β-catenin-dependent transcription is central to Bmp-mediated formation of venous vessels

β-catenin is a transcriptional regulator that mainly acts downstream of Wnt signaling and mediates diverse cellular functions including proliferation, survival and fate determination during embryo-genesis, tissue homeostasis and regeneration (Angers and Moon, 2009; Clevers, 2006; Logan and Nusse, 2004; MacDonald et al., 2009). In the absence of Wnt signaling, cytosolic β-catenin is phosphorylated by a destruction complex composed of axin, adenomatous polyposis coli and glycogen synthase kinase 3β, which leads to degradation of β-catenin via the ubiquitin-proteasome pathway. Wnt inhibits the activity of this destruction complex, resulting in the stabilization and nuclear translocation of β-catenin. In the nucleus, β-catenin associates with the T-cell factor (Tcf)/lymphoid enhancer factor family of transcription factors, thereby inducing the transcription of target genes. In addition, β-catenin regulates cell-cell adhesions by linking cadherin molecules to the actin cytoskeleton via α-catenin (Dejana et al., 2008).

β-catenin has been implicated in many aspects of vascular development and maintenance. Mice deficient in β-catenin in endothelial cells (ECs) exhibit embryonic lethality due to defective vascular patterning, increased vascular fragility and impaired cardiac valve formation (Cattelino et al., 2003; Liebner et al., 2004). The phenotypes caused by EC β-catenin deficiency are thought to be attributable to not only the impairment of EC junctions but also the loss of the Wnt/β-catenin pathway. Therefore, it is important to explore the involvement of β-catenin-dependent transcriptional regulation in vascular development by clearly distinguishing the role of β-catenin in gene regulation from that in cell-cell adhesion.

The Wnt/β-catenin pathway promotes sprouting angiogenesis in the central nervous system and controls blood-brain barrier development (Daneman et al., 2009; Liebner et al., 2008; Stenman et al., 2008). Wnt/β-catenin signaling also cooperates with the Notch pathway in vascular development. Notch-regulated ankyrin repeat protein acts as a molecular link between the Notch and Wnt/β-catenin pathways in ECs (Phng et al., 2009). Furthermore, the Wnt/β-catenin pathway potentiates Notch signaling by inducing the expression of Dll4, which modulates vascular remodeling and arterial EC specification during early development (Corada et al., 2010; Yamamizu et al., 2010; Zhang et al., 2011). However, a recent report indicated that Dll4 expression in ECs and subsequent arterial specification are not mediated by the Wnt/β-catenin pathway in vivo (Wythe et al., 2013). Therefore, the role of β-catenin-dependent transcription in vascular development has yet to be elucidated.

It is widely accepted that the optimal way to analyze the role of β-catenin-dependent transcription in ECs during vascular development is to directly visualize its activity exclusively in ECs in vivo. To date, several transgenic (Tg) murine models have been developed to detect the in vivo transcriptional activity of β-catenin (DasGupta and Fuchs, 1999; Maretto et al., 2003). However, these reporter mice do not faithfully reproduce endogenous β-catenin transcriptional activity (Ahrens et al., 2011; Al Alam et al., 2011). In addition, several zebrafish reporter lines that express fluorescent proteins under the control of a β-catenin/Tcf-responsive promoter have been generated (Dorsky et al., 2002; Moro et al., 2012; Shimizu et al., 2012). However, they might not be suitable for studying β-catenin-dependent transcription in ECs, since the reporter gene expression is not restricted to the vasculature.

We have developed a novel Tg zebrafish line that allows us to visualize the β-catenin transcriptional activity in ECs, revealing β-catenin-mediated transcription in specific parts of the vasculature, including the caudal veins (CVs), in developing embryos. During CV formation, bone morphogenetic protein (Bmp) induces the expression of Angiogenic factor with G patch and FHA domains 1 (Aggf1), which in turn stimulates β-catenin transcriptional activity. Subsequently, activated β-catenin mediates CV formation by promoting venous EC differentiation through the expression of Nuclear receptor subfamily 2, group F, member 2 (Nr2f2, also known as Coup-TFII) and by maintaining their survival.

To monitor the transcriptional activity of β-catenin, we constructed a plasmid encoding a β-catenin-dependent Gal4 driver (Gal4db-TΔC) that contains the β-catenin-binding domain of Tcf4 fused to the DNA-binding domain of Gal4. To ascertain whether this driver can be stimulated by β-catenin, HEK 293T cells were transfected with plasmid encoding Gal4db-TΔC and an upstream activation signal (UAS)-luciferase reporter together with a β-catenin-expressing plasmid. β-catenin overexpression induced luciferase expression in the cells expressing Gal4db-TΔC, but not in those expressing the mutant form Gal4db-TΔC (D16A), which lacks β-catenin binding capacity (supplementary material Fig. S1A). Treatment with BIO, a glycogen synthase kinase 3 inhibitor, potentiated the Gal4db-TΔC-driven reporter gene expression (supplementary material Fig. S1B). These results indicate that reporter gene expression driven by the β-catenin-dependent Gal4 driver reflects the transcriptional activity of β-catenin.

To visualize β-catenin transcriptional activity in ECs, we developed a Tg zebrafish line that expresses a fusion protein containing Gal4db-TΔC, 2A peptide and mCherry (Gal4db-TΔC-2A-mC) under the control of the fli1 promoter. Then, we crossed it with UAS:GFP reporter fish to generate the Tg(fli1:Gal4db-TΔC-2A-mC);(UAS:GFP) line, which we refer to as EC-specific β-catenin reporter fish (Fig. 1A). In EC-specific β-catenin reporter fish with the Tg(fli1:Myr-mC) background, GFP was observed in particular regions of ECs together with mCherry fluorescence (Fig. 1B). However, most of the mCherry-marked blood vessels emitted GFP fluorescence upon heat shock promoter-mediated overexpression of Wnt3a, an activator of canonical β-catenin signaling (Fig. 1B). Furthermore, in BIO-treated embryos most of the ECs exhibited GFP, whereas GFP expression was significantly suppressed by treatment with IWR-1, which promotes β-catenin degradation by stabilizing axin (Chen et al., 2009) (Fig. 1C,D; supplementary material Fig. S1C). These results indicate that the GFP in our EC-specific β-catenin reporter fish line reflects the transcriptional activity of β-catenin in ECs.

We investigated how β-catenin-mediated transcription in ECs is spatiotemporally controlled during vascular development by analyzing our EC-specific β-catenin reporter fish. The GFP signal was first detected in the cranial vessels at 24 h post-fertilization (hpf), and reached a maximum at 36 hpf (Fig. 1E). Consistent with the role of endocardial Wnt/β-catenin signaling in cardiac valve formation (Hurlstone et al., 2003), the GFP signal was detected in endocardial ECs at 48 hpf (Fig. 1E). The ECs in the common cardinal veins also expressed GFP at 36 hpf (Fig. 1E). Furthermore, GFP was observed in the caudal vessels starting at 26 hpf, but this fluorescence was confined to the CV at 48 hpf (Fig. 1E). These observations imply that β-catenin plays a significant role in vascular development.

CV formation involves venous fate specification/differentiation, venous sprouting and remodeling, making it suitable for studying venous vessel formation (Choi et al., 2011; Kim et al., 2012; Wiley et al., 2011). To address the role of β-catenin in CV formation, we first analyzed CV formation during development (supplementary material Fig. S2). The CV primordia initially formed just beneath the caudal artery (CA) at 24 hpf, as previously reported (Choi et al., 2011). Then, the ECs sprouted from the CV primordia to migrate ventrally at 28 hpf, and formed a CV plexus (CVP) at 36 hpf. At 48 hpf, the ECs in the ventral region of the plexus underwent remodeling to establish the CV, while some ECs in the dorsal part of the plexus migrated dorsally to form secondary sprouts. During the next few days of development, the CVP relocated ventrally and fused to the CV, while some ECs in the plexus regressed during this process.

To investigate when and where β-catenin-dependent transcription occurs during CV formation, we performed time-lapse imaging of EC-specific β-catenin reporter fish embryos. GFP expression was dramatically induced in the ECs that had just sprouted from the CV primordia (Fig. 2A; supplementary material Movie 1). These GFP-positive ECs migrated ventrally, became located in the ventral part of the CVP, and finally formed the CV (Fig. 2A,B; supplementary material Movie 1). However, GFP fluorescence gradually decreased once the CV had formed, suggesting a transient role of β-catenin-mediated gene expression in CV formation.

To address this hypothesis, we analyzed Tg(fli1:Gal4FF);(fli1:Myr-mC) embryos injected with a UAS:axin,NLS-GFP Tol2 plasmid, which drives the expression of axin and NLS-GFP simultaneously in ECs, and found that axin-overexpressing ECs, in which β-catenin degradation was facilitated, failed to contribute to CV formation (supplementary material Fig. S2B,C). Furthermore, we inhibited β-catenin/Tcf-dependent transcription in ECs by generating Tg(fli1:Gal4FF);(UAS:RFP);(UAS:TΔN-GFP) fish embryos. Expression of dominant-negative Tcf (TΔN-GFP) in these ECs resulted in impairment of CV formation, but did not affect CA formation (Fig. 2C,D). These results indicate the essential role of β-catenin-dependent transcription in CV development but not in CA formation. During CV formation, nuclear fragmentation occurred in the TΔN-GFP-expressing ECs, but not in those expressing NLS-GFP (Fig. 2E,F; supplementary material Movie 2). Among the TΔN-GFP-expressing ECs, most of the cells undergoing nuclear fragmentation were found in the CVP, although a small number of the ECs in the CA also underwent nuclear fragmentation (Fig. 2E,F; supplementary material Movie 2). In addition, TUNEL analyses revealed that TΔN-GFP expression induced EC apoptosis in the CVP, but not in the CA (Fig. 2G,H).

Collectively, these findings indicate that β-catenin-dependent transcription regulates CV formation by promoting the survival of venous ECs, whereas it is not necessary for the survival of arterial ECs.

We next sought to identify the upstream signaling molecules that stimulate β-catenin transcriptional activity during CV formation, starting with Wnt as the canonical stimulator of β-catenin activity. To inhibit Wnt signaling, we ubiquitously expressed the Wnt antagonist Dickkopf 1 (Dkk1) in the Tg(hsp70l:dkk1-FLAG) and Tg(hsp70l:dkk1-GFP) lines by activating the hsp70l heat shock promoter. When heat shocked at 12 hpf, these Tg embryos displayed a short body axis phenotype (supplementary material Fig. S3A-E), as previously reported (Caneparo et al., 2007). Embryos heat shocked at 24 hpf exhibited caudal fin defects, indicating that Wnt signaling can be functionally inhibited by heat shock promoter-driven expression of Dkk1. However, Dkk1 expression affected neither CV formation nor β-catenin transcriptional activity in the CV (Fig. 3A,B), suggesting that Wnt is not involved in β-catenin activation during CV formation.

We next focused on Bmp signaling, as Bmp regulates sprouting angiogenesis in CV development (Wiley et al., 2011). To investigate the role of Bmp in β-catenin-mediated CV formation, EC-specific β-catenin reporter fish embryos were injected with hsp70l:noggin3 plasmid at the one-cell stage and heat shocked at 24 hpf. Overexpression of the Bmp antagonist Noggin 3 suppressed the transcriptional activity of β-catenin in the CV primordia and the ECs that had just sprouted from the CV primordia at 28 hpf (supplementary material Fig. S3F,G). At 36 hpf, Noggin 3-overexpressing embryos exhibited impaired CVP formation and decreased transcriptional activity of β-catenin in the defective CVP (Fig. 3C,D). By contrast, heat shock promoter-mediated overexpression of Bmp2b resulted in ectopic sprouting of venous ECs from the CVP (Fig. 3C). Importantly, β-catenin-dependent transcriptional activity was clearly observed in these ectopic vessels (Fig. 3C). Furthermore, the number of apoptotic ECs in the CVP was significantly increased by Noggin 3 (Fig. 3E,F). These findings reveal that Bmp stimulates β-catenin activity to regulate CV formation by promoting the survival of venous ECs.

Vascular endothelial growth factor A (Vegfa) stimulates sprouting angiogenesis from the CA (Wiley et al., 2011). Thus, we hypothesized that Vegfa might promote EC survival in the CA. Treatment with Ki 8457 (Kwon et al., 2013), a Vegfa receptor inhibitor, induced nuclear fragmentation of ECs in the CA and intersegmental vessels (ISVs), but not in the CVP (supplementary material Fig. S4). These results indicate that distinct signaling pathways regulate EC survival in the CVP and CA; Bmp promotes EC survival in the CVP through β-catenin, whereas Vegfa maintains EC survival in the CA.

To address how Bmp stimulates β-catenin transcriptional activity and to identify the β-catenin target genes responsible for CV formation, we performed RNA-seq analyses. To identify the genes that are upregulated and downregulated in the ECs with high β-catenin transcriptional activity, GFP-positive and mCherry-positive [β-catenin (+)] ECs and GFP-negative and mCherry-positive [β-catenin (−)] ECs were isolated from the caudal parts of Tg(fli1:Gal4db-TΔC-2A-mC);(UAS:GFP);(fli1:Myr-mC) embryos and subjected to RNA-seq analyses (supplementary material Fig. S5). As expected, the β-catenin (+) ECs expressed high levels of β-catenin/Tcf target genes, such as axin2, ccnd2a, ccnd2b, ctnnb1 and tcf7 (Bandapalli et al., 2009; Kioussi et al., 2002; Lustig et al., 2002; Roose et al., 1999) (Table 1). In addition, some of the Bmp signal-responsive genes, such as id1, id2a, id2b, smad6a and smad6b, were highly expressed in the β-catenin (+) ECs (Hollnagel et al., 1999; Ishida et al., 2000) (Table 1), suggesting that β-catenin activity is stimulated by Bmp during CV development. Moreover, venous markers, including dab2, ephb4a, ephb4b, flt4, nrp2a and nr2f2, were highly expressed in the β-catenin (+) ECs (Covassin et al., 2006; Herzog et al., 2001; Sprague et al., 2006; Swift and Weinstein, 2009) (Table 1), suggesting that β-catenin plays a role in venous vessel development.

Analyzing the RNA-seq data revealed that the level of aggf1 expression is higher in β-catenin (+) ECs than in β-catenin (−) ECs (Table 1). AGGF1 was first described as a gene encoding an angiogenic factor and has been implicated as a causative gene for Klippel–Trenaunay syndrome (KTS), a disorder characterized by venous malformation and hypertrophy of bones and soft tissues (Tian et al., 2004). It was recently reported that Aggf1 mediates the specification of venous EC identity in zebrafish embryos (Chen et al., 2013). Importantly, Major et al. also characterized AGGF1 as a nuclear chromatin-associated protein that participates in β-catenin-mediated transcription in human colon cancer cells (Major et al., 2008). Therefore, we hypothesized that Aggf1 acts downstream of Bmp to stimulate β-catenin transcriptional activity during CV development. To address this possibility, we examined the expression pattern of aggf1 mRNA in zebrafish embryos. At 24, 30 and 36 hpf, aggf1 mRNA was predominantly expressed in the head region (Fig. 4A), as previously reported (Chen et al., 2013). Additionally, its expression was weakly but broadly detected in the trunk region at 24 hpf, and had become more restricted to the vasculature by 30 hpf (Fig. 4A). At 36 hpf, aggf1 expression was confined to the CVP, where β-catenin-mediated transcription actively occurred (Fig. 2A, Fig. 4A). Importantly, aggf1 expression in the CVP was suppressed by Noggin 3, whereas its expression domain was expanded in Bmp2b-overexpressing embryos (Fig. 4B). These results indicate that Bmp induces aggf1 expression in the CVP.

We next investigated whether Aggf1 is responsible for CV formation. Depletion of Aggf1 by two independent morpholino oligonucleotides (MOs) resulted in CV absence and defective CVP formation (Fig. 4C,D). Co-depletion of p53 (also known as Tp53) did not alleviate the CV defects caused by Aggf1 deficiency, although it partially rescued the defective formation of the CVP (supplementary material Fig. S6A,B), indicating that the phenotypes of aggf1 morphants are not due to any MO off-target effects. By contrast, CA formation was unaffected by knockdown of Aggf1 (Fig. 4C; supplementary material Fig. S6A). Although the aggf1 morphants showed some ISV defects, more than 80% of defective ISVs were venous vessels that sprouted from the CVP (venous ISVs, 83.3%; arterial ISVs, 16.7%; n=30) (supplementary material Fig. S6C), suggesting a role of Aggf1 in venous vessel development. Consistent with this, the ectopic formation of venous vessels induced by overexpression of Bmp2b was suppressed by the depletion of Aggf1 (Fig. 4E,F). Furthermore, aggf1 morphants exhibited an increase in EC apoptosis in the CVP as compared with control MO-injected embryos (Fig. 4G,H). Collectively, these results indicate that Bmp induces Aggf1 expression to maintain the survival of venous ECs during CV formation.

To investigate the role of Aggf1 in Bmp-induced activation of β-catenin transcriptional activity, we performed the luciferase assay using a Tcf-responsive reporter construct (TOPflash). Although Aggf1 alone did not stimulate the TOPflash reporter, it synergized with β-catenin to induce reporter gene expression (Fig. 5A). We also tested the effect of Aggf1 on the transcriptional activity of Gal4–β-catenin to explore whether Tcf is required for Aggf1-induced activation of β-catenin. Aggf1 potently increased the transcriptional activity of Gal4–β-catenin, but did not affect that of Gal4-VP16 (Fig. 5B). These findings suggest that Aggf1 stimulates the transcriptional activity of β-catenin by acting as a transcriptional co-factor.

We ascertained whether Aggf1 enhances β-catenin-mediated transcription in vivo by injecting aggf1-mCherry mRNA into EC-specific β-catenin reporter fish embryos. Overexpression of Aggf1-mCherry resulted in a significant increase in GFP in the CVP, although it did not induce ectopic GFP expression in other parts of the vasculature, indicating that Aggf1 enhances β-catenin transcription specifically in the CVP (Fig. 5C,D). We also found that depletion of Aggf1 suppressed β-catenin transcriptional activity in the defective CVP (Fig. 5E,F; supplementary material Fig. S6D,E). These findings suggest that Aggf1 acts downstream of Bmp to stimulate β-catenin transcriptional activity in the CVP.

RNA-seq analyses revealed that nr2f2 is highly induced in β-catenin (+) ECs (Table 1). Nr2f2 is an orphan nuclear receptor that plays a crucial role in venous specification (Aranguren et al., 2011; You et al., 2005). We investigated whether Bmp induces nr2f2 expression to regulate CV formation. Expression of nr2f2 was detected in both posterior cardinal veins (PCVs) and the CV in 48 hpf embryos (supplementary material Fig. S7A). Heat shock promoter-driven expression of Noggin 3 caused downregulation of nr2f2 in the CV (Fig. 6A). By contrast, the expression domain of nr2f2 in the CV was expanded by overexpression of Bmp2b (Fig. 6B), indicating that Bmp induces nr2f2 expression in the CV.

We next explored the role of β-catenin in nr2f2 expression. Expression of dominant-negative Tcf in ECs resulted in downregulation of nr2f2 in the CV (Fig. 6C), suggesting that β-catenin plays a role in nr2f2 expression. Consistent with this, the NR2F2 expression level in human umbilical vein ECs was reduced by knockdown of β-catenin and increased by treatment with BIO (supplementary material Fig. S7B-D). We also investigated the involvement of Aggf1 in β-catenin-mediated nr2f2 expression. The nr2f2 expression level in the CV was decreased by depletion of Aggf1, but slightly increased by overexpression of Aggf1-mCherry (Fig. 6D,E). Importantly, treatment with BIO further augmented nr2f2 expression in the CV of aggf1-mCherry mRNA-injected embryos (Fig. 6E). Collectively, these results indicate that Bmp induces nr2f2 expression in the CV through Aggf1-mediated activation of β-catenin.

Next, we investigated the role of Nr2f2 in CV formation. Knockdown of Nr2f2 did not affect EC sprouting from the CV primordia and formation of the CVP (Fig. 6F,G). However, subsequent remodeling of the CVP was severely inhibited by depletion of Nr2f2, causing defective formation of the CV (Fig. 6H,I; supplementary material Fig. S8A-D). The CV defects caused by Nr2f2 deficiency were not rescued by co-depletion of p53, indicating that the phenotypes of nr2f2 morphants are not due to MO off-target effects (supplementary material Fig. S8E,F). These results suggest that Nr2f2 is involved in CV formation. Unlike in the aggf1 morphants, EC apoptosis in the CVP was not increased by Nr2f2 depletion, reflecting the non-essential role of Nr2f2 in β-catenin-promoted survival of venous ECs (supplementary material Fig. S9). However, knockdown of Nr2f2 caused downregulation of the venous marker flt4 in the CVP without affecting expression of the endothelial marker fli1a (Fig. 6J,K). These results suggest that Nr2f2 promotes the differentiation of venous ECs required for CV development. Thus, β-catenin may regulate CV formation by promoting venous EC differentiation through Nr2f2 expression in addition to maintaining venous EC survival.

Analyses of EC-specific β-catenin knockout mice have confirmed the crucial role of β-catenin in vascular development. However, the precise function of β-catenin-mediated gene regulation in vascular development has not been clearly understood, since β-catenin regulates not only gene expression but also the formation of cell-cell junctions. Here, we have demonstrated a novel role of β-catenin in venous vessel development by generating a Tg zebrafish line that allows us to visualize EC-specific β-catenin transcriptional activity.

To date, several Tg animals have been developed to trace the transcriptional activity of β-catenin in vivo (DasGupta and Fuchs, 1999; Dorsky et al., 2002; Maretto et al., 2003; Moro et al., 2012; Shimizu et al., 2012). However, the reporter genes are ubiquitously expressed in these systems, making it difficult to study the function of β-catenin in a specific cell type or tissue. Our Tg line represents the first cell type-specific β-catenin reporter and as such is anticipated to contribute to a greater understanding of the diverse functions of β-catenin, since it is, at least in theory, applicable to any cell type or tissue.

GFP in the EC-specific β-catenin reporter line appears to reflect the transcriptional activity of β-catenin in ECs. It is noteworthy that this fluorescence is observed in venous ECs. However, this Tg line might not completely reproduce endogenous β-catenin activity in ECs. Although a previous report suggested that Wnt/β-catenin signaling promotes the proliferation of stalk cells during sprouting angiogenesis (Phng et al., 2009), we did not detect GFP signals in endothelial stalk cells. Therefore, the sensitivity of the EC-specific β-catenin reporter fish might not be high enough to detect weak transcriptional activity of β-catenin. However, at a minimum, the β-catenin-dependent transcription that we detected in this study reveals an essential role for β-catenin in CV formation.

Bmp signaling, but not Wnt, stimulates β-catenin transcriptional activity through Aggf1 to regulate CV development. Although Aggf1 was originally identified as a secreted angiogenic growth factor (Tian et al., 2004), it also functions in the nucleus to stimulate β-catenin-mediated gene expression (Major et al., 2008). Major et al. reported that AGGF1 associates with SWI/SNF chromatin remodeling complexes and participates in β-catenin-dependent transcription in human colon cancer cells (Major et al., 2008). Consistently, our study revealed that Aggf1 stimulates the transcriptional activity of β-catenin in vitro and in vivo. Thus, Aggf1 might act as a chromatin remodeling transcriptional co-factor for β-catenin. In addition to inducing the expression of Aggf1, Bmp might stabilize β-catenin and thereby lead to the induction of β-catenin-dependent transcription, since overexpression of Aggf1 enhanced β-catenin transcriptional activity and induced nr2f2 expression only in the CVP, but not in the other vascular structures. Indeed, Bmp2 induces β-catenin accumulation in pulmonary artery ECs in vitro (de Jesus Perez et al., 2009). Further studies are needed to confirm these hypotheses.

Consistent with the role of Aggf1 in venous vessel formation, AGGF1 is known to be mutated in KTS, a complex congenital disease characterized by varicosities and venous malformations of the lower limbs (Tian et al., 2004). Moreover, KTS patients develop hypertrophy of bones and soft tissues in addition to the venous abnormalities. It would be interesting to investigate whether AGGF1-mediated activation of β-catenin by Bmp is involved in the development of KTS.

Bmp promotes the survival of venous ECs through β-catenin-mediated gene expression during CV formation. A previous report revealed that Bmp promotes the survival of pulmonary artery ECs through β-catenin-mediated expression of survivin (de Jesus Perez et al., 2009). However, our RNA-seq analyses did not detect the expression of survivin (birc5a) mRNA in β-catenin (+) ECs. Therefore, other target genes are likely be induced by β-catenin to promote the survival of venous ECs during CV formation.

β-catenin induces nr2f2 expression in the CV. β-catenin has been shown to induce Nr2f2 expression in preadipocytes by acting directly on its −4.7 kb proximal promoter region (Okamura et al., 2009). However, our luciferase reporter gene analyses revealed that a 5 kb proximal zebrafish nr2f2 promoter is not stimulated by active β-catenin. Therefore, β-catenin is likely to indirectly promote nr2f2 expression in ECs during CV formation. In addition, β-catenin-dependent and -independent pathways might regulate nr2f2 expression, as inhibition of β-catenin/Tcf-mediated transcription suppressed nr2f2 expression in the CV but not in the PCV. Consistent with this, the chromatin remodeling enzyme BRG1 directly promotes murine Nr2f2 expression independently of β-catenin (Davis et al., 2013). Furthermore, it was recently shown that nr2f2 expression in the PCV depends on Sox7 and Sox18 (Swift et al., 2014). Therefore, different mechanisms of nr2f2 expression might be employed in the CV and PCV.

β-catenin-mediated expression of Nr2f2 regulates the differentiation of venous ECs during CV development. Nr2f2 depletion resulted in downregulation of the venous marker flt4 in the CVP, suggesting a role of Nr2f2 in venous EC specification. However, Nr2f2 is required for expression of some but not all markers of venous identity. For instance, expression of the venous markers ephb4a and dab2 in the PCV is only partially affected by depletion of Nr2f2 in zebrafish (Aranguren et al., 2011; Swift et al., 2014). Similarly, levels of endothelial Ephb4 are only slightly reduced in the cardinal vein of Nr2f2-deficient mice (You et al., 2005). Therefore, other factors, in addition to Nr2f2, might be required for full establishment of venous identity.

Although the present study showed that Bmp promotes venous EC differentiation through expression of Nr2f2, Bmp is also known to negatively regulate lymphatic development. Recently, Dunworth et al. reported that Bmp suppresses the expression of Prox1, a key regulator of lymphatic development, thereby preventing differentiation of lymphatic ECs (Dunworth et al., 2014). As such, Bmp might regulate venous vessel formation by inducing the differentiation of venous ECs through expression of Nr2f2 and by suppressing the differentiation of lymphatic ECs through downregulation of Prox1. However, the requirement of Nr2r2 for prox1 expression in embryonic veins has also been reported (Srinivasan et al., 2010; Swift et al., 2014). Thus, further studies are needed to delineate the mechanisms by which Bmp regulates venous and lymphatic development.

In summary, we have successfully visualized the EC-specific transcriptional activity of β-catenin in living animals, and have uncovered a novel role for β-catenin in vascular development (Fig. 7). During CV formation, Bmp induces the expression of Aggf1 in ECs, which leads to the activation of β-catenin-mediated transcription. Activated β-catenin regulates CV formation by promoting the Nr2f2-dependent differentiation of venous ECs and by maintaining their survival.

Zebrafish (Danio rerio) were maintained and bred under standard conditions. Embryos were staged by hpf at 28°C (Kimmel et al., 1995). Animal experiments were approved by the animal committee of the National Cerebral and Cardiovascular Center and performed according to the regulations of the National Cerebral and Cardiovascular Center. Heat shock, MO and mRNA injections and chemical treatments of embryos were undertaken as described in the supplementary Materials and Methods.

Expression constructs were generated as described in detail in the supplementary Materials and Methods.

Tg(fli1:Gal4db-TΔC-2A-mC), Tg(fli1:Myr-mC), Tg(UAS:TΔN-GFP), Tg(UAS:NLS-GFP), Tg(flk1:NLS-Eos), Tg(hsp70l:dkk1-FLAG) and Tg(flt1:mC) zebrafish lines were generated according to the protocol described previously (Kwon et al., 2013). Other Tg lines were obtained as described in the supplementary Materials and Methods. Genotyping of the Tg lines was by genomic PCR as described in the supplementary Materials and Methods.

Zebrafish embryos were mounted in 1% low-melting agarose poured onto a 35-mm diameter glass-based dish (Asahi Techno Glass) as previously described (Fukuhara et al., 2014) (see supplementary Materials and Methods).

Confocal images were taken with a FluoView FV1000 confocal upright microscope system (Olympus) equipped with water-immersion 10× (LUMPlanFL, 0.30 NA) and 20× (XLUMPlanFL, 1.0 NA) lenses. The 473 nm and 559 nm laser lines were employed. For confocal time-lapse imaging, images were collected every 10-20 min for 5-12 h. To avoid cross-detection of green and red signals, images were acquired sequentially at 473 nm and 559 nm. z-stack images were 3D volume rendered with fluorescence mode employing Volocity 3D imaging analysis software (PerkinElmer).

Quantitative analyses of fluorescence intensity, CV defects and ectopic vessel formation were performed as described in the supplementary Materials and Methods.

Tg(fli1:Gal4db-TΔC-2A-mC);(UAS:GFP);(fli1:Myr-mC) fish embryos were cut along the dorsoventral axis as described in supplementary material Fig. S4. Caudal parts of the embryos were collected, incubated in 0.25% trypsin in phosphate-buffered saline at 28°C for 15 min, and dissociated by gentle pipetting. The dissociated cells were sorted using a FACS Aria III cell sorter (BD Biosciences) and grouped into GFP-positive mCherry-positive [β-catenin (+)] and GFP-negative mCherry-positive [β-catenin (−)] EC populations.

Total RNA was purified from β-catenin (+) ECs and β-catenin (−) ECs using a NucleoSpin RNA XS kit (Macherey-Nagel) following the manufacturer's instructions, and subjected to RNA-seq analyses as described in the supplementary Materials and Methods. The RNA-seq data have been deposited at the Sequence Read Archive (SRA) database (NCBI) under accession number SRP051129.

Whole-mount in situ hybridization of zebrafish embryos was performed as described previously (Fukuhara et al., 2014).

EC apoptosis was quantified by TUNEL assay as described in the supplementary Materials and Methods. TOPflash reporter assays were carried out in HEK 293 cells as described in the supplementary Materials and Methods.

Data were analyzed using GraphPad Prism software. Data are expressed as the mean±s.e.m., as described in the figure legends. Statistical significance for paired samples and for multiple comparisons was determined by Student's t-test and by one-way analysis of variance with Tukey's test, respectively. Data were considered statistically significant at P<0.05.

We thank N. Lawson for the fli1 promoter and Tg(fli1:GFP) fish; K. Kawakami for the Tol2 system and Tg(UAS:GFP) and Tg(UAS:RFP) fish lines; M. Affolter for Tg(fli1:Gal4FF) fish; M. Hibi for the UAS-GFP reporter and the Gal4FF plasmid; D. Y. Stainier for the flk1 promoter; J. Kuwada for the hsp70l promoter; G. Felsenfeld for the chicken β-globin insulator; A. Kikuchi for the hTCF4 plasmid; K. Hiratomi, M. Sone, W. Koeda, E. Okamoto and T. Babazono for excellent technical assistance; and K. Shioya for excellent fish care.