von Hippel-Lindau tumor suppressor mutants faithfully model pathological hypoxia-driven angiogenesis and vascular retinopathies in zebrafish

Blood vessel formation is a dynamic process that is tightly regulated to ensure oxygen homeostasis in vertebrates. During embryogenesis, this involves the de novo formation of blood vessels by differentiation of endothelial precursors, or angioblasts, to form the primary vascular network (vasculogenesis), and the formation of new blood vessels that branch off from existing vessels (angiogenesis or neovascularization) (Carmeliet, 2000). Adult blood vessels are normally quiescent unless the balance of angiogenic inhibitors is tipped in favor of angiogenic activators (termed ‘angiogenic switch’) during normal physiological processes, such as in wound healing or during the female reproductive cycle (Carmeliet, 2000). Angiogenesis is also crucial for the pathological progression of tumor growth and angiogenesis-related disorders, including many retinopathies that lead to vision loss (Arjamaa and Nikinmaa, 2006), such as age-related macular degeneration and diabetic retinopathies.

The hypoxia-induced transcription factor (HIF) is perceived as the master transcriptional regulator of hypoxia-inducible genes that are involved in adaptive cellular responses to oxygen (Kaelin, 2005), including those genes that are involved in angiogenesis (reviewed in Pugh and Ratcliffe, 2003; Liao and Johnson, 2007; Dewhirst et al., 2008). HIF is a heterodimer composed of a hypoxia-regulated alpha subunit (HIF-1α, HIF-2α or HIF-3α), and an oxygen-insensitive HIF-1β subunit (also known as the aryl hydrogencarbon receptor nuclear translocator, or ARNT) (Semenza, 2001). Under normoxic conditions, HIF-α is hydroxylated on either of two prolyl residues by the family of prolyl hydroxylases (PHD1-3), and targeted for rapid proteasomal degradation by a multi-subunit E3 ubiquitin ligase complex, containing the von Hippel-Lindau (VHL) protein as the substrate recognition component. In the absence of oxygen or functional pVHL (the VHL gene product), HIF-α protein is stabilized and translocated to the nucleus where it partners with HIF-β to form functional HIF, triggering an increased expression of multiple angiogenic growth factors (Pugh and Ratcliffe, 2003), including vascular endothelial growth factor (VEGF, or VEGFA) (Liu et al., 1995; Forsythe et al., 1996), plasminogen activator inhibitor-1 (PAI-1) (Kietzmann et al., 1999), angiopoietin-2 (ANG-2) (Oh et al., 1999), platelet-derived growth factor (PDGF) (Gleadle et al., 1995), stromal-derived factor-1 (SDF-1, or CXCL12) and CXC chemokine receptor 4 (CXCR4) (Kryczek et al., 2005).

The secreted glycoprotein VEGFA is regarded as the key regulator of both physiological and pathophysiological angiogenesis as it facilitates blood vessel growth and remodeling processes by increasing vascular permeability and changing the extravascular matrix. Additionally, VEGFA provides mitogenic and survival stimuli for endothelial cells (reviewed in Thomas, 1996; Yancopoulos et al., 2000; Carmeliet, 2005). Recently, it was shown that autocrine VEGFA signaling is also required for vascular homeostasis (Lee et al., 2007). VEGFA acts through two high affinity receptor tyrosine kinases, FLT-1 (VEGFR1) and KDR (VEGFR2 or Flk-1). KDR is thought to mediate most of the angiogenic functions of VEGFA, whereas FLT1 may act as a sink for VEGFA, or as a decoy receptor to negatively regulate signaling through KDR (Yancopoulos et al., 2000).

Overproduction of hypoxia-inducible mRNAs, including VEGFA, is a hallmark of the highly vascularized neoplasms associated with biallelic inactivation of the VHL tumor suppressor gene, including hemangioblastomas of the retina and central nervous system (CNS), and clear cell renal cell carcinoma (ccRCC) (Maher et al., 1991; Maher and Kaelin, 1997; Van Poppel et al., 2000; Lonser et al., 2003). Hemangioblastomas are the most common manifestation of VHL disease, affecting 60–80% of all patients, and frequently causing neurological morbidity and mortality (Wanebo et al., 2003; Chew, 2005). These cystic tumors contain many thin-walled blood vessels that are separated by non-invasive neoplastic VHL^−/− stromal cells (Vortmeyer et al., 1997), which were shown to be embryonically arrested hemangioblasts that express brachyury, KDR and the stem cell leukemia factor (SCL) (Park et al., 2007). Besides retinal hemangioblastomas, VHL patients can develop retinal neovascularization that is similar to diabetic retinopathy (Chew, 2005). Although activated HIF is a major contributor to the vascularization phenotype of VHL-associated neoplasms, pVHL appears to be essential for endothelial extracellular fibronectin matrix deposition in a HIF-independent fashion (Ohh et al., 1998; Bluyssen et al., 2004; Tang et al., 2006). Furthermore, it has been shown that endothelial cells require pVHL for correct vascular patterning and maintenance of vascular integrity during development (Tang et al., 2006).

We recently reported that zebrafish (Danio rerio) with mutations in the VHL ortholog vhl display general hypoxic physiology, including the upregulation of hypoxia-induced genes, and a severe hyperventilation and cardiophysiological response (van Rooijen et al., 2009). Furthermore, vhl mutants have severe hematopoietic defects, closely resembling the human VHL-associated disorder Chuvash polycythemia.

Here, we investigate the effect of vhl loss on blood vessel formation. The zebrafish vascular system shows a remarkable functional conservation with humans (Vogel and Weinstein, 2000; Isogai et al., 2001; Zon and Peterson, 2005; Covassin et al., 2006). This has also been demonstrated in a pathological context, as tumor cells excreting human VEGFA strongly promoted zebrafish vessel remodeling and angiogenesis (Stoletov et al., 2007). Furthermore, the VEGF signaling pathway is highly conserved. Like all vertebrates, with the exception of placental mammals, zebrafish have four Vegf receptors (Flt1, Kdr, Kdr-like and Flt4); Kdr-like (also known as Kdrl) was lost during the evolution of placental mammals (Bussmann et al., 2007; Bussmann et al., 2008). Although Kdr and Kdr-like represent separate classes of Vegf receptor, the duplicated VEGFA orthologs Vegfaa and Vegfab can bind to, and activate, both receptors in vitro (Bahary et al., 2007).

Using blood vessel-specific transgenes, we show severe neovascularization of the vhl^−/− brain, eye and trunk. We observed vascular leakage in the vhl^−/− retina that results in severe macular edema and retinal detachment, resembling VHL-associated vascular neoplasms and diabetic retinopathies. Quantitative real-time PCR (qPCR) indicates that there is increased Flt1, Kdr and Kdr-like signaling in vhl mutants. Treatment with VEGFR tyrosine kinase inhibitors that are used in the cancer clinic halted excess blood vessel formation, demonstrating that Vegfaa/b and Vegfb act as key effectors of the vhl^−/− angiogenic phenotype by signaling through the Flt1, Kdr and Kdr-like receptors. We validate zebrafish vhl mutants as an important new in vivo model for pathological angiogenesis and vascular retinopathies.

Through HIF-α, VHL plays a crucial role in the adaptive cellular response to hypoxia, including the regulation of angiogenesis through factors such as VEGFA. We have previously shown that vhl loss-of-function mutants display a systemic hypoxic response, including the upregulation of several angiogenesis-related genes (van Rooijen et al., 2009). At around 7.5 days-post fertilization (dpf), whole-mount in situ hybridization results revealed an overexpression of vegfa (vegfaa) mRNA, predominantly in the vhl mutant brain, eye and glomerulus. Increased kdr and kdr-like (also known as kdrl) mRNA expression was observed in blood vessels, the liver and the glomerulus. Additionally, kdr was expressed highly in the vhl^−/− eye.

Here, we investigated the contribution of other Vegf signaling pathway components by qPCR. We identified the vertebrate VEGF ligands and performed synteny analyses using the Ensembl database (http://www.ensembl.org). We constructed a rooted neighbor-joining phylogenetic tree (Fig. 1A) showing the conservation of VEGF ligands in zebrafish. The duplicated zebrafish VEGFA orthologs vegfaa and vegfab (Bahary et al., 2007), as well as zebrafish vegfc (Ober et al., 2004) and vegfd (Song et al., 2007), have all been described previously. From our analyses, we identified zebrafish vegfb, which is syntenic to human VEGFB (not shown) and with orthologs that are present in other fish species. The placental growth factor (PGF) locus is duplicated in the zebrafish, fugu and stickleback. Zebrafish pgfa is syntenic with human PGF, and zebrafish pgfb is syntenic with pgfa (not shown), indicating that a fish-specific duplication event occurred during evolution.

To investigate the expression levels of all Vegf ligands and receptors, qPCR analyses were performed on embryos at 7.25 dpf (Fig. 1B). vhl mutants displayed a significant upregulation of flt1 (2.42-fold), kdr (4.53-fold) and kdr-like (2.65-fold) receptors, accompanied by increased vegfaa (3.04-fold), vegfab (9.16-fold) and vegfb (2.70-fold) ligand expression. flt4, vegfc, vegfd and pgfb levels were not altered significantly, and pgfa could not be detected. These data suggest that Flt1-, Kdr- and Kdr-like-mediated signaling is increased in vhl mutants.

We crossed vhl mutant alleles into a kdr-like:egfp transgenic (TG) background and observed a marked increase in the number of blood vessels throughout the vhl^−/− embryo. Although vasculogenesis is seemingly unaffected at the morphological level (not shown), angiogenesis is enhanced, as first evidenced by the increase of cranial blood vessels at around 58 hours post-fertilization (hpf) (Fig. 2A). The trunk vasculature is unaffected at this stage; however, blood vessels appear dilated (Fig. 2B). At 3.5 dpf, blood vessels can be observed throughout the forebrain and midbrain of vhl mutants. Furthermore, compared with their siblings, most vessels in vhl^−/− zebrafish appear dilated, including the ventral aorta (VA), branchial arches (AA), prosencephalic artery (PrA) and mesencephalic vein (MsV) (arrowheads, Fig. 2A IV). In the trunk, aberrant angiogenic sprouting emanates from the intersomitic vessels (ISV), mainly in regions that are dorsal to the horizontal myoseptum (arrowheads, Fig. 2B XI). By 5.75 dpf, strong neovascularization of the entire vhl^−/− brain is observed (Fig. 2A VI). Similarly, the trunk vasculature in vhl mutants is expanded, forming a complex of functional vessels (Fig. 2B XII) through which blood flow was observed (not shown).

To investigate whether lymphangiogenesis is affected, we examined 7.5-dpf vhlmutants expressing both the fli1:egfp (marking both blood and lymph vessels) and kdr-like:rfp (marking only blood vessels) transgenes. Although confocal analysis of the trunk revealed that vhl mutants display a slightly dilated thoracic duct (TD), lymphangiogenesis was not obviously affected (Fig. 2C). In accordance with our qPCR data showing unaltered flt4, vegfc and vegfd expression levels in vhl mutants, we conclude that the vhl^−/− angiogenic response is restricted to the blood vasculature at the stages analyzed.

Since retinal vascular lesions are the most common and earliest manifestations of VHL disease (Wong et al., 2007), we investigated the vhl mutant eye vasculature. The zebrafish embryonic eye is nourished by choroidal vessels overlying the retinal pigmented epithelium (RPE) and by a central retinal artery, from which hyaloid vessels branch by angiogenesis and attach to the back of the lens (Alvarez et al., 2007). Confocal analysis of vhl mutant eyes revealed a dramatic increase in both the choroidal (Fig. 3A I,III) and the hyaloid (Fig. 3A II,IV) vascular networks compared with siblings at 5.75 dpf.

VHL-associated retinal and CNS hemangioblastomas display increased expression of the chemokine receptor CXCR4 and its ligand SDF-1 in tumor and vascular cells (Zagzag et al., 2005). CXCR4 is regulated by HIF and VEGFA (Zagzag et al., 2006), and CXCR4/SDF-1 and VEGFA have been shown to synergize in the context of angiogenesis (Kryczek et al., 2005). In zebrafish, the CXCR4 gene is duplicated into cxcr4a and cxcr4b, with both fulfilling distinct functions (Chong et al., 2001). In Fig. 3B, we show that cxcr4a is expressed in blood vessels in the vhl^−/− brain and retina, whereas in siblings, no vascular expression could be detected. cxcr4b is expressed in the pronephric hematopoietic tissue (supplementary material Fig. S1), which is expanded in vhl mutants, as described previously (van Rooijen et al., 2009). Interestingly, although we see a marked increase in blood vessels throughout the vhl mutant embryo, cxcr4a-positive blood vessels can only be observed in the distinct areas where high vegfaa mRNA expression is detected (supplementary material Fig. S1). This is most striking in the vhl^−/− brain and in the vhl^−/− eye, where cxcr4a-positive blood vessels were detected in the retina and not in the neovascularized choroidal vessels overlying the RPE (Fig. 3B).

In mammals, primary retinal vessels branch to form intraretinal capillaries that form plexi within the inner and outer plexiform layers; by contrast, these networks of vessels are not present in the largely avascular zebrafish eye (Wittenberg and Wittenberg, 1974; Alvarez et al., 2007; Fruttiger, 2007). Accordingly, intraretinal vessels are not observed in the sibling eye at 7.5 dpf (Fig. 3B,C I–III). Histological analysis of the vhl^−/− retina, however, reveals the presence of cxcr4a-positive blood vessels (containing blood cells), predominantly in the inner plexiform layer (IPL) at 7.5 dpf (arrowheads, Fig. 3B,C IV).

A common clinical feature of VHL-associated and diabetic retinopathies is the development of macular edema, which is the accumulation of fluid in the central part of the retina (Tranos et al., 2004). Although a broad range of pathologies can induce macular edema, in these particular instances, edema is caused by fluid leakage from the neovascularized retinal capillaries (breakage of the inner retina-blood barrier) (Tranos et al., 2004). Owing to the hydrostatic characteristics of the distinct retinal nerve layers (with the plexiform layers being high-resistance barriers to flow) (Antcliff et al., 2001), macular edema is mainly observed in the outer retinal nerve layers, including the inner nuclear layer (INL) and the photoreceptor cell layer (PCL) (Ciulla et al., 2003; Tranos et al., 2004). If left untreated, it can cause detachment of the retinal layers from the RPE and lead to vision loss.

Histological analyses revealed multiple retinal lesions in the outer retinal layers of the vhl^−/− eye (asterisks, Fig. 3C IV–VI), suggestive of macular edema. This was most pronounced in the ventral half of the mutant eye. In 95.8% (n=23/24) of the examined vhl^−/− eyes, severe macular edema was observed predominantly in the INL and PCL (Fig. 3C IV–VI), and mostly in close proximity to the outer plexiform layer (OPL). In 14.3% (n=4/28) of the vhl^−/− eyes, additional sites of retinal detachment of the retinal nerve layers from the RPE were observed (Fig. 3C VI). These retinal abnormalities were never observed in age-matched sibling eyes (n=24/24). Electron microscopy analysis confirmed these vhl^−/− retinal lesions to represent macular edema since they contained electron-dense material that was not observed in siblings (n=2/2) at 7.5 dpf. As shown in Fig. 3D, accumulation of this material was observed between the nuclei of the INL, and reaching into the OPL and PCL. It even appeared that nuclei from the INL were pushed out into the OPL (arrowheads, Fig. 3D) or IPL (white arrow, Fig. 3C IV) by the accumulating fluid.

To investigate whether the observed macular edema is caused by vascular leakage, we performed fluorescent angiographies at 6 dpf with a high molecular weight rhodamine-dextran conjugate (2,000,000 MW TAMRA) that is normally too large to pass through the vessel wall. At approximately 20–25 hours post-injection, embryos were fixed and subjected to confocal analysis. In Fig. 3E, nuclear counterstaining revealed that while retinal abnormalities were never observed in siblings, TAMRA clearly accumulates in the vhl−/− retinal lesions and at distinct locations between the outer retinal cell layers, indicating sites of macular edema.

Confocal analysis of the optic artery and the inner optic circle after TAMRA injection in 6-dpf kdr-like:egfp transgenic embryos revealed severe vascular leakage through the vessel wall into the surrounding tissue (white arrows) in vhl mutants. Three-dimensional (3D) reconstructions from confocal stacks are presented in Fig. 4A–C (see supplementary material Movie 1 for a 3D movie animation of these images). Taken together, these data suggest that neovascularization of the choroidal vasculature in the vhl^−/− eye induces vascular leakage, leading to severe macular edema in the outer layers of the retina, irregularities in the retinal layers, and eventually retinal detachment. Although vhl mutant embryos do not develop typical hemangioblastomas, these data indicate that the vhl^−/−-associated angiogenic phenotype shares distinct characteristics with VHL-disease manifestations, and more generally with other vascular retinopathies.

To investigate whether the VEGF signaling pathway is necessary and sufficient to induce the observed vhl-associated neovascularization in our mutant zebrafish, we treated embryos with two multi-targeted VEGF receptor tyrosine kinase inhibitors: 676475 (Calbiochem) and sunitinib malate (or Sutent). Blocking VEGF has been shown to inhibit vasculogenesis, angiogenic formation of the ISVs, and formation of the DLAV, when administered very early during zebrafish development (Chan et al., 2002; Lee et al., 2002). Therefore, to ensure proper patterning of the cranial and trunk vasculature prior to treatment, we initiated exposure to each individual VEGFR inhibitor at 58 hpf, when the first vascular abnormalities were observed in vhl mutants. Embryos were incubated with 10 μM 676475, 0.2 μM sunitinib or DMSO (vessel control) in embryo medium for 3 days (Fig. 5). Confocal analysis of the head, eye and trunk vasculature at 5.75 dpf revealed that the pathological angiogenesis in vhl mutants was almost completely blocked by 676475, and completely blocked by sunitinib treatment. Unfortunately, owing to the accumulation and auto-fluorescence of the yellow substance from the sunitinib solution in the skin (which also occurs in human patients), it was not possible to analyze the choroidal vasculature by confocal microscopy. Our results indicate that vhl mutants provide a new and clinically relevant model that is suitable for large-scale compound screens for anti-angiogenic agents and combinatorial treatments.

VHL disease is characterized by the development of highly vascularized neoplasms that overexpress a variety of angiogenic factors, including VEGFA, KDR and CXCR4 (Chan et al., 2005; Zagzag et al., 2005; Liang et al., 2007; Park et al., 2007). Similarly, we show that loss of vhl in zebrafish embryos induces severe pathological neovascularization, with increased Vegf signaling and cxcr4a expression in the brain and retina. Comparable to VHL-associated retinal neoplasms, diabetic retinopathy and age-related macular degeneration, we show by TAMRA angiographies that vascular abnormalities in the vhl^−/− retina lead to vascular leakage, severe macular edema and retinal detachment.

Although hemangioblastomas were not found in our vhl mutant embryos or in heterozygote adult zebrafish, the observed vhl^−/− retinal neovascularization appears to be more similar to a new class of ocular VHL disease that was more recently identified in a large prospective case study (Chew, 2005). Of the VHL patients with ocular lesions, a novel ocular manifestation similar to diabetic retinal neovascularization was identified in 17 patients (8.3%). Although these neovascular lesions are less common than hemangioblastomas, the actual incidence might be higher owing to the fact that they have not been attributed to loss of pVHL.

Vascular abnormalities have been observed in murine Vhlh (also known as Vhl) models. Conventional Vhlh knockout mice (Gnarra et al., 1997), and mice harboring a targeted deletion of Vhlh in endothelial cells (Tang et al., 2006), are embryonic lethal [death occurs between embryonic day (E)9.5–12.5] owing to hemorrhagic lesions and vascular abnormalities in the placenta. Deletion of Vhlh in the epidermis (Boutin et al., 2008) and liver (Haase et al., 2001; Rankin et al., 2005) resulted in increased angiogenesis in the targeted organ systems, and mosaic Vhlh inactivation (Ma et al., 2003) or systemic Vhlh^+/– mice often develop hepatic cavernous hemangiomas. However, hemangioblastomas of the retina and CNS have not been found in any Vhlh mouse model. Therefore, the fact that the vhl zebrafish survives past embryogenesis, and that additional organ systems are affected, underscores the complementary nature of these two models in studying hypoxia-induced angiogenesis.

It was shown that the development of vascularized tumors in Vhlh-deficient livers was entirely dependent on HIF (Rankin et al., 2005). This finding supports our data, indicating that systemic stabilization of Hif through the loss of vhl is the most likely initiator of vegf-induced vhl^−/− neovascularization. However, it has been shown that, during mouse development, endothelial cells require Vhlh for correct vascular patterning and maintenance of vascular integrity (Tang et al., 2006), through HIF-independent regulation of fibronectin deposition (Ohh et al., 1998; Tang et al., 2006). Therefore VHL-induced changes in the extravascular matrix probably contribute to the characteristics of the vascular neoplasms in VHL patients, however, this needs to be investigated further.

In recent years, VEGF, its receptors and the HIF signaling pathway have, in general, received considerable attention as potential therapeutic targets for cancer and other angiogenesis-related disorders. However, the current in vitro and in vivo models for angiogenesis fail, to a certain extent, to recapitulate the complexity of angiogenesis and its microenvironment, or to allow in vivo functional evaluations (Taraboletti and Giavazzi, 2004). Recently, an adult zebrafish model for hypoxia-induced retinal angiogenesis was presented (Cao et al., 2008). Adult zebrafish that were exposed to 10% air-saturated water developed severe retinal neovascularization after 3–12 days of incubation. The anti-VEGF agents sunitinib and ZM323881 effectively blocked hypoxia-induced neovascularization, demonstrating the clinical relevancy of this retinal angiogenesis model. However, these studies are invasive and an adult zebrafish system is less suitable for high-throughput and cost-effective screening of novel compounds owing to the minimal water volume required.

In this study, we showed that vhl^−/− embryos develop severe neovascularization of the choroid and hyaloid vessels; this was blocked by treatment with the tyrosine kinase inhibitors 676475 and sunitinib. In diabetic retinopathy, age-related macular degeneration, and VHL-associated retinal neoplasms, the prominent vascular abnormalities induce vascular leakage, macular edema and retinal detachment, which can lead to a complete loss of vision (Arjamaa and Nikinmaa, 2006). Similarly, we also observed severe macular edema and retinal detachment in vhl mutants, which was not described in the adult zebrafish retinal angiogenesis model (Cao et al., 2008).

Additionally, we observed an early and fully penetrant increased angiogenic response throughout the vhl^−/− embryo. Wild-type zebrafish embryos that were kept under hypoxic conditions (partial pressure of oxygen of 8.7 kPa) from 1 to 15 dpf did not show significant changes in vascularization, or in the basic pattern of the vascular bed of the trunk and gut, as compared with normoxic animals, and the perfusion rate was increased in the trunk musculature, unchanged in the brain and reduced in the gut (Schwerte et al., 2003). Since, at lower oxygen levels, zebrafish embryos do not develop properly (Schwerte et al., 2003) and adult zebrafish die (Cao et al., 2008), the induced hypoxia in these studies might not have been low enough to induce a similar angiogenic response as in vhl^−/− embryos.

Therefore, our data indicate that vhl mutant zebrafish embryos represent a unique and clinically relevant angiogenesis model where, in the context of systemic hypoxia signaling, the process of pathological angiogenesis and vascular retinopathies can be studied non-invasively in a tissue-specific context under normoxic conditions. Using a blood vessel-specific transgenic background, vhl mutant embryos allow for a cost-effective and efficient way to screen pharmacological agents for anti-angiogenic properties and combinatorial treatments in a whole organism setting.

Zebrafish were maintained as described previosuly (Westerfield, 1995). Animal experiments were conducted in accordance with the Dutch guidelines for the care and use of laboratory animals, with the approval of the Animal Experimentation Committee (DEC) of the Royal Netherlands Academy of Arts and Sciences (KNAW). The mutant alleles vhl^hu2117(Q23X) and vhl^hu2081(C31X) (van Rooijen et al., 2009) were out-crossed to the transgenic lines TG(kdr-like:egfp)^s843 (Jin et al., 2005), TG(fli1a:egfp)^y1 (Lawson and Weinstein, 2002) and TG(kdr-like:ras-cherry)^s916 (Chi et al., 2008; Hogan et al., 2009). Unless indicated otherwise, transheterozygote embryos (vhl^hu2117/vhl^hu2081) were used in experimental assays. Where indicated, embryos were anesthetized with MS222 (final concentration of 0.17 mg/ml).

Phylogenetic analysis of the zebrafish Vegf ligands was performed with the MEGA3 software package (Kumar et al., 2004), using the PDGF and VEGF family domain (cd00135) from NCBI’s Conserved Domain Database (CDD). Amino acid sequences were aligned with ClustalW and a phylogenetic tree was constructed using a neighbor-joining algorithm. The resulting tree was tested using 1000 bootstrap resamplings. Pairwise distances were calculated with the PAM substitution matrix. Identification of vertebrate VEGF ligands and synteny analysis was performed using the Ensembl database (http://www.ensembl.org), release 44, April 2007. Sequences are available upon request.

For total RNA isolation, a maximum of 40 vhl mutants and siblings per clutch were homogenized by shredding in 600 μl of RTL lysis buffer (Qiagen RNeasy kit) containing 10% beta-mercaptoethanol. One volume of 70% ethanol was added and the homogenate was loaded onto a column for total RNA isolation according to the manufacturer’s protocol, followed by DNaseI (Promega) treatment. The RNA quality and concentration were determined using a NanoDrop, and verified by gel electrophoresis. cDNA was synthesized from total RNA (1–5 μg) with random hexamers (Biolegio, www.biolegio.nl) using reverse transcriptase M-MLV (Promega). Primer sets were designed using Primer3 (http://frodo.wi.mit.edu/) and Limstill (http://limstill.niob.knaw.nl/) software, with an optimal product size of 150–200 base pairs that, where possible, spanned two exons to avoid genomic contamination. The PCR efficiency and optimal melting temperatures were determined per primer set, and the specificity was verified by gel electrophoresis using standard real-time PCR on zebrafish cDNA. See supplementary material Table S1 for primer sequences and melting temperatures. qPCR was preformed using the MyIQ single color real-time PCR detection system and software (Bio-Rad). Each reaction contained: 12.5 μl SYBR Green fluorescent label (Bio-Rad), 3 μl of 1.5 μM primer mix, 4.5 μl MQ and 5 μl cDNA (10ng/μl). Cycling conditions were: 95°C for 3 minutes; 40 cycles of 95°C for 10 seconds and the optimal primer temperature for 45 seconds; followed by 95°C for 1 minute; and finally 65°C for 1 minute. All reactions were performed in triplicate on cDNA isolated from at least two different clutches of pooled vhl mutants and siblings. Ct values were corrected for the 18S housekeeping gene (Cooper et al., 2006). vhl mutant cDNA concentrations were calculated in arbitrary units compared with the sibling average, and are represented as the fold change with the sibling value set to 1. A paired Student’s t-test was used to determine significant differences in expression between vhl mutants and siblings.

Anesthetized embryos were embedded in 0.5% agarose on a coverslip, and confocal images were collected on a Leica DM IRE2 microscope (Leica Microsystems) using 10×, 20×, 40× and 63× objectives. 3D reconstructions of confocal stacks were generated using the Volocity software package (Improvision, www.improvision.com).

Transversal plastic sections (7 μm) were stained with periodic acid-Schiff (PAS) or eosin using standard protocols.

Anesthetized embryos were embedded in 0.5% agarose and were administered with 10 mg/ml of 2,000,000 MW tetramethyl rhodamine (TAMRA; Molecular Probes) by cardiac injection at 6 dpf. Only embryos exhibiting TAMRA throughout the cardiovascular system immediately after injection were analyzed further. At approximately 20–25 hours post-injection, embryos were fixed with 4% PFA for 2 hours at room temperature or overnight at 4°C. Retinal vascular leakage was determined by confocal analyses on sections cut on either a Microm HM650V (http://www.microm-online.com) vibratome (75–100 μm) or a Leica CM3050 cryostat (50 μm).

Whole-mount in situ hybridization was performed as described (Schulte-Merker, 2002) with minor modifications. Antisense digoxygenin (Roche)-labeled mRNA probes for cxcr4a (Chong et al., 2001), cxcr4b (Chong et al., 2001) and vegfaa (Liang et al., 2001) were generated as described, and purified using NucleoSpin RNA clean-up columns (Machery-Nagel, www.macherey-nagel.com). To improve probe penetration, 7.5-day-old larvae were partially cut open at the level of the yolk sac extension after proteinase K permeabilization. After in situ hybridization, pigmented embryos were bleached in a solution containing 0.5× SSC, 5% formamide and 10% H₂O₂ for 20–40 minutes. Embryos were mounted in 2% methylcellulose on a depression slide and imaged on a Zeiss axioplan microscope with a 5× objective.

Embryos were fixed in Karnovsky fixative (2% paraformaldehyde, 2.5% glutaraldehyde, 0.08 M Na cacodylate, pH 7.4, 0.25 mM calcium chloride, 0.5 mM magnesium chloride, set to pH 7.4) for at least 24 hours at 4°C. Samples were postfixed in 1% osmium tetroxide and embedded in Epon 812. Ultra-thin sections (60 nm) were contrasted with 3% uranyl magnesium acetate and lead citrate, and viewed with a Jeol JEM 1010 transmission electron microscope.

2.5-day-old embryos were treated with 10 μM of the VEGF receptor tyrosine kinase inhibitor 676475 (Calbiochem, http://www.merckbiosciences.co.uk/; single dose) or 0.2 μM of sunitinib malate (Selleck Chemicals, www.selleckchem.com; daily refreshment) in embryo medium (Westerfield, 1995), at 28°C, in 6-well culture plates containing 30 embryos per well. Control embryos were incubated with an equivalent amount of DMSO solution under the same conditions. Experiments were performed in triplicate.

This project was funded by the Dutch Cancer Society (UU-2006-3565). R.H.G. is supported by a VIDI award (NWO). F.J.v.E. is supported by Cancer Research UK (C23207/A8066), and J.B. by the Stichting Vrienden van het Hubrecht. We gratefully acknowledge Bart Weijts (Veterinary Medicine, Utrecht University) for help with quantitative real-time PCR; Roel Goldschmeding and Kevin van de Ven (Department of Pathology, UMC Utrecht) for transmission electron microscopy; Henk Verheul and Kristy Gotink (Department of Medical Oncology, VUMC Amsterdam) for helpful discussions; and Lasse Dahl Jensen (Karolinska Institute, Sweden) for sharing information on sunitinib treatments in zebrafish.