Semaphorin3a1 regulates angioblast migration and vascular development in zebrafish embryos

Organogenesis requires specific signaling between cells during embryogenesis. These signals include extrinsic signals that regulate locomotive behaviors, such as cell migration and neural growth cone extension. One family of molecules believed to act as guidance cues for both cell migration and growth cone extension is the semaphorin family. The semaphorins are a large and diverse gene family that is conserved from invertebrates to humans. These proteins are secreted, GPI-linked or transmembrane; may have an Ig or thrombospondin type 1 domain; and all share a large, conserved sema domain (reviewed by Tessier-Lavigne and Goodman, 1996; Kolodkin,1998; Raper,2000). The first member of this family that was identified to be repulsive for growth cones was chick Collapsin 1 (now called Sema3a), a secreted protein that causes the collapse of specific growth cones(Luo et al., 1993).

The receptor for Sema3a consists of a complex of neuropilin and plexin molecules (Tamagnone et al.,1999; Takahashi et al.,1999). Neuropilins are a small family of conserved proteins whereas plexins are a larger family of conserved proteins. Interestingly,neuropilins are also part of the receptors for a spliced isoform of vascular endothelial growth factor (Vegf), Vegf₁₆₅. Vegf₁₆₅ binds to a complex of neuropilin and Vegfr2 (KDR/flk1) on endothelial and tumor cells and elicits mitogenic and chemotactic responses(Soker et al., 1998). The fact that both Vegf and semaphorins use neuropilins as part of their receptors suggests the possibility of dynamic interactions between the formation of blood vessels and development of the nervous system. Indeed, Semas can inhibit the action of Vegf in vitro and vice versa via competition for binding to neuropilin. Sema3a can inhibit Vegf₁₆₅-mediated aortic endothelial cell migration and capillary angiogenesis, and Vegf₁₆₅ can inhibit Sema3a-induced collapse of dorsal root ganglion growth cones and apoptosis of neural progenitor cells (Miao et al.,1999; Bagnard et al.,2001). However, it is unknown whether interactions between semaphorins and Vegf occur in vivo.

In zebrafish, vascular endothelial precursors (angioblasts) along with hematopoietic progenitors arise initially within the lateral mesoderm at gastrula stages (Gering et al.,1998; Detrich et al.,1995; Brown et al.,2000). Subsequently, these cells converge to the midline where they give rise to axial blood vessels and blood cells(Al-Adhami and Kunz, 1977; Zon 1995; Childs et al., 2002; Zhong et al., 2001). At these stages, Vegf is expressed by the ventromedial region of each somite that the angioblasts migrate on (Liang et al.,1998; Liang et al.,2001) and is required for early vasculature formation(Nasevicius et al., 2000).

Zebrafish contain two copies of the sema3a gene, sema3a1and sema3a2 (Shoji et al.,1998; Roos et al.,1999; Yee et al.,1999). The expression of sema3a1 by the somites guides the growth cones of spinal motor and posterior lateral line neurons(Shoji et al., 1998; Yee et al., 1999; Halloran et al., 2000). sema3a1 is normally expressed by the dorsal and ventral regions of the somites but not the horizontal myoseptal region found in-between these regions. The horizontal myoseptal region is adjacent to the notochord and the dorsal aorta, suggesting that Sema3a1 may act to restrict migrating angioblasts to the vicinity of the notochord. Furthermore, some mutations that affect the notochord or notochord-derived factors lead to both expression of sema3a1 in the entire somite including the horizontal myoseptal region (Shoji et al., 1998)and selectively delete the dorsal aorta(Fouquet et al., 1997; Brown et al., 2000). These correlations suggest that Sema3a1 is involved in dorsal aorta formation in addition to growth cone guidance.

Here we demonstrate that Sema3a1 can regulate vascular development in vivo. A subset of mesodermal cells that are probably angioblasts express neuropilin 1, and these cells migrate dorsally toward the notochord. Antisense knockdown of Sema3a1 inhibits the formation of the dorsal aorta. Furthermore, induced ubiquitous expression of sema3a1 interferes with dorsal migration by angioblasts and adversely affects dorsal aorta formation.

Zebrafish (D. rerio) were maintained in a laboratory breeding colony at 28.5°C on a 14/10-hour light/dark cycle. Embryos collected from breeding fish were allowed to develop at 28.5°C and were staged as described by Kimmel et al. (Kimmel et al.,1995). Embryo age was defined as hours post-fertilization(hpf).

Zebrafish sema3a1 cDNA (Yee et al., 1999) was tagged with Egfp (Clontech) and 6 Myc epitopes(Roth et al., 1991). Egfp was inserted between the 25th and 26th amino acids, between the putative signal peptide and the sema domain, and the Myc epitope was added at the C-terminus. Following transfection of HEK 293 cells with the tagged sema3a1expression construct, the media conditioned by the transfected cells was compared with untagged Sema3a1 for activity using the chick dorsal root ganglion growth cone collapse assay (Luo et al., 1993). Both untagged and tagged Sema3a1 were effective in inducing collapse of DRG growth cones and there was no difference in the potency of the two recombinant proteins. The 1.5 kb zebrafish hsp70promoter (Halloran et al.,2000) was then linked to the tagged sema3a1 and injected into one blastomere of embryos at 1-4 cell stage. The injected embryos were raised to sexual maturity, pair-wise mated with wild-type fish, and their F1 progeny screened with PCR for the transgene using primers for the transgene(tcaagtccgccatgcccgaa/cgtccaygccgagagtgatc) to identify founder fish. F1 embryos were also examined for expression by GFP fluorescence and western blotting (see below) following heat induction. We established two independent lines in which sema3a1 is expressed ubiquitously after heat treatment. Embryos were heat-induced by raising the water temperature from 28.5°C to 38°C over a period of 15 minutes using a programmable water bath (BU150P, Yamato) and then holding the temperature at 38°C for another 30 minutes. Full-length fusion protein was detected by western blotting as early as 15 minutes after heat treatment. In all experiments, heat treatment was started at 15 hpf.

Digoxygenin-labeled riboprobes for sema3a1, sema3a2, neuropilin 1,fli1 and gata1 were synthesized by in vitro transcription and hydrolyzed to an average length of 200-500 base pairs by limited alkaline hydrolysis (Cox et al., 1984). Hybridization on wholemounted embryos was performed according to the protocol of Schulte-Merker et al. (Schulte-Merker et al., 1992). For double in situ hybridization, fli1 and neuropilin 1 riboprobes were labeled with FITC and digoxygenin,respectively. First the red color was developed with AP-conjugated anti-FITC and FAST Red (Sigma), then the green color was developed with HRP-conjugated anti-digoxygenin and the TSA system (Perkin Elmer Life Sciences). Sections were made with cryostat (Cryocut 1800, Leica) or microslicer (DTK-3000W,Dosaka EM) after wholemount hybridization. For cryostat sectioning, embryos were equilibrated in 30% sucrose, embedded in OCT compound (Sakura Finetechnical, Tokyo, Japan) and cut into 20 μm. For microslicer sectioning, embryos were embedded in 30% albumin, 0.5% gelatin, 0.8%glutaraldehyde in PBS and cut into 40 μm.

Whole zebrafish protein (10 μg) was separated with SDS-PAGE gel electrophoresis. The protein was transferred onto PVDF membrane (Millipore)and incubated with 1/100 dilution of anti-Myc (9E10, Roche) in 5% skim milk/PBS followed by an HRP conjugated anti-mouse IgG and ECL immunostain kit(Amersham).

Embryos were anesthetized in 0.01% tricaine (3-aminobenzonic acid ethylester, Sigma) and mounted in 1% agar on a microslide(Shoji et al., 1998). A 0.2%solution of diI(1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate, Molecular Probes) dissolved in dimethylformamide was passed iontophoretically from a micropipette onto approximately 10 mesodermal cells lateral and ventral to somites 10-12 at 18 hpf. Nomarski images and epifluorescence images were captured by 3CCD Video Camera System (DEI-750 Optronics), and combined using Adobe Photoshop. The position of the dorsal-most diI-labeled cells were measured from the dorsal boundary of the yolk tube at 18 hpf and 22 hpf following heat induction at 15 hpf.

Anesthetized embryos were embedded and their vascular system labeled with ink as previously described (Isogai et al., 2001). The sinus venosus was incised for drainage, and then 0.75% Berlin Blue solution was pressure-injected into the dorsal aorta. After perfusion, the embryos were fixed in 4% paraformaldehyde. In yotmutants, dye was injected into the heart cavity because the narrower dorsal aorta was difficult to visualize.

Morpholino oligonucleotides (MOs) were obtained from Gene Tools, LLC. The antisense sema3a1 morpholino sequence (25 mer) was complimentary to a sequence of the 5′UTR (–59 to –34). The control morpholino sequence had 4 bases mismatched compared with the sema3a1 antisense morpholino sequence. Sequences were as follows: sema3a1 antisense MO,5′-CTTGTAGCCCACAGTGCCCAGAGCA-3′; sema3a1 control MO,5′-CTTCTAGCCGACAGAGCCCAGTGCA3′. Morpholino oligonucleotides were solubilized in 1x Danieau Solution (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO₄, 0.6 mM Ca(NO₃)₂, 5 mM HEPES, pH 7.6)and injected into 1 cell-stage embryos. For the knockdown experiments, 0.3 pmol of these MOs were injected into hsp70:^gfpsema3a1^myc embryos.

In order to study the role of Sema3a1 for vascular development, we generated transgenic zebrafish by injecting recently fertilized embryos with plasmid DNA encoding the zebrafish sema3a1 gene(Yee et al., 1999) driven by the zebrafish hsp70 promoter(Shoji et al., 1998; Halloran et al., 2000). The sema3a1 cDNA was tagged with egfp and 6 mycepitopes. Egfp was inserted between the putative signal peptide and conserved sema domain near the amino terminus, and the Myc epitopes were fused to the carboxyl terminus. This fusion protein was expressed by HEK293 cells and found to be secreted and to collapse chick dorsal root ganglion growth cones in vitro (not shown).

Recently fertilized embryos were injected with the hsp70:^gfpsema3a1^myc construct. The injected embryos were raised to maturity and founder fish were identified by pairwise crosses with wild-type zebrafish and PCR of DNA isolated from the F1 embryos using primers for gfp. Of 96 injected fish, 4 transmitted the transgene to F1 offspring. Of the F1 progeny from different founders,2.8-24.8% expressed the transgene as assayed by GFP fluorescence and Myc immunohistochemistry, indicating that the germlines of the founders were mosaic. As expected, each transgenic F1 produced F2 offspring that were 50%transgenic when crossed with wild-type fish. We also generated lines of fish homozygous for the sema3a1 transgene, by incrossing pairs of F1 or F2 hemizygous fish. Mature homozygous offspring were identified based on their ability to generate 100% transgenic embryos when crossed with a wild-type fish. Homozygous transgenic embryos but not wild-type embryos showed strong,widespread induction of the sema3a1 transgene as assayed by GFP fluorescence following exposure to increased temperatures, whereas transgenic embryos were not induced to express the transgene without elevated temperatures (Fig. 1A). Western blots against Myc-tag revealed an approximately 150 kDa protein corresponding to the predicted size of the fusion protein. Expression peaked 1-7 hours following heat induction and began to decrease after 15 hours(Fig. 1B).

There are two copies of the sema3a gene in zebrafish, sema3a1 and sema3a2, that are equally homologous with sema3a in other species (Roos et al., 1999; Yee et al.,1999). The expression patterns of Sema3a1 and Sema3a2 and their receptor, neuropilin 1, were analyzed by wholemount in situ hybridization during early vasculogenesis in the trunk. sema3a2 is expressed transiently in the posterior half of each somite during early stages of somite maturation (Fig. 2A)(Roos et al., 1999). Shortly after the onset of sema3a2 expression, sema3a1 is expressed by the posterior somites (Fig. 2B). Interestingly, the putative angioblasts that give rise to the dorsal aorta are located posterior to the border of each somite (see below),i.e. adjacent to the region of the somites not expressing the semaphorins. Subsequently, sema3a1 expression changes so that it is expressed by dorsal and ventral regions of each somite but not in the horizontal myoseptal region that is in-between the expressing regions and adjacent to the site where the dorsal aorta forms (Fig. 2C,G) (Shoji et al.,1998; Yee et al.,1999). The change in expression of sema3a1 correlates with dorsal migration of vascular endothelial precursors (angioblasts) along the anterior region of each segment (see below) to the eventual site just ventral to the notochord. Furthermore, it suggests that the semaphorins act to restrict the route taken by the angioblasts.

If the hypothesis that Sema3a1 and Sema3a2 regulate migration of angioblasts is correct, then angioblasts should express neuropilin 1, the putative receptor for Sema3a1 and Sema3a2. In fact, neuropilin 1 is expressed by cells that are probably angioblasts(Lee et al., 2002). We confirmed that neuropilin 1 is expressed transiently from 18 to 22 hpf by cells ventral to the somites and by cells that appear to be extending dorsally along the anterior border of each somite(Fig. 2D,E). The apparent dorsal migration by neuropilin 1-positive cells is most apparent in transverse sections (Fig. 2I,J,L). At any one stage of development the neuropilin 1-positive cells are generally found in a more dorsal position in anterior segments compared with those in posterior segments(Fig. 2E). Because the somites form sequentially from anterior to posterior, this suggests that neuropilin 1-positive cells are migrating dorsally. The pattern of neuropilin 1-positive cells is similar to that of migrating angioblasts (Fouquet et al.,1997; Brown et al.,2000), and neuropilin 1-positive cells co-express the endothelial marker fli1 (Fig. 2L-N). These observations suggest that angioblasts express neuropilin 1.

Interestingly, by 21.5-22.5 hpf the dorsal-most neuropilin 1-positive cells have reached a site just ventral to the notochord in segments 10-15, whereas CaP axons labeled with monoclonal antibody Znp-1 are still located at the muscle pioneers (not shown), a site that is dorsal to the notochord. This suggests that the neuropilin 1-positive cells are not following the CaP axons from the ventrolateral somite to the notochord. Furthermore, following ubiquitous induction of sema3a1 in hsp70:^gfpsema3a1^myc transgenic embryos at 15 hpf, there were many fewer neuropilin 1-positive cells along the somites between the notochord and the yolk tube(Fig. 2F,K), suggesting angioblast migration was retarded.

The distribution of neuropilin 1-positive cells along the anterior-posterior axis suggested that the cells expressing neuropilin 1 migrate dorsally. To test this hypothesis, putative neuropilin 1-positive cells were labeled with the fluorescent dye, diI, to see if they migrate. DiI was injected into cells lateral and ventral to the somites. When diI was injected in 18 hpf wild-type embryos (n=5), some of the labeled cells were found in a more dorsal position several hours later,whereas others had not changed their positions suggesting that a subset of cells from ventrolateral mesoderm migrated dorsally(Table 1; Fig. 3A-C).

Following ubiquitous induction of sema3a1, many fewer neuropilin 1-positive cells appeared to migrate towards the notochord(Fig. 2F,K). To verify this,putative angioblasts were labeled with diI and examined following heat-induction of sema3a1. Following heat-induction of sema3a1 in hsp70:^gfpsema3a1^myc1embryos at 15 hpf, none of the labeled cells migrated dorsally to any significant amount (n=5; Table 1; Fig. 3D-F). The dorsal-most labeled cells were 2.4 μm more dorsal at 22 hpf compared with at 18 hpf in transgenics that were heat induced at 15 hpf. In comparison, the dorsal-most labeled cells were 23.2 μm more dorsal at 22 hpf compared with 18 hpf in wild-type embryos. Furthermore, heat treatment of wild-type embryos did not affect dorsal migration of labeled cells (n=3; not shown). These results demonstrate that some but not all cells from the ventro-lateral mesoderm migrate dorsally, and that Sema3a1 can regulate this migration in vivo.

Development of endothelial and hematopoietic lineages are closely related in early embryogenesis (Stainer et al., 1996; Thompson et al., 1998). They express common early markers, emerge simultaneously, and mutations in mice and zebrafish can cause the loss of both lineages(Shalaby et al., 1995; Stainer et al., 1996). Furthermore, in zebrafish both lineages migrate in a similar manner from the lateral mesoderm (Detrich et al., 1995; Gering et al.,1998; Zhong et al.,2001). In order to see which population is regulated by Sema3a1,we examined fli1, which is expressed by differentiating endothelial cells as well as by immature endothelial and hematopoietic progenitors(Brown et al., 2000), and gata1, which is selectively expressed by hematopoietic progenitors following misexpression of Sema3a1(Detrich et al., 1995).

Fli1 is an ETS domain transcription factor and is expressed by early vascular and hematopoietic precursor cells(Brown et al., 2000). At 20 hpf, fli1 is expressed diffusely by many cells found between the notochord and the yolk tube (n=4; Fig. 4A), but at 25 hpf expression is clearly seen in presumptive dorsal aorta cells just ventral to the notochord (n=5; Fig. 4B) (Brown et al.,2000). In the notochord mutant you-too (yot)that fails to develop the dorsal aorta(Fouquet et al., 1997; Chen et al., 1996), and that expresses sema3a1 in the entire somite(Fig. 2H)(Shoji et al., 1998), fli1 expression is restricted abnormally in patches rather than in cells just ventral to the notochord at 25 hpf (n=5; Fig. 4C). Thus aberrant expression of sema3a1 throughout the entire somite correlates with the abnormal distribution of fli1 cells. Finally, following induction of sema3a1 in hsp70:^gfpsema3a1^mycembryos at 15 hpf, the distribution of fli1-positive cells was abnormal with cells diffusely distributed between the notochord and yolk tube at 25 hpf (n=5; Fig. 4D), suggesting that ubiquitous misexpression of Sema3a1 partially inhibits dorsal migration of angioblasts. This result is consistent with our finding that misexpression of Sema3a1 interferes with dorsal migration by ventro-lateral mesodermal cells, and suggests that these cells are fli1-positive angioblasts that give rise to the dorsal aorta.

GATA1 is a transcription factor that regulates hematopoietic development(Pevny et al., 1991; Orkin, 1992), and is restricted to blood progenitors (Detrich et al., 1995). In contrast to fli1, the expression pattern of gata1 was not perturbed following misexpression of sema3a1 in transgenic embryos (n=5), nor in yotmutants (n=4; Fig. 4F-H). Thus, Sema3a1 appears to regulate migration by fli1-positive angioblasts but not migration by gata1-positive hematopoietic progenitor cells.

The previous sections showed that induced ubiquitous expression of sema3a1 interfered with dorsal migration of putative angioblasts from the ventrolateral mesoderm. This finding predicts that overexpression of sema3a1 should lead to defects in the dorsal aorta and thus blood circulation. Indeed, blood circulation was abnormal at 30 hpf in transgenic embryos (n=20) following induction of sema3a1 at 15 hpf(Fig. 5). Initially, the heart was seen to be beating in many transgenic embryos, but by 27 hpf the blood cells were restricted to the heart and yolk sac and tube and did not circulate into other parts of the embryo. Subsequently the heart cavity became swollen and eventually stopped beating (29-30 hpf). The dorsal aorta was present in hsp70:^gfpsema3a1^myc embryos (30 hpf) following induced ubiquitous expression, but the lumen of the dorsal aorta contained no blood cells and was constricted when visualized by dye injection(Fig. 5D,E). The axial vein and other cranial vessels at this stage appear relatively unaffected. Thus, it appears that constriction of the dorsal aorta is responsible for the lack of circulation in these embryos. Corroborating these findings, in yotmutants high pressures that actually ruptured the axial vein were required to fill the dorsal aorta, and the lumen of the dorsal aorta was constricted at various points (Fig. 5F). These results suggest that Sema3a1 in the somites acts to regulate angioblast migration and, thereby, the formation of the dorsal aorta.

To see whether Sema3a1 is required for normal dorsal aorta formation, we injected antisense MOs (Nasevicius et al.,2000) against sema3a1 into recently fertilized embryos. The efficacy of the sema3a1 antisense MO to effectively knockdown translation of Sema3a1 was determined by injecting the antisense or control MO into recently fertilized hsp70:^gfpsema3a1^mycembryos, heat inducing them at 18 hpf, and assaying for induction of ^GFPSema3a1^Myc by fluorescence at 25 hpf. GFP fluorescence was significantly reduced in the sema3a1 antisense MO-injected transgenic embryos (n=22), whereas sema3a1control MO-injected transgenic embryos (n=19) showed GFP fluorescence comparable to uninjected transgenic embryos(Fig. 6A).

Knocking down Sema3a1 appeared to interfere with normal migration of angioblasts and disrupted formation of the dorsal aorta and normal circulation(Fig. 6B-E). In control MO-injected wild-type embryos, neuropilin 1-positive cells were seen to be extending dorsally along the anterior border of the somites from the yolk tube to the notochord at 22 hpf as in uninjected wild-type embryos (see Fig. 2E) and circulation was normal at 27-28 hfp (not shown). In contrast, in antisense MO-injected wild-type embryos there were fewer neuropilin 1-positive cells in the region between the yolk tube and the notochord (n=20; Fig. 6B, compare with Fig. 2E; Fig. 6C, compare with Fig. 2J), suggesting that their migration was inhibited. Furthermore, high pressures were required to inject Berlin blue ink into the vasculature at 30 hpf. In these embryos, the cardinal vein but not the dorsal aorta (Fig. 6D) or the cardinal vein and a constricted dorsal aorta(Fig. 6E) were labeled. Furthermore, the heart was swollen and no circulation was observable (not shown). These results indicate that Sema3a1 expressed by the somites is necessary for normal development of the dorsal aorta.

Based upon previous work, we hypothesized that Sema3a regulates migration by angioblasts that give rise to the dorsal aorta. To test this hypothesis, we first showed that cells that are probably migrating angioblasts express neuropilin 1, a receptor for Sema3a. Second, the neuropilin 1-positive cells appear to migrate to regions that do not express Sema3a1 nor Sema3a2, suggesting that they avoid the Sema3a-expressing regions. Third,both misexpression and knockdown of Sema3a1 interfered with migration by putative angioblasts, and adversely affected development of the dorsal aorta and blood circulation. Both ubiquitous misexpression and knockdown of Sema3a1 would be expected to disrupt the normal dorsal/ventral gradient of Sema3a1,with overexpression masking the gradient and knockdown eliminating or dramatically decreasing the gradient. These results suggest that a gradient of Sema3a1 expressed by the somites does regulate migration by angioblasts that give rise to the dorsal aorta.

Recently, MO knockdown of neuropilin 1 was shown to lead to defective circulation in intersegmental vessels but not axial vessels, and MO knockdown of Vegf to defective circulation in both intersegmental and axial vessels(Lee et al., 2002). Because neuropilin 1 serves as a coreceptor for both Sema3a and Vegf, one would have expected a knockdown of neuropilin 1 to interfere with dorsal aorta formation. It is possible that defects of the dorsal aorta require higher levels of neuropilin 1 knockdown compared to defects of the intersegmental vessels. In fact, neuropilin 1 does appear to contribute to formation of the dorsal aorta because injection of neuropilin 1 antisense MO with a non-effective level of Vegf antisense MO did interfere with both intersegmental and axial circulation(Lee et al., 2002).

How does Sema3a1 regulate angioblast migration? One possibility is that migration by angioblasts is directed by CaP motor axons whose pathway partially coincides with that of the angioblasts. Consistent with this idea is that CaP growth cones are repulsed by Sema3a1(Halloran et al., 2000). Thus,it is possible that the deleterious effects on angioblast migration following manipulation of Sema3a1 are secondary effects mediated via erroneous outgrowth by the CaP motor axons. However, our evidence suggests that angioblasts do not appear to follow the CaP axons. Angioblasts start ventrolateral to the somites and first migrate medially and then dorsally to the notochord. The CaP motor axon starts out in the spinal cord and extends ventrally then laterally to the ventrolateral edge of the somite (Myers et al., 1986). In midtrunk segments angioblasts are arriving at their destination just ventral to the notochord whereas the CaP motor axons are still at the muscle pioneers, a site that is dorsal to the destination of the angioblasts. Thus, many angioblasts have completed their migration prior to extension of the CaP motor axon beyond the notochord, suggesting that angioblasts need not follow CaP axons to complete their migration.

A second possibility is that Sema3a1 may directly repulse angioblasts. First, class 3 semaphorins are secreted molecules that repulse specific growth cones (Raper, 2000). Second,class 3 semaphorins regulate migration by neural crest cells(Eickholt et al., 1999) and neurons (Marin et al., 2001). Third, Dev cells, a human medulloblastoma cell line, avoid migrating on a Sema3a substrate in vitro (Bagnard et al.,2001).

A third possibility is that because class 3 semaphorins and Vegf both use neuropilins as a component of their functional receptor(Takahashi et al., 1999; Soker et al., 1998), Sema3a1 may regulate migration of angioblasts by interfering with the chemoattractant activity of Vegf for these cells. Because semaphorins can inhibit the action of Vegf in vitro and vice versa via competition for binding to neuropilin(Miao et al., 1999; Bagnard et al., 2001), the effects on migration of putative vascular endothelial cells following misexpression of Sema3a1 could be accounted for by this mechanism. In fact,migration of angioblasts that form the dorsal aorta in Xenopus is guided by Vegf expressed by the hypochord cells ventral to the notochord(Cleaver and Krieg, 1998). In zebrafish, Vegf is expressed by the ventromedial region of each somite(Liang et al., 1998),neuropilin 1 and Flk1/Vegfr2 is expressed by putative angioblasts (this study)(Fouquet et al., 1997), and knocking down Vegf leads to defective vascular development including that of the dorsal aorta (Nasevicius et al.,2000). Interestingly, the expression domains of Vegf and Sema3a1 within the somites appear to partially overlap(Fig. 7). Thus it is possible that Sema3a1 secreted by the ventral third of the somite may create a functional gradient of secreted Vegf activity such that the level of Vegf is high dorsal and low ventral in the ventral half of the somite. Such a functional Vegf gradient would be consistent with the dorsal migration of the angioblasts. Furthermore, it is possible that competitive interactions between Vegf and Sema3a1 may also generate a functional gradient of Sema3a1 within the ventral third of the somites. Here the repulsive activity of Sema3a1 would be high ventral and low dorsal in the ventral half of the somites. Thus, there exists the intriguing possibility that complementary gradients of attractive and repulsive molecules may regulate dorsal migration by angioblasts.

The manipulations of Sema3a1 expression are consistent with the possibility that Sema3a1 may repulse angioblasts and/or regulate attraction by Vegf. In the case of repulsion, overexpression of Sema3a1 would be expected to mask directed repulsion by a gradient of endogenous Sema3a1 and thereby interfere with angioblast migration. Similarly, knocking down Sema3a1 would significantly decrease or eliminate the Sema3a1 gradient and disrupt directed angioblast migration. In fact, a gradient of Sema3a is hypothesized to direct cortical axons via a repulsive activity and both ubiquitous application of exogenous Sema3a and elimination of Sema3a lead to similar and aberrant outgrowth by cortical axons (Polleux et al., 1998). This demonstrated that cortical axons were not inhibited from extending despite encountering a uniform distribution of a repulsive molecule. Similarly, temporal retinal axons will avoid posterior tectal membranes that they find repulsive, but nevertheless are able to extend on posterior tectal membranes in the absence of a better substrate(Walter et al., 1990). Disruption of angioblast migration would also be expected if Sema3a1 acted as a competitive inhibitor for Vegf. Ubiquitous misexpression of Sema3a1 would inhibit Vegf action by binding to neuropilin 1 and thus interfere with attraction by dorsal sources of Vegf. Knockdown of Sema3a1 might lead to a shallower functional gradient of Vegf and thus impede angioblast migration.

Whether migration of angioblasts is regulated by complementary gradients of Vegf and Sema3a1 is an open question. What is clear is that both Vegf and Sema3a1 can regulate the formation of the dorsal aorta. That they do so in concert in vivo is suggested by the correlation of dorsal aorta defects and concomitant changes in expression of Vegf and Sema3a1 in the zebrafish floating head (flh) mutant. Vegf expression is missing in somites and Sema3a1 is expressed throughout the entire somite in flhembryos (Shoji et al., 1998; Liang et al., 2001). The defects of the dorsal aorta are more severe in flh(Brown et al., 2000; Liang et al., 2001) compared to that seen following antisense knockdown of Vegf (Nascevicius et al., 2000),knockdown of Sema3a1 or misexpression of Sema3a1 throughout the entire somite(this study). Furthermore, the fact that the notochord is missing in flh suggests that notochord signaling is critical for guiding angioblasts that form the dorsal aorta(Fouquet et al., 1997;Weinstein et al., 1999). Sonic hedgehog from the notochord appears to play a major role in this concerted process, because it downregulates Sema3a1(Shoji et al., 1998) and upregulates Vegf (Lawson et al.,2002) on the somites. A similar dorsal aorta phenotype is found in other midline mutants, which appear to lack hedgehog signaling(Chen et al., 1996; Brown et al., 2000) (W.S.,unpublished).

Our evidence demonstrates that in zebrafish Sema3a1 can affect migration by putative angioblasts and the formation of the dorsal aorta. Intriguingly,Sema3a is expressed in early somites of mammals in a pattern similar to that seen in zebrafish (Giger et al.,1996; Taniguchi et al.,1997). Endothelial precursors in avian embryos migrate from the somitic mesoderm and splanchnopleural mesoderm to form the dorsal aorta(Pardanaud et al., 1996). Furthermore, Sema3a knockout mice and neuropilin 1 knockout mice exhibit cardiovascular defects (Behar et al.,1996; Kawasaki et al.,1999). This suggests that Sema3a may also regulate migration of angioblasts in other vertebrates as well.

Other class 3 semaphorins also regulate cardiovascular development. sema3C is expressed along the path followed by migrating cardiac neural crest cells to the truncus arteriosus and the aortic arch(Feiner et al., 2001). In Sema3C knockout mice these vascular elements are defective suggesting that Sema3C is an attractive cue for neural crest migration(Feiner et al., 2001). Thus,Sema3a and Sema3C appear to regulate different processes in cardiovascular development.

Ephrins and their Eph receptors are also involved in patterning of the vascular system. Ephrin B2 (Efnb2) is expressed by arterial but not venous endothelial cells, whereas Ephb4 the proposed Efnb2 receptor is expressed at higher levels by venous endothelial cells(Wang et al., 1998; Adams et al., 1999; Gerety et al., 1999). Furthermore, mice lacking Efnb2 or Efnb4 exhibit aberrant vascular patterning(Wang et al., 1998; Gerety et al., 1999). Because ephrins and their receptors are well-known regulators of growth cone guidance and cell migration, it is possible that they may also regulate migration by angioblasts during vascular development. In the zebrafish efnb2 is expressed by somites (Durbin et al.,1998) in a pattern that is similar to that of sema3a1. Early the posterior half of each somite expresses efnb2, but later expression changes so that the dorsal and ventrolateral somite expresses efnb2. Thus the expression pattern of efnb2 is appropriate for the guidance of angioblast migration as well. efnb2 is also expressed by dorsal arterial angioblasts in zebrafish, but only after migration (Lawson et al.,2001). Similarly, Ephb4 is expressed by the venous angioblasts in the trunk after cell migration but is only diffusely expressed prior to vessel formation (Zhong et al.,2001). This later expression suggests that Efnb2/Ephb4 may be involved in the patterning and separation of dorsal arterial and posterior cardinal venous angioblasts in their final positions ventral to the notochord.

In conclusion, our experiments demonstrate that Sema3a1, a factor that guides growth cone extension, can also regulate migration of putative angioblasts. Interestingly, Vegf, which is an important regulator of migration by angioblasts, also affects growth cones (W.S., unpublished). Thus, factors that regulate migration/guidance in the nervous system also regulate migration in the vascular system and vice versa.

We thank Drs N. Yanai, S. Ikawa, T. Watanabe, S. Lyons and M. Halloran for helpful discussions and comments, Drs M. Kobayashi and M. Yamamoto for providing fli1 and gata1, and L. Barthel and P. Raymond for their in situ hybridization protocol. The research was supported by NINDS(NS36587) to J.Y.K., a Grant-in Aid for Scientific Research on Priority Areas(C) – Advanced Brain Science Project – and (A) – Research for Comprehensive Promotion of Study of Brain – from Ministry of Education, Culture and Technology (Japan) to W.S.