The VEGFC/VEGFR3 signaling pathway is essential for lymphangiogenesis (the formation of lymphatic vessels from pre-existing vasculature) during embryonic development, tissue regeneration and tumor progression. The recently identified secreted protein CCBE1 is indispensible for lymphangiogenesis during development. The role of CCBE1 orthologs is highly conserved in zebrafish, mice and humans with mutations in CCBE1 causing generalized lymphatic dysplasia and lymphedema (Hennekam syndrome). To date, the mechanism by which CCBE1 acts remains unknown. Here, we find that ccbe1 genetically interacts with both vegfc and vegfr3 in zebrafish. In the embryo, phenotypes driven by increased Vegfc are suppressed in the absence of Ccbe1, and Vegfc-driven sprouting is enhanced by local Ccbe1 overexpression. Moreover, Vegfc- and Vegfr3-dependent Erk signaling is impaired in the absence of Ccbe1. Finally, CCBE1 is capable of upregulating the levels of fully processed, mature VEGFC in vitro and the overexpression of mature VEGFC rescues ccbe1 loss-of-function phenotypes in zebrafish. Taken together, these data identify Ccbe1 as a crucial component of the Vegfc/Vegfr3 pathway in the embryo.
The lymphatic vasculature develops from pre-existing vessels in a dynamic process called lymphangiogenesis, and is necessary to preserve tissue fluid homeostasis, for fat absorption and normal immune function. The lymphatic vascular network originates in the mouse embryo from the cardinal vein, where lymphatic endothelial precursor cells first bud and migrate dorsolaterally away from the vein under the control of Vegfc and its receptor Vegfr3 (Flt4 - Mouse Genome Informatics). In mice, the overexpression of Vegfc in the skin results in dramatic hyperplasia of lymphatic vessels (Jeltsch et al., 1997), whereas Vegfc knockout mice develop lymphatic hypoplasia and lymphedema (Karkkainen et al., 2004). In developing Vegfc knockout embryos, endothelial cells are still specified to the lymphatic lineage, and express Prox1, but fail to migrate from the cardinal vein (Karkkainen et al., 2004; Hägerling et al., 2013). Reduced development, or impaired function of lymphatics, leads to tissue fluid accumulation and lymphedema in humans, which can be inherited as a result of mutations in key developmental genes (Karkkainen et al., 2000; Irrthum et al., 2003; Alders et al., 2009; Connell et al., 2010). A mutation in VEGFC has recently been shown to be responsible for inherited lymphedema (Gordon et al., 2013).
Vegfr3-deficient mice die early during gestation (embryonic day 9.5) owing to cardiovascular failure (Dumont et al., 1998), consistent with the expression of Vegfr3 in blood vascular endothelium in the early embryo. However, after midgestation, Vegfr3 expression becomes enriched in the developing lymphatic vasculature (Kaipainen et al., 1995) and Vegfr3 signaling is sufficient to initiate lymphangiogenesis (Veikkola et al., 2001). Importantly, patients with mutations in VEGFR3 (FLT4 - Human Gene Nomenclature Database) develop Milroy’s disease, characterized by reduced lymphatic drainage and lymphedema in the lower limbs (Karkkainen et al., 2000). These studies, and many more, have shown that the precise modulation of the Vegfc/Vegfr3 signaling pathway is crucial during lymphatic vascular development.
In zebrafish, lymphatic vascular development initiates from 32 hours post-fertilization (hpf), when precursor cells first migrate dorsally from the cardinal vein to colonize the horizontal myoseptum, generating a population of parachordal lymphangioblasts (PLs) (Küchler et al., 2006; Yaniv et al., 2006; Hogan et al., 2009a; Isogai et al., 2009). These PLs subsequently migrate alongside arteries, both dorsally and ventrally (from ∼60 hpf), and remodel into the major trunk lymphatic vessels (Bussmann et al., 2010; Cha et al., 2012), forming the thoracic duct, intersegmental lymphatic vessels and dorsal longitudinal lymphatic vessels (for reviews, see Koltowska et al., 2013; van Impel and Schulte-Merker, 2014). In addition, a complex craniofacial lymphatic network is formed (Okuda et al., 2012). Zebrafish vegfr3 (flt4 - Zebrafish Information Network) and vegfc function is essential for all secondary (venous) angiogenesis, including the sprouting of lymphatic precursors from the cardinal vein and subsequent formation of PLs (Isogai et al., 2003; Küchler et al., 2006; Yaniv et al., 2006; Hogan et al., 2009a).
We have previously reported two zebrafish mutants with an absence of lymphatic (but not blood vascular) development (Hogan et al., 2009a; Hogan et al., 2009b). These phenotypes were caused by mutations in vegfr3 and the previously uncharacterized gene collagen and calcium-binding EGF domains-1 (ccbe1) (Hogan et al., 2009a; Hogan et al., 2009b). ccbe1 was shown to act at stages identical to vegfr3 and vegfc (Hogan et al., 2009a). ccbe1 encodes an extracellular matrix (ECM) protein, and is composed of N-terminal calcium-binding epidermal growth factor (EGF)-like and EGF domains, and two collagen-repeat domains towards the C terminus of the protein. The protein localizes to secretory vesicles (Alders et al., 2009), and is secreted to the ECM, where it binds proteins such as collagens or vitronectin (Bos et al., 2011). Consistently, Ccbe1 functions non-cell-autonomously in zebrafish (Hogan et al., 2009a). There are no described receptors or pathways known to interact with Ccbe1, with progress limited at least in part by the fact that full-length CCBE1 protein is insoluble in many standard assays (S.S.-M. and B.M.H., unpublished observations).
Importantly, mice deficient for Ccbe1 lack lymphatic vasculature (Bos et al., 2011) and mutations in human CCBE1 are causative for generalized lymphedema and lymphangiectasia in Hennekam syndrome (Alders et al., 2009; Connell et al., 2010). Interestingly, treatment with Ccbe1 protein (a truncated soluble form) increased Vegfc-induced lymphangiogenesis in a cornea micropocket assay (Bos et al., 2011). Furthermore, mice heterozygous for both Vegfc and Ccbe1 show a genetic interaction in lymphatic development (Hägerling et al., 2013). These results were suggestive that Ccbe1 and Vegfc can function together in lymphangiogenesis, but mechanistic insights into the molecular function of Ccbe1 have remained elusive.
Here, we have investigated the relationship between Ccbe1 and the Vegfc/Vegfr3 signaling pathways. Using genetic epistasis and analysis of signaling events in vivo, we show that the embryo requires Ccbe1 for normal Vegfc/Vegfr3/Erk signaling. CCBE1 is capable of increasing the levels of mature, processed VEGFC in vitro, which can rescue ccbe1 loss-of-function phenotypes when overexpressed in vivo. Together, our findings suggest that Ccbe1 activates Vegfc through its processing and release from the ECM in order to regulate lymphangiogenesis and venous sprouting in the embryo. These findings suggest common mechanisms in development and vascular pathologies (e.g. lymphedema) and further suggest Ccbe1 as a new therapeutic entry point in treating pathological lymphangiogenesis.
Identification of four zebrafish vegfc mutant alleles
In a forward genetic screen for zebrafish lymphatic vascular mutants, we identified several mutants that were indistinguishable from vegfr3 and ccbe1 mutants at the level of gross embryo morphology (supplementary material Fig. S1A-D), and that displayed a severe reduction in secondary angiogenesis and block in lymphatic precursor sprouting from the cardinal vein (Fig. 1A-D; Fig. 2A-H; supplementary material Fig. S1E-K). At 5 days post-fertilization (dpf), the hu5055, hu6124, hu5142 and hu6410 mutants have a grossly normal, functional blood vasculature, but the lymphatic vasculature is absent (Fig. 1B,D). Using a positional cloning approach, we found that these mutants were linked to chromosome 1, containing the vegfc gene (Fig. 1E; data not shown). Sequencing of vegfc in hu5055 mutants revealed a missense mutation changing a cysteine into an arginine residue (C365R) in the highly conserved C-terminal propeptide (Fig. 1F,G). Additional mutations identified caused a premature stop in hu6410 (L107X), a phenylalanine to cysteine change in the highly conserved VHD-VEGF homology domain (F138C) and another cysteine-to-serine substitution in the C-terminal propeptide of Vegfc (C339S) (Fig. 1F,G). All mutations were predicted to be damaging by PolyPhen-2 (Adzhubei et al., 2010). These vegfc mutations segregate with the thoracic duct (TD) deficiency phenotype and phenocopy MO-vegfc knockdown or soluble Vegfr3 ligand trap-induced phenotypes (Küchler et al., 2006; Yaniv et al., 2006; Hogan et al., 2009a; Hogan et al., 2009b). Furthermore, these phenotypes are consistent with recently described phenotypes in a Vegfc truncation mutant (Villefranc et al., 2013).
Interestingly, scoring both TD extent and PL number, we noted variable penetrance of different alleles and haplo-insufficient phenotypes for the more penetrant alleles (Fig. 1I,J). The hu5055 allele segregated in a Mendelian, autosomal recessive manner and was thus used in subsequent assays. Crossing hu5055 mutant carriers to the previously described expando/vegfr3 mutant (Hogan et al., 2009b), we found an increased frequency of TD defects consistent with genetic interaction in the same pathway (Fig. 1K; Fig. 2I).
ccbe1, vegfr3 and vegfc zebrafish mutants genetically interact
The ccbe1hu3613, vegfr3hu4602 and vegfchu5055 mutants display a complete loss of the lymphatic vasculature when homozygous (Fig. 2A-H). Previous, non-quantitative observations (Hogan et al., 2009a; Hogan et al., 2009b) indicated that partial loss of ccbe1 or vegfr3 led to phenotypically wild-type lymphatic development. We took advantage of these mutants to determine more rigorously if combined heterozygous mutations led to enhanced lymphatic phenotypes. We crossed heterozygous carriers and quantified the extent of the TD across ten body segments in the trunks of individual embryos. Subsequent genotyping first confirmed the genetic interaction in vegfchu5055/+, vegfr3hu4602/+ double heterozygous animals (Fig. 2I). We found that ∼28% of scored embryos displayed ≤50% TD extent; within this population, 71% of the embryos were double heterozygotes for vegfc and vegfr3, a significant enrichment from total double heterozygosity of 27% (Fig. 2I). Importantly, similar robust interactions of vegfr3 and vegfc were observed with ccbe1. The population of embryos displaying ≤50% TD extent from a ccbe1hu3613/+, vegfr3hu4602/+ cross was enriched (54%) for double heterozygotes, as were embryos from a ccbe1hu3613/+, vegfchu5055/+ cross (61%) (Fig. 2J,K). This genetic interaction seems to be selective to the ccbe1hu3613, vegfr3hu4602 and vegfchu5055 mutants because in crosses to two other, as yet genetically uncharacterized lymphatic mutants, we did not observe any interaction (L.L.G. and B.M.H., unpublished observations). Finally, we generated triple heterozygote embryos and found that the population developing ≤50% TD extent contained most of the triple heterozygote embryos, demonstrating a strong interaction (Fig. 2L). This synergistic genetic interaction shows that lymphatic development is sensitive to the level of activation of the Vegfc/Vegfr3 pathway, and that altering ccbe1 dosage modifies phenotypes in this pathway. Combined with the fact that ccbe1, vegfc and vegfr3 act at the same stage during the cellular progression of lymphangiogenesis and that these mutants present the most selective loss of lymphatic vascular structures that we have observed, we considered this a strong indication that Ccbe1 could be a novel component of the Vegfc/Vegfr3 signaling pathway.
ccbe1 mutation suppresses phenotypes driven by ectopic Vegfc/Vegfr3 signaling in embryonic arteries
Previous studies have shown that the knockdown of dll4 leads to an excessive intersegmental artery (aISV) angiogenesis phenotype by 72 hpf (Leslie et al., 2007; Siekmann and Lawson, 2007; Hogan et al., 2009b). This phenotype occurs because in wild-type arteries, the Vegfc/Vegfr3 pathway is inhibited by Dll4. Depletion of Dll4 leads to increased activation of Vegfc/Vegfr3 signaling in aISVs, resulting in aISV hyperbranching (Leslie et al., 2007; Siekmann and Lawson, 2007; Hogan et al., 2009b). We took advantage of the hyperbranching phenotype to investigate ccbe1 function. In clutches of embryos derived from incrosses of ccbe1 heterozygous carriers, we activated the Vegfc/Vegfr3 pathway in arteries by knocking down dll4. Embryos were first sorted at 72 hpf according to the severity of arterial hyperbranching, and then subsequently genotyped. The population containing animals with vastly reduced or no hyperbranching (Fig. 3Aiii, graph) was significantly enriched in ccbe1hu3613 mutants (81%; P<0.0001) whereas the category containing severe aISV hyperbranching (Fig. 3Aii) was enriched in wild-type and heterozygote embryos (83%; P=0.0019; Fig. 3A, graph). Hence, ccbe1 mutation suppressed the dll4 loss-of-function phenotype.
To validate this observation further, we used a second phenotypic assay that has been previously described (Hogan et al., 2009b). When vegfc is overexpressed by mRNA injection and dll4 is simultaneously depleted, we observe ectopic turning of aISVs in the trunk from as early as 30 hpf (compare Fig. 3Bi with 3Bii) that is not seen in single treatment controls (supplementary material Fig. S2). aISVs deficient for dll4 are more responsive to vegfc RNA injection (Hogan et al., 2009b) and phenotypes appear the same as those observed when vegfc is expressed from the hsp70l promoter (Nicoli et al., 2012). In clutches of embryos derived from ccbe1 heterozygous incrosses, we ectopically activated the Vegfc/Vegfr3 pathway by injecting vegfc mRNA and a dll4 morpholino. Animals were sorted at 30 hpf according to the severity of aISV phenotype, and then genotyped. Of the embryos displaying a weak phenotype, or no phenotype at all (Fig. 3Biii), 70% were ccbe1 mutants (Fig. 3B, graph), a highly significant (P<0.0001) enrichment. The population displaying severe phenotypes was significantly enriched in wild-type and heterozygote animals (83%; P=0.0006; Fig. 3B, graph). Hence, ccbe1 mutation suppressed the dll4 loss-of-function, vegfc overexpression phenotype.
Ccbe1 knockdown suppresses phenotypes driven by ectopic Vegfc/Vegfr3 signaling in embryonic veins
We next generated a new zebrafish transgenic line (hsp70l:GAL4;UAS:vegfc), which ubiquitously expresses full-length vegfc under the control of the heat shock (hsp70l) promoter (supplementary material Fig. S3A,B) (Scheer et al., 2001). In these embryos (following two consecutive heat shocks at 28 and 56 hpf), veins sprout ectopically, resulting in hyperbranched intersegmental vessels (vISVs) as visualized in fli1a:GFP, flt1:tomato double transgenic animals (Fig. 3Ciii). ccbe1 morpholino injection rescued this phenotype, blocking excessive venous sprouting (Fig. 3Civ, graph; P<0.0001).
Taken together, we found in two blind, genotype-based experiments, that Vegfc- and Vegfr3-driven arterial phenotypes are suppressed in ccbe1hu3613 mutants, and additionally that a Vegfc- and Vegfr3-driven venous phenotype is rescued in ccbe1 morphant embryos. These findings suggest that Ccbe1 is necessary for the function of the Vegfc/Vegfr3 signaling pathway during embryonic development.
Vegfr3-dependent Erk activation in embryonic veins requires ccbe1
It is well established that VEGFR3 signals via intracellular kinases that include ERK (reviewed by Bahram and Claesson-Welsh, 2010). We decided to further investigate downstream signaling pathways to determine if Ccbe1-deficient embryos display a signaling block in venous endothelium. We used whole-mount immunofluorescence to examine phospho-Erk in the context of the embryonic vasculature. The signal observed was specific as it was reduced in embryos treated with the MEK inhibitor PD98059 (which inhibits Mek activation and hence Erk phosphorylation) and was ectopically induced in the ventral posterior cardinal vein (PCV) in the context of vegfc overexpression (Fig. 4A). In wild-type embryos, we found that Erk activity was broadly detected in the neural tube, muscle and epithelia, but also in the endothelium of the PCV. In PCV endothelium, Erk activation was segmented and dorsally enriched in individual cells at 32 hpf, concomitant with the induction of secondary sprouting. Importantly, this patterned activation of Erk was markedly reduced in both MO-vegfr3- and MO-ccbe1-injected embryos compared with wild-type controls (Fig. 4B,C). These findings confirm the presence of Vegfc- and Vegfr3-dependent Erk signaling in embryonic veins and demonstrate that Ccbe1 is necessary for activation of this pathway.
We also examined the activation status of Vegfr3 by western blot of embryonic lysates. The phosphorylation of Tyr1063/1068 in human VEGFR3 reflects the activation status of VEGFR3, promoting downstream signaling essential for lymphangiogenesis (Dixelius et al., 2003; Salameh et al., 2005). Antibodies against zebrafish Vegfr3 are unavailable, but we found two commercial antibodies directed against conserved regions of human VEGFR3 that cross-reacted with zebrafish Vegfr3. One of these was directed against the phosphorylated residues Tyr1063/1068 (human) or Tyr1071/1076 in zebrafish (supplementary material Fig. S4). Using this phospho-VEGFR3 antibody for western blotting we detected Vegfr3 in zebrafish lysates immunoprecipitated with the same antibody at 32 hpf, during initiation of secondary sprouting. The signal was increased in vegfc mRNA-injected embryos, but was reduced in MO-vegfr3- and MO-vegfc-injected embryos. A similar reduction in phospho-Vegfr3 was observed in MO-ccbe1-injected embryos quantified and normalized over three independent experiments (Fig. 4D). We were not able to precipitate Vegfr3 using another cross-reacting antibody directed against VEGFR3 intracellular domains. However, we were able to validate the specificity of the bands and result observed by blotting lysates immunoprecipitated by the phospho-Vegfr3 antibody with the VEGFR3 antibody directed against Vegfr3 intracellular domains. We detected the bands of the same molecular weight and intensities (supplementary material Fig. S5). Importantly, Ccbe1, Vegfc and Vegfr3 do not cross-regulate each other transcriptionally (Fig. 4E), indicating that Vegfr3 expression is unchanged in this assay. This is consistent with immunofluorescence-based observations in the Ccbe1 knockout mouse (Hägerling et al., 2013). We also detected normal expression of all other signaling Vegfr family members in the morphants examined (Fig. 4E).
Ccbe1 enhances Vegfc-driven sprouting
To characterize further the function of Ccbe1 in Vegfc/Vegfr3 signaling, we generated two transgenic lines that express ccbe1 or vegfc from the shh promoter in the floor plate (FP) from as early as 24 hpf (supplementary material Fig. S3C,D). At 32 hpf, vegfc- or ccbe1-overexpressing animals display no phenotype. However, when these two lines were crossed, we found that double transgenic animals display aberrant ectopic turning of ISVs at 32 hpf (Fig. 5A).
At 48 hpf, ccbe1-overexpressing animals still show no phenotype, whereas vegfc-overexpressing animals display vastly increased endothelial cell accumulation at the horizontal myoseptum, as well as dramatic ectopic sprouting of the ISVs (Fig. 5A), as also observed in the Tg(hsp70l:Gal4;UAS:vegfc) line (Fig. 3Cii). Interestingly, at 48 hpf in double ccbe1/vegfc-overexpressing animals, we observe a dramatic accumulation of endothelial cells in dorsal aspects of the embryo (Fig. 5A), combined with ectopic ISV sprouting. The accumulation of endothelial cells (arrow) occurs at the approximate level of the FP in double Tg(shh:ccbe1;shh:vegfc), but not in single Tg(shh:vegfc) or Tg(shh:ccbe1) animals. This could be taken to indicate that Ccbe1 expression is capable of locally enhancing Vegfc-driven sprouting in the embryo.
We also examined the activity of the Tg(shh:vegfc) transgenic line in ccbe1hu3613 mutants. ccbe1hu3613 mutant embryos showed a suppression of the Vegfc-driven hyperbranching of ISVs, which was consistent with our findings that ccbe1 is needed for Vegfc-driven hyperbranching in the hsp70l:GAL4;UAS:vegfc line and dll4 depletion models. When these animals were scored at 3.5 dpf for endothelial cells at the horizontal myoseptum or 5 dpf for the presence of the TD, we found that overexpression of full-length Vegfc could partially rescue ccbe1hu3613 mutant lymphatic phenotypes (supplementary material Fig. S6A,B). However, the shh-driven vegfc expression did not rescue the MO-ccbe1 lymphatic phenotype and, furthermore, we did not see any rescue of lymphatic development in MO-ccbe1-injected, hsp70l:GAL4;UAS:vegfc embryos (supplementary material Fig. S6C-E). We take this to indicate that partial rescue exclusively in the ccbe1hu3613 mutant occurs as a result of some retained capability of the ccbe1hu3613 D162E (hypomorphic) mutant protein.
CCBE1 overexpression leads to enhanced levels of processed, soluble VEGFC in vitro
VEGFC is produced as a secreted precursor protein requiring proteolytic cleavage of N- and C-terminal propeptides to produce the highly bioactive mature form, consisting of dimers of the VEGF homology domain (VHD), which potently activates VEGFR3 signaling (Joukov et al., 1997). Although proteases that can promote processing of VEGFC have been identified (McColl et al., 2003; Siegfried et al., 2003), other factors that might regulate the function of VEGFC in the developing embryo remain to be described. To determine if the capability of CCBE1 to modulate VEGFC activity in vivo may be at the level of VEGFC production, release from the cell surface/ECM or proteolytic processing, we turned to in vitro cell-based models. 293EBNA-1 cells stably harboring an expression construct encoding full-length human VEGFC allow analysis of secreted VEGFC polypeptides in conditioned cell culture medium. These cells were transfected with an expression vector encoding human CCBE1 (Fig. 5B). In control (empty vector) or mock transfected cells lacking detectable CCBE1, mature VEGFC was readily detected in conditioned cell culture media, with partially processed forms detected at lower levels (Fig. 5C,C′). CCBE1 transfection resulted in a significant enrichment of all forms of VEGFC in the conditioned cell culture media but in particular the relative abundance of the mature form was increased compared with other forms of VEGFC (Fig. 5C,C′). This indicates that although release of VEGFC from the cell surface or ECM is enhanced, processing to the mature form is particularly increased in the presence of CCBE1.
We confirmed the enhanced processing of VEGFC in an independent cell line, HEK293T cells, using transient transfections of CCBE1 and VEGFC plasmids; note the increased abundance of mature VEGFC in the cell culture media in response to CCBE1 in supplementary material Fig. S7B,C. However, the release of VEGFC was not generally enhanced in this system as the abundance of other forms of VEGFC did not increase. In both cell systems, we also examined the levels of VEGFC present in the cell lysates and found that in 293EBNA-1 cells we could detect partially processed VEGFC (i.e. the form containing the N-terminal propeptide and VEGF homology domain) (supplementary material Fig. S7A) whereas in HEK293T cells we detected both full-length and partially processed forms (supplementary material Fig. S7C). The relative levels of different forms of VEGFC in cell lysates were not markedly altered in the presence of CCBE1 (supplementary material Fig. S7A,C). We take this to indicate that the cell-associated VEGFC may be far more abundant than VEGFC in the medium, hence a small decrease in the proportion of cell-associated VEGFC, due to release from the cell surface or ECM, could lead to a significant increase in the relative abundance of VEGFC detected in the medium on western blots. Importantly, CCBE1 overexpression had no effect on VEGFC mRNA levels in vitro or in vivo in zebrafish embryos (Fig. 5D).
In the in vitro assays above, VEGFC and CCBE1 were produced and secreted from the same cells. To test if CCBE1 needs to be expressed in the same cell as VEGFC to promote VEGFC processing, we transfected two different populations of HEK293T cells with CCBE1 and/or VEGFC and mixed them before detection of VEGFC in the culture medium. We found that CCBE1 was indeed capable of enhancing VEGFC processing in trans (supplementary material Fig. S7B), indicating that CCBE1 can exert its effects on VEGFC outside the cell. These data, taken together, indicate that in these cell-based systems CCBE1 regulates VEGFC at a post-transcriptional level, that this regulation occurs extracellularly and that it is capable of increasing the release and processing of VEGFC. The degree to which each of these two mechanisms apply may be dependent on the cell type, given that the effect on VEGFC release was not observed in the HEK293T model system.
Constitutively secreted, processed VEGFC rescues ccbe1 loss-of-function phenotypes
To build on these in vitro findings, and given the observation that full-length Vegfc overexpression cannot rescue the MO-ccbe1 loss-of-function phenotype (supplementary material Fig. S5), we decided to investigate whether a constitutively secreted, mature form of VEGFC was sufficient to recue ccbe1 loss of function. To do this, we generated a truncated form of human VEGFC (ΔNΔCVEGFC), missing the N- and C-terminal domains but retaining the signal peptide for secretion, and placed it under the control of the hsp70l promoter. We used the human form because the proteolytic processing sites of zebrafish Vegfc are not experimentally validated whereas the human form has been examined in detail and processing sites are known (Joukov et al., 1997). We injected DNA for hsp70l:ΔNΔCVEGFC.t2a.mCherry and generated mosaic embryos. After a 1-hour heatshock at 26 hpf, embryos injected with hsp70l:ΔNΔCVEGFC.t2a.mCherry displayed ISV hyperbranching (at 54 hpf, Fig. 6A) as observed previously upon vegfc overexpression (Fig. 3C; Fig. 5A) and expressed mosaic mCherry. Co-injection of the hsp70l:ΔNΔCVEGFC.t2a.mCherry with MO-ccbe1 generated a robust, quantifiable rescue of PL formation in these embryos compared with MO-ccbe1 controls (Fig. 6A,B). Co-injecting hsp70l:ΔNΔCVEGFC.t2a.mCherry with MO-vegfr3 did not rescue PL formation at the horizontal myoseptum by 54 hpf (Fig. 6A,B). Finally, we found that ISV hyperbranching was present in ccbe1 and vegfr3 morphants injected with hsp70l:ΔNΔCVEGFC.t2a.mCherry (Fig. 6C), which is consistent with the ability of processed VEGFC to activate VEGFR2 (KDR - Human Gene Nomenclature Database) signaling. The ability of a processed form of VEGFC to rescue the MO-ccbe1 loss-of-function phenotype demonstrates that a deficit in Vegfc maturation is responsible for venous angiogenesis and lymphangiogenesis defects in the absence of Ccbe1.
Our results show that ccbe1 genetically interacts with vegfc and vegfr3 and that ccbe1 mutation or depletion suppresses the formation of excess and abnormal aISV and vISV sprouts driven by ectopic Vegfc/Vegfr3 signaling. These data highlight the necessity for functional Ccbe1 for the propagation of (ectopic) Vegfc-driven signals in the developing embryo. Furthermore, we show that in ccbe1 and vegfr3 morphants, venous Erk signaling and Vegfr3 activation at Tyr1071/1076 are reduced. Our finding of a block in Vegfc/Vegfr3 signaling is consistent with phenotypes observed in humans (Alders et al., 2009) and mice (Bos et al., 2011). Previous studies have used proximity ligation assays as a readout for Vegfr3 signaling (Bos et al., 2011), and did not observe the reduction of Vegfr3 signaling in the absence of Ccbe1 that we see here. However, the previous study was unable to examine the entirety of signaling in cells at equivalent developmental stages during the induction of sprouting of lymphatics from the vein. We overcame previous limitations by utilizing the immunofluorescence visualization of activated (phosphorylated) Erk in the PCV. We found that during the induction of secondary sprouting Erk activation is dorsally enriched and segmentally patterned in the PCV in a Vegfr3-dependent manner. This assay hence provides improved sensitivity to detect Vegfc-Vegfr3 induced signaling during secondary angiogenesis in vivo compared with previous approaches. Using this approach and also cross-reacting antibodies to phospho-Vegfr3, we found a crucial requirement for Ccbe1 for in vivo Vegfc-Vegfr3-Erk signaling.
Given this function of Ccbe1, its non-autonomous role in zebrafish (Hogan et al., 2009a) and its extracellular localization (Bos et al., 2011), we asked if Ccbe1 can directly modulate the levels of bioactive Vegfc. We did this in cell-based models, and found that upon CCBE1 overexpression, the amount of mature bioactive VEGFC is increased in the culture medium. CCBE1 can induce higher levels of VEGFC in trans and hence performs this function outside of the cell in this in vitro context. Fully processed VEGFC was increased more than the partially processed and full-length forms in the medium, indicating that CCBE1 regulates processing. However, CCBE1 can increase levels of the partially processed and full-length forms to some degree, also suggesting enhanced release from the cell surface or matrix. Confirming that this capability has relevance in vivo, we observed a rescue of the ccbe1 loss-of-function phenotype when we reintroduced a secreted and fully processed form of VEGFC in zebrafish. This rescue, combined with the observation that full-length Vegfc failed to rescue the phenotype, demonstrates that Ccbe1 regulates Vegfc maturation and bioavailability. Ccbe1 does not appear to be a protease and the precise molecular mechanisms involved, including the roles of additional proteins in the CCBE1/VEGFC/VEGFR3 pathway, require further analysis.
One pertinent question is: why would the developing embryo need an additional factor to regulate Vegfc-driven activation of Vegfr3? It is crucial to note that vegfc in zebrafish is transcribed in the hypochord and dorsal aorta in the midline of the embryonic trunk. Despite this, secondary sprouts from the vein migrate past this transcriptional source of ligand on the dorsoventral and mediolateral axes of the embryo to the horizontal myoseptum and give rise to PLs. This suggests that spatial presentation of active Vegfc protein during secondary sprouting must be regulated independently of the transcriptional source of vegfc. ccbe1 transcript expression precedes and predicts the migration route of lymphatic precursors in the trunk (Hogan et al., 2009a) suggesting that Ccbe1 serves to impart the necessary post-translational activation of Vegfc (see proposed model in Fig. 7) in a spatially and temporally regulated manner. Hence, a migrating lymphatic endothelial cell could be directed through the embryonic environment by a route pre-determined by the earlier, local concentration of Ccbe1. To test this idea directly is inherently difficult, but the hypothesis is supported by overexpression regimes of ccbe1 and vegfc in an ectopic location, the FP of the neural tube: here we found that Vegfc-driven sprouting of endothelial cells was enhanced in the presence of ectopic Ccbe1.
CCBE1 mutations have been shown to be responsible for generalized lymph vessel dysplasia in humans (Alders et al., 2009; Connell et al., 2010). Our finding that Ccbe1 is a crucial modulator of the Vegfc/Vegfr3 pathway during embryo development, suggests that a deficiency of VEGFR3 signaling may be responsible for the lymphatic aspects of Hennekam Syndrome. Other aspects of this syndrome that are distinct from Milroy’s disease might be due to other, yet to be identified, functions of CCBE1.
MATERIALS AND METHODS
Zebrafish strains and transgenesis
Animal work followed guidelines of the animal ethics committees at the University of Queensland, and the Royal Netherlands Academy of Arts and Sciences (DEC). Zebrafish transgenic lines Tg(fli1a:EGFP), Tg(-6.5kdr-l:mCherry), Tg(-0.8flt1:tdTomato) and Tg(hsp70l:Gal4)1.5kca4 lines were described previously (Scheer et al., 2001; Lawson and Weinstein, 2002; Hogan et al., 2009a; Bussmann et al., 2010). N-ethyl-N-nitrosourea (ENU) mutagenesis was performed as previously described (Wienholds et al., 2002).
The Tg(4xUAS:vegfc) line was generated using a 4xUAS promoter upstream of the full length zebrafish vegfc cDNA (supplementary material Fig. S3A) cloned using the Gateway system (Hartley et al., 2000; Walhout et al., 2000). The Tg(flt4:YFP) reporter line was generated from BAC DKEY-58G10 as previously described (Bussmann and Schulte-Merker, 2011) and is characterized in detail elsewhere (van Impel et al., 2014). The Tg(s1173:Gal4) line was kindly provided by Ethan Scott (School of Biomedical Sciences, University of Queensland). sonic hedgehog (shh) promoter and floorplate enhancer (Ertzer et al., 2007) was used to drive Ccbe1 IRES tagRFP or Vegfc IRES mturquoise (supplementary material Fig. S3C), cloned into the MiniTol2 vector (Balciunas et al., 2006); plasmids (25 ng/μl) were injected with tol2 transposase mRNA (25 ng/μl) into 1- to 2-cell-stage embryos. All genotyping details and primers are given in supplementary material Tables S1 and S3. KASPar (KBioscience) was used for the indicated vegfc alleles.
DNA, RNA and morpholino injections
Constructs for transient DNA injection of the Tg(hsp70l:ΔΔVEGFC.t2a.mCherry) transgene were generated by PCR amplifying the N- and C-terminally truncated form of human VEGFC, containing the endogenous secretion peptide (Joukov et al., 1997) followed by Gateway recombineering. Primer sequences are given in supplementary material Table S2. DNA was injected at a concentration of 90 ng/μl at single-cell stage. Morpholinos used are shown in supplementary material Table S4. The vegfc cDNA used was described previously (Hogan et al., 2009a).
Analysis of ERK and Vegfr3 signaling in zebrafish embryos
Phospho-ERK was detected by immunofluorescence, using the previously described protocol (Inoue and Wittbrodt, 2011) but modified as follows: embryos were blocked and incubated with p-ERK primary antibody (1/250; Cell Signaling, #4370) in TBS containing 0.1% Triton X-100, 1% bovine serum albumin, 10% goat serum and then with horseradish peroxidase-conjugated anti-goat secondary antibody (1/1000; Cell Signaling, #7074) in 2% blocking solution (Roche) dissolved in 100 mM maleic acid with 150 mM NaCl (pH 7.4). Phospho-ERK was subsequently detected using the TSA reagent (Perkin Elmer) as per manufacturer’s guidelines. As a control for antibody specificity, embryos were incubated in 100 μM PD98059 (Cell Signaling) from 24 hpf. Embryos were vibratome sectioned (100 μm) (Leica VT100S) and imaged as described below.
Zebrafish phospho-Vegfr3 was detected following immunoprecipitation (IP) with anti phospho-VEGFR3 (Cell Applications, CB5793) using the same antibody. One hundred and fifty embryos were lysed overnight at 4°C in modified RIPA buffer containing 1 mM EDTA, 4% protease inhibitor (Sigma), 1% phenylmethylsulfonyl fluoride (PMSF; Sigma) and 1% Halt phosphatase inhibitor cocktail (Thermo Fisher Scientific). IP was carried out overnight at 4°C with the phospho-Vegfr3 antibody (1/1000), followed by 2 hours at 4°C with protein A agarose beads (1/10) (Thermo Fisher Scientific). Standard western blotting approaches were used to detect Vegfr3 with both anti-phospho-VEGFR3 (1/1000; Cell Applications, CB5793) and anti (non-phosphorylated) VEGFR3 (1/1000; Cell Applications, CB5792). Mouse anti-myosin light chain 1 and 3f (1/100; Developmental Studies Hybridoma Bank, F310) was used to detect a loading control band. Quantification relative to IgG light chain (phospho-Vegfr3 only) or Myosin used ImageJ across three biological replicates. Only phospho-Vegfr3 was assessed and not total Vegfr3, owing to a limitation of the cross-reacting antibodies in only detecting bands post-IP for phospho-Vegfr3.
VEGFC processing/secretion in vitro
For analysis of VEGFC processing and secretion two approaches were used. Briefly, using 293EBNA-1 cells stably expressing a full-length form of human VEGFC (Karnezis et al., 2012) were transiently transfected [Lipofectamine 2000 (Invitrogen)] with a CCBE1 expression vector after a medium change to a serum-free chemically defined medium (Pro293a-CDM, Lonza). Both supernatant and cell lysates were subsequently collected for western blotting. 293EBNA-1 cells were maintained in Dulbecco’s Modified Eagle Medium, supplemented with 10% fetal bovine serum, 50 mM L-glutamine, 50 μg/ml penicillin, 50 μg/ml streptomycin and 100 μg/ml hygromycin B (Roche Diagnostics). Cells were lysed in 0.1% SDS, 50 mM Tris, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, pH 8.0, 1 mM PMSF, 10 μg/ml aprotinin and 10 μg/ml leupeptin. Lysates were incubated at 4°C for 20 minutes with gentle agitation and centrifuged at 13,000 g for 10 minutes. Western blotting was performed with antibodies targeting VEGFC (R&D Systems, BAF752) or CCBE1 (Abcam, ab101967).
For cell mixing experiments, separate populations of HEK293T cells were transfected with constructs expressing CCBE1, VEGFC or both. Cells were passaged and mixed as indicated after 24 hours. From mixed cultures, conditioned supernatants were collected and cells were lysed in RIPA/SDS buffer after another 48 hours. Supernatant and cell lysates were isolated and western blotting analysis performed using VEGFC (VEGFC isoform 103 antibody, antibodies-online) and CCBE1 (HPA041374, Sigma) antibodies.
qPCR detection of Vegfc and Ccbe1 expression
For expression analysis in the cultured cells, RNA was isolated using RNeasy Mini Kit (Qiagen), and cDNA synthesized using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). For expression analysis, Taqman Gene Expression Assays were employed [VEGFC (Hs01099203_m1), GAPDH (Mm99999915_g1)] using Taqman Fast Universal PCR Master Mix (2×) (Applied Biosystems) as per manufacturer’s instructions and analyzed on a 7300 RT-PCR machine (Applied Biosystems). For zebrafish qPCR, RNA was extracted at 24 hpf from 20 zebrafish embryos. cDNA was synthesized using the Superscript III Kit (Invitrogen) and reactions used SYBR Green kit (Applied Biosystems) analyzed on a 7500 RT-PCR machine (Applied Biosystems). Data were normalized using the geometric average of ef1α (eef1a1l1), rpl13 and rps29, which were found to be the most stable reference genes using GeNorm with data displayed as arbitrary units (A.U.).
Confocal imaging was performed on live embryos mounted laterally using a Zeiss 510 or a Leica SPE confocal microscope at the indicated stages. Sections were imaged using a Zeiss 710 FCS confocal microscope.
Statistical analysis and mutation effect prediction
For genetic interaction data, χ2 tests were performed using GraphPad (http://graphpad.com/quickcalcs/chisquared1.cfm). PolyPhen-2 was used to evaluate mutation pathogenicity (http://genetics.bwh.harvard.edu/pph2/).
We thank C. Neyt, G. van de Hoek, N. Chrispijn and M. Witte for technical assistance; Dagmar Wilhelm for advice; Nathan Lawson for providing an initial P-Erk immunofluorescence protocol; and Ethan Scott for providing the Tg(s1173:Gal4) line. Imaging was performed in the Australian Cancer Research Foundation’s Dynamic Imaging Facility at IMB and at the Hubrecht Imaging Center (HIC).
L.L.G. designed and performed experiments, analyzed data and co-wrote the manuscript; T.K., D.S., N.C.H., K.K., G.R., N.I.B. and A.v.I. designed and performed experiments, analyzed data and edited the manuscript; S.A.S. and M.G.A. designed experiments, analyzed data and edited the manuscript; S.S.-M. and B.M.H. designed experiments, analyzed data and co-wrote the manuscript.
This work was supported by an Australian Research Council Future Fellowship [FT100100165 to B.M.H.]; a European Molecular Biology Organization Long-Term Fellowship [LTRF 52-2007 to T.K.]; Veni grants from The Netherlands Organisation for Scientific Research (NWO) [916.10.132 to T.K.; 863.11.022 to G.R.]; Marie Curie Intra-European Fellowships (to D.S. and A.v.I.); National Health and Medical Research Council of Australia (NHMRC) project grants [631657 and APP1050138]; a Program Grant [487900 to M.G.A. and S.A.S.] and Research Fellowships (to M.G.A. and S.A.S.) from the National Health and Medical Research Council; and KNAW (S.S.-M.).
The authors declare no competing financial interests.