A Zebrafish Model to Study and Therapeutically Manipulate Hypoxia Signaling in Tumorigenesis

Oxygen homeostasis, a physiologic process essential and critical for the normal development and functioning of an organism, requires coordinate regulation of multiple pathways. Reduction in the normal level of oxygen tension in tissues (hypoxia) is observed in a range of disease conditions, including locally in peripheral vascular disease, myocardial infarction and stroke, and systemically in pulmonary disease. In addition, hypoxia is a well-described feature of tumor microenvironment. Intratumoral hypoxia is elicited by uncontrolled proliferation of tumor cells coupled with aberrant growth of tumor vasculature. Under hypoxic conditions 2 distinct mechanisms are activated, which promote cell/tissue survival in low oxygen environments and enhance oxygen delivery to the hypoxic sites. These adaptive responses to hypoxia are induced by a key transcription factor, hypoxia inducible factor (HIF), whose activity is regulated by the availability of oxygen (1). HIF is a heterodimeric protein complex consisting of basic helix-loop-helix Per/Arnt/Sim (PAS) domain containing α and β subunits (HIFα and HIFβ). HIF plays a crucial role in altering the cellular metabolism of tumors to thrive under hypoxic conditions and in facilitating malignant transformation of tumors (2). Several HIF target genes such as VEGF, PDGF, TGF, CXCR4, TGF-α, and MMP-1 play important roles in the development of cancer (3–5) and significant correlations have been observed between increased HIF-1α levels and patient mortality in many types of cancer (6).

The stability of HIF is regulated in an oxygen-dependent mechanism by the von Hippel-Lindau tumor suppressor protein (pVHL). pVHL is encoded by VHL whose biallelic loss lead to the development of a hereditary cancer syndrome involving multiple hemangioblastomas of CNS, clear cell renal carcinoma and pheochromocytoma (7), which occur when heterozygosity is lost either by mutation on the second allele of VHL or via epigenetic silencing (8). pVHL is an essential component of the E3 ubiquitin ligase complex and interacts with the HIFα subunit under normoxic conditions, targeting them for proteasomal degradation. The pVHL–HIFα interaction requires oxygen-dependent hydroxylation of either of 2 conserved prolyl residues (Pro402 and Pro564) within the oxygen-dependent degradation (ODD) domain of HIFα by prolyl hydroxylase domain (PHD) enzymes (9–12).

HIF prolyl hydroxylases belong to the 2OG-dependent dioxygenase superfamily (13), which also require molecular oxygen as a cosubstrate for their catalysis resulting in the contribution of one oxygen atom for the hydroxylation reaction and the second oxygen atom toward the oxidative decarboxylation of 2-oxoglutarate yielding succinate and CO₂. Prolyl hydroxylation of HIFα generates a binding site for pVHL leading to polyubiquitination of HIFα and subsequent hydrolysis. As oxygen tension decreases below the critical threshold, prolyl hydroxylases become quiescent, allowing HIFα to stabilize and migrate into the nucleus to form a functional transcription complex by binding with HIFβ subunit. Thus, PHD enzymes play a critical role in transducing the physiological signal of hypoxia to the nucleus (14–17). In vertebrates, there are 3 prolyl hydroxylase (PHD1, PHD2, and PHD3) isoforms capable of hydroxylating HIFα. Interestingly, the PHD2 and PHD3 genes are also HIF target genes, providing negative feedback and thus a mechanism to create a flexible threshold for activation of the HIF signaling pathway (17, 18).

By negatively regulating HIFs, pVHL plays an important role in the adaptive cellular response to hypoxia. Constitutive stabilization of HIFα subunits and enrichment of PHD transcripts have been observed in the VHL-deficient renal carcinoma cells under normoxic conditions (19, 20). We reported earlier the identification of 2 mutant lines carrying null mutations in the zebrafish vhl ortholog. The mutant embryos exhibit a general systemic hypoxic response with upregulation of genes involved in oxygen sensing and transport, angiogenesis, anaerobic energy metabolism, and display key features of the human VHL-associated disorder “Chuvash polycythemia” (21, 22). The vhl mutants also displayed a very strong upregulation of prolyl hydroxylase 3 (phd3) gene expression at various stages (21), suggesting phd3 as an ideal candidate for the readout of Hif activity. We characterized phd3 as a suitable response gene and created a Tg(phd3::EGFP) hypoxia reporter line and describe the utility of this transgene as an in vivo marker for intratumoral hypoxic regions in a zebrafish melanoma model system.

In zebrafish, the effects of vhl heterozygosity on tumor formation have not been analyzed in detail, but such fish are generally healthy and fail to show indications of increased susceptibility to tumor formation. Using the Tg(phd3::EGFP) reporter line as a tool we assayed the frequency of LOH in the vhl fish. However, the late average age of diagnosis of ccRCC in human VHL patients suggests that several additional changes are required to turn a VHL LOH cell into a cancer cell, thus the short lifespan of the fish may prevent development of renal cancer. We therefore sought to test their susceptibility when challenged with a wide-spectrum zebrafish carcinogen; 7,12-Dimethylbenz(a)anthracene (DMBA; ref. 23). We report here strong susceptibility of the DMBA-treated vhl^+/− fish to hepatic and intestinal tumorigenesis, showing for the first time that Vhl functions as a tumor suppressor in zebrafish. The treatment also led to an increased occurrence of proliferating renal tubules compared with appropriate control fish. Our observations establish that the Tg(phd3::EGFP) line provide a novel in vivo tool to monitor Vhl/Hif/hypoxic signaling under tumor pathologic conditions.

Zebrafish (Danio rerio) strains were maintained on 14-/10-hour light/dark cycle at 28°C in a UK Home Office approved aquarium facility at the University of Sheffield. The fish embryos were staged according to hours or days postfertilization (hpf or dpf; ref. 24).

We have utilized 2 different vhl mutant lines namely, vhl^hu2117/+ and vhl^hu2081/+, previously identified and described (21).

A EGFP-SV40pA-FRT-Kn-FRT recombineering cassette and red recombineering system in EL250 cells were used to insert EGFP with an SV40 polyadenylation site at the phd3 ATG start site in BAC clone CHORI73-277E22. This modified BAC was used to generate stable transgenic lines according to published protocols (25). We selected 2 different lines namely, Tg(phd3::EGFP)^i144/+ and Tg(phd3::EGFP)^i146/+ with the i144 exhibiting a stronger expression. All the experiments in this study were conducted on the i144 line.

We have observed phd3 transgene activation phenotype in the 2 different vhl mutant allele backgrounds viz. vhl^{hu2117/hu2117} and vhl^{hu2081/hu2081} as well as in the vhl^{hu2081/hu2117} transheterozygous embryos.

The embryos were raised in E3 medium in a 5% oxygen supplied sealed chamber (INVIVO₂ 200; Ruskinn Technology Ltd.) for stipulated periods of time. The chamber was deoxygenized by positive infusion of 5% CO₂/95% N₂ gas mixture with >90% relative humidity at 28°C. The continuous O₂ saturation and the pressure inside hypoxia chamber were maintained during the course of the experiment. The transgenic embryos raised under similar conditions in a well-aerated bag inside the chamber served as controls.

A splice blocking antisense morpholino (Genetools) sequence, namely 5′-GCATAATTTCACGAACCCACAAAAG-3′, designed to the exon1 splice-donor site of the vhl gene was utilized to perturb its function. The morpholino stock solution (10 mg/mL) was diluted appropriately in distilled H₂O and the 1-cell stage transgenic embryos were injected with ∼3 ng per embryo. Full-length zebrafish vhl gene was subcloned into the pCS2+ expression vector. mRNA from the sequence verified constructs was transcribed in vitro after NotI linearization, using SP6 mMESSAGE Kit (Applied Biosystems). Dominant active and dominant negative forms of hif-1αa and hif-1αb were synthesized and injected as previously described (26). For the rescue experiments, 6.5 pg vhl mRNA was injected.

The 3-dpf wild-type transgenic embryos were treated with 100-μmol/L dimethyloxaloylglycine (DMOG; Frontier Scientific) or 0.1% dimethylsulfoxide (DMSO; Sigma-Aldrich) for 1 to 2 days as previously described (21).

Whole-mount in situ hybridization was conducted as per standard protocols (27). The phd3 (BC066699) antisense digoxygenin-UTP labeled mRNA probe was synthesized from an expressed sequence tag (EST) clone (RZPD/Imagenes; ref. 21) and the in situ data were collected on a Zeiss Axioplan with a 5× or 20× objective using a Spot4 digital camera.

Wild-type and vhl^hu2117/+ fry at 3 weeks of age were immersed overnight in aquarium water containing 5 ppm of DMBA (Sigma-Aldrich-Chemie BV) dissolved in DMSO (23). The next day, all fry were rinsed several times in regular aquarium water and subsequently were returned to the aquarium. Fish were carefully grown and regularly monitored for signs of sickness or evidence of tumor formation. Eventually all fish were culled for analysis at 14 months after the DMBA treatment.

Adult fish were euthanized, culled, and fixed in 4% paraformaldehyde for 3 to 4 days at 4°C. The samples were decalcified in 0.25M EDTA pH 8 for 2 to 4 days at room temperature, embedded in paraffin, and 5-μm sections were made. For immunostaining, the primary antibodies mouse monoclonal anti-PCNA (1:1,000; Abcam), rabbit polyclonal anti-HIF-1α (1:100; ref. 28) and rabbit polyclonal anti-GFP (1:250; Torrey Pines Biolabs) were used. The stained sections were viewed in a Leica DM 2500 microscope and photographed using Leica DFC 420C digital camera. All the tumor sections were observed and analyzed by a clinical pathologist.

After the hypoxia exposure treatment, the embryos were harvested in the hypoxic E3 medium itself and protease inhibitor MG132 was added (25 μmol/L final concentration) to prevent proteasomal degradation of Hif-1α protein. The samples were processed as per standard protocols and the antibodies used were anti-Hif-1α (1:500; ref. 28) and anti-acetylated tubulin (1:1,000; Sigma-Aldrich).

Real-time RT-PCR was conducted on a BioRad MyiQ machine along with BioRad iQ SYBER Green Supermix according to the manufacturer's protocol. The sequences of the primers used in the study are as follows: vegf A: F—CAGTGTGAGCCTTGCTGTTC, R—CCATAGGCCTCCTGTCATTT; phd 3: F—CGCTGCGTCACCTGTATT, R—TAGCATACGACGGCTGAACT; vhl: F—ATGTCGGCTGTCTGGAGATT, R—GATGCACAGGTGTGCTCAGT; β-actin: F—GCAGAAGGAGATCACATCCCTGGC, R—CATTGCCGTCACCTTCACCGTTC. The primers were optimized using a standard curve of wild-type cDNA. The cycling conditions of the PCR program were as follows: step 1: 95°C × 1 minutes; step 2: 40 cycles of the following 95°C × 30 seconds, 60°C × 30 seconds, 72°C × 30 seconds; step 3: 95°C × 30 seconds; step 4: 60°C × 30 seconds; step 5: 40 cycles of 55°C × 10 seconds. The data were analyzed using the ΔΔCt method (29).

Statistical analysis and the P-value calculations were done using Chi-square test (30).

In microarray expression analyses of 7 dpf vhl^{hu2117/hu2081} transheterozygous and 3.5 dpf vhl^{hu2117/hu2117} homozygous mutant and vhl^+/+ sibling larvae, we observed upregulation of several Hif target genes, including epo, vegf, phd1-3 in the mutants, with phd3 consistently exhibiting the strongest increase overall (e.g., 23-fold at 7 dpf) (21, FvE unpubl.). When we did Western blot analysis of vhl^−/− embryos using Hif-1α antibody, we observed strong deposition of Hif-1α in the mutants (Fig. 1A). Subsequently, when we conducted qPCR on 3.5 dpf vhl^−/− mutant embryos, we observed 216-fold increase in phd3 expression in mutant embryos compared with wild types (Fig. 1B). In comparison, the classical HIF target vegf-a was increased 7-fold (Fig. 1B) in the mutants. To identify the earliest developmental time point at which phd3 expression is stimulated in the vhl^−/− mutant embryos, we conducted whole mount in situ hybridization on selected stages of mutant and sibling embryos. In wild-type embryos, phd3 expression could only be weakly detected in the brain around 30 hpf (data not shown). By 48 hpf, the expression became restricted to a population of cells in the tectum and midbrain–hindbrain boundary (MHB) region and the wild-type fish continued to preserve this confined phd3 expression pattern during the subsequent embryonic and larval development (Fig. 1C and E and Supplementary Fig. S1A). In contrast, vhl mutant embryos exhibited strong phd3 expression as early as 21.5 hpf (25-somite stage) in the brain and pronephros (Fig. 1D), suggesting an induction in Hif activity. By 36 hpf, the mutant embryos showed strong phd3 expression in the heart and other visceral organs indicating a systemic response. These embryos continued to exhibit robust expression of the phd3 gene all through their embryonic (Fig. 1F and Supplementary Fig. S1B) and larval development, until their death at 10 to 12 dpf. Presence of wild-type maternal vhl mRNA and Vhl protein during the early embryonic stages in mutant embryos (21), might explain the delay in the activation of phd3 expression until ∼21 to 22 hpf. These observations suggested that, although the physiologic and behavioral responses of vhl mutants only become evident by 3 dpf, the molecular onset of the hypoxic response can be detected from approx. 22 hpf.

In the human PHD3 gene, a hypoxia response element (HRE) was identified in the first intron, but its conservation in fish is unclear and further HREs could be present (31). Therefore, we utilized a recombineering strategy to insert an EGFP with an SV40 3′ UTR at the phd3 ATG start site in the BAC clone CHORI73-277E22. This BAC contains the complete phd3 gene in a large genomic context (>25 Kb 5′ and >90 Kb 3′). The phd3 start codon was destroyed by the recombination thus preventing the transgene construct from interfering itself with the hypoxic signaling pathway. This modified BAC was injected to generate 2 stable transgenic lines and, the stronger line was used in all subsequent experiments.

In wild-type embryos, similar to phd3 in situ staining, phd3::EGFP transgene-mediated EGFP fluorescence commenced weakly in the brain and spinal cord from 30 hpf (data not shown). The transgene expression became restricted to MHB region by around 60 hpf and the EGFP fluorescence was found to adhere to this region during subsequent stages of development (Fig. 1G). In contrast, the vhl mutant embryos commenced transgene expression from 25-somite stage onward in the brain, spinal cord, and pronephros (data not shown). Subsequently, the mutants began exhibiting strong induction of the transgene in several other organs like the heart, pectoral fins, pancreas, and the hatching gland (Fig. 1H). In situ hybridizations for gfp mRNA on phd3::EGFP transgenic embryos correlated well with the endogenous phd3 expression at various stages of development (Supplementary Fig. S2). We also quantified and compared the level of EGFP fluorescence between the vhl mutants and wild-type siblings at 4 dpf, and found a 3-fold increase in the amount of fluorescence in the mutants (Supplementary Table S1).

To prove that phd3::EGFP activation observed in the vhl mutant embryos is specifically affected by the loss of Vhl function, we tried to rescue this phenotype by injecting 6 pg of capped vhl mRNA into single-cell stage embryos. By around 28 hpf, both the uninjected and injected wild-type embryos showed no activation of the phd3::EGFP transgene. But while the uninjected vhl mutants exhibited strong expression of the transgene at 28 hpf, the mRNA-injected mutant embryos exhibited no EGFP fluorescence at the identical time point (Fig. 2A).

We also stimulated phd3::EGFP expression in the wild-type embryos by knocking down vhl function using antisense morpholino oligonucleotides. Injection of vhl splice morpholino into 1-cell stage Tg(phd3::EGFP)^i144/+ embryos led to strong induction of the transgene from 60 hpf onward and morphant embryos continued to display strong EGFP fluorescence until 120 hpf. But when we injected vhl morpholino and vhl mRNA together, persistence of wild-type expression pattern was observed during embryonic and larval development (Fig. 2B), suggesting a complete rescue of the morphant phenotype.

Our observations suggest that loss of function mutations in vhl can activate phd3::GFP reporter gene expression possibly through the stabilization of Hifα subunit but Vhl protein is reported to have numerous noncanonical functions as well (32). To verify the role of Hif transcription factor in eliciting transgene expression in vhl^−/− embryos, we generated dominant active (DA) forms of Hif-1αa and Hif-1αb by mutating the conserved proline and asparagine hydroxylation sites of Phd and Fih hydroxylases, respectively (26). When the DA hif-1αa and hif-1αb mRNAs were coinjected into 1-cell stage Tg(phd3::EGFP)^i144/+ embryos, we observed activation of the transgene in the injected embryos from 24 hpf, as evidenced by strong EGFP fluorescence, mainly in the brain and CNS. The enhanced phd3::EGFP expression persisted in the larvae expressing DA Hif-1α isoforms until at least 48 hpf (Fig. 2C). When the same amount of wild-type hif-1αa and hif-1αb mRNAs were injected this level of phd3::EGFP expression was not observed, confirming that the DA versions of hif-1α have successfully stabilized the protein. These observations clearly show the direct role of Hif-1α transcription factor in mediating the phd3::EGFP transgene activation (Supplementary Fig. S3A).

To further extend this observation, we activated Hif in the wild-type transgenic embryos by inhibiting Phd/factor inhibiting Hif (Fih) functions using the prolyl hydroxylase inhibitor DMOG, and thus preventing the degradation of the Hifα subunit even in the presence of functional Vhl. Transgenic wild-type embryos at 2.5 dpf were treated with 100 μmol/L DMOG for 2 days displayed an increase in phd3::EGFP expression in the pronephros, gall bladder, liver, and brain (Supplementary Fig. S4A) compared with very mild expression of transgene observed in the 0.1% DMSO-treated control embryos. Hif activity and hence the transgene induction were observed to be restricted to selected tissues/cells of DMOG-treated embryos, because limitations in the mode and timing of the treatment (21). When mRNAs encoding DN forms of hif-1αa and hif-1αb (26) were injected into Tg(phd3::EGFP)^i144/+ embryos and subsequently treated with DMOG at 24 hpf, we could block the DMOG-mediated induction in EGFP fluorescence at 48 hpf (Supplementary Figs. S3B and S4B).

Our results predict that the Tg(phd3::EGFP)^i144/+ embryos should act as a live reporter of hypoxia, and to validate this we exposed the transgenic embryos to reduced oxygen environment in a hypoxic chamber for stipulated periods of time. When we subjected 12 hpf transgenic embryos to 5% O₂ (hypoxia) for 12 hours, phd3::EGFP expression was triggered indicating the activation of Hif in these embryos (Fig. 3A). In contrast, the normoxia-raised control embryos showed no expression of the transgene at the corresponding stages. High levels of fluorescence observed in the CNS and pronephros of the hypoxia-raised embryos implies vigorous Hif activity in these organ systems in eliciting critical adaptive mechanisms. To understand whether Hif stabilization could be induced during early embryonic development itself, we subjected 128-cell stage transgenic embryos to hypoxia for 4 hours. We observed strong induction of the transgene in the hypoxia-raised embryos by 70% epiboly stage (Fig. 3B) compared with the normoxia-raised embryos, which showed no induction. When we assayed the hypoxia-raised embryos for Hif-1α levels using Western blot analysis, strong presence of Hif-1α was detected in them compared with normoxia-raised embryos, which showed only negligible levels of Hif-1α (Fig. 3C). These observations indicate the usefulness of Tg(phd3::EGFP)^i144/+ embryos as a robust sensor to detect hypoxia.

Because the presence of hypoxic regions is a hallmark of locally advanced solid tumors, we asked whether the phd3::EGFP transgene could be used as a novel reporter for detecting intratumoral hypoxic regions in fish tumor models. To test this hypothesis, we induced melanoma in the Tg(phd3::EGFP)^i144/i144 fish by injecting a human oncogenic HRAS^G12V (V12RAS::mCherry) construct into single-cell embryos and misexpressing the oncogene specifically in cells of the melanocyte lineage (33). Few days after injections, the V12RAS injected fish developed several clones of ectopic melanocytes and after few weeks they progressed into melanocytic naevi. Subsequently, after 3 months we observed few of these lesions to form small tumor nodules. As these melanoma nodules became expanded into melanomas, we periodically monitored them for phd3::EGFP expression. After 1 year, few fish with large tumors (Fig. 4A) exhibited distinct regions of phd3::EGFP-mediated fluorescence (Fig. 4C). These EGFP⁺ regions are likely to correspond to intratumoral hypoxic areas in these melanomas, essentially revealing a subpopulation of tumor cells that possess activated hypoxia-mediated Hif signaling. Histologic analysis of these samples using proliferating cell nuclear antigen (PCNA), Hif-1α and GFP antibody staining revealed clear demarcation of the hypoxic regions within these tumors (Fig. 4D–G). When taking into account the differences in expected half-lives of EGFP (Supplementary Fig S5) and HIF-1α (34) proteins, we observed good correlation between GFP and Hif-1α staining within the melanoma (Fig. 4F and G), thus validating that the EGFP⁺ regions correspond to intratumoral hypoxic regions with activated Hif signaling. These experiments show the potential of the phd3::EGFP transgene in tracking hypoxia under in vivo tumor pathologic conditions in zebrafish.

The vhl^+/− fish are not obviously predisposed to neoplasia formation. One reason for the lack of tumors might be that the lifespan of the fish is too short to accumulate the necessary changes. Hence we asked whether DMBA treatment of Tg(phd3::EGFP)^i144/+;vhl^+/− fish would make them susceptible to tumorigenesis. We treated 4 batches of 21 dpf fish fry with DMBA (23), with each batch consisting of an equal mix of Tg(phd3::EGFP)^i144/+;vhl^+/+ and Tg(phd3::EGFP)^i144/+;vhl^+/− genotypes. Two months after the treatment, we observed that a subset of fish in every batch possessed EGFP⁺ cell clones, primarily in retina (Supplementary Fig. S6A), skin (Supplementary Fig. S6B), gills and fins because of easy identification of fluorescent cells in these tissues. When we segregated such fish having EGFP⁺ cells and conducted genotyping analysis: 89% (102/114) was found to be vhl^+/− carriers (Supplementary Table S2). The activation of phd3::EGFP-mediated GFP fluorescence in random cell populations in the treated fish suggested DMBA-induced loss of the wild-type vhl allele in these cells. Expectedly, the frequency of occurrence of EGFP⁺ cells is significantly higher in the vhl^+/− heterozygotes where only a single wild-type copy needs to be lost.

All the DMBA challenged fish were grown to adulthood and were constantly monitored for tumor formation. Seven to eight months after the DMBA treatment, we observed a small group of fish in all 4 treated batches to exhibit strong EGFP fluorescence mainly in the trunk region as well as in the region adjacent to cloaca (Fig. 4H and I). At the end of 9 months, genotyping and statistical analysis of fish exhibiting strong EGFP fluorescence in all the 4 batches revealed a significantly (P < 0.002, Chi-square test) greater number (23/27; Supplementary Table S3) of them to be vhl^+/− heterozygotes.

After growing all the experimental fish for 14 months, they were culled and processed for histologic analysis. Serial sagittal sections were made for every sample and subsequently immunostained for the proliferation marker, PCNA. Screening of the PCNA-stained slide preparations showed a significant increase (P < 0.02, Chi-square test) in the occurrence of neoplastic growth in the DMBA-treated vhl^+/− samples (59%) than the similarly treated wild-type samples (23%; Table 1). In many cases, the tumor formation was found in the liver, bile duct, and gut tissues. Importantly, the analysis also showed that the fluorescing tissues in all strongly EGFP⁺ fish to correlate with neoplasias.

The wild-type adult zebrafish intestine showed a regular arrangement of villi and the proliferating epithelial cells were observed to be strictly restricted to intervillus pockets (Fig. 4J; refs. 35 and 36). However, in the DMBA-treated vhl^+/− fish, we observed a disorganized pattern of intestinal villi architecture (n = 8; Fig. 4K) and these villi were completely encompassed by PCNA⁺ proliferating epithelial cells (Fig. 4K) analogous to adenoma formation in the human gut. A few of the gut tumors showed abnormally proliferating intestinal epithelial cells in one half of the gut whereas the other half showed restricted presence of PCNA⁺ cells (Fig. 4L and M). A few samples exhibited lesions wherein PCNA⁺ epithelial cells were scattered all through the intestinal villi (Fig. 4M).

The liver tissue in zebrafish consists of hepatocytes, bile ducts, and portal vessels (35). In the wild-type liver tissue usually very few proliferating cells are present (Fig. 4N), but in the DMBA treated vhl^+/− fish we observed the presence of highly proliferating hepatocytes (n = 18) in the liver tissue (Fig. 4O). These tumors were identified histologically as primary hepatocellular carcinoma, wherein either the whole liver or selected regions of liver tissue showed strong staining for PCNA (Fig. 4P and Q). These liver neoplasias showed altered hepatic architecture as well as abnormal cell morphology along with apoptotic characteristics such as presence of fragmented nuclei (Fig. 4R). Similarly, compared with the wild-type bile duct tissue, which were not actively proliferating, the DMBA-treated fish possessed enlarged, abnormally shaped bile ductules as well as cysts. These bile ducts exhibited the presence of PCNA⁺ epithelial cells as well as showing abnormal growth of these ducts leading to the development of cholangiocarcinoma (Fig. 4S).

In addition, we also observed occurrence of testicular (Supplementary Fig. S7A) and ovarian (Supplementary Fig. S7B) tumors in the DMBA treated wild-type and vhl^+/− batches at equal frequencies.

Because ccRCC originates from renal epithelial cells and the human VHL gene is observed to exhibit strong tumor suppressor activity in renal tubules, we asked whether the DMBA treatment of vhl^+/− fish could trigger proliferation of renal tubular epithelial cells. Hence, we screened the kidney regions in the PCNA-stained slide preparations of DMBA treated vhl^+/− and wild-type groups. In the untreated samples (both wild-type and vhl^+/−), the hematopoietic cells in the kidney strongly expressed the proliferation marker, PCNA, and the epithelial cells of the tubules are found to be mostly quiescent. However, the DMBA treatment is observed to induce the transformation of the epithelial cells of renal tubules to a proliferative state as evidenced by robust PCNA⁺ staining (Fig. 5A–F). A normal renal tubule typically consists of a single organized layer of epithelial cells but, in contrast, most of the proliferating renal tubules exhibited abnormal organization of epithelial cells (Fig. 5A–F) as well as altered cellular morphology (Fig. 5C and E). Statistical analysis showed a highly significant increase in the number of proliferative renal tubules in DMBA treated vhl^+/− fish compared with the controls (26/30; P < 0.0002, Chi-square test; Supplementary Table S4). These observations suggest that DMBA can induce strong proliferation of renal epithelial cells in the vhl^+/− fish.

The observation that after DMBA treatment, ∼85% of the fish with EGFP⁺ tumors belonged to the vhl^+/− group suggested that such tumors might have originated from vhl null cells and hence most of the tumor cells might be possessing vhl^−/− genotype. To address this question, we isolated tumor cells, by laser capture microdissection (LCM), from paraffin tissue sections of 5 vhl^+/− EGFP⁺ tumor samples, extracted genomic DNA and conducted standard sequencing analysis. But, we were unable to identify unambiguously the loss of wild-type vhl allele in these samples, either at the vhl^hu2117 locus or elsewhere in the coding region (Supplementary Fig. S8).

To extend this analysis further, we wanted to quantify the levels of vhl mRNA in the vhl^+/− EGFP⁺ tumor cells. We dissected tumor tissue from 5 different strongly fluorescing EGFP⁺ fish (Supplementary Fig. S9A and S9B) selected from a DMBA treated vhl^+/− group, isolated total RNA and conducted quantitative real-time PCR analysis. In 4 of 5 tumor samples, we observed a medium to strong decrease in the vhl mRNA levels compared with control (Supplementary Fig. Fig. S9C), suggesting that in these samples the wild type Vhl function could be lost/reduced either by mutation(s) or by epigenetic inactivation. Because of technical limitations, we could not assay potential molecular lesion(s) on the wild-type vhl allele in these neoplastic cells with decreased vhl mRNA levels. The fact that DMBA treatment led to tumorigenesis in 55 of 94 vhl^+/− fish compared with neoplasias in only 18 of 77 experimental wild-type fish strongly indicate that the Vhl possesses a conserved tumor suppressor function in zebrafish.

In this study, we have generated and validated a novel reporter line, Tg(phd3::EGFP), for tracking VHL/HIF signaling in zebrafish. When we subjected the transgenic embryos to hypoxia induced by either genetic, physical, or chemical conditions, we could observe strong activation of the transgene. We have also showed the potential of the transgenic line as an in vivo reporter for intratumoral hypoxia in a melanoma model system. These phd3::EGFP transgenic embryos would be ideally suited for small molecule screens to identify novel chemicals that activate or suppress Hif activity in the wild-type or vhl^−/− background, respectively.

Several approaches are being utilized to monitor intratumoral hypoxia such as direct measurements of tumor oxygenation, assaying the levels of hypoxia marker proteins such as carbonic anhydrase IX, lysyl oxidase, etc. and by using small molecule markers such as EF5, pimonidazole, which are derivatives of 2-nitroimidazole, but many of these procedures are invasive and time consuming (37, 38). Moreover, these procedures and/or molecular markers have not been successfully employed for measuring hypoxia in zebrafish (39), indicating the need for novel markers for hypoxic signaling. The low background expression and powerful induction by HIF could make the Tg(phd3::EGFP)^i144/i144 fish a novel fluorescent marker for in vivo monitoring of hypoxic signaling in many zebrafish tumor model systems.

von Hippel-Lindau disease is characterized by formation of a variety of tumor types, most importantly clear cell renal cell carcinoma and retinal and central nervous system hemangioblastomas. Unfortunately, although a variety of mouse models have been made for this disease, a comparable tumor suppressor activity of VHL homologs has been difficult to show in rodent models. Therefore, we tested for such a function of vhl in the zebrafish. Indeed, we report increased susceptibility of the vhl^+/− zebrafish to hepatic, bileduct, and intestinal tumorigenesis upon exposure to dimethylbenzanthracene (DMBA) compared with the control fish. We also report significantly increased occurrence of proliferating renal tubules in the DMBA-treated vhl^+/− batch compared with appropriate controls.

We tracked for random occurrence of vhl LOH in live Tg(phd3::EGFP)^i144/i144; vhl^+/− fish by scoring for cells and/or cell clones emitting strong EGFP fluorescence. Although the 6-month-old untreated vhl^+/− fish had approx. 1 to 2 detectable EGFP⁺ cells in the skin and retinal epithelia, as they become older (∼1.5–2-yr-old) they accumulated few more LOH events in the skin epithelium (∼ 8–10 GFP⁺ cells/fish). Upon treatment with DMBA, the vhl^+/− fish showed a significant increase in the occurrence of vhl^−/− EGFP⁺ cells in the skin epithelium.

Probably owing to the low rate of spontaneous vhl LOH, vhl^+/− heterozygous fish exhibited normal development and showed no detectable signs of tumor or cyst formation. In the case of Vhlh^+/− heterozygous mice, depending on the type of genetic background, variable susceptibility to spontaneous tumorigenesis were observed (refs. 40, 41, and 42). Because many of the zebrafish laboratory strains are not highly inbred (43), the strain differences might be less pronounced in fish. However, the lifespan of the zebrafish is only a few years, so even after the spontaneous vhl LOH at few cells they may not be undergoing enough subsequent mutations in other genes to predispose such fish with a disease phenotype. But, when we challenged the vhl^+/− fish with DMBA, we observed a high incidence of tumorigenesis arising from liver, bile duct, and gut.

Interestingly, we observed decreased vhl mRNA levels in a set of vhl^+/− EGFP⁺ tumors, suggesting that in these samples the wild-type vhl allele might have undergone either loss-of-function mutation(s) or epigenetic silencing. Interestingly in human VHL patients, an epigenetic mechanism for inactivation of the wild-type VHL allele, mediated through the transcriptional silencing of the gene via hypermethylation of a 5′-CpG island, has been frequently observed (8, 44). DMBA in addition to inducing mutations is also known to cause aberrant methylation of 5′-CpG islands leading to epigenetic silencing of gene(s) (45). Our study exhibit the potential of the vhl^+/− zebrafish as a model to characterize molecular mechanisms involved in VHL induced tumorigenesis.

Although occurrence of ccRCC is the classical VHL disease phenotype, we did not observe development of renal cysts or ccRCC in the DMBA-treated vhl^+/− fish within the 14-month experimental period of our analysis. However, the presence of a significantly increased number of PCNA⁺ renal tubules suggests that these proliferating tubules might be at an early stage in the process of renal cyst formation. Some of these proliferating tubules were dysplastic and might be comparable to the VHL mutant foci that are observed in VHL patients (46). Interestingly, the presence of a duplicated copy of the vhl gene, a vhl-like (vll) gene in zebrafish, indicates a possible segregation of VHL functions between these 2 genes and it would be interesting to assess the combined role of these genes in renal tumor suppression.

No potential conflicts of interest were disclosed.

Conception and design: K. Santhakumar, S. McKee, F.J.M. van Eeden

Development of methodology: K. Santhakumar, S. McKee, S. Elworthy, E. van Rooijen, S.A. Renshaw, F.J.M. van Eeden

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Santhakumar, E.C. Judson, P.M. Elks

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Santhakumar, E.C. Judson, P.M. Elks, S.A. Renshaw, S.S. Cross, F.J.M. van Eeden

Writing, review, and/or revision of the manuscript: K. Santhakumar, S. McKee, E. van Rooijen, S.S. Walmsley, S.A. Renshaw, S.S. Cross, F.J.M. van Eeden

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.C. Judson

Study supervision: F.J.M. van Eeden

The authors thank Adam Hurlstone (The University of Manchester, UK) for providing V12RAS::mCherry construct; Yvonne Stephenson and Christopher Hill for help in histology; Louise Koblitz and Bernd Pelster (University of Innsbruck, Austria) for Hif-1α antibody; and aquarium facility staff for excellent fish care.

This work is supported by the following grants: Cancer Research UK (CRUK): C23207/A8066, European Commission FP7: HEALTH-F4-2010-242048, and Royal Society: 2006/R1. The CDBG is supported by a MRC grant G0700091.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.