Introduction

The blood vascular system is a dynamic, multicellular tissue that adapts to the metabolic demands of development, growth and homeostasis by altering its pattern and morphology. Macroscopic transformation in the vascular network is achieved by microscopic changes in endothelial cell (EC) behaviours. For example, vascular expansion through sprouting angiogenesis requires the coordination of EC migration, proliferation, anastomosis, lumen formation and cell rearrangements, all of which require specific cell shape changes. At the initial phase of sprouting angiogenesis, endothelial tip cells generate numerous membrane protrusions such as filopodia and lamellipodia that drive migration and anastomosis (Figueiredo et al., 2021; Gerhardt et al., 2003; Phng et al., 2013; Sauteur et al., 2017). During tubulogenesis, apical membranes deform by generating inverse blebs to expand lumens so that blood can be transported efficiently through the blood vascular network (Gebala et al., 2016). However, following vessel perfusion, ECs become less dynamic and instead develop resistance to the deforming forces of blood pressure to maintain vessel morphology (Kondrychyn et al., 2020). EC shape is therefore dynamic and changes depending on the morphogenetic state of vascular development.

It is well established that localized remodelling of actin cytoskeleton and non-myosin II contractility are instrumental in driving cell shape changes during tissue morphogenesis (Clarke and Martin, 2021; Munjal and Lecuit, 2014; Murrell et al., 2015). More recently, there is an accumulating body of work implicating a role of water flow and the resulting changes in hydrostatic pressure as another mechanical mechanism of cell shape change (Choudhury et al., 2022; Chugh et al., 2022; Li et al., 2020). For example, in the osmotic engine model of cell migration, water inflow at the leading edge and outflow at the rear leads to forward translocation of tumour cells in confined microenvironment (Stroka et al., 2014). Additionally, hydrostatic pressure from the extracellular environment can drive tissue morphogenesis, as demonstrated during mouse blastocyst formation (Chan et al., 2019; Dumortier et al., 2019), development of the otic vesicle in the zebrafish (Mosaliganti et al., 2019) and blood vessel lumen expansion (Gebala et al., 2016). Hydrostatic pressure can therefore generate sufficient forces to shape tissues and organs to their proper form and size during development.

We have previously discovered that EC migration persists after the inhibition of actin polymerization and in the absence of filopodia to generate new blood vessels in the zebrafish (Phng et al., 2013). This observation suggests the existence of an alternative mechanism of migration independent of actin polymerization. In this study, we sought to determine whether hydrostatic pressure regulates EC migration during sprouting angiogenesis in the zebrafish by investigating the function of Aquaporins (Aqp), which are transmembrane water channels that increase water permeability of cell membranes to promote transcellular water flow (Day et al., 2014; Farinas et al., 1997; Kozono et al., 2002; Preston et al., 1992). We discovered that ECs of newly formed intersegmental vessels (ISV) express aqp1a.1 and aqp8a.1 mRNA, and observed the enrichment of Aqp1a.1 and Aqp8a.1 proteins in the leading edge of migrating tip cells. Detailed single cell analyses showed that endothelial tip cells lacking aqp1a.1 and aqp8a.1 expression have reduced cell volume, membrane protrusions and migration capacity. As a result, there is defective sprouting angiogenesis and ISV formation in aqp1a.1;aqp8a.1 double mutant zebrafish. Notably, when actin polymerization is inhibited in aqp1a.1;aqp8a.1 double mutant zebrafish, there is a greater impairment in EC migration and sprouting angiogenesis, demonstrating the additive function of actin polymerization and hydrostatic pressure (that is created by water inflow) in generating membrane protrusions to drive EC migration in vivo.

Results

aqp1a.1 and aqp8a.1 are differentially expressed by tip and stalk cells during sprouting angiogenesis

Of the 18 number of aquaporin genes that are expressed in zebrafish (Tingaud-Sequeira et al., 2010), aqp1a.1 and aqp8a.1 mRNA have been detected in blood endothelial cells (ECs) (Koun et al., 2016; Rehn et al., 2011). We confirmed the endothelial expression of aqp1a.1 and aqp8a.1 at 30 hours post fertilization (hpf) and 2-, 3- and 4-days post fertilization (dpf) by whole-mount in situ hybridization (Fig. S1). The two endothelial aqp genes, however, differ in spatial expression. While aqp1a.1 is widely expressed in blood vessels of the head and trunk (dorsal aorta (DA), caudal artery (CA), ISVs, dorsal longitudinal anastomotic vessels (DLAV) and caudal vein plexus) at 30 hpf, aqp8a.1 expression is absent in cerebral blood vessels and is restricted to the DA, CA and the ventral regions of ISVs in the trunk. After 1 dpf, aqp8a.1 expression expands to the entire ISV and DLAV. Additionally, the expression of aqp1a.1 and aqp8a.1 gradually decreases in the DA and CA so that both are absent by 4 dpf. Single cell RNA sequencing (scRNAseq) of ECs isolated at 34 hpf and 3 dpf further confirmed the endothelial expression of aqp1a.1 and aqp8a.1 mRNA (Fig. S2 A - D), revealed differential expression of aqp1a.1 and aqp8a.1 mRNA in distinct endothelial clusters and highlighted higher aqp1a.1 transcript expression in all endothelial subtypes compared to aqp8a.1 (Fig. S2 E - H).

We next examined the expression pattern of aqp1a.1 and aqp8a.1 at the beginning of sprouting angiogenesis at higher resolution using RNAscope and confocal microscopy. At 20 hpf, aqp1a.1 mRNA is highly expressed in the DA with little expression in the PCV (Fig. 1A, C and Fig. S3). Closer inspection uncovered heterogeneous expression in the DA, with higher expression in ECs that would be specified as tip cells (Fig. 1A, C and Fig. S3B). Indeed, tip cells of newly formed vascular sprouts express higher levels of aqp1a.1 compared to adjacent ECs in the DA at 20 hpf (Fig. 1A – C, Fig. S3B) and the majority of ISVs (61%) display higher aqp1a.1 expression in tip cells than in trailing stalk cells at 22 hpf (Fig. 1G). Although we previously observed aqp8a.1 mRNA in the DA at 30 hpf (Fig. S1), its expression here is largely absent at 20 hpf (Fig. 1B, D and Fig. S3C). At 22hpf, aqp8a.1 mRNA expression in the DA increases (Fig. 1E and F). Interestingly, unlike aqp1a.1, in 78% of ISVs analyzed, aqp8a.1 mRNA is more highly expressed in stalk cells than tip cells (Fig. 1G). These observations highlight differential expression pattern of aqp1a.1 and aqp8a.1 in newly formed vascular sprouts, with aqp1a.1 expression enriched in tip cells and aqp8a.1 expression in stalk cells.

Figure 1.

VEGF-VEGFR2 signalling induces the expression of aqp1a.1 and aqp8a.1 mRNA

As aqp1a.1 and aqp8a.1 mRNA display differential tip-stalk cell expression, we explored whether their expression is regulated by the VEGFR2-Notch signalling axis, which controls endothelial tip-stalk cell specification during sprouting angiogenesis (Benedito and Hellström, 2013; Phng and Gerhardt, 2009; Siekmann et al., 2013). To inhibit VEGFR2 activity, embryos were treated with 1 µM ki8751 from 20 hpf. Relative quantitative polymerase chain reaction (qPCR) experiments showed that aqp1a.1 mRNA expression was suppressed by 52% after 6 hours (p < 0.0001) and by 40% after 24 hours (p < 0.0001) of ki8751 treatment when compared to DMSO-treated embryos (Fig. 1H). aqp8a.1 mRNA expression was also decreased by the inhibition of VEGFR2 activity, and its suppression was accentuated after 24 hours treatment (93% decrease, p < 0.0001) compared to after 6 hours (17% decrease, p = 0.0028). These results demonstrate that VEGFR2 activity upregulates the expression of aqp1a.1 and aqp8a.1, but the temporal dynamics is different for aqp1a.1 and aqp8a.1. To examine whether VEGFR2 activity also regulates Aquaporin expression in human ECs, we treated cultured human aortic endothelial cells (HAECs) with ki8751 for 6 hours and analyzed AQP1 expression by qPCR. AQP8 expression was not examined since it is not expressed in HAECs (data not shown). Inhibition of VEGFR2 activity with 1 nM and 10 nM ki8751 decreased AQP1 mRNA expression by 31% (p = 0.0023, Fig. 1I) and 40% (p = 0.1032, Fig. 1I), respectively, demonstrating cross-species conservation of VEGFR2 activity in inducing Aqp1 mRNA expression.

Notch signalling was inhibited using the gamma-secretase inhibitor, Dibenzazepine (DBZ), which prevents proteolytic cleavage and consequently, activation of the Notch receptor (Groth et al., 2010). Treatment of embryos with DBZ from 20 hpf for 6 and 24 hours significantly decreased the expression of hey1 and hey2, canonical targets of Notch signalling (Fig. 1J), confirming the effectiveness of the drug treatment in our experiments. When compared to DMSO-treated embryos, 6-hour treatment with DBZ did not cause a significant change in aqp1a.1 and aqp8a.1 expression levels (Fig. 1J). Longer treatment with DBZ for 24 hours, however, increased aqp8a.1 expression by 98% (p < 0.0001) while aqp1a.1 expression was relatively unchanged (Fig. 1J). These findings suggest that Notch signalling inhibits aqp8a.1, but not aqp1.a, expression.

Aqp1a.1 and Aqp8a.1 proteins are enriched at the leading front of migrating tip cells

To gain a better understanding of where within the cell Aquaporin protein functions, we generated stable Tg(fli1ep:aqp1a.1-mEmerald)^rk30 and Tg(fli1ep:aqp8a.1-mEmerald)^rk31zebrafish lines in which ECs express Aqp1a.1 or Aqp8a.1 tagged with mEmerald, respectively. At 25 hpf, we observed an accumulation of Aqp1a.1 (Fig. 2A) and Aqp8a.1 (Fig. 2B) in mCherryCAAX-positive membrane clusters in the cytoplasm and filopodia of tip cells. Interestingly, many of the Aqp1a.1- or Aqp8a.1-positive membrane clusters are located at the migrating front of the tip cell. Time-lapse imaging capturing the process of ISV formation between 24 to 30 hpf further showed sustained enrichment of Aqp1a.1 and Aqp8a.1 in clusters at the front of the tip cell as well as in filopodia (Fig. 2C and D, Movie S1 and S2), suggesting that Aqp1a.1 and Aqp8a.1 may promote localized water flux at the migrating edge of tip cells.

Figure 2.

Loss of Aqp1a.1 and Aqp8a.1 function leads to defective trunk blood vessel formation

We subsequently proceeded to generate zebrafish aqp1a.1 and aqp8a.1 mutants to investigate the function of Aquaporin-mediated water flow in blood vessel development. Using the CRISPR/Cas9 genome editing technique, we targeted the first exons of aqp1a.1 and aqp8a.1 genes. The CRISPR-generating allele aqp1a.1^rk28 carries a 14-bp deletion and an 8-bp insertion in the 5’ end of the gene leading to a premature termination codon at amino acid 76 after 1 missense amino acid (Fig. S4A-B). The aqp8a.1^rk29 allele carries a 4-bp deletion in the 5’ end of the gene that leads to a frameshift after T15 and premature termination codon at amino acid 73 after 58 missense amino acids (Fig. S5A-B). Premature termination codons are often more deleterious than missense mutations because they result in the loss of protein expression. Nonsense-mediated mRNA decay is a quality-control mechanism that selectively degrades mRNAs harboring premature termination codons (Chang et al., 2007). Using whole-mount in situ hybridization and qPCR analysis, we demonstrated that aqp1a.1 and aqp8a.1 mRNA expression is significantly decreased in aqp1a.1^rk28/rk28 (Fig. S4C-D) and aqp8a.1^rk29/rk29(Fig. S5C-D) zebrafish embryos, respectively, suggesting a rapid degradation of mutant mRNAs and as a result, the loss of Aquaporin protein expression.

Although aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos look morphologically normal when compared to wildtype or aqp1a.1^+/rk28;aqp8a.1^+/rk29 embryos (Fig. S8A, B, C and D), we observed defects in blood vessel formation and patterning at 1 to 3 dpf. Compared to 28 hpf control embryos whose ISVs have reached the dorsal roof of the neural tube and started to form the DLAV, ISVs of aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos are much less developed (Fig. 3A and D) with significantly reduced length (Fig. 3E and Fig. S6H and K). Single deletion of aqp1a.1 (Fig. 3B and Fig. S6F and I) or aqp8a.1 (Fig. 3C and Fig. S6G and J) also led to defective ISV formation, although the defects are less severe compared to the loss of both aqp1a.1 and aqp8a.1 expression. Comparison of ISV length in aqp1a.1^rk28/rk28, aqp8a.1^rk29/rk29 and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos shows that the loss of aqp1a.1 leads to a greater decrease in ISV length than the loss of aqp8a.1 (Fig. 3E, Fig. S6I and J), and that the loss of both aqp1a.1 and aqp8a.1 resulted in the greatest reduction. These observations indicate an additive effect of Aqp1a.1 and Aqp8a.1 function in blood vessel development and implicate Aqp1a.1 as the more dominant Aquaporin protein in EC function, in line with the higher expression of aqp1a.1 mRNA in tip cells.

Figure 3.

We next examined the embryos at 3 dpf, a stage when arterial ISVs (aISVs) and venous (vISVs) are fully established and lumenized, connecting the DLAV to either the DA or PCV (Fig. 3F and G), respectively. Such fully established ISVs are significantly reduced in embryos when aqp1a.1, aqp8a.1 or both aqp1a.1 and aqp8a.1 are deleted. Instead, these embryos display an increased number of aISVs and vISVs that fail to establish a connection to the DLAV, DA or PCV (Fig. 3H – J). Quantification shows that while 19.1% of wildtype embryos display at least 1 incomplete ISV per embryo, this increased to 65.8% in aqp8a.1^rk29/rk29embryos, 83.3% in aqp1a.1^rk28/rk28 embryos and 92.2% in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29 embryos (Fig. 3K). The number of incomplete ISVs per embryo in aquaporin mutants was variable, between 1 to 6, 1 to 16 and 1 to 22 incomplete ISVs in aqp8a.1^rk29/rk29, aqp1a.1^rk28/rk28and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29 embryos, respectively (Fig. S6A-D). We also found that incomplete ISVs appear with higher frequency in the zebrafish trunk (ISV number 6 to 15) rather than in the tail (ISV number 16 to 27, Fig. S6E). Detailed analysis revealed that most of the incomplete ISVs are vISVs that are connected to the PCV but not the DLAV (Class I, 75.2%), followed by the absence of aISV or vISV (Class II, 15.8%, Fig. 3L and Fig. S7A and B). A small fraction of the incompletely formed vessels are aISVs that are not connected to the DA (Class IV, 2.2%, Fig. S7D) or DLAV (Class V, 2.9%, Fig. S7E), or vISVs that are not connected to the DLAV (Class VII, 2.68%, Fig. S7F). At some regions, there is an aISV that is not connected to the DA and a vISV that is not connected to the DLAV (Class III, 1.2%, Fig. S7C).

As aqp1a.1 is expressed in brain vasculature during development (Fig. S1), we examined whether cerebral vasculature is also affected in aqp1a.1^rk28/rk28 embryos. At 4 dpf, we found that the length of the dorsal longitudinal vein (DLV), a main blood vessel that supplies blood to the choroid plexus (Bill and Korzh, 2014) was reduced in aqp1a.1^rk28/rk28 embryos (p = 0.0098, Fig. S9B). Also, in 2 out of 21 aqp1a.1^rk28/rk28embryos, the DLV was undeveloped (Fig. S9A and C), a defect that is also observed in 50% of vegfab^-/- and vegfc^-/- mutant zebrafish (Parab et al., 2021).

In summary, our analyses suggest that in the absence of Aqp1a.1 and/or Aqp8a.1 function, there is an initial delay in ISV formation during primary sprouting angiogenesis. By 3 dpf, some ISVs are completely formed (Fig. S8K and M), connecting the DA or PCV to the DLAV. However, the number of completely formed ISVs is significantly decreased, with the loss of Aqp1a.1 having a more severe impact compared to the loss of Aqp8a.1, and the combined loss of Aqp1a.1 or Aqp8a.1 having a greater deleterious effect.

Aquaporins promote endothelial tip cell protrusion and migration

To gain a better understanding of how incomplete ISV formation arise, we performed timelapse imaging in zebrafish depleted of both Aqp1a.1 and Aqp8a.1 function from 21 – 22hpf, when sprouting angiogenesis begins. In aqp1a.1^+/rk28;aqp8a.1^+/rk29 embryos, tip cells emerge from the DA between 20 and 24 hpf (Fig. 4A) and migrate dorsally between the somite boundaries over the notochord and neural tube to form primary ISVs (Fig. 4B, Movie S3). However, in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos, tip cell nuclei emerge between 22 and 28 hpf (Fig. 4A). Furthermore, in some aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29 embryos, tip cells retract whereby membrane protrusions do not extend dorsally in the direction of migration but withdraw into the DA so that a primary ISV does not form at this location (Fig. 4C, Movie S4). These observations indicate that tip cell protrusion from the DA is significantly delayed in the absence of Aqp1a.1 and Aqp8a.1 function.

Figure 4.

We have previously demonstrated that filopodia facilitate EC migration by serving as a template for lamellipodia formation (Phng et al., 2013). Lamellipodia frequently emanate laterally from stable filopodia at the leading edge of tip cells, leading to rapid expansion of the filopodium, increase in volume and the formation of a stable protrusion in the direction of migration (Phng et al., 2013). This was similarly observed in aqp1a.1^+/rk28;aqp8a.1^+/rk29 embryos (Fig. 4D, Movie S5) but not in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos (Fig. 4E, Movie S6). Comparison of filopodia dynamics in wildtype (Fig. 4F), aqp1a.1^+/rk28;aqp8a.1^+/rk29 and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29(Fig. 4G) embryos revealed a significant reduction in filopodia number (wildtype, 18.75 ± 5.1/100µm; aqp1a.1^+/rk28;aqp8a.1^+/rk29, 19.8 ± 6.3/100µm; aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29, 9.9 ± 5.2/100µm; Fig. 4H) and length (wildtype, 5.9 ± 3.2 µm; aqp1a.1^+/rk28;aqp8a.1^+/rk29, 6.2 ± 3.3 µm; aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29, 4.8 ± 2.4 µm; Fig. 4I) in ECs lacking both Aqp1a.1 and Aqp8a.1. function. Analysis of time-lapse images of tip cell behaviour further showed slower tip cell membrane expansion at the leading edge (Fig. 4J), decreased nuclei displacement (Fig. 4K) and a significant reduction in tip cell migration velocity (Fig. 4L) in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos.

In conclusion, Aqp1a.1 and Aqp8a.1 are required for sprouting angiogenesis by promoting endothelial tip cell emergence from the DA, the formation of membrane protrusions and endothelial cell migration.

Tip cell volume regulation depends on Aquaporin-mediated water influx

Aquaporin channels permit bidirectional flow of water across the plasma membrane, with the direction of flow determined by the osmotic gradient between the extracellular environment and cell cytoplasm (Agre et al., 2002). We next sought to determine whether water flows in or out of tip cells during sprouting angiogenesis. We hypothesize that water flux across the cell membrane can lead to changes in tip cell volume, with water influx increasing the volume while outflow decreases. To quantify tip cell volume, we injected a plasmid encoding kdrl:mEmerald transgene into 1-cell stage wildtype or aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos to label ECs in a mosaic manner. mEmerald-positive tip cells were imaged and their volume measured at 24 to 25 hpf. Wildtype ECs are on average 946.6 ± 299.8 µm³ in size (Fig. 5A and C). In the absence of both Aqp1a.1 and Aqp8a.1 function (Fig. 5B), tip cell volume significantly decreased by about 24% to 718.4 ± 308.5 µm³ (p = 0.0176, Fig. 5C) when compared to wildtype tip cells. This finding supports water influx as the direction of water flow in tip cells during migration, and that this is mediated through Aqp1a.1 and Aqp8a.1.

Figure 5.

It has been proposed that water influx through Aquaporin increases the diffusion of actin monomers and increases actin polymerisation (Loitto et al., 2007; Papadopoulos et al., 2008). Given that filopodia formation is dependent on actin polymerization (Mallavarapu and Mitchison, 1999; Ridley, 2011), we examined whether actin polymerization is perturbed in the absence of Aqp1a.1 and Aqp8a.1 by quantifying the growth rate of actin bundles in tip cell filopodia by tracking Lifeact-mCherry signal in the filopodia. In wildtype embryos, the average growth rate of actin is 0.054 ± 0.015 µm per second whereas in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos, the growth rate is significant decreased by 47% to 0.028 ± 0.010 µm per second (p < 0.0001, Fig. 5D). Our observations therefore support a role of zebrafish Aqp1a.1 and Aqp8a.1 in promoting actin polymerisation and consequently, filopodia number and length, which are reduced in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos (Fig. 4F and G).

In summary, our results demonstrate a role of endothelial Aquaporins in increasing water influx and as a consequence, endothelial tip cell volume and hydrostatic pressure, as well as in promoting actin polymerization.

Additive effects of actin polymerization and water influx in driving EC migration and sprouting angiogenesis

We have previously observed that ECs are still able to migrate to generate ISVs after the inhibition of actin polymerization (Phng et al., 2013). In these experiments, embryos were treated with a low concentration of Latrunculin B (Lat. B), an inhibitor of actin polymerization, that resulted in the suppression of filopodia formation. Under such conditions, ECs can generate small membrane protrusions and migrate more slowly in a directed manner to generate ISVs. How ECs are still able to migrate after the inhibition of actin polymerization was unclear. As our current study demonstrates a role of water inflow and hydrostatic pressure as another mechanism of cell migration, we hypothesize that ECs employ water influx to migrate when actin polymerization is compromised and that the depletion of both water influx and actin polymerization will result in a greater inhibition of EC migration.

To test this hypothesis, we examined ISV development in embryos with decreased actin polymerization and water inflow by treating aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos with Lat. B. At 28 hpf, ISVs of wildtype embryos treated with DMSO have reached the dorsal roof of the neural tube (Fig. 6A) and show an average length of 104.1 ± 13.47 µm (Fig. 6E). The treatment of wildtype embryos with 0.08 µg/ml Lat. B from 20 to 28 hpf (Fig. 6B) resulted in a significant decrease in ISV length (61.53 ± 23.27 µm) compared to control embryos (p < 0.0001, Fig. 6E). A similar decrease in ISV length was also observed in aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos treated with DMSO (63.75 ± 30.24 µm, Fig. 6C and E). When aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos were treated with 0.08µg/ml Lat. B from 20 to 28 hpf, there was a greater decrease in the length of ISV formed at 28 hpf (43.10 ± 24.42 µm, Fig. 6D and E) when compared to the inhibition of actin polymerization (p = 0.0028) or depletion of water influx (p < 0.0001) alone (Movie S7). These results therefore demonstrate additive functions of actin polymerization and water influx in driving EC migration during sprouting angiogenesis (Fig. 6F).

Figure 6.

Discussion

In this study, we demonstrate that endothelial tip cells utilize two mechanisms of migration concurrently during sprouting angiogenesis in the zebrafish - actin polymerization and hydrostatic pressure. The increase in hydrostatic pressure in endothelial tip cells is generated by Aqp1a.1- and Aqp8a.1-mediated water inflow. By performing detailed expression analyses, we showed that aqp1a.1 and aqp8a.1 are differentially expressed in newly formed vascular sprouts. Aqp1a.1 is expressed earlier in the DA than aqp8a.1 during development and in ECs destined to be tip cells. In newly formed ISVs, endothelial tip cells express higher aqp1a.1 mRNA levels while stalk cells express higher aqp8a.1 mRNA levels. In zebrafish harbouring mutations in both aqp1a.1 and aqp8a.1, there is delayed tip cell protrusion from the DA, decreased tip cell migration and increased occurrence of truncated ISVs at 3 dpf. We further discovered that aqp1a.1 and aqp8a.1 expression is upregulated by VEGFR2 activity, proposing that VEGFA-VEGFR2 signalling employs Aquaporin function to drive sprouting angiogenesis.

Several lines of evidence implicate Aquaporins in facilitating water flow into endothelial tip cells during sprouting angiogenesis. First, timelapse imaging showed enriched localization of Aqp1a.1 and Aqp8a.1 proteins at the migrating front of tip cells, suggesting that water flow occurs locally at the leading edge of the cell. Secondly, single cell analyses revealed Aqp1a.1- and Aqp8a.1-deficient tip cells have decreased cell volume, impaired expansion of cell membranes at the leading edge and decreased generation of stable cell protrusions. These observations implicate water flow into tip cells, leading to increased cell volume. Due to the incompressible property of water within an enclosed compartment such as the cell, water influx will lead to a rise in hydrostatic pressure. At the leading edge of tip cells, increased hydrostatic pressure would expand and deform membranes to generate protrusions. Additionally, water influx can reduce cytoplasmic viscosity and increase spacing between the plasma membrane and the underlying actin cytoskeleton, promoting actin monomer diffusion, actin polymerization and the generation of membrane protrusions to facilitate cell migration (Boer et al., 2023; Loitto et al., 2009).

A key finding in this study is the demonstration that endothelial tip cells employ hydrostatic pressure as a second mechanism of cell migration in vivo. In a previous study, we discovered that ECs continue to migrate when actin polymerization is inhibited and in the absence of filopodia (Phng et al., 2013). Here, we show that hydrostatic pressure is the driving force for EC migration in the absence of actin-based force generation. When both mechanisms are perturbed, there is a greater impairment in EC migration (Fig. 6D and E). This finding highlights the importance of hydrostatic pressure as an additional mechanism that reinforces actin-based cell shape changes and motility to ensure robust migration and formation of new blood vessels in complex three-dimensional environments. Such dual mode migration is advantageous when blood vessels develop in physically confined spaces such as in the zebrafish trunk, where ECs must emerge from the DA and migrate within somite boundaries and over tissues such as the notochord and neural tube to form ISVs. Such dependence on water flow in cell migration under confined microenvironment has been reported in human MDA-MB-231 breast cancer cells and mouse S180 sarcoma cells, which show cell migration persistence in narrow channels after inhibition of actin polymerization or myosin II-mediated contractility (Stroka et al., 2014). In this context, AQP5 mediates water flow and the decrease or increase in AQP5 expression suppresses or enhances, respectively, cancer cell migration (Chae et al., 2008; Jung et al., 2011; Stroka et al., 2014).

Our findings on the endothelial function of zebrafish Aqp1a.1, which is homologous to mammalian AQP1, corroborate previous studies demonstrating a function of AQP1 in promoting cell migration. In tumour bearing AQP1-null mice, there is reduced tumour angiogenesis and AQP1-deficient ECs display reduced migration in an in vitro wound healing assay (Maltaneri et al., 2020; Saadoun et al., 2005). In chick, AQP1 expression levels affect neural crest cell migration speed and direction by regulating filopodia length and stability (McLennan et al., 2019). The regulation of membrane protrusions has also been demonstrated for other Aquaporin proteins. For example, AQP9 weakens membrane-cytoskeleton anchorage and promote formation of membrane protrusions such as filopodia and blebs (Karlsson et al., 2013; Loitto et al., 2007). We also demonstrate for the first time a role of Aqp8a.1 in promoting EC migration and sprouting angiogenesis in the zebrafish. However, its mammalian ortholog, AQP8, was shown to inhibit migration of colorectal cancer cell lines in vitro (Wu et al., 2017), suggesting cell- and context-dependent function of AQP8 in the regulation of cell migration.

A central determinant of water movement is the osmotic gradient across the membrane, with water flowing in the direction of higher solute concentration. Given a crucial role of water flow in regulating cell volume, shape and migration, it is also important to understand the function and distribution of ion channels, exchangers and transporters and water channels that generate an osmotic gradient across the cell. This is reflected in a growing number of studies demonstrating a role of ion transporters in cell migration. In neutrophils, the increase in cell volume and potentiation of cell migration depend on the sodium-proton exchanger 1 (NHE1) and the chloride-bicarbonate exchanger 2 (AE2) (Nagy et al., 2023). In T cells, chemokine-induced migration depends on ion influx via SLC12A1, and water influx via AQP3 (Boer et al., 2023). In tumour cells migrating in narrow channels, the establishment of a polarized distribution of Na⁺/H⁺ pumps and Aquaporin proteins in cell membranes is required for net inflow of ions and water at the leading edge and net outflow at the trailing edge (Stroka et al., 2014). We have also observed enrichment of Aqp1a.1 and Aqp8a.1 proteins in ECs, at the leading edge of migrating endothelial tip cells as well as apical membranes during lumen formation (not shown). This leads to the question of how Aquaporin localisation is regulated as this will determine the site of increased water flow and hydrostatic pressure. The rapid changes in subcellular localization of mammalian Aquaporins upon stimulation occurs mainly via trafficking to and from the plasma membrane (Markou et al., 2022), and can lead to changes in the amount of Aquaporin protein in the plasma membrane. The regulation of AQP subcellular relocalization via calmodulin- and/or phosphorylation-dependent mechanisms has been implicated for AQP0-5 and AQP7-9 (Markou et al., 2022). For example, rapid translocation of AQP1 to plasma membrane upon a hypotonic stimulus is dependent on calmodulin activation and phosphorylation by protein kinases C (PKC) (Conner et al., 2012). Phosphorylation of AQP2 by protein kinase A (PKA) is required for vasopressin-stimulated relocalization of AQP2 from storage vesicles to the plasma membrane (Noda and Sasaki, 2005). Subcellular localization of AQP8 in hepatocytes is also regulated by PKA and PI3K signalling (Gradilone et al., 2005, 2003). In mouse neutrophils, exposure to a chemotactic gradient caused translocation of AQP9 to the leading edge of the cell that is mediated by PKC-dependent phosphorylation (Karlsson et al., 2011). Further work is needed to elucidate the mechanism of Aqp1a.1 and Aqp8a.1 protein distribution in ECs and which ion transporters are expressed in ECs to regulate the ion movement and cell migration.

In this study, we have assumed that the function of Aqp1a.1 and Aqp8a.1 is a result of water permeation based on previous studies demonstrating water molecules flowing through mammalian AQP1 (Preston et al., 1992; Zeidel et al., 1992) and AQP8 (Liu et al., 2006; Ma et al., 1997) as well as through their zebrafish orthologs Aqp1a.1 and Aqp8a.1 (Tingaud-Sequeira et al., 2010). However, AQP1 and AQP8 are not water-specific channels. AQP1 can conduct unpolar gases carbon dioxide (Nakhoul et al., 1998; Prasad et al., 1998; Wang et al., 2007) and nitric oxide (Herrera et al., 2006; Herrera and Garvin, 2007), and in zebrafish, Aqp1a.1 also mediates the transfer for small gaseous molecules such as carbon dioxide and ammonia (Talbot et al., 2015). AQP8 is permeable to ammonia (Jahn et al., 2004; Liu et al., 2006) and is present in the inner mitochondrial membrane of different cells (Calamita et al., 2005), where it is suggested to mediate ammonia transport rather than water fluxes (Soria et al., 2010). Among the three Aqp8 paralogs in teleost, Aqp8a.1 and Aqp8a.2 are also permeable to urea (Tingaud-Sequeira et al., 2010) and Aqp8b was shown to act mainly as a mitochondrial H₂O₂ channel (peroxiporin) in activated marine and freshwater teleost spermatozoa (Chauvigné et al., 2021, 2015). Nevertheless, because water movement flows up an osmotic gradient, increased entry of ions into the cell through Aquaporins will also trigger water inflow.

In summary, our study highlights the role of water influx and hydrostatic pressure as another force-generating mechanism utilized by cells to build tissues during animal development. We demonstrate that endothelial tip cells employ both actin polymerization and hydrostatic pressure for robust sprouting angiogenesis. As morphogenetic events are governed by a coordination of cell shape changes, cell migration and rearrangements, we envision that the shaping and formation of other tissues will also depend on Aquaporin function and changes in hydrostatic pressure.

Methods

Zebrafish maintenance and stocks

Zebrafish (Danio rerio) were raised and staged according to established protocols (Kimmel et al., 1995). Transgenic lines used in this work are Tg(kdrl:EGFP)^s843 (Jin et al., 2005), Tg(fli1a:H2B-EGFP)^ncv69 [re-named in ZFIN as Tg(fli1:zgc:114046-EGFP)] (Ando et al., 2019), Tg(fli1:Lifeact-mCherry)^ncv7 (Wakayama et al., 2015), Tg(fli1:myr-mCherry)^ncv1(Fukuhara et al., 2014), Tg(fli1ep:EGFP-PLC1δPH)^rk26 (Kondrychyn et al., 2020), Tg(fli1:h2bc1-mCherry)^ncv31 (Nakajima et al., 2017), Tg(kdrl:ras-mCherry)^s916 (Hogan et al., 2009). Zebrafish were maintained on a 12-hour light/12-hour dark cycle, and fertilized eggs were collected and raised in E3 medium at 28⁰C. To inhibit pigmentation in embryos older than 24 hpf, 0.003% N-Phenylthiourea (Sigma-Aldrich) in E3 medium was used. All animal experiments were approved by the Institutional Animal Care and Use Committee at RIKEN Kobe Branch (IACUC).

Plasmids and oligonucleotides

All DNA constructs used in this study were generated by using In-Fusion HD Cloning kit (Takara Bio Inc.). Plasmid mEmerald-N1 was a gift from Michael Davidson (Addgene plasmid #53976). Plasmids pDsChLexOG (Emelyanov and Parinov, 2008) containing terminal cis-required sequences of the Ds transposon from maize was a generous gift from Sergei Parinov. A detailed information regarding plasmids and primers used in this study can be found in Supplementary Tables 1 and 2, respectively.

Cloning of aqp1a.1 and aqp8a.1 genes

Total RNA was isolated from 1-day old zebrafish embryos with TRI Reagent using Direct-zol^TM RNA MicroPrep kit (Zymo Research) according to the manufacturer’s protocol. The first-strand cDNA was synthesized from 1 µg of a total RNA by oligo(dT) priming using the SuperScript III First-Strand synthesis system (Invitrogen) according to the manufacturer’s protocol. Amplification of cDNAs was performed using a high-fidelity KOD-Plus-Neo DNA polymerase (Toyobo, Japan) and resulting PCR products were cloned using NEB^® PCR Cloning kit (New England BioLabs). Positive clones and plasmids were verified by DNA sequencing.

Mosaic expression of DNA constructs and transgenic lines generation

To establish Tg(fli1ep:aqp1a.1-mEmerald)^rk30 and Tg(fli1ep:aqp8a.1-mEmerald)^rk31 transgenic lines, Tol2-based plasmids (10 ng/μl) were co-injected into one-cell stage zebrafish embryos with Tol2 transposase mRNA (100 ng/μl) transcribed from NotI-linearized pCS-TP plasmid (a gift from Koichi Kawakami, National Institute of Genetics, Japan) using the mMESSAGE mMACHINE SP6 kit (Invitrogen). Injected embryos were raised to adults and screened for germline transmission to establish a stable transgenic line. For mosaic expression of DNA construct, Ds-based kdrl:mEmerald plasmid (10 ng/μl) was co-injected into one-cell stage zebrafish embryos with Ac transposase mRNA (100 ng/μl) transcribed from BamHI-linearized pAc-SP6 plasmid (Emelyanov et al., 2006) using the mMESSAGE mMACHINE SP6 kit (Invitrogen). Embryos were analyzed at 25-26 hpf after injections.

In vivo cell volume analysis

Single-cell labelling of ECs in ISVs was achieved by mosaic expression of pDs-kdrl:mEmerald plasmid in wild type or aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29mutant Tg(fli1:Lifeact-mCherry)^ncv7 transgenic embryos. Embryos were imaged at 25-26 hpf with an Olympus UPLSAPO 60x/NA 1.2 water immersion objective with optical Z planes interval of 0.26 μm. Cell volume was measured with Huygens Essential 22.10 software (Scientific Volume Imaging B.V.) using object analyzer tool with manual adjustment of threshold.

Generation of zebrafish aqp1a.1 and aqp8a.1 mutants

Aquaporins mutants were generated by CRISPR/Cas9-mediated mutagenesis. A guide RNAs (gRNAs) targeting the first exons of aqp1a.1 and aqp8a.1 (Fig. S4 and S5) were generated by using cloning-independent protocol (Gagnon et al., 2014). Briefly, to generate templates for gRNA transcription, a 60-base oligodeoxynucleotides containing the SP6 promoter sequence, the gene-specific sequence (aqp1a.1: 5’-GACAGCTGGCCAGCAGACCC-3’; aqp8a.1: 5’-GATGTCTCCCCCATCGCCCG-3’) and 23-base overlap region was annealed to an 80-base constant oligodeoxynucleotide encoding the reverse-complement of the tracrRNA tail. The ssDNA overhangs were filled-in with T4 DNA polymerase (New England BioLabs) to generate dsDNA. The gRNAs were in vitro transcribed using MEGAScript SP6 kit (Invitrogen), DNase treated and purified with RNA Clean and Concentrator kit (Zymo Research). The gRNAs were quality-checked by running 1 µg of the product on a 2% TBE-agarose gel. Cas9/gRNA RNP complex was assembled just prior to injection and after 5 min incubation at room temperature 200 pg of Cas9 protein (Invitrogen) and 200 pg of gRNAs (100 pg of each, aqp1a.1 gRNA and aqp8a.1 gRNA) were co-injected into one-cell stage Tg(fli1:myr-mCherry)^ncv1 embryos. Such dose produced over 70% embryos survival post-injection showed CRISPR/Cas9-induced somatic aqp1a.1 and aqp8a.1 gene mutations. F₀ founders were identified by outcrossing CRISPR/Cas9-injected fish with wild type fish and screening the offspring for mutations at 2 dpf using Sanger sequencing. We found one F₀ fish with germline transmitted mutations in both aqp1a.1 and aqp8a.1 genes and outcrossed it with Tg(fli1:myr-mCherry)^ncv1 and Tg(fli1a:H2B-EGFP)^ncv69 transgenic fish to establish F₁ generation of heterozygotes. A 55 F₁ fish were genotyped to identify 4 aqp1a.1^+/rk28, 11 aqp8a.1^+/rk29 and 12 aqp1a.1^+/rk28;aqp8a.1^+/rk29 heterozygote fish. Subsequently, F₁ heterozygote fish were in-crossed to establish a single and double homozygote fish.

DNA isolation and genotyping

Genomic DNA was isolated from either embryos or fin clips using HotSHOT method (Meeker et al., 2007). To identify genomic lesions, the following primers were used: aqp1a1-fwd (5’-CGCCTCCAGATTCATTAGCAGGA-3’) and aqp1a1-rev (5’-GTAAGTGAACTGCTGCCAGTGA-3’) to amplify a 550 bp fragment, and aqp8a1-fwd (5-GGATCAATTGAGTTGCATAACAGAC-3’) and aqp8a1-rev (5’-CTGTAATGTAGACTTGTAAAGTGGA-3’) to amplify a 638 bp fragment. Mutations were assessed by direct sequencing of purified PCR products.

Quantitative real-time PCR

Total RNA from whole embryos was isolated with TRI reagent using Direct-zol RNA MicroPrep kit (Zymo Research) according to the manufacturer’s protocol, including on-column DNA digestion. RNA was quantified using a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific, USA). cDNA was synthesized from 300 ng of purified RNA using LunaScript RT SuperMix kit (New England BioLabs) according to the manufacturer’s protocol. Amplification of target cDNA was performed in technical triplicate of three biological replicates using the SYBR green methods. Each qPCR reaction mixture contained 5 µL 2x Luna Universal qPCR master mix (New England BioLabs), 1 µL cDNA (2-fold dilution), 0.2 µL antarctic thermolabile UDG (New England BioLabs) and 250 nM each primer to a final volume of 10 µL. Reactions were run in 384-well plates (Applied Biosystems) using the QuantStudio 5 Real-Time PCR system (Applied Biosystems) with the following thermal cycling conditions: initial UDG treatment at 25⁰C for 2 minutes, UDG inactivation at 50⁰C for 5 min and denaturation at 95⁰C for 1 minutes, followed by 45 cycles of 15 sec at 95⁰C, 30 sec at 60⁰C with a plate read at the end of the extension step. Control reactions included a no-template control (NTC) and a no-reverse transcriptase control (NRT). Dissociation analysis of the PCR products was performed by running a gradient from 60 to 95⁰C to confirm the presence of a single PCR product. Amplification data was analyzed using QuantStudio Design and Analysis software v1.5.3 (Applied Biosystems) and relative fold change was calculated using Pfaffl method (Pfaffl, 2001) and normalized to gapdh expression (as an internal control). Primers are listed in Supplementary Table 2. Two embryos were used in each biological replicate to generate an average value that was used to calculate the final mean ± SD from two or three independent experiments.

RNA in situ hybridization

RNAscope in situ hybridization

RNAscope^® in situ hybridization was conducted by using RNAscope Multiplex Fluorescent Reagent kit v2 (Advanced Cell Diagnostics). We adapted manufacture’s protocol designed for samples mounted on slides and protocol developed for whole-mount embryo samples (Gross-Thebing et al., 2014). For each experimental point, 5 embryos were processed in one 1.5 ml Eppendorf tube. Briefly, 20 and 22 hpf old embryos were manually dechorionated and fixed in freshly prepared 4% methanol-free PFA in PBS for 1 hour at room temperature. After fixation embryos were washed three times in PBS containing 0.01% Tween-20 (PBST), dehydrated stepwise in 5-min washes in a series of increasing methanol concentrations (25%, 50%, 75%, 100%) in PBS and then stored in 100% methanol at -20⁰C before use them for hybridization. The methanol-stored embryos were incubated with 5% H₂O₂ in methanol for 20 min at room temperature and then rehydrated stepwise in 5-min washes in series of decreasing methanol concentrations (75%, 50%, 25%) in PBS, followed by three times 5-min washes in PBST. Embryos were permeabilized in 2 drops of RNAscope^® Protease III for 15 min, rinsed three times with PBST and hybridized with RNAscope target probes overnight at 40⁰C in water bath. After hybridization probes were recovered and embryos were washed three times with 0.2x SSCT [0.01% Tween-20 in SSC] at room temperature, re-fixed with 4% PFA for 10 min, and washed three times with 0.2x SSCT. For RNA detection, the embryos were first hybridized with three different amplifier solution for 30 min (Amp1, and Amp2) and 15 min (Amp3) in a water bath at 40⁰C. After each hybridization step, the embryos were washed three times with 0.2x SSCT for 15 min at room temperature. To develop signal of each probe, the embryos were sequentially incubated in a water bath at 40⁰C with: (i) the horseradish peroxidase (HRP) for 15 min, (ii) TSA fluorophore for 30 min and (iii) the HRP blocker for 15 min. After each incubation step, the embryos were washed three times with 0.2x SSCT for 15 min at room temperature. HRP step is linked to a probe channel, we first developed C1 probe, Dr-aqp1a.1 (Advanced Cell Diagnostics) using HRP-C1 and TSA Vivid Fluorophore 520 (Tocris, dilution 1:1500), and then C3 probe, Dr-aqp8a.1 (Advanced Cell Diagnostics) using HRP-C3 and TSA Vivid Fluorophore 650 (Tocris, dilution 1:1500). Nuclei were counterstained with DAPI ready-to-use solution overnight at 4⁰C. Prior to imaging embryos were rinsed in PBST and kept in 70% glycerol in PBS. Images were acquired using Olympus FV3000 confocal microscope and an Olympus UPlanXApo 60x/NA 1.42 oil immersion objective.

Chromogenic in situ hybridization

Chromogenic in situ hybridization was conducted according to standard protocol (Thisse and Thisse, 2007) with minor modifications. For each experimental point, 10 embryos were processed in one 1.5 ml Eppendorf tube. Briefly, 30, 48, 72 and 96 hpf old embryos were manually dechorionated and fixed in freshly prepared 4% methanol-free PFA in PBS (pH 7.4) at 4⁰C overnight. After fixation embryos were washed three times in PBS containing 0.1% Tween-20 (PBST), dehydrated stepwise in 5-min washes in a series of increasing methanol concentrations (25%, 50%, 75%, 100%) in PBS and then stored in 100% methanol at -20⁰C before use them for hybridization. The methanol-stored embryos were rehydrated stepwise in 5-min washes in series of decreasing methanol concentrations (75%, 50%, 25%) in PBS followed by three times 5-min washes in PBST, and permeabilized 20 min with 10 µg/mL proteinase K (ThermoFisher Scientific). Hybridization was carried out in buffer [50% formamide, 5x SSC, 50 g/mL heparin, 500 g/mL tRNA (Roche) and 0.1% Tween-20] containing 5% dextran sulfate (MW=500 kDa) at 69⁰C overnight. After stringency wash, the specimens were blocked with 2% blocking reagent (Roche) in maleic acid buffer [100 mM maleic acid, 150 mM NaCl, 50 mM MgCl₂ and 0.1% Tween-20, pH 7.5] and incubated overnight with anti-digoxigenin antibody, conjugated with alkaline phosphatase (dilution 1:5000, Roche) at 4⁰C. For detection of alkaline phosphatase, specimens were incubated in staining buffer [50 mM Tris-HCl, 50 mM NaCl, 25 mM MgCl₂, 2% Polyvinyl alcohol and 0.1% Tween-20, pH 9.5] containing 375 µg/mL nitro-blue tetrazolium chloride (Roche) and 175 µg/mL 5-bromo-4-chloro-3’-indolyl-phosphate (Roche). Embryos were cleared in 70% glycerol overnight and imaged using Leica M205FA microscope. RNA probe synthesis. The cDNA-containing vectors were linearized with appropriate restriction enzymes (detailed maps of vectors will be provided upon request) and used as a template for RNA probe synthesis. Sense and antisense RNA probes were synthesized using MEGAscript SP6 or T7 kits (Invitrogen) and digoxigenin (DIG)-labeled rNTPs (Roche) according to the manufacturer’s protocol and purified using RNA Clean and Concentrator kit (Zymo Research). The following sense and antisense DIG-labeled riboprobes were generated: (i) aqp1a.1, 1.1 kb long probe comprised 121 nt of 5’UTR, 783 nt of the open reading frame (ORF) and 207 nt of 3’UTR; (ii) aqp8a.1, 0.8 kb long probe comprised 17 nt of 5’UTR, 783 nt of ORF and 30 nt of 3’UTR. Sense probes showed no specific or unspecific staining (data not shown).

Image processing

Chromogenic and RNAscope in situ hybridization images were processed using ImageJ software (NIH, version 2.9.0/1.53t) with brightness and contrast adjustments. Z-stacks of fluorescent images are presented as maximum intensity projections.

Analysis of mRNA expression in tip and stalk cells

Embryos were mounted on slide and imaged using Olympus FV3000 confocal microscope and Olympus UPlanXApo 60x/NA 1.42 oil immersion objective. Maximum intensity projection of z-stacks was used to determine the level of cellular fluorescence in ImageJ. The tip and stalk cells were selected using freehand selection tool, “mean grey value”, “area” and “integrated density” in each ROI were measured first for channel 1 (aqp1a.1). A region next to cell that has no fluorescence was selected as background and measured. The step was repeated for channel 2 (aqp8a.1) on the same ROIs. Two parameters were calculated for each channel, (i) the corrected total cell fluorescence, CTCF=Integrated Density – (Area of selected cell x Mean fluorescence of background) and (ii) ratio, R=CTCF of tip cell/CTCF of stalk cell. We consider that if R <0.9, expression is lower in tip cell, if R=0.9-1.1, expression is equal in tip and stalk cells, if R>1.1, expression is higher in tip cell.

Imaging

For live confocal imaging, embryos were mounted in 0.8% low-melt agarose (Bio-Rad) in E3 medium containing 0.16 mg/mL Tricaine (Sigma-Aldrich) and 0.003% phenylthiourea (Sigma-Aldrich) in glass bottom 35-mm dishes (MatTek). Confocal Z-stacks were acquired using an inverted Olympus IX83/Yokogawa CSU-W1 spinning disc confocal microscope equipped with a Zyla 4.2 CMOS camera (Andor) and Olympus UPLSAPO 40x/NA 1.25 or 30x/NA 1.05 silicone oil immersion objectives. Bright-field images were acquired on Leica M205FA microscope. Images were processed using ImageJ software (NIH, version 2.9.0/1.53t).

Analysis of cell and nuclei migration

aqp1a.1^+/rk28;aqp8a.1^+/rk29 and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos on Tg(fli1a:H2B-EGFP)^ncv69;(fli1:Lifeact-mCherry)^ncv7 double transgenic background were imaged from 21 to 30 hpf with time interval 6 min using Olympus UPLSAPO 30x/NA 1.05 silicon oil immersion objective. We tried to mount both heterozygote and mutant embryos on the same dish (1 het plus 2 mut, or 2 het plus 2 mut) to image embryos on the same experimental conditions. Acquired time-lapse images were registered using HyperStackReg plugin in ImageJ, tip cell and nuclei were tracked using Manual Tracking plugin and velocity was calculated.

Analysis of grow rate of actin bundles in filopodia

Wild type and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29embryos on Tg(fli1ep:EGFP-PLC1δPH)^rk26;(fli1:Lifeact-mCherry)^ncv7 double transgenic background were imaged starting from 25 hpf for 10 min with time interval 30 sec using Olympus UPLSAPO 60x/NA 1.2 water immersion objective. Actin bundles growth was tracked using Manual Tracking plugin and velocity was calculated.

Analysis of filopodia number and length

Wild type, aqp1a.1^+/rk28;aqp8a.1^+/rk29 and aqp1a.1^rk28/rk28;aqp8a.1^rk29/rk29 embryos on Tg(fli1ep:EGFP-PLC1δPH)^rk26;(fli1:Lifeact-mCherry)^ncv7double transgenic background were imaged starting from 24 hpf for 30 min with time interval of 50 sec using Olympus UPLSAPO 60x/NA 1.2 water immersion objective. In a single image we counted filopodia number per vessel, measured membrane length and calculated filopodia number.

Chemical treatment

Latrunculin B (Merck Millipore) was dissolved in DMSO to 1 mg/ml and stored at -20⁰C. Dibenzazepine (YO-01027) (Selleck Chemicals) was prepared as 10 mM solution in DMSO and stored at -80⁰C. Ki8751 (Selleck Chemicals) was prepared as 5 mM solution in DMSO and stored at -80⁰C. All compounds were diluted to the desired concentration in E3 medium.

Statistical analysis

Statistical analysis was performed using Prism software version 10.2.0 (GraphPad). The variance between the mean values of two groups was evaluated using the unpaired Student’s t-test. For assessment of more than three groups, we used one-way analysis of variance (ANOVA) test. A P value of <0.05 was considered statistically significant. Statistic details can be found in each figure legend.

Cell culture

Adult human aortic endothelial cells (HAEC) (Lonza, lot-20TL231227) were cultured in EGM medium (Lonza) and used at passage 3. Cells were seeded in EBM medium (Lonzo) supplemented with 2% FBS (Gibco) at 4 x 10⁵ cells/well on 24-well plate (Trueline) coated with 5 μg/ml fibronectin (Sigma-Aldrich). Cells were treated with either 0.01% DMSO or different concentrations of Ki8751 inhibitor for 6 hours. Cells were lysed with TRI reagent and RNA was isolated using Direct-zol RNA MicroPrep kit (Zymo Research).

Single-cell RNA sequencing

ECs were isolated from Tg(kdrl:EGFP)^s843 transgenic embryos. Cell sorting was carried out with FACSAriaII Cell Sorter (BD Bioscience). Single-cell suspension was loaded into the 10X Chromium system and cDNA libraries were constructed using Chromium Next GEM Single Cell 3′ GEM, Library and Gel Bead Kit v2 (10X Genomics) according to manufacturer’s protocol. Library was sequenced on the Illumina HiSeq 1500 Sequencer (Illumina, USA). Cell Ranger v2.1 was used to de-multiplex raw base call (BCL) files generated by Illumina sequencers into FASTQ files, perform the alignment, barcode counting, and UMI counting. Sequencing reads were mapped to the zebrafish genome assembly (GRCz11, Ensembl release 92). Further analyses were performed using Seurat package49 (version 4.3.0) in R software (https://www.r-project.org). The expression matrices were first filtered by keeping genes that are expressed in a minimum of 3 cells and cells that expressed a minimum of 200 genes for downstream analysis. The data were then normalized using NormalizeData function which normalizes the gene expression for each cell by the total expression counts. To correct for batch effect between the data from the two-time points, the rpca method in the Seurat package was applied to integrate the data. The top 2000 variable genes identified using the vst method in FindVariableFeatures function were used for principal component analysis in RunPCA function, and the first 30 principal components were used for visualization analysis with Uniform Manifold Approximation and Projection (UMAP) method, and in FindNeighbors function analysis. The cell clustering resolution was set at 0.5 in FindClusters function.

Acknowledgements

We thank members of the Phng Lab, Y-C. Wang and S. Thukral for discussions and suggestions; G. Chen, E. Taniguchi, RIKEN BDR Research Aquarium and RIKEN Kobe BioImaging Facilities & Factory for technical assistance; J. Vermot, M. Francois and A. Yap for comments on the manuscript. This work was supported by core funding from RIKEN BDR (to L.K.P.), RIKEN BDR-Otsuka Pharmaceutical Collaboration Center (I.K.), the Naito Foundation (L.K.P.), the JSPS Grants-in-Aid for Scientific Research (KAKENHI) grants (22H02624, 22H05168 to L.K.P; 22K06244 to I.K.), Swedish Research Council (2015-00550 to C.B.); Swedish Cancer Society (2018/449, 2018/1154 to C.B.); Knut and Alice Wallenberg Foundation (2020.0057, 2018.0218 to C.B.); Swedish Brain Foundation (ALZ2019-0130, ALZ2022-0005 to C.B.); the Leducq Foundation (22CVD01, 23CVD02 to C.B.) and the Innovative Medicines Initiative (IM2PACT-807015 to C.B.)

Authorship contributions

Conceptualization: L.K.P.; Methodology: I.K., H.W. and L.K.P.; Formal analysis: I.K. and L.H.; Resources: C.B. and L.K.P.; Writing: I.K. and L.K.P; Supervision: L.K.P.

Declaration of interests

The authors declare no competing interests.