GIPC proteins negatively modulate Plexind1 signaling during vascular development

Semaphorins (SEMAs) and their Plexin (PLXN) receptors are central regulators of metazoan cellular communication. SEMA-PLXND1 signaling plays important roles in cardiovascular, nervous, and immune system development, and cancer biology. However, little is known about the molecular mechanisms that modulate SEMA-PLXND1 signaling. As PLXND1 associates with GIPC family endocytic adaptors, we evaluated the requirement for the molecular determinants of their association and PLXND1’s vascular role. Zebrafish that endogenously express a Plxnd1 receptor with a predicted impairment in GIPC binding exhibit low penetrance angiogenesis deficits and antiangiogenic drug hypersensitivity. Moreover, gipc mutant fish show angiogenic impairments that are ameliorated by reducing Plxnd1 signaling. Finally, GIPC depletion potentiates SEMA-PLXND1 signaling in cultured endothelial cells. These findings expand the vascular roles of GIPCs beyond those of the Vascular Endothelial Growth Factor (VEGF)-dependent, proangiogenic GIPC1-Neuropilin 1 complex, recasting GIPCs as negative modulators of antiangiogenic PLXND1 signaling and suggest that PLXND1 trafficking shapes vascular development.


Introduction
Angiogenic sprouting, the formation of new vessels via branching of pre-existing ones, drives most of the life-sustaining expansion of the vascular tree. Sprouting angiogenesis is also pivotal for recovery from injury and organ regeneration and is often dysregulated in disease (Carmeliet, 2003;Ramasamy et al., 2015;Cao, 2013;Fischer et al., 2006). Among the pathways that modulate this process, SEMA-PLXN signaling plays a prominent role. In particular, the vertebrate-specific PLXND1 receptor transmits paracrine SEMA signals from the somites to the endothelium to shape the stereotypical and evolutionarily conserved anatomy of sprouts in the trunk's arterial tree (Torres-Vázquez et al., 2004;Zygmunt et al., 2011;Gitler et al., 2004;Zhang et al., 2009;Gu et al., 2005;Childs et al., 2002;Epstein et al., 2015). For instance, we have shown that antiangiogenic Plxnd1 signaling antagonizes the proangiogenic activity of the VEGF pathway to regulate fundamental features of vascular development (Torres-Vázquez et al., 2004;Zygmunt et al., 2011; see also Moriya et al., 2010;Fukushima et al., 2011;Kim et al., 2011). Specifically, Plxnd1 signaling acts cell autonomously in the endothelium to spatially restrict the aorta's angiogenic capacity to form sprouts, thereby determining both the positioning and abundance of the aortic Segmental (Se) vessels. Plxnd1 signaling also guides Se pathfinding, thus shaping these vascular sprouts.
To address this gap, we focused on elucidating the relationship between GIPCs and antiangiogenic PLXND1 signaling. Based on our identification of GIPC1/Synectin as a PLXND1-binding protein, we previously showed by X-ray crystallography and co-immunoprecipitation (CoIP) assays that all three GIPC proteins can physically interact with the intracellular tail of the PLXND1 receptor and elucidated the structural organization of the GIPC-PLXND1 complex Shang et al., 2017).
Here, we have further evaluated the molecular determinants of the PLXND1-GIPC interaction in the native environment of mammalian cells, firmly establishing a critical role for the last few amino acids of PLXND1 for the formation of the PLXND1-GIPC complex. Moreover, we defined the vascular role of PLXND1-GIPC association using zebrafish and cultured human umbilical vein endothelial cells (HUVECs) as model systems. Our results suggest that GIPCs act redundantly to limit the antiangiogenic signaling of the PLXND1 receptor.
Hence, our identification of GIPCs as negative intracellular regulators of PLXND1 signaling outlines a novel molecular mechanism by which these endocytic adaptors promote angiogenic development of the vertebrate arterial tree (Chittenden et al., 2006;Lanahan et al., 2010;Fantin et al., 2011;Lampropoulou and Ruhrberg, 2014;Herzog et al., 2011;Plein et al., 2014), highlighting the potential importance of PLXND1 trafficking for proper vascular development.
To understand how PLXND1 and GIPC1 interact physically we performed a crystallographic study with bacterially purified recombinant mouse proteins. This structural analysis revealed that the receptor's nonameric C-terminal sequence NIYECYSEA-COOH (residues 1917-1925 is the primary GIPC-Binding Motif (GBM) and makes N-and C-terminal contacts with GIPC1's GH1 and PDZ domains. A pair of receptor's helices (residues 1893-1908) and the GIPC1's PDZ domain form a secondary heteromeric interface. Pulldown experiments of GIPC1 by MYO6 in the presence of PLXND1 indicate that MYO6 associates with PLXND1-bound GIPC1 and that deletion of the receptor's canonical PBM (PLXND1DSEA) or I1918A-Y1919A GBM substitutions significantly reduces PLXND1-GIPC1 binding. Conversely, the GIPC1 G114Y mutation disrupting the interaction between PLXND1's I1918/Y1919 GBM residues and GIPC1's GH1 hydrophobic pocket also impairs PLXND1-GIPC1 binding (Shang et al., 2017).
Using exogenous expression of murine proteins and a eukaryotic cellular environment, we interrogated the requirement for PLXND1's canonical PBM and nonameric GBM and also tested the involvement of GIPC1's PDZ and GH1 domains. To this end, we performed CoIP experiments with COS7 cells, which lack endogenous expression of both PLXND1 (Gu et al., 2005;Uesugi et al., 2009;Sakurai et al., 2010;Takahashi et al., 1999) and NRP1 (Tordjman et al., 2002) (Figure 1; see also Supplementary file 1 and Supplementary file 2).
To address the requirement for PLXND1's canonical PBM and nonameric GBM, we used V5tagged PLXND1 cytosolic tails ( Figure 1A), either with (V5-C-mPLXND1 WT ) or without (V5-C-mPLXND1D CYSEA ) the canonical PBM. For GIPC1, we used a FLAG-tagged version of the full protein, FLAG-mGIPC1 WT ( Figure 1B). We found that deletion of PLXND1's canonical PBM (V5-C-mPLXND1D CYSEA ) significantly reduced binding to FLAG-mGIPC1 WT ( Figure 1C, lanes 2-3 and graph below), consistent with the existence of additional contacts, as suggested by our crystallographic findings (Shang et al., 2017; see also Lee and Zheng, 2010). Accordingly, the PLXND1 form lacking the GBM motif (V5-C-mPLXND1D GBM ) shows a further reduction in FLAG-mGIPC1 WT -binding capacity ( Figure 1C, lane 4 and graph below). Importantly, the minimum GIPC1-binding capacity of the V5-C-mPLXND1D GBM form observed under these protein over-expression conditions suggests that the secondary GIPC-binding interface formed by the C-terminal pair of helices of PLXND1 (Shang et al., 2017) plays a minor role as a molecular determinant for PLXND1-GIPC1 complex formation.
To test the involvement of the two GIPC1 domains that interact with PLXND1 in the crystal structure (Shang et al., 2017), we used FLAG-tagged GIPC1 fragments harboring the GH1 and PDZ domains individually (FLAG-mGIPC1 GH1 and FLAG-mGIPC1 PDZ , respectively) and V5-C-mPLXND1 WT . Our CoIP data confirmed that each of these domains is sufficient for PLXND1 association ( Figure 1D). The ability of GIPC1's GH1 domain to interact with PLXND1 is consistent with prior observations implicating GIPC1's N-terminal region in binding to other transmembrane proteins (Naccache et al., 2006;Giese et al., 2012).
Taken together, the data from our yeast two-hybrid  and crystallographic (Shang et al., 2017) studies demonstrate that PLXND1 and GIPC1 can interact directly and that NRP1 is dispensable for their interaction. These findings are consistent with our CoIP of PLXND1-GIPC1 complexes from COS7 cells, which lack endogenous NRP1 expression (see Tordjman et al., 2002). Together with our dissection of the molecular determinants of PLXND1-GIPC1 complex formation, the high sequence identity of GIPCs, and our observation that GIPC2 and GIPC3 also associate with PLXND1 (Shang et al., 2017), these results establish the critical importance of PLXND1's GBM and GIPC's GH1 and PDZ domains for PLXND1-GIPC complex formation. Diagrams of the wild-type (WT) and truncated V5-tagged (red) forms of the cytosolic tail of murine PLXND1 (V5-C-mPLXND1) used for coimmunoprecipitation experiments. Color-coding is used to highlight the following domains and motifs. GAP1 and GAP2 (Guanosine triphosphatase-Activating Protein domains 1 and 2; black), RBD (Rho GTPase-Binding Domain; green), T-segment (C-terminal segment, includes the GBM; blue) and, GBM (GIPC-Binding Motif; magenta); see . (B) Diagrams of the wild-type (WT) and truncated FLAG-tagged (purple) forms of murine GIPC1 (FLAG-mGIPC) used for co-immunoprecipitation experiments. Domains indicated as follows: PDZ (PSD-95/Dlg/ZO-1; green) and GH (GIPC Homology domain) 1 (blue) and 2 (orange); see (Katoh, 2013). (C) Western blots (top) and their quantification (bottom, bar graphs). Numbers indicate lane positions. Left-side Western blot (IP FLAG ): FLAG immunoprecipitates and V5 co-immunoprecipitates showing interactions between the indicated V5-C-mPLXND1 and FLAG-mGIPC forms. Right-side Western blot (TCL), expression levels of these proteins in total cell lysates as detected with V5 and FLAG antibodies. Quantifications. n = 3 independent experiments for each protein pair. Left-side bar graph (C). Means of percentual V5/FLAG relative binding [(V5 CoIP /V5 TCL )/(FLAG IP /FLAG TCL )] between the indicated protein pairs from the IP FLAG Western blot (top left) and the relative abundance of the expression levels of these proteins from the TCL Western blot (top right). Error bars, ± SEM. V5/FLAG relative binding was significantly different (p<0.05) between V5-C-mPLXND1 forms and FLAG-mGIPC, F(2, 6)=22.376, p=0.002, as determined by a one-way ANOVA test. A Tukey post hoc analysis was conducted to determine whether the percentual V5/FLAG relative binding between the three tested pairs of proteins was significantly different (p<0.05; asterisks). Right-side bar graph (C). Means of percentual V5 TCL /FLAG TCL relative abundance between the indicated protein pairs from the TCL Western blot (top right). Error bars, ± SEM. A Kruskal-Wallis H test was conducted to determine significant differences (p<0.05) in V5 TCL / FLAG TCL relative abundance between the indicated protein pairs. Distributions of V5 TCL /FLAG TCL relative abundance were not similar for all groups. The medians of V5 TCL /FLAG TCL relative abundances were 92.64 (for V5-C-mPLXND1 WT /FLAG-mGIPC1), 87.79 (for V5-C-mPLXND1D CYSEA /FLAG-mGIPC1), and 96.22 (for V5-C-mPLXND1D GBM /FLAG-mGIPC1), but were not statistically significantly different between them (Ramasamy et al., 2015), Zebrafish plxnd1 skt6 homozygous mutants, which express a Plxnd1 receptor with a predicted impairment in GIPC binding, display angiogenesis deficits with low frequency To determine the role that GIPC binding exerts on antiangiogenic PLXND1 signaling, we sought to specifically impair PLXND1's ability to associate with GIPC endocytic adaptors in an in vivo model of vascular development. To do this, we performed CRISPR/Cas9-based genome editing Chang et al., 2013;Cong et al., 2013;Cong and Zhang, 2015;Gagnon et al., 2014;Hill et al., 2014;Hruscha et al., 2013;Hwang et al., 2013;Irion et al., 2014;Kimura et al., 2014;Mali et al., 2013;Talbot and Amacher, 2014) of the last coding exon of the zebrafish plxnd1 locus to introduce disrupting mutations into the receptor's GBM (NIYECSSEA-COOH, canonical PBM underlined; Figure 2A). The resulting plxnd1 skt6 allele encodes a Plxnd1 receptor missing the PBM because of replacement of the five C-terminal residues by a stretch of 31 amino acids ( Figure 2B; see also Supplementary file 1 and Supplementary file 8). Because adding just a single C-terminal residue to the PBM of proteins that interact with PDZ domain-containing partners is sufficient to block their cognate association (Rickhag et al., 2013;Saras et al., 1997;Cao et al., 1999;Garbett and Bretscher, 2012), and deletion of PLXND1's PBM reduces GIPC binding significantly ( Figure 1A-C; see also Shang et al., 2017), the plxnd1 skt6 mutant allele is expected to encode a Plxnd1 receptor with reduced or null GIPC-binding ability.
In the trunk of WT embryos, Se sprouts arise bilaterally from the Dorsal Aorta (DA) just anterior to each somite boundary (SB) at 21 h post-fertilization (hpf). Se sprouts grow dorsally with a chevron shape connecting with their ipsilateral neighbors above the spinal cord's roof to form the paired Dorsal Longitudinal Anastomotic Vessels (DLAVs) by 32 hpf (Isogai et al., 2003) ( Figure 2C). In contrast, null plxnd1 alleles, such as plxnd1 fov01b , show a recessive and hyperangiogenic mutant phenotype characterized by a dramatic mispatterning of the trunk's arterial tree anatomy. As such, the homozygous mutants form too many, and misplaced, Se sprouts that branch excessively and interconnect ectopically ( Figure 2D) as a result of increased proangiogenic VEGF activity (Torres-Vázquez et al., 2004;Zygmunt et al., 2011).
A comparative analysis of WT embryos, plxnd1 fov01b homozygous mutants, and plxnd1 skt6 / plxnd1 fov01b transheterozygotes ( Figure 2C-E) revealed that, like the WT, the transheterozygotes display a properly patterned arterial tree and normal Se and DLAV angiogenic development ( Figure 2C,E; see also Supplementary file 3 and Materials and methods). These observations argue against the model that GIPCs directly enable PLXND1's antiangiogenic signaling (see Burk et al., Figure 2. The plxnd1 skt6 allele encodes a functional Plxnd1 receptor putatively impaired in GIPC binding, and its homozygosity induces angiogenesis deficits with low frequency. (A, B) Diagrams of the cytosolic tails of the zebrafish Plxnd1 proteins encoded by the WT (A) and plxnd1 skt6 mutant (B) alleles including their C-terminal amino acid sequences. Color-coding is used to highlight the following domains and motifs. GAP1 and GAP2 (Guanosine triphosphatase-Activating Protein domains 1 (left) and 2 (right); black), RBD (Rho GTPase-Binding Domain; green), T-segment (C-terminal segment, includes the GBM; blue), and GBM (GIPC-Binding Motif; magenta). In the WT protein diagram (A), the canonical PBM (PDZ-Binding Motif) is underlined. In the mutant protein diagram (B), the thin horizontal red bar denotes the amino acid sequence replacing the PBM. (C-H) Confocal lateral images of the trunk vasculature (green) of 32 hpf embryos (region dorsal to the yolk extension). Anterior, left; dorsal, up. Scale bars (white horizontal lines), 100 mm. Genotypes indicated on top of each image in yellow font. Angiogenesis deficits are indicated with asterisks as follows: white (DLAV gaps), magenta (truncated Se). In the WT image (C), the vessels are designated with the white font as follows: DLAV (Dorsal Longitudinal Anastomotic Vessel), Se (Segmental Vessel), DA (Dorsal Aorta), and PCV (Posterior Cardinal Vein). The homozygous WT and homozygous plxnd1 skt6 mutant embryos (F-H) are siblings derived from the incross of plxnd1 skt6 /+heterozygotes. (I) Bar graph. Percentage of Se-DLAV in 32 hpf embryos of the indicated genotypes belonging to each of the following four phenotypic classes. Truncated: maximal (red; includes missing Se), moderate (yellow), and minimal (gray). Non-truncated: Full (black). There was no statistically significant difference in the distribution of the four phenotypic classes between WT and plxnd1 skt6 mutants as assessed by a two-sided Fisher Exact test, p=0.05905. Quantifications. To determine whether plxnd1 skt6 complements the plxnd1 fov01b null, we analyzed vascular patterning (C-E) and scored Se-DLAV angiogenesis (C, E) in embryos of the following three genotypes: WT (124 Se-DLAV, 11 embryos, an average of 11.27 Se-DLAV/embryo), plxnd1 fov01b homozygotes (12 embryos), and plxnd1 fov01b /plxnd1 fov01b transheterozygotes (162 Se-DLAV, 16 embryos, an average of 10.13 Se-DLAV/embryo). All the WT and transheterozygotes displayed proper vascular patterns indistinguishable from each other and lacked Se-DLAV truncations. All the plxnd1 fov01b mutants displayed hyperangiogenic vascular mispatterning. To determine how Plxnd1's inability to interact with GIPCs impacts angiogenic growth, we scored Se-DLAV angiogenesis (F-I) in sibling embryos of the Figure 2 continued on next page 2017). We conclude that the novel plxnd1 skt6 allele complements the plxnd1 fov01b null and encodes a functional receptor capable of antiangiogenic signaling.
Further support for the notion that reducing Plxnd1's ability to associate with GIPCs does not inactivate the receptor comes from experiments driving mosaic transgenic endothelial expression of N-terminally tagged forms of zebrafish Plxnd1 in plxnd1 fov01b null mutants (Figure 2-figure supplement 2). Briefly, we found that the WT form (2xHA-Plxnd1 WT ) and a receptor lacking the entire GBM because of removal of the nine C-terminal residues (2xHA-Plxnd1D GBM ) had the same capacity to cell-autonomously rescue the vascular defects of plxnd1 fov01b homozygotes (Figure 2-figure supplement 2D-E,G-H,J). In other words, endothelial cells within Se exogenously expressing either receptor form were absent from ectopic Se, had WT shapes concordant with their position within Se and DLAVs, and, when found within the base of sprouts, were positioned correctly just anterior to somite boundaries. These are the same features that WT endothelial cells display when transplanted into plxnd1 fov01b hosts; see .
This experiment does not clarify whether impairing Plxnd1's GIPC-binding ability either increases or does not affect Plxnd1's signaling. To address this point, we compared homozygous WT and homozygous plxnd1 skt6 mutant sibling embryos derived from incrosses of plxnd1 skt6 /+heterozygotes. We found that plxnd1 skt6 homozygous mutants showed an adequately organized vascular tree and a low frequency of Se and DLAV truncations ( Figure 2G-I; see also the Materials and methods section 'Quantification of angiogenesis deficits in the trunk's arterial tree of WT and mutant zebrafish embryos'). However, the distributions of complete and truncated vessels and both the penetrance and expressivity of these defects were not statistically significantly different between WT and plxnd1 skt6 mutants ( Figure 2I Homozygous plxnd1 skt6 mutants are hypersensitive to the antiangiogenic drug SU5416 The results above show that the receptor encoded by the plxnd1 skt6 allele is active, but do not fully clarify the nature of the plxnd1 skt6 allele and the precise role of GIPCs in antiangiogenic PLXND1 signaling. Hence, we compared the sensitivity of homozygous WT and homozygous plxnd1 skt6 mutants to the antiangiogenic compound SU5416, a tyrosine kinase inhibitor that preferentially targets VEGR2 (Herbert et al., 2012;Covassin et al., 2006;Fong et al., 1999. If GIPCs limit PLXND1 antiangiogenic signaling, then plxnd1 skt6 homozygotes should be hypersensitive to SU5416. In contrast, plxnd1 skt6 mutants should be hyposensitive to SU5416 if GIPCs bind PLXND1 to directly enable the receptor's antiangiogenic signaling (as in Burk et al., 2017) or if PLXND1 indirectly exerts a GIPC-dependent antiangiogenic effect by sequestering rate-limiting GIPCs away from proangiogenic VEGFR2/Nrp1 signaling (see Gay et al., 2011;Lanahan et al., 2013;Lanahan et al., 2010;Wang et al., 2006;Horowitz and Seerapu, 2012;Lanahan et al., 2014;Chittenden et al., 2006). Finally, if GIPCs play no role in PLXND1 signaling, then WT and plxnd1 skt6 mutants should have similar SU5416 sensitivities.
As a first step, we evaluated the impact of 18-32 hpf SU5416 treatments (0.1-1 mM range) on WT angiogenesis (not shown), which defined a suboptimal SU5416 dose (0.2 mM; in 0.1% DMSO vehicle) capable of inducing minimal antiangiogenic effects on Se and DLAV development by 32 hpf (as in Covassin et al., 2006;Stahlhut et al., 2012; see also Torres-Vázquez et al., 2004; Childs et al., 2002;Isogai et al., 2003;Zygmunt et al., 2012;Yokota et al., 2015). Accordingly, we treated homozygous WT and homozygous plxnd1 skt6 mutants with 0.2 mM SU5416 (experimental group) or 0.1% DMSO (control group); see Figure 3, Figure  DMSO and SU5416 had no apparent effects on embryonic shape, size, somite number, cardiac contractility, and circulation (not shown). DMSO slightly impaired Se and DLAV angiogenesis in both genotypes, consistent with its effects at higher doses (Hallare et al., 2006;Chen et al., 2011;Maes et al., 2012). The weak angiogenic impairment induced by DMSO was greater in plxnd1 skt6 mutants, albeit without a significant penetrance difference ( Figure 3A As expected, SU5416 increased the frequency and severity of angiogenesis deficits in both genotypes ( Figure 3C-E, Figure 3-figure supplement 1 and Supplementary file 4). WT embryos and plxnd1 skt6 mutants treated with DMSO and SU5416 exhibited statistically significantly different distributions of Se and DLAV truncations ( Figure 3E and Supplementary file 4). Notably, DMSO-treated WT embryos and plxnd1 skt6 mutants showed angiogenic deficits at statistically significantly different proportions, but only when comparing the 'minimal' category or the collective grouping of the three defective categories. In contrast, SU5416-treated WT embryos and plxnd1 skt6 mutants displayed angiogenic deficits at statistically significantly different proportions when comparing the three defective categories, either individually or as a group ( Figure 3E and Supplementary file 4). For example, while SU5416 induced a non-significant penetrance difference in Se-DLAV truncations between WT embryos (89.7%) and plxnd1 skt6 mutants (100%) ( In particular, SU5416 disproportionally increased the frequency of the most severe class of angiogenesis deficit in plxnd1 skt6 mutants (maximal Se-DLAV truncation), a defect unique to SU5416-treated embryos of both genotypes ( Figure 3E).
These effects indicate that plxnd1 skt6 mutants are hypersensitive to SU5416. This conclusion fits the prediction of a single hypothesis, namely that GIPC binding limits PLXND1 signaling, and implies that the plxnd1 skt6 allele hypermorphically increases antiangiogenic PLXND1 signaling.
Given the reported coexpression of GIPC1 and GIPC2 in the endothelium (Wang et al., 2006;Chittenden et al., 2006;Jiang et al., 2013;Butler et al., 2016), we focused on the vascular phenotype of the corresponding single and double zebrafish mutants in both zygotic and maternal-zygotic (MZ) combinations at 32 hpf. We scored 112 gipc mutants (69 embryos without maternal and zygotic  gipc1 activity, including 19 animals devoid of zygotic gipc2 activity; plus 24 embryos lacking zygotic gipc1 activity, including 13 without gipc2 activity). This analysis revealed that gipc mutants display recessive Se angiogenic deficits of partial penetrance ( . Specifically, the Se vessels were sometimes missing or truncated, leading to corresponding DLAV gaps. At times, they were abnormally thin and straight-shaped. Importantly, every gipc mutant lacked supernumerary, ectopic, or misguided Se vessels, the hallmark defects of zebrafish and mice Plxnd1 nulls (Torres-Vázquez et al., 2004;Zygmunt et al., 2011;Gitler et al., 2004;Zhang et al., 2009;Gu et al., 2005;Worzfeld et al., 2014) and, reportedly, of murine Gipc1 mutants (Burk et al., 2017). We also found that as maternal or zygotic gipc dosage decreases, the penetrance and expressivity of Se truncations tended to increase (Figure 4-figure supplement 3). Together, these findings indicate that gipc1 and gipc2 act redundantly and that, at the zygotic level, loss of gipc1 causes a greater angiogenesis deficit than the removal of gipc2 ( Overall, the angiogenic deficits of the single (gipc1 and gipc2) and double (gipc1; gipc2) zebrafish mutants resemble those found in gipc1 morphants (but without the delayed vasculogenesis, small aortic lumen, and abnormal body shape of the latter) Hermans et al., 2010;Chittenden et al., 2006) and postnatal Gipc1 mutant mice (impaired arterial branching and arteriogenesis) Chittenden et al., 2006;Dedkov et al., 2007;Lanahan et al., 2010;Moraes et al., 2013;Lanahan et al., 2014;Paye et al., 2009). They are also similar to the vascular defects found in kdrl mutant fish (in which inactivation of a VEGFR2 ohnolog leads to reduced VEGF signaling) (Habeck et al., 2002;Covassin et al., 2009;Bussmann et al., 2008) and plxnd1 skt6 homozygotes.
Reducing Plxnd1 signaling ameliorates the angiogenic deficits of maternal-zygotic (MZ) gipc1 mutants The hypothesis that the angiogenesis deficits of gipc mutants result, at least in part, from increased antiangiogenic PLXND1 signaling predicts that reducing the latter via heterozygosity for the null plxnd1 fov01b allele might have a restorative effect on the angiogenic growth of zebrafish gipc mutants.
In contrast to the normal vasculature of WT embryos and the properly organized but hypoangiogenic vascular tree observed in gipc1 skt1(MZ) ; gipc2 skt(MZ) double MZ mutants (Fischer et al., 2006) ( Figure 6A,C), we found that all the plxnd1 morphants exhibited the hyperangiogenic phenotype of plxnd1 nulls, namely excessive and disorganized Se angiogenesis ( Figure 6B,D). The vascular Bar graph. Percentage of Se in 32 hpf embryos of the indicated genotypes belonging to each of the following four phenotypic classes. Truncated: severe (includes missing Se), medium (yellow), and weak (gray). Non-truncated: complete (black). Significance values were calculated using a two-sided Fisher Exact test and significant differences (p<0.0033) assigned using a Bonferroni-type adjustment for 15 pairwise genotype comparisons (0.05/ 15 = 0.0033). Brackets and asterisks indicate pairs of genotypes with significantly different distributions of these four phenotypic classes. Quantifications. We scored Se angiogenesis in embryos of the following six genotypes: WT (138 Se, 12 embryos; an average of 11.5 Se/embryo), gipc1 skt1 (130 Se, 11 embryos; an average of 11.8 Se/embryo), gipc1 skt1(MZ) (380 Se, 33 embryos; an average of 11.5 Se/embryo), gipc2 skt3/skt4 (130 Se, 11 embryos; an average of 11.8 Se/embryo), gipc1 skt1 ; gipc2 skt3/skt4 (152 Se, 13 embryos; an average of 11.6 Se/embryo), and gipc1 skt1(MZ) ; gipc2 skt3/skt4 (220 Se, 19 embryos; an average of 11.5 Se/embryo). For additional data, graphs, and statistical comparisons related to this figure, see  (C) Bar graph. Percentage of Se-DLAV in 32 hpf embryos of the indicated genotypes belonging to each of the following four phenotypic classes. Truncated: maximal (red; includes missing Se), moderate (yellow), and minimal (gray). Non-truncated: full (black). There was a statistically significant difference (bracket with an asterisk) in distribution of the four phenotypic classes between gipc1 skt1(MZ) and gipc1 skt1(MZ) ; plxnd1 fov01b /+ embryos, as assessed by a two-sided Fisher Exact test (p<0.05). Quantifications. We scored Se-DLAV angiogenesis in embryos of the following two genotypes:
Using this setup, we measured the effects of SEMA3E treatment on relative pERK abundance over time in control cells and cells with GIPC loss, PLXND1 loss, and GIPC-PLXND1 double loss ( Figure 7E). Importantly, we found that this effect is PLXND1-dependent because PLXND1 loss abrogated it ( Figure 7B, red bars in Figure 7E). GIPC loss significantly potentiated the SEMA3E-induced decrease in relative pERK abundance at 45 min ( Figure 7C, green bars in Figure 7E). Moreover, this potentiating effect is also PLXND1-dependent, because pERK levels failed to decrease under SEMA3E stimulation in GIPC-PLXND1 double loss cells ( Figure 7D, blue bars in Figure 7E). Importantly, cells with GIPC-PLXND1 double loss did not show an intermediate relative level of pERK abundance between that of cells with PLXND1 loss ( Figure 7B, red bars in Figure 7E) and cells with GIPC loss ( Figure 7C, green bars in Figure 7E). Instead, the relative pERK abundance of SEMA3E stimulated cells with PLXND1 loss ( Figure 7B, red bars in Figure 7E) and with GIPC-PLXND1 double loss ( Figure 7D, blue bars in Figure 7E) were not statistically significantly different. These quantitative findings align with the qualitatively similar vascular phenotypes of plxnd1 morphants in the WT and gipc1 skt1(MZ) ; gipc2 skt(MZ) backgrounds (Fischer et al., 2006) (Figure 6B,D).
The results of additional cell culture experiments with primary HUVEC agree with these findings ( In summary, the results of HUVEC assays (Figure 7 and Figure 7-figure supplement 3) support the hypothesis that GIPC depletion potentiates antiangiogenic PLXND1 signaling, consistent with the angiogenesis deficits and SU5416 hypersensitivity of plxnd1 skt6 homozygous mutants (Figures 2-3). Given the demonstrated capacity of GIPCs and PLXND1 to bind directly to each other independently of NRP1 Shang et al., 2017) (Figure 1), the alignment of the plxnd1 skt6 and HUVEC data argues that GIPCs bind to PLXND1 to directly limit its antiangiogenic signaling. Furthermore, the role of GIPCs as negative regulators of antiangiogenic PLXND1 signaling is also consistent with the angiogenesis deficits of gipc mutants (Figure 4) and the ameliorating effect that the partial removal of Plxnd1 signaling exerts on the angiogenic deficits of MZ gipc1 mutants ( Figure 5). (E) Bar graph. Means of percentual relative ERK activity (pERK/ERK Total ) under the described conditions (color coded as above) and treatments. Error bars, ± SEM. Relative ERK activity. Statistically significant differences between pairwise combinations of conditions and treatments are indicated (brackets and asterisks). Quantifications. n = 4 independent experiments per PLXND1 gRNA KO (for a pooled total of 8 experiments); n = 4 independent experiments per non-targeting gRNA (for a pooled total of 8 experiments). One-way ANOVA tests were conducted to determine whether relative ERK activity was significantly different between cells in the control, PLXND1 loss, GIPC loss, and GIPC-PLXND1 double loss conditions across each treatment. There were no outliers in the data, as assessed by inspection of a boxplot. Relative ERK activity data were normally distributed, for each treatment, as determined by Shapiro-Wilk's test (p>0.05) except for the SEMA3E 15 min treatment; p=0.031. There was homogeneity of variances, as assessed by Levene's test (p>0.05) for equality of variances in all conditions. One-way ANOVA tests summary. Relative ERK activity was not statistically significantly different between conditions under vehicle treatment (F(3, 28)=0.004, p=1). Relative ERK activity was statistically significantly different between conditions under SEMA3E 15 min treatment (F(3, 28)=10.291, p<0.0005), effect size was w 2 = 0.46. Relative ERK activity was statistically significantly different between conditions under SEMA3E 45 min treatment (F(3, 28)=28.738, p<0.0005), effect size was w 2 = 0.72. Summary of the four statistically significant differences revealed by Tukey post hoc analysis (between conditions under SEMA3E 15 min treatment
Beyond its involvement in developmental angiogenesis, the GIPC-based modulation of endothelial PLXND1 signaling might promote the stabilization, repair, homeostasis, and arteriogenic remodeling of vessels, particularly in contexts with minimal proangiogenic stimulation. These conditions are found in the quiescent vascular beds of adults and are a hallmark of several diseases (see Carmeliet, 2003;Carmeliet, 2005).
We highlight that our results and conclusion challenge the model, recently proposed by Burk et al., that GIPC1 enables PLXND1 activity (Burk et al., 2017). These authors primarily address how Gipc1 and Plxnd1 modulate axonal circuit development using mouse embryos and neuronal explants. They propose that GIPC1's PDZ domain and PLXND1's canonical PBM drive GIPC1-PLXND1 heteromerization and, based on ex vivo forced expression experiments, conclude that the PBM is essential for PLXND1 activity. Finally, using a Plxnd1 null allele and floxed Gipc1 allele that might yield a truncated protein retaining PLXND1 binding (see Shang et al., 2017;Moraes et al., 2013), they peripherally explore the vascular role of the GIPC1-PLXND1 interaction using a qualitative phenotypic analysis that omits Plxnd1 mutant homozygotes.
In all likelihood, the proangiogenic function of GIPCs is multifaceted and involves additional mechanisms besides the negative modulation of PLXND1 signaling. First, GIPCs physically interact with many other PDZ-binding transmembrane proteins and with the retrograde motor MYO6. Some of these GIPC-binding proteins, for example, NRP1, are expressed in endothelial cells and have vascular functions (Shang et al., 2017;Katoh, 2013;Gao et al., 2000;Naccache et al., 2006;Salikhova et al., 2008;Lou et al., 2001;Cai and Reed, 1999). Second, the angiogenesis deficits of gipc1; gipc2 double mutants are more penetrant and severe than those of plxnd1 skt6 homozygotes, in agreement with the notion that the vascular defects of gipc mutants are partially a result of enhanced Plxnd1 signaling.
We highlight that the identification of GIPCs as the first negative intracellular regulators of PLXND1 signaling expands the vascular roles of GIPCs beyond the prevailing notion that GIPC1 promotes arterial branching by facilitating, in an NRP1-dependent manner, proangiogenic VEGF signaling (Chittenden et al., 2006;Lanahan et al., 2010;Fantin et al., 2011;Lampropoulou and Ruhrberg, 2014;Herzog et al., 2011;Plein et al., 2014). Importantly, our findings do not argue against the possibility that GIPCs might also, either directly or indirectly, modulate PLXND1 signaling in connection with NRP1. For example, proangiogenic VEGF signaling via GIPC-NRP1-VEGFR2 complexes might trigger posttranslational modifications in GIPC and PLXND1 that counteract the antiangiogenic activity of the latter by promoting GIPC-PLXND1 interactions or the degradation of the PLXND1 receptor. Alternatively, GIPCs might mediate the formation of novel complexes containing receptors from both pathways (for instance, PLXND1-GIPC-NRP1 and PLXND1-GIPC-NRP1-VEGFR2) that perform still uncharacterized functions (see Chauvet et al., 2007;Bellon et al., 2010).
GIPCs (and MYO6) might limit PLXND1 signaling by restricting the amount of time that the activated receptor spends at the cell surface. This mechanism could serve a dual role of regulating access of the receptor to its catalytic targets such as the Rap1 GTPase, found at the inner cell membrane Worzfeld et al., 2014;Wang et al., 2013), as well as shaping the inactivation kinetics of ligand-bound PLXND1. We note that recent structural studies indicate that mammalian MYO6 also functions as a ubiquitin receptor, thereby suggesting a potential link between GIPCs and the proteasome-mediated degradation of their cargo . Alternatively, GIPCs might function in a non-endocytic manner to limit PLXND1 signaling independently of MYO6. For example, GIPCs might recruit a third protein, which in turn, directly antagonizes PLXND1 signaling (see Lampropoulou and Ruhrberg, 2014;De Vries et al., 1998a;Lou et al., 2001;Wieman et al., 2009;Cai and Reed, 1999;Kofler and Simons, 2016;Guo and Vander Kooi, 2015;De Vries et al., 1998b;Fischer et al., 1999;Fischer et al., 2003;Jeanneteau et al., 2004b;Jean-Alphonse et al., 2014;Lin et al., 2006). Finally, the interaction between PLXND1 and GIPCs might directly influence cell morphology via modulation of cytoskeletal dynamics. Support for this notion comes from genetic experiments in Drosophila and mammalian proteomic studies that implicate GIPCs and MYO6 in actin network stabilization (Djiane and Mlodzik, 2010;O'Loughlin et al., 2018;Isaji et al., 2011;Noguchi et al., 2006).
The mechanisms that regulate GIPC-PLXND1 interaction are unclear. SEMA3E stimulates colocalization of GIPC1 and PLXND1 (Burk et al., 2017;Shang et al., 2017), but the kinetics of their association are not yet defined. Another open question is whether non-canonical PLXND1 ligands (SEMA3A, SEMA3C, SEMA3D, SEMA3G, and SEMA4A) Liu et al., 2016;Hamm et al., 2016) also promote GIPC-PLXND1 complex formation. How SEMA-induced changes in the conformation and oligomerization of PLXND1 (see Pascoe et al., 2015) impact the receptor's ability to interact with GIPCs, or the specificity of their interaction is also unexplored. Finally, reversible posttranslational modifications might modulate the GIPC-PLXND1 interaction. Because phosphorylation of PDZ domains modulates the recruitment of their partners (Lee and Zheng, 2010;Liu et al., 2013), this is a plausible mechanism for GIPC-based control of GIPC-PLXND1 interactions. At the PLXND1 level, the phosphorylation and S-palmitoylation of the receptor's cytosolic tail could also play a modulatory role. Plxns are phosphorylated (Cagnoni and Tamagnone, 2014;Franco and Tamagnone, 2008), and the GBM of PLXND1 harbors a conserved tyrosine located within a consensus Src family kinase phosphorylation site. This residue plugs into a hydrophobic GIPC pocket between the GH1 and PDZ domains Shang et al., 2017). On the other hand, some type I transmembrane proteins, including Plxns, are S-palmitoylated (Holland and Thomas, 2017;Blaskovic et al., 2013), and S-palmitoylation of the carboxy tail of the GIPC1-binding dopamine Drd3 receptor buries the PBM within the cell membrane to prevent the Drd3-GIPC1 interaction (Arango-Lievano et al., 2016).
In conclusion, we have identified a novel role for the GIPCs as pioneer negative regulators of SEMA-PLXND1 signaling. Given the prominent role of this pathway in shaping organogenesis of cardiovascular, nervous, and other systems and its central importance in cancer biology (Moriya et al., 2010;Gay et al., 2011;Valdembri et al., 2016;Oh and Gu, 2013;Gaur et al., 2009;Gu and Giraudo, 2013;Neufeld et al., 2016;Bielenberg and Klagsbrun, 2007), our findings suggest a broad human health relevance for therapeutic targeting of the SEMA-PLXND1 pathway at the GIPC level.

Biochemistry and cell culture experiments
Yeast two-hybrid screen The pGBK-T7 and pACT2 vectors (Clonetech Laboratories, Inc) were used, respectively, for the bait and the preys. The bait consisted of a 96 kDa fusion protein harboring N-terminally the Myc-tagged DNA-binding domain of GAL4 and the cytosolic tail of mPLXND1 (632 aa) at its carboxyl end. A pretransformed cDNA library from E11 stage mouse embryos undergoing PLXND1-dependent vascular and neuronal development van der Zwaag et al., 2002) (Clonetech Laboratories, Inc) was used to make prey proteins fused to the C-terminus of an HA-tagged GAL4 activation domain. The screening (2.8 Â 10 6 preys) was performed by Dualsystems Biotech AG and yielded 15 bait-dependent clones.
Co-immunoprecipitation of mGIPC-mPLXND1 complexes in COS7 cells Details for the cell line used COS-7 cells (Monkey Kidney Fibroblasts) purchased from the American Type Culture Collection (ATCC); ATCC #CRL-1651 (unrecorded lot number, thus the Certificate of Analysis containing detailed authentication and mycoplasma contamination information is unretrievable from ATCC's website).
Cell culture and transfection COS-7 cells were cultured in 10 cm diameter culture dishes (Falcon #379096) with Dulbecco's modified Eagle's medium (DMEM; Corning cellgro #10-013-CV) and 10% fetal bovine serum (FBS; Gemini Bio-Products #100-106). Cells were grown using a 37˚C humidified 5% CO 2 atmosphere. Transfections were performed at 70% confluency with Lipofectamine2000 (Thermofisher Scientific #11668027) according to the manufacturer's specifications using 2.5 ug of each plasmid. Four hours post-transfection the medium was replaced with fresh complete DMEM. Expression vectors and their V5-tagged mPLXND1 and FLAG-tagged mGIPC protein payloads are described in Supplementary file 1 and Supplementary file 8.
Quantification of protein-protein interactions between V5-C-mPLXND1 forms and FLAG-mGIPC1 WT from FLAG immunoprecipitates and V5 coimmunoprecipitates Data were collected from three independent experiments for each protein interaction and analyzed using a one-way ANOVA test followed by a Tukey post hoc analysis. See Figure 1C (left-side bar graph and legend) and Supplementary file 2. Data analysis was carried out using the SPSS Statistics 23.0 software package and the Laerd Statistics tutorial (Statistics, 2015a).
Quantification of the relative abundance of V5-C-mPLXND1 forms and FLAG-mGIPC1 WT from total cell lysates (TCL) Data were collected from three independent experiments for each protein interaction and analyzed using a Kruskal-Wallis H test. See Figure 1C (right-side bar graph and legend) and Supplementary file 2. Data analysis was carried out using the SPSS Statistics 23.0 software package and the Laerd Statistics tutorial (Statistics, 2015b).

Experiments with HUVEC/TERT2 cells Details for the cell line used
Immortalized HUVEC cells (HUVEC/TERT2) purchased from the American Type Culture Collection (ATCC); ATCC #CRL-4053 (authenticated via STR profiling and mycoplasma negative).
Cell culture, lentivirus production, and infection with lentiCRISPR v2-Blast vectors for Cas9 and gRNA coexpression Immortalized HUVEC cells were cultured in low serum medium optimized for human endothelial cells and without human growth factors (VascuLife EnGS Endothelial Medium Complete Kit from Lifeline Cell Technology #LL-0002). Cells were grown in 10 cm diameter culture plates (Falcon #379096) precovered with 0.1% gelatin (SIGMA #G1890-100G) and incubated in a humidified 5% CO 2 atmosphere at 37˚C. Human embryonic kidney 293 T cells (HEK293T; gift of Matthias Stadtfeld; NYU) were cultured in high D-Glucose Dulbecco's modified Eagle's media (Gibco # 10313-021) containing 10% FBS and 2 nM L-glutamine (Gibco #25030-081). Plates (10 cm) of human embryonic kidney 293T (90% confluence) were individually transfected with one of the four lentiCRISPR v2-Blast constructs along with lentivirus packaging and envelope plasmids (gift of Matthias Stadtfeld; NYU) using TransIT-293 reagent (Mirus Bio #Mir2700). Each plate was transfected with 15 mg of the lentiCRISPR v2-Blast construct, 0.75 mg of each lentivirus packaging vector (tat, rev, and gag/pol) and 1.5 mg of pVSV-G. Viral supernatants were harvested and filtered through 0.45 mm at 48 and 72 h. HUVEC cells were infected at~50% confluency using 100 ml of viral supernatants and 5 mg/ml polybrene (Millipore #TR-1003-G) for 2 consecutive days. Pools of stable HUVEC cells were selected for 5 days with 4 mg/ml blasticidin (ThermoFisher Scientific # R21001). Pilot studies confirmed that these selection conditions killed 100% of the uninfected cells.

Isolation of monoclonal cell populations of PLXND1 knockout cells infected with lentiCRISPR v2-Blast vectors for Cas9 and gRNA coexpression
Pools of stable HUVEC cells were harvested using trypsin (Gibco #25300-054) and grown at a very low density in 15 cm diameter culture plates (Corning #430599). Twelve individual cells from each of the PLXND1-KO1 and PLXND1-KO2 gRNAs expressing lentiCRISPR v2-Blast vectors were grown to 100% confluence in 96-well plates. These cells were then expanded in larger plates until confluent 10 cm plates were obtained. The resulting monoclonal cell populations were used for characterizing the DNA sequence (GENEWIZ performed sequencing) and protein level via Western blot to identify PLXND1 knockout lines. Two of these (PLXND1 gRNA KO1 and KO2) were used for the experiments presented here.
Detection of phospho-active (pERK) and total ERK (ERK Total ) and quantification of relative ERK activity Cells were stimulated with either vehicle or 2 nM recombinant human Semaphorin 3E (R and D Systems #3239-S3B) for 15 and 45 mins. pERK and ERK Total levels were measured using Western blots from total cell lysates (TCLs) prepared as follows. Cell plates were placed on an ice bed, rinsed with PBS and lyzed in 250 ml of lysis buffer (50 mM Tris pH 7.5, 0.5 mM EDTA, 150 mM NaCl, 1 mM EDTA, 1% SDS) supplemented with protease (cOmplete, Roche #12683400) and phosphatase (1 mM sodium fluoride, 10 mM b-glycerophosphate, and 1 mM sodium vanadate) inhibitors. Lysates were loaded onto NuPAGE 4-12% Bis-Tris protein gels (Invitrogen #NP0336BOX) and proteins transferred to Immobilon-P membranes (Millipore; #IPVH00010). pERK and ERK Total were immunodetected with rabbit and mouse antibodies against ERK1/2 (CST #4370S and #4696S, respectively) and revealed with Western Lightning PLUS-ECL (PerkinElmer #NEL103001EA). Relative ERK activity was calculated ratiometrically as pERK/ERK Total .

Experiments with primary HUVEC cells Details for the cell line used
Normal primary human umbilical vein endothelial cells (HUVEC) were purchased from Lifeline Cell Technology; #FC-0003 (STR profiling data unavailable, mycoplasma negative).

Cell collapse assays
Cells between passages 2 and 6 were infected with shRNA lentiviral particles and puromycinselected for 48 h as described above. Cells were divided into two pools, one for validating the knockdown and the other for performing the cellular morphology assay. Cells for the latter experiment were starved for 6 h and then treated with either vehicle or 10 nM recombinant human Semaphorin 3E (R and D Systems #3239-S3B) for 45 min. Cells were washed with ice-cold 1xPBS, fixed at room temperature (RT) for 15 min with 4% PFA in PBS. Cells were then washed 2x with PBS and permeabilized with 0.1% Triton-X100 in PBS for 5 min at RT. Next, cells were incubated for 20 min at RT in 20 mg/ml phalloidin-tetramethylrhodamine B isothiocyanate (SIGMA #P1951) to label the F-actin cytoskeleton. Cells were next washed 3x in PBS, incubated for 5 min at RT in 0.5 mg/ml DAPI (Molecular Probes #D1306) to visualize nuclei, and kept in PBS at 4˚C until imaging. Fluorescent images were acquired with an Eclipse Ti-E inverted microscope (Nikon) using the 10X objective (NA 0.3).

Cell collapse data collection and evaluation
Images were collected from 50 to 100 cells from each of three independent experiments per knockdown and treatment condition. Images were processed with FIJI (https://imagej.net/Fiji). As automated extraction of cellular contours was unfeasible, cells were qualitatively classified as uncollapsed, collapsed, or hyper-collapsed based on their size and morphology (Figure 7-figure  supplement 3).
Western blot validation of GIPC knockdowns and PLXND1 knockout in primary and immortalized HUVEC Western blots of TCLs were performed with rabbit antibodies for GIPC1 (Proteintech Group; 14822-1-AP) and GIPC2 (Abcam #ab175272). A mouse antibody was used for PLXND1 (R and D Systems #MAB41601). GAPDH (loading control) was detected with a rabbit antibody (CST; #5174P).
Forced endothelial expression of 2xHA-Plxnd1 forms in plxnd1 fov01b mutants We used the GAL4/UAS system (Scheer and Campos-Ortega, 1999) and Tol2-based transient transgenesis (Kikuta and Kawakami, 2009) to drive forced mosaic endothelial-specific expression of 2xHA-tagged forms of zebrafish Plxnd1 (2xHA-Plxnd1 WT or 2xHA-Plxnd1D GBM ) and the green fluorescent marker EGFP. Briefly, 1 nl of a 100 pg of Tol2 mRNA and 20 pg of vector DNA solution was injected into the cytoplasm of one-cell plxnd1 fov01b mutants carrying both the endothelial Tg(fli1a: GAL4FF) ubs4 GAL4 driver  and the red nuclear arterial Tg(flt1:nls-mCherry) skt7 reporter (this study). Faithful coexpression of 2xHA-Plxnd1 and EGFP was accomplished using an IRES (internal ribosomal entry site element)-based bicistronic UAS cassette (Kwan et al., 2007). Embryos were fixed at 32 hpf, immunostained and imaged as described below.
Quantification of the vascular patterning activity of exogenous 2xHA-Plxnd1 forms in plxnd1 fov01b mutants Embryo treatments with SU5416 and DMSO SU5416 (SIGMA #S8442) was prepared and used as in Covassin et al. (2006) and Stahlhut et al. (2012). Briefly, a 200 mM SU5416 stock solution in DMSO (SIGMA #D8418; vehicle) was dissolved in fish water to a final concentration of 0.2 mM SU5416 (a suboptimal dose) and 0.1% DMSO. Control, vehicle-only (0.1% DMSO) treatments were also performed. Homozygous WT and homozygous plxnd1 skt6 mutant embryos were manually dechorionated before receiving the SU5416 and DMSO treatments using a common solution for both genotypes. Embryos were treated from 18 to 32 hpf to specifically target both Se and DLAV angiogenesis (see Torres-Vázquez et al., 2004;Zygmunt et al., 2011;Childs et al., 2002;Isogai et al., 2003;Zygmunt et al., 2012;Yokota et al., 2015), and then fixed for immunostaining.
Quantification of angiogenesis deficits in the trunk's arterial tree of WT and mutant zebrafish embryos Embryos carrying the Tg(fli1a:EGFP) y1 vascular reporter were fixed at 32 hpf, genotyped using a tail biopsy, immunostained to visualize both the vasculature and the somite boundaries and confocally imaged as described below. Scoring of angiogenesis deficits was performed bilaterally in the~six somite-long region dorsal to the yolk extension using one of the two following scoring methods and with knowledge of the genotype. Scoring of 'Se and DLAV truncations': this scoring method emphasizes truncations found within the dorsal side of the trunk's arterial tree. This method was used for Figure 2C,E,F-I (see also  . The classification of Se and DLAV truncations is based on the relative span of Se and DLAVs along the dorsoventral and anteroposterior axes, respectively, using as reference the following landmarks. The horizontal myoseptum and the actual (or expected) level of the DLAV. Four phenotypic classes are used to define Se and DLAV spans. The three Se-DLAV truncation categories are as follows: maximal (includes both Se that are missing and those that fail to grow dorsally past the horizontal myoseptum); moderate (includes Se that grow dorsally past the horizontal myoseptum but not further than half the distance between the horizontal myoseptum and the level of the DLAV); and minimal (Se that grow dorsally past half the distance between the horizontal myoseptum and the level of the DLAV, but which form an incomplete DLAV). An incomplete DLAV is one that fails to span the distance between the anteriorly and posteriorly flanking ipsilateral somite boundaries. Non-truncated Se-DLAV: full (Se that grow dorsally to the level of the DLAV and that form a complete DLAV). Scoring of 'Se truncations': this scoring method emphasizes truncations found within the ventral side of the trunk's arterial tree. This method was used for Figure 4 (see also Figure 4-figure supplements 2-3 and Supplementary file 5). The classification of Se truncations is based on the relative span of Se vessels measured along the dorsoventral axis. Briefly, the length of a perpendicular line traced between the actual (or expected) Se sprouting site and the level of the DLAV was assigned a value of 100%. Four phenotypic classes are used to define Se span. Se truncation categories: severe (0-25%, includes missing Se), medium (26-50%), and weak (51-75%). Non-truncated: complete (76-100%). Note that both the Se-DLAV and Se scales used for quantifying angiogenesis deficits are based on the relative span of the vascular structures scored, which eliminates the noise that variations in size between embryos with similar body proportions would otherwise introduce.

Comparisons of angiogenesis deficits between genotypes and treatments
Significance statistical values were calculated using a two-sided Fisher Exact test with the aid of the SISA web tool (http://www.quantitativeskills.com/sisa/statistics/fiveby2.htm). When performing more than one pairwise comparison (as in Figures 3-4), significant differences were assigned using a Bonferroni adjustment. This conservative adjustment reduces the likelihood of obtaining false-positive results (type I errors or the rejection of true null hypotheses) when simultaneously applying many statistical tests to a dataset (Statistics, 2016). The Bonferroni adjustment is commonly used to compare drug effects and in angiogenesis studies (see Hamada, 2018;Weichand et al., 2017;Basagiannis et al., 2016).

Confocal imaging of zebrafish embryos
Confocal images of fixed immunostained embryos were taken with a Leica TCS SP5 microscope using a water dipping 40x lens (0.8 NA). The 488, 561, and 647 nm laser lines were used. Images were processed with FIJI (https://imagej.net/Fiji).

Ethics statement
Zebrafish embryos and adults were kept and handled using standard laboratory conditions at New York University and under IACUC-app roved animal protocols (#151202-01 and #170103-01). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Additional files

Supplementary files
. Supplementary file 1. Miscellaneous tables listing the following information. Vectors for expressing PLXND1 and GIPC proteins/fragments, primers for genotyping Tg(fli1a:GAL4FF) ubs4 zebrafish, oligos for assembling DNA templates for in vitro transcription of gRNAs for zebrafish genome editing and for making lentiCRISPRv2-Blast vectors for Cas9 and gRNA coexpression for use in HUVEC, cognate sequences of WT alleles and mutant alleles generated in this study via genome editing, and primers for genotyping mutant alleles generated in this study via genome editing. Related to . Supplementary file 4. Tables comparing the Se-DLAV truncations of wild-type embryos and plxnd1 skt6 mutants (at 32 hpf) in animals treated with DMSO and SU5416. Related to Figure 3E and  Data availability All data generated or analysed during this study are included in the manuscript and supporting files.