Vegfd modulates both angiogenesis and lymphangiogenesis during zebrafish embryonic development

Signaling through the VEGF/VEGFR pathway is essential for multiple steps during development of the vasculature. In zebrafish, primary angiogenesis occurs when arterial sprouts emerge from the dorsal aorta (DA) from ∼22 hpf to give rise to the intersegmental arteries (aISV) in a process dependent on Kdr, Kdrl and Vegfa (Bahary et al., 2007; Covassin et al., 2006; Lawson et al., 2002; Lawson and Weinstein, 2002; Nasevicius et al., 2000; Olsson et al., 2006). During primary angiogenesis, blood vessel sprouting is led by tip cells, which express high levels of Flt4 (Vegfr3 in mammals) and display active Erk signaling (Gore et al., 2011; Shin et al., 2016a). In zebrafish, whole-genome duplication has resulted in the presence of two Vegfr2 paralogs, Kdr and Kdrl (Bussmann et al., 2008; Covassin et al., 2006). Tip cells are highly responsive to Kdr, Kdrl and Flt4 signaling, and have low Notch-signaling activity. In the zebrafish, the loss of Dll4, a Notch ligand, leads to increased angiogenic sprouting (Leslie et al., 2007; Siekmann and Lawson, 2007). Dll4 negatively regulates arterial Flt4 signaling and suppresses the arterial response to Vegfc (Hogan et al., 2009b; Villefranc et al., 2013). Likewise, Flt1 (Vegfr1) also plays a negative regulatory role, its soluble form (sFlt1) suppresses arterial Vegfr signaling and angiogenesis, with morpholino (MO) knockdown of Flt1 leading to aISV hyperbranching (Krueger et al., 2011). Although Vegfa plays the dominant role in controlling arterial sprouting and Vegfc has been shown to influence this process, Vegfd has not previously been implicated in the regulation of primary sprouts (Hogan et al., 2009b; Jakobsson et al., 2010; Siekmann and Lawson, 2007).

The sprouting of both venous intersegmental vessels (vISV) and lymphatic endothelial cells (LECs) occurs by secondary angiogenesis from ∼32 hpf in zebrafish (Isogai et al., 2003). Sprouts emerging from the cardinal vein can become vISVs or LECs when they either anastomose with existing aISVs to form vISVs, or fail to anastomose and migrate dorsally to form parachordal LECs at the horizontal myoseptum (HM) (Isogai et al., 2003; Koltowska et al., 2013, 2015). Parachordal LECs in the HM proliferate and migrate dorsally and ventrally to form the major trunk lymphatics: the thoracic duct (TD), dorsal longitudinal lymphatic vessel (DLLV) and the intersegmental lymphatic vessels (ISLVs) (Cha et al., 2012; Hogan et al., 2009b; Küchler et al., 2006; Yaniv et al., 2006). Facial lymphangiogenesis proceeds from the common cardinal vein (CCV) as lymphatic sprouts migrate to form the lateral facial lymphatic (LFL), medial facial lymphatic (MFL) and the otolithic lymphatic vessel (OLV) by 5 dpf (Okuda et al., 2012). The formation of zebrafish lymphatic vessels is dependent on the Ccbe1/Vegfc/Flt4 signaling axis (Hogan et al., 2009a; Le Guen et al., 2014; Villefranc et al., 2013). Vegfd has been shown by MO knockdown to compensate for loss of Vegfc during facial lymphatic development but not elsewhere in the embryo (Astin et al., 2014). Growth factor-induced signaling downstream of Flt4 occurs through the activation of Erk, which is necessary and sufficient in the trunk to induce the normal sprouting of LECs and expression of the LEC marker Prox1 (Koltowska et al., 2015; Shin et al., 2016b). During facial lymphangiogenesis from the CCV, Erk signaling controls LEC sprouting but not Prox1 expression, suggesting spatially distinct mechanisms that remain to be fully understood (Shin et al., 2016b).

VEGFD is structurally related to VEGFC and can signal through both VEGFR3 and VEGFR2 in humans but only through Vegfr3 in mice (Achen et al., 1998; Baldwin et al., 2001a). Overexpression of VEGFD in tumor cells induces tumor angiogenesis, lymphangiogenesis and the dilation of collecting lymphatic vessels, ultimately promoting the metastatic spread of tumor cells (Karnezis et al., 2012; Stacker et al., 2002, 2001; Von Marschall et al., 2005). Expression of VEGFD in solid human tumors is a strong indicator of metastasis and poor prognosis (Achen et al., 2005; Stacker et al., 2014). However, despite this potent capability of VEGFD, its function during embryonic development has remained unclear. Vegfd knockout mice on a mixed genetic background are viable with no obvious defects in blood or lymphatic endothelium (Baldwin et al., 2005). Vegfc, Vegfd double-knockout mice display a developmental phenotype that appears to be the same as that of Vegfc knockout mice (Haiko et al., 2008), hence it is considered that Vegfd is not involved during embryonic development. However, postnatal phenotypes have been observed in some tissues and include a reduction in the abundance of pulmonary lymphatics (Baldwin et al., 2005) and, in adult mice of a pure C57BL/6 background, smaller initial dermal lymphatics and compromised function (Paquet-Fifield et al., 2013).

We have recently found that vegfd is expressed throughout the embryonic trunk during zebrafish development (Duong et al., 2014), suggesting a potential function that remains to be fully explored. Here, we report the generation of a zebrafish vegfd mutant. vegfd is dispensable for development of trunk blood and lymphatic vessels. However, we find that vegfd is essential for normal facial lymphangiogenesis and combines with vegfc contextually during lymphangiogenesis throughout the trunk and face. Furthermore, Vegfc and Vegfd together influence aISV development in a number of knockdown or overexpression settings. In this context, Vegfd is likely to act through Kdr, which we find binds to Vegfd in zebrafish. These findings identify Vegfd as a developmental regulator of both lymphangiogenesis and angiogenesis.

To investigate the developmental role of Vegfd, we used TALEN mutagenesis to generate a zebrafish mutant harboring a 7 bp deletion at base 162 of the vegfd-coding sequence. The deletion site is in an exon common to both of the known protein-coding transcript variants that contain the Vegfd homology domain (VHD). The 7 bp deletion introduces a stop codon at amino acid 65, which is prior to the VHD of the Vegfd protein and is predicted to result in a null mutant (Fig. 1A,B). Analysis of the number of aISVs, vISVs and the extent of the TD in the trunks of vegfd^uq9bh mutant embryos revealed no differences between homozygous mutant, heterozygous or wild-type embryos (Fig. 1C,D, Fig. S1A,B). vegfd^uq9bh homozygous mutant embryos were raised to adulthood and produced viable offspring (Fig. 1E-H). Analysis of vegfd mRNA levels in embryos derived from in-crosses of vegfd homozygous mutant adults showed that these mutant embryos had less than 50% of the transcript levels of stage-matched wild-type embryos, indicative of nonsense-mediated decay (NMD) (Fig. S1C). Additionally, we did not observe any change in vegfc mRNA levels in vegfd^uq9bh homozygous mutants, indicating that compensation for the loss of vegfd is unlikely to occur (Fig. S1C).

Vegfd morpholino (MO) knockdown has previously been reported and revealed that Vegfd compensates for the loss of facial lymphatics in vegfc^hu5055 mutants (Astin et al., 2014). We performed the reciprocal experiment, injecting a vegfc MO into embryos produced from a cross of homozygous with heterozygous vegfd^uq9bh mutants. Using an approach blinded for genotype, we scored injected embryos as displaying ‘severe’ (defined as a complete loss of the MFL and LFL) or ‘partial’ (indicating reduced length of the MFL and LFL) phenotypes and found vegfd^uq9bh mutants were enriched in the ‘severe loss’ category of embryos, confirming previous findings and suggesting that these mutants did not express functional Vegfd (Fig. S1D,E). Taking into account this phenocopy of the MO-induced defects, the nature of the mutation and detection of NMD, we suggest that the vegfd^uq9bh allele is a loss-of-function allele and likely to be null.

To more accurately characterize the role of Vegfd during development of the facial lymphatics, we generated vegfd^uq9bh, vegfc^hu5055 double mutants in a Tg(fli1a:negfp), Tg(-5.2lyve1b:dsRed) transgenic background. This previously reported vegfc^hu5055 mutant is hypomorphic and displays a strong but variable loss of lymphatics (Le Guen et al., 2014). The number of individual nuclei in each of the different facial lymphatic vessels was quantified in embryos from vegfd^uq9bh, vegfc^hu5055 double heterozygous in-crosses (Fig. 1J-Q). In double mutant embryos, the facial lymphatics failed to develop (Fig. 1N). Strikingly, vegfd^uq9bh mutant embryos displayed fewer facial lymphatic nuclei in the LFL and MFL than in wild-type embryos, without any reduction in vegfc gene dose (Fig. 1L,O,P). There were also significantly fewer nuclei in vegfc^hu5055, vegfd^uq9bh double heterozygotes compared with wild-type embryos for both the LFL and MFL (Fig. 1K,O,P). Vegfd did not contribute to the formation of the otolithic lymphatic vessel (OLV), which was vegfc dependent (Fig. 1Q). Together, these data show that vegfd regulates development of the facial lymphatics, specifically the MFL and LFL, and that combined gene dose for vegfc and vegfd is essential for the formation of the complete facial lymphatic network.

vegfd^uq9bh mutant embryos develop normal trunk lymphatics in a vegfc wild-type background (Fig. 1D, Fig. S1A). To determine whether vegfd works together with vegfc in this context, we examined the formation of parachordal LECs in the vegfc^hu5055 hypomorphic mutant in a vegfd^uq9bh mutant background (Fig. 2A-F). In siblings that were homozygous wild type for the vegfc^hu5055 allele, the loss of a single copy or both copies of vegfd did not produce any lymphatic phenotype, indicating that Vegfd is dispensable for normal trunk lymphangiogenesis in the presence of wild-type levels of Vegfc (Fig. 2A,D). However, in a vegfc^hu5055 heterozygous background, vegfd^uq9bh homozygous and heterozygous mutants developed significantly fewer parachordal LECs than vegfd^uq9bh wild-type embryos, demonstrating a compensatory role for Vegfd with reduced gene dose for vegfc (Fig. 2E). Analysis of vegfd^uq9bh allelic loss in a vegfc^hu5055 homozygous mutant background did not reveal further significant interactions (Fig. 2F).

We next scored the extent of the TD across nine somites at 5 dpf. In siblings that were homozygous wild type for the vegfc^hu5055 allele, the loss of a single or both copies of vegfd again did not produce a lymphatic phenotype (Fig. 3A,B). In both vegfc^hu5055 heterozygous and homozygous mutant backgrounds, vegfd^uq9bh homozygous loss led to significantly reduced TD extent (Fig. 3C-F,H,I). To more accurately quantify the loss of TD in vegfd^uq9bh, vegfc^hu5055 double mutant backgrounds, we precisely scored the number of nuclei present in the TD across 9 somites using the Tg(fli1a:negfp), Tg(-5.2lyve1b:dsRed) transgenic strain (Fig. 3J-L). We found that many hypomorphic vegfc^hu5055 homozygous mutant embryos partially formed a TD; however, this was significantly reduced in the absence of vegfd (Fig. 3J-L). In the majority of cases, double mutants failed to form a single LEC in the region of the embryonic TD. Together, these data show a compensatory role for vegfd with vegfc during trunk lymphangiogenesis, although vegfc alone is sufficient to generate a trunk lymphatic vasculature during development.

Given the known and suggested biochemical interactions between Flt4, Ccbe1 and Vegfd in zebrafish and in mammals, we next examined whether there were genetic interactions between mutants for these components during the formation of the TD. We analyzed TD formation in embryos from vegfd^uq9bh, flt4^hu4602 and vegfd^uq9bh, ccbe1^hu3613 double heterozygous in-crosses at 7 dpf. ccbe1^hu3613 homozygous mutant embryos had a complete loss of the TD irrespective of vegfd^uq9bh genotypic status and we were unable to detect any genetic interaction in these crosses (data not shown). Analysis of TD formation in embryos from vegfd^uq9bh, flt4^hu4602 double heterozygous in-crosses revealed that, in a flt4^hu4602 heterozygous background, vegfd^uq9bh mutants had significantly reduced TD formation compared with vegfd^uq9bh heterozygous or wild-type embryos (Fig. 3M-O). Hence, the flt4^hu4602 mutation dominantly interacts with the vegfd^uq9bh mutation, which does not normally result in a TD phenotype.

VEGFC and VEGFD are capable of signaling through VEGFR3 (Flt4 in zebrafish) or VEGFR2 (Kdr and Kdrl in zebrafish) in different settings (Achen et al., 1998; Covassin et al., 2006; Hogan et al., 2009b; Joukov et al., 1996; Karnezis et al., 2012; Le Guen et al., 2014; Villefranc et al., 2013). Overexpression of vegfd or vegfc by mRNA injection in zebrafish causes bilateral turning of the aISVs at 28-32 hpf (Astin et al., 2014; Hogan et al., 2009b). We co-injected vegfd mRNA with a flt4 MO and found no decrease in the number of turned aISVs when compared with vegfd mRNA injection alone (Fig. 4B,F,I), whereas co-injection of flt4 morpholino with vegfc mRNA led to a rescue of the vegfc mRNA-driven phenotype, as previously reported (Hogan et al., 2009b) (Fig. S2A,B). To determine which receptor Vegfd is likely signaling through, we co-injected vegfd mRNA with morpholinos targeting other Vegf receptors. Importantly, we ensured that each of these MOs was efficacious by using previously published MOs and reproducing previously validated phenotypes. Specifically for the MOs used in this study, ISV hyperbranching is robustly induced using the flt1 MO (Krueger et al., 2011) and the dll4 MO (Hogan et al., 2009b; Siekmann and Lawson, 2007); the kdr and kdrl MO combined robustly block ISV development (but neither alone achieves this) (Covassin et al., 2006; Wiley et al., 2011); and the flt4 MO blocks the formation of lymphatic vessels in a highly reproducible manner (Hogan et al., 2009b). Co-injection of the flt1 MO with vegfd mRNA led to an increase in the number of turned aISVs at 28-32 hpf compared with embryos injected with vegfd mRNA alone (Fig. 4B,G,H,J). Flt1 is known to suppress Vegfa signaling through Vegfr2 (Kdr or Kdrl in zebrafish) (Krueger et al., 2011; Wiley et al., 2011), but enhancement of the hyperbranching phenotype by vegfd overexpression is unexpected. Co-injection of vegfd mRNA with kdr MO rescued the bilateral turning of the aISV phenotype at 28-32 hpf (Fig. 4B,C,I), whereas injection with kdrl MO failed to rescue (data not shown). We also co-injected vegfc mRNA with kdr MO and found no rescue of the vegfc-driven phenotype (Fig. S2B). It has previously been reported that overexpression of Vegfd leads to an upregulation of Vegfr2 in rabbit blood vessels (Rissanen et al., 2003), which could contribute to the ectopic turning of aISVs in this setting. To examine this, we performed qPCR on 28 hpf vegfd mRNA-injected embryos and controls; we found no significant changes in the expression of kdr, kdrl, flt4, vegfaa, vegfab, dll4, flt1 or fli1a (Fig. 4K). Together, these data suggest that in the context of zebrafish aISV development, Vegfc and Vegfd are capable of acting through Flt4 and Kdr, respectively.

The genetic evidence presented above suggests that zebrafish Vegfd influences primary angiogenesis via interaction with Kdr. To further confirm this, we tested the capacity of zebrafish Vegfd to bind Kdr. Comparison of zebrafish and human VEGFD revealed that the zebrafish protein has a similar primary domain structure with N- and C-terminal propeptides flanking a central VEGF homology domain (VHD) (Fig. 5A). Previous studies with human (Stacker et al., 1999) and mouse (Baldwin et al., 2001b) VEGFD demonstrated that the N- and C-terminal propeptides can be proteolytically cleaved to generate a mature form consisting of dimers of the VHD. The mature form of human VEGFD has much higher affinity for human VEGFR2 compared with the full-length form (Stacker et al., 1999); furthermore, the proteolytic processing of VEGFD is important for its activity in disease settings (Harris et al., 2011). We therefore expressed a form of full-length zebrafish Vegfd tagged at the N-terminus with the FLAG octapeptide, designated zVegfd-FULL-N-FLAG, in transiently transfected human HEK293T cells to assess whether it can be proteolytically processed. Western blot analysis of conditioned media (CM) revealed the presence of bands at ∼47, 36 and 15 kDa (Fig. 5B, left panel), consistent with the expected sizes of full-length Vegfd, of a partially processed form containing the N-terminal propeptide and the VHD, and of a form consisting of only the N-terminal propeptide, respectively. These findings confirm that zebrafish Vegfd can be proteolytically processed, at least when expressed in mammalian cells.

Given that zebrafish Vegfd can be processed, and that the mature form of human VEGFD exhibits higher affinity for VEGFR2 than other forms of the growth factor, we used a mature form of zebrafish Vegfd, consisting of the VHD, for receptor-binding studies. A version of mature zebrafish Vegfd tagged at the N-terminus with FLAG, designated zVegfdΔNΔC-FLAG, was expressed in transiently transfected HEK293T cells – the size of the subunits of this glycoprotein was in the 21-28 kDa range (Fig. 5B, right panel), as expected based on the sizes of mature human and mouse Vegfd (Achen et al., 1998; Baldwin et al., 2001b). CM containing zVegfdΔNΔC-FLAG were incubated with Ig-fusion proteins, consisting of the extracellular domains of zebrafish Kdr or Kdrl and the Fc region of mouse IgG, that were bound to protein A-sepharose (see Materials and Methods). Western blotting of material precipitated by the Ig-fusion proteins/protein A-sepharose revealed that the Kdr extracellular domain could bind zVegfdΔNΔC-FLAG, whereas the extracellular domain of Kdrl could not (Fig. 5C, left panel). Hence, in line with our epistasis data above, mature Vegfd can bind Kdr, but not Kdrl in zebrafish.

As our data suggest that Vegfd is capable of signaling through Kdr, it is possible that vegfd^uq9bh mutants, upon loss of Kdrl, would show arterial defects similar to the loss of Kdr and Kdrl together (Wiley et al., 2011). To test this, we injected the kdrl MO into embryos derived from a homozygous vegfd^uq9bh mutant crossed to a heterozygous vegfd^uq9bh carrier. We independently quantified the number of aISVs that extended fully from the DA to the DLAV and the number of aISVs that had failed to sprout from the DA at 28 hpf (Fig. S3A,B). We found no significant difference in the number of aISVs between mutants or siblings upon Kdrl knockdown, suggesting that Vegfa plays a dominant role in aISV sprouting (Fig. S3A,B).

Given that vegfc mutants have been shown to interact genetically with kdrl during primary angiogenesis (Villefranc et al., 2013), we also examined the number of cells in the aISVs in vegfc^hu5055, vegfd^uq9bh double mutants at 28 hpf using the Tg(fli1a:negfp) transgenic background; however, we did not find any significant difference (Fig. S3C). This implies that Vegfa plays a more potent role than Vegfd. Although Vegfd is dispensable for early primary angiogenic sprouting, the ability to signal through Kdr when overexpressed shows that Vegfd is capable of inducing angiogenesis, which may be important in pathological states or in some mutant or knockdown settings.

In zebrafish, loss of the Notch ligand Dll4 leads to increased tip cells and angiogenic sprouting because Dll4 negatively regulates Vegfc-driven arterial Flt4 signaling (Hogan et al., 2009b; Siekmann and Lawson, 2007; Villefranc et al., 2013). However, in vegfc mutants injected with dll4 MO some arterial hyperbranching was still observed, despite the vegfc^um18 allele used being shown to be a null allele (Villefranc et al., 2013). We injected dll4 MO into embryos from a vegfc^hu5055, vegfd^uq9bh double heterozygous incross and quantified hyperbranching of aISVs followed by retrospective genotyping of the embryos (Fig. 6A-E). Although there was no reduction in hyperbranching in vegfd^uq9bh mutants alone, there was a significant reduction of hyperbranching in vegfd^uq9bh, vegfc^hu5055 double mutants compared with single vegfc^hu5055 mutants (Fig. 6C-E). This suggests that Vegfc is sufficient to drive the phenotype, but Vegfd acts in a compensatory manner. We conclude that Dll4 suppresses both Vegfc and Vegfd signaling in developing intersegmental arteries.

Vegfr1 (Flt1 in zebrafish) plays a negative regulatory role during angiogenesis, with its soluble form (sFlt1) suppressing arterial Vegfr signaling and the knockdown of Flt1 leading to arterial hyperbranching (Hiratsuka et al., 1998; Krueger et al., 2011). Given the known interaction between Dll4/Notch signaling and Flt1 (Jakobsson et al., 2010), and our observation that Flt1 knockdown enhances the vegfd mRNA injection phenotype (Fig. 4H), we injected flt1 MO into embryos from a vegfc^hu5055, vegfd^uq9bh double heterozygous in-cross. In a vegfc^hu5055 heterozygous background, we found significantly fewer hyperbranched aISVs in vegfd^uq9bh heterozygous and homozygous mutant embryos compared with wild-type sibling embryos at 72 hpf (Fig. 6F-J). In a vegfc^hu5055 mutant background, there were also fewer hyperbranched aISVs in vegfd^uq9bh heterozygous and homozygous mutant embryos versus wild-type siblings (Fig. 6J). We also found less hyperbranching of aISVs when vegfc^hu5055 heterozygous and mutant embryos were compared with wild-type siblings (Fig. S4A). These data indicates that Flt1 inhibits Vegfc and Vegfd signaling during primary angiogenesis, either directly or through its modulation of Vegfa signaling.

To further examine this scenario, we generated a dll4 mutant harboring a 2 bp deletion at base 90 of the dll4-coding sequence. This deletion introduces a frame shift and a premature stop codon at amino acid 47 (Fig. S4B). We found that at 5 dpf dll4^uq10bh heterozygous embryos displayed mild hyperbranching of the intersegmental arteries, confirming dll4 MO and hypomorphic mutant phenotypes (Fig. S4C) (Leslie et al., 2007; Siekmann and Lawson, 2007). dll4 carriers had a poor survival rate of 6.25% in the F2 population. Given that the aISV hyperbranching in flt1 morphant embryos was reduced in the vegfc^hu5055, vegfd^uq9bh double mutants, we injected the flt1 MO into embryos from a dll4^uq10bh+/− outcross and quantified the extent of artery hyperbranching at 72 hpf. This is a time point where the flt1 MO injection induces hyperbranching of aISVs, but dll4^uq10bh heterozygous mutants do not show hyperbranching (Fig. 6G,L). We found that dll4^uq10bh heterozygosity significantly enhanced the hyperbranching of aISVs induced by flt1 MO knockdown (Fig. 6N,O).

Finally, following our observations that the bilateral turning of arteries by vegfc and vegfd overexpression is Flt4 or Kdr dependent, respectively, we co-injected kdr and flt1 MOs, and also flt4 and flt1 MOs and found a significant reduction in hyperbranching in each of these separate co-injection scenarios compared with flt1 MO only (Fig. S4D). Consistent with vegfc having the greatest impact on hyperbranching in mutant analyses (Fig. 6J), Flt4 knockdown produced the most significant reduction in this experiment (Fig. S4D). These data, combined with the findings above further demonstrate that flt1 genetically interacts with vegfc and vegfd.

Here, we describe a zebrafish vegfd mutant produced by Talen-induced mutagenesis. The vegfd^uq9bh mutant is viable and displays no obvious defects in primary or secondary angiogenesis in the trunk. This mutant is most likely a null, based on the nature of the mutation, an interaction with vegfc compared with published loss-of-function data (Astin et al., 2014) and NMD of the vegfd transcript. VEGFD has been widely considered to be dispensable for developmental lymphangiogenesis based on studies in knockout mice (Baldwin et al., 2005; Haiko et al., 2008). However, phenotypes have been observed in postnatal settings (Paquet-Fifield et al., 2013) and during pathological processes (Stacker et al., 2001). Furthermore, Vegfd genetically interacts with Sox18 during mouse blood vascular development (Duong et al., 2014), suggesting its potential during embryonic development. Our findings are made by high-resolution quantitative analysis of lymphangiogenesis, including determining total LEC numbers per vessel. These phenotypes demonstrate that Vegfd plays compensatory roles with Vegfc in both the face and trunk, and Vegfd alone is indispensable during facial lymphangiogenesis of the LFL and MFL (see Fig. 7). These findings warrant further detailed analysis of LEC number in different vessel beds in the Vegfc, Vegfd double-knockout murine model. Of relevance to human lymphatic disease, we show that vegfd gene dose broadly modifies Vegfc/Flt4 pathway phenotypes in a variety of mutant models and settings. Zebrafish with mutations in genes involved in this pathway are disease models of heritable lymphedema (Alders et al., 2009; Connell et al., 2010, 2009; Crawford et al., 2016; Evans et al., 2003; Gordon et al., 2013; Hogan et al., 2009a,b; Karkkainen et al., 2000; Shin et al., 2016b; Villefranc et al., 2013). Given that ∼30% of cases of primary (inherited) lymphoedema are caused by mutations in the Ccbe1/Vegfc/Vegfr3 (Flt4) pathway (Mendola et al., 2013), it may be that Vegfd variants can modify patient outcomes.

During primary angiogenesis in zebrafish, vegfc and vegfd are expressed in the DA and the trunk, respectively (Duong et al., 2014; Hogan et al., 2009b). Despite this, normal angiogenesis was observed in vegfc, vegfd double mutants and is also seen in flt4-null mutants (Kok et al., 2015). Both Dll4 and Flt1 negatively regulate arterial angiogenesis, with loss of either resulting in severe hyperbranching of the aISVs (Krueger et al., 2011; Leslie et al., 2007; Siekmann and Lawson, 2007). Here, we find that both Vegfc and Vegfd contribute to these phenotypes (see Fig. 7). Interestingly, although the aISV defects observed upon overexpression of vegfd and vegfc are grossly similar, the ligands appear to act through different mechanisms. The vegfc overexpression phenotype is strongly enhanced in the absence of Dll4, whereas the vegfd overexpression phenotype is only slightly enhanced (Hogan et al., 2009b). Consistent with this, we found that the bilateral turning of aISVs following injection of vegfc mRNA requires Flt4 and following injection of vegfd mRNA requires Kdr. Overall, our data suggest that zebrafish Vegfd can signal through Kdr (Vegfr2) during primary angiogenesis and most likely through Flt4 during secondary angiogenesis based on genetic interactions. Interestingly, a recent study by Shin et al. shows that Prox1 induction in zebrafish facial lymphatics can occur in the absence of Flt4 signaling (Shin et al., 2016b), but in the trunk the induction of Prox1 expression is Vegfc/Flt4 dependent (Koltowska et al., 2015; Shin et al., 2016b). These studies combined suggest that differential signaling events are at play in facial and trunk lymphangiogenesis. Perhaps, given our findings, further examination of Vegfd signaling may help to illuminate these different mechanisms.

Our biochemical data show that Vegfd is capable of binding to Kdr but not to Kdrl. Zebrafish Vegfd is capable of rescuing Vegfaa zebrafish mutants (Rossi et al., 2016), which strongly supports our data and together suggests that Vegfd can activate signaling downstream of Kdr. As mouse Vegfd does not bind to Vegfr2 but the human protein does (Baldwin et al., 2001a), zebrafish may present a very useful alternative model for studying the function of VEGFD/VEGFR2 signaling. Further to these binding studies, we also show that full-length zebrafish Vegfd is processed by proteolytic cleavage of the N- and C-terminal propeptides to generate the mature Vegfd peptide. This resembles the processing of mammalian/human VEGFD, via the proprotein convertases furin, PC5 and PC7 (McColl et al., 2003, 2007), and the processing of zebrafish Vegfc, which likely requires the proprotein convertases furin, PC5 and PC7 (Khatib et al., 2010), similar to requirements in mammalian systems (Joukov et al., 1996; Siegfried et al., 2003). Activation of VEGFC requires CCBE1 and ADAMSTS3, but VEGFD is not processed in a CCBE1-dependent manner in these mammalian models (Bui et al., 2016; Jeltsch et al., 2014). By contrast, in zebrafish-overexpression models it was shown that Vegfd activity is Ccbe1 dependent (Astin et al., 2014). Uncovering differences between species in VEGF processing, activation and signaling may reveal novel and selective capabilities that can be exploited therapeutically and further our fundamental understanding of vascular development.

All zebrafish strains were maintained and animal work performed in accordance with the guidelines of the animal ethics committee at The University of Queensland, Australia. The previously published transgenic and mutant lines are listed in Table S1.

Talen monomers (Table S2) for vegfd and dll4 were generated using the golden gate kit (Cermak et al., 2011), cloned into the pCS2TAL3RR and pCS2TAL3DD backbones and transcribed as previously described (Dahlem et al., 2012). Fish were screened by high resolution melt analysis (HRMA) using a Viia7 Real-Time PCR System (Applied Biosystems) and F1 embryos carrying a 7 bp deletion for vegfd and 2 bp for dll4 were outcrossed to Tg(flt:YFP);Tg(-5.2lyve1b:dsRed) to generate an F2 population. Primers for HRMA and sequencing of individual mutations are listed in Table S2.

Capped RNA was transcribed and injected as previously described (Hogan et al., 2009b). MOs used are listed in Table S3. DNA extraction from fin clips and embryos was performed as previously described (Dahlem et al., 2012). Primers for genotyping vegfd and vegfc mutants by KASP assays (LGC Genomics) are listed in Table S2. dll4 mutants were genotyped by EcoRI restriction digest of amplified PCR product (Table S2). Quantitative real-time PCR was performed as previously described (Coxam et al., 2014) using the primers indicated in Table S2.

Zebrafish embryos were mounted in 0.5% low melting agarose and imaged using a Zeiss LSM 710 FCS confocal microscope. Images were processed using Image J 1.47 software (National Institutes of Health). The number of nuclei expressing Tg(fli1a:negfp) and co-expressing Tg(-5.2lyve1b:dsRed) in the TD or facial lymphatics were manually counted through a z-stack. Nuclei in the MFL and LFL were counted from the branch point of these two vessels. Quantification of bilateral turning of aISVs at 28-32 hpf, hyperbranching of aISVs at 3 dpf and PLs in 54 hpf, and TD in 5 dpf embryos was performed by live scoring using a Leica M165FC microscope.

The extracellular domains for Kdr (amino acids 23-773) and Kdrl (amino acids 29-738) were PCR amplified and cloned using Infusion enzyme (Clontech, CA, USA) into the pFUSE-mlgG2Aa-Fc2 (IL2ss) vector (InvivoGen, CA, USA) using the primers listed in Table S2. Full-length vegfd cDNA encoding both N- and C-terminal propeptides flanking the Vegfd homology domain (VHD) (with the signal peptide amino acids 1-22 removed) and cDNA encoding the zebrafish Vegfd VHD (amino acids 85-198) were PCR amplified using primers listed in Table S2. Products were then cloned into pEFBOSSFLAG (Achen et al., 1998) immediately downstream from the DNA sequence for the interleukin 3 signal sequence using Infusion enzyme (Clontech, CA, USA) to generate zVegf-d-FULL-N-FLAG and zVegf-dΔNΔC-FLAG, respectively. A threonine and an arginine residue were included in the Vegfd constructs as artifacts from cloning, see Fig. 5 for details.

HEK293T cells were maintained in DMEM supplemented with 10% FBS, 2 mM glutamax (Life Technologies Australia, Victoria, Australia), 100 U/ml penicillin and 100 µg/ml streptomycin (Life Technologies Australia) at 37°C in a humidified atmosphere of 10% CO₂.

HEK293T cells were transfected with expression vectors, as indicated, using lipofectamine 2000 (Life Technologies Australia) according to the manufacturer's instructions. Briefly, cells in 5 ml of DMEM in 10 mm tissue culture dishes were transfected with 10 µg of plasmid DNA and 40 µl of lipofectamine 2000. Cells were then incubated for 6 h before addition of 5 ml of DMEM containing 0.4% BSA, and were then starved by further incubation for 24 h. Conditioned media (CM) were collected and clarified via centrifugation at 150 g for 5 min at room temperature. CM containing Vegfd derivatives were used either directly for western blotting or in binding studies with Ig-fusion proteins. For binding studies, Ig-fusion proteins were precipitated from CM with protein A-sepharose (Sigma Aldrich, MO, USA) by incubation at 4°C for 3 h followed by centrifugation (1500 g for 5 min at 4°C). Ig-fusion proteins bound to protein A-sepharose were then incubated with CM containing zebrafish Vegfd for 16 h at 4°C, and the sepharose was then washed once with 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.1% Triton X-100, once with 500 mM NaCl and once with 50 mM Tris-HCl (pH 8.0), before heating in reducing LDS sample buffer (Life Technologies Australia) at 95°C for 5 min. Samples were then analyzed by SDS-PAGE/western blotting using Bolt 4-12% Bis Tris plus gels (Invitrogen, CA, USA) and proteins were transferred to iBlot2 nitrocellulose membrane (Invitrogen) using an iBlot2 protein transfer apparatus (Life Technologies, CA, USA). Blots were probed with M2 anti-FLAG monoclonal antibody (catalogue number F3165; Sigma-Aldrich) that had been conjugated to 800 IR dye (LI-COR Biosciences, NE, USA) according to the dye manufacturer. Detection was performed using an Odyssey Cx imaging system (LI-COR Biosciences). Soluble Ig-fusion proteins used for binding studies were shown to be of expected sizes by western blotting with goat anti-mouse IgG antibodies (catalogue number 926-32210, LI-COR Biosciences).

Statistical analysis was performed using Prism software (GraphPad software). When data conformed to parametric assumptions, ANOVA using Fisher's individual error post-hoc test was used to identify significant differences. When parametric assumptions were not met, Kruskal–Wallis test was used.

Imaging was performed in the Australian Cancer Research Foundation's Dynamic Imaging Facility at IMB. We thank Kylie Georgas for help with figures and critical reading of the manuscript.