VHL Mutations Linked to Type 2C von Hippel-Lindau Disease Cause Extensive Structural Perturbations in pVHL^*,S⃞

EXPERIMENTAL PROCEDURES

Plasmids—pVB and pBB75 (14), pST39-HisTrxN-VHL-ElonginB-ElonginC (39), and pcDNA3-HA-VHL (23) were gifts of N. Pavletich (Memorial Sloan-Kettering Cancer Center, New York, NY), S. Tan (Penn State University, University Park, PA), and W. Krek (Swiss Federal Institute of Technology, Zurich, Switzerland), respectively. Human VHL (codons 54–213) was subcloned into pCMV-2C (Stratagene) and pBabe-Puro (40) (gift of W. Krek) using standard techniques. Human ElonginB (full length) and ElonginC (codons 17–112) were PCR-amplified using pST39-HisTrxN-VHL-ElonginB-ElonginC as template and cloned into a pET24a (Invitrogen) derivative for the expression of fusion proteins with the lipoyl domain of Bacillus stearothermophilus dihydrolipoamide acetyltransferase. VHL alleles coding for pVHL mutant proteins carrying the amino acid substitutions P81S, L188V, P81S/L188V, V84L, and C162F, respectively, were generated by site-directed mutagenesis using a QuikChange kit (Stratagene). The correct sequences of all inserts were confirmed by sequencing.

Cell Lines—HEK293T cell lines (ATCC) and the packaging cell line LinX-A (41) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 units/ml of penicillin, and 100 μg/ml streptomycin. RCC4 VHL^-/- renal cell carcinoma cell lines were a gift of P. Maxwell (Imperial College, London, UK). RCC4-derived cell pools expressing wild type or mutant VHL were generated by infection with retroviruses obtained by transfecting LinX-A cells with pBabe-Puro-VHL using FuGENE (Roche Applied Science). Stable cell pools were selected and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, 100 μg/ml streptomycin, and 2 μg/ml puromycin.

Protein Purification—Expression and purification of recombinant VCB complex were performed exactly as described (15). For isotopically enriched protein preparations, VCB or ElonginBC complexes were co-expressed in minimal medium containing ¹⁵NH₄Cl and purified as will be published elsewhere.4

Biochemical Analyses—In vitro transcription and translation using pcDNA3-HA-VHL as template and HIF-1α peptide binding measurements were performed exactly as described (15). Partial tryptic proteolysis was essentially performed as described (15), using 800 ng of VCB complex and 8 ng of trypsin. For the analysis of VCB aggregation state by size exclusion chromatography, 30 μg of VCB complex were incubated overnight at 4, 30, or 37 °C and subsequently loaded at 4 °C onto a Superdex S75 HR10/30 column (GE Healthcare) equilibrated in 40 mm Tris/HCl, pH 7.9, 150 mm KCl, 5 mm dithiothreitol. Eluted fractions were concentrated by trichloroacetic acid precipitation and analyzed by SDS-PAGE.

NMR Spectroscopy—¹H,¹⁵N HSQC spectra of ¹⁵N-labeled wild type VCB complex, pVHL^L188V-containing VCB complex, and ElonginBC complex (all 0.1 mm in 40 mm Hepes, pH 7.2, 150 mm NaCl, 5 mm dithiothreitol, 5% D₂O) were recorded at 25 °C on a Bruker AVANCE 700 MHz spectrometer equipped with a cryo probe head. Full assignments of the wild type VCB spectrum will be published elsewhere.4

Immunoprecipitation and Immunoblots—Expression of FLAG-pVHL in 293T cells and immunoprecipitation were performed as described (42). Bound proteins were analyzed by immunoblots using antibodies directed against the FLAG epitope (M2; Sigma), Cullin-2 (Zymed Laboratories Inc.), Rbx1 (Ab-1; NeoMarkers), and ElonginC (R-20; Santa Cruz). Protein levels in RCC4-derived cell pools were analyzed by immunoblots using antibodies directed against pVHL (VHL-CT; gift of W. Krek), HIF-1α (clone 54; BD Transduction), and tubulin (Sigma).

Quantitative PCR—To quantify mRNA levels in RCC4-derived cell pools, total RNA from 10⁶ cells was purified using a High Pure RNA isolation kit (Roche Applied Science). cDNA was synthesized from 1.5 μg of total RNA using a Transcriptor First Strand cDNA synthesis kit (Roche Applied Science) with a combination of oligo(dT)₁₈ primer (final concentration, 2.5 μm) and primer pBabe3 (5′-CTG ACA CAC ATT CCA CAG GG-3′; final concentration, 2 μm). Real time PCR was performed on a LightCycler 480 (Roche Applied Science) using a SYBR Green I Master kit (Roche Applied Science) according to the manufacturer's instructions with 1 μl of cDNA (diluted 1:10) and primer pairs (final concentration, 0.5 μm) specific for VHL (forward, 5′-TTG TCC GGA GCC TAG TCA AG-3′; reverse, 5′-CTG TGC TGG CGA ATT CAA TC-3′); GLUT1 (forward, 5′-GAG CAG CTA CCC TGG ATG TC-3′; reverse, 5′-CAG AAC CAG GAG CAC AGT GA-3′); HPRT (forward, 5′-TGG ACA GGA CTG AAC GTC TTG-3′; reverse, 5′-CAG TCA TAG GAA TGG ATC TAT CAC-3′); and PBG-D (forward, 5′-CCC TGG AGA AGA ATG AAG TGG-3′; reverse, 5′-TTC TCT GGC AGG GTT TCT AGG-3′). Crossing point raw data were quantified using as standard curve a serial dilution of cDNA from RCC4/pBabe-Puro-pVHL wild type, and subsequently normalized to the mean of the HPRT and PBG-D data. Negative controls performed by omitting reverse transcriptase from the cDNA synthesis reaction did not result in any PCR product (data not shown).

RESULTS

Type 2C-associated pVHL Mutant Proteins Are Proficient in Substrate Binding—Type 2C-associated mutations do not affect pVHL residues in close proximity to the HIF-1α-binding site (14, 43, 44) (supplemental Fig. S1). Consistently, no significant defects in HIF-1α binding were detected in previous, qualitative analyses of type 2C-associated mutant pVHL proteins (25, 26). To test this more rigorously, we used recombinant wild type and mutant VCB complexes to quantify the binding of a fluorescent, HIF-1α-derived substrate peptide (15). In kinetic measurements of substrate peptide release, all of the mutant VCB complexes were found to possess dissociation rate constants very similar to the wild type complex (Table 1). In equilibrium titration experiments, the pVHL^P81S-containing VCB complex and the type 2C-associated pVHL^L188V- and pVHL^P81S/L188V-containing VCB complexes bound substrate peptide with slightly (∼3-fold) lower affinity than the wild type complex (Table 1). Thus, type 2C-associated mutations affect substrate binding of pVHL only moderately. In particular, the affinities for the HIF-1α-derived peptide are still in the low nanomolar range and thus 1–2 orders of magnitude higher than for prototypical type 2A-band type 2B-associated mutant pVHL proteins (15).

Type 2C-associated pVHL Mutant Proteins Assemble into Unstable VCB Complexes—Next, the stability of the wild type and mutant VCB complexes was analyzed. To that end, VHL was transcribed and translated in vitro at 30 °C in reticulocyte lysate providing endogenous ElonginB and ElonginC, and the formation of stable VCB complex was monitored by nondenaturing polyacrylamide gel electrophoresis (15, 45, 46). When the electrophoresis was performed at 4 °C, wild type and type 2C-associated pVHL mutant proteins migrated in a defined VCB complex, unlike the assembly-deficient pVHL^C162F mutant protein that served as a negative control (Fig. 1). At 30 °C, in contrast, pVHL^P81S/L188V and pVHL^V84L failed to migrate in a stable VCB complex, and the level of stably assembled pVHL^L188V was significantly reduced. At 37 °C, the differences between wild type and mutant pVHL proteins were even more pronounced, because pVHL^L188V completely failed to migrate in a stable complex, and the amount of pVHL^P81S incorporated into VCB was strongly reduced. These data show that type 2C-associated mutant pVHL proteins are impaired in their ability to form or maintain stable VCB complexes at physiologically relevant temperature. Moreover, the destabilization caused by the L188V and P81S amino acid substitutions was additive in this assay, because the pVHL^P81S/L188V double mutant protein was significantly stronger affected than pVHL^L188V. It is important to note, however, that all type 2C-associated mutant proteins were fully proficient in forming a stable VCB complex at 4 °C after they had been in vitro translated at 30 °C, indicating that they, unlike the type 1-associated pVHL^C162F mutant protein, do not possess a general assembly defect. This conclusion is also supported by our ability to purify recombinant mutant VCB complexes to homogeneity over several chromatographic steps at 4 °C after co-expression of the subunits in Escherichia coli at room temperature.

To further analyze the temperature-dependent stability of the wild type and mutant VCB complexes, we performed size exclusion chromatography of recombinant VCB at 4 °C after overnight incubation at 4, 30, or 37 °C (Fig. 2). After incubation at 4 °C, wild type and mutant complexes showed indistinguishable elution profiles consistent with their molecular masses of 43 kDa, confirming that type 2C-associated pVHL mutant proteins do not cause gross structural perturbations or destabilization of the VCB complex at this temperature. After incubation at 30 °C, however, significant amounts of pVHL^L188V- and pVHL^P81S/L188V-containing VCB complex eluted in the void volume of the column, indicating a higher oligomerization/aggregation state and/or altered shape at this temperature. Preincubation at 37 °C aggravated this effect and resulted in the accumulation of ∼50% of pVHL^L188V-containing and more than 50% of pVHL^P81S/L188V-containing VCB complex in the void volume, as judged by the relative peak areas of the chromatograms. At this temperature, some pVHL^P81S-containing VCB complex also eluted in the void volume, whereas wild type VCB complex was exclusively detected as a 43-kDa species. Together, these results indicate that type 2C-associated mutations cause temperature-dependent aggregation or structural alterations of the VCB complex, consistent with the results obtained with the reticulocyte system.

Type 2C-associated pVHL Mutant Proteins Are Partially Unfolded—To specifically analyze the conformation of pVHL at 37 °C, the susceptibility of wild type and type 2C-associated mutant recombinant VCB complexes to tryptic degradation was tested (Fig. 3). While the ElonginB and ElonginC subunits were protease resistant during the time course of the experiment, wild type pVHL was gradually degraded and barely detectable after 2 h. The pVHL^P81S mutant protein was degraded only slightly faster than the wild type. In contrast, pVHL^L188V, pVHL^P81S/L188V, and pVHL^V84L were extremely susceptible to tryptic degradation and undetectable after 10–20 min. Of note, the protease concentration in these experiments was 10-fold lower compared with our previous analysis of type 2A/B-associated mutant pVHL proteins (15) in order to detect type 2C-associated pVHL mutant proteins at all. These results suggest that type 2C-associated mutations induce significant conformational changes in pVHL, either within the VCB complex or upon dissociation from ElonginB and ElonginC.

To analyze the folding state of wild type pVHL and the type 2C-associated pVHL^L188V mutant protein within the VCB complex in more detail in a noninvasive manner, we employed two-dimensional, heteronuclear NMR spectroscopy. The ¹H,¹⁵N HSQC spectrum of the wild type complex at 25 °C shows the expected number of predominantly well dispersed cross-peaks, with some spectral overlap in the central part and the side chain region (Fig. 4a, blue spectrum). The spectrum of the heterodimeric ElonginB-ElonginC complex overlaps extensively with the wild type VCB spectrum (Fig. 4a, yellow spectrum), illustrating that the overall structure of ElonginB and C is retained in the VCB complex. While the ¹H,¹⁵N HSQC spectrum of the pVHL^P81S-containing VCB complex is almost indistinguishable from the wild type complex (data not shown), the spectrum of the pVHL^L188V-containing complex is dramatically different (Fig. 4b, red spectrum). The most obvious difference is the almost complete lack of native pVHL resonances, highlighting the loss of structure of the pVHL subunit of the mutant VCB complex. Although pVHL^L188V shows extensive structural disorder, as evident from the intense, overlapping resonances in the central region of the spectrum, most ElonginB and C signals are retained, demonstrating the individuality of the Elongins within the VCB complex. Of note, however, some weak residual native pVHL resonances could be identified, including the β domain residues Gly¹²³, Asn¹³¹, Gln¹³², and Gly¹⁴⁴ (Fig. 4, c and d). These signals indicate conformational exchange between the folded and unfolded states of pVHL^L188V, with an emphasis on the unfolded state. In addition, they show that at least a subpopulation of pVHL^L188V is bound to ElonginB and C at equilibrium. Together, the NMR spectra provide direct proof that the fold of type 2C-associated pVHL^L188V protein is globally destabilized at room temperature even in the presence of ElonginB and C.

Type 2C-associated pVHL Mutant Proteins Assemble into Unstable CBC^VHL Complexes in Vivo—The pronounced alterations in pVHL conformation and VCB stability observed at physiologically relevant temperature in vitro suggested that type 2C-associated VHL mutations should also destabilize the full CBC^VHL E3 ubiquitin ligase complex in vivo. To test this prediction, we transfected human 293T cells to transiently express FLAG epitope-tagged wild type or mutant pVHL, immunoprecipitated FLAG-pVHL from cell lysates, and compared the amounts of co-precipitated, endogenous CBC^VHL subunits (Fig. 5). Similar amounts of Cullin-2, Rbx1, and ElonginC were co-precipitated with wild type pVHL and pVHL^P81S, but not with type 1-associated, assembly-deficient pVHL^C162F. Importantly, reduced amounts of ElonginC were co-precipitated with all type 2C-associated pVHL mutant proteins, indicating that type 2C-associated mutations indeed affect the stability of the CBC^VHL E3 ubiquitin ligase complex in vivo. Interestingly, pVHL^L188V and pVHL^P81S/L188V, but not pVHL^V84L, also bound significantly less Cullin-2 and Rbx1, a distal CBC^VHL subunit binding via Cullin-2. These results suggest that the V84L mutation primarily affects ElonginC binding, whereas the L188V mutation affects both ElonginC and Cullin-2 binding.

Type 2C-associated pVHL Mutant Proteins Are Mildly Defective in HIF-1α Regulation—To analyze potential physiological consequences of the observed CBC^VHL destabilization, we stably reintroduced wild type and mutant VHL into the renal cell carcinoma-derived RCC4 VHL^-/- cell line. A low copy number, episomal expression vector was used to avoid masking of potential subtle defects by overexpression. Surprisingly, the steady-state levels of all three type 2C-associated pVHL mutant proteins, but not of pVHL^P81S, were strongly reduced compared with wild type pVHL, and the level of pVHL^C162F was below detection (Fig. 6a). These differences were reproducibly observed in independently selected cell pools that had been infected with independently generated retroviral inocula. The differences in pVHL levels are much more pronounced than those observed in transiently transfected 293T cells (cf. input in Fig. 5), presumably because of the weaker, more physiological expression from the episomal constructs. Importantly, the reduced pVHL mutant protein levels in the RCC4-derived cell pools are not the consequence of reduced VHL expression, because VHL mRNA levels were comparable between wild type and mutant cell pools and did not correlate with pVHL protein levels (Fig. 6b). Rather, the reduced pVHL levels suggest that type 2C-associated pVHL mutant proteins are short-lived in vivo, presumably because they associate less stably with the other subunits of the CBC^VHL complex.

Next, the impact of reduced pVHL protein levels on the regulation of the key cellular CBC^VHL target, HIF-1α, was analyzed. Under normoxic conditions, HIF-1α was hardly detectable in cell pools expressing wild type pVHL or pVHL^P81S, whereas it strongly accumulated in pools carrying empty vector or expressing pVHL^C162F (Fig. 6a). Of note, cell pools expressing type 2C-associated pVHL mutant proteins showed slightly elevated HIF-1α levels under normoxic conditions, indicating that the reduced pVHL levels in these cell pools lead to an incomplete down-regulation of HIF-1α. The residual HIF-1α levels observed in cell pools expressing type 2C-associated pVHL mutant proteins were, however, insufficient to drive expression of the HIF-1α target gene, GLUT1 (Fig. 6b). Thus, at least in the setting of a renal cell carcinoma-derived cell line, the mild defects in HIF-1α down-regulation exhibited by type 2C-associated pVHL mutant proteins do not result in a measurable up-regulation of HIF-1α target genes.

DISCUSSION

This is the first study investigating defects of type 2C-associated pVHL mutant proteins at the molecular level. We show that pVHL^L188V, pVHL^P81S/L188V, and pVHL^V84L are structurally perturbed and impaired in forming stable VCB and CBC^VHL complexes.

Using biochemical and structural approaches, we were able to directly analyze pVHL folding and stability in the context of the purified, recombinant VCB complex. We found a significant destabilization at physiologically relevant temperature of VCB containing type 2C-associated pVHL mutant proteins consistent with the extensive structural alterations observed by NMR spectroscopy. Our results are compatible with two slightly different interpretations regarding the integrity of the VCB complex. In one scenario, the equilibrium between free pVHL and the VCB and CBC^VHL complexes, respectively, is shifted toward free pVHL by type 2C-associated mutations. Because free pVHL is unable to adopt its native conformation (14, 47), it is highly susceptible to degradation both by trypsin in vitro and by cellular proteases in vivo. Indeed, pVHL has been shown to be stabilized in vivo by incorporation into the CBC^VHL complex (48, 49), providing a straightforward explanation for the reduced levels of type 2C-associated pVHL mutant proteins in vivo (Fig. 6a). Alternatively, the mutant pVHL proteins may remain associated with VCB and CBC^VHL while having lost most of their native three-dimensional structure. Again, this would lead to high susceptibility toward proteolysis. In addition, this interpretation could explain the coelution of ElonginB and C with pVHL in high molecular weight size exclusion chromatography fractions upon incubation at elevated temperature (Fig. 2), and the presence of residual, native cross-peaks in the NMR spectrum (Fig. 4, c and d).

The destabilizing effect of type 2C-associated VHL mutations can at least partially be rationalized based on the known three-dimensional structure of the VCB complex (14, 43, 44) (see supplemental Fig. S1). Residue Leu¹⁸⁸ is located in one of the three helices of the pVHLα domain, which are complemented by a helix from ElonginC to form an intermolecular four-helix cluster (14). Consequently, the L188V mutation is expected to affect the packing of these helices and hence ElonginC binding. In addition to local effects on the α domain, our NMR data also reveal gross structural alterations of the β domain (Fig. 4b), most likely caused by disorder in the carboxyl-terminal region of pVHL and/or the α/β domain interface. According to a recently presented structural model of VCB in complex with Cullin-2 (50), a perturbation of the four-helix cluster would also be expected to affect critical contacts with Cullin-2. Indeed, we found that the interaction between pVHL^L188V and Cullin-2 is significantly weakened in vivo (Fig. 5), indicating that the proper orientation of the helices comprising the pVHL α domain is crucial for the ordered assembly of the entire CBC^VHL complex.

Residue Pro⁸¹ is located at the interface between the pVHL β domain and ElonginC. Although the P81S substitution does not cause significant structural perturbations on its own, it might affect the orientation of the neighboring residue Arg⁸², whose side chain forms the center of a hydrogen bond network stabilizing both the α/β domain interface of pVHL and the pVHL/ElonginC interface (14). Suboptimal hydrogen bonding might further destabilize the interaction with ElonginC upon mutation of Leu¹⁸⁸, consistent with the additive defects we observed in some, but not all, of our assays for the pVHL^P81S/L188V double mutant protein. We speculate that the P81S substitution might also aggravate molecular defects caused by concomitant VHL mutations in trichloroethylene-induced renal cell carcinomas (37, 38). In contrast, the reported link of the P81S substitution to type 1 VHL disease (36) remains enigmatic, given the lack of significant functional defects of the pVHL^P81S mutant protein in vivo.

The side chain of residue Val⁸⁴ contributes to the hydrophobic core of the pVHL β domain. Consequently, the V84L substitution destabilizes pVHL, presumably by weakening hydrophobic interactions within the β domain. In contrast to many type 2B-associated pVHL proteins mutated in neighboring residues, however, the binding of pVHL^V84L to HIF-1α is not severely affected (25, 26) (Table 1 and supplemental Fig. S2). Rather, this mutation affects binding to ElonginC, either by altering the interface of the β domain with loop 5 of ElonginC (14) or more indirectly by altering the intermolecular four-helix cluster via the Arg⁸² hydrogen bond network. Interestingly, the binding of pVHL^V84L to Cullin-2 is not significantly compromised (Fig. 5), suggesting that it is separable from ElonginC binding despite the requirement for an intact α domain (see above).

Previous studies did not report defects of type 2C-associated pVHL mutant proteins in the down-regulation of HIF-1/2α in vivo (25, 26). It is possible that the low VHL expression levels in our stable cell lines facilitated the detection of relatively subtle effects on HIF-1α regulation that otherwise might be masked by VHL overexpression. Nevertheless, a careful examination of published data reveals that defects of type 2C-associated pVHL mutant proteins in HIF-1/2α regulation were detectable in several studies. Examples include slightly elevated HIF-1α and/or HIF-2α levels in stable 786–0/pVHL^V84L (25), RCC10/pVHL^V84L (51), RCC4/pVHL^V84L, RCC4/pVHL^L188V (52), and 786–0/pVHL^L188V (27, 53) cell lines. Together, these pieces of evidence support our data demonstrating small, but detectable defects in HIF-1α regulation for pVHL^L188V, pVHL^P81S/L188V, and pVHL^V84L (Fig. 6). The risk of renal cell carcinomas in type 2 von Hippel-Lindau disease has been shown to strongly correlate with the ability to bind, ubiquitylate, and down-regulate HIF-1/2α (15–17). The residual down-regulation of HIF-1/2α by type 2C-associated pVHL mutant proteins is more efficient than that by type 2A-associated pVHL mutant proteins (26), consistent with the absence of renal cell carcinomas in type 2C von Hippel-Lindau disease. Consequently, no significant up-regulation of classical HIF-1/2α target genes is observed in renal cell carcinoma-derived cell lines expressing type 2C-associated pVHL mutant proteins (Fig. 6b; Refs. 25, 27, 52, and 53). This does not, however, exclude the possibility that mild defects in HIF-1/2α regulation lead to an up-regulation of target genes requiring relatively low HIF levels for expression in other cell types, thereby potentially contributing to the pathogenesis of pheochromocytomas in type 2 von Hippel-Lindau disease.

Irrespective of the role of HIF-1/2α in pheochromocytoma, the reason for the lack of pheochromocytoma in type 1 von Hippel-Lindau disease remains enigmatic. The most obvious difference between type 1- and all type 2-associated pVHL mutant proteins studied so far is the ability of the latter to assemble at least to some extent into CBC^VHL complexes (this study and Refs. 15 and 26), which protects them from rapid intracellular degradation (48, 49). Thus, development of pheochromocytoma appears to require some minimal cellular pVHL level and to be incompatible with total loss of pVHL function. In type 2B von Hippel-Lindau disease, pheochromocytomas are found in the context of complete loss of HIF-1/2α down-regulation, suggesting that some HIF-independent function of pVHL is critical for the pathogenesis of pheochromocytoma. Recently, the down-regulation of atypical PKC isoforms (aPKCs) and JunB, which is required for apoptosis of neuronal progenitor cells upon nerve growth factor withdrawal, was reported to be pVHL-dependent but partially HIF-independent (27). In a model taking into account the dual role of pVHL in HIF-1/2α and aPKC/JunB regulation, it was proposed that the balance of defects in the HIF-1/2α versus aPKC/JunB pathways determines whether VHL mutations are associated with pheochromocytoma (2). However, this model appears to be inconsistent with the finding that HIF-2α, its target gene product GLUT1, and JunB accumulate to very similar levels in VHL^-/- renal cell carcinoma cell lines stably expressing type 1 and type 2B mutant proteins, respectively (27). Another HIF-independent function of pVHL that has been implicated in pheochromocytoma pathogenesis is its involvement in extracellular matrix assembly by virtue of interactions with fibronectin (25) and collagen IV (54, 55). Again, however, type 1- and type 2-associated pVHL mutant proteins exhibit similar defects (25), making it unlikely that extracellular matrix assembly is critical with respect to pheochromocytoma.

Although further efforts are needed to clarify the as yet elusive HIF-independent pVHL function required for the development of pheochromocytoma, our study suggests the possibility that partially defective HIF-1/2α regulation contributes to the pathogenesis of pheochromocytoma in von Hippel-Lindau disease.

Supplementary Material