Reck enables cerebrovascular development by promoting canonical Wnt signaling

The cerebral vasculature is essential for brain development, activity and homeostasis (Vallon et al., 2014). It supplies the metabolic needs of this organ, which consumes a fifth of the oxygen and a quarter of the glucose used by the body (Mergenthaler et al., 2013; Rolfe and Brown, 1997). Assembly of the cerebral vasculature involves endothelial cells initially found within vasculogenic perineural vessels. Some of these cells form angiogenic sprouts that invade the brain, yielding intracerebral vessels that branch and interconnect with the perineural vasculature (Ruhrberg and Bautch, 2013). Morphogenesis of the cerebral vasculature is accompanied by barriergenesis – the process of endothelial differentiation yielding the blood-brain barrier (BBB). Key BBB hallmarks include contiguous intercellular tight junctions, lack of fenestrations and selective or enriched expression of particular tight junction components and nutrient/efflux transporters (Hagan and Ben-Zvi, 2014; Obermeier et al., 2013; Siegenthaler et al., 2013). The BBB regulates the brain's extracellular ion balance, facilitates nutrient transport into the parenchyma, prevents the entrance of harmful molecules and metastatic cells, and enables neuroepithelial immune surveillance (Carson et al., 2006; Hawkins and Davis, 2005; Mergenthaler et al., 2013; Ousman and Kubes, 2012; Rolfe and Brown, 1997). Cerebral angiogenic growth and barriergenesis are modulated by neuroepithelial cues that activate endothelial signaling cascades (Hagan and Ben-Zvi, 2014). These include canonical Wnt or Wnt/β-catenin signaling, which promotes both aspects of vascular development specifically in the central nervous system (CNS), and global regulators of angiogenic growth that have positive (VEGF and SDF1) or negative (Notch and TGF-β) roles (Arnold et al., 2014; Bussmann et al., 2011; Fujita et al., 2011; Gridley, 2010; Larrivee et al., 2012; Mackenzie and Ruhrberg, 2012; Masckauchan and Kitajewski, 2006; Reis and Liebner, 2013). Reck (reversion-inducing cysteine-rich protein with Kazal motifs) is a dimeric multi-domain glycosylphosphatidylinositol (GPI)-anchored protein isolated as a tumor suppressor whose overexpression normalizes the aberrant morphology of transformed fibroblasts. Reck is an inhibitor for metalloproteinases (MPs) of the MMP (matrix metalloproteinase) and ADAM (A disintegrin and metalloproteinase) families (Chang et al., 2008; Hong et al., 2014; Nagini, 2012; Omura et al., 2009). MPs promote cell migration by weakening the mechanical barrier properties of the extracellular matrix (ECM) and reducing cell-cell adhesion by destruction of intercellular junctions (Page-McCaw et al., 2007; Seals and Courtneidge, 2003). For example, fly Reck limits basement membrane degradation (Srivastava et al., 2007) and glioma migration (Silveira Correa et al., 2010). MPs also modulate signaling pathways by cleaving ligands, receptors, ECM components (Page-McCaw et al., 2007; Seals and Courtneidge, 2003) and adherens junctions (Rims and McGuire, 2014). They can also act non-proteolytically (Mantuano et al., 2008; Mori et al., 2013). Accordingly, Reck also modulates cell signaling with MP dependency (Miki et al., 2010; Muraguchi et al., 2007). Finally, consistent with its multi-domain structure, Reck associates with the ERBB2 receptor to block its activity in an MP-independent fashion (Hong et al., 2011). Reck modulates the development of forelimbs, dorsal root ganglia neurons (DRG), brain (Muraguchi et al., 2007; Park et al., 2013; Prendergast et al., 2012; Yamamoto et al., 2012) and vasculature. Reck-knockout mice show perineural vascular plexus disorganization and reduced intracerebral vascularization (Chandana et al., 2010; Miki et al., 2010; Oh et al., 2001). In this study, we provide key mechanistic insights into how Reck modulates cerebrovascular development at both the cellular and molecular levels.

In a zebrafish genetic screen (Shaw et al., 2006), we isolated the recessive-lethal mutant nft ^y72 for its brain-specific vascularization deficit. Although nft ^y72 mutants lacked intracerebral CtAs (Fig. 1A,B), its other cephalic blood vessels formed and carried circulation normally (Fig. 1A,B and Movies 1-4). Importantly, in the mutants, gross cerebral organization was undisturbed (Fig. S1). Cardiac contractility appeared normal (Movies 11,12). In the trunk, blood vessels formed and functioned properly (Fig. 1C,D and Movies 5-8) and the lymphatic thoracic duct was patterned correctly (Fig. 1F,H). However, the neural crest-derived DRG were missing (Fig. 1E,G). The shape, patterning and size of the head and body were unaffected (Fig. 1I-L), except for minor jaw defects (Prendergast et al., 2012).

nft ^y72 maps to a genetic interval spanning Df(Chr24:reck)^w15 (chromosome 24 deficiency removing reck and other genes), which was isolated as a sensory deprived (sdp; now reck) allele in a screen for DRG-deficient mutants. All four sdp alleles are recessive lethal and genetic nulls (Prendergast et al., 2012). Given the positional and/or phenotypic similarities between nft ^y72, sdp (Prendergast et al., 2012) and Reck-knockout (Reck^−/−) mice (Chandana et al., 2010) we tested nft ^y72 and sdp for complementation. We found that nft ^y72/sdp transheterozygotes and both nft ^y72 and sdp homozygotes have large CtA and DRG deficits (Fig. S2A-H; Table S1). To compare nft ^y72 and Df(Chr24:reck)^w15 with respect to additional cardiovascular phenotypes, see Fig. S2E-I, Fig. S3 and Movies 7-14.

DNA sequencing from nft ^y72 revealed a G-to-A transition at position 761 of the 2868 nt open reading frame of reck (Prendergast et al., 2012), yielding a missense, non-conservative substitution of the evolutionarily conserved Cys²⁵⁴ residue to Tyr at the fourth cysteine knot 4 (CK4; Fig. 2A). A similar Cys substitution occurs in sdp^w13 at CK1 (Prendergast et al., 2012; Fig. S3). To confirm that this reck transition is the causative mutation in nft ^y72 mutants we provided exogenous wild-type (WT) reck mRNA to one-cell stage embryos from nft ^y72/+ in-crosses (see Prendergast et al., 2012). This treatment rescued the CtA and DRG deficits of nft ^y72 (henceforth called reck ^y72) mutants without yielding a surplus of these structures (Fig. 2B-F), indicating that reck plays permissive roles in the formation of CtAs and DRG. Together with the results of experiments using tissue-specific gene expression to rescue CtA formation in reck ^y72 mutants (Fig. 3, Figs S6, S11), the identical intracerebral vascularization deficits of reck ^y72 and Df(Chr24:reck)^w15 mutant embryos (Fig. 3J) and the differential subcellular localization of the WT and Reck^y72 mutant proteins (Fig. 2G-J), our observations imply that reck ^y72 is an amorphic allele of reck.

Secretion of disulfide-bridged proteins is often impaired by Cys substitutions (Bodin et al., 2007; Boute et al., 2004; Claffey et al., 1995; Halliday et al., 1999; Mason, 1994; Schrijver et al., 1999). Thus, we hypothesized that the inactivity of Reck^y72 is due to failure to reach the outer cell surface (see Simizu et al., 2005). We thus assayed the cell surface localization of epitope-tagged (3×FLAG and 2×HA) WT (Reck) and mutant (Reck^y72) zebrafish proteins in non-permeabilized immunofluorescently stained cells (Imhof et al., 2008). In 293T cells epitope-tagged Reck was detected at the cell surface (Fig. 2G-H; these proteins are active: Fig. 3, Fig. S6). By contrast, Reck^y72 was undetectable at the cell surface (Fig. 2I,J). Quantification of COS7 cell lysates showed that the localization disparity was not due to differential abundance (Fig. 2K,L). We also co-expressed differentially tagged versions of Reck and Reck^y72 and found that the WT form still localized correctly. Our findings thus provide a simple molecular explanation for the recessive amorphic nature of the reck ^y72 allele: Reck^y72 fails to reach the outer cell surface without disrupting the targeting of its WT counterpart.

To elucidate the endothelial cellular bases of the intracerebral vascularization deficit of reck mutants (reck⁻) we exploited the advantages of the hindbrain (Hb) vasculature as a model for cerebrovascular development, as we reported previously (Ulrich et al., 2011). The WT Hb harbors both extracerebral (perineural) and intracerebral vessels. These lie, respectively, ventral to the Hb or inside it (Fig. 4A). The extracerebral vasculature comprises the two lateral primordial hindbrain channels (PHBCs) and, at the midline, the posterior communicating segments (PCSs) and basilar artery (BA). The PHBCs communicate with the midline vessels via arteriovenous connections (avcs). The intracerebral vessels or CtAs sprout dorsally from the PHBCs into each rhombomere center. The assembly of the Hb vasculature at 24-72 hours post-fertilization (hpf) primarily involves endothelial cell migration and follows a reproducible sequence. The PHBCs and PCSs form first (Fig. 4B), then, PHBCs launch ventral sprouts towards the midline, which coalesce into the BA (Fig. 4C,D). While the BA forms, the CtAs emerge from the PHBCs, penetrate the Hb and connect to the PCS and BA (Fig. 4D,E). Most BA-forming sprout remnants are lost by 48 hpf; those that persist become avcs (Fig. 4D,E; see Bussmann et al., 2011; Corti et al., 2011; Fujita et al., 2011; Fukuhara et al., 2014; Isogai et al., 2001; Ulrich et al., 2011).

At the cellular level, the intracerebral vascularization deficit of reck⁻ could be due to defects in the abundance and/or distribution of endothelial cells (Fig. 4F-K). Quantification of these parameters revealed that endothelial abundance was slightly reduced at 36 hpf but not at 50 hpf (Fig. 4J), consistent with a minor transient delay in the mutant's vascular development and eliminating the possibility that reduced endothelial cell abundance (as a result of impaired cell specification, proliferation and survival) causes the lack of CtAs in reck⁻. By contrast, endothelial cell distribution was abnormal in the mutant: the PHBCs (but not the BA) were hyperplastic (Fig. 4J). We also found that reck ^y72 mutants had overabundant avcs (Fig. 4I,K,M), reminiscent of the murine perineural vascular plexus disorganization of Reck^−/− embryos (Chandana et al., 2010). Notably, the mutant avcs harbored only few cells (Fig. 4J). Together with the results of our time-lapse imaging showing that the PHBCs in reck ^y72 mutants failed to form CtA sprouts (Fig. S4 and Movies 15, 16), our observations indicate that the primary cellular defect leading to the intracerebral vascularization deficit of reck⁻ is a dramatic reduction in CtA-forming endothelial cell emigration from the PHBCs, which, in turn, induce the hyperplasia of the latter (Fig. 4L,M).

Blood vessel perdurance and circulatory flow can be linked (Bussmann et al., 2011; Chen et al., 2012; Fish et al., 2008; Kochhan et al., 2013; Nicoli et al., 2010; Watson et al., 2013). For instance, drug-induced inhibition of the heartbeat reduces the abundance of avcs, which suggests that in WT embryos the few avcs that persist do so because they were carrying robust flow (Corti et al., 2011; Helisch and Schaper, 2003). Hence, we hypothesized that in reck⁻, the CtA deficit increases circulatory pressure through extracerebral vessels (which have robust flow: Movies 1, 2, Table S1), secondarily enhancing maintenance and/or formation of avcs (Fig. S4, Movies 1, 2). We thus asked whether the overabundance of avcs in reck ^y72 mutants is suppressed by lack of circulatory flow. We abrogated cardiac contractility with silent heart (sih^b109), a recessive mutation that inactivates cardiac troponin-t2a (Sehnert et al., 2002). Unlike drug-based inhibition of circulatory flow, genetic abrogation of circulation increases the number of avcs, even in reck ^y72 mutants, and has little impact on the abundance of CtAs (Table S2 and Fig. S5; Bussmann et al., 2011; Corti et al., 2011; Fujita et al., 2011; Fukuhara et al., 2014; Isogai et al., 2001; Ulrich et al., 2011). Thus, reck limits the number of avcs, even without flow, strongly suggesting that the overabundance of avcs in reck ^y72 mutants is unrelated to circulatory pressure gains.

To visualize the cephalic expression of reck during cerebrovascular development we performed RNA in situ hybridization chain reaction (HCR; Choi et al., 2014) using 36 and 48 hpf WT embryos. Consistent with prior reports (Chandana et al., 2010; Miki et al., 2010; Prendergast et al., 2012), we found that in WT embryos, reck is expressed in cerebral vessels and neural crest derivatives. For example, at 48 hpf, reck expression was found in the MtA, PHBCs and CtAs, as well as the branchial arches (BAx; see Fig. 5A-C). On the basis of these observations, we hypothesized that development of CtAs requires reck activity in the cerebral endothelium. To test this hypothesis, we performed cell transplants between WT donors and reck ^y72 hosts, making chimeras with mosaicism in the Hb and/or its vessels (Carmany-Rampey and Moens, 2006). Although cell transplants rarely target the Hb endothelium, despite yielding frequent mosaicism in both the Hb and the trunk vasculature, we found that CtAs, like other vessels (Zygmunt et al., 2011), were of mixed clonal origin (Fig. 4D-F). Moreover, WT endothelial cells formed chimeric CtAs in reck ^y72 hosts (Fig. 5G; n=3 chimeras), consistent with the notion that intracerebral vascularization requires endothelial reck activity in a non-cell autonomous manner. Accordingly, mosaic endothelial expression of WT Reck (but not Reck^y72 or EGFP) was sufficient to rescue the CtA deficit of reck ^y72 mutants (Fig. 3, Figs S6, S11).

The intracerebral vascular deficit of reck⁻ best fits the model that Reck is a positive modulator of pathways that foster cerebral vascular development (Wnt, VEGF, SDF1) and/or an inhibitor of cascades that antagonize it (Notch, TGF-β). Since our experiments eliminated the latter possibility, we focused on testing the hypothesis that Reck promotes Wnt, VEGF and/or SDF1 signaling. We visualized the expression of molecular markers in WT and reck ^y72 mutants, including components and/or targets of these pathways. We found no obvious differences between genotypes in the neuroepithelial expression of genes encoding Wnt (wnt1), VEGF-A (vegfaa and vegfab) and SDF (sdf1b; also known as cxcl12b) ligands and Reck-targeted MPs (mmp2 and mmp14a). Analysis of the expression of pan-endothelial genes [fli1a, tie1 and ve-cdh (also known as cdh5)] and markers of specific cerebral vessels (dab2, dll4 and cxcr4a) failed to reveal any obvious expression abnormalities, beyond the anticipated CtA labeling deficit in reck ^y72 mutants (Amoyel et al., 2005; Bai et al., 2005; Brown et al., 2000; Bussmann et al., 2011; Fujita et al., 2011; Janssens et al., 2013; Larson et al., 2004; Lyons et al., 1998).

However, consistent with the cerebrovascular role and expression of reck (Fig. 1A,B, Figs 3, 5, Fig. S6), we found that reck ^y72 mutants displayed abnormalities in the expression of both artificial and endogenous targets of the canonical Wnt signaling pathway in the cerebral endothelium (Figs 6, 7). For example, in WT embryos, the fluorescent reporter of canonical Wnt signaling Tg(7xTCF-Xla.Siam:GFP)^ia4 (Moro et al., 2013) was expressed both in brain vessels (PHBCs and nascent CtAs at 36 hpf; CtAs at 48 hpf) and non-vascular cephalic tissues. By contrast, reck ^y72 embryos displayed a selective loss of cerebrovascular expression of Tg(7xTCF-Xla.Siam:GFP)^ia4 (Fig. 6A-N″).

In addition, reck ^y72 mutants show altered expression of two genes that are endogenous targets of canonical Wnt signaling in the mammalian cerebral endothelium and serve as markers of BBB differentiation: glucose transporter 1 (glut1, also known as slc2a1a/b) and plasmalemma vesicle-associated protein (plvap, also known as vsg1) (Fig. 7). Glut1 is a BBB component with Wnt-activated expression and plvap encodes a Wnt-repressed marker of fenestrated endothelium that highlights the immature BBB (Daneman et al., 2009; Hallmann et al., 1995; Liebner et al., 2008; Posokhova et al., 2015; Qian et al., 2005; Tam et al., 2012; Zhou and Nathans, 2014; Zhou et al., 2014). We found that at 72 hpf, Glut1 immunostaining labeled the MtA, PHBCs and CtAs of WT embryos (Fig. 7A-D″), but in reck ^y72 mutants, Glut1 was undetectable in the PHBCs, despite the fact that it remained in the MtA (Fig. 7E-G″). Conversely, plvap mRNA was found in the dorsal aspect of the CtAs, but not the extracerebral PHBCs in WT embryos at 48 hpf (Fig. 7H). However, in reck ^y72 embryos, the PHBCs displayed ectopic plvap expression (Fig. 7I).

We next asked whether the intracerebral vascularization deficits and aberrant expression of BBB-related markers found in reck ^y72 mutants can be phenocopied by inactivation of canonical Wnt signaling in otherwise WT embryos. To do this, we used transgenes that enable heat-induced global inhibition of canonical Wnt signaling by distinct mechanisms. Tg(hsp70l:Xla.TCFΔC-EGFP) drives expression of a dominant-negative form of the transcription factor TCF3 (T-cell factor 3; also known as LEF3) that lacks its DNA-binding HMG domain and is fused to EGFP. These truncated forms of TCF3 (TCFΔC) act downstream of the destruction complex in Wnt-receiving cells by binding to β-catenin, preventing its interaction with endogenous TCF3 (Martin and Kimelman, 2012). Tg(hsp70l:Mmu.Axin1-YFP)^w35 expresses a fusion of Axin1 (a pivotal component of the β-catenin destruction complex) and fluorescent YFP to promote β-catenin degradation (Kagermeier-Schenk et al., 2011; MacDonald et al., 2009; Stamos and Weis, 2013). We found that blocking canonical Wnt signaling with these tools yields defects in intracerebral angiogenesis and endothelial expression similar to those displayed by reck ^y72 mutants (Fig. S7). For example, forced expression of TCFΔC greatly inhibited CtA angiogenesis and abrogated Glut1 expression (Fig. S7E-H″) without disrupting the trunk's vascular patterning (Fig. S8). Similarly, Axin1-YFP reduced the abundance of CtA by ∼50% (Fig. S9E,F) and induced ectopic plvap expression in the PHBCs (Fig. S7J).

We also attempted to rescue CtA angiogenesis in reck ^y72 mutants by forced activation of canonical Wnt signaling. Briefly, we overexpressed constitutive active (ca) forms of β-catenin (caβ-catenin), either ubiquitously with the heat-inducible Tg(hsp70l:caβ-catenin-2A-TFP)^w130 transgenic line (Veldman et al., 2013) or with endothelial specificity using flt1-driven mosaic expression (Wada et al., 2013; Wu et al., 2012; Yost et al., 1996). Forced expression of caβ-catenin inhibited CtA angiogenesis in WT animals and, unsurprisingly, failed to rescue CtA angiogenesis in reck ^y72 mutants. It is likely that in these experiments, the overexpression of caβ-catenin inhibited canonical Wnt signaling by transcriptional squelching (titration of endogenous transcription factors), as previously demonstrated (Prieve and Waterman, 1999). In addition, flt1-driven mosaic endothelial expression of a dominant-negative form of GSK3β (DN-GSK3-GFP; Taelman et al., 2010) had no effect on CtA angiogenesis in WT animals and failed to promote CtA angiogenesis in reck ^y72 mutants, consistent with the fact that chemical activation of Wnt signaling using the GSK3β inhibitors CHIR99021 and LiCl likewise failed to rescue both the DRG and CtA deficits of reck ^y72 mutants (Fig. S10; Klein and Melton, 1996; Ring et al., 2003; Vanhollebeke et al., 2015; Veldman et al., 2013). Overall, it is likely that the inability to achieve physiological levels of canonical Wnt signaling with proper spatio-temporal dynamics by forced expression of modified components of this pathway or with drugs explains why these treatments failed to rescue the defects of reck ^y72 mutants.

Nonetheless, our observations demonstrate that the role of canonical Wnt signaling in promoting intracerebral angiogenesis and proper expression of markers of barriergenic differentiation is evolutionarily conserved between zebrafish and mammals (Tam et al., 2012; Umans and Taylor, 2012). In addition, our other findings indicate that the cerebrovascular activity of canonical Wnt signaling is Reck dependent, thus uncovering this tumor suppressor as a key and novel modulator of this pathway.

VEGF signaling, which is crucial for intracerebral vascularization, is mediated primarily by receptors encoded by the zebrafish kinase insert domain receptor-like (kdrl) and mammalian KDR (also known as VEGFR2) ‘ohnologs’ (Bussmann et al., 2008; Habeck et al., 2002; Mackenzie and Ruhrberg, 2012; Sivaraj et al., 2013; Sohet and Daneman, 2013; Wittko-Schneider et al., 2013). We noted that the kdrl transcriptional reporters Tg(kdrl:RFP)^s896 and/or Tg(kdrl:GFP)^1a116 (Chi et al., 2008; Choi et al., 2007) were weakly expressed in PHBCs of reck ^y72 and Df(Chr24:reck)^w15 homozygotes, and reck ^y72/Df(Chr24:reck)^w15 trans-heterozygotes (Fig. S9A,B), but a similar reduction was not apparent in trunk vessels. By contrast, endothelial fluorescence from the fli1a transcriptional reporter Tg(fli1a:EGFP) ^y1 (Lawson and Weinstein, 2002) was unaffected in reck⁻ embryos. We verified these qualitative observations by performing quantitative confocal imaging to determine the normalized fluorescent intensity (NFI) of Tg(kdrl:RFP)^s896 using the signal of Tg(fli1a:EGFP) ^y1 as a reference (Venkiteswaran et al., 2013). These measurements confirmed that in reck ^y72 mutants, the fluorescence of Tg(kdrl:RFP)^s896 decreased in the PHBCs, but not in other cephalic vessels such as the MtA (Fig. S9D). Consistent with the existence of distinct genetic circuits governing the activity of VEGF signaling in arterial and venous vessels (Covassin et al., 2006) we found that in CtA-deficient kdrl ^y17 mutants (Covassin et al., 2006) Tg(kdrl:RFP)^s896 fluorescence was unaffected in the PHBCs but reduced in the MtA (Fig. S9C-D). These findings suggest that the expression of kdrl in the PHBCs, as in the trunk vessels (Wythe et al., 2013), is insensitive to VEGF signaling (but see also Lawson et al., 2003,, 2002; Liang et al., 2001). We next overexpressed Axin1-YFP to determine whether inhibition of canonical Wnt signaling reduces Tg(kdrl:RFP)^s896 expression (Veldman et al., 2013; Wang and Nakayama, 2009). Indeed, this treatment dramatically reduced Tg(kdrl:RFP)^s896 expression (Fig. S9E,F). These observations suggest that the cerebrovascular inactivation of canonical Wnt signaling in reck ^y72 mutants reduces kdrl expression, thereby secondarily impairing VEGF activity. Yet, additional observations are inconsistent with this possibility. First, the cerebrovascular phenotypes of reck⁻ and kdrl⁻ mutants are not identical. Both mutants lack CtAs, but in kdrl mutants, the BA is missing and cerebral endothelial abundance is reduced (Bussmann et al., 2011). Second, plvap was ectopically expressed in the PHBCs of reck ^y72 mutants (Fig. 7I) and in embryos in which canonical Wnt signaling was silenced by overexpression of Axin1-YFP (Fig. S7J). By contrast, in kdrl^um19 mutants, plvap was not ectopically expressed (Fig. S9G-H). This finding is consistent with the emerging notion that intracerebral angiogenesis and barriergenic differentiation are genetically uncoupled (Hagan and Ben-Zvi, 2014) and suggests that VEGF activity is dispensable for cerebrovascular canonical Wnt signaling. Third, we found no obvious changes in the expression of VEGF-responsive markers such as dll4. Finally, RNA in situ HCR failed to reveal any obvious reduction in the abundance of kdrl transcripts in the cerebral vasculature of reck ^y72 mutants, but these were nearly undetectable in kdrl^um19 mutants, probably as a result of nonsense-mediated mRNA decay (Brogna and Wen, 2009).

To investigate whether Reck can impact VEGF signaling in other contexts we used human umbilical vein endothelial cells (HUVECs) because they respond to VEGF and express RECK (Miki et al., 2010). In these experiments, we manipulated the availability of VEGF, the abundance of RECK and KDR kinase activity with the SU5416 inhibitor (Sakao and Tatsumi, 2011), we quantified VEGF signaling readouts (KDR, AKT and ERK1/2 phosphorylation levels) and measured the abundance of RECK and KDR (Fig. S9I-J′; Koch et al., 2011; Lanahan et al., 2013; Zachary, 2003). We found that knockdown of RECK impairs VEGF signaling by reducing pKDR and pAKT levels without affecting the abundance of pERK (Fig. S9I,J,I′), which is consistent with the different thresholds and kinetics of AKT and ERK phosphorylation (Olszewska-Pazdrak et al., 2009). Finally, knockdown of RECK, but not chemical abrogation of VEGF signaling, reduced the abundance of KDR (Fig. S9I-J′). Together, the results of our zebrafish and HUVEC studies suggest that Reck promotes VEGF signaling in contexts other than embryonic cerebrovascular development.

Our findings highlight Reck as a pivotal player in vascular development required for the branching morphogenesis and barriergenic differentiation of cerebral blood vessels (Fig. 8). At a cellular level reck acts in the first process by ensuring the migration-dependent formation of intracerebral vessels and remodeling of the perineural network. The results of our cell transplantation and tissue-specific reck ^y72 rescue experiments indicate that, consistent with its cerebrovascular expression pattern, reck is required in the brain endothelium in a non-cell autonomous manner to promote CtA angiogenesis (Chandana et al., 2010; Miki et al., 2010; Prendergast et al., 2012; Vanhollebeke et al., 2015).

Molecularly, reck is required for proper expression of glut1 and plvap, two barriergenic differentiation markers regulated by canonical Wnt signaling. Accordingly, the cerebrovascular expression of a transgenic reporter of canonical Wnt signaling is lost in reck ^y72 mutants. Our study thus reveals that canonical Wnt signaling promotes intracerebral angiogenesis and barriergenic gene expression with evolutionarily conservation from fish to mammals (Daneman et al., 2009; Liebner et al., 2008; Posokhova et al., 2015; Tam et al., 2012; Zhou and Nathans, 2014; Zhou et al., 2014) and shows that Reck plays a pivotal role in enabling cerebrovascular canonical Wnt signaling (Vanhollebeke et al., 2015).

These novel mechanistic insights about reck might illuminate the etiology and treatment of CNS vascular diseases – such as familial exudative vitreoretinopathy (FEVR), osteoporosis-pseudoglioma syndrome (OPPG) and Norrie disease – that are caused by aberrant Wnt signaling and suggest strategies for manipulating vascular barriers to enable drug delivery into the CNS (Daneman et al., 2009; Dejana, 2010; Hagan and Ben-Zvi, 2014; Hawkins and Davis, 2005; Liebner et al., 2008; Obermeier et al., 2013; Siegenthaler et al., 2013; Vallon et al., 2014; Zhou et al., 2014). Accordingly, future studies will determine whether reck, like canonical Wnt signaling, regulates vascular development and/or maintenance throughout the CNS (Daneman et al., 2009; Liebner et al., 2008; Stenman et al., 2008).

How reck promotes canonical Wnt signaling remains undefined at the biochemical level. Reck associates with Gpr124, another novel member of the canonical Wnt signaling pathway that acts downstream of Wnt7 ligands, which has a role in cerebrovascular and DRG development similar to Reck (Posokhova et al., 2015; Vanhollebeke et al., 2015; Wang et al., 2014; Zhou and Nathans, 2014). Thus, Reck-Gpr124 heteromers might modulate Wnt7 binding, the assembly of the Wnt signaling complex, its internalization and/or MP-mediated processing events that gate its activity. However, it is worth highlighting that the effects of inactivating reck and gpr124 are not fully equivalent. For example, gpr124 mutants begin to recover intracerebral angiogenesis at 5 dpf and 50% of them survive to become adults without any apparent deficits in cerebral vascularization (Vanhollebeke et al., 2015). By contrast, in mutants of the different reck alleles, CtA angiogenesis does not recover after 5 dpf and lethality is fully penetrant by 10 dpf. Similarly, the CtA and DRG deficits of gpr124 morphants (Vanhollebeke et al., 2015), but not of reck ^y72 mutants, can be rescued by chemical activation of Wnt signaling.

Finally, our zebrafish and endothelial cell culture studies suggest that Reck promotes VEGF signaling in contexts other than cerebrovascular development, consistent with the involvement of Gpr124 in VEGF-mediated tumor angiogenesis (Wang et al., 2014). RECK is expressed in tumor vessels (Clark et al., 2011; Miki et al., 2010; Oh et al., 2001; Rahmah et al., 2012) and epigenetically inactivated in human cancers. Reduced abundance of RECK is a poor prognosis tumor signature linked to high metastasis and short survival in patients with cerebral glioma, neuroblastoma and other tumors (Nagini, 2012; Noda and Takahashi, 2007). Given the involvement of both canonical Wnt and VEGF signaling in both tumor vascularization and metastasis (Carmeliet, 2005; Goel and Mercurio, 2013; Hu et al., 2009; Klaus and Birchmeier, 2008), our findings might also prove relevant in these pathological settings.

Zebrafish (Danio rerio) were handled under protocols approved by the New York University IACUC/IBC.

The following transgenic fluorescent reporters were used. Endothelial: Tg(kdrl:RFP)^s896 (Chi et al., 2008), Tg(kdrl:EGFP)^la116 (Anderson et al., 2008) Tg(kdrl:eGFP-NLS)^zf109 (Blum et al., 2008), Tg(fli1a:eGFP) ^y1 (Lawson and Weinstein, 2002); erythrocytes: Tg(gata1a:DsRed)^sd2 (Traver et al., 2003); DRG (and other neuronal populations): Tg(neurog1:eGFP)^w61 (McGraw et al., 2008); canonical Wnt signaling activity: Tg(7xTCF-Xla.Siam:GFP)^ia4 (Moro et al., 2013). Transgenes for inhibiting canonical Wnt signaling were Tg(hsp70l:Mmu.Axin1-YFP)^w35 (Kagermeier-Schenk et al., 2011) and Tg(hsp70l:Xla.TCFΔC-EGFP) (Martin and Kimelman, 2012). Transgene used for activating canonical Wnt signaling was Tg(hsp70l:caβ−catenin-2A-TFP)^w130 (Veldman et al., 2013). Mutant transgenes used were kdrl ^y17 and kdrl^um19 (Covassin et al., 2009; Meng et al., 2008), reck ^y72 (this study), reck^w14 and Df(Chr24:reck)^w15 (Prendergast et al., 2012), sih^b109 (Sehnert et al., 2002). Genotyping protocols are provided in supplementary Materials and Methods.

Tg(hsp70l:Xla.TCFΔC-EGFP), Tg(hsp70l:caβ−catenin-2A-TFP)^w130 and their HS controls were heat-shocked at 40°C for 30 min at 30 hpf and fixed at 48 hpf. Tg(hsp70l:Mmu.Axin1-YFP)^w35 and its HS controls were heat-shocked at 39°C for 1 h at 24 hpf and fixed and/or imaged live at 48 hpf. Tg(hsp70l:Xla.TCFΔC-EGFP) and Tg(hsp70l:Mmu.Axin1-YFP)^w35 transgenes were provided paternally.

Primary antibodies used in zebrafish experiments: rabbit anti-RFP (1:500; Clonetech, 632496), rabbit anti-GFP (1:2000; Life Technologies, A11122), mouse anti-phospho-FAK y397 (1:100; Millipore, MAB1144), rabbit Glut1 (1:200; Novus Biologicals, NB300666), mouse zrf1 (1:10; ZIRC), mouse anti-3A10 (1:100; DSHB), mouse anti-HuC (1:100; Life Technologies, A21271) and mouse anti-HA (1:500; Cell Signaling, 2367S). Primary antibodies used in cell culture experiments purchased from Cell Signaling: pVEGFR-2 (1:1000; 2478), VEGFR-2 (1:5000; 2479), pAKT (Ser473; 1:1000; 4058), AKT (1:10,000; 9272), pERK1/2 (Ser473; 1:15,000; 4370), ERK1/2 (1:15,000; 4695), Reck (1:5000; 3433), GAPDH (1:10,000; 4058) and HA (1:10,000; 2367). GFP (1:5000; Life Technologies, A11121) and FLAG (1:20,000; Sigma, F1804) secondary antibodies were all used. Secondary antibodies for zebrafish and cell culture experiments were donkey Alexa Fluor-labeled anti-mouse or anti-rabbit antibodies from Life Technologies (1:1000; A31571, A10040, A21206 and A100036).

Fixed embryos were incubated in a Phalloidin-488 (Sigma) solution for 2 h at room temperature as described by Snow et al. (2008).

Non-fluorescent chromogenic WISH was carried out as in Zygmunt et al. (2011). Fluorescent hybridization chain reaction (HCR-WISH) was performed as in Choi et al. (2014). See supplementary Materials and Methods for further details.

HUVEC (Lifeline Cell Technology, FC-0003) cells were infected with anti-RECK shRNA lentiviral particles, puromycin-selected, starved overnight and stimulated with or without VEGFA. SU5416 (Sigma, S8442) was used for VEGFA signaling inhibition. HEK 293T (ATCC, CRL-11268) and COS-7 (ATCC, CRL-1651) cells were treated with Lipofectamine 2000 (Life Technologies) to transfect constructs for expressing epitope-tagged Reck and Reck^y72. Cells were tested for contamination by the commercial provider. Cells were used for western blotting or immunofluorescence studies 48 h post-transfection (see supplementary Materials and Methods for details).

Cell transplantation was carried out as previously described (Carmany-Rampey and Moens, 2006; Zygmunt et al., 2011).

Tol2 transgenesis and Tol2/Gateway-based vectors were used for flt1-driven mosaic endothelial expression in zebrafish. Gateway-based vectors were used for CMV-driven expression in cultured cells (Bussmann et al., 2010; Hogan et al., 2009; Kwan et al., 2007; Villefranc et al., 2007).

reck mRNA (100 pg) was injected into one-cell stage embryos.

Dechorionated embryos were incubated with DMSO vehicle (0.7% in egg water; negative control) or GSK3β inhibitors starting at the 16-somite stage (16 hpf) until 72 hpf. The following GSK3β inhibitors were used: CHIR99021 (10 μM, diluted in DMSO) and LiCl (100 mM, diluted in egg water) as in Vanhollebeke et al. (2015) and Veldman et al. (2013).

Image were acquired using a Leica SP5 confocal microscope with 40× dipping/water immersion objectives (NA=0.8 or 1.1), bi-directional scans at 200 lines/s in 1024×1024 pixel windows; z-stacks at 1 µm z intervals. Images are maximum intensity projections. Whole-head images were assembled from combining separately collected anterior and posterior regions using ImageJ and Photoshop. Live quantification of Tg(kdrl:RFP)^s896 fluoresence in Tg(kdrl:RFP)^s896/+; Tg(fli1a:eGFP) ^y1/+ embryos (WT and reck ^y72) was carried out by normalizing RFP signals to those of GFP at every z-level with a custom ImageJ macro (Venkiteswaran et al., 2013), and the resulting values were averaged over all z-levels. Embryos were fixed in 1% agarose/PBS or fixed live at single time points in 1% agarose/fish medium with Tricaine and PTU or in 0.1% agarose under 1% agarose for live time-lapse imaging (Kaufmann et al., 2012; Lawson and Weinstein, 2002).

Densitometry data from western blot assays were analyzed using a two-tailed Student's t-test. Differences were considered significant when P<0.05. Error bars indicate s.e.m.

We thank David Kimelman, Deborah Yelon, Didier Stainier, Donghun Shin, Duc Si Dong, Eleanor Y. Chen, Edward De Robertis, Enrico Moro, Francesco Argenton, Gilbert Weidinger, H. Joseph Yost, James Amatruda, Jau-Nian Chen, Jeanot Muster, Jenna Galloway, Leonard Zon, Koichi Kawakami, Lilianna Solnica-Krezel, Markus Affolter, Michael Taylor Nathan Lawson, Richard Dorsky, Tatjana Piotrowski, Timothy Hla and Tohru Ishitani for zebrafish and/or reagents. Thanks to Harry Choi and Niles Pierce for HCR-WISH advice. We thank the Knaut lab and Michael Taylor for useful discussions and Dolly Chan, John Grosso and ‪Wilberto Ortiz-Batista for administrative support.