Abstract

1. Introduction

The von Hippel–Lindau gene (VHL) was initially identified as the tumor suppressor gene that is mutated in patients with von Hippel–Lindau disease, a rare familial tumor syndrome characterized by the development of highly vascularized tumors in multiple organs. Typical clinical manifestations of VHL disease include hemangioblastomas of the retina and central nervous system, multifocal and bilateral renal cell cancer of the clear cell type often preceded by preneoplastic renal cysts, pancreatic cysts and tumors, pheochromocytomas and others [1]. More importantly, VHL was also found to be mutated in the majority of sporadic clear cell renal cell carcinomas (CC-RCC) indicating that loss of VHL function is a critical event during renal carcinogenesis [2]. This notion was supported by experiments in which reintroduction of wild type VHL into VHL deficient renal cancer cell lines resulted in suppression of tumor growth in vivo and in vitro [3–5]. The mechanisms by which the VHL gene product, pVHL, carried out its tumor suppressor function were initially obscure as homologies to known proteins did not exist when the gene was identified by positional cloning in 1993 [6]. Observations that oxygen dependent regulation of hypoxia inducible genes such as vascular endothelial growth factor (VEGF) and glucose transporter 1 (GLUT-1) was lost in VHL deficient cell lines and could be restored after reintroduction of wild type pVHL suggested a role for pVHL in oxygen sensing [4,7,8]. After pVHL was found to be associated with the elongins B and C, cullin2 and Rbx [9–14] in a complex that had E3 ubiquitin ligase activity [15,16], Maxwell et al. reported in a seminal paper that pVHL was critical for targeting the oxygen sensitive α-subunit of hypoxia-inducible factor (HIF) for oxygen dependent proteolysis [17], thus providing a direct link between VHL associated tumorigenesis and cellular adaptation to hypoxia via HIF. While it had already been reported that HIF-α was degraded by the ubiquitin-proteasome [18], pVHL mediated polyubiquitylation of HIF-α was subsequently established by several groups [19–22].

In order to understand the role of pVHL mediated HIF proteolysis in normal tissue physiology, tumorigenesis and development, we have used cell type specific gene targeting based on Cre/loxP mediated recombination. This review article will, after providing a brief synopsis of pVHL functions, summarize the findings from a variety of developmental and physiological studies with conditional VHL knock out mice and discuss their relevance for the successful generation of a VHL tumor model.

2. pVHL is part of a hypoxic switch

As a normal physiological response to hypoxia, HIF-1 and -2 facilitate both oxygen delivery and adaptation to oxygen deprivation by regulating genes that are involved in glucose uptake and metabolism, angiogenesis, erythropoiesis, cell proliferation and apoptosis [23,24]. HIFs belong to the PAS (Per-ARNT-Sim) family of basic helix-loop-helix (bHLH) transcription factors and bind DNA as heterodimers composed of an oxygen-sensitive α-subunit and a constitutively expressed β-subunit, also known as the arylhydrocarbon receptor nuclear translocator (ARNT) or HIF-1β [25]. HIF-1α is ubiquitously expressed [23,24,26], while expression of HIF-2α seems to be restricted to certain cell types, including endothelial cells, cardiomyocytes, hepatocytes, renal interstitial cells but not differentiated renal epithelial cells, type II pneumocytes and glial cells [27–29].

All three known HIF alpha subunits, HIF-1α, -2α and -3α have been shown to interact with pVHL [17,30]. This highly conserved interaction requires hydroxylation of defined HIF-α proline residues by HIF specific proline hydroxylases (PHDs) and is absolutely required for the execution of HIF proteolysis under normoxia [31–37]. Therefore, lack of pVHL results in maximal HIF-α stabilization, increased HIF transcriptional activity and up-regulation of HIF target genes such as VEGF, GLUT-1 and erythropoietin (EPO) irrespective of oxygen levels, explaining some of the clinical features associated with VHL deficient tumors; e.g. the high degree of vascularity or erythrocytosis. A second hypoxic switch operates in the carboxy terminal transactivation domain of HIF-α with the hydroxylation of an asparagine residue; in hypoxia asparagine hydroxylation is blocked and CBP/p300 recruitment is facilitated [38,39].

3. HIF independent functions of pVHL

Although significant erythrocytosis is only seen in a relatively small fraction of patients with VHL disease [40,41], it is the clinical hallmark of patients with Chuvash disease, a congenital form of polycythemia mostly caused by homozygosity for a specific amino acid substitution at pVHL codon 200 (R200W) [42]. Interestingly this mutation, although it results in increased HIF transcriptional activity, does not result in a predisposition to classic VHL associated tumors [43], suggesting that additional HIF independent mechanisms must be important in VHL associated tumorigenesis, in particular renal carcinogenesis. This notion is also supported by VHL genotype–phenotype correlations, as in the case of type 2C VHL disease, where patients develop pheochromocytomas despite the fact that mutated pVHL retains the ability to polyubiquitylate HIF-α [44,45].

Besides regulating the degradation of HIF-α subunits, pVHL has been shown to play a role in fibronectin extracellular matrix assembly and matrix turnover [46–49], microtubule stability [50], and in regulating the stability and/or activity of other proteins such as plant homeodomain protein Jade-1 [51,52] and atypical protein kinase C isoforms [53–57]. Other pVHL targets have been described, most notably transcription factor Sp1 [58], a KRAB-A domain protein, VHLak, repressing HIF transcriptional activity [59], de-ubiquitinating enzymes [60], the large subunit of RNA polymerase II [61] and the RNA-binding protein hnRNP A2 [62]. It is unclear at present, how other pVHL functions and recently discovered protein interactions contribute to the development of the VHL phenotype and to renal carcinogenesis, certainly, further investigations are needed.

4. pVHL is essential for normal embryonic development

Mouse embryos with homozygous deletion of Vhlh (gene symbol for murine VHL) die during midgestation between embryonic day (E) 10.5 and 12.5 and are characterized by a lack of placental vasculogenesis. In a study by Gnarra et al. [63], a neomycin selection cassette was used to replace most of the Vhlh coding sequence in 129 J-1 ES cells, thus generating a complete null allele. Although the placenta and embryos of Vhlh null mice appeared to develop normally until E9.5, a severe disruption and loss of normal placental labyrinth architecture (absent development of syncytiotrophoblasts) associated with hemorrhage was noted by E11.5 to 12.5, suggesting that abnormal placental development had led to the demise of Vhlh−/− embryos. Whether this phenotype is mediated by HIF has not been studied, but could be investigated by a direct phenotypic comparison of Vhlh−/− and PHD−/− knock out mice.

Independent of this important role in placental vascular development, conditional disruption of Vhlh in neuro-epithelial progenitor cells alone resulted in abnormal neuronal differentiation and embryonic lethality during late gestation (Haase et al., unpublished data and see Table 1), suggesting that proper pVHL function is not only required for vascular development of the placenta, but also for normal development of other tissues.

5. Mice which lack one copy of Vhlh are predisposed to the development of vascular tumors

While Gnarra et al. did not observe a phenotype in Vhlh+/− mice bred in a mixed C57BL/6-129 genetic background, we observed that Vhlh heterozygotes in a BALB/c-129 background were predisposed to the development of cavernous liver hemangiomas with high phenotypic penetrance by 15 months of age [64]. By contrast hepatic vascular tumors are rare manifestations of VHL disease in humans [65,66]. The discrepancy in mouse phenotypes implies a strong dependence on genetic background. In deed, Ma et al. showed that the incidence of hepatic cavernous hemangiomas in a BALB/c background was 88% but only 18% in a C57BL/6 background, possibly a result of polymorphic differences in certain modifier genes as has been shown for other knock out mice [67].

Although the molecular mechanism for the liver phenotype in Vhlh−/− mice has not been systematically investigated, it is most likely that, following Knudson’s two-hit hypothesis, inactivation of the remaining Vhlh wild type allele in hepatocytes by, for example, loss of heterozygosity or promoter methylation, resulted in the formation of cavernous hemangiomas through increased HIF dependent vascular growth factor production. Hepatocyte specific inactivation of Vhlh (Fig. 1) resulted in the same phenotype, thus providing evidence that hepatic cavernous hemangiomas in Vhlh heterozygotes are the result of pVHL loss in this cell type and not in other liver cell types, such as endothelial cells. Endothelial cells in this model do however respond to uncontrolled, constitutive production of vascular growth factors by hepatocytes and proliferate while expressing functional pVHL. This is mechanistically similar to human VHL hemangioblastomas, in which “stromal cells” and not endothelial cells have lost pVHL and represent the neoplastic component of this tumor [68]. The histogenesis of stromal cells is uncertain; however, there is evidence in the literature that these cells may be of angiogenic origin [69,70].

Open in a separate window

Fig. 1

Hepatic vascular tumors in mice with hepatocyte specific inactivation of Vhlh. (A and B) Gross photography of formalin fixed livers from hepatocyte-specific Vhlh knock out mice. Arrows in (A) point to cavernous hemangiomas that appear deflated after resection of the liver. The gallbladder is labeled by a star. (B) shows a section of the same liver. Arrows point to blood filled cavities containing fibrin clots. (C and D) Paraffin sections of livers with cavernous hemagiomas stained with H&E. (C) Large endothelial cell lined, blood filled cavities are depicted by stars. Arrow depicts an area of hemorrhage. (D) Cavernous hemangiomas are typically associated with macro- and microvesicular hepatocellular steatosis, which results from the accumulation of neutral lipid [64], here depicted by arrows. Lipid accumulation is also a feature of stromal cells, the neoplastic component of VHL associated central nervous system hemangioblastomas [69,70]. Magnification in (C) is ×100 and in (D) is ×400.

Other more classic VHL associated lesions such as renal cysts were only found at very low frequency in Vhlh heterozygotes; frequency of 1/30 in a study by Haase et al. [64] and 1/28 as reported by Kleymenova et al. [71]. In these reports renal cell carcinomas were not observed, nor were central nervous system hemangioblastomas or retinal angiomas. Treatment with streptozotocin did not result in increased susceptibility to renal carcinogenesis in Vhlh heterozygotes, but instead, in increased susceptibility to hemangiomas and hemangiosarcomas in the liver and other organs such as the uterus and ovaries [71].

6. Conditional inactivation of pVHL by Cre/loxP mediated recombination

In order to study pVHL in adult mice we have generated a conditional allele that allows tissue specific inactivation of Vhlh by Cre/loxP mediated recombination [64]. In this allele loxP sites flank the Vhlh promoter and exon 1 which are deleted upon Cre-mediated recombination, thus generating a Vhlh null allele (Fig. 2). Gene targeting by Cre/loxP mediated recombination allows inactivation of genes in a temporally and tissue restricted fashion, as opposed to conventional gene knock out protocols where the gene of interest is always disrupted in the germline [72,73]. This technology exploits the Cre/loxP system, a bacteriophage P1-derived site-specific recombination system, in which Cre-recombinase (Cre) catalyzes the recombination between two loxP sites which flank a genomic DNA sequence of interest, resulting in its deletion [74]. Genomic sequences or genes flanked by loxP sites are also referred to as “floxed” genes or 2-lox alleles and mice with tissue-specific gene inactivation can be generated by breeding the floxed allele to Cre-transgenic mice which express Cre-recombinase under the control of a tissue specific promoter. Similar to our approach Ma et al. generated a conditional allele in which Vhlh exons 2 and 3 were floxed [67].

Open in a separate window

Fig. 2

Genomic structure of the Vhlh conditional allele. Shown are the genomic structures of Vhlh in wild type (wt) configuration, the Vhlh 2-lox allele and the recombined Vhlh 1-lox allele. LoxP sites are depicted as red triangles, Vhlh exons are shown as black boxes. Exon 1 and the Vhlh promoter are deleted by Cre-mediated recombination, deleted area is shown in red. This allele was generated in 129 J-1 ES cells by homologous recombination with a targeting vector containing a “floxed” hygromycin-thymidine-kinase selection cassette, which was removed after transient transfection with Cre-recombinase [64]. Abbreviations: H, HindIII; N, NcoI.

7. Phenotypes generated by conditional inactivation of pVHL and their dependence on HIF

We have used a variety of Cre transgenic lines to inactivate pVHL in different tissues. Table 1 summarizes the phenotypes that were observed. In most tissues inactivation of pVHL resulted in abnormal differentiation, which was associated with either a decrease in proliferation or an increase in apoptosis (see Table 1).

In order to determine the role of the different HIF transcription factors in the development of the VHL phenotype in mice, we have generated mice that are either deficient for both Vhlh and Hif-1α, or double deficient for Vhlh and Hif-1β (Arnt). Using this double knock out approach, Biju et al. was able to show that a caspase-8 dependent pro-apoptotic phenotype in Vhlh−/− thymocytes was completely Hif-1 mediated [75]. Hif-2α, on the other hand, although expressed in thymocytes, was found to be non-functional. Inactivation of Hif-1α alone was sufficient to completely suppress the increase of VEGF mRNA levels and vascularity in Vhlh−/− thymi [75]. Using the same approach, Rankin et al. [76] was able to determine that the development of vascular tumors in Vhlh deficient livers was entirely dependent on Hif. While inactivation of Hif-1α in the liver did not affect vascular tumorigenesis, inactivation of Hif-1β (Arnt), the binding partner for both Hif-1α and -2α, was sufficient to prevent the development of cavernous hemangiomas. It is important to note here that Hif-2 is fully functional in hepatocytes.

8. pVHL inactivation in apoptosis, cystogenesis and tumorigenesis: a central role for HIF

Biju et al. have investigated pVHL’s role in the regulation of apoptosis and cell viability in thymocytes, a cell type that is normally not affected by VHL disease [75]. This study showed that tissue-specific inactivation of Vhlh in thymocytes resulted in highly vascularized thymi and in increased apoptosis rates through a Hif-1 dependent mechanism involving caspase-8. Although the development of a hypervascular phenotype was not surprising, as it is a general hallmark of VHL associated neoplasms, the finding of impaired cell survival was unexpected. In another study by Pfander et al. chondrocyte specific inactivation of Vhlh in the growth plate resulted in a reduction of proliferation through a probably Hif-1 dependent increase in cell cycle inhibitor p57^kip2 [78]. In contrast, Hif-1α inactivation in chondrocytes resulted in a lack of p57^kip2 mediated growth arrest followed by massive apoptosis [82].

Although pVHL’s function as a tumor suppressor in CC-RCC and other VHL associated neoplasms has been well established [3–5,83], the unexpected finding that loss of pVHL can negatively affect cell survival and proliferation, raises interesting questions in regard to pVHL’s role in tumor suppression. In CC-RCC, re-introduction of wt pVHL protected VHL deficient CC-RCC cells from apoptosis under certain conditions via upregulation of BCL-2 [84,85]. Of note, BCL-2 knock out mice are prone to develop renal cysts [86], and altered regulation of BCL-2 expression may play a role in the development of VHL associated renal cysts.

In a non-RCC tumor model, Mack et al. found significant growth retardation in embryonic stem (ES) cell derived Vhlh−/− teratocarcinomas [87]. Interestingly, an increase in apoptosis rates was not reported. Inactivation of Vhlh may therefore have differential effects on proliferation and apoptosis in cell types of diverse histogenetic origin and genetic backgrounds. This notion is furthermore supported by the fact that VHL is widely expressed [6,88,89], yet the spectrum of VHL associated tumors is limited [83].

The role of HIF in cell death and tumor progression has been intensely studied (for a recent review see [90]) and there is substantial evidence that HIF is important for VHL tumorigenesis and metastasis. In regard to HIF-1 and -2 function in VHL renal carcinogenesis, it is of interest that a significant number of VHL defective CC-RCC cell lines do not express HIF-1α, but do express HIF-2α [17]. A bias towards HIF-2α expression was also found in clinical CC-RCC samples (94% versus 69% express HIF-2α in CC-RCC with confirmed VHL defect) [91]. Thus, VHL associated tumor development may depend on a shift in the ratio of HIF-1 versus HIF-2 expression towards HIF-2. VHL reconstituted 786-O CC-RCC cells, for example, transfected with a non-degradable form of HIF-2α were still able to form tumors in nude mice thereby overriding pVHL’s tumor suppressor function [92], while, the expression of non-degradable HIF-1α in a similar experimental setting did not produce tumors [93]. Consistent with these data is the finding that inactivation of HIF-2α by RNA interference in a VHL deficient background suppressed tumor formation [94,95]. Taken together these reports suggest that HIF-1 and -2 have diverse functions in regard to VHL associated renal tumorigenesis. Whether differential effects on signaling pathways critical for renal epithelial cell growth may explain the functional differences between the two HIFs remains to be investigated. For example, the expression of transforming growth factor-α (TGF-α), a potent renal epithelial mitogen, is HIF dependent [96,97] and HIF-2 dependent activation of the TGF-α/epidermal growth factor receptor pathway may be important in renal cystogenesis and carcinogenesis [98,99]. The exact role of chemokine receptor CXCR4, a direct HIF target, in renal cystogenesis and tumorigenesis is unclear, but high levels of expression correlate with poor survival and the ability to metastasize [100]. Another direct HIF target is EPO, which has been shown to stimulate renal tumor cell growth [101,102]. It is of interest in this context that EPO appears to be a preferential HIF-2 target, at least in hepatocytes [103].

Taken together, it is possible that reduced growth of Vhlh−/− teratocarcinomas as reported by Mack et al. [87], may have been consequence of constitutively active Hif-1 in conjunction with nonfunctional Hif-2 (Celeste Simon, personal communication). Koshiji et al. have shown that increased levels of HIF-1 in colon cancer cells can result in a suppression of Myc transcriptional activity through direct protein to protein interaction, thus inhibiting cell cycle progression by up-regulating p21^cip1 [104]. These findings are in concordance with observations made by Carmeliet et al. in a teratocarcinoma model derived from Hif-1α deficient ES cells. In this model, tumors that lacked Hif-1 had a growth advantage [105]. Whether increased HIF-1 in a pVHL deficient background in the presence or absence of functional HIF-2 has different effects on cell cycle regulation remains to be investigated. However, growth suppression in Vhlh deficient teratocarcinomas may not be entirely dependent on Hif-1. In a follow-up study to Mack et al. [87], Rathmell et al. showed that introduction of a mutated pVHL species (Y112H) into a Vhlh−/− background remained growth suppressive despite a restoration of Hif-α regulation [106].

9. Conditional inactivation of pVHL in the kidney: towards the development of a mouse model for VHL renal carcinoma

Besides its role in VHL associated renal cancer, pVHL plays a major role in the pathogenesis of sporadic renal cell carcinoma of the clear cell type (CC-RCC), the most common form of kidney cancer [2]. VHL associated CC-RCCs are often preceded by pre-neoplastic renal cysts, which are multifocal and bilateral in patients with VHL disease [107], and are generally thought to arise from the proximal renal tubule (PRT) [1,107]. The worldwide mortality from renal cancer is expected to exceed 100,000 per year [108] and therapeutic options are limited; the generation of a mouse model for VHL associated renal cancer therefore would be invaluable.

To inactivate of Vhlh in the PRT, we have generated Cre transgenic mice using a mutated version of the rat phosphoenolpyruvate carboxykinase (PEPCK) promoter (Rankin et al., unpublished). Analysis of Cre expression with ROSA26-lacZ Cre-reporter mice [109] demonstrated that PEPCK-Cre is expressed in all S3- and most S1- and S2-segments (>80%) of the PRT (Rankin et al., unpublished observation). PEPCK-Cre mutant mice developed renal cysts (Fig. 3) at low frequency (4/25), but renal tumors were not observed by 18 months of age (Rankin et al., unpublished observation). Whether renal cyst formation in these mutants is a direct result of Vhlh gene inactivation in the PRT remains to be investigated. Similar to our findings, renal tumors were not found by Ma et al. when β-actin promoter driven Cre-recombinase was used to generate mice which lacked pVHL in a mosaic pattern [67].

Open in a separate window

Fig. 3

Inactivation of Vhlh in proximal renal tubule (PRT) cells results in renal cysts at low frequency. (A) Gross in situ photograph of the right kidney from a conditional knock out mouse with inactivation of Vhlh in the PRT. Arrows point to two superficial cysts. Evidence of renal cell cancer or epithelial dysplasia was not detected (Rankin et al., unpublished observation). (B) Paraffin section of kidney with renal cysts stained with H&E. Shown is a cluster of smaller cysts lined with renal epithelial cells (arrows). Magnification is ×200.

There are several potential explanations for this observation: (a) additional mutations in other tumor suppressor genes or oncogenes are required before tumors develop; (b) increased activity of Hif-1 may prevent the development of VHL associated renal lesions and loss or diminished activity of Hif-1 is necessary (see discussion in Section 8); (c) a shift towards high levels of Hif-2 expression may be required before cysts or tumors develop but has not occurred (as discussed in Section 8); (d) a specific gain of function mutation in pVHL may be required; (e) the genetic background of mice may play a role; and (f) VHL associated renal lesions arise from nephron segments that were not targeted (Fig. 4).

Open in a separate window

Fig. 4

HIF plays a central role in VHL associated renal cystogenesis and tumorigenesis. Above: proposed sites of origin of VHL renal cycts and tumors. A schematic representation of a nephron is shown above. The area shaded in blue represents the proximal renal tubule, the area shaded in red represents the THP expressing medullary thick ascending limb of Henle and early distal tubule. Below: overview of selected hypoxia/HIF dependent and independent pVHL targets with potential importance for renal cystogenesis and tumorigenesis. Ovals identify transcriptionally regulated targets, rectangles identify targets, which are regulated by non-transcriptional mechanisms; vertical arrows indicate up- or down regulation. HIF-1α counteracts Myc activity through direct protein-protein interaction and has been shown to induce cell cycle arrest in colon cancer cells [104], what this interaction means for renal tumorigenesis is unclear. BNIP-3 is a pro-apoptotic factor and is directly up-regulated by HIF-1 [116–118]. BCL-2, an anti-apoptotic protein, has been shown to be HIF-1 dependent in ES cell derived teratomas [105]. Up-regulation of cyclin D1 (CCND1), an important regulator of cell cycle progression, was found in VHL-deficient renal cancer cell lines and shown to be hypoxia-dependent in a cell type specific manner [119–122]. Chemokine receptor 4 (CXCR4) is a direct HIF target and has been shown to correlate with poor survival and metastatic ability of CC-RCC [100]. Jade 1 is a novel pVHL interacting protein that may play a role in renal cytogenesis through interaction with polycystin-1 [123]. Abbreviations: gl, glomerulus; cd, collecting duct.

Although it is widely believed that VHL associated renal tumors originate from the proximal nephron [1,107], it has been shown that CC-RCCs can express markers of both, the distal and the proximal nephron [110–113], implying a multipotential capacity of neoplastic renal cells to synthesize antigens that can be associated with different tubular segments and different stages of nephrogenesis. More recently Maxwell’s group reported that in kidneys from VHL patients “early” multicellular lesions appeared to be more frequently derived from Tamm–Horsfall protein (THP) expressing tubular segments than from the proximal tubule. Tamm–Horsfall glycoprotein is expressed in the medullary thick ascending loop of Henle (mTAL) and in the early distal renal tubule (for a recent review see [115]). On the other hand, single cell foci of HIF-target gene expression were found more frequently in the proximal tubular cells [114], suggesting a site-specific tumor suppressor function for pVHL in the kidney. This finding supports the notion that VHL associated renal tumors may be of a different histogenetic origin than commonly believed and arise from more distal nephron segments (Fig. 4). Systematic targeting of Vhlh in different nephron segments using site specific promoters may therefore be necessary in order to reproduce classic VHL associated renal lesions in mice. Fig. 4 provides an overview of pVHL’s and HIF’s role in renal cysto- and carcinogenesis.

10. Concluding remarks and future directions

In this review I have summarized findings from the analysis of VHL knock out mice and have discussed the role of HIF in the context of VHL associated tumorigenesis. Proper targeting of HIF-α subunits for proteolysis by the pVHL E3-ubiquitin ligase complex is critical for normal embryogenesis and tissue physiology. In the context of VHL tumorigenesis HIF plays a central role. While both HIF-1 and -2 seem to be capable of mediating the highly vascular nature of VHL deficient tissues, they differ substantially in regard to their effects on apoptosis and proliferation, especially in the context of CC-RCC tumorigenesis. The negative effects of pVHL inactivation on cell survival and proliferation in thymocytes, chondrocytes and ES derived teratocarcinomas were largely unexpected for a protein with tumor suppressor function. They most likely emphasize tissue specific differences in HIF-1 and -2 activity levels and target gene expression, as well as tissue specific aspects of HIF-independent pVHL functions. Clearly, for a more comprehensive understanding of pVHL’s role in normal physiology and tumor suppression, further knock out and transgenic studies are needed that investigate cell type specific functions of individual HIF transcription factors in wild type and VHL deficient backgrounds, as well as the role of specific VHL mutations.

Acknowledgments

Abbreviations

References