Distinct Notch signaling outputs pattern the developing arterial system

Establishment of arterial and venous identity is among the earliest endothelial differentiation steps to occur in the developing vascular system. Although these differences were historically attributed to physiological forces associated with blood flow, more recent studies demonstrated that conserved developmental signaling pathways govern artery and vein identity (Swift and Weinstein, 2009). Furthermore, numerous studies have found that specification of arterial and venous endothelial cell fates is essential for normal vascular morphogenesis and function in vertebrate embryos (Seo et al., 2006; Wang et al., 1998; You et al., 2005). Importantly, the failure to maintain artery identity through conserved developmental signaling pathways may lead to vascular malformations in adults (Carlson et al., 2005; Murphy et al., 2009). Thus, patterning artery and vein identity is essential for normal embryonic development and subsequent circulatory function.

By early-somitogenesis stages of embryonic development, endothelial progenitors, or angioblasts, initiate expression of endothelial cell-specific genes and coalesce to form nascent cords that will give rise to the major blood vessels (Cleaver and Krieg, 2010). Subsequently, expression of artery-specific genes can be detected (Zhong et al., 2001), suggesting that artery and vein specification is an early step during vascular development. Indeed, lineage tracing of progeny from lateral mesoderm progenitors at mid-somitogenesis stages reveals artery or vein lineage commitment by this stage (Kohli et al., 2013; Zhong et al., 2001). Subsequently, artery differentiation appears to be important for vascular morphogenesis. For example, loss of the artery-specific transmembrane ligand, ephrinB2, or its receptor EphB4, which is preferentially expressed in venous endothelial cells, causes similar defects in vascular morphogenesis (Gerety et al., 1999; Wang et al., 1998). Notably, ephrinB2 and EphB4 play important roles in sorting cells at boundary domains within tissues such as the hindbrain (Xu et al., 1999), suggesting that these receptors play a similar role in artery and vein formation. Indeed, ephrinB2 or EphB4 deficiency in zebrafish disrupts segregation of arterial and venous endothelial cell during vessel morphogenesis (Herbert et al., 2009). Thus, differentiation of artery and vein identity governs vascular morphogenesis, in part by establishing the appropriate complement of guidance molecules to distinctly mark and sort these cell populations.

The first genes to be implicated in arterial endothelial specification were components of the Notch signaling pathway (Lawson et al., 2001). Notch proteins are large transmembrane receptors that govern fate decisions in numerous cell types in all animal species studied to date (Guruharsha et al., 2012). Notch receptors are activated by binding to transmembrane molecules known as Delta-like or Jagged ligands. Following binding, the intracellular domain of the processed Notch receptor translocates into the nucleus where it interacts with the Suppressor of hairless/recombination signal binding protein for immunoglobulin kappa J (Rbpj) DNA-binding protein to form a transactivation complex that induces target gene expression (Bailey and Posakony, 1995; Fortini and Artavanis-Tsakonas, 1994). In most cases, the net outcome of this signaling event is that ligand- and receptor-expressing cells will adopt distinct cell fates or exhibit different behaviors in the context of a developing tissue (Guruharsha et al., 2012). During vascular development, Notch receptors and ligands are preferentially expressed in arterial endothelial cells (Lawson et al., 2001; Villa et al., 2001). Accordingly, Notch deficiency leads to a failure in artery marker gene expression, along with the appearance of arteriovenous shunts, in both mouse and zebrafish embryos (Krebs et al., 2004; Lawson et al., 2001). Loss of Notch also leads to a ‘hyper-angiogenesis’ phenotype characterized by increased proliferation and migration of endothelial cells and the excessive formation of non-functional blood vessels (Hellström et al., 2007; Leslie et al., 2007; Siekmann and Lawson, 2007b; Suchting et al., 2007). In these cases, Notch acts to appropriately coordinate tip and stalk cell behaviors in a sprouting blood vessel, in part by dampening the response to surrounding pro-angiogenic factors. In addition to its requirement for vascular development, Notch signaling is also essential for vessel homeostasis in the adult. For example, similar to Delta-like 4 (Dll4) in the embryonic vasculature, Dll1 is restricted to arterial endothelial cells in adult blood vessels, where it is required for collateral artery remodeling (Limbourg et al., 2007). Thus, Notch signaling plays a number of roles within endothelial cells to govern artery differentiation, angiogenesis, and vascular homeostasis during both embryonic and adult stages.

Notch signaling is highly dynamic and can be used reiteratively during the commitment and differentiation of cell lineages (Guruharsha et al., 2012). Although we have learned a great deal concerning the role of Notch in vascular development from genetic manipulations in model systems, we still know little about when and where Notch is activated in endothelial cells during embryogenesis. We previously generated zebrafish lines that allow visualization of Notch-responsive cells and lineage tracing of progeny cells in which Notch has been activated (Clements et al., 2011; Lorent et al., 2010; Parsons et al., 2009; Quillien et al., 2011; Wang et al., 2011). In the study reported here, we have applied these transgenic lines together with genetic manipulation of Notch receptors and ligands, to investigate the dynamic role of Notch signaling during vascular development. Our results suggest that Notch acts at several distinct developmental stages through the use of different receptor and ligand combinations to govern multiple stages of artery formation and patterning.

To determine when Notch is activated in endothelial cells during development, we performed two-color fluorescence in situ hybridization on embryos with a Notch-responsive promoter (referred to as ‘tp1’) driving enhanced green fluorescent protein [Tg(tp1:egfp)^um14 (Parsons et al., 2009)]. At the five somite stage (ss), we observed egfp transcript in presumptive somitic mesoderm and proneural clusters (Fig. 1A), as well as neural crest cells located dorsal to cells expressing ets variant gene 2 (etv2) (Fig. 1A,B), an early marker for endothelial progenitors (Sumanas and Lin, 2006). We noted rare etv2-positive cells expressing egfp at this stage, but not earlier (Fig. 1B,C, yellow arrows; data not shown). By the 12 ss, increasing numbers of etv2-positive cells expressed egfp and they were intermingled with egfp-negative angioblasts (Fig. 1D,E). At the 16 ss, we noted graded levels of egfp expression in fli1a-positive endothelial cells forming the aortic cord (Fig. 2A,B). At this stage, endothelial cells located more rostrally expressed high levels of egfp, whereas caudal cells exhibited lower expression (Fig. 2B, compare cells indicated by yellow arrowheads and white arrows, respectively). Egfp expression in fli1a-positive cells in the developing aorta continued to increase in a rostral to caudal pattern until 24 hours postfertilization (hpf) when all cells within the dorsal aorta expressed high levels of egfp (compare Fig. 2A,C,D). We observed continued expression of Egfp in endothelial cells lining intersomitic arteries and the dorsal aorta in Tg(fli1a:dsredex)^um13;(tp1:egfp)^um14 double transgenic embryos until at least 7 days postfertilization (Fig. 2E). In addition to endothelial expression, we also noted rare egfp-positive gata1a-expressing cells in the early lateral mesoderm, whereas egfp and cmyb expression were largely non-overlapping during mid-somitogenesis stages (supplementary material Fig. S1). Egfp expression did not expand in these populations, as we observed with etv2- or fli1a-positive cells, although egfp-positive cmyb-expressing cells were apparent in the ventral wall of the dorsal aorta at the 18 ss (supplementary material Fig. S1; data not shown). These observations suggest that Notch activation initiates in etv2-positive endothelial progenitors during the earliest stages of vasculogenesis. Subsequently, Notch activation was restricted to and persisted in arterial endothelial cells throughout later embryonic and larval stages.

To determine the fate of cells in which Notch activation occurred during early somitogenesis, we performed lineage tracing using Tg(tp1:creert2)^jh12 transgenic zebrafish that express a tamoxifen-inducible form of Cre recombinase driven by the tp1 element (Wang et al., 2011). For this purpose, we pulsed Tg(tp1:creert2)^jh12 embryos bearing a ubiquitously expressed conditional hmgb1-mcherry fusion [Tg(actb2:loxP-stop-loxP;hmgb1-mcherry)^jh15] with tamoxifen for 1 hour at the 5 or 10 ss. After 48 hours, we observed red fluorescent cells lining the dorsal aorta, but not the cardinal vein in Tg(tp1:creert2)^jh12;(actb2:loxP-stop-loxP;hmgb1-mcherry)^jh15 embryos (Fig. 3A; n=28 embryos). We similarly observed aorta-restricted expression in Tg(tp1:creert2)^jh12 embryos bearing an endothelial-specific conditional red fluorescent transgene [Tg(fli1ep:loxP-nblue-loxP;mcherry)^um43] at 30 hpf following a 1 hour exposure to tamoxifen at the 10 ss (Fig. 3B; n=28 embryos from three separate clutches with contribution to the aorta, but not the vein). Taken together, these results suggest that Notch activation marks endothelial progenitors committed to a dorsal aorta fate.

Within the developing zebrafish vascular system, notch1a, notch1b and notch3 are expressed in arterial endothelial cells (Geudens et al., 2010; Lawson et al., 2001; Leslie et al., 2007) (supplementary material Fig. S2A). To determine their contribution to endothelial Notch activation and arterial differentiation, we assessed expression of tp1:egfp and endogenous efnb2a (ephrin B2a) in embryos lacking different combinations of Notch receptors. Knockdown of notch1a or notch1b alone had no effect on Notch activation at early stages (supplementary material Fig. S2B; data not shown), whereas notch3^fh332 mutant embryos lost tp1:egfp expression in endothelial progenitors compared with wild-type siblings (Fig. 4A,B). Despite this defect, notch3^fh332 mutant embryos exhibited both tp1:egfp and efnb2a expression at 24 hpf in dorsal aorta endothelial cells, albeit at lower levels compared with wild-type siblings (Fig. 4C,D), suggesting a compensatory role for notch1a or notch1b. Indeed, tp1:egfp and efnb2a expression was extinguished in notch3^fh332 mutant embryos injected with a notch1b morpholino oligonucleotide (MO), but reduction of notch1b alone had no effect (Fig. 4E,F). Knockdown of notch1a alone also had no major effect on tp1:egfp or efnb2a expression (supplementary material Fig. S3A,B), similar to notch1a^tp37 mutant embryos (data not shown), which are adult viable (van Eeden et al., 1996). However, combined loss of notch1a and notch1b, or notch1a and notch3, further reduced tp1:egfp and efnb2a expression (supplementary material Fig. S3C,D).

Loss of general Notch signaling components (e.g. Rbpj) causes circulatory defects, including cranial hemorrhage and arteriovenous shunts between trunk blood vessels, and increased endothelial cell numbers in sprouting intersomitic vessels (ISVs) (Lawson et al., 2001; Leslie et al., 2007; Siekmann and Lawson, 2007b). However, embryos lacking each individual receptor exhibited normal trunk circulation (supplementary material Movies 1-4 and Fig. S4A), whereas only notch3^fh332 mutant embryos exhibited cranial hemorrhage (supplementary material Fig. S4A), as recently reported (Wang et al., 2014; Zaucker et al., 2013). We also noted increased numbers of ISV endothelial cells only in embryos deficient for notch1b (supplementary material Fig. S4B). Although loss of single receptors did not affect circulatory patterns, combined loss of different receptor pairs caused qualitatively distinct defects. For example, reduction of notch1b in notch3^fh332 mutant embryos increased the penetrance of cranial hemorrhage and caused loss of trunk blood flow (supplementary material Fig. S2C and Fig. S4A). In some of these cases, a poorly formed dorsal aorta and cardinal vein accompanied the circulatory block (supplementary material Movie 5), although active arteriovenous shunts were not generally observed (supplementary material Fig. S4A). We also noted embryos in which aorta and vein morphology appeared normal, but few blood cells were in circulation (supplementary material Movie 6), possibly due to the occurrence of severe cranial hemorrhage. Injection of notch1b MO into notch1a^tp37 mutant embryos also caused a block in trunk circulation, although we never observed hemorrhage in these embryos (supplementary material Fig. S4A). Furthermore, arteriovenous shunts between the aorta and vein were apparent in notch1a^tp37;notch1b MO embryos (supplementary material Movie 7 and Fig. S4A). Taken together, these observations suggest multiple Notch receptors together contribute to artery differentiation during vascular development. However, there appear to be distinct roles for receptors in governing different aspects of subsequent vascular morphogenesis and circulatory function.

In mouse embryos, the Notch ligand Dll4 is required for both artery differentiation and angiogenesis (Hellström et al., 2007; Krebs et al., 2004). However, zebrafish embryos deficient for dll4 display hyperangiogenesis but normal artery differentiation (Leslie et al., 2007; Siekmann and Lawson, 2007b), suggesting a role for other ligands. A possible candidate is Deltac (Dlc), which is similar to both Dll1 and Dll4, and is expressed in posterior somitic mesoderm and in arterial endothelial cells (Smithers et al., 2000). In Tg(tp1:egfp)^um14 embryos, Deltac exhibits punctate localization, indicative of its engagement with Notch receptors (Wright et al., 2011), within somite cells adjacent to Etv2/EGFP-positive cells (Fig. 5A,B). These observations suggest that Deltac may be responsible for Notch activation in endothelial progenitors. Accordingly, egfp expression was lost in etv2-positive angioblasts of Tg(tp1:egfp)^um14 embryos injected with deltac MO, compared with control MO (Fig. 5C,D). Despite this effect, zebrafish mutant for dlc survive to adulthood (Jülich et al., 2005) and exhibit normal artery identity and cardiovascular function (supplementary material Fig S4C and Movie 8; data not shown) as described previously (Shaw et al., 2006). However, reduction of dll4 in dlc^tit446 mutant embryos, but not wild-type siblings, caused loss of both efnb2a and tp1:egfp transgene expression in arterial endothelial cells (Fig. 5E,F; data not shown). Likewise, circulation was normal in dll4-deficient wild-type embryos, whereas heterozygous and homozygous dlc^tit446 embryos lacking dll4 showed increasing penetrance of trunk circulatory defects, including arteriovenous shunts (supplementary material Fig. S4C and Movies 9-11). Additionally, combined loss of dlc and dll4 led to low penetrance of cranial hemorrhaging, similar to notch3^fh332 mutant embryos (data not shown).

ISV endothelial cells sprout from the dorsal aorta and initially exhibit artery identity (Bussmann et al., 2010; Geudens et al., 2010; Isogai et al., 2003; Siekmann and Lawson, 2007b). However, approximately half of the ISVs later regress from the aorta and connect to intersomitic vein (ISVe) sprouts from the posterior cardinal vein to establish a functional circulatory loop (Isogai et al., 2003). During this process, arterial endothelial cells in remodeled ISVes appear to progressively lose arterial identity (Bussmann et al., 2010). Furthermore, notch1b and dll4 are required for maintenance of arterial ISV connections to the dorsal aorta, suggesting a role for Notch in regulating this process (Geudens et al., 2010; Leslie et al., 2007). Because our initial lineage-tracing observations suggested that early Notch activation led to arterial endothelial commitment, we wanted to determine whether early activation might influence the identity of ISV endothelial cells. For this purpose, we pulsed Tg(tp1:creert2)^jh12;(actb2:loxP-stop-loxP;nred)^jh15 embryos with tamoxifen at different stages and observed the distribution of labeled cells in intersomitic arteries (ISA) or ISVe at 60 hpf. Interestingly, cells labeled at the 5-10 ss were equally distributed between ISA or ISVe (Fig. 6A,B), despite their initial restriction to the dorsal aorta (see Fig. 3). Thus, early Notch activation can bestow positional determination but does not irrevocably commit cells to an arterial fate. A similar distribution in ISV contribution was apparent until 36-48 hpf, when there was a clear shift toward a contribution to only intersomitic arteries (Fig. 6B; data not shown).

Our results suggest that ISV endothelial cells derive from an early Notch-positive angioblast and that persistence or downregulation of Notch activity in these cells determines identity as an artery or a vein, respectively. To further investigate this possibility, we utilized Tg(tp1:kaede)^um15 embryos in which the tp1 element drives expression of the photoconvertible Kaede fluorescent protein (Clements et al., 2011). Following exposure of Tg(tp1:kaede)^um15 embryos to UV light at 36 hpf, we observed photoconverted red fluorescence in ISVs at 60 hpf (Fig. 6C), consistent with initial Notch activation in these cells at 36 hpf. Furthermore, ISVs showing green fluorescence, indicative of persistent Notch signaling following photoconversion, retained their connection to the dorsal aorta (Fig. 6C,D, white arrows), whereas those displaying only red fluorescence were connected to the posterior cardinal vein (Fig. 6C,D, yellow arrows). These results suggest that ISV endothelial cells derive from angioblasts in which Notch was activated but subsequent reduction in Notch activity in some cells caused them to lose artery identity and become incorporated into veins. Given the importance of notch1b and dll4 in the control of ISV identity (Geudens et al., 2010; Leslie et al., 2007), we reasoned that they maintain Notch activation at later developmental stages in endothelial cells lining the ISVs. Accordingly, loss of either notch1b or dll4 completely eliminated tp1:egfp expression in ISV endothelial cells, at 30 hpf (Fig. 6E-G), whereas loss of notch3 had no effect (Fig. 6H). We also observed decreased egfp levels within the neural tube of notch1b-deficient embryos (Fig. 6F). Interestingly, tp1:egfp levels in the dorsal aorta were also decreased at 30 hpf following loss of notch1b or dll4, although they were individually dispensable for egfp expression at earlier stages (compare Fig. 6F,G; Fig. 4E, left panel; data not shown). Consistent with our observations and previously published studies (Geudens et al., 2010), most ISVs in dll4- or notch1b-deficient embryos retain venous connections, whereas arterial connections are normal in dlc^tit446, notch1a^tp37 and notch3^fh332 mutant embryos (supplementary material Fig. S5). These observations suggest that dll4 and notch1b are specifically required for maintaining Notch activation and arterial identity in endothelial cells at later stages of vascular development.

Our observations using zebrafish bearing tp1-driven transgenes, together with genetic manipulation of Notch receptors and ligands, suggest a series of Notch-dependent decisions that pattern the arterial system in the zebrafish embryo (Fig. 7). We first observe Notch activation in etv2-positive endothelial progenitors at the 5 ss and lineage tracing demonstrates that these cells, or their progeny, are committed to contribute to the dorsal aorta. These results are consistent with previous lineage analyses in zebrafish, which demonstrated that cells within the lateral mesoderm are restricted to artery or vein fate as early as the 7 ss (Kohli et al., 2013; Zhong et al., 2001). We postulate that cells committed to the aorta in those studies probably exhibited Notch activation. At this early stage, endothelial Notch activation appears to be induced by Deltac-positive somitic mesoderm. However, this signaling event does not irrevocably determine identity because progeny can later be found in intersomitic veins. Surprisingly, notch3 and dlc, which are both required for early Notch activation in etv2-positive cells, alone are dispensable for arterial differentiation. Rather, early Notch activation might simply provide a positional cue to separate presumptive dorsal aorta and cardinal vein cell populations. Even in the absence of this signal, a general migration of all mesodermal types, including endothelial progenitors, to the midline occurs during somitogenesis. At this point, stochastic sorting could allow sufficient numbers of endothelial cells to be in position to receive a Notch-dependent, pro-artery differentiation signal in the absence of an initial directed guidance cue. This might provide an explanation of why early Notch activation by notch3 and dlc is dispensable for later artery differentiation.

Temporal analysis of tp1:egfp expression suggests that a second wave of Notch activation occurs after cells reach the midline and begin to form the dorsal aorta. Our results suggest that at least three different Notch receptors (notch1a, notch1b and notch3), together with two separate ligands (dll4 and dlc) contribute to overall activation in this developmental window and that this activation contributes, in part, to artery differentiation. This is consistent with our previous work demonstrating that endothelial cell-autonomous Notch activation in this window induces artery differentiation and preferential dorsal aorta localization (Lawson et al., 2002; Siekmann and Lawson, 2007b). Whether the requirement for multiple receptors reflects simultaneous signaling to a required activation threshold, or is reflective of qualitatively distinct outputs is not clear. The partial reduction in arterial tp1:egfp and efnb2a expression at 24 hpf following loss of single Notch receptors, along with complete loss accompanying deficiency in different receptor combinations might suggest the former. However, differences in circulatory defects associated with distinct receptor knockdowns, as well as unique phenotypes associated with single receptor deficiencies clearly suggest there are qualitative differences between receptor outputs (Fig. 7). In any case, a better understanding of the downstream regulatory cascades that are affected by each receptor may begin to provide some insight into how they work together to contribute to artery differentiation and vascular morphogenesis.

Although multiple notch receptors appear to contribute to arterial differentiation in the aorta, there are also subsequent stage- and cell-type-specific requirements for notch1b. Together with dll4, notch1b is essential for maintaining Notch activation in arterial endothelial cells in both the dorsal aorta and developing ISAs at later stages. Previous studies have shown that hyperangiogenesis defects are apparent in embryos deficient only for notch1b or dll4 (Leslie et al., 2007; Siekmann and Lawson, 2007b), suggesting a specific role of this ligand-receptor pair in controlling angiogenic cell behaviors. Taken with previous observations, our current work suggests that dll4 and notch1b promote the decision of an arterial endothelial cell to remain as such rather than becoming a tip cell and sprouting out of the dorsal aorta (Fig. 7, ∼24-30 hours). We envisage that this process is probably governed by lateral inhibition, in which Notch acts to define distinct fates in two adjacent equipotent cells. This is consistent with previous models in which high levels of Vegfa, experienced by cells in the dorsal aorta near the somite boundary, induces expression of the Notch ligand Dll4 (Siekmann and Lawson, 2007a). This would in turn activate Notch1b in the adjacent cell to prevent migration out of the dorsal aorta and reinforce an arterial endothelial cell fate. Thus, the second wave of tp1:egfp induction observed as cells reach the midline might be, in part, a consequence of Notch activation in the selection of patent arterial endothelial cells versus sprouting tip cells (Fig. 7). It is likely that this process is then reiteratively used following cell division of the tip cell in the sprouting ISV. Whether this decision point is linked to the later fate of the ISV as arterial or venous is unknown. Based on the pattern of tp1:egfp reporter expression and lineage tracing of Notch-activated cells during sprouting, we would propose that these are distinct processes.

An interesting finding from our lineage-tracing studies is that cells in which Notch was activated at early stages can later contribute to ISVes. Despite the initial positional commitment of these Notch-activated cells to the dorsal aorta, our observations suggest that arterial endothelial cells can later de- or trans-differentiate to become venous tissue. These findings are consistent with those in chick embryos, which found that arterial endothelial cells transplanted into a vein would adopt venous identity, although this plasticity was limited to the earliest stages of development (Moyon et al., 2001). Accordingly, veins grafted into coronary arteries in rats exhibit de-differentiation of venous identity, but fail to acquire artery marker gene expression, suggesting limited plasticity in adult vessels (Kudo et al., 2007). Despite this finding, ectopic activation of Notch in adult vasculature is capable of modulating artery identity, leading to arteriovenous malformations (Carlson et al., 2005), suggesting that mature endothelial cells retain the ability to re-program their identity under some settings. Our results suggest that venous reprogramming is associated with downregulation of Notch activity, consistent with previous work in which notch1b or dll4 deficiency causes the formation of excessive venous ISV connections (Geudens et al., 2010). Interestingly, venous ISVs in which Notch signaling is downregulated do not appear to gain a hyper-angiogenic phenotype, as is observed when Notch signaling is lost in all ISVs [e.g. dll4 deficiency (Leslie et al., 2007; Siekmann and Lawson, 2007b)]. This might reflect an early requirement for Notch in limiting angiogenesis only during initial ISV sprouting when there is probably an increase in pro-angiogenic signaling, compared with later stages when ISV remodeling occurs. Thus, it is likely that notch1b and dll4, which are required to both limit angiogenesis and maintain ISV artery identity, are acting in distinct contexts during each of these processes (Fig. 7).

A number of approaches have been applied to determine the spatial and temporal control of Notch activation during development. These include transgenic mice bearing a Notch-responsive element comprising three Rbpj-binding sites (Duncan et al., 2005) or the tp1 element driving lacZ (Souilhol et al., 2006). Both lines have been helpful to visualize Notch activation in the embryonic and postnatal vascular system. Although tp1-driven Cre lines have not been developed, Vooijs et al. generated mice in which the intracellular domain of the Notch1 receptor was fused in frame with Cre recombinase (NIPCre) to allow lineage tracing of progeny, and these cells exhibit Notch1 activation (Vooijs et al., 2007). However, NIPCre-labeled cells are permanently marked by the initial Notch signal, making it difficult to assess dynamic changes. These mouse models also suffer from the difficulty in making serial temporal observations during development. In this regard, our application of the tp1 element in zebrafish embryos, which are transparent and develop externally, provides a major benefit. Using a combination of tp1-driven transgenes and genetic manipulation of Notch signaling components, we were able to ascertain when and where particular Notch ligands and receptors are required for activation. In parallel, our use of a tp1:CreERT line allowed us to follow progeny cells after Notch activation. In this case, the ability to limit tamoxifen treatment to a narrow developmental stage allowed precise definition of when Notch signaling is affecting lineage decisions. Despite the benefits of the tp1 promoter, a drawback is that it reports signaling mediated through Rbpj, which is commonly used by all Notch receptors for target gene expression (Guruharsha et al., 2012), making it challenging to discern receptor-specific signaling events. Although combining genetic manipulation of Notch signaling with the tp1-driven transgenes has been helpful, generation of receptor-specific sensors, similar to NIP-Cre, or Gal4 fusions used successfully in Drosophila (Struhl and Adachi, 1998), will be needed moving forward.

Zebrafish (Danio rerio) and their embryos were handled according to standard protocols and in accordance with accepted University of Massachusetts Medical School IACUC protocols. The establishment and characterization of Tg(fli1ep:dsredex)^um13, Tg(tp1:egfp)^um14, Tg(tp1:creert2)^jh12, Tg(tp1:kaede)^um43, Tg(actb2:loxP-stop-loxP;hmgb1-mcherry)^jh15 [referred to hereafter as Tg(actb2:loxP-stop-loxP;nred)^jh15], notch1a^tp37, deltac^tit446, dll4^j16e1 and notch3^fh332 have been described elsewhere (Clements et al., 2011; Covassin et al., 2009; Holley et al., 2002; Jülich et al., 2005; Leslie et al., 2007; Parsons et al., 2009; Zaucker et al., 2013). Lines generated in this study are described below.

To construct a conditional endothelial marker transgene, we cloned the fli1ep fragment (Villefranc et al., 2007) into pDestTol2pA (Kwan et al., 2007) using PspOMI after removing ATGs between the PspOMI site and attR4 site to give pTol-fli1epR4-R3pA. In parallel, a monomeric form of mCerulean fused to H2B was cloned between direct loxP repeats flanked by attL4 and attR1 to give p5E-loxP-nBlue-loxP. We performed a multisite LR reaction using p5E-loxP-nBlue-loxP, pME-mCherry and an empty p3E-mcs-1 vector with pTol-fli1epR4-R3pA to give pTol-fli1ep:loxP-nBlue-loxP;mcherry.

deltac, dll4 and notch1a morpholino oligonucleotides (MOs) are described elsewhere (Holley et al., 2002; Siekmann and Lawson, 2007b). The sequence of the splice blocking Notch1b MO, which targets the exon 8 splice donor site, is 5′-GTTCCtCCgGTTACCTGGCATACAG-3′; lowercase letters denote two mismatches compared with Zv9 reference sequence. MOs were dissolved in nuclease-free water (Ambion). Embryos from incrosses of indicated heterozygous carriers or wild-type adults were injected at the one- to two-cell stage with 15 ng dll4, 2.5 ng notch1b MO, or 1 ng notch1a MO, respectively, or comparable amount of scrambled (control) MO. We confirmed the effect of the notch1b on splicing by RT-PCR of notch1b from RNA isolated from MO-injected embryos using the following primers: forward, 5′-GTTGGGCCGGCCTGTGAC-3′; reverse, 5′-CTCGCACAGACGGCCC-3′. The Tg(fli1ep:loxP-nblue-loxP;red)^um43 line was generated using pTol-fli1ep:loxP-nBlue-loxP;mcherry by Tol2 transposition as described previously (Covassin et al., 2009).

Whole-mount in situ hybridization was performed using standard protocols (Hauptmann and Gerster, 2000). efnb2a riboprobe preparation has been described elsewhere (Lawson et al., 2001). The antisense egfp riboprobe was prepared by linearizing pENTR-EGFP2 (Villefranc et al., 2007) with NcoI followed by in vitro transcription using T7 polymerase. Stained embryos were imaged by differential interference contrast (DIC) microscopy using an Axiophot2 (Zeiss) and a W Plan-Apochromat 20×/1.0 (NA) DIC D=0.17 M27 75 mm, and images were captured using an AxioCam hRC (Zeiss). For fluorescence in situ hybridization, etv2 and egfp templates were generated by PCR with a custom SP6 (5′-gcttgatttaggtgacactat-3′) and a standard M13 reverse primer from pCS2-etv2 (Moore et al., 2013) and pCS2-egfp (Villefranc et al., 2007) templates, respectively. Antisense etv2 or egfp riboprobes were synthesized from these templates using T3 RNA polymerase and DIG-dUTP or fluorescein-dUTP, respectively. Hybridization and detection was carried out using the TSA Plus system (PerkinElmer) as follows: Tg(tp1:egfp)^um14 embryos at 10-12 ss were fixed with 4% paraformaldehyde in 1× phosphate-buffered saline (PFA/PBS) overnight at 4°C, transferred to methanol, and washed with 3% H₂O₂/methanol for 1 hour. Embryos were washed with methanol and stored at -20°C until use. Standard hybridization and washing was performed (Hauptmann and Gerster, 2000) and embryos were placed in maleic acid buffer (MAB; 150 mM maleic acid/100 mM NaCl (pH 7.5)/2% blocking reagent; Roche). Embryos were incubated overnight at 4°C in MAB containing 1:1000 sheep anti-DIG antibody conjugated to horseradish peroxidase (POD; Roche). Embryos were washed with MAB/0.001% Tween 20 six times at room temperature, once with PBS, and placed in 1:50 tyramide-Cy3/1× amplification diluent buffer for 3 hours to detect etv2 mRNA. Embryos were washed with PBS/0.1% Tween 20 (PBST) overnight at room temperature and placed in methanol. To quench POD activity, stained embryos were treated with 3% H₂O₂/methanol for 30 minutes at room temperature and washed with PBST, followed by MAB/0.001% Tween 20. Egfp mRNA was detected as above, except that 1:500 POD-conjugated anti-fluorescein (Roche) and 1:50 tyramide-FITC were used. Double-stained embryos were briefly washed with PBST, flat mounted, and imaged with a Leica confocal TCS SP2 microscope using an oil immersion 40× HC x PL APO CS (NA=1.25). Vertical projections were assembled using Imaris or the Fiji package of ImageJ.

Tg(tp1:egfp)^um14 embryos were fixed in 4% PFA/PBS for 15 minutes at room temperature, washed with methanol, quenched with 3% H₂O₂ in methanol for 1 hour at room temperature and stored in methanol at -20°C before use. Embryos were washed with TBST (Tris-buffered saline/0.1% Tween 20) plus 0.1% Triton X-100 and permeabilized with TBST/0.5% Triton X-100 for 30 minutes at room temperature. After blocking with TBST containing 1% BSA and 10% goat serum, embryos were incubated with 1:250 mouse anti-DeltaC antibody (zdc2; Abcam) and 1:500 rabbit anti-zebrafish Etv2 (Moore et al., 2013) in blocking buffer overnight at 4°C. Embryos were washed briefly and incubated with 2% blocking reagent/MAB, followed by HRP-conjugated goat anti-mouse IgG (H+L) at 1:1000 and goat anti-rabbit IgG (H+L) conjugated to Alexa Fluor 633 (Invitrogen, cat. no. A-21070) at 1:500 overnight at 4°C. The embryos were washed with MAB, followed by 1× PBS and incubated with tyramide-Cy3 diluted 1:50 in 1× amplification diluent (PerkinElmer) for 3 hours at room temperature. Subsequently, HRP activity was quenched with 3% H₂O₂/methanol for 1 hour at room temperature. Embryos were washed in TBST, followed by incubation with 2% blocking reagent/MAB and then 1:1000 goat anti-GFP-HRP (Rockland). After washing with MAB and 1× PBS, embryos were incubated with 1:50 tyramide-FITC in 1× amplification diluent (PerkinElmer). Embryos were flat mounted and imaged using confocal microscopy as above.

Embryos from crosses between Tg(tp1:creert2)^jh12 and either Tg(actb2:loxP-stop-loxP;nred)^jh15 or Tg(fli1ep:loxP-nblue-loxP;nred)^um43 carriers were placed in standard egg water containing 0.5 μM 4-hydroxy-tamoxifen (Sigma) diluted from a 10 mM stock in ethanol for 1 hour at the indicated developmental stages. Embryos were washed successively with 800 ml and 50 ml egg water and grown to the indicated stage. Tamoxifen-treated Tg(tp1:creert2)^jh12;(actb2:loxP-stop-loxP;nred)^jh15 embryos were imaged at 30 hpf by DIC and epifluorescence microscopy using an Axioplan2 and a water immersion 20× lens (NA=1) or subjected to microangiography using blue fluorescent 0.02 μm carboxylate-modified Fluospheres (Life Technologies) and imaged with a LSM 7MP laser scanning microscope (Zeiss) equipped with a Chameleon Ti:Sapphire pulsed laser (Coherent) using ZEN 2009 software (Zeiss). Red, green and blue fluorescence was sequentially excited at 1034, 890 and 760 nm, respectively. Emissions were captured using separate detectors with 565-610 (red), 500-550 (green) and 400-480 nm (blue) band pass filters. Labeled progeny cells in Tg(tp1:creert2)^jh12;(fli1ep:loxP-nblue-loxP;red)^um43 embryos were similarly imaged by two-photon microscopy using sequential excitation at 1034 and 890 nm. Blue fluorescence was detected using a band pass filter at 440-475 nm and mCherry was detected as above. For assessment of contribution to arterial or venous ISVs, Tg(tp1:creert2)^jh12;(fli1ep:loxP-nblue-loxP;red)^um43 embryos were observed at 60 hpf after tamoxifen treatment as described in the text. ISVs in which labeled cells were found were defined as arterial or venous based on the direction of circulatory flow as determined by light microscopy.

Tg(tp1:kaede)^um43 embryos were mounted on depression slides in 5% methylcellulose at 36 hpf and illuminated for 20 seconds with UV-wavelength light on an Axiophot2 (Zeiss) microscope using a 20× lens (NA=1). Both red and green emissions were immediately checked following UV exposure to confirm complete photoconversion. Embryos were returned to egg water at 28°C until 60 hpf. Red and green fluorescence were detected on a Leica confocal TCS SP2 microscope or embryos were subjected to microangiography and two-photon imaging as above.

Embryos deficient for different combinations of Notch receptor or ligand function were observed, using a dissection microscope at 72 hpf, for circulation through the trunk blood vessels and presence of cranial hemorrhage. Following phenotypic assessment, embryos were subjected to genotyping using primers flanking mutations in dlc (forward, 5′-GAAACAAACATTGACGACTGCTCAAGCGACCCC-3′; reverse, 5′-GTCCTTGCCTGAGAAACC-3′) or notch3 (forward, 5′-CCCTGAAGGGTTCCATGATCCCTACTACTA-3′; reverse, 5′-TCCAGGCTCACAGTCACACCGATA-3′). Amplicons were digested with either PshAI (dlc) or SpeI (notch3) and separated by gel electrophoresis. Circulatory defects were captured by video micrography on a Zeis Axophot2 microscope equipped with a DMK 21AU04 digital video camera (Imaging Source). Movies were labeled and compressed using Flash (Adobe) and Quicktime (Apple) software, respectively.

We would like to thank present and past members of the Lawson laboratory for helpful insights in the course of this work. We thank George Kourkoulis and Ira Male for technical assistance. We also thank John Polli and Ed Jaskolski for fish care and maintenance.