NAD(P)H Oxidases Regulate HIF-2α Protein Expression*

Biallelic inactivation of the von Hippel-Lindau tumor suppressor gene (VHL) is linked to the development of hereditary and sporadic renal cell carcinoma (RCC). In the absence of VHL, the α subunits of heterodimeric hypoxia-inducible transcription factors (HIF-1α and HIF-2α) are stabilized. Reactive oxygen species, generated by NAD(P)H oxidases, are involved in signaling cascades of malignant growth. We show that in VHL-deficient cells p22phox, Nox4 protein levels and NADPH-dependent superoxide generation are increased. Reintroduction of VHL into the VHL-deficient cells down-regulates the expression of p22phox and NADPH-dependent superoxide generation. Inhibition of the 26 S proteasome in VHL-expressing cells increased p22phox protein levels, which correlated with an increase of NADPH-dependent superoxide generation. We also show that p22phox co-immunoprecipitates with VHL in vivo. Moreover, p22phox is a target of ubiquitination. Importantly, in VHL-deficient cells, diphenyleneiodonium chloride (DPI), an inhibitor of Nox oxidases, decreased the expression of HIF-2α. Down-regulation of Nox1, Nox4, and p22phox expression by small interfering RNA also decreased HIF-2α protein expression and inhibited Akt and 4E-BP1 phosphorylation, suggesting that a translational mechanism is involved in maintaining HIF-2α in VHL-deficient cells. Colony formation by RCC 786-O in soft agar was markedly inhibited by DPI. Moreover, DPI significantly inhibited RCC 786-O tumor formation in athymic mice. Collectively, the data demonstrate that VHL protein exerts its tumor suppressor action, at least partially, via inhibition of p22phox-based Nox4/Nox1 NADPH oxidase-dependent reactive oxygen species generation.

Epithelial tumors comprise the majority of renal carcinomas in which ϳ75% are histologically of the clear cell type (RCC). 2 Biallelic inactivation of the von Hippel-Lindau (VHL) tumor suppressor gene is linked to the development of hereditary and sporadic RCC. Mutations within the VHL gene occur in 80% of RCC and maintain the expression of HIF-1␣ and HIF-2␣ (1), which leads to downstream activation of genes involved in angiogenesis, enhanced glucose transport, cellular proliferation, and apoptosis, major features of RCC (2)(3)(4)(5)(6). Reactive oxygen species (ROS) regulate hypoxia-dependent and -independent activation of HIF-1␣ (7,8). NAD(P)H oxidase systems are major sources of ROS. The Nox family of NAD(P)H oxidases have a core structure consisting of six transmembrane domains, including two heme-binding regions located at the N terminus and a cytoplasmic C terminus containing FAD-and NADPH-binding regions. There are five NAD(P)H oxidase isoforms in mammalian cells. Recent evidence indicates an important role for redox-regulated signaling in neoplastic growth. Nox1 overexpression transforms normal fibroblasts and creates a cell that is tumorigenic in athymic mice (9). Furthermore, Nox1 triggers an angiogenic switch and converts tumors from dormant to aggressive growth (10). Nox4 regulates growth of malignant melanoma cells and prevents pancreatic cancer cell apoptosis in the presence of growth factors (11,12). Nox5 mediates growth of prostate cancer cells (13). In this study, we provide the first evidence that VHL regulates NADPH-dependent ROS generation by regulating the protein expression of p22 phox , a key component in the activation of NAD(P)H oxidases. In addition, we show that generation of ROS activate the phosphoinositide 3-kinase (PI3K)/Akt-dependent translational pathway mediating HIF-2␣ protein accumulation in VHL-deficient cells. Furthermore, specific inhibition of ROS generation via flavoprotein enzymes down-regulates HIF-2␣ protein expression, cell growth in vitro, and tumor formation in vivo.
Cell Culture, Transfection, Adenovirus Infection, and Preparation of Cell Lysates-Human renal proximal tubular cells (HK2). VHL-deficient cells (RCC 786-O), and VHL-expressing ACHN cells (from American Type Culture Collection) were maintained in RPMI 1640 (HK2 cells) or in Dulbecco's modified Eagle's medium (RCC 786-O and ACHN) (Invitrogen) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin sulfate at 37°C with 5% CO 2 . Dominant-negative Akt pCMV6-HA-Akt (K179M) was kindly provided by G. G. Choudhury. RCC 786-O cells were infected with adenovirus vectors at the indicated multiplicity of infection. As a control for the effects of adenovirus infection alone, Ad GFP, containing green fluorescence protein, was used. NADPH-dependent superoxide anion generation in cell homogenates was measured by lucigenin-enhanced chemiluminescence as described previously (14). For transient transfection assays, HK2 or RCC 786-O cells were transiently transfected with p22 phox , pVP-FLAG (empty vector), or VHL-FLAG (generous gift from W. G. Kaelin) using the Lipofectamine 2000 (Invitrogen) as per the manufacturer's instructions. 48 h post-transfection, cell homogenates were prepared for lucigenin-enhanced chemiluminescence as described below. For Western blot analysis, cells were lysed in a modified radioimmune precipitation assay buffer (100 mM Tris-Cl, pH 8, 10 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, supplemented with protease inhibitor mixture (Sigma) and rotated at 4°C for 1 h. Insoluble debris was removed by centrifugation at 10,000 ϫ g for 15 min at 4°C. Protein concentration was determined by the Bradford assay (Bio-Rad).
Small interfering RNAs (siRNAs), designed by Dharmacon, were transfected at 200 nM in a double transfection using X-treme Gene. Briefly, RCC 786-O cells were plated in antibiotic-free medium to obtain ϳ40% confluency on the day of transfection. 200 nM scrambled control or specific siRNA was added to the cells. 24 h later, the medium was aspirated, and fresh medium minus antibiotics was added to the cells. The transfection was repeated, and 48 h later the cells were harvested for Western blot analysis.
Antibodies and Immunoblotting-Anti-Nox4 antibody was generated as described previously (15). Anti-HIF-2␣ and GAPDH were from Novus Biologicals, and anti-actin, anti-FLAG (M2), and ␤-tubulin were from Sigma. Nox1 and p22 phox were from Santa Cruz Biotechnology, pAkt and p4E-BP1 from Cell Signaling, and anti-HA from Roche Applied Science. Between 25 and 70 g of total protein was typically analyzed by immunoblotting. After exposure with the indicated primary antibody, the immunoblots were washed and incubated with goat anti-rabbit/mouse-coupled horseradish peroxidase (Bio-Rad) followed by chemiluminescence using ECL reagent (Amersham Biosciences).
Immunoprecipitation-Western-Assays were conducted using whole cell lysates generated with radioimmune precipitation assay buffer. For immunoprecipitations, the amount of lysate used was normalized to an equal amount of total protein, as determined by Bradford analysis (Bio-Rad) or direct Western blot of the protein of interest, and ranged between 200 and 500 g depending on expression levels. The lysates were immunoprecipitated using 8.8 g of FLAG M2 antibody (Sigma) or 2 g of HA antibody (Roche Applied Science). The immunoprecipitates were bound to protein A-or G-Sepharose (Amersham Biosciences), washed with radioimmune precipitation assay buffer, boiled, and analyzed by SDS-PAGE. Immunoblots were performed on the immunoprecipitated material or on 30 g of total cell lysate per gel lane.
Measurement of Intracellular ROS Production-Intracellular ROS were measured as described previously (16). Briefly, subconfluent cells were washed with Hanks' balanced salt solution without phenol red and then incubated for 15 min in the dark at 37°C with the same solution containing the peroxide-sensitive fluorophore 2,7-dichlorofluorescein diacetate (DCF-DA; Molecular Probes) at 5 mol/liter. DCF-DA fluorescence was detected at excitation and emission wavelengths of 488 and 520 nm, respectively, as measured with a multiwell fluorescence plate reader (Wallac 1420 Victor 2 , PerkinElmer Life Sciences).
NADPH Oxidase Assay-NADPH oxidase activity was measured by the lucigenin-enhanced chemiluminescence method as described (14). Briefly, cultured cells or tumor tissue was homogenized in lysis buffer (20 mM KH 2 PO4, pH 7.0, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 0.5 g/ml leupeptin) by using a Dounce homogenizer (100 strokes on ice). Homogenates were centrifuged at 800 ϫ g at 4°C for 10 min to remove the unbroken cells and debris, and aliquots were used immediately. To start the assay, 100-l aliquots of homogenates were added to 900 l of 50 mM phosphate buffer, pH 7.0, containing 1 mM EGTA, 150 mM sucrose, 5 M lucigenin, and 100 M NADPH. Photon emission in terms of relative light units was measured in a luminometer every 30 s for 5 min. There was no measurable activity in the absence of NADPH. Superoxide anion production was expressed as relative chemiluminescence (light) units (RLU)/mg protein. Protein content was measured using the Bio-Rad protein assay reagent. S.E. was calculated for each set of samples.
Soft Agar Growth Assays-Anchorage-independent growth was determined by the ability of cells to form colonies in soft agar. RCC 786-O cells were treated for 24 h in buffer alone (Me 2 SO) or DPI (10 M). After 24 h treatment, viable cells of both buffer-and DPI-treated (11,000 cells/plate) were grown in 0.3% agar on a cushion of 0.6% agar containing Me 2 SO (buffer) or DPI (10 M) in 6-well plates. RCC 786-O cell growth was allowed to proceed for ϳ25 days. Colonies were stained using a cell stain solution (Chemicon).
Tumorigenesis Assays-Nude mouse xenograft assays were performed as described (18) with minor modifications. Briefly, six-week-old male athymic nude (BALB/c, nu/nu) mice were injected with RCC 786-O cells. The cells, (1 ϫ 10 6 ) were resuspended in 100 l of phosphate-buffered saline and injected subcutaneously. Tumors were allowed to grow for 10 weeks after which treatment with DPI (1 mg/kg body weight/day) or vehicle alone was initiated. Tumor volume (mm 3 ) was monitored every week for 3 weeks using calipers. Mice were then euthanized and the tumors excised and snap-frozen. All experiments using mice were approved by the University of Texas Health Science Center at San Antonio Institutional Animal Care and Use Committee.

ROS Generation Is Increased in VHL-deficient RCC 786-O Cells as Compared with HK2 Renal Proximal Tubule Cells-Superoxide anion (O 2
Ϫ ) production in renal cancer cell lines was measured as lucigenin-enhanced chemiluminescence using NADPH as the substrate. Fig. 1, A and B, shows that the rate of superoxide formation in the VHL negative renal carcinoma cell line RCC 786-O is higher than in the HK2 cells or VHL positive renal carcinoma ACHN cells. The increased ROS generation in RCC 786-O cells was confirmed using the peroxide-sensitive fluorescence probe DCF-DA (Fig. 1C). These results indicate that VHL deficiency is associated with increased NADPH-dependent superoxide generation. To confirm a role for VHL in the regulation of ROS production, wild-type VHL-FLAG or empty vector were introduced into the VHL-deficient RCC 786-O cells. Anti-FLAG immunoblotting of cell lysates showed significant expression of VHL protein in VHL-transfected cells (Fig. 1D, compare lanes 2 and 1). ROS generation was measured in parallel by NADPH-dependent lucigenin-enhanced chemiluminescence. NADPH-dependent superoxide generation was suppressed in RCC 786-O cells transiently transfected with VHL-FLAG as compared with vector-transfected cells (Fig. 1E).
RCC 786-O cells constitutively express HIF-2␣ but do not express HIF-1␣ (19). We analyzed HIF-2␣ protein expression in lysates from 786-O cells transfected with VHL-FLAG or control vector-transfected cells. Fig. 1F shows a marked decrease in HIF-2␣ protein expression in VHL-FLAG-transfected RCC 786-O cells compared with the vector-transfected cells. Taken together, these data demonstrate that VHL inhibits NADPH-dependent ROS generation and down-regulates HIF-2␣ protein expression.
ROS Generated by Flavoprotein Oxidase(s) Maintain HIF-2␣ Protein Expression, an Effect of Diphenyleneiodonium-To analyze the effect of NADPH-dependent ROS generation on HIF-2␣ protein, we used DPI, an inhibitor of flavoprotein-containing oxidases. Fig. 2A shows that DPI decreases HIF-2␣ protein expression in a time-dependent manner. HK2 cells were used as a negative control for HIF-2␣ protein expression. These data indicate that ROS maintain HIF-2␣ protein expression. DPI inhibits electron transport by NAD(P)H oxidases as well as other flavin-containing enzymes. Major contributors to ROS generation that are NADPH-dependent and contain a flavoprotein domain include respiratory complexes of the mitochondria, nitric-oxide synthase (NOS), and the Nox family members. To confirm the source of intracellular ROS that are involved in HIF-2␣ protein expression, we used pharmacologic inhibitors to block other sources of ROS: rotenone, an inhibitor of complex I of the mitochondrial respiratory chain; antimycin, an inhibitor of complex III of the mitochondrial respiratory chain; L-NAME, an inhibitor of NOS; and DPI, a flavoprotein inhibitor. Fig. 2B shows that inhibitors of mitochondria, rotenone and antimycin, and the NOS inhibitor L-NAME did not affect HIF-2␣ protein expression after a 24-h incubation in RCC 786-O cells. On the other hand, DPI did cause downregulation of HIF-2␣ protein expression. These data indicate that the activity of NAD(P)H oxidase(s) is required for HIF-2␣ protein expression in VHL-deficient cells.
VHL Deficiency Is Associated with Up-regulation of NAD(P)H Oxidases-To determine which NAD(P)H oxidase isoforms are expressed in renal tubular epithelial cells and RCC cells, mRNAs were analyzed by RT-PCR, and the end products were resolved on an agarose gel. Both Nox1 and Nox4 mRNAs are expressed in HK2 cells as well as in RCC 786-O cells, whereas Nox2, Nox3, and Nox5 mRNAs were not detected (Fig. 3A). Expression of Nox1 mRNA is comparable in RCC 786-O and HK2 cells. However, Nox4 mRNA levels were higher in RCC 786-O cells compared with HK2 cells (Fig. 3A, lower panel versus upper panel). The steady state protein levels of these oxidases were also determined by Western analysis. Consistent with the RT-PCR results, Nox4 protein expression is higher in VHL-deficient RCC 786-O cells than in HK2 cells (Fig. 3B, Nox4). We did not detect any difference in protein expression of Nox1 in HK2 versus RCC 786-O cells by Western analysis (data not shown). We next analyzed the expression of p22 phox , the membrane-bound subunit that binds to and potentiates the activities of both Nox1 and Nox4. Strikingly, p22 phox levels were higher in the RCC 786-O VHL-deficient cell line compared with HK2 cells (Fig. 3B). It is likely that the increased ROS generation in VHL-deficient cells is due to the increased expression of p22 phox subunit required for activation of Nox1 and Nox4. To examine the involvement of VHL in regulation of p22 phox , VHL was transfected into VHL-deficient cells. Introduction of VHL resulted in a decrease in p22 phox levels compared with vector-transfected RCC 786-O cells (Fig. 3C).

Proteasome Inhibitors Increase NADPH Oxidase Activity and
Increase the Expression of p22 phox -VHL is part of an E3 ubiquitin ligase that regulates protein degradation via the 26 S proteasome. We next determined whether p22 phox protein expression was regulated through the 26 S proteasome. VHLexpressing HK2 cells were treated with proteasomal inhibitors. In the presence of MG132, NADPH-dependent superoxide generation was increased (Fig. 4A). Moreover, the proteasome inhibitors epoxomicin and MG132 increased the expression of p22 phox (Fig. 4, B and C). We did not observe changes in the steady state protein levels of Nox1 and Nox4 (data not shown). Taken together, these data suggest that a proteasomal pathway regulates p22 phox expression. p22 phox Interacts with VHL in Vivo-To examine whether p22 phox and VHL can interact in vivo, 293 cells were transfected   A, homogenates were prepared from buffer (Ϫ)-or MG132 5 M (ϩ)-treated HK2 cells and examined for NADPH-dependent superoxide anion generation using the lucigenin assay. B, HK2 cells were treated with increasing amounts of proteasome inhibitor epoxomicin or with buffer alone (Ϫ) for 24 h. Cell lysates were evaluated by Western analysis for p22 phox protein levels. GAPDH was used as a loading control. C, HK2 cells were treated with increasing amounts of proteasome inhibitor MG132 for 24 h or with buffer alone (Ϫ). Equivalent amounts of total protein from cell lysates were evaluated by Western analysis for p22 phox protein levels. GAPDH was used as a loading control.
with HA-tagged p22 phox (p22 phox -HA) and FLAG-tagged VHL (VHL-FLAG) followed by immunoprecipitation with HA antibodies and immunoblot analysis with FLAG antibodies. The data show that VHL-FLAG co-precipitate with p22 phox -HA (Fig. 5A). Importantly, when the same blot was probed for p22 phox , an immunoreactive band was detected at ϳ23 kDa, representing p22 phox -HA. In addition, several higher molecular mass species of p22 phox were observed (Fig. 5B). This suggested a posttranslational modification of p22 phox . To examine whether the higher molecular mass species could be an ubiquitinated form of p22 phox , we probed the blot for ubiquitin. Fig. 5C shows that antiubiquitin immunoreacts with the higher molecular mass species, which matches the pattern observed with anti-p22 phox . This suggests that p22 phox is modified by ubiquitin in vivo. In the reciprocal experiment, antibodies against FLAG were used for immunoprecipitation of VHL followed by immunoblot analysis for p22 phox . Western analysis of the immunoprecipitate from the transfected cell lysates indicates a p22 phox -immunoreactive band representing HA-tagged p22 phox (Fig. 5D). These data demonstrate that p22 phox and VHL interact in vivo. p22 phox and Nox Oxidases Regulate HIF-2␣ Protein Accumulation-To examine the contribution of Nox oxidase components to HIF-2␣ expression, we employed knockdown strategy. siRNAs to hNox1, hNox4, and hp22 phox were transiently transfected into RCC 786-O cells. Successful knockdown of Nox1 and Nox4 (Fig. 6A) and p22 phox (Fig. 6C) was evaluated by Western analysis. HIF-2␣ protein levels were decreased in siNox1 as well as in siNox4 knockdown cell lysates compared with scrambled control cell lysates. In cells transfected with both siNox1 and siNox4, we also observed a decrease in HIF-2␣ protein levels (Fig. 6B). Importantly, knockdown of p22 phox alone significantly decreased HIF-2␣ protein levels (Fig. 6D). These data suggest that Nox oxidases, including the subunit p22 phox , are important mediators that contribute to HIF-2␣ protein expression in VHL-deficient cells.
Nox Oxidases Increase HIF-2a Protein Accumulation through a Translational Mechanism-We determined the potential mechanisms of Nox-dependent HIF-2␣ protein accumulation in VHL-deficient cells. Cycloheximide, an inhibitor of protein translation, but not actinomycin D, an inhibitor of transcription, decreased HIF-2␣ protein (Fig. 7A). To further investigate a translational pathway in HIF-2␣ protein regulation, we evaluated a known pathway linked to translational regulation often activated in cancer, the PI3K/Akt pathway. PI3K/Akt-dependent signaling pathway culminates in phosphorylation of components essential for the cap-dependent translational machinery such as 4E-BP1. Incubation of RCC 786-O cells with the PI3K inhibitor LY29004 blocks HIF-2␣ protein accumulation (Fig. 7A, left panel), indicating that PI3K activity is required for HIF-2␣ protein expression. Inhibition of PI3K was also associated with decreased phosphorylation of Akt and 4E-BP1 (Fig.  7B). The role of Akt in HIF-2␣ expression was examined by FIGURE 5. p22 phox and VHL interact in vivo. A, human 293 cells were co-transfected with vector control or mammalian expression plasmids encoding VHL-FLAG tag and C-terminal HA-tagged p22 phox (p22 phox HA). Equivalent amounts of total protein were immunoprecipitated (IP) from transfected cell lysates (vector control or p22 phox HA) with anti HA followed by Western blot analysis (IB) with FLAG antibody. In the input lane, 15% of the total cell lysate were loaded. The solid arrow indicates the protein band corresponding to VHL-FLAG. B, the same blot as used for A was probed with an antibody against p22 phox (anti-p22 phox ). Note that the p22 phox antiserum immunoreacts with a band of ϳ22 kDa (p22 phox HA) as well as higher molecular mass species. C, the same blot as used for A was probed with an antibody against Ubiquitin (Ub). Note that the ubiquitin antiserum correlates with p22 phox . D, human 293 cells were co-transfected as described in A, and equivalent amounts of total protein were immunoprecipitated from transfected cell lysates (IP; vector control or p22 phox HA) with anti-FLAG followed by Western blot analysis with anti p22 phox antibody. Input shows the total cell lysate analyzed by Western blotting alone. infecting RCC 786-O cells with an adenovirus vector expressing a dominant-negative mutant of Akt (Ad DN-Akt). GFP adenovirus was infected as a control. Expression of Ad DN-Akt inhibited 4E-BP1 phosphorylation and blocked HIF-2␣ protein accumulation similar to the effect of inhibition of PI3K and protein synthesis (Fig. 7, A and B, CHX). Cycloheximide, LY29004, and wortmannin had no effect on NADPH oxidase activity, suggesting that ROS generation is upstream of the translational pathway (data not shown).
To elucidate the role of Akt as a potential mechanism by which p22 phox -based Nox oxidases regulate HIF-2␣ protein accumulation, siRNA-mediated knockdown of p22 phox , Nox1, and Nox4 was performed in RCC 786-O cells, and the phosphorylation of Akt was evaluated by Western blot analysis. Knockdown of p22 phox was as efficient as either Nox alone or in combination (siNox1 ϩ siNox4) in decreasing the phosphorylation of Akt (Fig. 7C). Importantly, the phosphorylation of the downstream target of PI3K/Akt pathway, 4E-BP1, was also efficiently inhibited by the down-regulation of p22 phox (Fig. 7D). These data suggest that p22 phox is a pivotal component of the Nox oxidase that regulate the PI3K/Akt/4E-BP1 signaling cascade leading to enhanced translation of HIF-2␣ protein expression in VHL-deficient cells. However, other mechanisms of HIF-2␣ regulation cannot be excluded.

DPI Blocks RCC 786-O Colony Formation in Soft
Agar-Because HIF-2␣ expression is maintained in cells with increased flavoprotein oxidase-dependent ROS generation, we determined the effect of inhibition of ROS generation on tumor cell growth in vitro. Colony formation in agar gel was used as an in vitro assay for anchorage-independent growth. Colony formation by RCC 786-O cells in soft agar medium was assessed in the presence or absence of DPI. Fig. 8A shows RCC 786-O cells forming colonies in soft agar (left panel), which was inhibited in the presence of DPI (right panel). The number of colonies were counted and normalized as percent of control (Fig. 8B). Moreover, we found that cellular proliferation was markedly decreased in DPI-treated cells compared with cells treated with vehicle alone (Fig. 8C). Taken together, these data indicate that inhibition of ROS generation results in down-regulation of HIF-2␣ protein expression in RCC cells in the absence of VHL and is associated with a decrease in the growth of tumor cells.
DPI Blocks RCC 786-O Tumor Growth in Vivo-We next examined the ability of DPI to block tumor formation in an in vivo model. RCC 786-O cells were administered subcutane-  ously into nude mice. After 10 weeks of growth, tumor size was measured, and DPI or vehicle was injected subcutaneously at 1 mg/kg body weight/day. Animals were followed for an additional 3 weeks, and the tumor size was measured each week. Tumor enlargement was measured as -fold increase from the start of treatment (10 weeks) until sacrifice (13 weeks). There was a significant decrease in tumor mass in the DPI-treated animals Fig. 9A. Fig. 9B confirms that DPI did block NADPHdependent oxidase activity in the tumors. Therefore, blocking ROS generation by NAD(P)H-dependent oxidases, most likely Nox1 and Nox4, inhibits tumor growth in athymic mice.

DISCUSSION
This study demonstrates that VHL deficiency increases the expression and activity of NAD(P)H oxidases. ROS generation through p22 phox and activation of Nox1 and Nox4 help maintain HIF-2␣ expression and thereby contribute to renal carcinogenesis. Nox oxidases represent targets for treatment of renal cell carcinoma. Different approaches were used to detect ROS generation in normal renal epithelial cells (HK2), VHL(ϩ/ϩ) renal carcinoma cells (ACHN), and VHL(Ϫ/Ϫ) RCC 786-O cell lines. NAD(P)H oxidase activity and ROS production in the VHL-deficient renal cancer cell line RCC 786-O is much higher than in normal renal epithelial HK2 cells or VHL (ϩ/ϩ) ACHN cells. Importantly, we have shown that reintroducing VHL into the VHL-deficient cells reduces ROS production and decreases p22 phox protein levels. To our knowledge, this is the first evidence that VHL regulates NADPH-dependent ROS generation.
VHL is mutated in 80% of sporadic renal carcinomas, and as a result, HIF transcription factors are constitutively expressed in an oxygen-independent manner. HIF expression in VHLdeficient cells leads to the up-regulation of genes involved in renal carcinogenesis. In nude mice, down-regulation of HIF-2␣ is sufficient to impair the growth of tumors (18,20). Because HIF-2␣ is a critical downstream target of VHL with respect to suppression of renal carcinogenesis, it is important to identify the pathways involved in the regulation of HIF-2␣ protein expression. VHL-deficient cells provide a unique opportunity to discover and evaluate these upstream mediators. The increased ROS generation in VHL-deficient cells, maintains the expression of HIF-2␣ protein levels, because inhibition of ROS generation by the flavoprotein inhibitor, DPI, results in a decrease of HIF-2␣ protein levels in RCC 786-O cells. Similar results were obtained in another VHL-deficient cell line that expresses HIF-2␣ (A498; data not shown). Taken together, our data suggest a novel association between NADPH-dependent ROS generation and HIF-2␣ protein expression. We have extensively characterized the source of enzymatic activity involved in the expression of HIF-2␣. Flavoprotein-containing ROS-generating enzymes include NOS, complexes of the mitochondrial respiratory chain, and the NAD(P)H oxidases of the Nox family. Previous studies have implicated the generation of ROS by mitochondria or NAD(P)H oxidases as mediators of HIF-1␣ protein stabilization and activation (8,21). Our data strongly support a role for the Nox proteins in maintaining HIF-2␣ expression based on several observations. Inhibitors of complex I and III of the mitochondrial respiratory chain or inhibition of NOS had no effect on HIF-2␣ protein expression. DPI is often used to delineate the role of the Nox family of flavoproteins. Indeed, DPI elicited a marked decrease of HIF-2␣ expression in RCC 786-O cells. We have identified Nox1 and Nox4 as the major Nox catalytic sources of superoxide anion production. Nox4 and p22 phox (a known interacting protein with Nox1 and Nox4) are highly expressed in VHL-deficient cells compared with the HK2 cells. The levels of Nox4 and p22 phox are also elevated in another VHL-deficient cell line (A498; data not shown). It has been shown previously that a subset of brain tumor cell lines derived from human glioblastomas aberrantly express Nox4 (17). It is interesting to note that VHL is often mutated in glioblastomas (22). p22 phox is an integral membrane protein that directly binds to and is required for the activation of both Nox1 and Nox4 (23,29). In particular, Nox1 activity is dependent on a stable complex formation with p22 phox and other regulatory subunits (23)(24)(25)(26)(27)(28)(29). Nox4, on the other hand, produces a significant amount of superoxide in a constitutive manner; this production is increased upon its association with p22 phox . Knockdown of endogenous p22 phox by RNAi results in reduced Nox4 activity (23,29). Our data show a decrease in p22 phox protein levels and NADPH-dependent ROS generation upon reintroduction of VHL into VHL-deficient cells. The regulation of p22 phox is currently unknown. We show a role for the 26 S proteasome in p22 phox regulation, because proteasomal inhibitors increase p22 phox protein levels and NADPH-dependent superoxide generation. We also show for the first time that p22 phox binds VHL in vivo. Moreover, we provide evidence that p22 phox is a target for ubiquitination, providing a potential mechanism of p22 phox regulation in VHLdeficient cells. We hypothesize that VHL binding to p22 phox inhibits the activation of the Nox oxidase complex and leads to destabilization of p22 phox protein.
The data also demonstrate an important role for specific Nox oxidases in downstream HIF-2␣ protein accumulation. In RCC 786-O cells transiently transfected with siRNA to Nox4, we observed a small but significant decrease in HIF-2␣ protein levels. Knockdown of Nox1 produces a clear inhibition of HIF-2␣ protein levels. Interestingly, it has been shown that up- Athymic mice were sacrificed 13 weeks after subcutaneous administration of RCC 786-O cells. DPI was injected daily subcutaneously 10 weeks after introduction of the cells into nude mice. A, the data were expressed as the means Ϯ S.E. of the -fold increase in tumor size calculated for each group. *, p Ͻ 0.05 compared with control. B, to test the efficiency of DPI treatment, NAD(P)H oxidase activity was evaluated in tumor homogenates from both groups, vehicle alone (Ϫ) or DPI-treated animals. NADPH-dependent ROS generation was measured by lucigenin-enhanced chemiluminescence (RLU, relative light units). The data were quantitated and the results expressed as the means Ϯ S.E. **, p Ͻ 0.01 compared with control. regulation of Nox1 activity by hypoxia enhances HIF-1␣-dependent gene expression via increased ROS (21). However, our data show that even down-regulation of p22 phox alone can markedly decrease the protein expression of HIF-2␣. The mechanism by which ROS generation maintains HIF-2␣ protein expression in VHL-deficient cells requires p22 phox -based Nox1 and Nox4 regulation of the PI3K/Akt pathway, culminating in the phosphorylation of 4E-BP1 for ongoing protein translation. Similar regulation of HIF-1␣ has been demonstrated. For example, attenuation of PI3K/Akt decreased HIF-1␣ protein in VHL-deficient renal cell carcinoma (RCC4) without altering the steady state of HIF-1␣ mRNA (30). Additionally, it has been shown that maintenance of the nonhypoxic ROS-dependent increase of HIF-1␣ protein expression by angiotensin II relies on ongoing protein translation in a p22 phox /PI3K-dependent manner (31). We determined the functional consequences of inhibition of ROS generation. Colony formation by RCC 786-O in soft agar was markedly inhibited by DPI. To determine the relevance of our in vitro observations to tumor growth in vivo, athymic mice harboring RCC 786-O tumors were treated with DPI or vehicle. DPI-treated animals had a significant decrease in tumor size. Our data indicate that increased expression and activation of NAD(P)H oxidases result in enhanced ROS generation in VHL-deficient cells. Specific catalytic NAD(P)H oxidases Nox1 and Nox4 and their regulatory subunit p22 phox are up-regulated and activated by VHL deficiency, leading to increased HIF-2␣ expression and tumorigenesis in vitro and in vivo. Thus, pharmacological inhibition of flavoprotein oxidases, and in particular the blockade of p22 phox binding to these oxidases, represents an alternative approach to the down-regulation of HIF-2␣ expression and its target pathways involved in renal carcinogenesis.