Zebrafish neuropilins are differentially expressed and interact with vascular endothelial growth factor during embryonic vascular development



Neuropilin1 (Nrp1) and Neuropilin2 (Nrp2) are nonkinase vascular endothelial growth factor receptors (VEGFR) identified in several vertebrates, which function as coreceptors for the receptor tyrosine kinases VEGFR1 and VEGFR2. We identified four zebrafish nrp genes, nrp1a, nrp1b, nrp2a, and nrp2b, and characterized their function in vascular development. We show that all nrp genes display distinct expression patterns and that nrp1a and nrp1b are expressed in the dorsal aorta, while nrp2a and nrp2b transcripts could be detected in the region of the posterior cardinal vein. Knockdown of nrp1a, nrp1b, and nrp2a resulted in improper arteriovenous connections and irregular intersegmental vessel patterning, indicating that these Nrps function in the same process. Nrp2b knockdown also caused vessel malformations and a pericardial defect. In addition, we provide evidence that the newly identified Nrps synergistically interact with VEGF in vivo. Taken together, our results show that, in zebrafish, all four neuropilins are involved in VEGF-mediated vessel development. Developmental Dynamics 231:33–42, 2004. © 2004 Wiley-Liss, Inc.


Blood vessels develop by either one of two different processes. The de novo formation of blood vessels by endothelial progenitors is termed vasculogenesis, while angiogenesis describes the sprouting from pre-existing vessels (reviewed in Carmeliet, 2000). Arteries and veins also have to differentiate during these processes. Recent studies provide insight into the molecular signals involved in arteriovenous differentiation of endothelial cells. A role for vascular endothelial growth factor (VEGF) in promoting arterial identity of endothelial cells during embryonic development has been described, placing VEGF downstream of Sonic Hedgehog and upstream of Notch signaling (Lawson et al., 2002). In a similar way, VEGF contributes to the arteriovenous differentiation of endothelial cells in the neonatal retina and in the adult heart (Mukouyama et al., 2002; Stalmans et al., 2002; Visconti et al., 2002).

Signaling of VEGF is mediated through the receptor tyrosine kinases VEGFR1 (Flt1) and VEGFR2 (Flk1; reviewed in Neufeld et al., 1999). In addition, the Neuropilin (Nrp) family of type I transmembrane proteins was identified as splice-form–specific receptors for VEGF (Soker et al., 1998). In humans, this small family includes two known members, Nrp1 and Nrp2 (Chen et al., 1997; Kolodkin et al., 1997; Takagi et al., 1991). Both Nrp proteins were initially identified as receptors for the class 3 family of Semaphorins. In this context, Nrps form complexes with plexins (Takahashi et al., 1999) and mediate chemorepulsive axon guidance (reviewed in Neufeld et al., 2002).

Both nrp genes show endothelial cell-specific expression. In chick and mouse, expression of nrp1 is localized in arteries, while nrp2 is expressed in veins (Herzog et al., 2001; Moyon et al., 2001; Yuan et al., 2002). In vitro studies suggest that Nrps act as coreceptors for the endothelial cell-specific VEGF receptors. When coexpressed with VEGFR2 in porcine aortic endothelial cells, Nrp1 enhanced the binding affinity of VEGF165 to VEGFR2 (Soker et al., 1998). Similarly, Nrp2 forms complexes with VEGFR1 in vitro (Gluzman-Poltorak et al., 2001).

Functional evidence that Nrps are required during physiological vascular development stems from knockout mice deficient for either nrp1, or nrp2, or for both genes. Nrp1-deficient mice die at embryonic day (E)13.5 and displayed partially impaired vascularization in the nervous system, improper development of branchial arch arteries, and a disorganized vascular network in the yolk sac (Kawasaki et al., 1999). In contrast, overexpression of nrp1 resulted in reverse defects of the vasculature, including excess capillary and blood vessel formation, malformed hearts, and hemorrhaging (Kitsukawa et al., 1995). In nrp2 mutants, the defect of the cardiovascular system was less severe. The formation of small lymphatic vessels and capillaries was abnormal, but many nrp2 homozygous mutant mice were viable until adulthood (Yuan et al., 2002). Double knockout analysis of both nrp genes revealed a more severe effect on the embryonic phenotype than the single nrp knockouts (Takashima et al., 2002). Double nrp1/nrp2 homozygous mutant mice die at E8.5 and the vasculature was severely abnormal, including a completely avascular yolk sac. These findings indicate an important role for Nrps in early embryonic vessel development and a partially overlapping requirement for both Nrp1 and Nrp2.

In zebrafish, so far only one Nrp1 has been described, mediating VEGF-dependent angiogenesis (Lee et al., 2002). We identified three novel zebrafish nrp genes and provide in this study the expression patterns and knockdown phenotypes of all four nrp genes. We show that both nrp1 and nrp2 genes show complementary expression in the dorsal aorta and around the posterior cardinal vein, respectively. Arteriovenous malformations characterized the knockdown phenotypes. In addition, we provide evidence that all Nrps interact synergistically with VEGF signaling in vivo. Our findings support the idea of Nrps playing a role during embryonic vascular development in a VEGF-dependent manner.


Cloning of Four nrp Genes in Zebrafish

To study the role of Nrps in blood vessel formation, we cloned the zebrafish homologues of nrp. A reverse transcriptase-polymerase chain reaction (RT-PCR) approach using degenerate primers led to the identification of three nrp genes. A subsequent blast search on genomic zebrafish sequences revealed the existence of a fourth zebrafish nrp gene. Full-length cDNAs of these four genes were obtained by rapid amplification of cDNA ends (RACE). All four genes encode a domain structure that is characteristic for members of the Nrp family, including five extracellular domains, a transmembrane domain, and a short cytoplasmic domain.

A phylogenetic tree resulting from BLAST search data showed that two of the cloned genes are more similar to the nrp1 family members, whereas the other two genes show more similarity to the nrp2 family members. Therefore, the respective genes were designated nrp1a and nrp1b as well as nrp2a and nrp2b (Fig. 1). The gene designated as nrp1a corresponds to the previously published znrp1 (Lee et al., 2002). A comparison of the predicted protein sequences revealed an overall amino acid sequence identity of 59% between Nrp1a and Nrp1b as well as 73% between Nrp2a and Nrp2b. Zebrafish Nrp1a and Nrp1b are more closely related to human Nrp1 (64 and 60% identical to human Nrp1, compared with 45 and 43% amino acid identity to human Nrp2, respectively), whereas zebrafish Nrp2a and Nrp2b share more sequence identity with human Nrp2 (57 and 59% identical to human Nrp2, compared with 51 and 45% amino acid identity to human Nrp1, respectively). Thus, the zebrafish genome encodes two nrp1 and two nrp2 genes.

Figure 1.

Phylogenetic tree of Neuropilin amino acid sequences. The zebrafish genome contains two genes each that belong to the nrp-1 and nrp-2 families, respectively. h, human; m, mouse; z, zebrafish.

Nrps Show Differential Expression in Arterial and Venous Endothelial Cells

To investigate the expression of the nrp genes, whole-mount in situ hybridization analysis was carried out. The localization of nrp1a transcripts (Fig. 2A–C) has been described previously (Lee et al., 2002). Expression was reported in the major arterial and venous as well as intersegmental vessels at 48 hrs postfertilization (hpf). However, we exclusively detected vascular nrp1a expression in the dorsal aorta as early as 26 hpf (Fig. 2C, see below), whereas after 48-hpf expression was restricted to neural tissues (data not shown; Martyn, 1999). In addition, nrp1a was also expressed in the epiphysis, the pharyngeal region, the otic vesicles, and the pronephric ducts.

Figure 2.

Neuronal- and vessel-specific expression of neuropilins. A–M: In situ hybridization on wild-type embryos with antisense probes of nrp1a (A–C), nrp1b (D–F), nrp2a (G–I), and nrp2b (K–M). Stages analyzed were 18-somite stage (A,D,G,K), 26 hours postfertilization (hpf; B,C,E,F,H,I,L), and 48 hpf (M). C,F,I: Close up of trunk region such as outlined in B, E, and H, respectively. White arrowheads point to the expression in the dorsal aorta. Nrp1a is also expressed in the motoneurons (white arrow in C), nrp1b shows expression in the middle cerebral vein (black asterisk in E), the posterior intermediate cell mass (gray arrowhead in E), and in the intersegmental vessels (black arrows in F) along the somite boundary, whereas nrp2a is expressed medial of each somite (white asterisks in I) between the somite boundaries. The epiphysis expresses nrp2b (outlined black arrowheads in K and L). M: Note expression of nrp2b in the heart (black arrowhead). All embryos lateral view, anterior to the left. Scale bars = 50 μm in F,I,M, 100 μm in A–E,G,H,K,L.

The expression of nrp1b resembled the characteristic pattern of vascular markers, such as fli, flk1, and flt4 (Thompson et al., 1998). During late segmentation stage, nrp1b was expressed throughout the pharyngeal region, in endothelial precursors and sites of blood formation in the midline of the trunk and tail as well as in the dorsal-most aspect of the neural tube (Fig. 2D). In 1-day-old embryos, nrp1b expression was found in blood vessels of the head (such as the middle cerebral vein), in the developing trunk vasculature (albeit restricted to the dorsal aorta and intersegmental vessels), and in the posterior intermediate cell mass (Fig. 2E,F, see below).

The nrp2a gene was predominantly expressed in the head and somitic regions (Fig. 2G,H). Vertical stripes of expression in the trunk were localized in the medial portion of each somite (Fig. 2I).

Nrp2b was also strongly expressed in neuronal tissues, including a striking expression in the epiphysis (Fig. 2K,L). At 1 day postfertilization (dpf), expression of nrp2b became restricted to the ventral half of the trunk and tail as well as the pharyngeal region. In contrast to the other nrp genes, nrp2b transcript could be detected in the heart at 2 dpf (Fig. 2M). Taken together, the four nrp genes displayed different, yet partly overlapping expression patterns, mainly in neuronal and vascular tissues.

To obtain more detailed information about the vessel-specific expression of nrps and to address the question of whether zebrafish nrps are differentially expressed in arterial and venous endothelial cells as seen in other species, transverse sections through the trunk of 1-day-old embryos stained by in situ hybridization were analyzed. By using this approach, differential expression of nrp1a and nrp1b in the dorsal aorta but not in the posterior cardinal vein could be confirmed (Fig. 3A,B). In the case of nrp2a, in addition to its strong expression in the somites, transcripts were localized in the ventral region of the trunk, indicating expression in tissues surrounding the posterior cardinal vein, while expression ventral to the notochord in the region of the dorsal aorta was barely detectable (Fig. 3C). Sections through embryos stained for nrp2b expression revealed that the ventral expression in the trunk indeed localized around the posterior cardinal vein (Fig. 3D). Thus, vessel expression of nrp1a and nrp1b is restricted to the dorsal aorta, while nrp2a and nrp2b are not expressed in the aorta, but in cells surrounding the posterior cardinal vein.

Figure 3.

Neuropilins are differentially expressed in arterial and around venous vessels. A–D: Transverse sections through the trunk of 26 hours postfertilization whole-mount stained embryos. A,C,D: Plastic sections. Trunk is outlined with a black line. B: Thick section. Note expression of nrp1a (A) and nrp1b (B) in the dorsal aorta (arrows), whereas nrp2a (C) and nrp2b (D) show expression in cells surrounding the posterior cardinal vein (arrowheads). m, motoneurons; n, notochord; s, somites. Scale bars = 20 μm in A–D.

Knockdown of nrp1a, nrp1b, or nrp2a Results in Arteriovenous Malformations

To learn about the role of nrps during vascular development, knockdown phenotypes were analyzed by live observation and the microangiography technique (Weinstein et al., 1995). For each nrp gene, two antisense morpholino oligos, targeting the start codon and the 5′ untranslated region (UTR), were injected independently. Whereas control injections with buffer had no effect (Fig. 4A), larvae injected with either one of the nrp morpholinos displayed shunts at 2 dpf (Fig. 4B). Here, shunt describes an arteriovenous malformation that persists in the tail of the larvae and is characterized by a disruption of the vessel wall between the caudal artery and the caudal vein leading to a premature venous return of the circulating blood. In the lumen posterior to this turning point, blood cells accumulate and pulsate, indicating a functional connection with the circulation. In the knockdown larvae of the nrp1a, nrp1b, or nrp2a genes, shunts persisted from 2 to 4 dpf (indicated by arrowheads in Fig. 5).

Figure 4.

Knockdown of nrps result in arteriovenous shunts in the tail. A: A vessel boundary separates the caudal artery and caudal vein in 2 days postfertilization wild-types larvae (arrowhead). B: In nrp1a morpholino (MO) -injected larvae, lack of this vessel boundary results in a premature return of the blood (arrow) and blood accumulation in the open space posterior to the shunt. Lateral view, anterior is to the left. Scale bar = 100 μm in B (applies to A,B).

Figure 5.

Knockdown phenotypes of nrps. A–K: Live observations at 2 days postfertilization (dpf; A,C,E,G) and 4 dpf (I) as well as microangiographies at 3 dpf (B,D,F,H) and 4 dpf (K) of larvae injected with injection buffer (A,B) or antisense morpholino (MO) targeting nrp1a (C,D), nrp1b (E,F), nrp2a (G,H), and nrp2b (I,K). Insets in B,D,F,H,K show outlined regions of the larvae. Note that the points of blood return in the nrp1a, nrp1b,and nrp2a knockdown larvae is shifted anteriorly compared with wild-type larvae (black arrowheads in live observation and white arrowheads in microangiographies). In addition, these larvae display a varying degree of irregularly shaped intersegmental vessels (insets in B,D,F,H). Larvae injected with nrp2b antisense morpholino display either a heart edema (black arrow in I) with beating heart but lack of circulation or an interrupted caudal artery (white arrow in K). All larvae lateral view, anterior is to the left. Scale bars = 200 μm (whole-mounts) and 100 μm (insets).

Compared with control-injected larvae at 2 dpf (Fig. 5A), nrp1a knockdown larvae showed a reduction in body size, a small hindbrain edema, and a severe shunt that was localized in the anterior part of the tail and characterized by a large accumulation of blood cells (Fig. 5C). At 3 dpf, blood flow in patent vessels of the larvae was visualized by using the microangiography technique. Whereas the intersegmental vessels were regularly arranged along the trunk and tail of control embryos (Fig. 5B), an irregular pattern was observed in nrp1a morpholino-injected larvae (Fig. 5D). Either single intersegmental vessels were missing (inset in Fig. 5D; endogenous alkaline phosphatase staining, data not shown) or extra sprouts were present, which connected across the somites to the dorsal longitudinal anastomotic vessel (DLAV), while somite boundaries appeared morphologically normal. In larvae injected with morpholinos targeting nrp1b, the arteriovenous shunt was most often accompanied with reduced blood accumulation in the lumen posterior to the shunt compared with the nrp1a knockdown embryos, a short and bent body axis, as well as a small edema in the hindbrain (Fig. 5E). Compared with nrp1a, the nrp1b knockdown larvae displayed rather subtle intersegmental vessel irregularities with only few extra sprouts (Fig. 5F). The knockdown of nrp2a also led to shunts with little blood in the posterior lumen. However, body size was almost unaffected, and much edema in the posterior head region was characteristic for these larvae (Fig. 5G). Microangiographies of larvae injected with nrp2a morpholino also revealed an irregular pattern of the intersegmental vessels, including sprouts without through connections with the DLAV (Fig. 5H).

The vascular-specific knockdown phenotypes of nrp1a, nrp1b, and nrp2a could be observed at a penetrance of 54% (2 ng of nrp1a_ATG morpholino), 42% (18 ng of nrp1b_UTR morpholino), and 58% (6 ng of nrp2a_ATG morpholino), respectively (Table 1). Injections of a second morpholino for each nrp gene confirmed the specific knockdown phenotypes with a penetrance between 22 and 50% (Table 1). As a further control, embryos were injected with 18 ng of a control morpholino (n = 21). This amount corresponded to the maximum amount injected in the gene-specific knockdown experiments. Control embryos displayed wild-type phenotypes.

Table 1. Efficiency of nrp Antisense Morpholino Oligos
Morpholino injectedAmount injected (ng)Penetrance of phenotype (%)n
  • a

    Shunt in tail at 3 days postfertilization.

  • b

    Caudal artery interrupted.

  • c

    No circulation.


Taken together, the knockdown of nrp1a, nrp1b, or nrp2a led to similar vascular phenotypes, including arteriovenous malformations in the tail region and irregular intersegmental vessel patterning. However, subtle differences in the localization of the shunts and in the severity of intersegmental vessel irregularities were identified.

Knockdown of nrp2b Results in Vascular and Heart Defects

In contrast to the nrp knockdown phenotypes described above, the arteriovenous shunt of nrp2b morpholino-injected larvae disappeared after 2 dpf. From 3 dpf onward, two different phenotypic classes occurred when 3 ng of nrp2b_ATG morpholino were injected (Table 1). First, 13% of the larvae displayed heart edema (Fig. 5I) and lack of circulation, although the heart was still beating. Alkaline phosphatase staining of these larvae revealed a wild-type-like vasculature (data not shown). Second, in 62% of the injected larvae, an interruption of the caudal artery was observed (Fig. 5K). This defect led to a detour of the circulating blood through the intersegmental vessels and the DLAV, either entering the caudal artery again posterior to the interruption, or returning to the heart by means of the caudal vein. Except for the interrupted caudal artery, vessel patterning was wild-type-like in nrp2b knockdown larvae. Injections of higher amounts of this morpholino resulted in an increase of larvae displaying the more severe heart phenotype (see below, Table 4). We could confirm the specificity of the nrp2b knockdown phenotype by injection of a second morpholino, targeting a different sequence within the 5′ UTR of this gene (Table 1). Thus, unlike the other nrp genes, nrp2b knockdown resulted in malformation specifically of the caudal artery and in a heart phenotype.

Table 4. Analysis of 3-Day-Old Larvae Coinjected With Morpholinos against vegf and nrp2ba
Morpholino injectedAmount injected (ng)Phenotypic class (%)n
Wild-typeCA interruptedNo circulation/HeShunt or HeNo circulation/blood in trunkAbnormal
  • a

    CA, caudal artery; He, heart edema.


Double Knockdown Phenotypes of nrps and vegf Reveal Synergistic Interactions

Studies in zebrafish provided the first in vivo evidence that Nrp1a regulates angiogenesis in a VEGF-dependent pathway (Lee et al., 2002). We confirmed the cooperation of VEGF and Nrp1a by coinjection of the respective morpholinos (data not shown). To investigate the question whether the same is true for the newly identified Nrps, we coinjected vegf and nrp antisense morpholino oligos. Injected larvae were analyzed by live observation at 3 dpf, a stage when embryos with arteriovenous shunts could be evaluated most reliably.

Vegf knockdown embryos develop blood vessel deficiencies that range from weak to major phenotypes, depending on the amount injected (Nasevicius et al., 2000). Phenotypes ranged from loss of very few intersegmental vessels to nonfunctional vasculature accompanied by a pericardial edema and blood accumulation in the ventral trunk. We also observed shunts when lower amounts of vegf morpholino were injected. In our experiments, vegf morpholino-injected embryos were classified into moderate shunt- and heart edema- as well as strong no circulation-phenotypes. The abnormal class included embryos with gastrulation defects or necrotic phenotypes and lysed embryos.

While single injections of 3 ng vegf or 9 ng nrp1b morpholino had no significant effect on the vasculature (vegf, 94% wild-type; nrp1b, 100% wild-type; Table 2), in coinjections of the same amounts of vegf and nrp1b morpholinos more than half of the injected larvae displayed either a shunt (38%) or lacked circulation with blood accumulating in the trunk (20%). Doubling the amount of injected morpholino resulted in 12% wild-type, 21% arteriovenous shunts, and 67% without circulation, compared with 37% wild-type and 45% without circulation in vegf and 71% wild-type and 27% shunts in nrp1b single knockdowns (Table 2). Thus, compared with single morpholino injections of each gene, double knockdown of vegf and nrp1b resulted in a phenotypical shift to more severe vascular defects, suggesting an interaction of VEGF and Nrp1b in vivo.

Table 2. Analysis of 3-Day-Old Larvae Coinjected With Morpholinos Against vegf and nrp1ba
Morpholino injectedAmount injected (ng)Phenotypic class (%)Abnormaln
Wild-typeShuntHeart edemaNo circulation/blood in trunk
  • a

    Note that these injections were performed on different days than the other vegflnrp morpholino coinjections explaining the subtle differences in the penetrance of vegf knockdown phenotypes compared to the other experiments (see Tables 2 and 3).


In double knockdown experiments of vegf and nrp2a, similar results were obtained. Whereas single injections of 1.5 ng of vegf or 3 ng of nrp2a morpholino on their own again had no visible effect (100% wild-type, respectively; Table 3), coinjection of the same amounts of vegf and nrp2a morpholinos led to a reduction of the wild-type proportion to 76% and an increase of injected larvae displaying a shunt to 18%. Doubling the amount of each injected morpholino confirmed this synergistic effect of VEGF and Nrp2a. Only 4% of injected larvae were phenotypically wild-type compared with 79 and 39% in the vegf or nrp2a single injections. The phenotypic class of embryos without circulation increased to 74% compared with 4% and null in the single injections, respectively. When higher amounts of morpholinos were injected, results were consistent with the effect described above (Table 3). However, 12 ng of nrp2a morpholino caused toxic effects, which increased the number of abnormally shaped embryos. Thus, these coinjection experiments demonstrate that vegf and nrp2a act synergistically.

Table 3. Analysis of 3-Day-Old Larvae Coinjected With Morpholinos Against vegf and nrp2aa
Morpholino injectedAmount injected (ng)Phenotypic class (%)n
Wild-typeShuntHe or CA interruptedNo circulation/blood in trunkAbnormal
  • a

    He, heart edema; CA, caudel artery.


Likewise, we carried out double knockdown experiments of vegf and nrp2b. In single injections with 1.5 ng nrp2b morpholino, 88% of the injected embryos showed wild-type phenotypes, 3% displayed an interrupted caudal artery, and 6% exhibited heart edema phenotypes (Table 4). Coinjection of embryos with the same amount of vegf morpholino, which had no effect in single injections, resulted in a major decrease of phenotypically wild-type larvae (25%; Table 4), and an increase of larvae with severe defects of the vasculature. Specifically, 34% showed interruptions within the caudal artery accompanied with irregular shaped intersegmental vessel, 22% displayed shunts or heart edemas, and 19% lacked circulation with accumulated blood in the trunk. We could not detect the nrp2b single knockdown heart edema phenotype, which lacks blood accumulations throughout the larvae and displays wild-type vessel patterning. Injection of higher amounts of vegf and nrp2b morpholinos resulted in a phenotypical shift toward more severe vascular phenotypes, which was more strongly pronounced in the double injections compared with the corresponding single injections (Table 4). Thus, vegf and nrp2b also interacted synergistically when both were knocked down in vivo. Taken together, Nrp1b, Nrp2a, and Nrp2b are required in proper vessel development and interact with the VEGF-pathway during this process.


Identification of Four nrp Genes in Zebrafish

We have identified four cDNAs encoding zebrafish homologues of the neuropilin gene family. The sequences of the cDNAs predict two proteins each that show homology to Nrp1 and Nrp2, respectively (Takagi et al., 1991; Chen et al., 1997). Based on sequence similarities, we designated the cloned zebrafish genes nrp1a, nrp1b, nrp2a, and nrp2b. We could not identify any further genomic sequence read in the ZFv3 Sanger assembly that showed similarity to the nrp gene family. Therefore, we predict that the zebrafish genome encodes for a total of two nrp1 and two nrp2 genes, whereas in other vertebrates only one of each has been described.

Expression Domains and Knockdown Phenotypes of Zebrafish nrps Suggest a Role in Arteriovenous Differentiation

The four nrp genes show overlapping, yet distinct expression patterns predominantly in neuronal and vascular tissues. The transcript distribution is consistent with a role of the respective genes as receptors for both Semaphorins and VEGF (He and Tessier-Lavigne, 1997; Soker et al., 1998) and with the expression described in other species (Kawakami et al., 1996; Chen et al., 1997; Herzog et al., 2001). With respect to the vasculature, nrp1 transcripts are localized in the dorsal aorta in chick and mouse embryos, whereas nrp2 is expressed in the posterior cardinal vein (Herzog et al., 2001; Moyon et al., 2001; Yuan et al., 2002). We confirm that zebrafish nrp1 and nrp2 are complementarily expressed in or around the corresponding vessels.

Differential expression of the nrp1 and nrp2 genes supports the idea of Nrps playing a role in arterial and venous blood vessel differentiation. In addition, we observed arteriovenous malformations when nrp function was knocked down. For the interpretation of this result, the following explanation is conceivable: Nrps function in a VEGF-dependent way (Lee et al., 2002, this study). Lawson et al. (2002) showed that VEGF is involved in the determination of arterial cell fate by acting upstream of the Notch pathway. Zebrafish mindbomb (mib) mutant embryos, which are defective in Notch signaling, display molecular defects in arteriovenous differentiation and exhibit ectopic connections between the dorsal aorta and the posterior cardinal vein within the trunk (Lawson et al., 2001). Thus, the shunt phenotypes of nrp knockdown larvae is in accordance with the speculation that Nrps might be involved in arteriovenous differentiation as coreceptors for the VEGF ligands.

In addition, a recent study in zebrafish provides in vivo evidence that Semaphorin3a1 affects the migration of angioblasts and the formation of the dorsal aorta (Shoji et al., 2003). Putative migrating angioblasts express nrp1a. Semaphorin3a1 is a possible ligand for Nrp1a. Based on these results, a defect in the dorsal aorta when nrp1a is knocked down was expected (Shoji et al., 2003). Our results together with the short-circuit blood flow of nrp1a morpholino-injected embryos reported by Lee et al. (2002) support a possible role of Nrp1a in proper formation of the dorsal aorta.

Remarkably, shunts were uniformly located within the tail of nrp knockdown larvae, whereas in mib mutant embryos shunts occur within the trunk. This observation could indicate a more sensitive response of the caudal artery and vein to a loss of nrps than the axial vessels within the trunk.

While the distinct expression patterns of the zebrafish nrp genes and the different phenotypes of the knockout of mouse nrp1 and nrp2 (Kawasaki et al., 1999; Yuan et al., 2002) would have suggested diverged functions of Nrps in zebrafish, three of four nrp knockdown larvae displayed rather similar shunt phenotypes. These results indicate that Nrps may have different responsibilities within the same process but that at least in the case of Nrp2a and Nrp2b the functions have diverged. Therefore, if either one of the components of this process is deficient, endothelial cells differentiate in the same improper way, resulting in arteriovenous shunts of the affected larvae.

In the nrp1a, nrp1b, and nrp2a knockdown larvae, the shunts were accompanied by defects in the intersegmental vessels. One could speculate about a role for Nrps in proper pathfinding of these vessels, similar to its function as a receptor for Semaphorins in axon repulsion (He and Tessier-Lavigne, 1997; Kitsukawa et al., 1997). However, many larvae with deficient genes relevant for vessel development display an intersegmental vessel phenotype (Nasevicius et al., 2000; Lawson, 2001; Childs et al., 2002; Habeck et al., 2002), indicating that these vessels are rather sensitive to molecular changes in their environment.

nrp2b is the only nrp gene whose inactivation leads to vascular phenotypes other than a shunt when knocked down by injections of antisense morpholino oligo. The vascular malformation in these larvae specifically affected the caudal artery. It is striking that nrp2b is expressed in cells ventral to the aorta; however, the neighboring artery is disrupted in the knockdown situation. Possibly, Nrp2b receptors on the caudal vein are necessary to guide arterial cell fate in a repulsive manner, such as reported in neuronal tissues. Here, binding of Semaphorin ligand to Nrp receptor proteins led to repulsion of growth cones (He and Tessier-Lavigne, 1997; Kitsukawa et al., 1997). When high amounts of morpholino were injected, nrp2b knockdown larvae displayed a heart phenotype that resulted in a lack of blood circulation. This finding is consistent with its expression in the heart. In contrast to this observation, nrp2-deficient mice were viable to adulthood and the vessel phenotype was rather subtle, effecting primarily the small lymphatic vessels and capillaries (Yuan et al., 2002). Of interest, lymphatic vessels have not been described in zebrafish to date. Taken together, both results, the differential expression and the knockdown phenotypes, are consistent with a model of Nrps being involved in proper embryonic vascular development.

All Four Nrps Interact With VEGF In Vivo

Lack of a single vegf allele results in abnormal vascular development and lethality of mouse embryos (Carmeliet et al., 1996). When zebrafish vegf is knocked down, injected larvae display a variety of phenotypes, ranging from defects in the intersegmental vessels to nearly complete loss of axial and intersegmental vessels (Nasevicius et al., 2000). A dosage sensitivity could also be observed in the nrp/vegf double knockdowns. Coinjections of morpholinos directed against vegf and each of the nrps led to severe vascular defects, whereas the same amounts of morpholino had no significant effect when injected alone (Lee et al., 2002, this study). These results are consistent with in vitro experiments showing that Nrp1 is a coreceptor for VEGFR2, thereby enhancing the binding affinity of VEGF165 ligand (Soker et al., 1998; Lee et al., 2002), while Nrp2 forms complexes with VEGFR1 (Gluzman-Poltorak et al., 2001). In addition, studies in mouse and zebrafish show that VEGF is required for arterial blood vessel development (Lawson et al., 2002; Stalmans et al., 2002). Together with the arteriovenous malformations observed in the nrp knockdown larvae, our data fit the idea of VEGF being an in vivo ligand for Nrps.

In summary, we provide strong evidence that Nrps are required for the correct development of the major axial vessels, the caudal aorta and the caudal vein. The results also demonstrate physiological interdependence of Nrps and VEGF during embryonic vascular development.



Zebrafish stocks were maintained as described previously (Haffter et al., 1996). All experiments were performed by using the Tübingen wild-type strain. Embryos were staged according to Kimmel et al. (1995).

Cloning of Zebrafish neuropilin Sequences

Based on highly conserved amino acid sequences of Nrp1 from human, rat, Xenopus, and chick degenerate primers were designed. The forward primer UM3 (5′-atcatgathaayttyaayccnca-3′) and the reverse primer UM4 (5′-gcraadatdatrtangtrcaytc-3′) were located in the a1 and a2 domains, respectively (corresponding to a 112 amino acid fragment in rat Nrp1 starting from amino acid 67). The forward primer UM5 (5′-atgacnttytggtaycayatg-3′), reverse primer UM7 (5′-acsarytcrarttrtartytc-3′), and nested reverse primer UM6 (5′-gttrtgccarcangcrcarta-3′) were located in the c and cytoplasmic domains, respectively (corresponding to a 164 amino acid fragment in rat Nrp1 starting from amino acid 722).

By using cDNA from a pool of 1- to 3-day-old zebrafish as a template, PCR products of approximately 300 bp and 500 bp in size were amplified and cloned into pBluescript SK+ (Stratagene). Several clones were sequenced. Two different clones for the N-terminal fragment and one clone for the C-terminal fragment were identified.

The 5′- and 3′- ends of the nrp mRNAs were amplified by using primers designed against the sequences of the three clones (Nrp1a-5′1: 5′-ggctctgctgggacatctgggtccttgc-3′; Nrp1a-5′2: 5′-gtccgctgacggtccacagcagcacatc-3′; Nrp1a-3′1: 5′-atggccgactgcaaggacccagatgtcc-3′; Nrp1a-3′2: 5′-cgccatgagtgcgcttggtgttttcctg-3′; Nrp1b-5′1: 5′-cagccgagcggtcggtgaagtttctgg-3′; Nrp1b-5′2: 5′-gcgattttgccgcagaacctcccaagag-3′; first RACE: Nrp1b-3′1: 5′-catggctgatccaggctcctgatccacag-3′; Nrp1b-3′2: 5′-cgcattttgacctggaaagccgggaatg-3′; second RACE: Nrp1b-3′1: 5′-cccgcagggcctagtgaagaaccagtcg-3′; Nrp1b-3′2: 5′-tctaccctgagcgtggctctcccgatgg-3′; Nrp2a-5′1: 5′-gactcgatcactccggttggggaacttg-3′; Nrp2a-5′2: 5′-ggaccggaggagatgatggcgggaggag-3′; Nrp2a-3′1: 5′-gccagcggagccggctacatcacatctc-3′; and Nrp2a-3′2: 5′-ccctctggaatacccgccccatcagagc-3′), the Marathon cDNA amplification kit, and Advantage 2 Polymerase mix (both from Clontech Laboratories). RACE products were cloned into pBluescript SK+ and sequenced. Sequence comparison and BLAST search revealed that the three clones belong to three different nrp genes, named nrp1a, nrp1b, and nrp2a.

A BLAST search of each single exon of human Nrp1 and Nrp2 against the zebrafish genomic sequences (Sanger Institute) identified several exons of an additional nrp gene. The full-length cDNA of putative nrp2b was amplified by using cDNA from 1-day-old embryos and primers that bind to the 5′- and 3′-UTRs (Nrp2b_FLCfor: 5′-cgagaacttcagctttttcctc-3′; Nrp2a_FLCrev: 5′-gcgtgatgtgatgtccattc-3′), cloned into pBluescript SK+ and sequenced. The full-length sequences of nrp1a, nrp1b, nrp2a and nrp2b have been submitted to GenBank (accession nos. AY351910, AY351911, AY351912, and AY351913, respectively).

In Situ Hybridization

Whole-mount in situ hybridizations were performed in principle as described (Schulte-Merker et al., 1994). Antisense RNA probes of neuropilins were generated by in vitro transcription using the digoxigenin RNA labeling mix (Roche). As templates for in vitro transcription, the nrp1a 5′ RACE product, the first nrp1b 3′ RACE product, a 1580-bp nrp2a cDNA fragment (amplified using primers Nrp2a-3′1 and Nrp2a_walk: 5′-gctctggggtgtcgtagttc-3′), and the nrp2b full-length cDNA were used.

For plastic sections, embryos stained by in situ hybridization were embedded in JB4 (Science Services), and 5-μm sections were made. For thick sections, embryos stained by in situ hybridization were transferred into 87% glycerol and cut manually using a scalpel.

Injection of Antisense Morpholino Oligos

Antisense morpholino oligos (AVI Biopharma) were designed complementary to either the start codon or the 5′ UTR of the neuropilin and the vegf gene sequences (nrp1a_ATG: 5′-tgcatcctcgaatcctggaag-3′; nrp1a_5UTR: 5′-cgtatttgggtaattcctctcg-3′; nrp1b_ATG: 5′-ccagtacatcctcaaacgaaatc-3′; nrp1b_5UTR: 5′-cagcatccgatatgaagactcc-3′; nrp2a_ATG: 5′-tatccagaaatccatctttccg-3′; nrp2a_5UTR: 5′-ctccattgccttatcggtctc-3′; nrp2b_ TG: 5′-gcgaataaatccatctttcctg-3′; nrp2b_5UTR: 5′-c tttgtaaatgttgttgtcacccc-3′; vegf: 5′-gtatcaaataaacaaccaagttcatg-3′). Morpholinos were dissolved in injection buffer (0.4 mM MgCl2, 0.6 mM CaCl2, 0.7 mM KCl, 58 mM NaCl, 25 mM Hepes, pH 7.6). A total of 1, 2, and 4 nl was injected at the one- to two-cell stage. Single and double injections were performed at least twice on different injection days.


Larvae were mounted and injected essentially as described (Weinstein et al., 1995).


We thank AVI BioPharma for providing morpholinos; the fish core unit for raising the fish with constant enthusiasm; H. Habeck, T. Koblizek, G. Stott, and A. Vogel for critical reading of the manuscript; and all members of the lab for helpful discussions and technical support throughout the making of this manuscript.