Vegfc acts through ERK to induce sprouting and differentiation of trunk lymphatic progenitors

The lymphatic system comprises a series of blind-ended vessels that are connected to the circulatory system and play a central role in fluid homeostasis (Alitalo, 2011). Lymphatic vessels are also an important conduit for the immune system and facilitate lipid absorption in the digestive system. Among the important regulators of lymphatic vessel growth and maintenance are vascular endothelial growth factor c (Vegfc) and its receptor tyrosine kinase, Flt4 (also known as vascular endothelial growth factor receptor-3, Vegfr-3; Bahram and Claesson-Welsh, 2010). Flt4 is expressed specifically and dynamically on endothelial cells in the blood vascular and lymphatic systems. In both mouse and zebrafish embryos, flt4 is expressed in actively sprouting endothelial cells, as well as endothelial cells lining large caliber veins and lymphatic vessels (Kaipainen et al., 1995; Siekmann and Lawson, 2007; Tammela et al., 2008; Thompson et al., 1998). Loss or reduction of flt4 or vegfc leads to severe defects in lymphatic development in zebrafish and mouse (Hogan et al., 2009b; Karkkainen et al., 2004; Kuchler et al., 2006; Villefranc et al., 2013; Yaniv et al., 2006). In both cases, initial sprouting of lymphatic progenitors from large veins is blocked. Importantly, mutations in human FLT4 or VEGFC are associated with congenital forms of lymphedema that are characterized by lymphatic hypoplasia (Gordon et al., 2013; Karkkainen et al., 2000). Thus, Vegfc signaling through Flt4 is essential for lymphatic development and this pathway is conserved in vertebrates.

Despite our knowledge of lymphatic vessel morphogenesis, less is known about what acts downstream of Flt4 in vivo to drive this process. In cell culture, common downstream effectors of receptor tyrosine kinases (RTKs), such as the serine/threonine kinases ERK (also known as MAP kinase) and AKT, can be activated in response to Flt4 dimerization (Deng et al., 2013; Makinen et al., 2001; Pajusola et al., 1994; Salameh et al., 2005; Wang et al., 2010). Moreover, specific tyrosine residues in the human FLT4 cytoplasmic domain have been implicated in the activation of distinct downstream targets and cellular outputs (Salameh et al., 2005). For example, Y1063 binds the adaptor molecule CRK and is required for activation of Jun N-terminal kinase (JNK; MAPK8), which facilitates endothelial cell survival. Similarly, Y1230 and Y1231 associate with the Grb2 adaptor to activate AKT and ERK, which promotes endothelial sprouting and proliferation (Salameh et al., 2005). Thus, FLT4 dimerization activates distinct downstream kinases to elicit different cellular outputs. However, the in vivo relevance of these studies is unclear and what acts downstream of Flt4 during lymphatic development is not well known.

An elegant approach to dissect RTK outputs in vivo is through the generation of animals bearing targeted mutations in residues within the cytoplasmic domain (Klinghoffer et al., 2002; Tallquist et al., 2000). This approach has been successful in implicating downstream effectors of vascular endothelial growth factor receptor-2 (Vegfr-2; also known as Kdr), a receptor for Vegfa that is similar to Flt4 and is essential for vascular development and angiogenesis (Koch et al., 2011). Mice bearing a point mutation in Vegfr-2 Y1173, which is essential for activating phospholipase c gamma 1 (Plcg1; Takahashi et al., 2001), exhibit defects similar to Vegfr2-null embryos (Sakurai et al., 2005). Zebrafish embryos mutant for plcg1 also recapitulate loss-of-Vegfa phenotypes, underscoring the conserved nature of downstream Vegfa signaling components in vivo (Covassin et al., 2009, 2006; Lawson et al., 2003). Interestingly, loss of plcg1 specifically affects artery development in the zebrafish, but lymphatic morphogenesis is normal (Lawson et al., 2003; Yaniv et al., 2006). This observation indicates that Plcg1, a presumed general RTK effector molecule, is selectively required for Vegfr-2, but not Flt4 signaling at this stage of development. Thus, there are clearly distinct downstream signaling molecules activated by Vegfa and Vegfc in the different contexts of vascular and lymphatic development. However, targeted strategies to identify downstream Flt4 effectors in vivo have been limited.

The zebrafish is an ideal model in which to study lymphatic development (Kim and Jin, 2014). The external development and transparency of the zebrafish embryo permits direct visualization of lymphatic morphogenesis and forward genetic screens have yielded mutants affecting lymphatic development (Hogan et al., 2009a,b; Villefranc et al., 2013). In addition, the recent advent of programmable site-specific nucleases has enabled targeted deletions to be easily introduced into the zebrafish genome (Blackburn et al., 2013). In this study, we used transcription activator-like effector nucleases (TALENs; Christian et al., 2010) to generate embryos bearing targeted in-frame deletions in flt4 to define downstream pathways important for lymphatic development. By comparing zebrafish embryos bearing a deletion that eliminates tyrosines 1226 and 1227 in the Flt4 cytoplasmic domain with embryos bearing a null flt4 allele and with a vegfc mutant, we find that there are distinct requirements for signaling during facial and trunk lymphatic development. In particular, ERK is an essential effector that drives sprouting and differentiation of trunk lymphatic progenitors downstream of Vegfc. Moreover, phenotypic analysis of flt4 alleles suggests that the facial lymphatic system is the first functional lymphatic network, whereas the thoracic duct is largely dispensable during zebrafish larval stages.

To define signaling molecules that act downstream of Flt4 during early lymphatic development, we applied TALENs to generate targeted deletions in the zebrafish flt4 gene. In this study we focused on Y1226 and Y1227 (Y1226/7), which are homologous to Y1230 and Y1231 (Y1230/1) in human FLT4 (Fig. 1A) where they are important for ERK and AKT activation in cell culture studies (Salameh et al., 2005). Unfortunately, we failed to identify active TALENs that could introduce lesions directly at Y1226/7 (data not shown). Therefore, we used two pairs of TALENs flanking the genomic sequence encoding Y1226/7 to generate zebrafish in which exons 27 and 28 were directly fused. We identified a founder fish bearing a lesion that results in a 23 amino acid in-frame deletion and eliminates Y1226/7 in the Flt4 cytoplasmic domain (Fig. 1A-C). The official allele designation for this allele is flt4^um200, but it is hereafter referred to as flt4^Y1226/7Δ. Despite this deletion, we did not observe obvious changes in flt4 transcript or protein levels, as assessed by whole-mount staining (Fig. 1D). Furthermore, dimerization of the Flt4^Y1226/7Δ cytoplasmic domain induced tyrosine autophosphorylation similar to an Flt4 cytoplasmic domain bearing Y1226/7F point mutations and to wild type (Fig. 1E). Dimerization of Flt4^Y1226/7Δ also resulted in a significant attenuation of basal and induced phosphorylated ERK (pERK) levels in HEK293 cells compared with the wild-type Flt4 cytoplasmic domain, similar to the effect of Y1226/7F mutations (Fig. 1F,G). This effect is similar to previous studies of human FLT4 bearing Y1230/1F mutations (Salameh et al., 2005). Unlike previous studies, we did not observe pAKT induction following Flt4 dimerization, possibly due to the already high levels of AKT phosphorylation in HEK293 cells (Fig. 1F). Together, these observations suggest that the 23 amino acid deletion in flt4^Y1226/7Δ mutant embryos can attenuate ERK activation, but does not generally affect Flt4 catalytic activity.

To assess the importance of Y1226/7, we compared phenotypes in flt4^Y1226/7Δ mutant embryos with those associated with vegfc^um18 or flt4^um131 mutations. The vegfc^um18 allele causes a premature stop codon that truncates the Vegfc protein and prevents its secretion, causing fully penetrant defects in lymphatic morphogenesis (Villefranc et al., 2013). The um131 mutation truncates Flt4 in the extracellular domain (Kok et al., 2015) and causes loss of flt4 transcript and protein (Fig. 1C,D). Therefore, flt4^um131 is likely to be a null allele and is hereafter referred to as flt4^null.

One of the earliest processes that requires Vegfc/Flt4 signaling is the formation of the primordial hindbrain channel (PHBC), a transient vein that returns cranial circulation to the heart (Covassin et al., 2006). In wild-type embryos, the PHBC forms by sprouting and fusion of two distinct endothelial cell populations, one that originates immediately caudal to the eye, the other near the Duct of Cuvier (Siekmann et al., 2009). Consistent with previous studies (Covassin et al., 2006; Villefranc et al., 2013), nearly all flt4^null mutant embryos fail to form a complete PHBC by 26 hpf, similar to vegfc^um18 mutants (Fig. 2A,B,G; data not shown). Likewise, flt4^Y1226/7Δ mutant embryos displayed PHBC defects at a penetrance similar to the flt4^null allele (Fig. 2C,G).

We next assessed the severity of lymphatic defects in flt4 mutant embryos by observing formation of the thoracic duct (TD), which is located immediately ventral to the dorsal aorta in the trunk of wild-type larvae at 5 days post fertilization (dpf; Fig. 2D). In both flt4^null and flt4^Y1226/7Δ mutant embryos we noted a fully penetrant deficit in TD formation at 5 dpf (Fig. 2E,F,H). Both alleles also displayed weaker partially penetrant TD defects in heterozygous embryos (Fig. 2H; data not shown). Similar results were observed in vegfc^um18 mutant embryos, although a much higher proportion of heterozygous embryos displayed TD defects (Fig. 2H).

TD formation entails a complex migratory process in which Prox1-positive endothelial cells sprout from the posterior cardinal vein (PCV) beginning at 30 hpf (Koltowska et al., 2015; Nicenboim et al., 2015; Yaniv et al., 2006). After sprouting dorsally, Prox1-positive lymphatic endothelial cells branch laterally and sprout along the horizontal myoseptum as parachordal cells (PACs). The PACs then sprout ventrally, usually along intersomitic arteries (ISAs) in adjacent somites, to give rise to the TD. During the same developmental window, Prox1-negative venous endothelial cells sprout to form intersomitic veins (ISVs) in parallel and fuse with ISAs, thereby establishing a circulatory network with alternating arterial and venous connections (Isogai et al., 2003). In wild-type embryos at 36 hpf, we observed sprouts emerging from the PCV using Tie2 (Tek) immunostaining and these were absent in flt4^null and flt4^Y1226/7Δ mutant embryos (Fig. 3A-C), suggesting a block in both venous and lymphatic sprouting. Consistent with loss of venous sprouts, both flt4^null and flt4^Y1226/7Δ mutant embryos displayed a significant increase in ISA connections, an effect that was more severe in vegfc^um18 mutants (Fig. 3D). In parallel, we noted a failure to form PACs in both flt4^null and flt4^Y1226/7Δ mutant Tg(fli1a:egfp)^y1 embryos (Fig. 3E-G), suggesting that sprouting of lymphatic progenitors was similarly affected. Accordingly, flt4^Y1226/7Δ mutant Tg(fli1a:egfp)^y1 embryos lack Prox1-positive PCV endothelial cells, similar to vegfc^um18 and flt4^null mutants (Fig. 3H-J; data not shown). In all cases, Prox1 was still apparent in non-vascular tissue (Fig. 3H,I; data not shown). A time course demonstrated that the population of Prox1-positive endothelial cells in the PCV normally expanded from 30 to 38 hpf in wild-type embryos (Fig. 3K), whereas in flt4 mutants these cells were rarely observed during this period. In this case, we utilized the flt4^C562Δ allele, which would be expected to disrupt Vegfc binding (Tvorogov et al., 2010) and causes phenotypes similar to flt4^null (Fig. S1). Together, these results suggest that signaling through Y1226/7 is essential for the initial sprouting of lymphatic and venous endothelial cells, and is required for differentiation and expansion of lymphatic progenitors.

The tyrosines in human FLT4 that are homologous to zebrafish Flt4 Y1226/7 are also important for ERK and AKT activation (Salameh et al., 2005). We observe robust pERK immunostaining in endothelial cells both in the dorsal wall of the PCV, as well as cells sprouting from the PCV, in wild-type Tg(fli1:egfp)^y1 embryos (Fig. 4A). By contrast, pERK staining was absent from the PCV in flt4^Y1226/7Δ mutant siblings, but unaffected in non-endothelial cells (Fig. 4B). Furthermore, we find that Prox1 and pERK colocalize in lymphatic progenitors in the dorsal wall of and sprouting from the PCV (Fig. 4C). These observations suggest that ERK activation may be essential for lymphatic and venous sprouting downstream of Flt4. Accordingly, treatment with SL327 (a MEK inhibitor) beginning at 28 hpf to block ERK phosphorylation (Fig. S2A,B) prevented lympho-venous sprouting (Fig. 4D,G; we use the term ‘lympho-venous’ to include all sprouts emanating from the PCV because Tie2 immunostaining cannot distinguish between lymphatic and venous sprouts). In addition, ERK inhibition prevented the appearance of Prox1-positive endothelial cells in the PCV and blocked PAC formation in Tg(fli1a:egfp)^y1 embryos (Fig. 4E,H,F,I). We could not assess TD formation as prolonged SL327 treatment led to loss of circulation and severe non-vascular defects at later stages (data not shown). In contrast to SL327, AKT inhibition did not affect Prox1 expression, lympho-venous sprouting, or PAC formation (Fig. S2C-F; data not shown).

To determine if ERK activation was sufficient to drive lymphatic sprouting and differentiation in a cell-autonomous manner in the absence of Vegfc/Flt4, we transiently expressed activated Map2k2b (actMap2k2b) in endothelial cells of flt4^C562Δ mutant embryos using the GAL4/UAS system (Fig. 5A). In this case, we utilized transiently injected Tol1 transposons (Koga et al., 2008) to avoid mobilizing the Tg(cdh5:gal4ff)^mu101 transgene, which is flanked by Tol2 direct repeat elements (Bussmann et al., 2011). flt4^C562Δ;Tg(cdh5:gal4ff)^mu101 mutant embryos injected with a control Tol1 transgene did not form lympho-venous sprouts and failed to induce Prox1 expression in the PCV while displaying robust endothelial-expressed nlstdTomato (Fig. 5B,B′). An activated Akt transgene also failed to rescue Prox1 expression or venous sprouting (Fig. 5C,C′; data not shown; 0 Prox1-positive cells out of 501 transgene-expressing cells in 24 embryos). By contrast, actMap2k2b rescued Prox1 expression and lymphatic sprouting in the PCV of flt4^C562Δ mutant embryos (Fig. 5D,D′; Fig. S2G). Taken together, these results demonstrate that activation of ERK is sufficient downstream of Flt4 to initiate lymphatic sprouting and differentiation in the PCV.

The lymphatic system maintains fluid homeostasis by collecting interstitial fluid and returning it to the circulatory system. Consequently, perturbations to lymphatic morphogenesis prevent interstitial fluid collection and lead to lymphedema. Indeed, both vegfc^um18 and flt4^null mutant larvae displayed edema around the digestive tract, heart and eyes and failed to inflate their swim bladder by 5 dpf, whereas wild-type larvae were normal (Fig. 6A-C). These defects were observed at nearly full penetrance and resulted in poor adult survival (Fig. 6E; data not shown). Despite their failure to form a TD (Fig. 2F,H), flt4^Y1226/7Δ mutants were indistinguishable from wild type until 10 dpf (Fig. 6A,D,E) and were viable and fertile as adults (data not shown). These results suggest that the flt4^null and flt4^Y1226/7Δ alleles display distinct effects on lymphatic function and that TD formation is dispensable for early lymphatic function.

Because trunk lymphatic defects were the same in flt4^null and flt4^Y1226/7Δ mutants, yet overt lymphedema was only observed with complete loss of flt4, we investigated whether this was due to distinct effects on facial lymphatic development (Okuda et al., 2012). Facial lymphatic progenitors sprout from the common cardinal vein (CCV) to fuse with lymphatic endothelial cells that emerge from the primary head sinus (PHS) giving rise to an initial facial lymphatic vessel (FLV; see Fig. S3; Okuda et al., 2012). As in the PCV, lymphatic endothelial progenitors emerging from the CCV and PHS are positive for Prox1 (Fig. 7A,D). However, we did not observe pERK immunostaining in CCV endothelial cells at the time during which Prox1-positive lymphatic progenitors are beginning to sprout, and flt4^null mutants appeared indistinguishable from wild type in this regard (Fig. S4A,B). Furthermore, numbers of Prox1-positive endothelial cells were unchanged in embryos mutant for either flt4^null or flt4^Y1226/7Δ compared with wild-type siblings (Fig. 7B,C,E-G). In addition, Prox1 expression in CCV endothelial cells was not affected by treatment with 15 µM SL327, although these same embryos display loss of Prox1 in trunk PCV endothelial cells (Fig. S4C-G).

To visualize facial lymphatic morphology in vegfc and flt4 mutants, we relied on the Tg(lyve1b:DsRed2)^nz101 transgenic line (Okuda et al., 2012). Because developing facial lymphatics lie directly lateral to the PHS, we color-coded vertical projections from confocal stacks based on depth. This approach allowed us to distinguish more easily the PHS from the lateral facial lymphatic (LFL), otolithic lymphatic vessel (OLV) and the initial budding of lymphatic branchial arch (LAA) sprouts (Fig. 7H). In all Tg(lyve1b:DsRed2)^nz101 embryos mutant for either vegfc^um18 or flt4^null, we observed severe truncation of the LFL and a failure to form LAA buds or the OLV, whereas formation of the PHS and posterior cerebral vein (PCeV) was unaffected (Fig. 7I-L). By contrast, nearly all flt4^Y1226/7Δ mutant embryos exhibited formation of the LFL and LAA buds, although the OLV appeared to be delayed (Fig. 7K,L). Interestingly, treatment with SL327 from 28 to 50 hpf blocked facial lymphatic morphogenesis, although initial sprouting of lymphatic progenitors from the CCV was apparent (Fig. S3H,I).

The facial lymphatic network should retain a connection to the vascular system to allow excess fluid to return back into circulation. Accordingly, we observed perfusion of the LFL with Qdots following angiography, but not with gata1:dsRed-expressing red blood cells (Fig. 7M). By contrast, both red blood cells and Qdots were readily apparent in the PHS (Fig. 7M). In vegfc^um18 and flt4^null mutant embryos, we observed incomplete perfusion of the LFL compared with wild-type siblings (Fig. 7N-P) consistent with reduced formation of the facial lymphatics. By contrast, flt4^Y1226/7Δ mutant embryos were indistinguishable from wild type (Fig. 7N,Q). Taken together, these results demonstrate distinct signaling requirements for facial and trunk lymphatic development and suggest that the facial lymphatics are the first functional lymphatic network in the zebrafish.

In this study, we have applied TALENs to generate a flt4 allele that affects signaling through tyrosine residues previously implicated in AKT and ERK activation in cell culture (Salameh et al., 2005). By comparing phenotypes associated with deletion of Y1226/7 with that of a null flt4 mutant we find evidence of distinct genetic requirements for Vegfc/Flt4 signaling in the development of the facial and trunk lymphatic networks. In the trunk, Flt4 signals through ERK to drive progenitor differentiation, expansion and sprouting at the earliest stages of lymphatic development. These observations are consistent with recent findings that demonstrate an important role for Vegfc in the direct induction of Prox1 expression in lymphatic progenitors within the posterior cardinal vein, as well as their subsequent expansion (Koltowska et al., 2015). Our results further show that ERK plays an essential role downstream of Vegfc and Flt4 in this crucial early step in lymphatic development.

Despite the early role for a Vegfc/Flt4/ERK signaling cascade in trunk lymphatic development, we do not observe a similar genetic requirement for Y1226/7 in facial lymphatics. Recent work has demonstrated a differential requirement for vegfd (figf) and vegfc, as well as chemokine receptor signaling, between facial and trunk lymphatic vessels (Astin et al., 2014). Thus, our observations are consistent with the likelihood that there is anatomical heterogeneity in lymphatic endothelial progenitors during early embryonic development. Unlike the PCV, ERK activation does not appear in Prox1-positive cells within the CCV. However, later sprouting of facial lymphatic vessels is blocked by SL327 treatment, suggesting a later role for ERK, albeit independent of signaling through Flt4 Y1226/7. Interestingly, Prox1 is retained in the CCV even in flt4^null embryos, suggesting that initial lymphatic specification is independent of both Vegfc and ERK in this particular anatomical location. This phenotype is similar to that observed in mouse, where loss of Vegfc blocks sprouting of lymphatic progenitors from the cardinal vein, but initial Prox1 expression appears normal in these cells (Karkkainen et al., 2004). Thus, several of the discrepancies that have been noted between Vegfc signaling mutants in zebrafish and mouse embryos might be due to anatomical distinctions between lymphatic networks rather than species-specific differences. In any case, it will be important to identify the molecular basis of heterogeneity in these lymphatic networks to gain a better understanding of the requisite signaling effectors responsible for their development in different anatomical locations.

The differential requirement for the trunk and facial lymphatic networks strengthens the likelihood that the Y1226/7 deletion specifically affects particular downstream effectors, rather than acting as a general hypomorphic allele that simply reduces overall Flt4 signaling output. If the latter were true, we would expect intermediate levels of expressivity and penetrance, especially when considering the trunk lymphatic and lymphedema phenotypes. Rather, we observe a stark difference in phenotypic penetrance when comparing the flt4^null and flt4^Y1226/7Δ alleles: whereas both alleles behave identically regarding trunk lymphatic development, there are clear distinctions in facial lymphatic morphogenesis and function, for which signaling through the domain encoding Y1226/7 appears to be dispensable. Coupled with our demonstration of normal transcript and protein expression levels, and no general effect on autophosphorylation, these observations provide strong evidence that the flt4^Y1226/7Δ allele is specifically affecting downstream Flt4 signaling. However, given the nature of the allele, we cannot rule out the possibility that this effect may not be dependent on tyrosines 1226/7, but rather relies on other residues within this domain. Moving forward, it will be important to develop more robust approaches to establish point mutations using available genome-editing tools in zebrafish.

In human FLT4, tyrosines homologous to zebrafish Y1226/7 are required for activation of both AKT and ERK in endothelial cell culture (Salameh et al., 2005). However, our chemical inhibition and autonomous transgenic rescue experiments suggest that activation of ERK, but not AKT, is necessary and sufficient to induce trunk lymphatic development in the zebrafish. It remains a possibility that signaling through AKT may be utilized downstream of Flt4 at later stages of lymphatic morphogenesis or maintenance. Indeed, signaling through the phosphoinositide 3-kinase (PI3K) pathway, an essential upstream activator of AKT, has been implicated in both mice and humans at later stages of lymphatic morphogenesis and maintenance, and in an organ-specific manner (Boscolo et al., 2015; Mouta-Bellum et al., 2009). Thus, although these studies suggest that it is likely that AKT plays an important role in lymphatic development, our findings rule out this effector during the earliest steps of trunk lymphatic development. Rather, Vegfc signals specifically through ERK to initiate sprouting and differentiation of trunk lymphatic progenitors in the zebrafish. Generation of targeted alleles that specifically disrupt PI3K signaling will be helpful to improve our understanding of the role of this pathway in zebrafish lymphatic morphogenesis.

Consistent with our findings, several observations in mouse have provided convincing indirect evidence that implicate ERK in lymphatic morphogenesis and differentiation. Mice lacking both Spred1 and Spred2, or deficient in the RasGAP Rasa1, all of which are negative regulators of the MEK/ERK signaling cascade, display increased numbers of lymphatic endothelial cells (Lapinski et al., 2012; Taniguchi et al., 2007). Moreover, lymphatic endothelial cells in these mice exhibit ectopic ERK signaling and abnormal lymphatic function. Similarly, inducible endothelial-driven expression of an activated Raf1S259A transgene, which can lead to increased ERK activation, induces ectopic lymphatic formation in mouse. In this case, Raf1S259A also drives expression of lymphatic regulators Prox1 and Sox18 in endothelial cells in vivo (Deng et al., 2013). This finding is consistent with our results that demonstrate that ERK activation can potently induce Prox1 expression and induce lymphatic sprouting downstream of Flt4. The similarity between flt4^Y1226/7Δ, flt4^null and vegfc^um18 embryos further suggests that ERK activation occurs downstream of Vegfc and Flt4, rather than being targeted by other lymphatic RTKs. Thus, our studies provide a crucial link to observations in mouse and demonstrate that Vegfc signaling through Flt4 activates ERK to initiate lymphatic morphogenesis and differentiation.

The discovery and initial characterization of an early lymphatic system in zebrafish was only made a decade ago (Yaniv et al., 2006). The majority of subsequent work has focused on the development of and genetic signals that govern formation of the TD, an anatomically conserved lymphatic vessel (Mulligan and Weinstein, 2014). With the advent of lymphatic- and vein-specific transgenic zebrafish, more recent work has described the development of a facial lymphatic network between 2 and 5 dpf (Okuda et al., 2012). Given that flt4^Y1226/7Δ mutants specifically lack a trunk lymphatic network, our analysis would suggest that the facial lymphatic network, but not the TD, is the first functional lymphatic vessel during zebrafish development. Moving forward, it will be important to apply similar genetic approaches to dissect the distinct signaling requirements that pattern these lymphatic networks.

Zebrafish were handled according to approved University of Massachusetts Medical School Institutional Animal Care and Use Committee protocols. Tg(fli1a:egfp)^y1, Tg(cdh5:gal4ff)^mu101, Tg(kdrl:egfp)^la116, Tg(gata1:DsRed)^sd2 and Tg(lyve1b:DsRed2)^nz101 are described elsewhere (Bussmann et al., 2011; Choi et al., 2007; Lawson and Weinstein, 2002; Okuda et al., 2012; Traver et al., 2003). The vegfc^um18 and flt4^null alleles have been described previously (Kok et al., 2015; Villefranc et al., 2013).

To delete Y1226/7, we designed TALENs to the following sites: exon 27,

5p target: 5′-CACAAGAGGAGATCAGAT-3′, 3p target: 5′-GTTGCTGTACCGTGTTGGC-3′, separated by a 16-nt spacer; exon 28, 5p target: 5′-GCCGTTTGCTGGCTGTGTGA-3′, 3p target: 5′-TCACTCGAGAATGACATGTG-3′ separated by a 16-nt spacer. To introduce mutations in flt4 at C562, we targeted the following sites in exon 12 using TALENs: 5p target: 5′-GCCAGTGTGCCAGCTAT-3′, and 3p target: 5′-CTTTTCCCACTTTGTTCT-3′, separated by a 17-nt spacer. TALENs were constructed as described elsewhere (Kok et al., 2014). TALEN mRNAs were injected into wild-type embryos at the one-cell stage. To establish the Y1226/7Δ allele, exon 27 and exon 28 TALEN pairs were co-injected. Founders were identified as previously described (Meng et al., 2008). For C562, we identified a founder bearing a 9-nt deletion (flt4^C562Δ). For Y1226/7, we identified a founder bearing a 4031-nt deletion that eliminates 23 amino acids encoded by exons 27 and 28, including codons encoding Y1226/7. Each founder was outcrossed to Tg(fli1a:egfp)^y1 adults and heterozygous carriers for flt4^C562Δ or flt4^Y1226/7Δ were confirmed by sequencing. All embryos were genotyped following all phenotypic analysis. Genotyping for C562Δ was performed by PCR using the following primers: 5′-tgtcggccgtgtggttattcag-3′ and 5′-agcatgaaactcactggttacgtag-3′, followed by digestion with MwoI (NEB), which cuts the mutant allele. Genotyping for Y1226/7Δ was carried out using two primer sets, one (5′-tcagattggcctgcacatgtc-3′ and 5′-ctggaggacgtaatcgacgtacatg-3′) recognizes Y1226/7Δ allele and the other (5′-atcagattggcctgcaacacac-3′ and 5′-ccggcttacatagctgagaaggctag-3′) recognizes wild-type allele. flt4^um131 genotyping was carried out using 5′-ttcagaacttataggcagctgaac-3′ and 5′-catttcctttggcttggaaa-3′ followed by Taqα1 (NEB), which digests wild type but not the mutant allele. Owing to the presence of a single nucleotide polymorphism at the flt4^um131 target site in some wild-type fish that eliminated the Taqα1 site, we also performed KASP genotyping assay (LGC Company) using the following primers: wild-type allele FAM primer (5′-atcttcataacagactctgaaacacac-3′), um131 allele HEX primer (5′-catcttcataacagactctgaaacacat-3′) and common primer (5′-gtcaggatctgaaaccagacatggta-3′). KASP assays were performed on a StepOnePlus Real-Time PCR machine (Life Technologies) using the primers above with 2× KASP Master Mix High ROX V4.0 (LGC Company).

A multisite Gateway destination cassette (attR4-ccdb-cam^r-attR3) followed by a polyadenylation sequence and flanked by Tol1 direct repeats (Koga et al., 2008) was synthesized in pUC57amp (Genewiz) and is referred to as pToneDest. A cDNA encoding full-length Tol1 transposase (Koga et al., 2008) and a polyadenylation sequence was synthesized in pUC57 downstream of a T7 promoter and is referred to as pToneTpase. 6x myc-tagged tdTomato with a nuclear localization signal (MPKKKRKV) was subcloned into pDONR221 or pDONRP2P3 using BP clonase II (Invitrogen) and are referred to as pME-myc-nlstdTomato and p3E-2a-myc-nlstdTomato, respectively. An activated myc-tagged form of MEK was generated from zebrafish map2k2b by mutating serines 211 and 225 to aspartic acid (Schramek et al., 1997) by PCR, followed by BP cloning to generate pME-myc-act-map2k2b. For expression constructs, p5E-uasE1b (Kwan et al., 2007), pME-myc-nlstdTomato, pME-myc-act-map2k2b or pME-mAkt (Covassin et al., 2009) and an empty p3E-mcs-1, p3E-2a-myc-nlstdTomato or p3E-2a-tdTomato were cloned into pToneDest using LR clonase II (Invitrogen) to give pTol1:uas-myc-nlstdTomato, pTol1:uas-myc-act-map2k2b-2a-tdTomato and pTol1:uas-mAkt-2a-myc-nlstdTomato. To test Flt4 autophosphorylation, cDNA encoding the Flt4 intracellular domain (amino acids 820-1354) was ligated into the SpeI site of pC₄M-F_v2E (Ariad) and subcloned into pDONR221 (pME-F_v2E-flt4-HA). This construct was used to generate F_v2E-flt4^Y1226/1227F by substituting tyrosine 1226 and 1227 with phenylalanine or by deleting amino acids 1220-1242 to generate F_v2E-flt4^Y1226/7Δ (pME-F_v2E-flt4^Y1226/1227F-HA and pME-F_v2E-flt4^Y1226/7Δ-HA, respectively). Each pME-F_v2E-flt4 plasmid, along with p3E-HA, was LR cloned into pCS-Dest2 (Villefranc et al., 2007) to generate pCS2-F_v2E-flt4-HA, pCS2-F_v2E-flt4^Y1226/1227F-HA and pCS2-F_v2E-flt4^Y1226/7Δ-HA. All plasmids are available at Addgene.org (https://www.addgene.org/browse/article/14538/).

Embryos derived from an incross of flt4^C562Δ/+;Tg(cdh5:gal4ff)^mu101 carriers were injected with 20 pg of pTol1:uas-myc-nlstdTomato, pTol1:uas-mAkt-2a-myc-nlstdTomato or pTol1:uas-myc-act-map2k2b-2a-tdTomato with 50 pg Tol1 transposase mRNA. Injected embryos were incubated in egg water with 1× phenylthiourea (PTU) until 38 hpf then fixed for immunostaining (see below). Tol1 transposase mRNA was synthesized using T7 polymerase (mMessageMachine, Ambion) following NotI digestion of pToneTpase.

To block phosphorylation and activation of ERK1/2, embryos were incubated in egg water with 1× PTU and 15 µM SL327 (EMD Millipore) from 28 hpf to 36 hpf or 49 hpf. To block phosphorylation of AKT1/2/3, we used 25 µM MK2206 (Santa Cruz). Lysates were prepared from embryos treated with 1, 10, 25 and 100 µM MK2206 or DMSO from 24 hpf to 28 hpf and separated by 10% SDS-PAGE, followed by western blotting. Blots were incubated with antibodies against phospho-AKT (Ser473) XP Rabbit monoclonal antibody (1:2000; #4060, Cell Signaling) and pan AKT Rabbit monoclonal antibody (1:5000; #4691, Cell Signaling). Phosphorylation level of AKT was measured using ImageJ.

pCS-myc-mCherry, pCS2-F_v2E-flt4-HA-2a-tdTomato, pCS2-F_v2E-flt4^Y1226/1227F-HA-2a-tdTomato and pCS2-F_v2E-flt4^Y1226/7Δ-HA-2a-tdTomato were transfected in HEK293T cells with DMEM/10% fetal bovine serum media on 10 cm dishes using Effectene (Qiagen). Cells were treated for 30 min with 500 nM AP20187 (Ariad) to induce F_v2E dimerization or ethanol as negative control, and lysed using RIPA buffer (150 mM NaCl, 50 mM Tris-HCl pH 8.0, 1% NP-40, 0.1% SDS) with protease and phosphatase inhibitors (Roche and Sigma, respectively). Lysates were immunoprecipitated using monoclonal anti-HA antibody (H3663, Sigma) and blotted with anti-phospho-Tyrosine (4G10, EMD Millipore), -HA, -panERK, -pERK, -panAKT or -pAKT antibody.

In situ hybridization was carried out using flt4 riboprobe as previously described (Siekmann and Lawson, 2007). GFP immunostaining was performed as previously described (Covassin et al., 2006). Immunostaining for Prox1, ERK and Flt4 was performed using a protocol described elsewhere (Nicenboim et al., 2015) with some modifications. Briefly, embryos were permeabilized sequentially with water and PBSTw (PBS containing 0.1% Tween 20)/1% Triton X-100. For antigen retrieval, embryos were incubated in citrate buffer (0.1 M sodium citrate dehydrate/0.05% Tween 20, pH 6.0) for 20 min at 98°C. The following antibodies and dilutions were used: 1:5000 rabbit anti-Flt4 polyclonal antibody (Villefranc et al., 2013), 1:1000 anti-Prox1 antibody (ab11941, Abcam) or 1:250 phospho-ERK1/2 (Thr202/Tyr204) XP rabbit monoclonal antibody (#4370, Cell Signaling) with 1:500 monoclonal anti-GFP antibody (sc-9996, Santa Cruz) or 1:500 monoclonal anti-Myc antibody (9E10, Sigma) in blocking buffer (PBSTw/0.1% Triton X-100/2% DMSO/5% bovine serum albumin/1% goat serum). GFP and Myc were detected with goat anti-mouse IgG(H+L)-Alexa488 (A11029, Thermo Fisher Scientific). For Tie2 staining, embryos were fixed with 4% paraformaldehyde in PBS (PFA/PBS) overnight at 4°C, quenched with 3% H₂O₂ in methanol and stored in methanol at −20°C until use. Embryos were treated with 30 µg/ml proteinaseK (NEB) in PBSTw (PBS/0.1% Tween 20) for 1 min and fixed in 4% PFA/PBS. After blocking with PBSTw/2% blocking reagent (Roche), embryos were incubated with 1:100 goat anti-zTie2 antibody (AF928, R&D Systems) overnight at 4°C, washed with PBSTw, and incubated with 1:5000 ZyMAX rabbit anti-goat IgG(H+L)-HRP (81-1620, Thermo Fisher Scientific) overnight at 4°C. Embryos were washed with PBSTw, then PBS, and incubated with 1:50 tyramide-Cy3 in 1× amplification diluent (PerkinElmer) for detection. For simultaneous detection of Prox1 and pERK1/2, embryos were immunostained with Prox1 antibody followed by incubation in citrate buffer for 20 min at 98°C to denature the primary and secondary antibodies. After blocking, embryos were incubated in pERK1/2 (Thr202/Tyr204) XP rabbit monoclonal antibody overnight at 4°C, then with goat anti-rabbit IgG(H+L)-HRP, and signals was detected with tyramide-FITC (PerkinElmer).

Embryos were placed in 1.5% low melt agarose on a 35 mm glass-bottomed culture dish filled with egg water. For live embryos, 1× tricaine was included. Images were acquired using a Zeiss LSM710 equipped with a W Plan-APOCHROMAT 20×/1.0 DIC D=0.17 M27 70 mm objective lens. GFP, Alexa488 or FITC, and Cy3 were simultaneously excited at 488 nm and 561 nm of confocal lasers, respectively. Emissions were split by a MBS 488/561/633 beam splitter and captured with two detection ranges (ch1: 493-536 nm, ch2: 576-685 nm). PHBC formation was imaged using a two-photon Chameleon Ti:Sapphire pulse laser (Coherent) at 900 nm through a 500-550 nm band pass filter into a BiG detector. Angiography using Qdot655 was performed as previously described (Siekmann and Lawson, 2007). Vertical projections of image stacks were assembled and quantified using ImageJ with Fiji or Imaris ver.7 (Bitplane). For depth colorization of the facial lymphatic stacks, we utilized the Temporal Color Code function in Fiji.

PHBC formation was scored in Tg(fli1a:egfp)^y1 embryos between 25 and 26 hpf on a M165FC microscope (Leica), or in two-photon images following GFP immunostaining. For quantification of intersegmental artery and vein connections at 72 hpf, image stacks were surface-rendered using Imaris and assessed for connection to the dorsal aorta or PCV between somites 9 and 16 on both sides of the embryo. The number of Prox1-positive endothelial cells between somites 8 and 16 was counted in embryos derived from at least three clutches of embryos. For rescue experiments, Prox1- and Myc-positive cells between somites 4 and 16 were scored in angiogenic sprouts and the PCV. All embryos were subjected to genotyping following phenotypic analysis. Student's t-test was used to calculate significance for quantification of cell numbers. Fisher's exact test was used to determine significance when comparing the degree of penetrance of observed phenotypes.

We are grateful to Arndt Siekmann for providing the cdh5:gal4ff line. We thank Phil Crosier and John Astin for the kind gift of the lyve1b:dsRed transgenic line. We thank members of the Lawson Lab for technical and intellectual input and the UMass Med Aquatics staff for excellent fish care.