Renal Tubular HIF-2α Expression Requires VHL Inactivation and Causes Fibrosis and Cysts

The Hypoxia-inducible transcription Factor (HIF) represents an important adaptive mechanism under hypoxia, whereas sustained activation may also have deleterious effects. HIF activity is determined by the oxygen regulated α-subunits HIF-1α or HIF-2α. Both are regulated by oxygen dependent degradation, which is controlled by the tumor suppressor “von Hippel-Lindau” (VHL), the gatekeeper of renal tubular growth control. HIF appears to play a particular role for the kidney, where renal EPO production, organ preservation from ischemia-reperfusion injury and renal tumorigenesis are prominent examples. Whereas HIF-1α is inducible in physiological renal mouse, rat and human tubular epithelia, HIF-2α is never detected in these cells, in any species. In contrast, distinct early lesions of biallelic VHL inactivation in kidneys of the hereditary VHL syndrome show strong HIF-2α expression. Furthermore, knockout of VHL in the mouse tubular apparatus enables HIF-2α expression. Continuous transgenic expression of HIF-2α by the Ksp-Cadherin promotor leads to renal fibrosis and insufficiency, next to multiple renal cysts. In conclusion, VHL appears to specifically repress HIF-2α in renal epithelia. Unphysiological expression of HIF-2α in tubular epithelia has deleterious effects. Our data are compatible with dedifferentiation of renal epithelial cells by sustained HIF-2α expression. However, HIF-2α overexpression alone is insufficient to induce tumors. Thus, our data bear implications for renal tumorigenesis, epithelial differentiation and renal repair mechanisms.


Introduction
Mammalian cells require oxygen for energy homeostasis and thus for maintenance of cellular function and integrity. On the molecular level, adaption to reduced oxygen concentrations (hypoxia) depends on the activation of the Hypoxia-inducible Factor (HIF), which enables critical processes such as glycolysis, angiogenesis and erythropoiesis [1]. HIF is a transcriptional heterodimer, consisting of a constitutive ß-subunit and an oxygen sensitive a-subunit, HIF-1a or HIF-2a. Both a-subunits are regulated similarly [2], mainly by oxygen dependent hydroxylation leading to ubiquitination and proteasomal destruction [3]. However, knockout experiments [4,5], tissue expression patterns [6,7] and target gene specificity [8,9] indicate isoform specific roles at least to some extent. Of note, in hypoxic rat kidneys HIF-1a and HIF-2a display a strikingly separate expression pattern. The former shows expression in tubular epithelia, whereas the latter shows expression in interstitial and glomerular cells [6].
For a number of reasons, the kidney has played a seminal role in understanding oxygen sensitive gene regulation. Despite a high oxygen transport rate to the kidney, oxygen tensions are very heterogeneous and in part lower as 10 mmHg [10]. Teleologically this may explain why the prototype of oxygen regulated genes, erythropoietin (EPO), is mainly induced in the kidney. Acute renal failure by ischemia-reperfusion injury is greatly attenuated if pharmacological preconditioning with HIF stabilizers is performed [11,12]. Finally, the recognition component of the oxygen responsive ubiquitin ligase complex, the von Hippel Lindau (VHL) tumor suppressor [13], is a gatekeeper for growth control of tubular epithelial cells in the kidney [14]. In humans, biallelic inactivation of VHL leads to the development of renal cell carcinoma (RCC) of the clear cell type, which occurs in the hereditary VHL syndrome as well as in sporadic RCC. In mouse tissue specific VHL knockout was not found to induce tumors, but generated cysts either alone [15] or in conjunction with a PTEN knockout [16].
Human clear cell RCCs typically show global oxygen independent activation of HIF-1a and HIF-2a [17,18]. Stabilization of HIFa subunits in early lesions of human RCC may be a decisive step in renal tumorigenesis [19]. Experimental studies have shown that HIF-2a, and not HIF-1a, appears to be the decisive subunit mediating tumorigenic features [20][21][22][23]. Mechanistically this may be due to a cellular proliferative effect of HIF-2a, whereas HIF-1a may have opposite effects [22,24]. Nevertheless, overexpression of HIF-1a in murine proximal tubuli has recently been shown to lead to RCC [25]. Constant activation of HIF by genetic inactivation of VHL in tubular epithelia has further shown to induce renal fibrosis [26], which may indicate a common pathway of epithelial dedifferentiation.
In summary, HIF effects play an important role in the kidney, which can be beneficial or deleterious, depending on the setting and the timing. The specific roles of the different HIFa isoforms in this context are not well defined. Furthermore, there is no knowledge of the differential expression patterns of HIF-1a and HIF-2a in the human kidney. Experimental data originates mostly from VHL knockout studies in the mouse, where HIFa stabilization is an inevitable consequence. However, other HIF independent effects with oncogenic potential are known to be released when VHL is inactivated [14]. We therefore aimed to clearly define the physiological expression patterns of HIFa subunits in normal kidney epithelial cells, as well as in response to experimental VHL knockout and human VHL disease. Furthermore, we aimed to study the tumorigenic potential of tubular HIF-2a overexpression in a novel mouse model.

Distinct expression patterns of HIF-1a and HIF-2a in hypoxic mouse and human kidneys
Identically to the predescribed situation in the rat, we see distinct expression patterns for the two HIFa subunits in mouse ( Figure 1) and man ( Figure 2). Hypoxic human kidneys were investigated from victims of carbon monoxide intoxication, where the tissues are severely hypoxic due to the reduced oxygen transport capacity of the haemoglobin and in kidney tissue adjacent to RCCs, where hypoxic microenvironment is often caused by expansive tumor growth. In all cases investigated, hypoxia leads to HIF-1a accumulation exclusively in tubular epithelial cells, whereas HIF-2a is stabilized in interstitial cells and in glomeruli. Hypoxia does not enable HIF-2a expression in renal tubular cells, in any species.

Inducible VHL knockout in mouse renal tubular cells enables HIF-2a expression
We next studied HIF expression in an inducible murine VHL knockout (VHL 2/2 ) model, which was implied by the Pax8 promoter and confers an inducible Cre-driven knockout in the complete tubular system of the kidney [27,28]. In these animals VHL expression is lost after doxycycline treatment. Control animals without doxycycline showed no expression of either HIFa subunit (data not shown). In contrast to the physiological HIF expression, the knockout mice showed tubular expression of HIF-2a (and HIF-1a) in the tubules upon a 3-day treatment with Biallelic inactivation of VHL releases HIF-2a expression in distinct early lesions of the distal tubule in kidneys of the human VHL disease Previously, we reported multiple premalignant lesions in tubules of patients with VHL germline mutations, which were identified on the basis of carbonic anhydrase 9 (CAIX) expression [19]. We detected that these express HIF-1a, and had inactivation of the wild type VHL allele. Subsequently we demonstrated, that these foci in the distal tubule also displayed decreased expression of E-cadherin [29]. In the light of the mouse data, we re-examined the kidneys of VHL patients and observed additional foci of reduced E-cadherin labelling, which did not express CAIX and little or no HIF-1a, but did express HIF-2a (Figure 4 A). Thus there are (at least) two distinct classes of foci of VHL inactivation, which we now designate Type I (high CAIX, high HIF-1a) and Type II (no CAIX and high HIF-2a) foci (see table 1). We also stained for other markers, which are not expressed in normal renal tubules but do show expression in clear cell RCCs; the glucose transporter-1 (Glut1) and the intermediate filament vimentin. In line with the hypothesis that these lesions contain precancerous cells, the type II lesions stain clearly positive for vimentin and Glut1. Finally, in contrast to type I lesions the type II lesions show intense labelling for cyclin D1 (Figure 4 B), which has already been implicated to be a candidate of HIF-2 mediated tumorigenesis [22].

Transgenic HIF-2a overexpression in renal tubular cells leads to renal fibrosis
We generated a transgenic mouse model with constitutively stable and active HIF-2a derived from a cDNA (tmHIF-2a.HA) under the control of the kidney specific Ksp-promoter which enables mainly distal tubular expression (Fig. 5, S1 and S2). Kidneys of tmHIF-2a.HA(+) mice at the ages of 3, 6 and 9 months displayed no pathology regarding gross morphology and histology (data not shown). We therefore decided to let the mice grow to an age of 14-16 month. Figure 6 A shows representative photographs of the kidneys from tmHIF-2a.HA(2) and tmHIF-2a.HA(+) mice at this age. Obvious morphological differences can already be seen on the surface of kidneys from these two strains. The kidneys of tmHIF-2a.HA(2) control mice displayed a uniform smooth surface (Figure 6 A, left), whereas the kidneys of tmHIF-2a.HA(+) mice had an irregular and rough appearance (Figure 6 A, right). tmHIF-2a.HA(+) mice had a significantly reduced kidney weight as judged by kidney to body weight ratio ( Figure S3 A). Histological analysis of the kidneys from aged animals was performed. The retracted organ surface implicated fibrotic events within the kidney. Therefore, total collagen as a fibrosis marker was stained by SiriusRed (Figure S3  Real-time PCR analysis for HIF target genes from whole kidney lysates of the aged transgenic mice showed a strong induction of PHD3 and a moderate induction of TGFa and Glut1 in the tmHIF-2a.HA(+) mice (Figure S3 C). Since VEGF expression was induced as well and angiogenesis is believed to be a classical function of HIF, we histologically analyzed kidney capillarization in our transgenic mice. Kidney sections were stained for the endothelial marker MECA-32 by immunohistochemistry and the vascularization level was determined by counting positively stained capillaries in a blinded fashion. Compared to the tmHIF-2a.HA(2) control group, tmHIF-2a.HA expression led to a significant increase in kidney vascularization ( Figure S3 D). However, EPO expression was not induced in tmHIF-2a.HA(+) mice and hematocrit not increased (data not shown).

Transgenic HIF-2a overexpression in tubular cells leads to renal cyst formation
In addition to the fibrotic phenotype, the histological analysis of the tmHIF-2a.HA(+) transgenic mice revealed frequent formation Transgene expressing tubular segments appear to have normal numbers of cilia, where HIF-mediated reduction of cilia could have been a mechanism leading to cysts ( Figure S4).

Discussion
The transcription factor HIF is widely recognized as a critical mediator of many (patho)physiological processes and is currently under evaluation as a putative therapeutic target either by  inhibition for tumor therapy or activation for organ protection. However, the spectrum of short and long term effects of HIF manipulation are difficult to foresee. Our study shows that constitutive aberrant overexpression of HIF-2a is sufficient to induce a complex type of kidney disease associated with tubular cyst formation, interstitial fibrosis and declining renal function, yet no development of RCC. These data bare implications for improved understanding of cellular control of HIF expression and its effects, as well as development and progression of kidney disease.

Renal tubular HIF expression
We show that the distinct expression pattern for HIF-1a and HIF-2a in the kidney is stable across species and whatever stimulus is used (oxygen deprivation, application of prolyl hydroxylase inhibitors [11,12] or neighbourhood to RCC). The molecular mechanism of differential control of HIFa expression remains elusive, where transcriptional, translational or posttranslational mechanisms could contribute. Importantly, when VHL is inactivated in renal tubular cells in mice or human VHL disease, we do see HIF-2a expression in these cells. The same has been described previously in another conditional VHL deletion model, although the authors did not compare the HIF-2a expression to microenvironmental stimulation [16]. This could mean, that VHL in some way specifically suppresses the expression of HIF-2a in tubular epithelial cells, which may have interesting implications for renal tumorigenesis. This is further supported by our findings in the early renal lesions of the human VHL disease. Type II foci show expression of HIF-2a, which is followed by upregulation of the proliferative target gene cyclin D1, which has been implicated in renal tumorigenesis before [22]. Accordingly, early studies have described a VHL dependent expression in RCC cells, which was more inclined to influence HIF-2a, rather than HIF-1a [18,30].

Renal cell carcinoma
The development of VHL associated clear cell RCC appears to be closely linked to HIF activation. It has been demonstrated that accumulation of HIF-1a and HIF-2a is a feature of very early tumorigenesis in kidneys derived from patients with the hereditary VHL syndrome, which can be verified already at a single cell level [19]. HIF-2a (and not HIF-1a) appears to be the decisive factor for growth initiation of experimental tumors. Functionally this may occur via pro-proliferative effects of HIF-2a, which may be mediated by cMyc signalling [24] and which can be verified in a subgroup of RCCs expressing HIF-2a only [31]. Thus, accumulating data has been published over the last decade implicating that HIF-2a activation is a seminal oncogenic hit in renal tumorigenesis as well as numerous other tumor entities, playing a further important role in tumor growth and behaviour [32]. We show that biallelic VHL inactivation releases HIF-2a expression in renal tubular cells. Therefore, this process could be an important event in the development of RCC. Nevertheless, our data also show that HIF-2a activation on its own is not sufficient to induce renal tumors in the mouse, which is similar to two further murine models who have deleted VHL in tubular cells [15,16]. The lack of tumors in our and most other mouse models may be a result of the absence of crucial additional oncogenic events. Accordingly, it has been estimated that numerous mutational events are necessary to establish a malignant tumor, which is capable of enforcing itself in a hostile microenvironment [26]. However, HIF-1a overexpression in the proximal tubule seems sufficient to induce tumors in mice [25]. Thus, the particular segment of the renal tubule may also be of importance.

Renal cystic disease
Regulation of renal tubular cell polarity and proliferation is strongly linked with the functional integrity of the primary cilium and ciliary defects often cause the development of renal cysts [33]. Intriguingly, recent data have shown that VHL and/or HIF play an important role in ciliary biology (for review see [34]). Consequently, two studies have demonstrated the development of renal cysts when VHL is inactivated in tubular cells with or without consecutive deletion of PTEN, respectively [15,16]. It is not clear from these studies, which HIFa isoform mainly contributed to cyst development, although one study suspected HIF-2a to be the driving force, since simultaneous HIF-2a knockouts did not show cysts [15].
In human autosomal dominant polycystic kidney disease (ADPKD) HIF is also activated, but follows the ''physiological'' expression pattern of tubular HIF-1a and peritubular HIF-2a [35]. Of note, primary cystic diseases such as ADPKD rarely show the development of RCC, whereas acquired renal cysts and cysts in the hereditary VHL syndrome are regarded as precancerous lesions [34]. Whether the reason for these differences lie within HIF-2a expression of cystic epithelia remains speculative to date and is under investigation.

Chronic kidney disease (CKD)
A large body of evidence exists showing that HIF activation can be beneficial in acute kidney injury models [11,36,37]. In CKD the situation is less clear although chronic hypoxia is considered to play a role in the development of progressive tubulointerstitial fibrosis and the development of CKD [38,39]. Pharmacological HIF activation has been shown to be protective in chronic models of the remnant kidney by increase of peritubular angiogenesis [40], in the Thy1 nephritis [41] or models of diabetic nephropathy [42]. Nevertheless, HIF-1 appears to be an important mediator of CKD by chronic hypoxia, which is caused by EMT and subsequent fibrosis [26]. Furthermore, stable expression of HIF-1a in tubular epithelial cells seems to be sufficient to promote interstitial fibrosis [43]. Our mouse model has now shown that sustained overexpression of HIF-2a alone is sufficient to induce tubulointerstitial Figure 5. Generation of Ksp/tmHIF-2a.HA transgenic mice. A. Schematic representation of the pcKsp/tmHIF-2a.HA expression vector used for pronucleus injection consisting of a 1.3 kb Ksp-promoter fragment and an HA-tagged mutated HIF-2a triple mutant cDNA (Ksp, kidney specific; tm, triple mutant; UTR, untranslated region; HA-tag, influenza hemagglutinin epitope tag; poly-A site, poly-adenylation site). Injection of the Ksp/tmHIF-2a.HA construct successfully produced transgenic mice. Two of these were chosen and bred into homozygous strains, which was confirmed by genotyping for Ksp/tmHIF-2a integration and referred as Ksp/tmHIF-2a strain 33 and strain 39, respectively (data not shown). B. Analysis by RT-PCR detected transgenic tmHIF-2a.HA mRNA expression only in whole kidney RNA extracts from mice of strain 39. C. tmHIF-2a.HA protein expression was analyzed in whole kidney extracts from both strains by immunoblot. In parallel to the RNA expression, transgenic tmHIF-2a.HA protein was only expressed in strain 39. Based on these data strain 39 has been termed as tmHIF-2a.HA(+), whereas mice from strain 33 have been defined as tmHIF-2a.HA (2) and served as control strain, having the transgene integrated but not expressed. D. The localization of tmHIF-2a.HA in the mouse kidney was next analyzed by immunohistochemistry against HIF-2a and the HA-tag of the transgene on consecutive sections. Transgenic tmHIF-2a.HA expression was detected in tubular epithelial cells for HIF-2a (left hand panels) as well the HA-tag (right hand panels). E. Kidneys of the tmHIF-2a.HA(2) control strain were negative for both antibodies. doi:10.1371/journal.pone.0031034.g005 fibrosis and renal insufficiency, perhaps partly mediated by induction of TGFb1. The molecular mechanism of TGFb stimulation and subsequent fibrogenesis in our model is unclear to date. However, a number of fibrogenic pathways have been described which appear to be influenced after prolonged hypoxia in the kidney, such as PAI-1, CTGF and TGFß [39] or lysyl oxidases [44]. Interestingly, all these effects have so far been attributed to HIF-1a, rather than HIF-2a. However, since renal tubular cells do not physiologically express HIF-2a this effect may have been impossible to observe in vivo. The origin of renal fibrogenesis is discussed controversially in the literature (transdifferentiated epithelia [45] or extrarenal neurocrest derived fibroblasts [46]. Of note, in our model fibrogenesis clearly stems from a characteristic change in epithelial cells. The irregular and rough appearance of the transgenic kidneys probably stems largely from fibrotic retractions, as can be observed in severe human CKD. However, in some cases cysts could also cause the rough appearance wherever these structures bulge out of the surface. We believe that the former is the leading cause for this phenotype, since the latter only occurs regionally.
Our study was undertaken in order to improve the understanding of differential effects of HIF-1a and HIF-2a in the kidney. There are a number of limitations to our approach, such as the use of different promotors of our genetic models resulting in the targeting of differing nephron segments. Furthermore, it is unclear whether the observed phenotype is specific for.
HIF-2 activation, or whether it could also be achieved by overexpression of HIF-1, possibly by overlapping target genes. Finally, we have seen the described phenotype only in ageing mice, which suggests that the effects are rather mild in nature. Therefore, further studies will be needed to confirm and expand on our data. However, our data imply that the switch towards tubular HIF-2a expression is an important mediator of renal tumorigenesis, which can be achieved by VHL inactivation. The observed phenotype with renal cysts, tubulointerstitial fibrosis and CKD may share common pathways, but it could also be caused by parallel mechanisms. Of note, at least in mice HIF-2a appears not to be an independent oncogene. Hence, its overexpression may be an important early step in tumorigenesis but is not sufficient to induce RCCs, which may also account for the majority of early lesions in human VHL kidneys.

Generation of plasmid DNA, mouse strains and experimental animal protocols
For the generation of HIF-2a transgenic mice, a mutated and HA-tagged cDNA of HIF-2a was used as previously described (triple mutant tmHIF-2a; [8,47]). The product is constitutively expressed and active, independent of oxygen levels. A 1.3 kb fragment of the kidney-specific (Ksp-) cadherin 16 promoter (described in [48]) was fused at its 39-end to a b-globin intron and cloned into pcDNA3 generating the vector pcKsp/betaGl. Thereby the internal CMV promoter of the pcDNA3 backbone was deleted. The HA-tagged tmHIF-2a cDNA was cloned into pcKsp/betaGl at the 39-site of the b-globin intron, generating the expression vector pcKsp/tmHIF-2a.HA. Pronucleus injection of the Ksp/tmHIF-2a.HA construct successfully produced transgenic mice in a C57Bl10xCBA/Ca hybrid background.
Conditional VHL knockout mice were generated by crossbreeding three different transgenic mouse lines: a) floxed VHL mice (Jackson Laboratories, Sacramento, CA) [49], b) LC-1 transgenic mice bearing Cre-recombinase expression under the control of the bidirectional P tet promoter [50] and c) mice which express a reverse tetracycline-dependent transactivator under control of the Pax8 promoter [28] and are described previously [27]. For the induction of the VHL knockout, mice received 0.2 mg/ml doxycycline/5% glucose in the drinking water for 3 days.
All procedures involving mice were approved by the institutional review board (file number 31/25/04, University of Erlangen, Germany) and were in accordance with the National Institutes of Health guidelines. All data provided in the manuscript are pooled data from male and female mice.

Renal function
Blood samples were taken immediately from sacrificed animals by aspiration from the heart. Plasma creatinine and urea were enzymatically measured using the Integra 800 (Roche).

Collection of human tissues
RCC specimens were from historical collections [17,19] collected from mostly radical tumor-nephrectomies and fixed in 4% paraformaldehyde (PFA). Written informed consent was obtained from each patient before nephrectomy. Paraformaldehyde fixed kidney specimens from victims of CO intoxication were from forensic medicine.

Immunohistochemistry and antibodies
Paraffin sections (2 to 4 mm) were dewaxed in xylene and rehydrated in a series of ethanol washes. Immunohistochemistry was performed as described previously [51]. Detailed information for all primary antibodies is given in supplementary table S1. For detection of HIF-1a, HIF-2a and HA the signal amplification system (CSA) from DAKO (Glostrup, Denmark) was used according to the manufacturer's instructions and described in Rosenberger [6]. Biotinylated secondary anti-mouse, -rabbit,sheep or -rat antibodies were from DAKO. Slides were partially counterstained with hematoxylin. Signals were analyzed with a Leica DMRB microscope (Leica, Bensheim, Germany). Photographs were digitally recorded by means of a Visitron system (Visitron, Puchheim, Germany).

In vitro generation of recombinant protein
Recombinant tmHIF-2a.HA protein was generated from the expression vector pctmHIF-2a.HA by using the TNTH Quick coupled transcription/translation system (Promega) according to manufacturer's instruction. pcDNA3 empty vector was used as a negative control.

RT-PCR and quantitative real-time PCR
Total RNA extracts were prepared by the use of RNAzol B (Biogenesis, Poole, U.K.) according to manufacturer's instructions. cDNA was generated using Superscript II RT (Invitrogen) according to manufacturer's instructions from 200 ng of total RNA.
The mRNA expression of Glut1, TGF-a and TGF-b1 was determined by quantitative real-time PCR in duplicates with the Taqman Gene Expression System (Applied Biosystems) according to manufacturer's instructions. Normalization was to b-2 microglobulin housekeeping gene and fold-expression level was calculated using the DDct method. The following Taqman Gene Expression Assays were used: Glut1 Assay ID Mm01192270_m1; TGF-a Assay ID Mm00446231_m1; TGF-b1 Assay ID Mn00441727_g1; b2-m Assay ID Mm00437762_m1.

Determination of fibrosis and vascularization score
The fibrosis level was quantified after SiriusRed staining of kidney sections. The score was determined by the percentage of staining in 15 randomly selected optical fields per kidney section at a 2006 magnification according to the following ranking: 0%score 0; 1-25% -score 1, 25-50% -score 2; 50-75% -score 3, more than 75% -score 4. Vascularization was quantified by counting signals in 15 randomly selected optical fields per kidney section at a 2006 magnification after immunohistochemical staining against MECA-32. The origin of the samples was blinded to the pathologist doing the scoring.