Müsch, and Le Bivic 2004; Buckley and St Johnston 2022)). Importantly, the fate of this apical-luminal membrane, after the release, may lead to cells potentially endowed with specific functional features. For example, this membrane surface may be directly and fully exposed to the extracellular space or released in the cytoplasm of EHT cells for recycling and/or degradation, after emergence completion (for examples of different scenarios, see Figure 1A). Overall, this could lead to precursors of hematopoietic stem cells that, ultimately, may be differentially fated. To address the polarity status of EHT cells, we raised transgenic fish lines that express endogenous Podocalyxin (Podxl2, (Herwig et al. 2011)), a sialomucin of the CD34 family (Nielsen and McNagny 2008). Podocalyxin was shown to take part in the formation of the preapical domain during polarization and in the regulation of its oriented organization, in tissue culture (Meder et al. 2005; Bryant et al. 2014). Its contribution to lumenization in vivo, in the mouse aorta, has been described and it involves negative charge repulsion induced by their glycosyl residues (Strilić et al. 2009; 2010).

Figure 1:

We first attempted to express transiently, in the vascular system, the full-length Podxl2 zebrafish protein fused to eGFP at its extreme N-terminus. We failed to detect the fusion protein at any time point relevant for observing easily the EHT process (a time window ranging from 48 to 60 hpf (hours post-fertilization)); therefore we designed a N-ter truncated form deleted of the mucin domain and that retains amino-acids 341-587 fused to eGFP at its N-terminus (Figure 1B and see Materials and Methods). Transient transgenesis revealed that this truncated version is detected and is targeted to the luminal membranes of EHT cells. We then raised 2 transgenic (Tg) fish lines that express the N-ter truncated form of Podxl2 fused to either eGFP (thereafter abbreviated eGFP-Podxl2) and under the control of the Kdrl:Gal4 driver (Tg(Kdrl:Gal4;UAS:RFP; 4xNR:eGFP-Podxl2)), or to mKate2 and under the control of the Kdrl promoter (Tg(Kdrl:mKate2-Podxl2)). We observed that eGFP-Podxl2 is enriched at the luminal side of crescent-shaped EHT undergoing cells (see Figure 1C, Figure 1 -figure supplement 1A and Figure 1 – video 1 (z-stack at t=0) and their legends for the details of the luminal/apical membrane evolution through time). Thereafter, these cells will be referred to as EHT pol+ cells.

We also followed the cell after emergence and observed the evolution of the luminal/apical membrane appearing as internal pseudo-vacuoles. We illustrate the reduction of their volume via membrane retrieval and, ultimately, their remanence as an intracellular membrane compartment which we define as a post-EHT signature (Figure 1-figure supplement 1A and Figure 1-video 2 (z-stack at t=80 min)); we propose a model for the intracellular evolution of the luminal/apical membrane which, unfortunately, cannot be traced after 2-3 hrs post-emergence because of the apparent short half-life of eGFP-Podxl2 and of the drop in activity of the Kdrl promoter (Figure 1-figure supplement 1B). Of notice, the pseudo-vacuoles are reminiscent of the cystic formations observed in EHT cells in the mouse and visualized by electron-microscopy (T. North et al. 1999; Marshall and Thrasher 2001) and also of the vacuolar structures recently described in EHT cells in avian embryos (Sato et al. 2023); in both cases, these vacuoles appear to emanate from the abluminal membrane (facing the sub-aortic space) and not from the lumen. In the latter case, they disappear before EHT completion, suggesting that they contribute to exert mechanical force to initiate the detachment of the cell from the endothelial layer. In addition, their dynamics appears to depend on the activity of aquaporins and it is very possible that aquaporins are active in zebrafish too, although rather in EHT cells late in their emergence and/or in post-EHT cells, for water chase and vacuolar regression as proposed in our model (Figure 1 – figure supplement 1B).

While imaging the EHT using the Podxl2 expressing lines that clearly delimitate the luminal membrane, we unambiguously identified a second type of emergence. This second cell type is primarily characterized by a round-to-ovoid morphology (cells never bend as crescent-shaped EHT pol+ cells, see Figure 1D and Figure 1 – video 3 for a time-lapse sequence). Importantly, these cells do not show any enrichment of eGFP-Podxl2 at the luminal membrane and will be referred to as EHT pol-cells. EHT pol-cells were observed in all other Tg fish lines that we are routinely imaging, including the Tg(Kdrl:Gal4;UAS:RFP) parental line that was used for transgenesis, thus excluding the possibility that these cells result from an artefact due to the expression of a deleted form of Podxl2 and/or to its overexpression.

Finally, we have estimated that the ratio between EHT pol+ and EHT pol-cells is of approximately 2/1, irrespective of the imaging time window and of the localization of emergence along the aortic antero-posterior axis (starting from the most anterior part of the AGM (at the limit between the balled and the elongated yolk) down to the caudal part of the aorta facing the CHT). We observed that both EHT pol+ and EHT pol-cells divide during the emergence and remain with their respective morphological characteristics after completing abscission (hence appearing as pairs of cells that exit the aortic wall sequentially, as shown Figure 1C, D). We also observed that both EHT pol+ and EHT pol-cells express reporters driven by the hematopoietic marker CD41 (data not shown), which indicates that they are both endowed with hematopoietic potential.

Altogether, our results show that hemogenic cells emerging from the aortic floor do so with heterogeneity in their morphodynamic characteristics. This suggests that this may be one of the sources underlying the complexity of pre-hematopoietic stem cell identity and of their downstream cell fate.

The immature HE is not polarized

EHT pol+ and EHT pol-cells appear to emerge from the hemogenic endothelium (HE), the latter constituting the virtually exclusive cell population of the aortic floor just prior to the initiation of the EHT time-window (around 28 hpf, see (Zhao et al. 2022)). In this context, we addressed the polarity status of HE cells. Surprisingly, confocal microscopy using the eGFP-Podxl2 expressing fish line revealed that HE cells do not appear to be polarized, based on the absence of eGFP-Podxl2 enrichment at luminal membranes, at the initiation of the EHT time-window (at approximately 28 - 30 hpf, Figure 2 – figure supplement 1). Interestingly, the cytoplasm of characteristic elongated HE cells located on the aortic floor is filled with more-or-less large membrane vesicles that carry eGFP-Podxl2 (the largest vesicles reaching approximately 30µm in diameter). This suggests that HE cells contain a reservoir of eGFP-Podxl2 membranes that may be subjected to exocytose; as such, HE cells may be comparable to endothelial cells organizing a vascular lumen and that have been proposed to exocytose large intracellular macropinocytic-like vacuoles when cultured in 3D extracellular matrices (Bayless and Davis 2002; Davis, Bayless, and Mavila 2002) or, in vivo, in the zebrafish model (Kamei et al. 2006; Lagendijk, Yap, and Hogan 2014). This finding is unexpected since HE cells are assumed to possess aortic cell characteristics (i.e exhibit an apicobasal polarity) as they are supposedly integrated in the aortic wall contemporarily to aortic precursors (Jin et al. 2005) and may have been taking part in the lumenization of the ancestral vascular cord, a process that takes place around 18-20 hpf. Consequently, loss of apicobasal polarity features of HE cells at 28-30 hpf may be part of the programme that initiates the EHT process.

Figure 2:

Although technically difficult for long hours (because of important variations in the volume of the balled yolk that trigger drifting of embryos), we have been able to follow through time non-polarized HE cells and to visualize the evolution of their vesicular content, starting at the initiation of the EHT time-window (around 35 hpf, see Figure 2 and Figure 2 – figure supplement 2). Interestingly, we could follow a dividing HE cell for which the vesicular content labelled with the polarity marker Podxl2 (eGFP-Podxl2) appeared to partition unequally between daughter cells (Figure 2A, EHT cell 1’ inherits the largest macropinocytic-like vacuole and emerges unambiguously as an EHT pol+ cell). This suggests that asymmetric inheritance of cytosolic vesicles containing apical proteins may contribute, presumably after delivery to the luminal membrane, to the acquisition of apicobasal polarity by EHT pol+ cells.

Altogether, these results support the idea that the HE, at the initiation of the EHT time-window, is not polarized. Subsequently, HE cells establish – or not – apicobasal polarity, thus leading to the emergence of either EHT pol+ or EHT pol-cell types (see our model Figure 2B). In the case of EHT pol+ cells and while emergence is proceeding, apicobasal polarity is maintained (if not reinforced) until the release.

Tuning of apicobasal polarity is sensitive to interfering with the Runx1 transcription factor and with blood flow

To substantiate our findings on the significance of controlling apicobasal polarity establishment on EHT sub-types, we explored the potential involvement of proteins of the Pard3 family; these proteins, recruited by transmembrane receptors via their PSD-95/Dlg/ZO-1 (PDZ) domains (Buckley and St Johnston 2022), are at the root of apicobasal polarity initiation and are essential for the maintenance of apical membrane functional properties (Román-Fernández and Bryant 2016). We anticipated that interfering directly with Pard3 proteins would hamper the development of the aorta (in addition to other functions of polarized tissues that are essential for embryonic development) and rather searched for correlative evidence for the differential expression of Pard3 isoforms - including natural loss or gain of function variants -, at critical time windows of the EHT process (i.e between 30-32 hpf (the timing at which the HE, as we show, does not appear to be polarized) and 48-50 hpf (the timing at which, as we show, part of HE cells have become competent to undergo emergence)). In addition, to provide functional support to this analysis and ensure that differences relate to what is taking place in hemogenic cells, our experimental conditions also included interfering with the gradual maturation of the HE (we have estimated that HE cells represent approximately 30% of total aortic cells at 48 hpf and others have estimated that they reach up to approximately 50% of total aortic cells at 28 hpf (Zhao et al. 2022)). We altered the maturation of the HE by abrogating blood flow and, more restricted to the EHT biological process per se, by interfering with the function of the Runx1 transcription factor whose expression is sensitive to fluid sheer stress (Adamo et al. 2009) and whose function is essential for EHT completion (Kissa and Herbomel 2010) and for regulating HSC number (Adamo et al. 2009; T. E. North et al. 2009).

Firstly, to reach these objectives, we sought to identify Pard3 isoforms that may be functionally relevant during EHT progression, using whole-mount in situ hybridization (WISH), RT-PCR and cloning. WISH performed at approximately 35 hpf allowed us to detect, in the aorta, mRNAs encoding for all Pard3 members, including Pard3ab, b, ba and bb, with prominent detection of Pard3ab and bb isoforms (see Figure 3A and Materials and Methods). Messenger RNAs, in the trunk region and encoding for all four isoforms, were also detected by qRT-PCR during the 30-32 and 48-50 hpf time windows (data not shown). We also obtained and cloned partial or complete cDNAs encoding for all four members after reverse transcription of mRNAs expressed in the trunk region of 48 hpf embryos. Sequencing unveiled yet unreferenced variants and, of relevance regarding our objectives, potential loss of function proteins for Pard3ab and Pard3bb (see Figure 3B and Materials and Methods); these include, for Pard3ab, a deletion of most of the sequence but that retains the amino-terminus containing the conserved region 1 (CR1) as well as the PDZ1 domain and, for Pard3bb, a carboxy-terminal truncation of the PDZ3 domain that nonetheless maintains the carboxy-terminal sequence of the protein (see Figure 3B and its legend for more details).

Figure 3:

Secondly, to interfere with Runx1 function, we generated a Tg fish line that expresses a truncated form of Runx1 (dt-runx1; fish line Tg(Kdrl:Gal4;UAS:RFP;4xNR:dt-runx1-eGFP) thereafter abbreviated Tg(dt-runx1)) deleted from its transactivation domain and carboxy-terminus but retaining its DNA-binding domain (the Runt domain (aa 55-183 (Kataoka et al. 2000; Burns et al. 2002; Kalev-Zylinska et al. 2002), see Figure 3C). Importantly, in this Tg fish line, dt-runx1 expression is restricted to the vascular system, hence excluding expression in the brain region that express Runx1 endogenously (see the aforementioned articles) and preventing biases owing to potential interference with neuronal functions. In addition, owing to the expression of eGFP concomitantly to dt-runx1 (eGFP is cleaved from dt-runx1 via a T2a site for endopeptidase), this fish line allows for the easy selection of embryos for imaging and for phenotypic analysis. In preliminary experiments aimed at addressing the localization of dt-runx1 as well as its stability, we expressed it transiently and measured its proper targeting to the nucleus, see Figure 3C).

To characterize further the Tg(dt-runx1) fish line, we performed time-lapse experiments using spinning disk confocal microscopy (with some of the experiments addressing apicobasal polarity performed with embryos obtained from outcrossing the Tg(dt-runx1) and Tg(Kdrl:mKate2-Podxl2) fish lines). Overall, we made the following observations: (i) hematopoiesis is affected, as attested by the significant increase in the number of hematopoietic cells in the thymus, at 5 dpf (days post-fertilization), in comparison to control siblings (Figure 3-figure supplement 1); (ii) at 30-32 hpf and contrarily to what we observed with our fish lines expressing either eGFP-Podxl2 or mKate2-Podxl2, HE cells appear to be polarized, based on enrichment of the polarity marker at luminal membranes. In addition, HE cells do not appear to contain any visible intra-cytosolic Podxl2-containing vesicles (Figure 3-figure supplement 2, compare with Figure 2 and Figure 2-figure supplement 1); (iii) at 48-55 hpf, we observe an accumulation of EHT pol+ cells both in the aortic floor and in the lateral sides of the aortic wall as well as of cells of uncharacterized nature (the latter may, hypothetically, represent EHT cells at an early phase of their emergence, including cells that should have evolved as EHT pol-cells (see Figure 3 - figure supplement 3A)); (iv) occasionally, we observe - for EHT cells exhibiting invagination of the luminal membrane -, scattered cytosolic and sub-plasmalemmal pools of Podxl2-containing membranes and, consistently, the apparent decrease of Podxl2 enrichment at the apical/luminal membrane (see Figure 3 - figure supplement 3B and compare with Figure 1C). In addition and occasionally as well, we observe the reversion of apparent EHT pol+ cells into apparent EHT pol-cells (data not shown). These two last observations suggest that perturbing the tuning of apicobasal polarity in the HE alters the morphodynamic characteristics of emergence types, particularly in the case of EHT pol-cells whose biomechanical features appear to require turning down apicobasal polarity establishment.

Altogether, the results obtained upon expression of dt-runx1 show impairment of hematopoiesis and suggest that, for both EHT cell types, the progression throughout EHT is perturbed, and so until the release. The accumulation of morphologically characterized EHT pol+ like cells that we observe may result from the sustained apicobasal polarity of the HE at early time points (around 30 hpf). Consistently, this appears to be detrimental to the emergence of EHT pol-cells that we could not identify unambiguously and that may be part of the cells that are of uncharacterized morphologically. In addition, our results suggest that precursors of HE cells are polarized, as is expected to be the case for non-hemogenic aortic cells, and that Runx1 is involved in controlling the molecular events that are tuning apico-basal polarity, starting at the initiation of the EHT time-window.

We then pursued our aim to correlate, using qRT-PCR, the expression levels of Pard3 isoforms with the apicobasal polarity status of the hemogenic tissue during the EHT critical time windows (30-32 versus 48-50 hpf). To complement interference with Runx1 function upon expression of dt-Runx1, we also prevented blood flow initiation using the well characterized silent heart (sih) morpholino which inhibits the expression of cardiac troponin T (Sehnert et al. 2002).

In a first series of 3 independent experiments performed with cDNAs obtained from dissected trunk regions of embryos at 30-32 and 48-50 hpf (Figure 4A), we measured the expression levels of the four Pard3 isoforms using heterozygous Tg(dt-runx1) fishes that were incrossed so as to obtain embryos expressing the highest possible amounts of dt-runx1, based on eGFP fluorescence resulting from proteolytic cleavage of the dt-runx1-eGFP fusion. To validate the efficiency of our interference conditions, we also measured expression levels of cmyb (an early marker of hematopoiesis) and of Scrib (a polarity protein involved in the functional delimitation of the cellular basal domain (Buckley and St Johnston 2022)). The qRT-PCR results show, for all Pard3 isoforms, a reproducible tendency for opposite effects of dt-runx1 expression between the 30-32 and 48-50 hpf time-windows, with the significant decrease of values for Pard3ab and bb at 48-50 hpf in comparison to controls. This suggests that the expression levels of Pard3 polarity proteins are differentially controlled between the two time-windows, indicating that the expression level of Pard3 polarity proteins is positively controlled in the wild type situation when emergence is proceeding at 48-50 hpf, but that it is rather negatively controlled, and thus decreased, at early time points (during HE maturation, at 30-32 hpf). These data and conclusions are also supported by results obtained with Scrib, with a comparable outcome. Importantly, cmyb expression was significantly decreased for both time-windows, consistently with the impairment in Runx1-dependent control of hematopoiesis.

Figure 4:

In a second set of 3 independent experiments, we interfered with blood flow using the Tg(Kdrl:Gal4;UAS:RFP) fish line after injection of the sih morpholino (MO) in 1 cell stage embryos. We focused on Pard3ab and Pard3bb isoforms (Figure 4B). We reproduced the results obtained for Pard3ab and Pard3bb upon dt-runx1 expression, with less variability and consistently with the function of blood flow in controlling Runx1 expression (Adamo et al. 2009). Finally, we also investigated expression levels of the +/-PDZ2 and +/-PDZ3 variants of Pard3ab and Pard3bb, respectively (using, as for experiments reported in Figure 4B, incrosses of heterozygous Tg(dt-runx1) fishes, see Figure 4C - left). We found that, at 48-50 hpf and for both proteins, isoforms including the full-length PDZ domains are the most decreased, hence being positively controlled in the wild type Runx1 context. This reinforces the consistency of our findings as well as the importance of controlling expression levels of fully active polarity proteins when emergence is proceeding. We also confirmed our results using sih MO injections, at least for the Pard3ab +/-PDZ2 variants (Figure 4C – right). Interestingly, the ratios between the Pard3ab +/-PDZ2 variants are opposite between the 30-32 hpf and 48-50 hpf time windows, reinforcing the idea of a fine tuning of apicobasal polarity functions mediated by aortic environmental constraints imposed by blood flow.

Junctional recycling is differentially controlled between EHT cell types

To investigate the functional links between apicobasal polarity and the peculiarities of EHT pol+ and EHT pol-emergence processes, we sought to follow the dynamics of Junctional Adhesion Molecules (JAMs) that belong to tight junction complexes (Garrido-Urbani, Bradfield, and Imhof 2014). During apicobasal polarity establishment in epithelial and endothelial tissues, these molecules recruit the Pard3/aPKC complex, via a PDZ-binding peptide located at their extreme carboxy-terminus (Figure 5A and (Itoh et al. 2001; Ebnet et al. 2001; 2018)). In addition, JAMs are expressed in the vascular system and, in relation to this work, JAM3b is expressed in the aorta of the zebrafish embryo and promotes endothelial specification and hematopoietic development (Kobayashi et al. 2020).

Figure 5:

We envisaged a scenario whereby EHT pol+ cells, whose longitudinal interface with endothelial neighbours shrinks along the X axis of the X, Y 2D-plane, may have less mobile junctional pools than EHT pol-cells whose entire junctional interface moves along the X, Y, Z 3D-axes (Figure 5 – figure supplement 1). In the case of EHT pol-cells, the consumption of the junctional interface with adjoining endothelial cells appears to result from the converging migration of endothelial neighbours crawling over the luminal membrane, based on interpretation of our time-lapse sequences (Figure 1-video 3). In this context, we favoured the analysis of junctional pools localized at antero-posterior sites of emerging cells as we have shown that they are enriched with tight junction components (Lancino et al., 2018). In addition and in the case of EHT pol+ cells, it is conceivable that these adhesion pools – spatially restricted owing to apicobasal polarity –, contribute to anchoring the emerging cell in the 2D-plane (Figure 5 - figure supplement 1).

To achieve our goal, we designed 2 constructs in which eGFP is introduced in the extracellular domains of the two JAM2a and JAM3b molecules (Figure 5A). To investigate their localization and proper targeting at junctional interfaces, these constructs were expressed transiently and ubiquitously, using the Hsp70 heat shock promoter. We observed that the 2 fusion proteins are efficiently targeted at cellular contacts and, more specifically, at the apical side of polarized epithelia such as for example cells of the ependymal canal (Figure 5 -figure supplement 2). We then established Tg fish lines expressing eGFP-Jam2a and eGFP-Jam3b under the control of the vascular kdrl promoter (Tg(kdrl:eGFP-Jam2a) and Tg(kdrl:eGFP-Jam3b)). Using these fish lines and spinning disk confocal microscopy, we observe a remarkable efficiency of targeting to intercellular junctions, for both proteins (data not shown for eGFP-Jam2a; for eGFP-Jam3b, see Figure 5B, C). Deployment of the aortic wall into 2D cartographies allows to point precisely at junctional pools established at the interface between endothelial and EHT cells and emphasizes on the enrichment of eGFP-Jam3b at antero-posterior poles of EHT pol+ and EHT pol-cells (Figure 5B, C bottom).

Using double transgenic Tg(kdrl:eGFP-Jam3b; kdrl:nls-mKate2) embryos at 48-55 hpf, we addressed the recycling capacity of junctional pools using Fluorescence Recovery After Photobleaching (FRAP). The labelling of nuclei with nls-mKate2 allowed to point at EHT pol+ and EHT pol-cells unambiguously because eGFP-Jam3b is essentially targeted to junctional contours and not labelling peripheral membranes, except in few cases (and nuclei of EHT pol+ cells have a crescent shape, see Figure 6A).

Figure 6:

First and to set up our protocol, we spotted bi- and tri-junctional contacts between endothelial cells (Figure 6 – figure supplement 1). Recycling parameters (fluorescence intensity recovery with time and maximum recovery amplitude (that addresses the mobile pool), Figure 6C) showed that bi-junctional contacts are less mobile than tri-junctions, with a higher dispersion of maximum recovery amplitude values for the latter (Figure 6E); this introduced a clear limitation for statistical significance of the results, although clear tendencies were observed for mean fluorescence intensity recovery (Figure 6C) and median values for maximum recovery amplitude (Figure 6E). We then focused at antero-posterior sites of EHT pol+ and EHT pol-cells and more specifically in the region of tricellular junctions (shared by 1 EHT cell and 2 endothelial neighbours, see Figure 6 – figure supplement 1) that are clearly the most enriched with eGFP-Jam3b. In each experiment, eGFP-Jam3b pools at tricellular junctions between endothelial cells were also spotted for a comparative analysis. Measurements of fluorescence recovery intensities revealed a significant increase in the mobile fraction of eGFP-Jam3b at EHT pol+ – EC versus EHT pol-– EC junctional interfaces (median values of 50% and 39% maximum recovery amplitudes after 10 min, respectively; Figure 6E)). Differences in recovery were also measured between EHT pol+ – EC and EHT pol-– EC junctional interfaces when focusing on the earliest time points (the first 30 seconds, Figure 6F), although with smaller significance, with the median value of early recovery slopes for the EHT pol+ - EC versus EHT pol-– EC junctional interfaces increased by 128% (Figure 6G).

Altogether and unexpectedly regarding our initial scenario, these results indicate that tri-junctional pools localized at antero-posterior poles of EHT cells and enriched with eGFP-Jam3b molecules are significantly more dynamic for EHT pol+ cells in comparison to EHT pol-cells. Since EHT pol+ cells - by virtue of apicobasal polarity establishment possibly assemble an apical endosome -, this should favour the rapid recycling of eGFP-Jam molecules (Ebnet et al. 2003). This recycling may entertain a certain rate of delivery of Pard3 proteins at apical junctional regions and contribute to the maintenance/reinforcement of polarity throughout the entire emergence process. In addition, this turnover may be adapted to the mechanical tensions EHT pol+ cells are exposed to, in comparison to EHT pol-cells, owing to their specific morphodynamic features (ex: cellular bending); these tensions may be sensed by tri-cellular junctions which would be preferential sites for the renewal of junctional components whose half-life is reduced owing to submission to mechanical forces.

ArhGEF11/PDZ-Rho GEF plays essential functions during EHT progression

Junctional maintenance and recycling are dependent on intracellular membrane trafficking, supported by sub-cortical actin remodelling and actomyosin contractility, which are controlled mainly by GTPases of the Rho family (Ridley 2006; Olayioye, Noll, and Hausser 2019). Owing to the significance of apicobasal polarity control on EHT features, as suggested by our work, we investigated which proteins may be essential for actin/actomyosin regulation and focused on regulators of Rho GTPases, in particular Rho GEFs that catalyse exchange of GDP for GTP to positively regulate their activity (Rossman, Der, and Sondek 2005). As for Pard3 proteins, several of these Rho GEFs contain one or several PDZ domain(s) that target most proteins to complexes acting at the apical side therefore interlinking actin/actomyosin regulation with cell polarity (Mack and Georgiou 2014; Ebnet and Gerke 2022). We focused on 9 PDZ-domain containing Rho GEFs, all encoded by different genes in the zebrafish (Figure 7-figure supplement 1): ArhGEF11/PDZ-RhoGEF (thereafter shortened as ArhGEF11), ArhGEF12a, ArhGEF12b, PRex1, PRex2, Tiam1a, Tiam1b, Tima2a, Tiam2b. We first investigated their expression by Whole mount In Situ Hybridization (WISH) and found that all 9 mRNAs are detected in the aorta, and for the vast majority at 30-32 and 48-50 hpf (Figure 7-figure supplement 2). Then, using qRT-PCR (Figure 7-figure supplement 1B), we measured and compared their expression levels in the trunk region (at 35 and 48 hpf), for dt-runx1 expressing embryos and controls. We found that, in comparison to controls, ArhGEF11, ArhGEF12b, Tiam1b and Tiam2a are significantly reduced upon dt-runx1 expression at 48 hpf (hence being positively controlled in the wild type condition) and that their expression appears to be oppositely controlled at 35 hpf; this mirrors the variations measured for polarity proteins (Figure 4). This is consistent with a functional link between these Rho-GEFs and apicobasal polarity control during EHT progression. We finally decided to focus on ArhGEF11 for the following reasons: (i) in comparison to Tiams that act on Rac1, ArhGEF11 and ArhGEF12 (which are close relatives and can form heterodimers (Chikumi et al. 2004)), are mostly acting on RhoA which is controlling apical constriction via the RhoA-Myosin II signalling axis; as we have shown previously (Lancino et al. 2018), EHT progression requires the constriction of circumferential actomyosin; (ii) ArhGEF11 was shown to regulate the integrity of epithelial junctions by interacting directly with the scaffolding protein ZO-1, hence connecting inter-cellular adhesion with RhoA-Myosin II (Itoh et al. 2012); (iii) ArhGEF11 mediated apical constriction is essential during tissue morphogenesis such as, for example, the neural tube formation in which epithelial cells, submitted to mediolateral contractile forces, constrict at their apical side thus triggering inward bending of the neural plate which leads to the formation of the neural tube (Nishimura, Honda, and Takeichi 2012). The EHT may share features with this process, i.e the anisotropic distribution of contractile forces controlling the plane of emergence; (iv) mammalian ArhGEF11 exhibits alternative splicing in its C-terminal region that controls tight junction stability via the regulation of RhoA activity (Lee et al. 2018) as well as cell migration and invasion (Itoh et al. 2017).

To confirm the potential function of ArhGEF11 at the junctional interface between HE/EHT and endothelial cells, we first investigated its intracellular localization. We attempted to detect the full-length form upon expression of a GFP fusion protein, in the vascular system, and failed to do so. We then generated a truncated form that retains the N-terminal fragment encompassing the PDZ and RGS domains (see Figure 7 – figure supplement 3A) fused with eGFP in its C-terminus. Upon transient expression in the vascular system, we visualized its localization at the interface between endothelial and hemogenic cells progressing throughout the EHT, with an apparent increase in density at antero-posterior regions between adjoining EHT cells (Figure 7 -figure supplement 3B).

In line with the formerly described function of a splicing variant of ArhGEF11 in controlling tight junction integrity, particularly during egression of cells from the skin epithelium in the mouse (Lee et al. 2018), we questioned the potential role of such variant in EHT. This variant, referred to as the mesenchymal form, results from the inclusion of a peptide encoded by exon 37 or exon 38 in mouse and human, respectively (Shapiro et al. 2011; Itoh et al. 2017; Lee et al. 2018). This insertion locates in the degenerated C-terminal region of the protein, which is predicted to be relatively poorly organized in its 3D structure. Upon amplifying the full-length sequence of zebrafish ArhGEF11 for investigating its localization, we cloned fragments encoding – or not – for the insertion of a peptide of 25 amino-acid residues (75 base pairs corresponding to exon 38). Although variable when compared with other species (see Figure 7 – figure supplement 4C, bottom panel), this peptide is inserted in the same region as it is for mammals and may correspond to an ArhGEF11 variant functionally equivalent and involved in the regulating of junctional stability (see Discussion).

To investigate the function of ArhGEF11 on the junctional interface between HE/EHT and endothelial cells, and more specifically of the isoform containing the exon 38 encoded peptide, we used both morpholino (MO) and CRISPR-based strategies. We designed a splicing MO at the 3-prime exon/intron boundary of exon 38 that interferes with its inclusion in the encoding mRNA (Figure 7 -figure supplement 4A). This MO did not trigger any deleterious effect on the gross morphology of zebrafish embryos (Figure 7 – figure supplement 4B) and blood circulation was normal in comparison to control embryos. We attempted to generate CRISPR-based genomic deletion mutants of exon 38, both using Cpf1 and Cas9 enzymes and failed (see Materials and Methods). However, using CRISPR/Cas9, we obtained a deletion mutant triggering a frame shift downstream of exon 38 and introducing a premature stop codon few amino-acid residues downstream thus leading to a sequence encoding for an ArhGEF11 C-terminal deleted form (see Materials and Methods and Figure 7-figure supplement 4C)). Unlike the variant skipping exon 38 induced by the splicing MO, expression of the CRISPR/Cas9 C-ter deletion mutant triggered, around 24 hpf, a severe retardation of blood circulation initiation in approximately 80% of the embryos obtained from incrossing ArhGEF11^{CRISPR-Cterdel-/+} heterozygous mutant fishes. From 24 to 48 hpf, approximately 50% of these embryos recovered a normal blood circulation (suggesting that these embryos are probably heterozygous for the mutation), approximately 35% remained with a severe phenotype (characterized by a large pericardiac oedema) and approximately 15% died, see Figure 7 – figure supplement 4D and Materials and Methods). This indicates essential functions of the C-ter region of ArhGEF11, in agreement with previously published data on the mammalian protein (Chikumi et al. 2004).

We then characterized more in depth the MO and CRISPR phenotypes and performed a quantitative analysis of the number and morphology of HE, EHT and adjoining endothelial cells, based on confocal images and subsequent segmentation of cell contours (using 2D deployment of the aortic cylinder).

For the morphants, we found approximately a double amount of hemogenic cells in the aortic floor (altogether localized on the floor and on the lateral sides of the aorta, see Figure 7A and Figure 7 -figure supplement 5); increase in the number of hemogenic cells also concerned cells oriented more perpendicularly to the blood flow axis (toward the Y-axis, Figure 7Aa’’), consistently with variation in emergence angle, as we have shown previously (Lancino et al. 2018). This is accompanied by a significant reduction in the number of morphologically characterized EHT cells (Figure 7A and Figure 7 – figure supplement 5). Similar results were obtained for the CRISPR homozygous deletion mutants, particularly regarding the accumulation of hemogenic cells, although with a smaller amplitude (Figure 7B and Figure 7 -figure supplement 6). Consistently, a significant increase in the aortic perimeter was found, at least for the morphants (Figure 7C). Of notice also, hemogenic cells in morphants are on average less elongated in comparison to control and to the CRISPR homozygous mutants (as well as less dispersed according to this parameter), with the accumulation of a cell population standing below the median value (Figure 7Aa’’’; to some extent, this is reproduced as well when measuring hemogenic cell area). Finally, the population of small cells in the morphants are of uncharacterized morphology, which means that they appear neither as EHT pol+ nor as EHT pol-cells and may represent abortive emergences (Figure 7A and Figure 7 – figure supplement 5 in which cellular morphology can be assessed from z-stacks).

Figure 7:

Altogether, the morphants and CRISPR mutant phenotypes show that the progression throughout EHT is significantly impaired supporting the idea that ArhGEF11 exerts important functions in the process. Increase in the frequency of elongated cells for the CRISPR mutants in comparison to morphants suggests that interference triggered by the C-ter deletion of ArhGEF11 is more effective at early time points. Interference with inclusion of exon 38 in morphants triggers an increase in the number of hemogenic cells of smaller area and reduced elongation indicating a more efficient progression throughout emergence in comparison to the CRISPR mutants, albeit with impairment at later stages. This may be due to alteration in the dynamics of endothelial cell intercalation which is required for emergence completion (which is also compatible with the accumulation of pairs of cells that may be post-mitotic and that do not spread apart because this also requires aortic cell intercalation).

ArhGEF11/PDZ-RhoGEF and its variant containing the exon 38 encoded sequence control junctional dynamics to support reduction of the hemogenic/endothelial interface and cell-cell intercalation

To investigate how ArhGEF11 controls the dynamic interplay between HE/EHT and endothelial cells, we characterized further the MO and CRISPR phenotypes, at the junctional level. To do so, we performed FRAP experiments using the Tg(Kdrl:eGFP-Jam3b) fish line that was injected at the one cell stage with the exon 38 splicing MO, or incrossing (Kdrl:eGFP-Jam3b x ArhGEF11^{CRISPR-Cterdel-/+}) heterozygous fishes to obtaining homozygous mutants and wild type alleles for control siblings.

FRAP experiments were focused on the hemogenic (HE)/endothelial (EC) membrane interface, with the support of images obtained after 2D deployments of aortic segments (Figure 8A). Pools of eGFP-Jam3b molecules localized at bi-junctional (HE-EC) and at tri-junctional (HE-EC-EC and HE-HE-EC) interfaces were bleached and subsequently imaged for FRAP analysis. In the case of the MO treatment (in which case HE cells accumulate on the aortic floor (see Figure 7A and the 2D cartography of the aortic floor Figure 8A)) we measured, in comparison to control, an increase in FRAP recovery parameters for eGFP-Jam3b pools at the tri-junctional HE-HE-EC interface (Figure 8B), which is the most significant for the recovery speed at early time points (Figure 8B, b’’’). Results are different for the ArhGEF11^{CRISPR-Cterdel+/+} homozygous mutants for which we did not observe any obvious effect at HE-HE-EC tri-junctions (Figure 8C) but, rather, a small tendency for an increase in the mobile pool at HE-EC-EC tri-junctions (that has the particularity to concern 2 endothelial cells that contact the longitudinal membrane of one HE cell, see Figure 8A, D).

Figure 8:

Altogether and including our results presented Figure 7, these results reinforce the idea that ArhGEF11 and its variant encoded by exon 38 are involved in critical steps of EHT progression, partly by controlling the dynamics of junctional pools at the level of tri-junctions established between adjoining hemogenic and endothelial cells (see the model Figure 8D and its legend for a detailed analysis of mechanistic issues). The results also suggest that the +exon 38 peptide encoding variant is prominently involved in controlling the HE-HE-EC interface and more specifically during intercalation, consistently with the accumulation, upon MO interference, of EHT cells having progressed throughout the emergence but appearing to be slowed down when reaching completion (cells of smaller area and less elongated, see Figure 7). Finally, ArhGEF11 is controlling RhoA and, most probably, its function in EHT progression also involves actin and acto-myosin activities that are involved in the contraction of HE and EHT cells (see Lancino et al. 2018; this is coupled with the consumption of longitudinal and transversal membrane interfaces, see also the model Figure 8D and its legend).

Discussion

Characteristics of the HE and complexity of pre-hematopoietic stem cell emergence

Heterogeneity in the identity and hematopoietic potential of cells emerging from the HE in the AGM has been recently shown in co-culture systems and in vivo in the mouse (Ganuza et al. 2017; Dignum et al. 2021) and in the zebrafish model (Tian et al. 2017; Xia et al. 2023); however, knowledge on the fundamental cell biological and molecular events leading to this complexity is still lacking. Here, we highlight functional aspects that may significantly contribute to this complexity and that take root in the cellular plasticity of the HE. We bring evidence suggesting that HE cells that become competent to initiate the EHT time-window (around 28-30 hpf) tune their apicobasal polarity status which endows them with the ability to support two emergence processes with radically different morphodynamic characteristics; eventually, this would lead to differentially fated cells, which remains to be determined.

While the ability of the HE to regulate its polarity features may be inherited from molecular cues involved in its upstream intra-aortic specification (such as for example NOTCH (Robert-Moreno et al. 2008; Gama-Norton et al. 2015; Bonkhofer et al. 2019), TGFβ (Monteiro et al. 2016), Gata2 (Butko et al. 2015; Daniel et al. 2019)), or any other of the more upstream factors involved in hemogenic specification (Zhao et al. 2022)), our results suggest that Runx1 is taking part in this regulation, consistently with its hematopoietic and EHT-inducing activity (North et al. 1999; Kalev-Zylinska et al. 2002; Kissa and Herbomel 2010; Lancrin et al. 2012). Our experiments using an interfering form of Runx1 (dt-Runx1) that produces a relatively weak an subtle phenotype in comparison to morpholino treatment that prevents aortic cells to convert into HE cells (Bonkhofer et al. 2019) or genetic mutation and gene knockout abrogating hematopoiesis (T. E. North et al. 2009; Sood et al. 2010; Gao et al. 2018), provided the opportunity to observe that, at 30 hpf, elongated HE cells on the aortic floor appear to be maintained and to exhibit apicobasal polarity, with the absence of apparent cytosolic vacuolar structures (as has been described in the mouse in Runx1 interfering condition (T. North et al. 1999)), which correlated with the subsequent accumulation of EHT pol+ cells that appeared to be at the expense of EHT pol-cells. This is in favour of the idea that the two cell types arise from a common precursor capable of evolving toward the one or the other, and that has lost apicobasal polarity (see the model Figure 2B). Recent work that has identified Runx1 targets in the zebrafish embryo at 29 hpf does not point at any of the conventional apicobasal polarity organizers as being direct targets although Pard3 and Pard6 mRNAs are clearly downregulated in the HE in comparison to aortic roof cells (Bonkhofer et al. 2019) which is consistent with our results at early time points of the EHT time-window.

A key question arising from our observations is what would control the balance orienting toward an EHT pol+ or an EHT pol-type of emergence. Cell division may be orienting toward EHT pol-emergence type (which would also be the reason why these cells have a round-shape morphology in comparison to EHT pol + cells). The frequency of EHT pol+ versus EHT pol-cells that we have estimated and that is of approximately 2/1 is not incompatible with this possibility since divisions take place in the HE at a relatively high frequency. Indeed, it has been estimated that 50% of HE cells that constitute the aortic floor at 28 hpf will divide and will undergo emergence thereafter during the 28-72 hpf time window (Zhao et al. 2022). However, we observe that both EHT pol+ and EHT pol-cells can divide during the emergence process (see also our former work Lancino et al. 2018 for EHT pol+ cells), which is followed by the full-recovery of their respective morphodynamic characteristics (not shown). Hence, this excludes the possibility of a short term and purely mechanical effect of cell division on emergence morphodynamics. In link with cell division however, there exist the possibility that the pool of large vesicular structures potentially awaiting for exocytosis in HE cells having lost their apicobasal polarity may become asymmetrically inherited after division. This would endow the cell remaining with the largest pool to undergo EHT pol+ emergence after exocytosis and expansion of the apical/luminal membrane. Our results Figure 2 would support this possibility but what controls this asymmetry would remain to be established.

Still, regarding the control of the balance orienting toward an EHT pol+ or an EHT pol-type of emergence, the intrinsic nature of the HE, i.e. its acquaintance with the aortic differentiation programme (Bonkhofer et al. 2019), raises the possibility of a partial recovery of an aortic cell phenotype in the case of EHT pol+ emergence, thus supporting its very unique mechanobiology. Therefore, the regulation of local signalling between aortic and HE cells may tune the balance toward the one or the other fate; this could be the case, for example, for adjoining HE cells in comparison to isolated ones that would then receive unequal signalling cues owing to the difference in homotypic versus heterotypic contacting surfaces. In this scenario, NOTCH signalling which is determinant for aortic and HE specification (Gama-Norton et al. 2015) may play a significant role, particularly via the ratio of the Jag1/Dll4 ligands that is fundamentally involved in HE versus aortic specification (Robert-Moreno et al. 2008; Bonkhofer et al. 2019).

Finally, we cannot totally exclude that EHT pol+ and EHT pol-cells would originate from different precursors. On this line, it has been shown that the aortic wall can home a transient veinous-fated progenitor cell population that will lead to the underlying vein via selective sprouting from 21 to 23 hpf (Herbert et al. 2009). This option would require a massive expansion of a minor fraction of the cell population remaining in the aortic floor since it is expected that most of it has been moving out toward the sub-aortic region, which is not in favour of this possibility. In addition, recent scRNAseq analysis of floor and roof aortic cells at 21 and 28 hpf has led, for both timing points, to two well defined clusters defined only by aortic and hematopoietic signatures (Zhao et al. 2022).

The complexity of pre-HSC emergence and heterogeneity of hematopoietic stem cell and progenitor populations

Currently, we do not know what would be the features that would endow the two cell types to behave differently after their release. Since post-EHT cells have to migrate in transient developmental niches (i.e the CHT, the thymus, the pronephric region), these features may provide them with more or less ability to migrate in the sub-aortic space and pass throughout the wall of the underlying vein to conquest more distant niches (Murayama et al. 2006), or to remain in a local AGM niche and be exposed to specific signalling there, in the proximity to the aorta as it is the case for hematopoietic clusters in mammals and avians (Ciau-Uitz et al. 2014; Jaffredo and Yvernogeau 2014). Beside their potential differences in colonizing different niches which would impact on their subsequent fate, these cells may also contribute to more or less transient waves of hematopoietic stem cells and progenitors that would be specific to the developmental period such as, for example, a restricted sub-set of T-lymphocytes that was described before (Tian et al. 2017). In the specific case of the Tian et al. work, the AGM was shown to produces transient T-lymphocytes proposed to be independent of long-lived HSCs according to a gradient that increases from the anterior to the posterior part of the trunk region, as well as from the aorta in the tail (the posterior blood island in the tail region that was proposed to give rise to hematopoietic precursors related to mammalian erythro-myeloid progenitors, (Bertrand et al. 2007)). Since the EHT pol+ and EHT pol-cells that we describe in our work are equally produced from the aorta in the trunk and in the tail regions (thus irrespective of the antero-posterior axis of the aorta), the one or the other of the two EHT types most probably does not directly relate to the first wave of T-lymphopoiesis arising from the aortic endothelium described in Tian et al 2017.

The cell biology behind emergence control and mechanics

The intrinsic capacity of the HE, that we show here to be able to support two highly distinct emergence processes, in the zebrafish embryo, is very unique. Indeed, not only are they taking place contemporarily, but also can the two cell types proceed throughout emergence at a distance shorter than 50 µm, which accommodates the intercalation of a single adjoining endothelial cell. This may reflect the remarkable properties of the vascular system that needs to allow cellular extrusion and transmigration. The latter is particularly relevant in the context of developmental hematopoiesis since hematopoietic precursors that travel via blood circulation need to pass throughout the aortic wall to conquest their distant niches.

In the present work, we provide some of the mechanistic insights into the specificities of the two emergence types and that rely on two fundamental aspects of cellular biology, i.e apicobasal polarity establishment and junctional mobility. The establishment and maintenance of apicobasal polarity in a cell extruding from its environment, as is the case for EHT pol+ cells, is at odds with the fundamental mechanisms of cell extrusion irrespective of the context into which the process is taking place (Nieto et al. 2016; Gudipaty and Rosenblatt 2017; Pei et al. 2019; Staneva and Levayer 2023); this places the emergence of EHT pol+ cells as an extreme case of extrusion and it is somehow unexpected to observe this type of emergence when the aortic wall can extrude EHT pol-cells that appear to emerge contemporarily, apparently according to a more conventional mode. During the EHT time-window, Runx1 is directly controlling the emergence programme by regulating Rho-dependent cytoskeletal functions in HE cells derived from mouse ES lines (Lie-A-Ling et al. 2014) and inducing the expression of Gfi1ab in the zebrafish (Bonkhofer et al. 2019) that belongs to the family of transcriptional repressors of the arterial programme expressed in the HE (Gfi1 and Gfi1b in the mouse, (Lancrin et al. 2012)). This programme downregulates key molecules involved in the maintenance of the aortic endothelium among which Cdh5 (VE-Cadh) and the downregulation of cadherins is a hallmark of extrusion. Thus, for EHT cells, and particularly EHT pol+ cells, downregulation of VE-Cadh should be compensated by other mechanisms that allow for the maintenance of aortic integrity. In this context of adhesion downregulation, the JAMs (that are expressed endogenously in the vascular system and that were used in this study) appear to be a well-suited substitute. As part of tight junction complexes, they can strengthen adhesion and, in addition, since they are also involved in cell migration (in particular leukocyte trans-endothelial migration, see (Ebnet 2017)), they could in theory support the extrusion of EHT pol-cells that do so according to a seemingly migration-type emergence mode (inferred from our time-lapse sequences) more or less related to trans-endothelial migration. In the case of EHT pol+ cells, we propose that JAMs may significantly contribute to the re-establishment of apicobasal polarity and to the strengthening of adhesion while EHT pol+ cells should undergo strong mechanical tension (owing to cellular bending). Interestingly, Jam3b appears to be a direct and positively regulated target of Runx1 (Bonkhofer et al. 2019), which is supporting its aforementioned potential functions during EHT.

Finally, regarding inter-cellular junctions and the control of their dynamics all along the EHT process, we also addressed the dynamic interface between aortic and HE cells. As for EHT cells, we focused more specifically on tri-junctional interfaces that have been proposed to sense tension and transmit the information to other types of junctional complexes (Bosveld, Wang, and Bellaïche 2018); in the EHT context, the contracting hemogenic endothelium is additionally exposed to high mechanical tension owing to wall sheer stress imposed by the blood flow and that was proposed to regulate emergence efficiency (Lundin et al. 2020; Campinho et al. 2020; Chalin et al. 2021). In our work, we highlight one essential aspect of controlling junctional recycling which is during cell-cell intercalation which concerns either contacting HE cells lying on the aortic floor or HE/EHT cells after cytokinesis (as mentioned before, cell division is rather frequent during the EHT time window (Zhao et al. 2022)). In this context, we addressed the function of ArhGEF11/PDZ-RhoGEF which provides a functional link between cell polarity (with its PDZ-binding motif and its RGS domain that couples to G-protein coupled receptor signalling to regulate planar cell polarity (Nishimura, Honda, and Takeichi 2012)), cell adhesion (it binds to the tight junction associated regulatory protein ZO-1, (Itoh et al. 2012)), and cytoskeletal regulation/contraction (Itoh et al. 2012; Itoh 2013), all three essential mechanistic components of morphogenetic movements and tissue remodelling. Since morphodynamic events leading to cell extrusion and to the control of cell migration do not only rely on the regulation of mRNAs levels but also on alternative splicing (Pradella et al. 2017) our interest for investigating the function of ArhGEF11 in EHT was raised by the fact that the protein undergoes changes in its biological properties via the alternative splicing of a small exon which leads to a modification of its C-terminus. We identified a splicing event in this region (concerning exon 38) for a potential zebrafish isoform ortholog. However, the precise function of this variant for the zebrafish protein, that was shown in the case of the mammalian isoform to induce cell migration and motility (Itoh et al. 2017) and to bind to the Pak4 kinase thus leading to the subsequent destabilization of tight junction complexes via loss of RhoA activation (Lee et al. 2018), remains to be established. On this line, our results consistently indicate that preventing exon 38 insertion interferes with EHT progression (as well as deleting the C-terminus of the protein as shown with our CRISPR mutant) and appears to impair intercalation of aortic cells between adjoining HE cells. Hence, the +exon 38 isoform, by regulating the dynamic interplay between HE and endothelial cells, would favour one of the essential steps of the EHT process leading to its completion and requiring the complete isolation of EHT cells to ensure the ultimate sealing of the aortic floor (see our former work (Lancino et al. 2018)). It remains to be established, as our measurements indicate, why the ArhGEF11 +exon 38 isoform appears to slow down junctional recycling which is counterintuitive with increasing the HE/endothelial dynamic interface. However, upon interference, we have been following Jam3b recycling which is only a sub-population of tight junction components that should be at play in the system. Finally, the EHT process on its whole may be submitted to a complex regime of alternating contraction and stabilization phases as we have shown previously in the context of EHT pol+ emergence and that may reflect the necessity to adapt to mechanical constraints imposed by the environment as well as the rearrangements of the HE/EHT/aortic cellular interface (Lancino et al. 2018). In this context, back and forth regulatory mechanisms, particularly involved in the control of the RhoA-Myosin II signalling axis partially regulated by ArhGEF11, may locally and at specific timing points change the turnover of junctional molecules, thus blurring the correlation between gross phenotype (here the organization of the HE/aortic interface) and very local and dynamic molecular events (for a complement to the discussion, see also the legend of the model presented Figure 8C).

Overall, our work highlights the complexity of pre-hematopoietic stem cell emergence as well as some of the essential molecular and mechanistic aspects involved. We show that the aorta, in the zebrafish embryo, produces two fundamentally different types of EHT cells and propose that this results from specific functional features of the HE regulated by the transcription factor Runx1. This would support the production of cell types potentially endowed with different cell fate potential, subsequently reinforced by the type of niche into which these cells will establish homing. It now remains to be established if the different cell types that we describe are indeed leading to cells that will eventually hold specific hematopoietic fate and properties.

Materials and Methods

Key Resources Table

For the exhaustive list of all primers, see Materials and Methods - Supplementary Table.

Contact for reagent and resource sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding Author, Anne A. Schmidt (anne.schmidt@pasteur.fr).

Zebrafish husbandry

Zebrafish (Danio rerio) of the AB background and transgenic fish carrying the following transgenes Tg(kdrl:ras-mCherry) (Chi et al. 2008); Tg(kdrl:nls-mKate2); Tg(kdrl:Gal4;UAS:RFP); and the fish lines generated in this study: Tg(kdrl:Gal4;UAS:RFP;4xNR:eGFP-Podxl2); Tg(kdrl:mKate2-Podxl2); Tg(kdrl:eGFP-Jam2a); Tg(kdrl:eGFP-Jam3b); Tg(kdrl:Gal4;UAS:RFP;4xNR:dt-runx1-eGFP); Tg(ArhGEF11^{CRISPR-Cterdel-/+}) were raised and staged as previously described (Kimmel et al. 1995). Adult fish lines were maintained on a 14 hr light/10-hr dark cycle. Embryos were collected and raised at either 28.5 or 24°C in N-Phenylthiourea (PTU, Sigma Aldrich, Cat# P7629)/Volvic source water (0.003% final) to prevent pigmentation complemented with 280 µg/L methylene blue (Sigma Aldrich, Cat# M4159). Embryos used for imaging, extracting mRNA, or for WISH ranged from developmental stages 28-to-60 hpf and larvae used for imaging were of 3-5 dpf, precluding sex determination of the animals. The fish maintenance at the Pasteur Institute follows the regulations of the 2010/63 UE European directives and is supervised by the veterinarian office of Myriam Mattei.

mRNA extraction and cDNA synthesis

Total RNA was extracted from whole 48 hpf embryos for cDNA cloning or from pooled trunks at the desired developmental stages (30-32 hpf, 48-50 hpf time windows for qRT-PCR experiments) (∼30 individuals per tube). Briefly, embryos were anesthetized using balneation in embryo medium supplemented with tricaine methanesulfonate (MS-222, Sigma-Aldrich Cat# A5040), at a final concentration of 160 µg/ml. RNA was extracted via organic extraction using TRIzol reagent (Invitrogen Cat# 15596026) according to the manufacturer’s guideline. gDNA contaminant was removed using TURBO™ DNase (Invitrogen Cat# AM2238) treatment according to the manufacturer’s guideline. Total RNA was stored at -80°C. Reverse transcription of mRNA was performed using SuperScriptIV (Invitrogen Cat# 18090010) with OligodT primer. cDNA samples were stored at -20°C until further processing (for quantitative real-time PCR, PCR, cloning and sequencing).

Transient and stable transgenesis

The kdrl promoter (flk) (Jin et al. 2005) was used to drive endothelial expression of mKate2-Podxl2, eGFP-Jam2a and eGFP-Jam3b, with eGFP (Clontech) and mKate2 (Evrogen) cDNAs amplified using overlapping primers for Gibson cloning. eGFP-Podxl2, ArhGEF11(PDZ-PRD-RGS-eGFP) and dt-runx1-eGFP were amplified for cloning into a pG1-4XNR vector (built from the 4XNR (non-repetitive) 4X UAS sequence less susceptible to methylation than the 14X UAS described in (Akitake et al. 2011)). For the sequence of all the designed cloning primers, see the Key Resources Table.

Podocalyxin-like 2 (Podxl2), ArhGEF11/PDZ-RhoGEF, JAM2a and JAM3b full length sequences were amplified from pools of 48 hpf whole embryo cDNA (see before for mRNA extraction and cDNA synthesis methodologies). All constructs generated in this study were obtained using the Gibson assembly assay (NEB, Cat # E2611S).

For eGFP-Jam2a and eGFP-Jam3b cloning, we used the JAM2a signaling peptide sequence (included in the 5-prime amplification primers and encoding for the signal peptide: mlvcvsllilihsvpvspvtvssr) and, for both constructs, the eGFP sequence was inserted in frame upstream of the sequence encoding for the transmembrane domain; for the eGFP-Jam2a construct, the eGFP was inserted upstream of amino-acid D221 (DLNVAA) of the NCBI Reference Sequence: NP_001091734.1; for the eGFP-Jam3b construct, the eGFP was inserted upstream of amino-acid D243 (DINIAG) of the NCBI Reference Sequence: NP_001076332.2. For transient ubiquitous and inducible expression, the eGFP-Jam2a and eGFP-Jam3b fusion constructs were inserted into a KpnI pre-digested pG1 plasmid containing tol2 sites and the HSP70-4 promoter. For stable transgenesis using the kdrl (flk1) promoter, constructs cloned into the pG1-HSP70 plasmids were extracted using XhoI/NotI restriction enzymes and inserted into the XhoI/NotI pre-digested pG1-flk1 vector using DNA T4 ligation (NEB, Cat # M0202S).

For the Podxl2 N-ter deletion mutant that has been used for establishing stable Tg fish lines, the cDNA encoding for the peptide sequence starting with amino-acid G341 (GGTEYL; fragment 341-587) of the NCBI Reference Sequence: XP_692207.6 was fused with either eGFP or mKate2 cDNAs. The sequence encoding for the signal peptide of the human CD8 (malpvtalllplalllhaarpsqfrvs) was introduced upstream (5-prime) of the eGFP or mKate2 sequences (included in the 5-prime primers designed for the Gibson cloning strategy). For the construct designed to obtain the Tg(kdrl:Gal4;UAS:RFP;4xNR:eGFP-Podxl2) fish line, 2 overlapping fragments were amplified for cloning into a pG1-tol2_4xNR vector after digestion with NcoI, with the following couples of primers: 4XNR_CD8-eGFP_fw and delPodxl2_eGFP_rev; delPodxl2_fw and Podxl2_pG1-4XNR_rev. For the construct designed to obtain the Tg(kdrl:mKate2-Podxl2) fish line 2 overlapping fragments were amplified for cloning into a pG1-flk1 vector after digestion with EcoR1, with the following couples of primers: pG1-flk1_CD8_mKate2_fw and delPodxL2-mKate2_rev; mKate2-delPodxL2_fw and Podxl2_PG1-flk1_rev.

For the ArhGEF11(PDZ-PRD-RGS-eGFP) construct, the cDNA encoding for the peptide ending upstream of the DH domain at Valine 709 of the amino-acid sequence (GenBank: AY295347.1; PALDEDV; fragment 1-709) was fused upstream of the sequence encoding for eGFP, for expression using the pG1-4xNR vector. The cDNA was amplified using the GB_4xUAS-gef11_fw; GB_eGFP-RGS_rev couple of primers.

For the construct designed to obtain the Tg(kdrl:Gal4;UAS:RFP;4xNR:dt-runx1-eGFP) fish line, the Runx1 sequence amplified omits the ATG (replaced by an nls) and ends downstream of the Runt domain (ending at amino-acid sequence PAHSQIP). The construct allows for the expression of the dt-runx1 protein under the control of the 4xNR driver using the same strategy as for the above PodxL2 constructs. The dt-runx1 cDNA construct was amplified from the zebrafish Runx1 cDNA sequence (Kalev-Zylinska et al. 2002) using the couple of primers GB-4xnr-nlsRunx1_fw; GB-eGFP-T2A-2xHA-delRunx1_rev. The sequence of these primers each allowed for the introduction, in the protein sequence, of a nls (encoded by the fw primer) and of a 2xHA-T2A peptide (encoded by the rev primer that overlaps with the 5-prime sequence of eGFP).

For all constructs, cloning reactions were used to transform NEB 5-alpha Competent E. coli (NEB, Cat # C2987) and resistant clones then used for miniprep plasmid preparations (Nucleospin plasmid, Macherey-Nagel, Cat # 740588.50), followed by sequencing. Plasmid DNA from clones containing correct insertions where then used to transform E. coli TOP10 competent cells (Invitrogen, Cat # C404010) and purified using the endotoxin free NucleoBond Xtra Midi kit (Macherey Nagel, Cat# 740420.10). Transgenesis was then performed by co-injecting 1 nl of plasmid (at 25 ng/µl) mixed with tol2 transposase mRNA (at 25 ng/µl) transcribed from linearized pCS-zT2TP plasmid using the mMESSAGE mMACHINE SP6 kit (Ambion, Cat# AM1340). For stable transgenesis, embryos were screened for fluorescence between 24 hpf and 48 hpf and raised to adulthood. Founders with germline integration were then isolated by screening for fluorescence after outcrossing with AB adults and for establishment of the F1 generation.

CRISPR methodology and GEF11 mutant transgenic line screening

CRISPR/Cas9 mutagenesis and isolation of founders

A sgRNA was designed against the splice donor site at the end of exon 38 with the aim to interfere with the production of the splicing variant including its encoding sequence, using the CRISPRscan web tool (Moreno-Mateos et al. 2015). The sgRNA was obtained by annealing for 5 min at 95°C equal volumes of 100µM specific Alt-R® crRNAs (IDT, Cat# sequence: AGCCAATCGTCTGAGGACGG) with 100µM Alt-R® generic tracrRNA (IDT, Cat# 1072532). Cas9 Nuclease Reaction Buffer (NEB Cat# B0386A) was added to obtain a final 45µM sgRNA stock solution. For the generation of the sgRNAs–Cas9 complex, a mix containing 18µM sgRNA and 12µM Cas9 protein (EnGen™ Spy Cas9-NLS S. pyogenes, NEB Cat# M0646T) was incubated for 10 min at room temperature. For mutagenesis, 1 nl of the sgRNA-Cas9 complex was injected into one-cell stage embryos, that were subsequently raised until 48 hpf. Bulk genomic DNA (gDNA) was extracted from 60 injected and 60 non-injected control embryos. gDNA extraction proceeded as follows: pooled samples were incubated for 3 hrs at 55°C in lysis buffer (100mM NaCl, 20mM Tris-HCl pH 8, 25mM EDTA, 0.5% SDS and 2 μg/μl proteinase K), purified using phenol:chloroform. Precipitation was done in 0.1 volume of 3M NaAc and 2.5 volume of 100% ethanol for 20 min at -20°C. Samples were centrifuged for 30 min at 4°C before washing with 70% ethanol and resuspension in water. gDNA was used for PCR, using DNA polymerase Platinum™ SuperFi™ (Invitrogen Cat# 12351010) and the SM31-SM32bis couple of primers (TTTCACTTTCTCTCTGCCTCTTACA; AACTCTGTCCAGATGATTGAGGAGC) flanking the targeted region. PCR products were run on gel electrophoresis before gel extraction of bands of interest (with sizes corresponding to either a wild-type allele or a deleted allele) (Nucleospin gel and PCR clean-up, Macherey-Nagel Cat# 740609.50). Fragments were cloned into blunt-end TOPO vector (Invitrogen Cat# 45024), transformed into E. coli TOP10 competent cells and grown on selective medium. 12 colonies for each PCR product were sent for sequencing. Mutant sequences contained either a 7 nucleotides deletion, or a mix of 1,4 or 7 nucleotides deletions located at the same site (at the end of exon 38, see Figure 7 -figure supplement 4A for the 7 nucleotides deletion), attesting of the proper functioning of our guide.

A stable homogeneous mutant line was generated by injecting the sgRNA:Cas9 complex into one-cell stage Tg(kdrl:Gal4;UAS:RFP) embryos similarly to what was described above. 100 embryos were raised until adulthood. Adult founders were identified by crossing adult F0 with AB fishes, and single embryo genotyping was performed on their progeny (10 embryos per adult fish). Briefly, single embryos were incubated in 40µl 25mM NaOH 0.2mM EDTA and heated for 10 min at 95°C, before cooling for 10 min at 4°C and addition of 40µl 40mM Tris-HCl pH5 to stop the reaction. PCR was performed using 5µl of template and specific primers (couple SM31-SM34 –TTTCACTTTCTCTCTGCCTCTTACA; ATAAATGAAGCCCCACCTCCGTCC – for wild-type (WT) allele and couple SM31-SM40 – TTTCACTTTCTCTCTGCCTCTTACA; ATGAAGCCCCACCTCAGACGATTGGC– for mutant allele). Presence of WT or mutant allele for each single embryo was assessed by the presence/absence of amplification band on gel electrophoresis. Four F0 founders were isolated, that had a transmission rate in the germline ranging from 20% to 80%. Progeny of the F0 founders and subsequent generations were raised separately until adulthood, and genotyped. After complete anesthesia (in fish water supplemented with 160µg/ml tricaine methanesulfonate), a small tissue sample was collected from the caudal fin, placed into DNA extraction buffer (10mM Tris pH8.2, 10mM EDTA, 200mMNaCl, 0.5% SDS, 200µg/ml proteinase K) and incubated 3 hrs at 50°C. DNA precipitation and resuspension was performed as described above. 7µL gDNA was used as template for PCR amplification using the primer couple SM31-SM32bis. Sequenced PCR products showed that all founders harbored the same 7 nucleotides deletion (see Figure 7 -figure supplement 4A). All experiments relative to Figure 7 and Figure 8 were performed using the progenies of one heterozygous mutant F3 couple. For these experiments, genotyping on single embryos was performed on gDNA as mentioned above. Only WT homozygous and mutant homozygous embryos were kept for analysis.

Alternative Cpf1 approach

As an alternative approach to the CRISPR/Cas9 deletion, we attempted to generate a second mutant line, using to CRISPR/Cpf1(Cas12a) system to delete a 315 nucleotides region encompassing exon 38 (with specific guides located in the introns before and after exon 38). The sequence of guides that have been tested are: TATCACACACACATCACCTTCTA and TTTCTCAGCGCTCCTGACAGATG. However, due to the structure of the intronic regions (repeat rich) and the necessary presence of specific PAM sequence, we only generated off-target deletions of intronic regions, without deletions of exon 38, leading us to focus on the CRISPR/Cas9 approach.

Morpholinos and injections

Morpholinos were obtained from GeneTools (see Key Resources Table for sequences). The sih Tnnt2 translation start codon and flanking 5-prime sequence MO (Sehnert et al. 2002) as well as the ArhGEF11 exon 38 splice blocking MO (overlapping the exon 38/intron 38-39 boundary) were resuspended in ddH2O to obtain stock solutions at 2mM and 1mM, respectively. The sih MO (1.5ng) and ArhGEF11 exon 38 MO (3ng) were injected into 1-cell stage zebrafish embryos, after dilution in ddH2O. For sih morphants, embryos that were used in the experiments were checked for absence of heart beating 24 hrs after injection as well as before being used for dissection and RNA extractions at the 30-32 and 48-50 hpf time-windows. For measuring the efficiency of the ArhGEF11 exon 38 MO, total RNA was extracted from pools of injected and control embryos followed by mRNA reverse transcription. PCR on cDNA flanking the exon 38 was performed using DNA polymerase Platinum™ SuperFi™ (Invitrogen Cat# 12351010) and analyzed on gel electrophoresis before extraction, cloning and sequencing to verify the absence of the DNA sequence encoded by exon 38 (see Figure 7 -figure supplement 4A).

Whole mount in situ hybridization

For the sequence of all the primers used to amplify probes, see the Key Resources Table. The probes for WISH concern Pard3ab, Pard3bb (Figure 3) and PDZ-RhoGEFs (Figure 7 – figure supplement 2).

Whole-mount chromogenic in situ hybridization was performed as described in Lancino et al 2018. Probes were synthetized using the manufacturer recommendation (T7 RNA polymerase, Promega, Cat# P2075, DIG-nucleotides, Jena Bioscence, Cat# NU-803-DIGX).

For Pard3ab (GenBank: AY524776.2), a probe of 717bp (nucleotides 1258-1974) encompassing the nucleotide sequence starting in the 3-prime of the sequence encoding for PDZ1 and ending between PDZ2 and PDZ3 was amplified with the couple of primers pard3ab_WISH_Nter_fw and pard3ab_WISH_Nter_T7_rev. For Pard3bb (NCBI Reference sequence: XM_021478865.1), a probe of 747bp (nucleotides 468-1214) encompassing the nucleotide sequence starting in the 5-prime of the sequence encoding for the N-ter of the protein and ending at the 3-prime of PDZ1 was amplified with the couple of primers pard3bb_WISH_Nter_fw and pard3bb_WISH_Nter_T7_rev (for the position of the primers respectively to the 2 pard3 proteins sequences and domains, see Figure 3B).

Images were captured with the Zeiss Axio ZOOM V16 microscope with the Zen Pro2 software, with a brightfield transmitted light optics. Post-processing steps were performed using the Extended-Depth Focus method to combine in focus regions from multiple z-planes and convert into in a transmitted light z-stack to generate a unique in-focus image.

Immunofluorescence detection of the HA-tagged dt-runx1 deletion mutant

Whole mount immunostaining was performed on hand-dechorionated 48 hpf Tg(dt-runx1) embryos to assess the localization of the 2XHA-tagged dt-runx1 protein in aortic cells. Briefly, embryos were fixed in 4% methanol free PFA (Polysciences Cat# 040181) for 3 hrs at room temperature (RT) and washed in PBS/0.1% tween20 (PBT). Embryos were treated in successive baths of milliQ water supplemented with 0.1% tween20, then in cold acetone for 10 min at -20°C and again in milliQ water plus 0.1% tween20. They were then washed in 1X HBSS (Invitrogen Cat# 14025), and permeabilized for 45 min at RT in 1X HBSS, 0.1% tween20, 5mM CaCl2 and 0.1mg/ml collagenase (Sigma-Aldrich Cat# C9891). Embryos were then rinsed in PBSDT and incubated in blocking solution 1× sheep serum (SS, Sigma Cat# S2263) for at least 4 hrs at RT. They were then incubated overnight at 4°C with primary antibodies in blocking solution: rabbit anti-GFP (MBL Cat# 598; 1/300) and mouse anti-HA (Sigma, Cat# 12ca5; 1/50). On the next day, embryos were washed several times in PBSDT and endogenous peroxidase activity was inactivated by treatment with 6%H2O2 (Sigma-Aldrich, Cat# H1009) for 1 hr at 4°C. Embryos were then incubated in the NGS for at least 4 hrs and incubated overnight at 4°C with secondary antibodies: goat anti-rabbit Alexa Fluor 488 (Invitrogen Cat# A11070; 1/400) and goat anti-mouse HRP-conjugated (Thermo Fisher Scientific Cat# G-21040; 1/300). Finally, embryos were rinsed several times first in PBSDT and after in PBT before HRP fluorescent revelation. Embryos were incubated for 45 min at RT in the dark in imidazole buffer (0.1 M imidazole (Sigma-Aldrich, Cat# I5513) in PBT supplemented with 1% H2O2) with Cy3 Tyramide Reagent (Cy3 NHS sigma Cat# PA13101, tyramide sigma Cat# T-2879, dimethyl formamide sigma Cat# T-8654 and triethylamine sigma Cat# T-0886). Final washes in PBT and in 6% H2O2 for POD inactivation were performed before embryos were mounted in low-melting agarose in 1X PBT for fluorescence confocal imaging.

Quantitative real-time PCR

Preparation of samples

To determine the impact of runx1 interference and blood flow on pard3 family genes and splice variants expression, total RNA from dissected trunk regions was extracted (as described previously) from 30-32 hpf and 48-50 hpf embryos from either incross of the Tg(kdrl:dt-runx1(eGFP)) fish line, outcross of the Tg(kdrl:dt-runx1(eGFP)) and Tg(kdrl:gal4;UAS:RFP) fish lines - with eGFP positive embryos as the interfering condition and eGFP negative embryos as sibling controls -, or Tg(kdrl:gal4;UAS:RFP) embryos injected at the single cell stage with sih morpholino. Analysis was carried on 3 biological replicates.

Technical procedure

For the sequence of all the designed qRT-PCR primers, see the Key Resources Table.

qRT-PCR was performed using Takyon Rox SYBR 2x Master mix blue dTTP kit (Eurogentec Cat# UF-RSMT-B0701), with starting concentrations of template of around 10ng/µl and primer concentrations of 0.5µM, on a QuantStudio3 system (Applied Biosystems). Each reaction was performed in technical triplicates and in at least 3 biological replicates. Ct was determined automatically by the QUANTSTUDIO™ DESIGN & ANALYSIS 2 software on the Thermofisher cloud and exported to be analyzed manually. Standard deviation was calculated within the technical triplicates, and when superior to 0.3, obvious outliers were removed (Dixon’s Q test). The delta-delta-Ct method (double normalization to endogenous control — zebrafish elongation factor ef1α — and to control sample) was used to compare gene expression in control and altered conditions.

In vivo confocal imaging

Embryos, dechorionated manually, were anesthetized using tricaine (Sigma Aldrich, Cat# A5040). They were then embedded on the side position in 1% low melting agarose (Promega, Cat# V2111) in a glass bottom 60μ-Dish (35mm high; Ibidi, Cat# 81156). To avoid movements and pigmentation during image acquisitions, 1x tricaine /1x PTU Volvic water was incorporated to the low melting agarose and, after solidification, 1ml of 1x tricaine /1x PTU Volvic water was added before sealing the dish.

Embryos were imaged with 2 confocal microscope systems. For the results presented Figure 1, z-stacks and time-lapse sequences were obtained using a Leica TCS SP8 inverted confocal microscope as described in Lancino et al 2018.

For all the other confocal microscopy results, embryos were imaged using an Andor (Oxford Instruments) spinning disk confocal system (CSU-W1 Dual camera with 50µm Disk pattern and single laser input (445/488/561/642nm), LD Quad 405/488/561/640 and Tripl 445/561/640 dichroic mirrors), equipped with a Leica DMi8 fluorescence inverted microscope and CMOS cameras (Orca Flash 4.0 V2+ (Hamamatsu)). Imaging was performed using a 40x water immersion objective (HC PL APO 40x/1.10 Water CORR CS2), a LEDs light source (CoolLED pE-4000 16 wavelength LED fluorescence system), and the acquisitions were piloted with the support of the MetaMorph software.

Morphological and morphometric analysis of aortic and hemogenic cells

For the morphometric analysis of the aorta and aortic cells, large z-stack of 48-55 hpf Tg(kdrl:eGFP-JAM3b; kdrl:nls-mKate2) or Tg(kdrl:eGFP-JAM2a; kdrl:nls-mKate2) encompassing the whole aortic depth were acquired with optimal z-resolution (0.3µm z-step). For each embryo, 3 contiguous z-stacks of 330µm width were acquired, allowing us to image the entirety of the AGM region, from the anterior to the posterior end of the elongated yolk.

Samples

For ArhGEF11 splicing morpholino experiment, 3 splicing morpholino injected embryos and 2 control non injected siblings were kept for the analysis.

For ArhGEF11 CRISPR Cter deletion and when performing the experiments, imaging was done, for putative mutants, on embryos that had shown a delay in initiation of circulation at 24 hpf, regained circulation at 48 hpf albeit with a clearly visible oedema in the cardiac region (non-altered embryos were kept as sibling controls). All imaged embryos whose signal allowed us to generate segmentable 2D-cartography from the whole length of the aorta were kept and processed further for genotyping. Genotyping confirmed that the circulation problems at 24 hpf and the remaining oedema at 48 hpf corresponded to embryos bearing the mutation.

For CRISPR mutant phenotype analyses, 2 mutant embryos and 2 wild type siblings were kept, after genotyping.

Image analysis

The following image analysis was performed on Icy software.

For each segment, we generated a 2D-cartography of the eGFP-Jam signals at intercellular contacts of aortic cells, using the Tubeskinner plugin, according to the protocol described in Lancino et al 2018. Briefly, the aorta was rotated to face the lumen, a circular ROI fitting the interior of the aorta was initialized by the experimenter on the first plane, and for each plane sequentially, the algorithm will more accurately fit the ROI to the aortic perimeter and collect the fluorescence intensity all along its longitudinal axis. The collected intensity was then projected on a 2D-plane, where the Y-axis of the 2D-cartography corresponds to the perimeter of the aorta, and the X-axis corresponds to the X-axis of the original z-stack.

Before further processing, a metadata table recapitulating the dates and conditions of experiments for each embryo was generated, and a random label was assigned to rename each acquisition. This allowed for the later part of the analysis, that involves semi-automated segmentation and classification by the user, to be blind and avoid potential bias.

Each 2D-cartography was then manually pre-segmented using the ROI Polygon tool to draw the contour of each aortic cell, using the original 3D z-stack as reference. Due to technical limitation (decreased fluorescence signal collection on the farthest lateral side of the aorta), only about 2/3 of the 2D-cartographies could be segmented (comprising the floor, the right lateral side of the aorta and the roof). The manual segmentation was then improved using the “Active Contour” plugin, that improved the fit of the cellular contour to the fluorescence signal and slightly dilated the cell contours until the ROIs formed a joint paving. All morphometric descriptors were extracted for each cell. The number of neighbours was automatically generated for each cell by counting the number of touching adjacent ROIs.

Additionally, cells were manually classified into cell types according to morphometric parameters using as reference the original 3D z-stack and in particular the nls-mKate2 nuclear signal. EHT-undergoing cells were positioned on the floor of the aorta, have reduce antero-posterior axis length, luminal membrane inward bending for EHT pol+ cells and luminal membrane outward bending for EHT pol-cells, and a thickened nucleus. Hemogenic cells whose criteria were established in Lancino et al 2018 (elongated in the antero-posterior axis, with lateral membranes not trespassing the equatorial plane of the aortic wall), are positioned on the floor of the aorta, systematically have a reduced width perpendicular to the blood flow and, importantly, a thickened nucleus that protrudes out of the aortic plane. Finally, endothelial cells have a flattened nucleus positioned high on the lateral side or on the roof of the aorta.

Measurement of the aortic perimeter

An accurate measurement of the aortic perimeter was obtained using the z-stack generated for the aortic cells’ morphometric analysis. Briefly, the z-stack was rotated to face the lumen of the aorta and a ROI fitting the aortic perimeter was drawn by the user and optimized by the “Cell Contour” plugin. The length of the perimeter was then extracted from the ROI statistics.

Fluorescence Recovery After Photobleaching (FRAP) measurements and analysis

Measurement of junctional turn-over was performed on 48-55 hpf Tg(kdrl:eGFP-JAM3b; kdrl:nls-mKate2) embryos using a Nikon Ti2e spinning disk microscope equipped with a sCMOS camera (Photometrics, Prime 95B, pixel size 11µm, 1200×1200 pixels, QE 95%) and a 40x water objective (Numerical Aperture 1.15, Working Distance 0.6mm, xy pixel size of 0.27µm). Embryos were mounted as previously described.

Samples

For ArhGEF11 splicing morpholino phenotypic analysis, 10 different embryos (injected with the splicing morpholino) obtained from 3 separate experiments and 30 control embryos (non-injected siblings or control morpholino injected siblings) obtained from 6 separate experiments were analyzed.

For EHT analysis, 16 embryos over 4 separate experiments were analyzed.

For ArhGEF11 CRISPR/Cas9 Cter deletion phenotypic analysis, 8 mutant embryos and 12 wild type siblings over 3 separate experiments were analyzed.

Overall, the exact number of interfaces analyzed is written on each figure panel.

Acquisitions

For each acquisition, a wide two channels z-stack of the aorta was generated to capture the whole aortic volume and generate a 2D-cartography of the aortic environment (Icy, plugin Tubeskinner). This 2D-cartography of the cellular contours and the z-stack with the cellular contours and nuclei were used jointly to identify un-ambiguously cellular and junctional types and define one or multiple single point regions of interest (ROIs) for subsequent bleaching and fluorescence recovery measurement.

The sequence of acquisitions before/after bleach focused only on the eGFP-Jam3b signal to speed up the acquisition and improve the temporal resolution. The 5 steps of the acquisition can be recapitulated as follows and we obtained/performed: (1) a single z-stack of 20µm centered on the plane of focus, (2) a 30 sec time-lapse acquisition on the plane of focus to measure average pre-bleach intensity with an image every second, (3) a high intensity stimulation of the ROIs (25% 300ms 488 FRAP laser exposure, leading to the bleaching of the Jam3b-eGFP signal), (4) a 3 min time-lapse acquisition on the plane of focus to measure fast post-bleach recovery with an image every second, and (5) a 7 min time lapse acquisition of 8µm z-stack centered around focus plane with a 20 sec time interval to measure slow post-bleach recovery. Each step was save separately as .nd2 files.

Signal quantification and recovery analysis

File were converted from .nd2 to .tiff using FIJI batch converter, scale factor 1.00, bilinear interpolation, no average when downsizing, images read using bioformat.

To collect fluorescence intensity, ROIs for each stimulation point was manually drawn using Fiji and named with an individual label using Fiji ROI manager. A metadata table was manually curated with all information for each ROI (junction type, condition, date of experiment, embryo number). An additional ROI was created to collect background intensity for subsequent corrections. Acquisitions with clear embryo movement or drifting were either completely removed from the dataset (∼5% of the acquisitions) or only kept for early recovery analysis (see below) if the drift only occurred in the last step (fifth) of the acquisition.

Fluorescence intensity for each ROI was collected and saved automatically as a .csv file for each acquisition using a homemade Fiji macro.

All further analysis were performed using R for automatization.

Briefly, for each ROI: the fluorescent signal was normalized to the background signal (I_{background normalized} = I_roi – I_background), then normalized to mean intensity pre-bleach (I_normalized= I_{background normalized ROI post-bleach} / I_{background normalized ROI pre-bleach}), and finally scaled (I_{full scaled normalized} = (I_normalized – min (I_normalized))/ (1– min (I_normalized)).

A single exponential equation was used to fit the data using the nls:multistart package for parameter optimization:

With A the maximum amplitude of the curve, t the time, and τ a constant defining the growth rate of the curve.

The fitted curve allowed to extract quantitative information describing the recovery of each type of junction in different conditions: the mobile fraction of the bleached pool (corresponding to A, the amplitude of recovery) and the time of half recovery, related to the speed of recovery, that can be extracted from τ:

A more precise description of the recovery speed was extracted from the observation of the early time period (until 30s post bleach), were the fluorescence intensity increases uniformly. A first-degree polynomial equation was fitted to the data (using lm() function from the stats package), and the coefficient corresponding to the slope of the fitted curve was used as a metric for speed of fluorescence recovery.

Statistical analysis

Statistical tests used are described for each figure panel in the legend. Since our samples are usually small, the test used is an unpaired two samples non-parametrical Wilcoxon mean comparison test, without a priori knowledge on sample normal distribution or homoscedasticity.

Data and software availability

All graphics and plots were generated using RStudio and packages including: ggplot2 (Wickham 2016), readr (Wickham, Hester, and Bryan 2022), dplyr (Wickham et al. 2023), stringr (Wickham 2022), ggstatsplot (Patil 2022), ggpubr (Kassambara 2020), wesanderson (Ram and Wickham 2018), cowplot(Wilke 2020), ggsci (Xiao 2018), ggbeeswarm (Clarke and Sherrill-Mix 2017), viridis (Garnier 2021), DT (Xie, Cheng, and Tan 2022), SciViews (Grosjean 2019), nls.multstart (Padfield and Matheson 2020), matrixStats (Bengtsson 2022).

Acknowledgements

We wish to warmly thank Philippe Herbomel for his encouragements and for ensuring the funding of this study.

We thank our fish facility members Y. Rolin and K. Sebastien for their daily help, R. Patient (University of Oxford) for the Hsp70 promoter. This work was supported by Institut Pasteur, CNRS, grants to P. Herbomel from the Fondation pour la Recherche Médicale (#DEQ20160334881), the Fondation ARC pour la Recherche sur le Cancer, and the Agence Nationale de la Recherche Laboratoire d’Excellence Revive (Investissement d’Avenir, ANR-10-LABX-73), and a grant to AS from the Institut Pasteur (PTR 439-2021). LT was recipient of a PhD fellowship from the Collège Doctoral, Sorbonne Université, and from the Labex Revive (ANR-10-LABX-73).

Authors contributions

LT designed, performed and analyzed experiments, and wrote the Method section and the Figure legends. SM designed and conducted experiments. CV conducted experiments. AS conceived the project, designed, performed and analyzed experiments, wrote the manuscript with input from LT, and obtained funding.

Declaration of interests

The authors declare no competing interests.

Legends to supplement figures

Figure 1 – figure supplement 1.

Figure 2 – figure supplement 1.

Figure 2 – figure supplement 2.

Figure 3 – figure supplement 1.

Figure 3 – figure supplement 2.

Figure 3 – figure supplement 3.

Figure 5 – figure supplement 1.

Figure 5 – figure supplement 2.

Figure 6 – figure supplement 1.

Figure 7 – figure supplement 1.

Figure 7 – figure supplement 2.

Figure 7 – figure supplement 3.

Figure 7 – figure supplement 4.

Figure 7 – figure supplement 5.

Figure 7 – figure supplement 6

Supplementary files

Supplementary file 1: Materials and Methods - Supplementary Table

Figure 1 – video 1. EHT pol+ cells at early timing. Z-stack (37 planes) corresponding to the time point t = 0 min whose plane 21 is shown Figure 1C.

Figure 1 – video 2. EHT pol+ cells at late timing. Z-stack (37 planes) corresponding to the time point t = 80 min whose planes 16 (for eht cell 1) and 22 (for eht cell 2) are shown Figure 1 – figure supplement 1B.

Figure 1 – video 3. Emergence of EHT pol-cells. Time-lapse sequence using a Tg(Kdrl:Gal4;UAS:RFP;4xNR:eGFP-Podxl2) embryo, obtained with spinning disk confocal microscopy (38 timing points, each acquired with 10 min intervals, starting at 55 hpf) and from which the images of the 2 emerging EHT pol-cells shown Figure 1D were extracted.