In vivo modulation of endothelial polarization by Apelin receptor signalling

Endothelial cells (ECs) respond to shear stress by aligning in the direction of flow. However, how ECs respond to flow in complex in vivo environments is less clear. Here we describe an endothelial-specific transgenic zebrafish line, whereby the Golgi apparatus is labelled to allow for in vivo analysis of endothelial polarization. We find that most ECs polarize within 4.5 h after the onset of vigorous blood flow and, by manipulating cardiac function, observe that flow-induced EC polarization is a dynamic and reversible process. Based on its role in EC migration, we analyse the role of Apelin signalling in EC polarization and find that it is critical for this process. Knocking down Apelin receptor function in human primary ECs also affects their polarization. Our study provides new tools to analyse the mechanisms of EC polarization in vivo and reveals an important role in this process for a signalling pathway implicated in cardiovascular disease.

V ascular endothelial cells (ECs) form a continuous cell layer that lines blood vessels. ECs are constantly subjected to a variety of shear stresses from the flow of blood and display profound morphological adaptations to their environmental conditions. These morphological adaptations include planar cell polarization, cell elongation and alignment of microtubules in the direction of blood flow 1,2 . Recent studies have shown that defects in ECs' adaption to blood flow are highly correlated with cardiovascular diseases including atherosclerosis 3,4 . The alignment of ECs in the direction of flow requires the sensing of mechanical shear stress and the conversion of these stimuli to biochemical signals that mediate cytoskeletal remodelling. A number of mechanosensitive molecules have been identified including mechanically gated channels [5][6][7][8] , mechanosensitive receptors 9-11 , G-protein-coupled receptors (GPCRs) [12][13][14][15][16] , G-proteins [16][17][18][19] , mechanosensitive enzymes 20 and cilia 21 . The observation that certain G-proteins are activated within seconds following mechanical perturbation by shear stress suggests that GPCRs are involved in mediating mechanochemical signal transduction in ECs.
Here we generated a new zebrafish transgenic line in which the Golgi apparatus is labelled with a fluorescent protein. Using this line in combination with an endothelial nuclear line allows one to follow EC polarization in real time. We analysed EC polarization during development and investigated the role of blood flow in this process. We further found that the Apelin receptor (Aplnr), a GPCR involved in EC migration, also regulates EC polarization.
Result EC polarization during development. To investigate the dynamics of endothelial polarization by blood flow in vivo, we generated a new zebrafish transgenic line that labels the Golgi apparatus specifically in ECs 22 . We fused the complementary DNA of human b-1, 4-galactosyltransferase1 (B4GALT1), which is mainly localized to the trans Golgi network 23 , to mCherry and expressed it under the control of the fli1a endothelial promoter 24 . To validate this transgenic line, we performed confocal imaging during vascular development of Tg(kdrl:NLS-EGFP);Tg(fli1a: B4GALT1-mCherry) embryos, which express nuclear enhanced green fluorescent protein (EGFP) and Golgi-specific mCherry specifically in ECs. We found that during EC migration the Golgi was localized at the leading edge of cells ( Supplementary Fig. 1a,b and Supplementary Movie 1). However, when ECs stopped migrating, the Golgi quickly moved to a middle position relative to the nucleus ( Supplementary Fig. 1b,d). In addition, the Golgi localized to the luminal side of ECs during lumen formation ( Supplementary Fig. 1c,c 0 ). These data indicate that this reporter line allows real-time monitoring of EC polarity in vivo and can be used to track changes in EC polarization.
We next examined Tg(kdrl:NLS-EGFP);Tg(fli1a:B4GALT1-mCherry) embryos to determine whether and when ECs become polarized in response to blood flow during development. In accordance with a previous report 25 , we classify ECs as polarized when the Golgi apparatus lies upstream of the nucleus with respect to the direction of flow (Fig. 1j(iii)). At 30 hours post fertilization (h.p.f.), when blood circulation has just started ( Fig. 1a-b 0 ,c-c 0 ,j(i)), ECs in the dorsal aorta (DA) and the posterior cardinal vein (PCV) remained unpolarized. However, the ECs in the DA (Fig. 1d-e 0 ,g-h 0 ,j(ii)) and the intersegmental vessels (ISVs) (Fig. 1f, (Fig. 1k,l). In contrast, most ECs in the PCV (98.2% at 30 h.p.f., 96.4% at 48 h.p.f. and 91.4% at 72 h.p.f.) failed to polarize during this time (Fig. 1k). To better understand the EC polarization differences in various vascular beds, we examined ECs in several different regions, that is, the brain, eye, pharyngeal arch, common cardinal vein (CCV) and caudal vein plexus 26 . ECs in the brain, eyes and pharyngeal arch were mostly polarized (white arrowheads), whereas those in the CCV and caudal vein plexus were not polarized as observed in the PCV (yellow arrowheads) ( Supplementary Figs 2a-k and 3g,h). Interestingly, venous ECs were usually less polarized than arterial ECs. These data reveal that there is a clear heterogeneity between vascular beds in terms of EC polarization and especially between arteries and veins.
EC polarization by blood flow. To better understand the dynamics of EC polarization by flow, we performed time-lapse confocal imaging of ECs starting at 30 h.p.f., that is, shortly after the onset of blood flow (Fig. 2a-d). Interestingly, the Golgi apparatus in DA ECs gradually moved from a downstream to an upstream position on initiation of vigorous blood flow (Fig. 2d, Supplementary Fig. 3a-c,f(i) and Supplementary Movie 2) and it took B4.5 h for them to become polarized after the onset of blood flow (Fig. 2e). As blood flow gradually increases in speed after the heart starts contracting 21 , we decided to look at dividing ECs at a time when vigorous blood flow was established, to exclude this variable. EC division plane is normally oriented perpendicular to the long axis of the vessel 27 . By measuring the time required for ECs to become polarized after cell division, we observed that most resulting daughter cells became polarized in the direction of blood flow within 105 min following division ( Supplementary Fig. 3d,e,f(ii) and Supplementary Movie 3). For comparison, we also analysed the correlation of EC division and polarization in the PCV by time-lapse confocal imaging. ECs within the PCV did not polarize under established blood flow, even though they divided as the DA ECs do ( Supplementary  Fig. 3g,h). Arterial and venous ECs have been reported to exhibit different polarization patterns in vivo 28,29 . To investigate whether arterial or venous identity plays a role in the polarization response to blood flow, we examined ECs in the arterial ISVs and vISVs, respectively. Interestingly, ECs in both vessel types polarized ( Fig. 2f-j) within 30 min after the onset of blood flow, suggesting that flow-induced EC polarization may not simply be dependent on arterial or venous identity.
EC polarization by blood flow is reversible. To further investigate the role of blood flow in EC polarization, we eliminated cardiac function by injecting morpholinos (MOs) for tnnt2a, a gene encoding the thin-filament contractile protein cardiac troponin T 30 , into one-cell stage Tg(kdrl:NLS-EGFP);Tg(fli1a: B4GALT1-mCherry) embryos. Most ECs in the DA of 48 h.p.f. tnnt2a morphants failed to polarize (89.1% non-polarized), as they displayed a random Golgi orientation, in contrast to ECs within the DA of uninjected embryos (76.6% polarized) ( Fig. 3a-g). Likewise, when we stopped the heartbeat of 96 h.p.f. larvae with a 10 mM 2, 3-butanedione monoxime (BDM) treatment, we observed the gradual movement of the Golgi apparatus from an upstream to a downstream position within minutes after the cessation of blood flow and the ECs eventually became completely depolarized (Fig. 3h-j and Supplementary Movie 4) 31 . We observed a similar loss of polarization in the ISV ECs after stopping the blood flow ( Supplementary Fig. 4a-c).
To investigate the extent of the flexibility of EC polarization, we examined ECs through a cycle of blood flow manipulation. We first stopped blood flow to depolarize the ECs and then   started flow again, to analyse whether ECs were able to repolarize (Fig. 3k). We found that most ECs in the DA completely depolarized in larvae treated with BDM for 12 h (86.4%; ) or 24 h (93.5%; Fig. 3p,q,t). However, 12 h after BDM removal to reinitiate blood flow, ECs repolarized to an extent (72.5%; Fig. 3n,o,t) similar to control larvae treated with dimethylsulfoxide (84.0%; Fig. 3r-t). These data indicate that EC polarization can be dynamically modulated by blood flow in vivo. We also investigated the role of red blood cells (RBCs) in EC polarization and depleted RBCs by injecting a gata1 MO 32,33 . We found no significant differences in EC polarization between gata1 morphants and uninjected controls at 96 h.p.f. (Supplementary Fig. 4d-  Aplnr signalling modulates EC polarization. Based on our in vivo observations, EC migration ( Supplementary Fig. 1a,b,d) and blood flow (Fig. 2d,i-k) induce the same polarization response. Thus, we decided to investigate the role of signalling pathways known to regulate EC migration during EC polarization. The Aplnr is a GPCR implicated in the migration of ECs and cardiac precursor cells [34][35][36][37] . In addition, Aplnr is strongly expressed in the cardiovascular system of mouse, frog and zebrafish, and it plays an important role in cardiovascular function [38][39][40] . In zebrafish, there are two paralogues of aplnr, aplnra and aplnrb, and two ligand-encoding genes, apelin (apln) and apela 41,42 . We examined the expression of these genes by in situ hybridization and found that aplnra, aplnrb and apln were highly expressed in the zebrafish vasculature during development ( Supplementary Fig. 5a-c), whereas apela was not ( Supplementary Fig. 5d), consistent with previous reports 41 . As we did not detect apela expression in developing blood vessels, we decided to focus on aplnra, aplnrb and apln. We first analysed aplnr, aplnra and aplnrb mutant larvae. We did not observe any obvious phenotype in aplnra þ / À or aplnra À / À larvae ( Supplementary Fig. 5e). As previously reported 34,35 , aplnrb À / À animals develop pericardial oedema (red arrow, Supplementary  Fig. 5e), whereas aplnrb þ / À animals exhibit no obvious phenotype ( Supplementary Fig. 5e). However, analysis of Golgi localization in aplnrb þ / À larvae at 72 h.p.f. revealed defects in EC polarization (23.1%) despite vigorous blood flow ( Fig. 4g-i,j and Supplementary Fig. 5f-i). In contrast, EC polarization in aplnra þ / À larvae (59.6%; Fig. 4d-f,j) appeared similar to wildtype siblings (66.7%; Fig. 4a-c,j). These data suggest that aplnrb modulates EC polarization. To test whether Aplnrb signalling modulates EC polarization cell autonomously, we generated mosaic embryos by cell transplantation (Fig. 4k). We observed a reduction in the polarization frequency of aplnrb þ / À ECs transplanted into wild-type larvae (14.3%; n/n' ¼ 6/28), compared with that of wild-type ECs transplanted into wild-type larvae (63.4%; n/n' ¼ 9/41) ( Fig. 4l-o,p), indicating that Aplnrb signalling modulates EC polarization in a cell-autonomous manner.
To investigate whether Apelin ligands also regulate EC polarization, we first examined apln morphant larvae at 72 h.p.f. 38 . To identify a dose of apln MO that causes minimal presumed off-target effects, we injected apln MO into embryos from an apln þ / À incross 43 (Supplementary Fig. 6a-f,g) and selected 4.5 ng. At this dose, apln morphants exhibited a complete block in parachordal vessel formation (Supplementary Fig. 6h-j) as previously reported 38 , but also minimal off-target effects, for example, pericardial oedema (Supplementary Fig. 6a-g). Interestingly, apln morphants showed a similar frequency of EC polarization (60.6%, Supplementary Fig. 6n-p,q) as uninjected controls (70.6%; Supplementary Fig. 6k-m,q). We further investigated whether EC polarization was dependent on Apln by evaluating apln À / À larvae at 72 h.p.f. (Fig. 4q-u). apln À / À larvae showed a lack of parachordal vessels similar to apln morphants ( Fig. 4s and Supplementary Fig. 6h-j). In addition, no significant effect on in EC polarization was observed in apln À / À larvae (68.5%; Fig. 4s-u) compared with wild-type siblings (67.4%; Fig. 4q,r,u). These data suggest that Aplnrb modulates EC polarization independently of Apln function. We next examined the expression of apela in apln À / À animals, to see whether there was an upregulation of apela expression that might compensate for the loss of Apln function. However, apela expression did not appear to be significantly increased in apln À / À embryos or larvae compared with wild-type siblings ( Supplementary Fig. 7a,b).
In mouse, Apln has been linked to tip/stalk cell differentiation during sprouting angiogenesis and Apln is dysregulated in Dll4 loss-of-function models 44 . Loss of Apln or Aplnr causes delayed blood vessel development in the mouse retina and reduces EC proliferation during fin regeneration in zebrafish 44 . We thus counted EC numbers in ISVs of aplnrb þ / À and apln À / À larvae, to determine whether cell proliferation was reduced. As previously reported, both aplnrb þ / À and apln À / À larvae showed reduced EC numbers in ISVs ( Supplementary  Fig. 7c,d). However, and as previously mentioned, apln À / À larvae do not show an endothelial polarization defect ( Fig. 4q-u), suggesting that reduction in cell proliferation is not sufficient to cause a polarization defect in ECs. In addition, we investigated whether loss of Aplnrb affected arterial-venous identity or Notch signalling. We did not observe any obvious differences between wild-type and aplnrb þ / À embryos and wild-type siblings in the expression of the arterial reporter line Fig. 7g-i). Furthermore, no differences in the expression of the Notch reporter line Tg(EPV.TP1-Mmu.Hbb: hist2h2l-mCherry) 45 were observed between aplnrb þ / À embryos and wild-type siblings ( Supplementary Fig. 7j-k 0 ).

APLNR signalling modulates the polarization of human ECs.
To test whether the role of Aplnr in EC polarization was conserved in humans, we investigated EC polarization in vitro after knocking down APLNR (also known as APJ and AGTRL1) in primary human ECs, human umbilical venous ECs (HUVECs), under laminar flow ( Fig. 5a-b 0 ,d). Interestingly, only 27.6% of APLNR knockdown (KD) HUVECs under laminar flow showed a polarized localization of their Golgi apparatus compared with 64.8% of control cells (white arrowheads; Fig. 5a-b 0 ,d), as defined by the position of the Golgi apparatus within a 45°angle of the direction of flow (Fig. 5c). In addition, APLNR KD HUVECs showed severely disturbed cytoskeletal structures compared with control cells, which show stress fibres aligned along the direction of flow (Fig. 5a 0 ,b 0 ). We observed similar defects in Golgi localization and stress fibre assembly when we knocked down APLNR in human arterial ECs (HUAECs) (Supplementary Fig. 8a-b 0 ,c). These results suggest that the role of Aplnr in EC polarization is conserved in humans.
The APLNR has been reported to function in mouse cardiomyocytes as a dual receptor for Apln via G-protein ai (Gai) and mechanical stretch via b-Arrestin (Arrb) 46 . To analyse whether the role of APLNR in laminar flow-induced EC polarization was dependent on its ligand APLN and Gai, we knocked down APLN ( Fig. 5e-g) and also inhibited Gai  (Fig. 5h-j). We did not find a significant effect on EC polarization or the actin cytoskeleton in APLN KD HUVECs (56.5%) under laminar flow, compared with controls (59.6%) ( Fig. 5e-g). However, we observed a moderate decrease in EC polarization under laminar flow in PTX-treated HUVECs (36.6%) compared with controls (56.6%) (Fig. 5h-j). Given that the exposure to PTX was relatively long (430 h due to the experimental design), it is likely to be that the effect of PTX is not specific to inhibiting APLNR downstream signalling. Next, we investigated whether laminar flow-induced EC polarization was dependent on ARRB. We found depolarized Golgi localization as well as disrupted actin cytoskeleton when both ARRB isoforms, ARRB1 and ARRB2, were knocked down compared with controls ( Fig. 5k-m 0 ,n). These data suggest that EC polarization may be modulated in an APLN-independent manner, and that APLNRdependent EC polarization may function, at least in part, via ARRB. We also analysed whether ARRB is an effector of APLNR in laminar flow-induced signalling by transfecting ARRB-GFP into HUAECs. We found that ARRB-GFP translocated to the membrane when HUAECs were exposed to laminar flow, but not after APLNR KD (Supplementary Fig. 8d-g 0 ,h). These data suggest that ARRB functions downstream of APLNR in a flowdependent signalling.
KLF2 regulation by APLNR and EC polarization in klf2 mutants.
Aplnr mutant mice show decreased endothelial expression of Krüppel-like factor 2 (Klf2), which encodes a master regulator of vascular adaptation to blood flow 47 . Thus, we examined whether KLF2 expression was reduced in APLNR KD HUVECs under laminar flow and in aplnrb heterozygous and homozygous mutant zebrafish embryos. We found that KLF2 expression was reduced in APLNR KD HUVECs under laminar flow compared with controls ( Supplementary Fig. 9a). However, there did not appear to be a dramatic difference in klf2a expression between wild-type and aplnrb þ / À zebrafish embryos, and although klf2a expression in aplnrb À / À is clearly reduced compared with wild type, these animals have no blood flow that probably accounts, at least in part, for this reduction in expression ( Supplementary  Fig. 9b-d). To investigate whether loss of Klf2 function leads to endothelial polarization defects to a similar extent as that observed in aplnrb þ / À larvae, we examined zebrafish mutants for the two klf2 paralogues, klf2a and klf2b. Both klf2a and klf2b single mutants appear to develop normally ( Supplementary  Fig. 9e) in agreement with a recent report 48 . However, we found that 50% of klf2a À / À larvae (n ¼ 6), 83% of klf2b À / À larvae (n ¼ 6) and 80% of klf2a MO-injected klf2b À / À larvae showed EC polarization defects as observed in aplnrb þ / À larvae ( Supplementary Fig. 9f,g-k 0 ). We also observed that loss of klf2 function reduced the diameter of the DA and PCV ( Supplementary Fig. 9l-n 0 ,o) and increased blood flow velocity ( Supplementary Fig. 9p,q), indicating that the polarization defects are not due to decreased blood flow velocity. Altogether, these data suggest that Aplnr may modulate EC polarization and vascular remodelling through Klf2.

Discussion
In this study, we found that EC polarization was heterogeneous between different vascular beds ( Fig. 1a-i,k,l and Supplementary  Fig. 2a-k). Previous studies have reported different polarization patterns in arterial and venous ECs in fixed tissue samples 28,29 . Consistent with these reports, we observed that ECs in the DA polarize shortly after the onset of vigorous blood flow (Fig. 2a-d), whereas ECs in the PCV do not ( Supplementary Fig. 3g,h). Interestingly, some non-polarized ECs in the PCV became polarized after migrating dorsally to form the vISVs (Fig. 2f,h,j). These data suggest that ECs' ability to respond to flow is not fixed by arterial or venous identity, but dynamically adjusted according to their environment. Factors such as the velocity and pulse patterns of flow are known to influence EC polarization in vivo and in vitro 41,49 . Anton et al. 50 and Goetz et al. 21 analysed the differences in blood cell motion in the DA, PCV and ISVs of zebrafish embryos. Blood cell motion in the DA (mean velocity of B300 mm s À 1 ; shear stress of B0.43 dynes cm À 2 ) reflects the contractions of the heart, that is, with regular intermittent propagation of fluid and almost no flow between pulses. ISV flow is also pulsatile but with a profile significantly different from that of the DA. Its peak flow is slower than that of the DA. By contrast, blood flow in the PCV (mean velocity of B100 mm s À 1 ; shear stress of B0.14 dynes cm À 2 ) is relatively steady with a 57% reduction in pulse amplitude. Thus, changes in blood flow velocity and pulse patterns may affect endothelial response to flow, as ECs move from the PCV into vISVs. In addition, Baeyens et al. 51 reported that Vegfr3 signalling regulates the shear stress response of ECs in different blood vessels including arteries, veins and lymphatics. However, the relative contribution of different extrinsic and intrinsic signals in ECs' response to shear stress remains to be investigated. Using live imaging, we observed that ECs become polarized during their migration ( Supplementary Fig. 1a-d) and also after exposure to flow (Fig. 2a-d,k). Aplnr is known to regulate EC migration during development 36 . Previously, Scimia et al. 46 reported that in mouse cardiomyocytes, the Aplnr transduces Apln binding or mechanical stretch in different ways. The ARTICLE response to Apln seems to involve G-proteins, as it is PTX sensitive, whereas that induced by stretch occurs in the absence of Apln, which is PTX insensitive and G-protein independent but Arrb dependent 46 . However, whether the Aplnr acts in ECs as a dual receptor for Apln and mechanical stimuli as it does in cardiomyocytes is unclear. Our studies on aplnr mutants reveal that ECs in aplnrb þ / À larvae fail to polarize under blood flow ( Fig. 4g-i,j). Consistently, APLNR depleted primary human ECs showed polarization defects under laminar flow (Fig. 5a-b 0 ,d), indicating that the role of Aplnr in EC polarization is conserved. Interestingly, ECs in apln mutant and apln morphant larvae did not show a significant difference in polarization compared with controls ( Fig. 4q-u and Supplementary Fig. 6k-p,q). Similarly, APLN KD HUVECs did not show a significant reduction in polarization compared with controls ( Fig. 5e-f 0 ,g). Furthermore, HUVECs deficient for ARRBs showed a reduction in EC polarization under laminar flow (Fig. 5k-n). Altogether, these data suggest an Apln-independent function of Aplnr in flowinduced EC polarization and one that requires at least in part Arrb function. However, even though EC polarization appeared mostly unaffected in apln mutants and morphants, as well as in APLN KD HUVECs under laminar flow, we cannot exclude the possibility of ligand dependency, for example, via Apela or an unknown ligand. Ligand dependency could still be the case if the residual levels of Apela in apln mutants and morphants or in APLN KD HUVECs are sufficient to induce EC polarity, or if another ligand, even possibly a heretofore unknown ligand, can activate Aplnr during EC polarization.
In summary, we have developed a new tool that allows the in vivo visualization of the Golgi apparatus and thereby the polarization status of ECs. We observed a dynamic and reversible localization of the Golgi apparatus in ECs in response to blood flow. Interestingly, haploinsufficiency in zebrafish of the Aplnrb hampered EC polarization in the presence of normal blood flow and in a cell-autonomous manner. These results indicate that ECs process local signals to regulate their polarization through Aplnrb. Furthermore, we found this mechanism to be operating in human cells, as Aplnr KD abrogated Golgi polarization under laminar flow. Therefore, this new zebrafish transgenic line represents a valuable tool to investigate in vivo the pathways that modulate flow sensing and response. A better understanding of these mechanisms will probably have implications towards treating numerous pathological conditions including cardiovascular diseases.

Methods
Zebrafish. All zebrafish husbandry was performed under standard conditions in accordance with institutional (Max Planck Society) and national ethical and animal welfare guidelines. Embryos were staged by h.p.f. at 28.5°C 54 . aplnra mu296 , aplnrb mu281 and apln mu267 mutant lines were previously reported 36 . klf2a bns11 and klf2b bns12 were generated using TALENs. We used wild-type (AB), Tg(kdrl:NLS-EGFP) zf109 (ref. 55 Assessment of EC polarization. Polarization patterns of ECs were defined as follows: (1) upstream, if the Golgi was located on the upstream side of the nucleus; (2) none, if the Golgi was located by the middle of the nucleus; or (3) downstream, if the Golgi was located on the downstream side of the nucleus. For quantitative analyses of EC polarization, we analysed ECs of the DA, PCV and ISVs at the 5th and 6th somites at 30, 48 and 72 h.p.f. (Fig. 1k,l). To assess EC number, we counted the number of ECs located in the upper half of the DA. For quantitative analyses of EC polarization in aplnr heterozygous mutant larvae, we analysed ECs of the DA between, and including, the 5th and 11th somites at 72 h.p.f.
Tg(fli1a:B4GALT1-mCherry) bns9 fish were generated by injecting pTol ISceI fli1a:B4GALT1-mCherry (25 ng), along with Tol2 transposase mRNA (25 ng) into one-cell stage AB embryos. Larvae were selected at 3 days post fertilization for high expression and grown to adulthood. Germline founders were identified by specific expression of B4GALT1-mCherry in the blood vessels of their progeny. In vivo imaging and image processing. Pigmentation of embryos and larvae was inhibited by 1-phenyl-2-thiourea (Sigma). The embryos were treated with 100 mg ml À 1 tricaine (Sigma), mounted in a drop of 1.5% low melting agarose in egg water and placed onto a glass dish. Fluorescence images were obtained using a Zeiss LSM 780 confocal microscope. Three-dimensional-rendered z-stack images and three-dimensional surface-rendered images and movies were analysed and assembled using the IMARIS software (BITPLANE).
Cells and reagents. HUVECs and human umbilical artery ECs (HUAECs) were obtained from Lonza. HUVECs were cultured on collagen-I-coated dishes and  supplied with the endothelial growth medium EGM-2 (Lonza). Confluent cells at passages oP3 were used for all experiments.
HUVECs were transfected with siRNAs using Opti-MEM and Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. A repeated KD was performed the next day and assays were started at 72 h after the first KD. RNA isolation was performed using Qiagen RNeasy kit. Reverse transcription was done using the Transcriptor high-fidelity cDNA synthesis kit from Roche. Primers were designed with the online tool provided by Roche and quantification was performed using the LightCycler 480 Probe Master System (Roche).
For ARRB localization assays, HUAECs were transfected with predesigned siRNAs against human APLNR (si APLNR) or non-targeting control scrambled siRNA (si SC) or the pcDNA3.1 ARRB-GFP plasmid using Opti-MEM and Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. A repeat transfection was performed the next day. To assess ARRB localization in response to laminar flow, HUAECs were grown in a m-slide VI 0.4 flow chamber (Ibidi, Germany) until confluence. Unidirectional laminar flow was generated by a computer-controlled setup containing an air-pressure pump and a two-way switching valve (Ibidi pump system) 59 . For quantitative analysis of ARRB localization in HUAECs, we exposed cells treated with si SC or si APLNR to laminar flow at 20 dynes cm À 2 for 15 min. We then analysed the plasma membrane of the GFP-positive cells.
Flow experiments and PTX treatment. HUVEC cells were grown in a m-slide VI 0.4 flow chamber (Ibidi) until confluence. Unidirectional laminar flow was generated by a computer-controlled setup containing an air-pressure pump and a two-way switching valve (Ibidi pump system).
For PTX experiments, HUVEC cells were pretreated with 100 ng ml À 1 PTX (P7208, Sigma) for 15 h then subjected to 18 h laminar flow with PTX (100 ng ml À 1 ) in the flow medium.
Whole mount in situ hybridization. For in situ hybridization, zebrafish embryos and larvae were fixed in 4% PFA overnight at 4°C and subsequently dehydrated in methanol and stored at À 20°C until required. Before hybridization, embryos were rehydrated to PBS/0.1% Tween and then digested in 10 mg ml À 1 Proteinase K (Roche) followed by fixation in 4% PFA in PBS/0.1% Tween. Embryos were washed in PBS/0.1% Tween, pre-incubated in hybrization buffer at 70°C for 4 h and then incubated with Dig-labelled RNA probes in hybridization buffer at 70°C overnight. After washing, the hybridized probes were detected with alkaline-phosphataseconjugated, alkaline-phosphatase-labelled anti-digoxigenin antibody (11093274910, Roche, dilution 1:1,000) at 4°C overnight and the signal was visualized with BM purple (1144207001, Roche).
To assess blood flow velocity, we performed time-lapse imaging of the trunk region of 84 h.p.f. wild type, klf2a À / À and klf2a À / À ;klf2b þ / À larvae using a Zeiss Spinning disc CSU-X1 confocal microscope with a high-speed camera and calculated blood flow velocity in the DA by measuring the time needed by RBCs to move 100 mm at the level of the fifth and sixth somites.
Statistical analysis. Statistical significance for paired samples and for multiple comparisons was determined by Student's t-test and one-way analysis of variance with Tukey's test, respectively. Data were considered statistically significant if the P-value o0.05 (***).
Data availability. The authors declare that data supporting the findings of this study are available within the article and its Supplementary Information files.