Wnt/β-catenin signaling regulates VE-cadherin-mediated anastomosis of brain capillaries by counteracting S1pr1 signaling

Canonical Wnt signaling is crucial for vascularization of the central nervous system and blood-brain barrier (BBB) formation. BBB formation and modulation are not only important for development, but also relevant for vascular and neurodegenerative diseases. However, there is little understanding of how Wnt signaling contributes to brain angiogenesis and BBB formation. Here we show, using high resolution in vivo imaging and temporal and spatial manipulation of Wnt signaling, different requirements for Wnt signaling during brain angiogenesis and BBB formation. In the absence of Wnt signaling, premature Sphingosine-1-phosphate receptor (S1pr) signaling reduces VE-cadherin and Esama at cell-cell junctions. We suggest that Wnt signaling suppresses S1pr signaling during angiogenesis to enable the dynamic junction formation during anastomosis, whereas later S1pr signaling regulates BBB maturation and VE-cadherin stabilization. Our data provides a link between brain angiogenesis and BBB formation and identifies Wnt signaling as coordinator of the timing and as regulator of anastomosis.

T he central nervous system (CNS) depends on nutrient and oxygen delivery from blood vessels during the development and homeostasis, but also requires protection from bloodborn toxins and pathogens. Endothelial cells (ECs) of CNS blood vessels acquire characteristic properties in order to fulfill the tasks of this blood-brain barrier (BBB), such as expression of a specific subset of junction molecules and nutrient transporters, downregulation of vesicular transport and establishment of cell-cell interactions within the neurovascular unit 1 . During a defined time window of embryonic development, molecular cues from neuronal and perineuronal tissues orchestrate CNS angiogenesis and barriergenesis [2][3][4] . The process of brain angiogenesis is well conserved in vertebrates: After acquiring a pre-sprouting signature (hereafter called pre-tip cell), these specified cells migrate out from the resident vessel, with tip cells guiding sprout formation, and invade into the neuronal tissue, where they form cell-cell contacts and anastomose with other sprouts or extracerebral vessels in order to establish circulatory loops 1,3,5 . In zebrafish (Danio rerio), hindbrain capillaries (intra-cerebral central arteries, CtAs) invade the brain parenchyma at around 32 h post fertilization (hpf) from the primordial hindbrain channel (PHBC), extend dorsally and connect ventrally to the basilar artery (BA) or laterally to other CtAs (Fig. 1a) [4][5][6][7] . At 48 hpf, most CtAs carry blood flow and by 72 hpf most of the respective BBB properties are established 8,9 . Studies in mice and zebrafish demonstrate β-catenindependent Wnt signaling through Wnt7a-Wnt7b to be essential for brain vascularization [10][11][12] . Additionally, Wnt signaling has been shown to be involved in the establishment of BBB characteristics, such as upregulation of tight junction components (e.g., Claudin 1 and 3) or nutrient transporters (e.g., Glut1) and decrease of transcellular transport processes (e.g., Plvap) 10,[13][14][15] . Recently, the Gpr124-Reck complex has been identified as an EC autonomous regulator of Wnt signaling essential for brain vascularization and BBB formation 12,[15][16][17][18][19][20] . Single-cell analysis of brain vascular development through live imaging approaches in zebrafish further revealed that the control of pre-tip cell function by the Gpr124-Reck complex operates within the parental vessels (PHBCs) at pre-invasive stages. This early function of Wnt signaling has so far largely precluded the analysis of Wnt signaling functions at later stages of CNS vascular development, including the invasion process itself and subsequent patterning events.
Vascular endothelial (VE)-cadherin is the major endothelialspecific member of the cadherin protein family and is involved in EC migration, cell-cell contact formation, anastomosis, and barrier formation 5,21,22 . Together with the EC-selective adhesion molecule a (Esama), VE-cadherin is essential for EC-EC recognition and contact formation during intersegmental vessel anastomosis 23,24 . Experiments using mice and tissue culture revealed that VE-cadherin is subject to extensive post-transcriptional regulation, including intracellular complex formation and trafficking, membrane localization as well as association in cell-cell junctions. This dynamic regulation of VE-cadherin in adherens junctions is not only crucial for vascular patterning, but also for the function of mature blood vessels, e.g., for maintenance of EC integrity or during leukocyte extravasation 25,26 , and can be regulated by Sphingosine-1-phosphate (S1p) signaling 27,28 .
ECs express S1p receptor (S1pr) 1, 2, and 3 29 and in zebrafish, two paralogues exist for S1pr3 (S1pr3a and S1pr3b). S1pr1 signaling promotes BBB integrity in mice 30 . Additionally, postnatal EC-specific knockout of S1pr1 induces loss of VEcadherin and vascular endothelial growth factor receptor 2 (VEGFR2) from the cell junctions 31 . To date, the contribution of neither S1pr nor VE-cadherin signaling has been addressed during brain capillary angiogenesis, leaving a missing link between brain EC angiogenesis and BBB formation.
In this study, we clarify the distinct requirements for Wnt signaling during brain angiogenesis. Whereas Wnt signaling is essential before sprouting to regulate yet to be defined early function of the future tip cells (pre-tip cells) within the parental vessel 12 , it is surprisingly dispensable during sprout elongation and migration, although it remains continuously active in the invading sprout. We show that Wnt signaling is regulating brain capillary anastomosis and lumen formation. Interestingly, Wnt signaling is crucial for VE-cadherin and Esama localization at cell-cell junctions in a transcription-independent manner. During active brain capillary angiogenesis, Wnt signaling counteracts S1pr signaling, which enables VE-cadherin-dependent anastomosis and lumen formation. In contrast, when angiogenesis is completed at later stages, S1pr signaling regulates BBB formation. Our data therefore reveal an important functional link between the intertwined processes of brain capillary angiogenesis and BBB formation.

Results
Wnt signaling is required for hindbrain capillary patterning. Studies in mice have shown that Wnt signaling is essential for hindbrain capillary angiogenesis 10,11,14 . Studies using zebrafish revealed that this control by Wnt signaling reflects an essential function in defining the pre-tip cell during early, pre-invasive angiogenic events within the parental vessels ( Fig. 1a 12 ). As a result, no hindbrain capillaries (central arteries, CtA) are formed in the absence of Wnt signaling. In order to bypass this early requirement, we blocked Wnt signaling pharmacologically using IWR-1 32 and generated transgenic lines, which allow for temporally and spatially controlled expression of a dominantnegative Tcf transcription factor. Treatment with Wnt signaling inhibitor IWR-1 after the pretip cell specification phase (29-72 hpf) resulted in severe vascular defects and hemorrhages in the brain, but not in the trunk vasculature ( Supplementary Fig. 1a), indicating that there are further requirements for Wnt signaling during brain angiogenesis. We therefore analyzed during which stages of brain capillary angiogenesis active Wnt signaling can be observed using confocal live imaging and two different transgenic Wnt signaling reporter lines, which express short-lived fluorophores controlled by Wntresponsive promoters: Tg(axin2BAC:Venus-Pest) mu288 and Tg (14TCF:loxP-STOP-loxP-dGFP) mu20233 . Both reporter lines showed robust fluorophore expression within the migrating CtA sprouts at 34 hpf ( Supplementary Fig. 1b,c), which we confirmed to be EC specific ( Supplementary Fig. 1d). We next performed time-lapse analysis using cre mRNA-injected Tg (14TCF:loxP-STOP-loxP-dGFP) mu202 embryos starting from 29 hpf (Fig. 1b, Supplementary movies 1, 2). We detected high Wnt reporter activity in the determined CtA pre-tip cells, which are forming within the PHBC. Despite a putative half-life of about 2 h, the destabilized GFP (dGFP) expression remained high in the emerging CtA sprouts during the following 5.5 h of development, indicating continuous Wnt signaling activity during CtA sprout migration and cell-cell contact formation (open arrow, Fig. 1b). With the onset of lumen formation about 2 h after cell-cellcontact formation, the dGFP signal decreased to a baseline level, which was maintained in the perfused CtAs, pointing to a postdetermination role of Wnt signaling.
After pre-tip cell formation, CtA angiogenesis proceeds by sprouting and sprout invasion into the brain parenchyma. We asked whether endothelial Wnt signaling regulates these processes.
We therefore generated a transgenic zebrafish line expressing a dominant-negative Tcf transcription factor fused to mCherry (mCherry-dnTcf) under the control of a heat shock inducible promoter (hsp70l). To restrict Wnt inhibition to ECs, we placed a loxP-flanked (floxed) STOP-cassette upstream of the mCherry-dnTcf coding sequence ( Supplementary Fig. 2a, c). This STOP cassette was removed only in ECs by mating to fish with endothelial-specific Cre recombinase expression (Tg(kdrl: cre) s898 ). Validation of Wnt inhibition by heat shock controlled mCherry-dnTcf expression is documented in Supplementary  Fig. 2.
To address whether Wnt signaling regulates sprout elongation or EC migration, we blocked Wnt signaling by EC-specific mCherry-dnTcf expression or IWR-1 treatment starting from 26 hpf and analyzed CtA sprouts at 32 hpf (Fig. 1c, d). Surprisingly, we did not observe differences in CtA sprout number or sprout length after Wnt signaling inhibition (Fig. 1c, d). Furthermore, sprout formation, behavior, and filopodia appearance were not affected by IWR-1 treatment (Fig. 1e, 4). Only the formation of long filopodia (>10 µm) was marginally decreased after IWR-1 treatment (Fig. 1f). Similarly, heat shockinduced EC-specific overexpression of mCherry-dnTcf (mCherry-dnTcf iEC ) did not result in differences in filopodia morphology, but instead in high filopodia motility, which resulted in slightly blurry images due to fast filopodia movement (Fig. 1g).
We conclude that after being required early within the parental vessel, Wnt signaling is continuously active during CtA angiogenesis, but dispensable for efficient sprout elongation and invasion of CtA tip cells into the brain parenchyma.

Wnt regulates brain capillary anastomosis and lumenization.
To address the role of the active Wnt signaling in the developing CtAs, we inhibited Wnt signaling by inducing mCherry-dnTcf expression in ECs at 26 hpf or by applying IWR at 29 hpf and analyzed CtA pattern and lumen formation at 48 hpf (Fig. 2a, b). We observed a reduced CtA number, reduced connections to the BA and ectopic connections between CtAs. CtA patterning appeared more disorganized, and CtAs had non-lumenized protrusions. To address whether the reduced number of CtAs ( Fig. 2a, b) correlated to a reduced number of EC within CtAs, we blocked Wnt signaling as before, and quantified the number of cell nuclei in Tg(fli1a:nGFP) y7 embryos. Both, treatment with IWR-1 or heat shock-induced EC-specific mCherry-dnTcf expression, reduced the EC number within CtAs (Supplementary Fig. 3a, b). However, blocking cell proliferation via the administration of aphidicolin and hydroxyurea (AHU) did not result in any patterning defects in CtAs, and the combined inhibition of Wnt signaling and proliferation did not aggravate the phenotype of Wnt deficiency ( Supplementary Fig. 3c). We therefore conclude that Wnt signaling is required for CtA patterning, but does not act via cell proliferation.
To understand the processes underlying the CtA patterning defects observed after Wnt signaling inhibition (Fig. 2a, b), we performed angiography following IWR-1 treatment (Fig. 2c). In control embryos, more than 80% of the CtAs were perfused at 48 hpf, whereas after IWR-1 treatment only about 25% of the CtAs carried the fluorescent tracer. We next performed time-lapse analysis following IWR-1 treatment or EC-specific mCherry-dnTcf expression using Tg(fli1a:lifeact-GFP) mu240 embryos to illustrate the actin cytoskeleton of ECs ( Fig. 2d-f). We focused on laterally connecting CtAs for better visualization and quantified the time from cell-cell contact until completion of lumen formation until 48 hpf (Fig. 2d). In control embryos, the average time from cell-cell contact to lumen formation is less than 2 h (Fig. 2d, e). After inhibition of Wnt signaling by IWR-1, we noted a drastic delay in lumen formation or even completely impaired lumenization within the observation window ( Fig. 2e; Supplementary movie [5][6][7][8]. Additionally, tip cells in CtA sprouts of IWR-1-treated embryos failed to establish stable cell-cell contacts, but instead continuously produced filopodia, leading to the formation of multiple cell-cell contact sites (asterisk in Fig. 2e, f). This phenotype was fully reproduced by EC-specific mCherry-dnTcf expression ( Fig. 2f; Supplementary movies 9, 10). However, the time difference was not as prolonged in the latter as in IWR-1-treated embryos, presumably due to the turnover time of mCherry-dnTcf, as mCherry is no longer detectable at the endpoint of analysis.
In summary, our analysis reveals that Wnt signaling is required for anastomosis of brain capillaries.
Block of Wnt signaling reduces VE-cadherin at cell junctions. To elucidate how Wnt signaling regulates CtA anastomosis, we investigated the abundance of the EC-specific adhesion molecules VE-cadherin and Esama at cell-cell junctions of the CtAs 23,24 . VE-cadherin, a member of the cadherin protein family expressed in ECs, has been shown to be crucial for EC contact formation during blood vessel anastomosis 5,22 . Esam belongs to the protein family of Junction adhesion molecules (Jam) and two paralogues exist in zebrafish, esama and esamb, of which esama is mainly expressed in the vasculature 24 . To address, whether Wnt signaling inhibition affects protein localization of VE-cadherin or Esama, we performed immunostaining for VE-cadherin, Esama and ZO-1 in 42 hpf-old embryos ( Fig. 3a-d). In control embryos (DMSO), VE-cadherin and Esama were expressed continuously along cell-cell junctions of ECs within CtAs and in anastomosis rings, which form at positions where two CtAs fuse (arrows in Fig. 3b). After IWR-1 treatment (Fig. 3b) or EC-specific mCherry-dnTcf expression (Fig. 3c, d), we found that the protein abundance of VE-cadherin, Esama, and ZO-1 at cell-cell junctions was dramatically reduced. Furthermore, we frequently observed clustered protein aggregates (stars in Fig. 3b) and the formation of many small adhesion spots (arrows in Fig. 3b) in Wnt-inhibited embryos, which is in line with our previous observations of multiple filopodia contacts (Fig. 2e, f). Hence, reduction and aberrant localization of VE-cadherin and Esama at the cell junctions could cause the anastomosis defects. To validate these findings in another model, we isolated primary brain endothelial cells (BECs) from microvascular fragments 34 of either P3 or adult mice. We were able to block Wnt signaling using 10 µM IWR-1 for P3 and 20 µM for adult brain ECs (Fig. 3e, Supplementary  Fig. 2e). To analyze the effect on VE-cadherin distribution we measured the length of the potential cell-cell contact area in form of opposing or touching membrane length and the actual Fig. 1 Wnt signaling is not required for migration of brain capillary sprout ECs. a Illustration of hindbrain angiogenesis in zebrafish embryos. Black box indicates localization of the hindbrain. Brain capillary (central artery, CtA) pre-tip cells can be detected within two bilateral primordial hindbrain channels (PHBCs, light gray) and sprout dorsally from PHBCs around 32 hpf. Between 32 and 36 hpf, CtA sprouts extend long filopodia and migrate in an arch toward the basilar artery (BA, dark gray). At 48 hpf CtAs have fused with either the BA or neighboring CtAs and carry blood flow. b Time-lapse analysis of a cre mRNA-injected Tg(14TCF:loxP-STOP-loxP-dGFP) mu202 embryo starting from around 29 hpf showed continuously active Wnt signaling in CtAs (arrowheads) before and during sprouting, during invasion, cell-cell contact (open arrowhead) and lumen formation. The bottom panel represents single channel images in inverted color for better visualization. c, d Inhibition of Wnt signaling by EC-specific dnTcf expression after heat shock at 26 hpf (mCherry-dnTcf iEC , c) or pharmacologically by IWR-1 treatment (d) resulted in normal CtA sprouting at 32 hpf. CtA sprout number or length of the sprouts was not affected by mCherry-dnTcf iEC expression in Tg(kdrl:GFP) s843 embryos (c: mCherry-dnTcf iEC : n = 11; mCherry iEC : n = 12) or IWR-1 treatment (d: DMSO: number n = 22, length n = 17; IWR-1: number n = 24, length n = 18). e Still images from time-lapse movies starting at around 28 hpf displayed normal CtA sprout formation in Tg(kdrl:ras-mCherry) s896 ; (fli1a:nGFP) y7 double transgenic embryos treated with IWR-1. f, g CtA sprout morphology was not affected by IWR-1 treatment (f) or dnTcf expression (g) in Tg(fli1a:lifeact-GFP) mu240 embryos. IWR-1 treatment slightly reduced the formation of long filopodia (>10 µm) compared to DMSO control (f: IWR-1: n = 15; DMSO: n = 15). Images are displayed in inverted color for better visualization. Confocal images show dorsal (b) or lateral views (c-g), anterior to the left. Values represent mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, Student's t-test; n, number of analyzed embryos; BA, basilar artery; CtAs, central arteries; ECs, endothelial cells; hpf, hours post fertilization; PHBC, primordial hindbrain channel; Scale bars: 30 µm (b), 50 µm (c-f), 6 µm (g) coverage of that area by VE-cadherin. We observed a marked reduction in the length of VE-cadherin junctions and the coverage of cell-cell contact sides by VE-cadherin in ECs isolated from P3 mice. Interestingly, Wnt signaling inhibition did not affect junctional VE-cadherin in adult brain ECs. We conclude that brain capillary ECs from P3 mice are capable of responding to Wnt signaling, presumably because these brain ECs are in a more angiogenic state, whereas ECs from adult mice gain mature BBB properties and are no longer sensitive to Wnt signaling inhibition.
We next analyzed how a complete loss of VE-cadherin affects CtA lumen formation using homozygous ve-cadherin ubs835 mutant embryos. In ve-cadherin ubs8 -deficient mutants, CtAs form and invade similarly to wild-type siblings (Supplementary movies 11,12), but fail to establish stable cell-cell contacts, continuously produce filopodia toward the ECs in neighboring CtAs or BA and fail to form continuous lumens ( Supplementary  Fig. 4a, b, Supplementary movies 11, 12), a behavior, which mimics our observations after Wnt signaling inhibition (Fig. 2e  We also analyzed lumen formation of CtAs in esama ubs1924 mutant embryos. Homozygous esama ubs19 mutants displayed fewer lumenized CtAs compared to heterozygous controls ( Supplementary Fig. 4c). We conclude that VE-cadherin is required for CtA anastomosis and that Esama contributes to efficient CtA lumen formation, which is in agreement with previous reports 23,24 . Therefore, we reason that the lumen formation defects caused by Wnt signaling inhibition were the result from decreased VE-cadherin and Esama protein levels at the cell-cell contact sites in CtAs.

Wnt does not regulate transcription of ve-cadherin and esama.
We hypothesized that the reduced protein levels of VEcadherin and Esama could result from reduced transcription in embryos lacking Wnt signaling. In order to analyze vecadherin and esama mRNA expression levels, we performed whole mount fluorescent in situ hybridization (FISH) after Wnt signaling inhibition by IWR-1 or EC-specific mCherry-dnTcf expression ( Fig. 4a, b, Supplementary Fig. 4e). We quantified the fluorescence intensity within CtAs at 32 or 42 hpf and normalized the fluorescence intensity to the respective CtA volume (generated from Tg(kdrl:GFP) s843 signal). Surprisingly, we did not find severe changes of the overall ve-cadherin or esama mRNA abundance in CtA ECs at 32 or 42 hpf (Fig. 4a, b, Supplementary Fig. 4e). At 32 hpf, ve-cadherin mRNA was slightly decreased following IWR-1 treatment and slightly increased following mCherry-dnTcf iEC expression. At 42 hpf, at which we detected reduced VEcadherin protein levels at the cell-cell junctions (Fig. 4b), vecadherin mRNA was increased using either of the Wnt signaling inhibition approaches. This indicates, that ve-cadherin mRNA levels in CtAs were not decreased (as expected from a putative Wnt target gene), but rather slightly increased over time, presumably due to a compensatory mechanism. The mRNA level of esama in CtA ECs was slightly decreased after Wnt signaling inhibition by IWR-1 or mCherry-dnTcf expression both at 32 and 42 hpf (Fig. S4e).
Additionally, to gain insights into transcriptional changes in CtAs mediated by Wnt signaling inhibition, we isolated Kaedephotoconverted CtA ECs from 1.5 days-old Tg(fli1a:Gal4) ubs3 ; (UAS:Kaede) rk8 embryos, which were treated with IWR-1 or DMSO (Fig. 4c, d) and performed gene expression profiling by RNA sequencing. We confirmed that ve-cadherin and esama expression levels were not regulated by Wnt signaling in sharp contrast to classical Wnt target genes (e.g., axin2 and lef1, Fig. 4d). We also analyzed the effect of Wnt inhibition on VEcadherin mRNA levels by qPCR in BECs isolated from P3 and adult animals. In line with our zebrafish data, we observed no transcriptional regulation of mouse VE-cadherin by Wnt signaling (Fig. 4e).
We therefore conclude that post-transcriptional mechanisms account for the changes in protein abundance at brain capillary cell-cell junctions following Wnt signaling inhibition.
Wnt signaling counteracts S1pr1 signaling in angiogenesis. To address how VE-cadherin localization at cell-cell junctions can be regulated during brain angiogenesis, we focused on Sphingosine-1-phosphate (S1p) signaling, as it is known to regulate EC integrity and behavior by modulating junction protein localization and cell contractility 27,28 . In particular, S1p-induced signaling mainly via S1p receptor 1 (S1pr1) has been shown in vivo and in vitro to increase recruitment of Cadherin protein family members, including VE-cadherin and N-cadherin, to the plasma membrane 31,38,39 . Additionally, S1p receptor 3 (S1pr3) was found to function cooperatively with S1pr1 in promoting adherens junction formation in ECs 38 , whereas S1p receptor 2 (S1pr2) signaling disrupted EC junctions, thereby increasing vascular permeability 40 . We confirmed that similar to what is seen in mice 30 , blocking of S1prs using the pharmacological antagonist VPC23019 (VPC) during barriergenesis (48-72 hpf) resulted in extravasation of red blood cells into the surrounding brain tissue, indicating breakdown of the BBB ( Supplementary Fig. 5a).
Our RNA sequencing data showed expression of s1pr1 and s1pr5a as well as very low level expression of s1pr2 and s1pr3b in zebrafish brain capillary ECs ( Supplementary Fig. 5b). In order to functionally dissect the contribution of these different S1pr's we validated various pharmacological antagonists by their potency to rescue phenotypes of S1pr overexpression (Supplementary Fig. 5c-g). In the zebrafish, VPC seems to antagonize the tested S1pr's (S1pr1, S1pr2, S1pr3a, and S1pr5a), whereas TY52156 (TY) acts specifically on S1pr1. In our hands the S1pr2 antagonist JTE013 (JTE) did not rescue S1pr2 overexpression (Supplementary Fig. 5c-g). Wnt signaling is crucial for brain capillary patterning and anastomosis. a, b Inhibition of Wnt signaling by IWR-1 treatment (a) or dnTcf expression (b) impaired CtA patterning. a IWR-1 treatment from 29 to 48 hpf reduced CtA number (n = 39) and the proportion of CtAs connecting to the BA (n = 39), but increased the proportion of interconnecting CtAs (n = 33), compared to DMSO treated embryos (CtA number: n = 35; BA connection: n = 35; interconnected: n = 29). b Expression of mCherry-dnTcf iEC through heat shock at 26 hpf reduced the number of CtAs (n = 40) and decreased the proportion of CtAs connecting to the BA (n = 32) or interconnecting (n = 32), compared to mCherry iEC control at 48 hpf (CtA number: n = 39; BA connection: n = 33; interconnected: n = 33). c Analysis of lumen formation by fluorescent tracer injection into the blood stream at 48 hpf. Treatment with IWR-1 decreased the proportion of lumenized CtAs (n = 16) compared to DMSO control (n = 18). d Schematic representation of CtA anastomosis (dorsal view,~36 hpf). CtA sprouts (gray) migrate, extend filopodia and form cell-cell contacts to neighboring CtA sprouts (asterisk) or the BA. Cell-cell contact formation triggers anastomosis, cell rearrangements and lumen formation through the connected sprouts in a distinct time window (Δt). e, f Inhibition of Wnt signaling drastically extended the time window from cell-cell contact formation until lumen formation. Still images from time-lapse movies embryos treated with DMSO (e: n = 58, N = 9) or IWR-1 (e: n = 33, N = 9) or expressing mCherry iEC (f: n = 36, N = 6) or mCherry-dnTcf iEC (f: n = 39, N = 8) after heat shock treatment at 26 hpf. Note that for events, where lumen formation was not completed until the end of the time-lapse recording (~48 hpf), the last measured time point was used for quantification. Confocal images show dorsal views (anterior to the left) of Tg(fli1a:lifeact-GFP) mu240 (a, b, e, f) or Tg (kdrl:ras-mCherry) s896 (c) embryos. Values represent mean ± SD (a-c) or ± SEM (e, f). *p < 0.05, **p < 0.01, ***p < 0.001, Student's t-test; n number of analyzed embryos (a-c) or n number of CtA fusion events analyzed; N, number of embryos analyzed (e, f). BA, basilar artery; CtAs, central arteries; ECs, endothelial cells; Scale bars: 50 µm To test the effects of S1pr inhibition on lumen formation and VE-cadherin localization, we next applied the antagonists of S1pr signaling during CtA angiogenesis at 29 hpf and assessed patterning and lumen formation at 48 hpf. In contrast to barriergenesis, treatment during brain angiogenesis with either VPC or TY did not affect CtA development and lumen formation in zebrafish (Fig. 5a). It is therefore likely that in wild-type embryos S1pr signaling is required only after the onset of blood circulation in CtAs, presumably through delivery of high amounts of its ligand S1p by the blood 41 . To our surprise, co-treatment of embryos with the Wnt signaling inhibitor IWR-1 and either VPC or TY increased the proportion of lumenized CtAs compared to single IWR-1 treatment and rescued CtA patterning and lumen formation (Fig. 5a). S1pr inhibition by VPC was also able to  rescue EC-specific loss of Wnt signaling in embryos overexpressing mCherry-dnTcf iEC (Fig. 5b).
We further analyzed, whether not only lumen formation, but also VE-cadherin and Esama levels at cell-cell junctions were rescued by inhibiting S1pr signaling in Wnt-depleted embryos. We observed that the reduced VE-cadherin and Esama localization at cell-cell contact sides and in anastomosis rings after IWR-1 treatment was restored by co-treatment with the S1pr1 antagonist TY and similar to control embryos when analyzed either by immunostaining for VE-cadherin and Esama or by live imaging of the Venus-fluorophore distribution of the VEcadherin-TS ( Supplementary Fig. 6). Hence, blocking of S1pr1 in parallel to IWR-1 treatment was sufficient to rescue VEcadherin and Esama protein levels at cell-cell junctions. In line with these results, inhibition of S1pr signaling by VPC restored VE-cadherin-positive junctions in IWR-1 treated P3 BECs (Fig. 5c).
As an alternative approach to pharmacological S1pr1 inhibition in zebrafish, we used transient CRISPR-Cas9-mediated s1pr1 knockdown in combination with IWR-1 treatment. This manipulation restored lumen formation of CtAs (Fig. 5d) similar to VPC or TY treatment (Fig. 5a). Furthermore, Wnt signaling inhibition was unable to affect VE-cadherin localization in S1pr1 iEC knockout BECs isolated from P3 mice (Fig. 5e).
Therefore, loss of S1pr1 signaling rescued Wnt depletion phenotypes, such as impaired lumen formation and VE-cadherin localization. Taken together, the studies reveal that Wnt signaling regulates brain capillary angiogenesis by counteracting S1pr1 signaling in zebrafish and mouse ECs, presumably to prevent premature barrier formation, which would impair brain angiogenesis (Fig. 5f).
Wnt signaling prevents S1pr1-mediated Rac1 activation. So far we have shown that Wnt signaling regulates brain capillary angiogenesis in an EC-specific manner by post-transcriptionally influencing VE-cadherin localization. Furthermore, the mislocalization of VE-cadherin caused by Wnt signaling inhibition was dependent on S1pr1 signaling (Fig. 5, Supplementary  Fig. 6). In order to understand the regulatory interaction of these two signaling pathways, we addressed at which level Wnt signaling interferes with S1pr1 signaling activation. We therefore stimulated S1pr1 signaling using a combination of two agonists (CYM5541 and CYM5542; validation for agonist activity, see Supplementary Fig. 5h). Overstimulation of S1pr1 signaling activity did not induce lumen formation defects (Fig. 6a) and therefore did not overcome the intrinsic counteraction by Wnt signaling. However, block of S1pr signaling in ve-cadherin or esama-deficient embryos did not result in a rescue of lumen formation defects ( Supplementary Fig. 7a).
Therefore, Wnt signaling antagonizes S1pr1 signaling downstream of receptor activation, but upstream of VE-cadherin and Esama function.
One potential mechanism that could account for the downregulation of VE-cadherin could be the regulation of VE-cadherin degradation or internalization. We therefore inhibited proteasomal degradation by MG132, dynamindependent internalization by Dynasore and lysosomal degradation by Chloroquine; however, none of these treatments rescued Wnt-deficiency ( Supplementary Fig. 7b, c). Interestingly, we also observed a mislocalization of VE-cadherin to seemingly intracellular vesicular structures in Wnt signaling-deficient BECs (Fig. 6b), indicating changes in cellular architecture rather than VE-cadherin degradation following Wnt signaling inhibition.
S1p-induced signaling in cultured ECs has been linked to activation of the Rho/Rac-pathway for stabilization of the cytoskeleton and recruitment of junction proteins to the cell membrane 38,42,43 . Similar to S1pr signaling, Rac1 activation as well as Rho inactivation is known to promote EC barrier function in mature blood vessels 44 ; however, their contribution to brain capillary angiogenesis has not been addressed yet. We therefore applied the pharmacological Rac1 inhibitor (NSC23766, NSC) or Rho inhibitor (CT03) alone or together with IWR-1 as described. Single treatment with either Rac1 or Rho inhibitors, did not affect CtA development or lumen formation. Indeed, co-treatment of IWR-1 and Rac1 inhibitor, but not of IWR-1 and Rho inhibitor, robustly restored the lumen formation defects caused by IWR-1 (Fig. 6c, Supplementary Fig. 8a).
Hence, Rac1 activation downstream of S1pr1 signaling is required and sufficient to mediate lumen formation defects caused by Wnt signaling inhibition.
In summary, we identified a temporal-specific function of Wnt signaling for brain capillary angiogenesis by post-transcriptionally and EC-specifically regulating junction protein localization upstream or at the level of Rac1 activity, but downstream of S1pr1 signaling. We suggest that EC-specific Wnt signaling counteracts S1pr1 signaling during CtA angiogenesis, presumably to prevent premature barrier formation processes, which would impair brain angiogenesis (Fig. 6f,g). Interestingly this mechanism seems to be specific to the role of Wnt signaling in regulating anastomosis and lumen formation, but not to the essential function of Wnt signaling during the earlier pre-tip cell phase 12,15 . Treatment of embryos lacking Wnt signaling following IWR-1 treatment or of gpr124 mutants with S1pr antagonists did not restore CtA sprout formation (Supplementary Fig. 8b, c).  In conclusion, we describe a temporal control of different steps of brain angiogenesis and BBB formation, which are differentially regulated by Wnt signaling. We also discover an interplay between Wnt and S1pr signaling during brain angiogenesis.

Discussion
We addressed the role of β-catenin-dependent Wnt signaling during early brain capillary angiogenesis in vivo. We found that (1) Wnt signaling is highly active in brain capillary sprouts, but after pre-tip cell formation it is dispensable for brain capillary invasion into the brain parenchyma. Instead, (2) Wnt signaling is required for brain capillary anastomosis. During this process, (3) Wnt signaling influences the post-transcriptional regulation and cell-cell junction localization of the EC-specific adhesion molecules, VE-cadherin, and Esama, by counteracting S1pr1 signaling (Fig. 6f, g).
We and others have shown that Wnt signaling is active in brain capillaries during brain angiogenesis (Fig. 1, 11,12,14 ). Recent reports demonstrate that Gpr124-Reck-mediated Wnt signaling is essential for brain vascularization, as homozygous grp124 or reck mutants are completely devoid of brain capillaries 12,15-17 . We hypothesized that Wnt signaling might be involved in the formation of pre-tip cells within the parental vessel, which most likely occurs at earlier developmental stages 12 . In line with this, a role of Wnt signaling for EC specification has already been suggested to drive cell fate from the Fli1a-positive precursor cell population into ECs at the expense of primitive erythrocytes 33 . At later developmental stages, Wnt signaling is required for subspecification of EC populations within the vasculature, which has been shown for hemogenic EC formation 45,46 and lymphatic cell fate specification 47 .
Here we demonstrate that Wnt signaling is continuously activated in CtAs during all stages of angiogenesis. Surprisingly, Wnt signaling is not involved in the process of CtA sprout invasion and filopodia formation (Fig. 1).
However, our data demonstrate that Wnt signaling is essential for brain capillary anastomosis (Fig. 2). During the anastomosis process in intersegmental vessels, VE-cadherin is required for cell-cell contact formation and together with Esama has been shown to be essential for efficient blood vessel fusion 23,24 . We detected reduced protein levels of VE-cadherin, Esama, and ZO-1 at the cell-cell junctions following Wnt signaling inhibition (Fig. 3), which presumably impair anastomosis and lumen formation. In line with these findings, knockouts of VE-cadherin (by ve-cadherin ubs8/ubs8 ) or Esama (by esama ubs19/ubs19 ) phenocopied CtA lumen formation defects (Supplementary Fig. 4). Surprisingly, neither ve-cadherin nor esama are transcriptional targets of Wnt signaling (Fig. 4). Similar to Hupe and colleagues, we detected mildly increased amounts ve-cadherin mRNA when Wnt signaling inhibition persists for longer time intervals (>12 h), indicating a compensatory response to the reduced protein levels of VE-cadherin at cell-cell junctions 48 . We therefore conclude that VE-cadherin and Esama downregulation occurs via a posttranscriptional mechanism, which interferes with their membrane localization. Using a series of inhibitors, we excluded internalization as well as proteasomal and lysosomal degradation as potential mechanisms (Supplementary Fig. 7). Previously, Wnt signaling has been shown to be required for the junction localization of Claudin 1 and 3 in murine brain ECs during BBB formation, but no direct transcriptional regulation has been detected 13,14 . As there have been no mechanisms proposed of how Wnt signaling could regulate these junctional molecules, we speculate that they might also be downregulated posttranscriptionally as a consequence of imbalanced S1pr signaling.
The concept that S1pr signaling regulates EC integrity by modulating translocation of junction proteins and actomyosin contractility is widely accepted 27,28 . So far, S1pr signaling has not been linked to brain capillary angiogenesis and our data also indicate that the endogenous S1pr signaling is neither required for CtA tip cell specification ( Supplementary Fig. 5) nor for other aspects of CtA angiogenesis (Fig. 5). However, as recently shown in mice 30 , we demonstrate that S1pr signaling indeed regulates BBB formation and integrity in zebrafish ( Supplementary Fig. 5). Moreover, our data points to a novel regulatory mechanism of Wnt signaling during brain capillary angiogenesis, which is to counteract S1pr-induced BBB formation in order to ensure efficient anastomosis (Fig. 5). We propose that Wnt signaling suppresses S1pr1-mediated signaling in CtAs during early stages of brain vascularization. After lumen formation, Wnt signaling activity within brain capillaries decreases to a low baseline level (Fig. 1 14 ). In lumenized vessels, activation of S1pr signaling would occur and be most likely enhanced by increased supply of the ligand S1p through the blood circulation 41 , which would enable functional S1pr signaling for BBB formation and maintenance ( Supplementary Fig. 5).
Additionally, over-activation of Rac1 could destabilize junction molecule complexes, as has been shown for initial stages of pharyngeal pouch formation in zebrafish 50 . In this study, the authors suggested that Wnt signaling mediates activation of Rac1 by an unknown mechanism 50 ; however, a possible role of S1pr signaling has not been addressed. The authors also observed a Siblings reduced membrane localization of the adherens junction molecules E-Cadherin and Alcama after Wnt signaling inhibition by morpholino-mediated knockdown of Disheveled, which is in line with our findings during brain angiogenesis. We therefore speculate that similarly to brain capillary ECs, Wnt signaling regulates Rac1-mediated S1pr signaling and thereby modulate junction localization of adhesion molecules also during pharyngeal pouch formation. During brain neo-vascularization, we and others report divers functions of Wnt signaling, which regulate many steps of brain capillary angiogenesis and BBB formation and crosscommunicate with other signaling cues 51 . First, Wnt signaling is required for the formation of the brain capillary pre-tip cells within the parental vessel, which is dependent on a functional Gpr124-Reck co-receptor complex 12,[15][16][17][18] . Second, Wnt signaling limits S1pr signaling during brain capillary sprouting and promotes brain capillary anastomosis through efficient VE-cadherin and Esama localization to the membrane (as shown in this work). And third, Wnt signaling leads to the induction of BBB components, such as glucose transporter Glut1 and Claudin family members, already during brain angiogenesis in order to subsequently establish a functional BBB after brain angiogenesis has been completed 10,26,48,52 and sealing of the BBB is likely to be promoted by S1pr signaling after the onset of blood circulation ( Supplementary Fig. 5 30,41 ). The complexity of the signaling network regulating EC behavior during brain angiogenesis and BBB formation is far from being understood. Our study provides a regulatory function of Wnt signaling during brain angiogenesis, which was only unraveled by interfering with Wnt signaling in a temporally and spatially restricted manner in combination with in vivo time-lapse imaging. We were able to identify a posttranscriptional level of regulation, which is essential for cell-cell junction formation and the modulation of EC barrier establishment or tightness. Therefore, interfering with Wnt or S1p signaling during treatment of infectious and neurodegenerative diseases as well as anti-cancer therapy remains a major clinical obstacle as adverse effects and cross-reactions are not well predictable 30,53-55 . The FDA approved sphingosine analog and agonist of S1pr1 and S1pr3-5 called Fingolimod (FTY720/Gilenya) has been reported to cause side effects of increased of vascular permeability in the lung and cardio-vascular complications. Moreover, cases with neuroencephalitis were noted when the drug was applied at higher doses. Therefore, the evaluation of the potential clinical impact for CNS development and pathology of these and future substances requires a greater understanding of signaling cues regulating EC properties and behavior.

Methods
Zebrafish husbandry and strains. Zebrafish (Danio rerio) husbandry and embryo maintenance were carried out under standard conditions at 28.5°C 56 . Embryonic developmental stages were determined according to ref. 57 .
The Tg(UAS:CA-Rac1) mu211 and Tg(UAS:DN-Rac1) mu212 lines were generated by injection of plasmids, which were a gift from Gage Crump 50 , and according to standard transgenesis protocol 68 . Both constructs harbor α-crystallin:cerulean, inducing blue fluorescence in the lenses at 48 hpf. To overexpress CA-Rac1 and DN-Rac1 in the vasculature the respective fish were mated to Tg(fli1ep:Gal4ff) ubs3 ; (kdrl:Hras-mCherry) s896 . Please note, that in the case of DN-Rac1 overexpression, all embryos were presorted for established circulation in PHBC and BA prior to analysis by confocal microscopy as we most likely obtained multiple integrations of the construct and thereby varying phenotypes in the offspring embryos.
For both lines, standard transgenesis was performed 68 .
Microinjections. For ubiquitous removal of the floxed STOP cassette by Cre recombinase, 2 nL of cre mRNA (200 pg/nL) were injected at single cell stage. Therefore the cre coding sequence 69 was cloned into pCS2+ and transcribed into mRNA by using the Sp6 mMessage mMachine kit (Ambion) following NotI-digest as previously described 33 .
Heat shock and pharmacological treatments. In order to induce gene expression by heat shock, 26 hpf-old dechorionated embryos were incubated 40 min at 39°C in pre-warmed E3 medium and analyzed at indicated developmental stages. For validation of the Tg(hsp70l:loxP-STOP-loxP-mCherry-dntcf) mu201 and Tg(hsp70l: loxP-STOP-loxP-mCherry) mu279 lines, embryos were incubated at 10 hpf and 15 hpf for 1 h at 39°C and analyzed at 24 hpf.
The antisense probes for ve-cadherin 80 and esama 24,81 were generated from plasmid templates using T3 or Sp6 polymerase and DIG-labelled UTP (2 h, 37°C) and purified by precipitation 81 .
Photoswitching and FACS isolation of brain capillary ECs. Photoswitching of Tg (fli1:gal4) ubs3 ;(UAS:kaede) rk8 embryos was performed using a Zeiss LSM 710 confocal microscope (Carl Zeiss, objective lenses: Plan-Apochromat ×20/0.8 M27) after embryo anesthesia with a low dose of tricaine. Embryos were mounted laterally in 1% low-melting agarose in glass-bottom Petri dishes (MatTek Corporation, Ashland, MA). Photoconversion of Kaede fluorescent protein was performed by scanning the selected region of interest with a 405 nm diode laser (100% laser, 5 iterations, 50 s). After photoconversion, embryos were manually removed from the agarose and placed in fresh egg water (0.3 g/L Instant Ocean Salt, 75 mg/L CaSO 4 ; 1 mg/L Methylene Blue) additionally supplemented with N-Phenylthiourea (30 mg/ L, Sigma-Aldrich) and covered with aluminum foil to protect from the light before trypsin treatment. After several washes in HBSS (Hank's Balanced Salt Solution, Gibco) without Ca 2+ /Mg 2+ , zebrafish embryos dissociation was performed at 28.5°C during 30 min using 2 mL of trypsin-LE select (Gibco). Dissociation was stopped by added 200 µL of FBS solution. After centrifugation, pellet was resuspended with HBSS with Ca 2+ /Mg 2+ +5% FBS and filtered. Photoswitched ECs were isolated on FACS Aria II. Cells were immediately centrifuged, frozen by liquid nitrogen and stored at −80°C.
RNA sequencing of zebrafish brain capillary ECs. For RNA-seq, RNA was isolated from FACS sorted zebrafish CtAs ECs cells using the miRNeasy micro Kit (Qiagen) combined with on-column DNase digestion (DNase-Free DNase Set, Qiagen) to avoid contamination by genomic DNA. RNA and library preparation integrity were verified with a BioAnalyzer 2100 (Agilent) or LabChip Gx Touch 24 (Perkin Elmer). RNA amount was adjusted on number of isolated cells by FACS and~250-500 pg total RNA was used as input for SMART-Seq® v4 Ultra® Low Input RNA Kit (Takara Clontech) for cDNA pre-amplification. Obtained fulllength cDNA was checked on LabChip and fragmented by Ultrasonication by E220 machine (Covaris). Final Library Preparation was performed by Low Input Library Prep Kit v2 (Takara Clontech). Sequencing was performed on the NextSeq500 instrument (Illumina) using v2 chemistry, resulting in minimum of 30M reads per library with 1 × 75 bp single end setup. The resulting raw reads were assessed for quality, adapter content, and duplication rates with FastQC (available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Reaper version 13-100 was employed to trim reads after a quality drop below a mean of Q20 in a window of 10 nucleotides 82 . Only reads between 30 and 150 nucleotides were cleared for further analyses. Trimmed and filtered reads were aligned versus the Zebrafish genome version DanRer10 (GRCz10.87) using STAR 2.4.0a with the parameter "-outFilterMismatchNoverLmax 0.1" to increase the maximum ratio of mismatches to mapped length to 10% 83 . The number of reads aligning to genes was counted with featureCounts 1.4.5-p1 tool from the Subread package 84 . Only reads mapping at least partially inside exons were admitted and aggregated per gene. Reads overlapping multiple genes or aligning to multiple regions were excluded. The Ensemble annotation was enriched with UniProt data (release 06.06.2014) based on Ensembl gene identifiers (activities at the Universal Protein Resource (UniProt)).
IWR-1 (Sigma-Aldrich) and VPC23019 (VPC, Cayman Chemical) were dissolved in DMSO to a 10 mM concentration. Inhibition of Wnt signaling was achieved after 48 h treatment with 10 or 20 μM IWR-1 for endothelial cells from young pups or adult animals, respectively. Alternatively, 10 μM IWR-1-treated cells from young pups were exposed in parallel to 2 μM VPC to assess the effect of S1pr1 signaling blocking upon Wnt inhibition. Vehicle (DMSO) treated was used in all experiments as control.
RT-qPCR of mouse brain primary ECs. After inhibitor treatment, cells growing in 24-well cell culture plates were collected in 350 μL RLT Plus Buffer (Qiagen, 1053393) supplemented with 1% β-mercaptoethanol (Sigma, M6250) and total RNA was isolated using the RNeasy Plus Micro Kit (Qiagen, 74034) following the manufacturer's instructions. Whole RNA was assessed in terms of quantity and quality using Bioanalyzer 2000 (Agilent), reverse transcribed and converted to complementary DNA (cDNA) using the iScript cDNA synthesis kit (BioRad, 170-8890). The following FAM-conjugated TaqMan gene expression probes (all from Thermo Fisher Scientific) were used: VE-cadherin (Cdh5, Mm00486938_m1), Axin2 (Mm01265780_m1), Lef1 (Mm01310389_m1), and Ccnd1 (Mm00432360_m1). VIC-conjugated Actb (4352341E) TaqMan probe was used to normalize gene expression. Quantitative PCR (qPCR) reactions were performed on a CFX96 Touch Real-Time PCR Detection System (BioRad) using the Sso Advanced Universal Probes Supermix (BioRad, 1725281). All relative gene expression analyses were performed using the 2 −ΔΔCt method in a minimum of four animals per group with triplicate reactions for each gene evaluated.
Confocal microscopy. Confocal microscopy was performed using LSM780 and LSM880 microscopes (Carl Zeiss Microscopy GmbH; objective lenses: Plan-Apochromat ×20/0.8; LD C-Apochromat ×40/1.1 W Korr M27). For filopodia analysis in zebrafish at high resolution, images were acquired with the LSM880 Airyscan module. For life imaging of VE-cadherinTS embryos, the Online Fingerprinting mode of the LSM880 microscope was utilized. In general, PFA-fixed or living zebrafish embryos were embedded in 0.3% agarose, which was dissolved in E3 medium and additionally supplemented with N-Phenylthiourea (30 mg/L, Sigma-Aldrich) and Tricaine (19.2 mg/L, Sigma-Aldrich) for living embryos as previously described 59 . For time-lapse analysis, the agarose was additionally supplemented with IWR-1 or DMSO and a stable temperature of 28.5°C was maintained using a heating chamber. Assembly of confocal stacks and time-lapse movies was conducted using Imaris 8/9 software (Bitplane). Quantification of signal intensity and volume was done using the Imaris surface-rendering algorithm.
Statistical analysis. For all quantifications, statistical analysis was performed using Prism6 software (GraphPad). Graphs show mean ± standard deviation (SD) or standard error of the mean (SEM) and p-values were calculated using unpaired two-tailed Student's t-test or one-way ANOVA for single and multiple comparisons as indicated for each experiment.