Role of Exon 2-encoded b -Domain of the von Hippel-Lindau Tumor Suppressor Protein*

,

Inactivating mutations of the von Hippel-Lindau (VHL) 1 tumor suppressor gene are associated with inherited VHL cancer syndrome, of which afflicted individuals are at risk to develop a wide variety of tumors including clear cell renal cell carcinoma (RCC) (1)(2)(3). Biallelic inactivating mutations of the VHL gene are also associated with sporadic RCC, the most common form of kidney cancer in humans (4,5). Reintroduction of wild-type VHL in VHL Ϫ/Ϫ RCC cells is sufficient to suppress their ability to form tumors in nude mice (6). The VHL gene contains three exons that code for a 213-residue protein. VHL protein assembles with at least four other associated proteins: elongin B, elongin C, cullin-2, and Rbx (the complex will be hereafter referred to as VBC/Cul-2) (7)(8)(9)(10)(11). VBC/Cul-2 is an E3 ubiquitin ligase that targets the ␣-subunits of hypoxia-inducible factor (HIF␣) for oxygen-dependent degradation (12)(13)(14)(15)(16)(17)(18). HIF␣ are stabilized by hypoxia and play an important role in the activation of hypoxia-inducible genes such as the vascular endothelial growth factor and glucose transport-1 (Glut-1) (8, 13, 19 -22). HIF␣ are stable in VHL Ϫ/Ϫ RCC cells bringing about a constitutive "hypoxia-like" response, regardless of oxygen concentration. The exact mechanism by which VHL can mediate the degradation of HIF␣ is still unknown but might be related to its ability to shuttle between the nucleus and the cytoplasm (6,(23)(24)(25)(26)(27). Another key functional characteristic of VHL is that it binds to fibronectin and is involved in the assembly of an extracellular fibronectin matrix (28).
The crystal structure of VHL has been reported. VHL mainly consists of two independent domains: a large ␤-domain that spans residues 64 -154 and an ␣-helical domain (␣-domain) that encompasses most of the C-terminal part of the protein (residues 157-189) (29). Tumor-derived inactivating mutations (279 entries; Ref. 30) are found across the VHL protein, indicating that both domains play a critical role in VHL tumor suppressor function (29). There is, however, an interesting correlation between the nature and localization of inactivating mutations and the clinical consequences in patients afflicted with inherited VHL syndrome. Individuals with type II VHL syndrome develop pheochromocytoma and have generally inherited a mutation in the exon 3-encoded ␣-domain. Type I VHL syndrome differs from type II in that patients do not develop pheochromocytoma and are likely to have inherited a mutation in the hydrophobic core of the ␤-domain (31). There is also an intriguing disparity in the distribution of tumor-derived missense mutations between the inherited and sporadic form of RCC. Mutations in the ␣-domain of VHL are much more frequent in the inherited form of RCC (5). The role of a few of these residues is well understood, since they correspond to the ones required for VHL binding to elongin C and formation of the VBC/Cul-2 complex (7,8,10,32). In contrast to inherited RCC, sporadic RCC frequently harbor inactivating point mutations in exon 2. This includes point mutations at the exon 2 boundary that cause a splice defect producing a mRNA that lacks exon 2 sequences altogether (5). Exon 2 mutations are rare in VHL patients, and it has been suggested that such mutations might not be easily tolerated and thus not transmit-ted in the germ line (1). The discrepancy in the distribution of mutations between sporadic and inherited RCC suggests that exon 2-associated mutations might inactivate VHL function in a different way than exon 3-associated mutations. Exon 2-encoded residues 114 -154 are mostly hydrophobic and form three of the seven ␤-strands of the ␤-domain (29). These residues are hypothesized to play a role in substrate protein recognition, although recent in vitro experiments have revealed that they might not be required for VHL binding to HIF␣ (33). Therefore, the role that exon 2-encoded sequences play in VHL-mediated tumor suppression is still poorly understood. Here, we further examine the role of these sequences in cells by comparing a tumor-derived VHL mutant that lacks residues 114 -154 with the known biochemical properties of wild-type VHL and mutant VHL lacking the exon 3-encoded ␣-domain. We show that the exon 2-encoded ␤-domain plays two independent roles in binding to HIF␣ and fibronectin and mediating transcriptiondependent nuclear/cytoplasmic trafficking of the VBC/Cul-2 complex. The use of a novel method to examine the energy requirement for nuclear import of proteins will also be discussed in this report. The results presented here support the model that the ␤-domain of VHL is involved in substrate recognition and nuclear/cytoplasmic trafficking.

Cell Culture, Transfections, and Adenoviral Infections
The VHL Ϫ/Ϫ 786-0 RCC cells and HeLa cells were obtained from the American Type Culture Collection (Manassas, VA). The VHL-GFP cell line corresponds to 786-0 cells stably transfected with the VHL-GFP fusion protein as described elsewhere (25). The 117-4 (VHL Ϫ/Ϫ ; referred to as 117) cells were a kind gift from Dr. James R. Gnarra (LSU Health Sciences Center, New Orleans, LA). The mouse embryonic fibroblasts (MEFs) nullizygous for HIF-1␣ were a kind gift from Dr. Randy Johnson (Department of Biology, University of California, San Diego) (34). Cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum in a 37°C, humidified, 5% CO 2 -containing atmosphere incubator. Transient transfections were performed overnight using a standard calcium phosphate method. Viruses were used as freeze/thaw lysates, and all infections were also performed overnight.

Expression Vectors and Constructs
The human VHL cDNA, which codes for a 213-amino acid VHL protein, was subcloned into pcDNA3.1(Ϫ) (Invitrogen) vector. A FLAG epitope tag (DYKDDDDK) was added to the N terminus of the VHL cDNA open reading frame. A cDNA coding for an enhanced fluorescence version of GFP (Fred 25; Ref. 35) was subcloned at the C terminus VHL to produce the VHL-GFP fusion protein. A deletion mutant of the last 56 amino acids was fused to GFP to produce the ⌬E3-GFP fusion protein. Another deletion mutant of VHL lacking amino acids 114 -154 was fused to GFP to produce the ⌬E2-GFP fusion protein. Two GFPs in tandem were cloned into pcDNA 3.1(Ϫ) to produce the GFP-GFP fusion protein. VHL-GFP-NES, ⌬E2-GFP-NES, and GFP-GFP-NES were produced by fusion of VHL-GFP, ⌬E2-GFP, and GFP-GFP at their C terminus to the strong nuclear export signal (NES) of human immunodeficiency virus REV, LPPLERLTL (NES) (36). All constructs were verified by standard DNA sequencing.

Construction of Adenovirus Vectors through
Cre-lox Recombination CRE8 and 293 cells were obtained from Dr. David Park (University of Ottawa, Ottawa, Ontario, Canada) and cultured in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal calf serum (FCS). The construction and properties of CRE8 cells are described elsewhere (37). The pAdlox vector and the 5 viral DNA were also obtained from Dr. Park. The three different VHL constructions (VHL-GFP, ⌬E2-GFP, and ⌬E3-GFP) were previously subcloned in the pAdlox vector. Transfections were done according to Graham and van der Eb (38). Typically, a confluent 10-cm diameter dish of CRE8 cells (1.6 ϫ 10 7 ) was split into 5-6-cm diameter dishes for transfection 2-4 h later. Each dish received 3 g of pAdlox vector (containing the foreign VHL construction) and 3 g of 5 viral DNA in a final volume of 0.5 ml of CaPO 4 , which was applied to the cells for 16 h. The 10% fetal calf serum DMEM was changed for 2% fetal calf serum DMEM 16 h following the transfection. Cells were fed with fresh 2% DMEM after 64 h. Between day 8 and 10, we had a sizable infection in each dish; almost all of the cells were rounded up or floating. Cells were harvested and subjected to freeze/ thaw three times with an alternating liquid N 2 /37°C water bath. The virus was then passed sequentially through CRE8 cells twice. Finally, a plaque purification assay was performed in order to isolate the recombinant virus expressing adVHL-GFP, ad⌬E2-GFP, or ad⌬E3-GFP. The recombinant viruses were then amplified in CRE8 cells to high titer.

Nuclear Import Assay in Living Cells
HeLa cells were plated on a 35-mm dish with a hole at the bottom replaced by a glass coverslip and transfected overnight with VHL-GFP-NES, ⌬E2-GFP-NES, and GFP-GFP-NES. Cells were washed with PBS and incubated for 2 h in DMEM at 37°C with or without metabolic poisons (6 mM 6-deoxyglucose and 0.02% sodium azide), at 4°C, or at 37°C with 10 M leptomycin B (39 -40).

In Vitro Ubiquitination Assay
VHL Ϫ/Ϫ 786-0 RCC cells infected with the three different adenoviruses and 786-0 cells stably expressing VHL-GFP were lysed in the presence of 1% Triton X-100, 20 mM Tris-HCl, pH 8.0, 137 mM NaCl with protease mixture for 30 min. at 4°C. Whole cell lysates were first immunoprecipitated with anti-FLAG M2 monoclonal antibody. Precipitates were washed five times with a buffer containing 20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 2 mM dithiothreitol. The total volume of the reaction mixture was adjusted to 20 l. E1 ubiquitin-activating enzyme (100 ng), E2 ubiquitin-conjugating enzyme (200 ng), 0.5 g of ubiquitin aldehyde, 0.5 g of ubiquitin, and an ATP-regenerating system (0.5 mM ATP, 10 mM creatine phosphate, and 10 g of creatine phosphokinase) were added to the reaction mixture (complete mixture). The reaction was stopped after 2 h of incubation at 37°C by adding 4ϫ SDS loading buffer. Samples were boiled 10 min and separated on an 8% SDS-PAGE and blotted onto a PVDF membrane. Blots were blocked and incubated in the presence of a mouse anti-ubiquitin antibody (Berkeley Antibody Company). The E1 ubiquitin-activating enzyme and the E2 ubiquitinconjugating enzyme were a kind gift from Dr. Kazuhiro Iwai (Kyoto University, Kyoto, Japan).
Immunoprecipitation of Cullin-2, NEDD8, and Fibronectin-VHL-GFP cells and infected 786-0 cells were lysed in 100 mM NaCl, 0.5% Igepal CA630, 20 mM Tris-HCl (pH 7.6), 5 mM MgCl 2 , and 1 mM sodium orthovanadate with 2 g/ml leupeptin, 2 g/ml aprotinin, and 1 g/ml pepstatin A. After clearance by centrifugation, 1-mg aliquots of lysate were incubated for 2 h at 4°C with anti-FLAG M2 beads. Beads were washed, boiled, and loaded on an 8% SDS-PAGE and blotted onto PVDF membranes using standard methods. Blots were blocked with 3% milk powder in PBS containing 0.2% Tween 20 and were then incubated in the presence of anti-cullin-2 (Ref. 41; provided by Dr. Arnim Pause, Max-Plank Institute, Germany), anti-NEDD8 (Alexis), anti-fibronectin (Dako Diagnostic), or anti-FLAG M2 monoclonal antibody (Sigma). For total cell lysates, cells were washed several times in PBS, scraped from the Petri dishes, centrifuged, and resuspended in 4% SDS in PBS (24). The samples were boiled for 5 min, and the DNA was sheared by passage of lysates through 19-gauge needles. Protein concentration was determined by bicinchoninic acid method (Pierce) and was used to normalize protein loading in a whole-cell immunoblot assay.

Immunofluorescence Staining
For Fibronectin-VHL Ϫ/Ϫ RCC 786-0 cells or VHL-GFP cells were infected and were grown on coverslips for 6 days, washed three times with PBS, and fixed/permeabilized in prechilled 95% ethanol at Ϫ20°C for 30 min. Ethanol was then aspirated, and the residual ethanol was allowed to air dry at 4°C. Cells were stained with polyclonal antifibronectin antibody (5 g/ml) (Dako Diagnostic) for 1 h at room temperature. The coverslips were then washed with PBS three times and incubated with Cy TM 3-conjugated anti-rabbit antibody (Jackson Immu-noResearch, PA) diluted 1:1000 for 1 h at room temperature. Coverslips were washed three times with PBS, incubated for 2 min with Hoechst 33342, and mounted with fluoromount-G (Southern Biotechnology Associates) on slides.
For HIF-1␣-117 cells or transiently transfected 786-0 with HIF-1␣ were grown on coverslip and infected with the three different adenoviruses overnight. Cells were washed three times with PBS, fixed/permeabilized in PBS containing 4% formaldehyde for 30 min at room temperature, washed again three times with PBS, and incubated for 1 h at room temperature with anti-HIF-1␣ antibody (Transduction Laboratories, Lexington, KY) diluted 1:1000 in PBS, 1% Triton X-100, 10% FCS. The cells were washed in PBS and incubated for 60 min in the presence of a Cy TM 3-conjugated anti-mouse antibody (Jackson ImmunoResearch) diluted 1:1000. The cells were washed in PBS, incubated for 2 min in Hoechst 33342, and mounted with fluoromount-G on slides.

Fluorescence Analysis and Image Processing
GFP fluorescence images were captured using a Zeiss Axiovert S100TV microscope with a C-Apochromat 40 ϫ water immersion objec-tive, equipped with an Empix digital charge-coupled device camera using Northern Eclipse software. Images were manipulated with Northern Eclipse and Adobe Photoshop software as described elsewhere (25). GFP images were always taken before Hoechst images to minimize any possible bleaching effect.

RESULTS
Biochemical Characterization of the Exon 2-Encoded ␤-Domain of VHL-The VHL protein, encoded by the VHL gene that contains three exons, can be divided into three independent domains: an acidic domain, a ␤-domain, and an ␣-domain (Fig.  1A). Sporadic RCC frequently harbor inactivating mutations in the exon 2-encoded part of the ␤-domain, whereas these mutations are relatively rare in individuals afflicted with inherited VHL syndrome (5). To study the role of exon 2-encoded ␤-domain in VHL tumor suppressor function, a cDNA encoding a tumor-derived truncation of residues 114 -154 was fused to GFP to produce the ⌬E2-GFP fusion protein (Fig. 1B). This truncation mutant is the consequence of point mutations that cause a splice defect producing a mRNA that lacks exon 2 sequences altogether (5). ⌬E2-GFP is predicted to have a partial, if not total, loss of ␤-domain function while retaining an intact, exon 3-encoded ␣-helical domain. A tumor-derived truncation of the exon 3-encoded ␣-helical domain (last 56 C-terminal residues), which retained intact the sequences of the ␤-domain, was also fused to GFP (⌬E3-GFP) (Fig. 1B). ⌬E2-GFP, ⌬E3-GFP, and wild-type VHL-GFP were cloned in pAdlox vector, and adenoviruses (ad⌬E2-GFP, ad⌬E3-GFP, and adVHL- GFP) were produced to high titers (Fig. 1B) (42). Adenovirus was chosen as a method to reintroduce VHL, since it eliminates the necessity to produce stable clones of different VHL Ϫ/Ϫ RCC cell lines. VHL Ϫ/Ϫ RCC cells were infected with very high efficiency, with essentially 100% of cells displaying GFP fluorescence (Fig. 1C). In adenovirus-infected cells, adVHL-GFP was mostly localized to the cytoplasm with some nuclear signal, consistent with data obtained with stable transfectants. In contrast to VHL alone (without GFP; Ref. 43), adVHL-GFP did not restrain proliferation of VHL Ϫ/Ϫ RCC cells or other cell lines such as 293 cells, even when expressed to very high levels (data not shown). Glut-1 protein levels were significantly decreased in VHL Ϫ/Ϫ 786-0 RCC infected with adVHL-GFP in normoxia compared with uninfected cells or cells infected with an adenovirus that expressed GFP alone (data not shown). Western blot analysis indicated that ad⌬E2-GFP accumulated to levels similar to those of adVHL-GFP and ad⌬E3-GFP, suggesting that ad⌬E2-GFP is a stable protein (Fig. 1D). We conclude that the adVHL-GFP protein produced from an adenovirus is a functional molecule and shares similar characteristics with VHL.
We next examined the biochemical properties of ad⌬E2-GFP in comparison with adVHL-GFP and ad⌬E3-GFP. The ␤-domain mutant ad⌬E2-GFP still retained the ability to assemble with cullin-2 ( Fig. 2A) and to exhibit E3 ubiquitin ligase activity in vitro (Fig. 2B) to levels similar to those observed for adVHL-GFP. The ␣-helical domain deletion mutant (ad⌬E3-GFP) failed to assemble with cullin-2 and to display E3 ubiquitin ligase activity in vitro, as expected. While the experiments described above were being performed, it was noticed that a second band, which migrated slower than cullin-2, was found in the adVHL-GFP lane but was lacking from the ad⌬E2-GFP lane (Fig. 2A). NEDD8 is a ubiquitin-like molecule, which is conjugated to cullin-2 in a VHL-dependent manner (41,44). Western blotting with an anti-NEDD8 antibody revealed that the slower migrating form of cullin-2 is conjugated to NEDD8 (Fig. 2C). Therefore, an intact exon 2-encoded ␤-domain is not required for VHL ability to assemble with cullin-2 and to function as an E3 ubiquitin ligase in vitro but is necessary for VHL-mediated NEDD8 conjugation on cullin-2.
Exon 2-encoded ␤-Domain Is Required for VHL Binding to Fibronectin and Proper Assembly of a Fibronectin Extracellular Matrix-VHL Ϫ/Ϫ RCC cells are unable to promote assembly of an extracellular fibronectin matrix, and the reintroduction of VHL was shown to be sufficient to correct this defect (28). Adenovirus-mediated reintroduced adVHL-GFP displayed similar activity as VHL and restored the ability of VHL Ϫ/Ϫ RCC cells to properly produce a fibronectin extracellular matrix ( Fig. 3A; VHL-GFP). In contrast, ad⌬E2-GFP was unable to rescue this defect (Fig. 3A). Fibronectin was observed in an endoplasmic reticulum-like intracellular distribution in uninfected cells as well as in cells expressing ad⌬E2-GFP. Immunoprecipitation analysis revealed that adVHL-GFP was able to assemble with fibronectin, whereas ad⌬E2-GFP failed to do so (Fig. 3B). The ad⌬E3-GFP was also unable to bind to fibronectin and correct the fibronectin deposition defect of VHL Ϫ/Ϫ RCC. Therefore, VHL requires an exon 2-encoded ␤-domain to bind to fibronectin and mediate proper extracellular matrix formation.
Role of Exon 2-Encoded ␤-Domain of VHL in Oxygen-dependent Degradation of HIF␣-It was recently shown that one of the major defects of VHL Ϫ/Ϫ RCC cells is their inability to mediate oxygen-dependent degradation of HIF␣, and reintroduction of wild-type VHL was sufficient to correct this defect (13). In vitro studies have also revealed that truncation mutants of exon 2 and exon 3 of VHL are still able to bind to HIF␣ (33), which probably assemble with sequences encoded by exon 1 (residues 64 -113) (12). Adenovirus-mediated reintroduction of adVHL-

FIG. 2. Biochemical characterization of an exon 2-encoded ␤-domain mutant of VHL.
A, an intact ␤-domain is not required for VHL ability to assemble with cullin-2. Stable VHL Ϫ/Ϫ RCC 786-0 cells stably expressing FLAG-tagged VHL-GFP or 786-0 cells infected or not infected with the adenoviruses ad⌬E2-GFP, ad⌬E3-GFP, and adVHL-GFP were lysed and immunoprecipitated with anti-FLAG M2 beads. Precipitated proteins were run on SDS-PAGE (8% acrylamide) and transferred on PVDF membranes. The membranes were then blocked and incubated in the presence of a rabbit anti-cullin-2 (top panel) or a mouse anti-FLAG M2 (bottom panel) antibody. Notice that a second band migrates slower than cullin-2 in the VHL-GFP and adVHL-GFP lanes only. This represents NEDD8 conjugation to cullin-2 (further confirmed in C). B, an intact ␤-domain is not required for VHL ability to function as an E3 ubiquitin ligase in vitro. In vitro ubiquitination reactions were performed as described under "Materials and Methods" (complete mixture) except for two negative controls; adVHL-GFP was immunoprecipitated and incubated with the complete mixture except the E1 enzyme (first lane from the left) or the E2 enzyme (third lane from the left). Reactions were stopped by adding 4ϫ sample buffer. Samples were electrophoresed in 8% SDS-PAGE and transferred to a PVDF membrane. The membrane was then blocked and incubated in the presence of a mouse anti-ubiquitin antibody. C) Exon 2-encoded ␤-domain is required for VHL-mediated NEDD8 conjugation to cullin-2. Immunoprecipitations were performed exactly like for cullin-2. Immunoprecipitated proteins were run on an 8% SDS-PAGE and transferred to a PVDF membrane. The membrane was then blocked and incubated in the presence of a rabbit anti-NEDD8 antibody.
GFP was sufficient to restore VHL Ϫ/Ϫ RCC cell line 117 (HIF-1␣) and 786-0 (HIF-2␣) ability to mediate degradation of HIF␣ in normoxia (Fig. 4A). HIF␣ levels were not affected by expression of ad⌬E2-GFP or ad⌬E3-GFP (Fig. 4A). We notice that adVHL-GFP assembled with a significant amount of HIF␣ (1␣ and 2␣) in hypoxia and in the presence of the proteasome inhibitor CI. In contrast to data obtained in vitro, immunoprecipitation analysis revealed that ad⌬E2-GFP and ad⌬E3-GFP failed to bind to HIF␣ in adenovirus-infected cells (Fig. 4B, top  panels). We did not detect binding of HIF␣ to ad⌬E2-GFP and ad⌬E3-GFP in cells expressing low to very high levels of the fusion proteins (data not shown). These results indicate that an intact exon 2-encoded ␤-domain, as well as the ␣-domain, is required for VHL assembly with HIF␣ in cells.
It has been hypothesized that HIF-1␣ requires a hypoxic environment to import in the nucleus most likely assembled into complexes that contain VBC/Cul-2 (33,45). To further examine the role of hypoxia and VHL in nuclear import of HIF␣, the subcellular localization of endogenous HIF-1␣ was examined by immunofluorescence in VHL Ϫ/Ϫ 117 RCC cells uninfected or infected with different VHL constructs. Data shown in Fig. 4C revealed that endogenous HIF-1␣ accumu-lated exclusively in the nucleus of uninfected VHL Ϫ/Ϫ RCC 117 cell line although these cells were incubated in normoxia (Fig.  4C, a, e, and i). This demonstrates that HIF-1␣ is able to import in the nucleus even in the presence of oxygen and in the absence of VHL. A strong HIF-1␣ nuclear signal was also observed in cells expressing ad⌬E2-GFP (Fig. 4C, c, g, and k) as well as ad⌬E3-GFP (Fig. 4C, d, h, and l), whereas it was essentially undetectable in cells expressing reintroduced ad-VHL-GFP (Fig. 4C, b, f, and j). We then examined the subcellular localization of overexpressed HIF-1␣ in RCC VHL Ϫ/Ϫ 786-0 cells (which do not express endogenous HIF-1␣). A strong HIF-1␣ signal was detected exclusively in the nucleus of normoxic RCC VHL Ϫ/Ϫ 786-0 cells transiently transfected with HIF-1␣ cDNA that were either uninfected (Fig. 4C, m), infected with GFP alone (data not shown), or infected with ad⌬E2-GFP (Fig. 4C, o) and ad⌬E3-GFP (Fig. 4C, p). The addition of proteasome inhibitors or incubation in hypoxia led to nuclear accumulation of endogenous or overexpressed HIF␣ regardless of the presence of adVHL-GFP or mutants, as expected (data not shown). HIF-1␣ was also detected in the nucleus of normoxic RCC VHL Ϫ/Ϫ 786-0 cells when co-transfected with different smaller deletion mutants of exon 2, with a substitution at residue 117 in exon 2 or at residue 98 in exon 1, fused to GFP (Fig. 4D). These results demonstrate that HIF-1␣ is able to import in the nucleus regardless of oxygen concentration or assembly with VHL.

Exon 2-encoded Residues Mediate Transcription-dependent Nuclear/Cytoplasmic Trafficking of VHL Independently of Assembly with HIF␣ and Oxygen Concentration-
We recently demonstrated that VHL mediates transcription-dependent nuclear/cytoplasmic trafficking of the VBC/Cul-2 complex (25,46). The addition of 5,6-dichlorobenzimidazole riboside (DRB), an inhibitor of RNA polymerase II activity, causes an important increase of nuclear VBC/Cul-2 by blocking VHL-mediated nuclear export of the complex. The dependence on transcription for trafficking is abolished by a deletion on exon 2-encoded sequences (25). We next wanted to determine if exon 2-encoded residues also regulate subcellular trafficking of VHL in conditions known to affect HIF␣ stability, such as oxygen concentration, and if it is able to do so independently of assembly with HIF␣. Since ad⌬E2-GFP is a small molecule (40 kDa), its presence in the nucleus (Fig. 5c) might be simply the outcome of unregulated passive diffusion through the nuclear pore com-plex rather than by the utilization of signal-mediated and -regulated energy-dependent processes. Therefore, the first step consisted of determining if the ␤-domain mutant required energy expenditure for nuclear import before further investigating its role in VHL-mediated shuttling of BC/Cul-2. To do so, we developed a new assay to test for energy requirement for nuclear import in living cells based on fusing proteins to the energy-dependent human immunodeficiency virus REV NES. NES confers strong nuclear export properties to fusion proteins, leading to their cytoplasmic accumulation at steady state (Fig. 5; compare a with d, b with e, and c with f; see Ref. 36). GFP-GFP-NES rapidly accumulated in the nucleus upon inhibition of NES function at 4°C or with metabolic poisons, as expected, since this fusion protein is able to passively diffuse in and out of the nucleus (Fig. 5, g and j, and Ref. 46). In contrast, VHL-GFP-NES and ⌬E2-GFP-NES strictly remained in the cytoplasm at 4°C or in the presence of metabolic poisons (Fig.  5, h, i, k, and l), indicating that both fusion proteins are unable to passively diffuse in the nucleus. ⌬E2-GFP-NES and VHL-GFP-NES (Fig. 5, n and o) accumulated in the nucleus upon incubation with leptomycin B, a drug that specifically inhibits NES function (39,40) at 37°C, but not at 4°C, indicating that both fusion proteins contain energy-dependent nuclear import signals. These observations demonstrate that VHL ability to confer energy-dependent nuclear import properties to a reporter GFP is independent of assembly with HIF␣ and exon 2-encoded ␤-domain residues.
Exon 2-encoded ␤-domain mediates transcription-dependent trafficking of VHL and VBC/Cul-2, and the next step was to test if this domain was sensitive to conditions known to affect HIF␣ stabilization. GFP fluorescence analysis of living cells indicated that the steady state distribution of adVHL-GFP was unaffected by oxygen tension (Fig. 6, a and j). The addition of the RNA polymerase II inhibitor DRB caused nuclear accumulation of adVHL-GFP, regardless of oxygen concentration (Fig.  6, b and k). It has been recently suggested that proteasome inhibitors, which prevent proteasome-mediated degradation of ubiquitinated proteins, might also act as general inhibitors of nuclear export (47,48). Interestingly, a strong shift in the steady state distribution toward the nucleus of adVHL-GFP was observed upon incubation with the proteasome inhibitor CI, or lactacystin (data not shown) in normoxia and hypoxia (Fig. 6, c and l). ad⌬E3-GFP steady state distribution is more nuclear than adVHL-GFP and is unaffected by oxygen concentration (Fig. 6, g and p). The addition of DRB or CI also caused an important nuclear accumulation of ad⌬E3-GFP with few cells displaying exclusive nuclear signal (Fig. 6, h, i, q, and r). In contrast, the localization of the ␤-domain mutant ad⌬E2-GFP remained unchanged regardless of oxygen tension, proteasome inhibitors, or RNA polymerase II inhibitors (Fig. 6, d-f and m-o). One possible explanation for ad⌬E2-GFP insensitivity to DRB and CI is that this mutant is unable to bind to HIF␣. These observations led us to test if the effect of DRB and CI on shuttling of VHL are intrinsic to exon 2-encoded residues or if this activity is mediated by HIF␣. To test this, VHL shuttling was analyzed in mouse embryonic fibroblasts that do not express endogenous HIF␣ (Fig. 7). We noticed that adVHL-GFP steady state subcellular localization was unaffected by the absence of HIF-1␣ (Fig. 7, a and c). Likewise, the addition of DRB caused nuclear accumulation of adVHL-GFP in HIF-1␣ Ϫ/Ϫ as well as in HIF-1␣ ϩ/ϩ cells (Fig. 7, b and d). The localization of both mutants was unaffected by the absence or presence of HIF-1␣ (Fig. 7, e, g, i, and k). The ␣-domain mutant ad⌬E3-GFP accumulated in the nucleus upon incubation with DRB, whereas ad⌬E2-GFP was unaffected by this treatment in HIF-1␣ Ϫ/Ϫ and HIF-1␣ ϩ/ϩ cells. The effect of CI was essentially the same as DRB (data not shown) on the three fusion proteins (data not shown). The same data were obtained in hypoxia (data not shown). Put together, these results demonstrate that oxygen tension and HIF␣ have no affect on VHL nuclear/cytoplasmic shuttling properties. They also indicate that exon 2-encoded ␤-domain plays a role in nuclear/cytoplasmic trafficking of VHL, which is independent of its role in binding to HIF␣. DISCUSSION Inactivating mutations of the VHL tumor suppressor gene are distributed equally between the ␤and ␣-domains, suggesting that both domains play a key role in tumor suppression (29). Yet, the nature and localization of the mutations has a profound effect on the clinical manifestations in inherited VHL syndrome (31). Likewise, sporadic RCC tumors are much more likely to harbor mutations in exon 2, mutations that are rarely found in individuals afflicted with inherited VHL syndrome (5). The discrepancy in the distribution of inactivating mutations between sporadic and inherited RCC implies that exon 2-associated mutations might inactivate VHL function in different ways than exon 3-associated mutations. We show here that loss of exon 2 or exon 3 function essentially gives rise to the same cellular defects in RCC, which includes aberrant nuclear accumulation of HIF␣ in normoxia and inability to produce an extracellular fibronectin matrix. However, loss of exon 2 function appears to have a lesser effect on the overall activity of the VHL protein compared with loss of ␣-domain activity. The major defects of the ␤-domain mutant that we were able to identify were its inability to bind to HIF␣ and fibronectin and to mediate transcription-dependent shuttling of VHL. The binding results are similar to those recently reported by two other groups, which demonstrated that missense mutations in exon 1-encoded portion of the ␤-domain also abrogated VHL assembly to HIF␣ but not to BC/Cul-2 (12,18). A deletion of the ␣-domain caused a more complete loss of function, since this mutant failed to assemble with BC/Cul-2 as well as with substrate proteins and act as an E3 ubiquitin ligase. This is not the consequence of a truncation of the ␣-domain, since a missense mutation at residue 162 in the elongin C-binding box has recently been reported to cause similar defects (8,29). There is a discrepancy between data obtained in vitro and in culture inasmuch as truncations of exon 2-and exon 3-encoded sequences of VHL are still able to assemble with HIF␣ in vitro (12,18,33). Either ⌬E2-GFP and ⌬E3-GFP fold in a different way in vivo compared with in vitro, or these mutants have a yet uncharacterized defect that prevents their assembly with HIF␣ in cells. Interestingly, an alternative spliced mRNA of the VHL gene that lacks exon 2 sequences has been reported to be produced in several independent tissues and cell lines (1). A VHL protein without exon 2 sequences might change substrate specificity from HIF␣ to another unidentified protein while still acting as an E3 ubiquitin ligase. An endogenous protein product originating from a mRNA lacking exon 2 sequences still remains to be identified. Nevertheless, the data presented in this report are in good agreement with the proposed model predicted by the crystal structure of VHL that the ␤-domain of VHL is involved in substrate protein, as well as fibronectin, recognition (29). They also demonstrate that tumor-derived mutations inactivate VHL functions in different ways, which may lead to distinct cellular phenotypes.
The study of ad⌬E2-GFP has also revealed other interesting biochemical aspects of the function of exon 2-encoded sequences, one of which is that it is required for VHL-mediated NEDD8 conjugation on cullin-2. The functional relevancy of this post-translational modification is still unknown, but it has been suggested that it might play a role in protecting cullin-2 from self-ubiquitination (49). Data shown here are somewhat in disagreement with this model, since equal amounts of cul-lin-2 can be found bound to VHL and ad⌬E2-GFP, regardless of conjugation to NEDD8. NEDD8 conjugation is reported to be a nuclear event (44). ad⌬E2-GFP can be detected in the nuclear compartment at steady state, and the lack of NEDD8 conjugation activity cannot be simply explained by a defect in nuclear import of the VBC/Cul-2 complex. This argument is supported by a novel assay presented here, which enables the analysis of energy requirement for nuclear import of proteins in living cells. Energy expenditure for nuclear import is a hallmark of signal-mediated and -regulated nuclear/cytoplasmic trafficking processes (50 -52). The observation that ad⌬E2-GFP retains the ability to import in the nucleus in an energy-dependent manner suggest that other protein/protein interactions involved in nuclear import of the VBC/Cul-2 complex are not affected by loss of function of exon 2-encoded sequences. Likewise, we noticed HIF␣ signal exclusively in the nucleus of normoxic VHL Ϫ/Ϫ cells, indicating that HIF␣ is able to import even in the absence of hypoxic conditions and assembly with VHL. These data are somewhat surprising, since it is generally believed that HIF␣ contains a nuclear import signal that is activated only in hypoxia (45). One possible interpretation of these data is that the hypoxia-inducible nuclear import of HIF␣ is regulated by VHL, which might play a role in retaining HIF␣ in the cytoplasm in normoxia.
Results shown here suggest that transcription-dependent nuclear/cytoplasmic shuttling and steady state distribution of VHL are not affected by oxygen tension and do not require assembly with HIF␣. However, we did find that adVHL-GFP accumulated in the nucleus upon incubation with proteasome inhibitors, similar to the effect obtained with DRB treatment. Drugs that inhibit proteasome-mediated degradation of proteins have been hypothesized to also interfere with general nuclear export processes (47,48). Sensitivity to proteasome inhibitors is mediated by exon 2-encoded ␤-domain in a manner reminiscent of DRB. We have previously shown that VHL transcription-dependent shuttling domain acts dominantly on the VBC/Cul-2 complex and that DRB is a good inhibitor of VHL-mediated VBC-Cul-2 nuclear export in living cells and in vitro (46). It is conceivable that CI also blocks exon 2-mediated nuclear export of VHL, leading to nuclear accumulation of VBC/Cul-2. It is unlikely that the observed nuclear accumulation of adVHL-GFP is the consequence of HIF␣-mediated nuclear retention, since proteasome inhibitors and DRB have similar effects on VHL in HIF-null MEFs. The presence of ad⌬E2-GFP in the cytoplasm at steady state might be explained by a fraction of VHL that is not importable at a given time. Alternatively, the existence of other nuclear export signals within the complex might gain dominance upon loss of function of exon 2-encoded residues. Taken together, these results support the model that exon 2-encoded residues are involved in two independent functions: mediating nuclear export of the VBC/Cul-2 complex and binding to substrate proteins. We are still in the process of identifying relevant sequences involved in signal-mediated and Ran-dependent nuclear/cytoplasmic trafficking of the VBC/Cul-2 complex. Identification of these sequences will surely provide important clues in the elucidation of VHL-mediated tumor suppressor function.