VHL has proved a unique and enigmatic tumor suppressor gene. Most of the tumor suppressor genes associated with familial cancer syndromes can be assigned specific cellular functions; for example, Rb is a cell cycle inhibitor, WT1 and TP53 are transcription factors, adenomatous polyposis coli is a Wnt signaling suppressor, and NF1 and NF2 are GTPase-activating protein regulator and cytoskeletal protein merlin, respectively. The role of VHL in tumorigenesis is more complex. The protein encoded by the VHL gene (pVHL) is best known as the substrate-binding subunit of a SCF (Skp1-Cdc53/Cul-1-F-box protein) type E3 ubiquitin ligase containing, besides pVHL, Cullin-2, elongin B and C and Rbx-1 (Figure 1; Pause et al., 1997; Lonergan et al., 1998; Kamura et al., 1999; Lisztwan et al., 1999; Stebbins et al., 1999). The best-known degradation target of VHL-containing E3 ligase is the α-subunit of hypoxia-inducible factor (HIF-α) in normal physiological conditions (Kaelin, 2005). At normal oxygen level, HIF-α is hydroxylated at the proline residues within an oxygen-dependent degradation domain. The prolyl-hydroxylated HIF-α is recognized by pVHL, leading to poly-ubiquitination and degradation. The hydroxylation reaction is mediated by the prolyl-hydroxylase domain proteins (PHDs; Berra et al., 2003). In hypoxic conditions, the prolyl hydroxylases are inactive and HIF-α-subunit is stabilized. HIF-α then dimerizes with the β-subunit (HIF-β), also termed aryl hydrocarbon receptor nuclear translocator (ARNT), which is constitutively expressed. The HIF heterodimer translocates to the nucleus where it functions as a transcription factor. Its best-known target genes encode proteins involved in glycolysis (phosphoglycerate kinase), glucose transport (Glut-1), angiogenesis (vascular endothelial growth factor (VEGF)) and erythropoiesis (erythropoietin); that is, proteins that mediate the cellular response and adaptation to hypoxic conditions. In addition, chemokine receptor CXCR4 was also identified as a HIF target (Staller et al., 2003), which indicates that HIF activation may contribute to the metastatic potential of cancer cells. These functions support a critical role of pVHL in regulating tumor progression, especially in hypervascularized tumors such as ccRCC. However, accumulated evidences have indicated that many HIF-independent activities of pVHL also exist (Figure 1), including regulation of extracellular matrix (Ohh et al., 1998; Tang et al., 2006; Feijoo-Cuaresma et al., 2008; Kurban et al., 2008), senescence (Young et al., 2008), apoptosis (Guo et al., 2009), phosphorylation enhancer (Yang et al., 2007), microtubule stability and cilia formation (Hergovich et al., 2003; Esteban et al., 2006; Schermer et al., 2006; Thoma et al., 2007, 2009), RNA stability (Datta et al., 2005; Danilin et al., 2009), endocytosis (Hsu et al., 2006), gene transcription (Mikhaylova et al., 2008) and many others (Frew and Krek, 2007, 2008). Interestingly, many of these HIFindependent functions of pVHL are mediated through stabilization of its binding proteins, contrary to its known E3 ligase activity. It therefore appears that pVHL is an adaptor protein that, depending on the interacting partners, can promote protein degradation or serve as a chaperon. Some of these diverse activities likely also contribute significantly to the tumor suppressor and other physiological functions.

The most significant early finding on pVHL function is the demonstration of its role in targeting prolyl-hydroxylated HIF-α (Ivan et al., 2001; Jaakkola et al., 2001). The VHL–HIF-prolyl hydroxylase triad was soon validated in C. elegans (Epstein et al., 2001). This study provided genetic and biochemical proof that prolyl-hydroxylated HIF-α is regulated by pVHL. More importantly, the study identified the nematode prolyl hydroxylase (encoded by the 2-oxoglutarate-dependent oxygenase gene egl-9) as the mediator of HIF-α degradation, as well as the mammalian ortholog EGLN/PHD (egl-9 homolog/prolyl hydroxylase domain). This cross-species validation was critical for the rapid acceptance of the model. Subsequent C. elegans studies of note came from whole-organism analysis of gene expression profiling. The C. elegans genome encodes a single VHL (vhl-1) and a single HIF-α (hif-1) gene. Interestingly, both vhl-1 and hif-1 loss-of-function mutants as well as double mutants are viable. The viability of these mutants made it possible to compare altered gene expression patterns in vhl-1-defective animals in the otherwise wild-type or hif-1 mutant backgrounds (Bishop et al., 2004). Thus, hif-dependent and independent vhl-responsive genes could be identified. Most of the hif-dependent genes upregulated in vhl-1 mutant are also upregulated by hypoxia. These include egl-9 that targets HIF-α during normoxia, indicating a possible feedback control. This mechanism is evolutionarily conserved because the mammalian ortholog of egl-9, the EGLN/PHD family of genes, was also upregulated in VHL mutant ccRCC cells (Harten et al., 2009). Another prolyl hydroxylase (encoded by phy-2) that modifies collagen is upregulated in both vhl-1 mutant and hypoxia. Curiously, no other familiar mammalian VHL–HIF target genes were revealed, such as homologs of VEGF, IGFBP3, Glut-1 and CXCR4, suggesting a different requirement for coping with hypoxia in nematodes as compared with mammals. Interestingly, among the vhl-dependent but hif-independent targets, one encodes an aldose 1-epimerase that is involved in gluconeogenesis, and three encode the solute carrier (slc) family of proteins that may be involved in sugar and lactate transport. These Vhl-dependent genes imply that the ancestral Vhl function may be involved in modulation of metabolic homeostasis.

A second whole-genome analysis focused on the hypoxia response (Shen et al., 2005). The analysis was designed to identify early response genes in acute hypoxia (4 h at 21°C in 0.1% oxygen). The analysis also compared three genotypes, wild type, hif-1 mutant and vhl-1 mutant, to identify hif-dependent and independent hypoxia-responsive genes. Among the hypoxia-inducible genes, only slightly more than half (63 out of 110) are hif-dependent. Most of the hif-dependent hypoxia-inducible genes are also vhl regulated, that is, they are upregulated in vhl-1 mutant and cannot be further induced by hypoxia. This indicates that in C. elegans, vhl is the major regulator of hif in normoxic cells and that hypoxia is the main activator of hif function. Prominent among the hypoxia-induced genes encode components of energy metabolism. For example, pyruvate carboxylase (pyc-1) is induced in a hif-1-independent (and vhl-dependent) manner. The enzyme provides oxaloacetate precursor for gluconeogenesis. gei-7 is induced in a hif-1- and vhl-dependent manner. It encodes isocitrate lyase that coverts isocitrate to succinate and glyoxylate in the glyoxylate cycle, a carbon-conserving pathway alternative to the carbon-consuming tricarboxylic acid cycle that converts isocitrate to succinyl-CoA. It is interesting to note that these gene products either promote energy store (glucose production) or conserve carbon sources, instead of the glycolysis switch observed in mammalian Vhl mutant (or hypoxic) cells. This may reflect the fact that there is no thermal homeostasis in nematodes, therefore, emergency energy production is not a priority. As in vhl-1 mutant gene profiling (Bishop et al., 2004), both egl-9 and phy-2 were upregulated in hypoxia in a hif-1- and vhl-1-dependent manner. Interestingly, phy-2 mutation shows <10% survival (to adults) in 0.5% oxygen. Thus, modification of extracellular matrix may be important for maintaining cellular function during hypoxic stress. Taken together, these studies demonstrated that the VHL–HIF axis is evolutionarily conserved to serve as a hypoxia responder, although the exact hypoxic response at the molecular level is specific to individual species.

The first phenotypic analyses (Adryan et al., 2000) utilized the then new but promising RNA interference technology. dVHL was in fact one of the first Drosophila genes studied using siRNA-mediated knockdown. Drosophila contains a vascular system, the trachea, for transporting oxygen, which is capable of oxygen sensing (Jarecki et al., 1999; Wingrove and O’Farrell, 1999). On the other hand, the Malpighian tubes perform excretory and osmoregulatory functions similar to those of the vertebrate kidneys. There was therefore much speculation as to which of these systems will be affected in dVHL mutant. As it turned out, injection of dVHL-specific siRNA into pre-blastoderm embryos resulted in ectopic looping and branching in the trachea, but not in the Malpighian tubes (Adryan et al., 2000). Importantly, injection of either dVHL or human VHL sense RNA into wild-type embryos blocked tracheal branch migration. These reciprocal experiments demonstrated that VHL could negatively regulate vascular cell motility.

The function of VHL was further explored in the Drosophila system by analyzing the genomic knockout of the endogenous dVHL gene (Hsouna et al., 2010). Homozygous dVHL mutant is lethal at the early larval stage and exhibits profound tracheal phenotypes, consistent with the earlier RNA interference study. Surprisingly, the lethality could be rescued by re-expressing dVHL specifically in the trachea, indicating that modulation of vascular development is the main developmental function of dVHL. This notion is consistent with the earlier mouse knockout phenotype (Gnarra et al., 1997) and was later confirmed in zebrafish (van Rooijen et al., 2010). In both flies and vertebrates, the key defining process in vasculogenesis is the outgrowth of branches, regulated by growth factor signaling pathways. In Drosophila embryo, the fibroblast growth factor (FGF) signaling pathway is utilized (Ghabrial et al., 2003; Uv et al., 2003; Akis and Madaio, 2004) to pattern the trachea in a stereotyped fashion. The tracheal phenotype in dVHL mutant resembled that resulting from over-active FGF receptor (FGFR) signaling (Dammai et al., 2003). Indeed, FGFR-ERK-Ets1 signaling pathway is over-activated in dVHL mutant tracheal cells, which is the result of accumulated FGFR on the mutant tracheal cell surface (Hsouna et al., 2010). Genetic epistasis analysis indicates that the surface accumulation of FGFR is the result of defective endocytic pathway and that dVHL functionally interacts with and stabilizes Abnormal Wing Discs (AWD), the homolog of human anti-metastasis factor Nm23, which has been shown to regulate endocytosis (Dammai et al., 2003; Hsu et al., 2006; Nallamothu et al., 2008, 2009; Woolworth et al., 2009) in part by stabilizing Rab5 (Woolworth et al., 2009). The detailed phenotypic analysis also revealed that the endocytic function of dVHL not only regulates cell motility but also controls the size and length of the tubule lumen (Hsouna et al., 2010). Formation of the tracheal lumen, and likely of other tubule systems in mammals, is controlled by secretion (exocytosis) of matrix material and, at the maturation phase, reabsorption through endocytosis of these structural materials (Behr et al., 2007; Tsarouhas et al., 2007). Mutations in dVHL result in defective endocytosis of the lumenal chitin and consequently, enlarged and tortuous lumen of the trachea. This phenotype can be reduced or exacerbated by expression of constitutively active Rab5 or dominant-negative Rab5, respectively. Interestingly, ectopic branching and dilated tubule phenotypes were also observed in the mouse Vhl knockout kidney tubules (Hsouna et al., 2010).

We also showed that specific regulation of FGFR internalization is evolutionarily conserved. In VHL mutant human ccRCC cells and primary human endothelial cells knocked down with VHL-specific RNA duplex, FGFR over-accumulates on the cell surface, leading to increased FGFR, ERK and Ets1 transcription factor activation, elevated cell motility and increased angiogenic potential in vitro (Hsu et al., 2006; Champion et al., 2008). In addition, heterozygous Vhl knockout mouse shows increased angiogenic response toward bFGF in vivo (Champion et al., 2008). Importantly, this endocytic function is at least partly independent of HIF function, as Hif-α knockdown could not rescue the cell motility phenotypes in multiple cell systems. However, pVHL has recently been shown to regulate endocytosis of EGF receptor via HIF-dependent suppression of a Rab5 effector, rabaptin-5, in ccRCC cell lines (Wang et al., 2009). It is not known whether these different mechanisms are specific for different surface receptors.

In a separate study (Mortimer and Moberg, 2009), the tracheal branching phenotype in dVHL knockdown flies (mediated by trachea-specific expression of dVHL shRNA) was shown to be partly the result of over-expression of sima, the Drosophila homolog of HIF-1α. sima in turn stimulates the transcription of btl (FGFR), leading to ectopic branching. sima-dependent btl transcription has also been demonstrated in terminal branching in the larval trachea, which is induced by localized hypoxia (Centanin et al., 2008). Thus, both canonical and non-canonical functions of dVHL are involved in the tubule formation in Drosophila.

The role of dVHL in regulating motile epithelial cells was further studied in Drosophila border cells (Doronkin et al., 2010), which have been considered an in vivo model for epithelial cell invasion and epithelial–to–mesenchymal transition (Duchek et al., 2001; McDonald et al., 2003). During oogenesis, a specialized group of follicular epithelial cells (border cells) delaminates from the epithelium and invades through the germ cell complex until they reach the anterior end of the oocyte (Rorth, 2002; Montell, 2003). The precise movement of the border cells is guided by the Drosophila platelet-derived growth factor/VEGF signaling pathway. Interestingly, Doronkin et al. (2010) showed that the dVHL-sima system is involved in border cell migration. However, the exact correlation between border cell migration and dVHL function appears to be complicated, because the extents of hypoxia and the expression levels of sima or dVHL could cause either slowed or accelerated migration. Nonetheless, this study presented the first in vivo demonstration of the role of pVHL in regulating epithelial cell invasion.

Because VHL mutant ccRCC involves pathological transformation of epithelial tissues, whether and how VHL regulates epithelial morphogenesis has been of great interest. One useful epithelial model is again the Drosophila follicular epithelium in the egg chamber (Duchi et al., 2010). Epithelial cells are characterized by asymmetrical specification of membrane domains. One crucial step in establishing epithelial polarity is the specification of the apical domain, which is defined by the localized accumulation of a complex containing atypical PKC (aPKC), Bazooka (Baz; mammalian and worm PAR-3) and PAR-6 (Suzuki and Ohno, 2006). The PAR complex is required for subsequent localization of the basolateral complex that consists of known tumor suppressors Discs Large, Lethal Giant Larvae and Scribble (Bilder et al., 2003; Betschinger et al., 2005), and the adherens junction (Yamanaka et al., 2003). In mutant dVHL clones of follicle cells, microtubule bundles are disrupted and aPKC is mislocalized, leading to adenoma-like piling up of the epithelium (Duchi et al., 2010). Consistent with the known human VHL activity (Hergovich et al., 2003), wild-type dVHL can bind to microtubules. The direct influence of microtubule stability by dVHL leading to aPKC localization is supported by the ex vivo culture of dVHL mutant egg chambers, in which the aPKC mislocalization and epithelial phenotypes could be rescued by treatment with microtubule stabilizing agent paclitaxel, whereas treatment of wild-type egg chambers with microtubule destabilizing agent nocodazole could recapitulate the dVHL mutant phenotypes (Duchi et al., 2010). Furthermore, although wild-type dVHL could rescue the phenotypes, the type 2A mutant Y98H (Y51H in Drosophila), which has been shown to lose the microtubule-stabilizing function but retain partly the HIF-degradation function (Hergovich et al., 2003), could not. Therefore, the initial apical localization of aPKC requires intact microtubule bundles, which is maintained by dVHL protein (Figure 2). The Drosophila study is the first demonstration of physiological and developmental significance of the microtubule-stabilizing function of VHL in vivo.

More detailed phenotypic analyses revealed that the homozygous mutants display increased blood vessel formation (van Rooijen et al., 2010). Significant increase in neovascularization was observed in the brain, eye and trunk. These abnormal vessels are leaky and can lead to macular edema and retinal detachment. Increased neovascularization is at least in part mediated by increased VEGF signaling, because treatment with VEGFR tyrosine kinase inhibitors could relieve the phenotypes. One important novel observation in the fish model is the increased activity of hematopoietic stem cells (HSCs), which not only increases the number of red blood cells, but also results in increased levels of endothelial progenitors in peripheral blood. This is highly interesting because in human hemangioblastoma, loss of heterozygosity of VHL occurs in stromal cells neighboring the proliferating blood vessels (Vortmeyer et al., 1997) and these VHL mutant stromal cells display multipotent cell characteristics (Park et al., 2007). The fish model may provide a clue to the origin of these mutant stromal cells.

The next knockout was a conditional allele (Haase et al., 2001). Surprisingly, both heterozygous null (one lox allele knockout in embryonic stem cells) and hepatocyte-specific knockout (using albumin promoter-directed Cre) resulted in cavernous hemangiomas of the liver, which can be correlated with increased level of HIF-2α and increased expression of VEGF. In addition, hepatocyte-specific knockout mutant mice developed polycythemia due to increased erythropoietin production in the liver. Although no hyperplasia was induced in these early Vhl knockout models, they did provide the in vivo confirmation of the VHL–HIF pathway. Subsequently, the first kidney proximal tubule conditional knockout was generated using the phosphoenolpyruvate carboxykinase (PEPCK) driver. Proximal tubule-specific knockout was attempted because proximal tubules have long been considered the origin of ccRCC. Somewhat surprisingly, this model generated only modest phenotype of renal microcysts in about 25% of >12-month old mice but no tumors. This was nonetheless significant because cysts are believed to be the precursor of ccRCC. It is formally possible that other genetic factors are needed for tumorigenesis in Vhl mutants. On the other hand, a recent seminal study suggests that proximal tubule cells may not be the origin of ccRCC, because in human VHL patient samples, VHL mutant proximal tubule cells do not show capacity to proliferate (Mandriota et al., 2002). More recently, Ksp1.3-Cre (expressed in distal tubules, ascending and descending loops of Henle and collecting ducts, but rarely in proximal tubules) was used to knockout Vhl outside of the proximal tubule domain (Frew et al., 2008b). This conditional knockout alone yielded hydronephrosis but no abnormalities in the tubule epithelium. However, when combined with another tumor suppressor gene knockout Pten, the kidney displayed, in addition to hydronephrosis, hyperproliferation of the urothelium and high penetrance of enlarged kidneys due to multiple epithelial tubule cysts in the cortex and medulla. This model strengthened the notion that loss of Vhl alone is insufficient for tumor initiation. Similarly, Vhl-Pten double conditional knockout in male genital track generated cystadenoma and squamous metaplasia (Frew et al., 2008a).

The vascular function of VHL was verified using endothelial cell-specific (Tie2-Cre-driven) knockout (Tang et al., 2006). The endothelial knockout also confirmed the defect in fibronectin deposition previously demonstrated in cell culture (Ohh et al., 1998; Feijoo-Cuaresma et al., 2008). Acute tamoxifen-induced mosaic inactivation of the Vhl gene at E10.5 resulted in embryonic lethality between E14.5 and E15.0 with extensive hemorrhage and necrosis (Hong et al., 2006). Liver damage and placental abnormalities were also observed, which confirmed the phenotypes of the previous Vhl knockouts. A number of subsequent conditional knockouts in other organs have shown defects in spermatogenesis, thymus cell survival and bone development due to reduced cell proliferation and/ or increased cell death (Ma et al., 2003; Biju et al., 2004; Pfander et al., 2004). More recently, Vhl has also been shown to influence the activity of HSCs (Takubo et al., 2010), which reside in hypoxic niches in the bone marrow and maintain a quiescent state. In Hif-1α-conditional allele (driven by interferon-inducible Mx1-Cre) mice, the HSCs lost cell cycle quiescence and their numbers decreased during various stress settings including bone marrow transplantation. Over-stabilization of HIF-1α in conditional biallelic loss of Vhl induced cell cycle quiescence in HSCs and their progenitors, as expected, but unexpectedly also resulted in impairment in bone marrow transplantation capacity, probably due to over-quiescence. In contrast, heterozygous loss of Vhl induced cell cycle quiescence in HSCs but improved (over-wild type) bone marrow engraftment after transplantation. It is possible that VHL modulates HSC activity by both HIF-dependent and independent mechanisms, and that the precise levels of its target proteins are critical. In this regard, it will be important to take into account the heterozygosity of VHL in the VHL disease patients, especially concerning the hematopoietic and vascular components of the disease manifestations.

The metabolic imbalance in VHL mutant cells due to HIF-mediated glycolic switch has attracted intensive interest, because this may present a new therapeutic approach by forcing oxidative metabolism on the mutant cells. A recent knockout model provided the physiological proof of glycolic switch in Vhl mutant cells (Zehetner et al., 2008). In this study, Vhl was conditionally knocked out in β cells of the pancreas. The mice developed impaired systemic glucose tolerance. This, intriguingly, can be attributed to the false sense of glucose shortage by the β cells because these Vhl mutant cells undergo glycolic switch and respond by increasing insulin secretion. In another study (Shen et al., 2009), intending to model the pancreatic cystadenoma of VHL disease, Vhl knockout in either α or β cells did not result in cysts formation. However, when Vhl was knocked out in the pancreatic progenitor cells using Pdx1-Cre, highly vascularized microcystic adenomas and hyperplastic islets did develop, thus representing the success modeling of one VHL disease-associated tumor phenotype.

One other intriguing disease model is the human polycythemic Chuvash disease. These patients inherit a homozygous VHL mutation that is incapable of down-regulate HIF-α, but they show no increased tumor incidences (Gordeuk et al., 2004). A mouse model for the Chuvash disease was generated by knocking in the R200W (R166W in mouse) mutant Vhl allele (Hickey et al., 2010). This mouse recapitulates the polycythemic phenotype of the human disease and exhibits pulmonary hypertension and lung fibrosis in a Hif-2α-dependent manner. The latter phenotypes also manifest in some Chuvash patients (Bushuev et al., 2006; Smith et al., 2006), which can be induced in hypoxic conditions (Stenmark et al., 2006). The fibrotic phenotype recalled an earlier knockout strain specific for the podocyte in kidney, which exhibited glomerulomegaly and occasional glomerulosclerosis (Brukamp et al., 2007). Interestingly, the earlier PEPCK-driven Vhl knockout could be induced to develop renal fibrosis after subtotal nephrectomy of one kidney and complete removal of the other (Kimura et al., 2008). In addition, hypoxia and increased Hif-1α activity have been linked to kidney fibrosis in mouse (Higgins et al., 2007). The connection between Vhl mutant cells and inflammation is worthy of more in-depth investigation, because prolonged inflammation can promote proliferation through the action of secreted cytokines, and importantly, can induce genetic changes in pre-cancerous cells by induction of reactive oxygen species or by oxidative inactivation of mismatch repair enzymes (Hussain et al., 2003; Colotta et al., 2009; Kraus and Arber, 2009; Grivennikov et al., 2010).