Inactivation of VHL by tumorigenic mutations that disrupt dynamic coupling of the pVHL.hypoxia-inducible transcription factor-1alpha complex.

The von Hippel-Lindau (VHL) gene product, pVHL, targets the alpha subunit of the hypoxia-inducible transcription factor (HIF-alpha) for ubiquitin-dependent degradation. This tumor suppressor function is mediated by the alpha- and beta-domains responsible for assembling the pVHL E3 ubiquitin ligase complex and for recognizing the prolyl-hydroxylated HIF-alpha, respectively. The molecular basis for a large number of tumor-derived mutations can be attributed to alterations that directly compromise the ability of pVHL to assemble the E3 or to contact the substrate. Here we describe a new mechanism of oncogenic inactivation by VHL missense mutations that lie in the L1 and L7 linker regions distal to the HIF-alpha-binding pocket. Employing molecular dynamics simulations, we show that the tumorigenic L1 loop mutation of Ser(65) to Leu, deficient in promoting the degradation of HIF-alpha, disrupts the coordination of internal motions of the pVHL.HIF-1alpha complex. Furthermore, we demonstrate that in addition to S65L, five other tumor-derived VHL mutations located within the L1 loop are each defective in mediating proteolysis of HIF-2alpha. Moreover, dynamic organization of pVHL.HIF-1alpha recognition is focally centered on Gln(145) within the L7 loop, and its tumorigenic mutant Q145H abolishes almost all of the correlated dynamic motions. Intriguingly, Q145H, whereas defective in targeting cellular HIF-alpha for degradation, had an attenuated hydroxylation dependence in binding to HIF-1alpha in vitro. Taken together, our results suggest that specific association between pVHL and the hydroxylated HIF-alpha requires both the L1 and L7 loops to coordinate dynamic coupling among distant pVHL regions, whose mutational disruption inactivates VHL and is hence responsible for tumorigenesis.

The VHL gene product, pVHL, is the targeting subunit of a multiprotein E3 ubiquitin ligase complex, VEC, which is comprised of pVHL, the elonginB/C heterodimeric adaptor, the cullin family member CUL2, and RING finger protein ROC1/ Rbx1 (reviewed in Ref. 5), and is activated by NEDD8 conjugation (6 -8). The VEC is responsible for polyubiquitination of the ␣ regulatory subunit of the heterodimeric hypoxia-inducible transcription factor, HIF-␣, and its subsequent degradation through the proteasomal pathway in an oxygen-dependent manner (9 -13). HIF-1␣ and HIF-2␣ are the two most extensively studied isoforms and both are found to be regulated in a similar manner (14). pVHL recognizes HIF-␣ only when the conserved proline residues (Pro 564 and Pro 402 in HIF-1␣) within the oxygen-dependent degradation (ODD) domain of HIF-␣ are hydroxylated in the presence of oxygen and iron (12,13,(15)(16)(17). Conversely, under hypoxia HIF-␣ subunits remain unhydroxylated and escape degradation mediated by the VEC. Stabilized HIF-␣ subunits bind to constitutively stable HIF-␤/ ARNT subunits forming an active HIF transcription complex that binds to the hypoxia-responsive elements in the promoters of hypoxia-inducible genes, thus activating the cellular responses to hypoxia (12,18). HIF target genes are involved in a variety of physiological processes such as angiogenesis, iron homeostasis, anaerobic metabolism, and cell growth and survival, including vascular endothelial growth factor, erythropoietin (EPO-1), glycolytic enzymes, glucose transporter 1 (GLUT 1), insulin-like growth factor 2, and cyclin D1 (19,20).
pVHL is comprised of two domains: the ␣-domain, which is required for binding ElonginC to assemble the VEC complex, and the ␤-domain, which recognizes and targets hydroxylated HIF-␣ subunits for ubiquitination. In the ␣-domain, the H1 helix plays a major role in establishing the interface with ElonginC and residues in the H2 and H3 helices contribute to the interaction as well (21). In the ␤-domain, the HIF-1␣ hydroxylated Pro 564 (HyP 564 )-binding pocket is formed through a coordinated network of hydrogen bonds between residues located in S2, S3, S4, and S5 ␤-strands, as well as in the S4 -S5 linker. The specificity of the binding pocket for only HyP 564 is mediated through pVHL Ser 111 and His 115 , which form one hydrogen bond each with the HyP 564 4-hydroxyl group (22,23).
In addition to its direct role in targeting HIF-␣ for ubiquitindependent degradation, pVHL was found to co-fractionate with fibronectin and required for proper assembly of the extracellular fibronectin matrix (24). Moreover, pVHL was identified as a microtubule-associated protein that can protect microtubules from depolymerization, and therefore was speculated to play a role in the regulation of microtubule dynamics (25).
VHL mutations associated with tumor development have been identified throughout the open reading frame and map to residues on both ␣and ␤-domains. Previous studies have documented that while frequent tumorigenic mutations in the ␣-domain, such as C162F and R167Q, lead to disassembly of the VEC complex, ␤-domain mutations that map to the HIF-␣ binding pocket of pVHL, such as Y98H, disrupt interaction with HIF-␣ (11, 21-23, 26, 27). In this study, we employed both biochemical and computational studies to characterize oncogenic VHL mutations that occur in the L1 and L7 linker regions, both of which are distant to the HIF-␣ binding pocket. Our results provide evidence suggesting a significant role for the L1 and L7 loop residues in coordinating dynamic motions that are required for specific interaction between pVHL and prolyl-hydroxylated HIF-␣. Dysregulation of such dynamic coordination by VHL mutations in these loops results in stabilization of HIF-␣ and may hence be responsible for human cancers.

EXPERIMENTAL PROCEDURES
Cells-Human embryonic kidney cells (293T) and renal clear cell carcinoma cells 786-O (ATCC) were maintained in a humidified tissue culture incubator at 37°C with 5% CO 2 . The cells were cultured in Dulbecco's modified Eagle's medium, 10% heat-inactivated fetal bovine serum, and 100 units/ml penicillin/streptomycin.
Assembly of the VEC E3 Complex in 293T Cells-Cells (293T) were co-transfected with pcDNA3-HA-CUL2 (5 g) and pcDNA3.1-Flag-VHL (5 g) that encodes either the wild type or mutant pVHL, using Fu-GENE 6 transfectant reagent according to the manufacturer's instructions (Roche Diagnostics). Transfected cells were harvested 48 h later and washed with phosphate-buffered saline and pelleted. Cell pellets were resuspended in 0.5 ml/plate with buffer C (20 mM HEPES, pH 7.5, 5 mM KCl, 1.5 mM MgCl 2 , 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin and antipain) and Dounce homogenized on ice. Salt (NaCl) was added slowly while stirring to a final concentration of 0.2 M, and the lysates were centrifuged at 20,000 rpm for 30 min at 4°C. The supernatant was ultracentrifuged in a Beckman Ti45 rotor at 37,000 rpm for 1 h at 4°C, and the extracts were stored in aliquots at Ϫ80°C.
To immunopurify the VEC complex, extracts (1 mg of protein) were mixed with M2 beads (20 l) for 16 h at 4°C. The beads were washed three times with buffer D (25 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 0.1% Nonidet P-40, 1 mM EDTA) and phosphate-buffered saline. Bound proteins were released by boiling the beads in SDS loading buffer and separated by SDS-PAGE. Proteins were transferred onto nitrocellulose membrane and immunoblotted with antibodies as indicated.
Degradation of HIF-2␣-Cells (786-O) were transfected with pcDNA3-Flag-VHL (5 g) that encodes either the wild type or mutant pVHL using FuGENE 6 Transfectant reagent. Proteasomal inhibition was performed by treatment with MG132 (5 M) for 4 -6 h before harvesting. Cells were harvested 24 h after transfection in phosphatebuffered saline and pelleted. Cell pellets were resuspended in buffer E containing 25 mM Tris-HCl, pH 7.5, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin and antipain, and sonicated. Following centrifugation, soluble extracts (50 g of protein) were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membrane, which was probed with antibodies as indicated.
In Vitro Hydroxylation-dependent HIF-1␣ Binding-HIF-1␣ was hydroxylated based on a protocol by Bruick and McKnight (18). Briefly, the in vitro translated 35 S-HIF-1␣ from the rabbit reticulolysates ( 35 S-IVT-HIF-1␣, 10 l) was incubated with MBP-PHD1 (0.25 g) in a reaction mixture (50 l) containing 20 mM Tris-HCl, pH 7.5, 5 mM KCl, 1.5 mM MgCl 2 , 1 mM DTT, 2 mM ketoxoglutarate, 2 mM ascorbic acid, and 250 M FeSO 4 . The incubation was for 30 min at 30°C. To measure the interaction between the VEC E3 and hydroxylated HIF-1␣, the hydroxylation mixture was mixed with the E3 complex immobilized on M2 beads (20 l), prepared as described above, and maintained in buffer F (200 l) that contains 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, and 0.1% Nonidet P-40. Following incubation for 2 h at 4°C, the resulting beads were washed 3 times with buffer D and once with phosphate-buffered saline. Bound proteins were released by boiling the beads in SDS loading buffer, separated by SDS-PAGE, and visualized by autoradiography.
In Vitro Ubiquitination of HIF-1␣-The in vitro ubiquitination of HIF-1␣ was carried out using conditions similar to those for the IB␣ reaction (28,29). 35 S-HIF-1␣ bound to the VEC E3 on M2 beads, prepared as described above, was incubated in a reaction mixture containing 40 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 4 mM ATP, 0.6 mM DTT, ubiquitin (Ub, 500 pmol), E1 (2 pmol), and Ubc5c (30 pmol or otherwise specified), or Cdc34 (115 pmol or otherwise specified) at 30°C for 30 min or for other times as indicated. The reaction was terminated by the addition of Laemmli loading buffer and boiled for 3 min prior to 6% SDS-PAGE analysis followed by autoradiography.
Molecular Dynamics Simulations and Analysis of Coupling between HIF-1␣ and Distant pVHL Sites-To examine the effects of tumorderived mutations in pVHL that are distant to the HIF-1␣ binding site, and that do not disrupt the folding and assembly of the pVHL-Elong-inC-ElonginB complex, we calculated the canonical ensembles of the pVHL⅐HIF-1␣ complex expected under physiological conditions. Crystallographic structure of the HIF-1␣ peptide bound to pVHL-ElonginC-ElonginB (23) was used as the starting configuration. ElonginB, Elong-inC, as well as crystallographically disordered residues 54 -60 of pVHL and 556 -560 of HIF-1␣ were omitted. The resulting structure was solvated under periodic boundary conditions using a 80 ϫ 80 ϫ 80 Å 3 cubic box of equilibrated TIP3 water (30), and energy minimized using the CHARMM27 potential energy function in the presence of randomly placed 40 chloride and 42 sodium ions to yield electroneutrality and physiological ionic strength of about 0.15 M (31). Parameters for 4-hydroxyproline were obtained from Klein and Huang (32). The resulting system was heated using molecular dynamics with a linear gradient of 20 K/ps to 310 K, and equilibrated for 1.1 ns in the isothermal-isobaric (NPT) ensemble at 310 K and 1 atmosphere, using a 2-fs integration time step, SHAKE to constrain hydrogen atoms, and particle mesh Ewald (PME) summation to calculate electrostatic interactions, as implemented in NAMD2 (33). The calculated molecular dynamics trajectory was sampled every 0.1 ps, while omitting the first 100 ps during which thermalization was still occurring as judged from analysis of energy fluctuations.
To understand the dynamic and thermodynamic coupling between the HIF-1␣ binding site and pVHL regions distant to it, we calculated the covariance of backbone C␣ atom positions in the pVHL⅐HIF-1␣ complex, according to: cov( with the resultant normalized covariance matrix containing elements between Ϫ1 and ϩ1, indicating positions that are coupled negatively (hinge-like motion) and positively (breathing-type motion), respectively. Positions that are not dynamically or thermodynamically coupled yield correlation coefficients of near zero.
Such an analysis is expected to reveal the effects of mutation on the fs-ps atomic motion of the bound pVHL⅐HIF-1␣ complex. Thus, whereas it captures the entropic contributions to the bound state, it does not take into account effects on either the unbound or transition states, both of which may contribute to the recognition and binding of HIF-1␣ in vivo. In this sense, the dynamical basis of the mechanism of S65L and Q145H mutants presented below should be considered as a component of a likely more complex phenomenon.
It should be pointed out that because of the short simulation length (1 ns), we observed only dynamical perturbations but not structural rearrangements. This is consistent with the absence of structural rearrangements in the course of the simulations of the wild type pVHL (root mean structural deviation of backbone C␣ Ͻ1.7 Å), and of significant differences between the wild type and pVHL S65L mutant (root mean structural deviation of backbone C␣ Ͻ1.8 Å). In addition, as shown in Figs. 3, A and B, 6A, and 8B, the VHL mutations examined in this study are all capable of interacting with ElonginC to form the VHL E3 complex. These observations are in accord with the notion that there are no major structural alterations with VHL S65L or Q145H mutants that may influence the overall stability of the protein.
As shown by the titration experiment, Ubc5c efficiently ubiquitinated HIF-1␣ in a concentration-dependent manner (Fig.  1A, lanes [1][2][3][4][5]. To assess the kinetics of Ubc5c in this reaction, we examined the efficiency of HIF-1␣ ubiquitination as a function of time. Very high molecular weight substrate-Ub conjugates formed after 9 min, and the level of polyubiquitinated substrate increased over time with a concomitant decrease of the unmodified substrate (Fig. 1A, lanes 6 -10). We next tested Cdc34, an E2 of the SCF complexes, for its ability to support VEC-catalyzed ubiquitination of HIF-1␣ in the reconstituted in vitro system. In comparison to Ubc5c, Cdc34 was able to ubiquitinate HIF-1␣ but required a 4-fold increase in enzyme concentration ( Fig. 1B, lanes 1-4). Graphically, the kinetics of HIF-1␣ ubiquitination by Cdc34 and Ubc5c show a similar rate ( Fig. 1, A and B). Both E2s formed highly conjugated polyubiquitinated chains on the substrate, although Cdc34 showed more heterogeneity in the chains compared with those of Ubc5c (compare Fig. 1, A, lanes 9 and 10, to B, lanes 8 and 9).
The linkage specificity in polyubiquitination of HIF-1␣ was investigated by substitution of the wild type Ub with Ub K48R , which inhibits Lys 48 -linked Ub chain assembly. The ability of Ubc5c to form polyubiquitin chains on HIF-1␣ was undeterred with the mutant Ub, suggesting that this E2 can conjugate Ub moieties through lysines other than Lys 48 ( Fig. 2A, lanes 1-5). We performed a kinetic analysis using Cdc34 with the Ub K48R mutant. In contrast to Ubc5c, in the presence of Ub K48R , Cdc34 failed to support formation of polyubiquitinated HIF-1␣ species that migrated above 250 kDa, suggesting a specificity of this enzyme in selectively assembling Lys 48 -linked polyubiquitin chains onto HIF-1␣ (compare Figs. 1, B, lane 9, to 2, B, lane 5). Given the heterogeneous migration pattern of HIF-1␣ displayed on SDS-PAGE (Fig. 2B, lane 1), we were unable to precisely identify the monoubiquitinated HIF-1␣ species linked with Ub K48R . This difficulty precluded us from assessing the rate of the initial ubiquitination reaction that ligated Ub K48R to HIF-1␣. However, it was evident that Cdc34 synthesized HIF-1␣ conjugates containing a limited number of Ub K48R moieties, as reflected by the formation of discrete HIF-1␣ species that migrated slower than the unmodified form (in the range of 100 -200 kDa) (Fig. 2B, lane 5).
Tumor-derived Mutations at Ser 65 in VHL Abolish HIF-1␣ Binding-Whereas previous studies have documented the structural basis of a large number of tumor-derived mutations within the ␣and ␤-domains (21)(22)(23), it remains poorly understood how VHL is inactivated by mutations situated in locations relatively distant from the HIF-␣-binding pocket. One such site, Ser 65 , located in the L1 loop that is N-terminal to the binding pocket of HIF-1␣ (21)(22)(23), is frequently mutated in VHL (S65C, S65P, S65L, and S65W) and displays highly malignant forms in human cancer (Table I and referenced therein).
We investigated two of these tumor-derived mutations, S65W and S65L, as well as an alanine substitution of Ser 65 , S65A, and compared the E3 complex formation and the HIF-1␣ binding activity of these point mutants to that of the wild type pVHL. A glutathione bead pull-down assay was performed to assess the ability of the pVHL mutants to form a pVHL-Elong-inC-ElonginB subcomplex assembled in bacteria. As determined by Coomassie stain analysis, all three pVHL mutants retained a comparable level of ElonginC/ElonginB binding activity compared with the wild type pVHL (Fig. 3A). To examine formation of the VEC E3 complex, each FLAG-tagged pVHL mutant and HA-CUL2 constructs were co-transfected in 293T cells and then analyzed for E3 assembly by immunoprecipitation and Western blot. As shown in Fig. 3B, the pVHL point mutants all efficiently associated with VEC components, HA-CUL2 and ElonginC, compared with the wild type. To investigate the ability of the pVHL mutants to bind HIF-1␣, we set up an in vitro hydroxylation-dependent binding assay. The VEC E3, immunopurified on M2 beads was incubated with 35 S-IVT-HIF-1␣ either pretreated or untreated with PHD1, and then washed extensively. Bound proteins were eluted in SDS loading buffer and analyzed by Western blot and autoradiography. Our results revealed that: 1) treatment of HIF-1␣ by PHD1 stimulated pVHL binding by 3-fold (Fig. 3C, lanes 1 and 2); 2) S65A retained the wild type binding activity (lanes 3 and 4); and 3) both S65L and S65W mutants completely lost their ability to bind hydroxylated HIF-1␣ (lanes [5][6][7][8]. S65A also showed efficient ubiquitination activity, and a kinetic analysis of S65A demonstrated a similar rate of HIF-1␣ ubiquitination to that of the wild type pVHL (data not shown). In all, these studies demonstrate that despite their full capacity to form the VEC E3 complex, tumor-derived Ser 65 mutations abolished pVHL function in binding to HIF-1␣.
Clifford et al. (27) have previously investigated the molecular defect of the pVHL S65W mutation. Whereas their study agreed with ours in that this mutant is deficient in binding to HIF-1␣, Clifford et al. (27) showed that S65W was defective in forming a complex with ElonginC in RCC4 cells. However, as revealed by the above studies using both bacterial and 293T cell transfection systems, S65W was found to interact with ElonginC (Fig. 3, A and B). It remains to be determined whether this discrepancy is because of the use of different cell systems.
The pVHL⅐HIF-1␣ Complex Exhibits Coordination of Dynamic Motions among Distant pVHL Regions-The above observation that tumorigenic mutations at the L1 loop-located Ser 65 eliminated the interaction with HIF-1␣, suggests a role played by regions outside of the HIF-␣ binding pocket in supporting formation of a stable complex with this substrate. (Ser 65 is more than 20 Å away from the HIF-␣ binding pocket as well as ElonginC.) Importantly, there appear to be no statistically significant structural rearrangements of pVHL upon binding HIF-1␣, with the root mean square deviation in C␣ positions of 1.7 Å (21, 23). Consequently, we sought to FIG. 1. In vitro ubiquitination of HIF-1␣ catalyzed by both Ubc5c and Cdc34. Using protocols described under "Experimental Procedures," hydroxylated 35 S-HIF-1␣ and the VEC complex immobilized on M2 beads were prepared separately and then mixed to form the substrate-E3 complex followed by addition of ubiquitination agents to initiate the modification reaction catalyzed by Ubc5c (panel A) or Cdc34 (panel B). Aliquots of each reaction were separated by 6% SDS-PAGE followed by autoradiography. Unmodified HIF-1␣ (110 kDa) and high molecular mass HIF-1␣-Ub n reaction products (Ͼ250 kDa) were quantitated using a PhosphorImager (Amersham Biosciences). The results are presented in graphs.
determine whether oncogenic mutations of pVHL that lie distant to the HIF-1␣ binding pocket, such as S65L, can perturb substrate binding by disrupting the dynamics of molecular recognition between pVHL and the substrate. Molecular dynamics make important contributions to the binding and recognition of biomolecules, where dynamics play crucial roles in determining specificity and modulating entropy of binding (reviewed in Ref. 34). Here, we present nanosecond scale molecular dynamics simulations of pVHL⅐HIF-1␣ under physiological conditions, using the crystal structure of the pVHL⅐HIF-1␣ complex (23) as the starting conformation. The simulated system is a fully solvated model of pVHL⅐HIF-1␣ in the presence of over 16,000 water molecules and 80 counterions under periodic boundary conditions. The wild type and mutant systems were energy minimized, heated, and equilibrated to 37°C and 1 atmosphere pressure in the isothermalisobaric ensemble, as judged from the stabilization of fluctuations of kinetic energy (data not shown). Systems were then simulated for 1 ns, while being sampled every 0.1 ps. Such an approach allows for an examination of atomic level dynamics of the interaction between HIF-1␣ and pVHL and its mutants in the bound state in a computationally efficient manner (see "Experimental Procedures").
To examine the dynamic coupling between the HIF-1␣ binding site and pVHL regions distant to it, we calculated the covariance of C␣ positions, yielding positive and negative correlation coefficients that denote breathing-type (Fig. 4, red) and hinge-like (blue) motions, respectively. The covariance matrix of backbone C␣ atom positions of the wild type pVHL⅐HIF-1␣ is shown in Fig.  4A. In addition to the structural ␣and ␤-domains and motifs contained therein as revealed by the crystal structure (23), the pVHL⅐HIF-1␣ complex also contains a number of dynamic modules, interacting in specific ways. Fig. 4B shows the schematic mapping of breathing (red circles) and hinging (blue diamonds) motions, as derived from the analysis of C␣ covariance patterns for the wild type pVHL⅐HIF-1␣. In particular, the structural ␤-domain of pVHL contains a breathing-type dynamic module constituted by the N-terminal L1 loop containing Ser 65 , as well as S3, S4, and S6 ␤-strands that contain Ser 111 and largely constitute the HIF-1␣ binding pocket. The positive coupling of this  module is accomplished both using through-space (L1 loop to S3-S4 linker, S3-S4 linker to S6, L1 loop to S6) and throughbond (S3-S4 linker to S3 and to S4) interactions (Fig. 4B). Furthermore, this breathing-type dynamic module within the structural ␤-domain comprises a larger hinge-type dynamic module at the interface of the ␣and ␤-domains, whereby its motions are negatively coupled to the motions of H1 and H3 helices in the ␣-domain of pVHL (Fig. 4B). Such coordination of dynamic motions play a role in stability of the pVHL⅐HIF-␣ recognition, as discussed below.
Tumor-derived S65L Mutation Disrupts Dynamic Organization of pVHL⅐HIF-1␣ Recognition-We next examined two Ser 65 variants, S65A and S65L, using computational analysis of dynamics as described above. To allow for the comparison with the wild type protein, analysis of the effects of pVHL substitutions on the molecular recognition of HIF-1␣ were performed in the context of the pVHL⅐HIF-1␣ complex (see "Experimental Procedures"). Results from the schematic mapping of S65A show minor alterations in the overall dynamics of pVHL⅐HIF-1␣ recognition as compared with the wild type (compare Figs. 4, A and B, with 5, A and B), consistent with its ability to bind HIF-1␣ in vitro (Fig. 3C), and to direct HIF-2␣ degradation in cells (Fig. 6D). On the other hand, dynamical organization of pVHL S65L ⅐HIF-1␣ is perturbed as compared with the wild type (Fig. 5C), as evidenced by a marked disruption of the breathing-type module in the ␤-domain, and complete absence of the hinge-like module in the ␣-domain (Fig.  5D). Whereas S65L exhibits positive coupling between the motions of L1 loop and S3-S4 linker, these dynamics fail to be transmitted to the rest of the module in general and to the HIF-1␣ binding site in particular (Fig. 5D). The wild type behavior of S65A is likely because of the isosteric nature of the S65A substitution that preserves packing of the Ser 65 -containing L1 loop with the S4 and S6 ␤-strands and H3 ␣-helix. In contrast, dynamic disorganization of S65L is likely because of the disruption of these interactions (compare Fig. 5, B to D).
Of prominent note in the dynamical organization of pVHL⅐HIF-1␣ recognition is the presence of a focal hinge point between the breathing-type module in the ␤-domain and hingelike module in the ␣-domain, centered around Gln 145 (Fig. 4B). The dynamic organization of this focal hinge is prominently destroyed by S65L (Fig. 5, C and D), and suggests that Gln 145 may constitute a control node in the pVHL⅐HIF-1␣ dynamics, as investigated below. In summary, analysis of molecular dynamics simulations of pVHL⅐HIF-1␣ indicates the presence of a unique and non-trivial dynamic organization of the complex, the integrity of which correlates with the loss-of-function of pVHL mutants in vitro and in vivo, with the N-terminal L1 loop playing an important role.

Tumor-derived Mutations of Residues in the N-terminal L1 Loop of pVHL Lead to Stabilization of HIF-␣ Subunits-In
further support of a critical contribution by the L1 loop to VHL function, genetic data from familial VHL syndrome and sporadic RCC demonstrate that of the eight L1 loop amino acids, six residues (Ser 65 included) are mutated in human cancers (Table I and referenced therein). We therefore analyzed molecular defects of the five tumor-derived pVHL L1 loop mutants that include L63P, R64P, S68W, R69C, and E70K. As shown in Fig. 6A, all five mutants were able to nucleate the VEC complex with efficiency comparable with that of the wild type pVHL, as determined by immunoprecipitation/immunoblot analysis. The effects of these mutations on in vitro binding to HIF-1␣ are shown in Fig. 6B. Among the five mutants examined, R64P, S68W, and R69C consistently exhibited the significantly re- Coomassie staining analysis of the GST⅐pVHL⅐ElonginC⅐ElonginB complexes containing either the wild type (Wt) or mutant pVHL as indicated, which were assembled in bacteria, isolated as described under "Experimental Procedures," and separated by 12.5% SDS-PAGE. B, Ser 65 mutants form complexes with ElonginC and CUL2 in transfected 293T cells. Cells (293T) were co-transfected with HA-CUL2 and FLAG-VHL or FLAG-VHL mutants as indicated. Cell extracts were immunoprecipitated on M2 beads and the resulting immunoprecipitates were separated by 4 -20% SDS-PAGE and Western blot with antibodies as indicated. C, tumor-derived Ser 65 mutants are defective in hydroxylation-dependent HIF-1␣ binding. The VEC E3 was immunopurified on M2 beads as described in A. 35 S-HIF-1␣ was incubated either with or without PHD1 and then reacted with the VEC E3 on M2 beads ("Experimental Procedures"). Bound 35 S-HIF-1␣, resolved on 4 -20% SDS-PAGE and visualized by autoradiography, was quantitated using a PhosphorImager (Amersham Biosciences), which was expressed as arbitrary units displayed below each lane. The amount of FLAG-VHL in each precipitate was detected by Western blot. duced ability to bind to HIF-1␣ in comparison to the wild type pVHL (Fig. 6B, compare lanes 2, 6, 8, and 10). On average, R64P, S68W, and R69C had 50, 20, and 55%, respectively, of the wild type activity of pVHL to bind to HIF-1␣. Both L63P and E70K had a modest reduction (20%) in HIF-1␣ binding (Fig. 6B, lanes 4 and 12). Thus, among the six pVHL L1 loop residues that contain tumor-derived mutations, S65L/S65W had the most pronounced effect in abolishing HIF-1␣ binding (Fig. 3C), followed by S68W, R64P, and R69C with each exhibiting significantly reduced binding activity.
To investigate the ability of the pVHL N-terminal loop mutants to degrade HIF-␣ subunits in vivo, we utilized a VHL-null cell line, RCC/786-O, which expresses HIF-2␣ (10). We first confirmed the ability of the wild type pVHL to degrade HIF-2␣ in this system in a proteasome-dependent mechanism. FLAG-VHL was transfected into cells, and its effect on the endoge-nous HIF-2␣ level was determined by direct Western blot of cell extracts. HIF-2␣ was efficiently degraded by pVHL, and this degradation was prevented by addition of the proteasome inhibitor, MG132 (Fig. 6C, lanes 1-4). We also tested the S65A mutant in this system and obtained the same results as the wild type pVHL (Fig. 6C, lanes 5 and 6), in keeping with the ability of this variant to interact with HIF-1␣ (Fig. 3). We next transfected each of the pVHL N-terminal loop mutants into 786-O cells and analyzed their ability to degrade HIF-2␣. Our results showed that all of the tumor-derived L1 loop pVHL mutants, while expressed at levels comparable with the wild type, were defective in HIF-2␣ degradation (Fig. 6D).
These results demonstrate that tumor-derived mutations at six positions within the pVHL L1 loop all have profound defects in targeting HIF-2␣ for degradation, thus providing compelling biochemical evidence strongly suggesting a critical role for this region in the VHL function. Notably, mutants R64P, S65L/ S65W, S68W, and R69C were deficient in both targeting cellular HIF-2␣ for degradation and binding to HIF-1␣ in vitro, thus correlating loss of in vitro binding activity with defective degradation of HIF-␣ subunits. However, future investigation is required to determine how L63P and E70K fail to degrade HIF-2␣, which cannot be attributed to a slight reduction of HIF-1␣ binding activity by either mutant as revealed by the in vitro assay.
Q145H Mutant Disrupts the Dynamic Coupling in pVHL and Is Defective in HIF-␣ Degradation-Analysis of molecular dynamics simulations of the pVHL⅐HIF-1␣ peptide complex revealed that Gln 145 may constitute a control node in the dynamics of pVHL⅐HIF-1␣ recognition. Gln 145 lies in the L7 loop that is close to the interface of the ␣and ␤-domains of pVHL, but is more than 20 Å away from the HIF-␣ binding pocket as well as ElonginC. It is, therefore, unlikely that this residue is directly involved in binding either ElonginC or HIF-␣ subunits. However, genetic data indicates that this residue is mutated in the VHL syndrome (35). Using molecular dynamics simulations, we assessed the effect of the tumorigenic mutant of this residue, Q145H, on the dynamic organization of the complex. In agreement with the prediction of the model of pVHL⅐HIF-1␣ recognition dynamics above (Fig. 4), Q145H dynamics are grossly disorganized, with an almost complete absence of positively or negatively correlated motions (Fig. 7).
We characterized the effects of the tumor-derived mutant Q145H in pVHL function. When introduced into the 786-O cell by transfection, Q145H was found to be unable to promote the degradation of HIF-2␣ (Fig. 8A), despite having an ability to form the VEC complex (Fig. 8B). Moreover, as revealed by immunoprecipitation/immunoblot analysis, wild type pVHL interacted with HIF-2␣ in 786-O cells when the activity of the 26 S proteasome was inhibited by MG132 (Fig. 8C, compare lanes  2 and 3). In contrast, under the same condition pVHL Q145H failed to form a complex with HIF-2␣ (Fig. 8C, compare lanes 3  and 5). These findings thus suggest that the Q145H mutation rendered pVHL incapable of interacting with HIF-␣ in vivo, resulting in stabilization of this regulatory subunit.
Unexpectedly, when analyzed by the in vitro binding assay, while the wild type pVHL interacted with HIF-1␣ in a hydroxylation-dependent manner (Fig. 8D, compare lanes 1 and 2), Q145H efficiently bound to HIF-1␣ in the absence of PHD1 (Fig.  8D, compare lanes 1 and 3). Subsequent titration experiments showed that at all levels of PHD1 tested, Q145H was able to bind to HIF-1␣ more efficiently than the wild type (Fig. 8E). As shown by quantitative analysis (Fig. 8E, graph), on average Q145H bound 4 -6 times more HIF-1␣ than wild type pVHL in the absence of PHD1. In the presence of high levels of PHD1, wild type pVHL displayed a 4 -7-fold increase in binding to HIF-1␣, FIG. 4. Dynamic coupling between ␣and ␤-domains of pVHL and their linkage to the HIF-1␣ destruction sequence-binding site. A, normalized covariance matrix of backbone C␣ atom positions for the wild type (WT) pVHL⅐HIF-1␣. The red to blue color gradient (right) indicates regions that are positively (breathing-type, squares) and negatively coupled (hinge-like, circles), respectively. B, schematic mapping of positive (red circles) and negative (blue diamonds) coupling, as derived from the analysis of C␣ covariance patterns for the wild type pVHL⅐HIF-1␣. The N-terminal L1 loop of pVHL is dynamically coupled to the spatially juxtaposed S3-S4 hairpin, which is coupled through bond to the ␤-sheets comprised by S3, S4, and S6, containing the HIF-1␣ destruction sequence-binding site. This positive dynamic coupling is transmitted to the HIF-1␣ peptide, including the 4-hydroxyproline binding pocket comprised by pVHL Ser 111 and His 115 . The breathing-type motions of pVHL ␤-domain that contribute to HIF-1␣ binding are negatively coupled to ␣-domain dynamics, most prominently present between S3, S4, S6 ␤-sheet and H1 helix and between pVHL N-terminal L1 loop and H3 helix. This hinge-like motion of ␤and ␣-domains is centered around pVHL Gln 145 (triangle) in the loop located in the vicinity of the ␣/␤-domain interface.
whereas Q145H only increased substrate binding by less than 2-fold. Thus, it appears that the Q145H mutation diminished a requirement for prolyl-hydroxylation by pVHL to contact HIF-␣ in vitro.
In summary, the above results underscore a critical role for pVHL Gln 145 in coordinating the dynamic organization of the pVHL⅐HIF-␣ recognition and in targeting cellular HIF-␣ for degradation. Our studies further suggest that this residue may have a unique role for establishing the selective interaction of pVHL with prolyl-hydroxylated HIF-1␣. Loss of such coordinating activity by the Q145H mutation leads to stabilization of HIF-␣ and may hence be responsible for the malignant RCC phenotype.

DISCUSSION
Ubiquitination of HIF-1␣-The ubiquitination of HIF-1␣ commences with the recognition of ODD by the VEC E3 ligase complex, which recruits an E2 conjugation enzyme to catalyze the transfer of Ub from E2 to the E3-bound substrate. We demonstrate that both Ubc5c and Cdc34 can function as E2 for VEC-dependent ubiquitination of HIF-1␣ in vitro, albeit substantially more Cdc34 than Ubc5c is required for the reaction (Fig. 1). However, only Cdc34 selectively modifies HIF-1␣ with Lys 48 -linked polyubiquitin chains (Fig. 2), which are signaling polypeptides well known for their function to recognize the 26 S proteasome (reviewed in Ref. 36), where substrate degrada-tion takes place. Whether Cdc34 functions in cells to mediate the degradation of HIF-1␣ requires future investigation.
Distinct to HIF-1␣ ubiquitination is the presence of two hydroxylated proline-containing motifs (NODD with Pro 402 and CODD with Pro 564 ), placed at the N-and C-terminal end of ODD, each capable of interacting with the pVHL substratetargeting subunit of VEC (14). We have observed in our reconstitution assays that substitution of Pro 564 with glycine significantly reduced the binding of HIF-1␣ P564A to VEC (data not shown). However, ubiquitination still occurred with HIF-1␣ P564G bound to VEC, via hydroxylated Pro 402 (data not shown), suggesting that the Pro 402 -containing NODD is at least partially functional for mediating Ub ligation. No significant effect was observed when Pro 402 in HIF-1␣ was replaced by alanine (data not shown), suggesting that the Pro 564 -containing CODD alone was sufficient in supporting ubiquitination in vitro.
Little is known with respect to identity of HIF-1␣ lysine receptor residue(s) for ubiquitination. Substitution of Lys 389 , Lys 391 , or Lys 547 with arginine did not affect HIF-1␣ ubiquitination in vitro (data not shown). Given the presence of two pVHL-interacting motifs (NODD and CODD), it is likely that multiple lysine residues can be used for Ub ligation. However, it is not known whether NODD and CODD can simultaneously engage in contacts with the VEC to initiate ubiquitination. In this context, it is intriguing to note that Cdc34 formed heterogenous lower molecular weight HIF-1␣-Ub K48R species (Fig.  2B), which could be attributed to multiple mono-Ub K48R conjugates assembled on HIF-1␣.
Role of the pVHL N-terminal L1 Loop-Genetic studies from familial VHL syndrome and sporadic RCC underscore a critical role played by the pVHL N-terminal L1 loop, comprised of eight amino acid residues (21). Of the eight residues, six (Leu 63 , Arg 64 , Ser 65 , Ser 68 , Arg 69 , and Glu 70 ) are mutated in human cancers (Table I and (22) suggests that Ser 65 provides van der Waals contacts that contribute to the proper orientation of pVHL His 115 , a residue playing a central role in interacting with hydroxyproline (see below). In addition, the His 115 -containing, water-mediated hydrogen-bonding network involves the side chain of Asn 67 and the main chain amide nitrogen of Arg 69 (22).
The results presented in this study confirm and extend the critical contribution by the pVHL N-terminal L1 loop to the binding of HIF-1␣. We demonstrate that: 1) all tumor-derived mutations at these six amino acid positions, while preserving  Table I). We provide further insights by demonstrating that S65L, a mutation causing a malignant type of VHL syndrome, disrupted internal organization of molecular dynamics that are responsible for coordination of dynamic coupling within the ␤-domain, as well as between the ␣and ␤-domains, of pVHL, respectively (compare Figs. 4 and 5).
Analysis of molecular dynamics simulation revealed extensive dynamic organization of the pVHL molecule (Fig. 4), which is largely disrupted by the S65L mutation (Fig. 5, C and D). Our findings thus suggest that in addition to its localized van der Waals contribution to the positioning of His 115 , pVHL Ser 65 possesses a previously unrecognized dynamic role in VHL function. We propose that the pVHL N-terminal, Ser 65 -containing L1 loop region is responsible for dynamically coupling the HIF-1␣ binding site with the ␤-sheet core of ␤-domain, as well as coupling the recognition of HIF-1␣ by the ␤-domain to the dynamics of the ElonginC-interacting ␣-domain. Mutation of the L1 loop, such as S65L, results in drastic attenuation of these motions and consequent decoupling of the pVHL⅐HIF-␣ complex as shown by HIF-1␣ interaction and degradation studies (Figs. 3 and 6). Hence, by coordinating the dynamic motions of the pVHL⅐HIF-␣ complex, the pVHL N-terminal L1 loop may play an entropic role in stabilizing this complex required for targeted HIF-␣ degradation and its mutation-imposed dysregulation may directly be responsible for human cancers.
Role of pVHL Gln 145 -Our studies revealed intriguing properties of pVHL Gln 145 , a residue in the proximity of the ␣/␤-domain interface (21) that has been found mutated (to His) in human RCC (Table I and referenced therein). Based on molecular dynamics simulations this residue is crucial for stabilizing the hinge-like motion of the ␣and ␤-domains of pVHL (Fig. 4), and is required for coordinated motions throughout the protein and in the pVHL⅐HIF-1␣ interface, as evidenced by a total loss of dynamic coupling observed with the Q145H mutant (Fig. 7). Importantly, the Q145H mutant was unable to direct the degradation of HIF-2␣ in 786-O cells (Fig. 8A), most likely because of its inability to form a complex with this regulatory subunit in vivo (Fig. 8C). These findings thus provide another example of correlation between loss of dynamic organization and inactivation of the VHL function in targeting HIF-␣ for degradation.
However, unlike the S65L mutation whose disruption of dynamic coupling is associated with a loss of interaction with HIF-1␣ (Figs. 3C and 5, C and D), Q145H appears to alter the specificity of binding as evidenced by its association with HIF-1␣ in vitro with a markedly reduced dependence on hydroxylation (Fig. 8, D and E). Presumably, the formation of productive complex between pVHL and HIF-␣ in vivo strictly depends on stabilization of the hydroxyproline binding pocket. Contacts to HIF-␣ by pVHL Q145H are likely provided by 4-hydroxyl group-independent interactions (see below), which may not be sufficiently stable in cells.
Contribution of Dynamic Atomic Motions to the Interaction between pVHL and HIF-␣-In recent years, molecular dynamics simulations of biological molecules have provided many insights concerning the importance of internal motions in specific biomolecular function. This approach has been used to elucidate the mechanism of bacterial chaperonin GroEL (37), to elegantly reveal a role for dynamic coupling of protein modules in regulating the activity of Src family tyrosine kinases (38), to advance the understanding of the dynamic gating of acetylcholinesterase (39), and to explore the concerted motions in the DNA-human topoisomerase I complex (40). In this study, we provide evidence suggesting a significant role for dynamic atomic motions in contributing both to the specificity as well as affinity of pVHL⅐HIF-␣ binding.
Structural analysis of pVHL bound with the HIF-1␣ HyP 564peptide (22,23) has established that the specificity and affinity of the pVHL-HIF-␣ interaction is primarily provided by the hydroxyproline side chain, which is buried in a deep pocket formed by conserved pVHL residues Trp 88 , Tyr 98 , Ser 111 , His 115 , and Trp 117 . Within this pocket, the exposed Ser 111 hydroxyl group and His 115 imidazole amino group serve as hydrogen-bonding partners to the HyP 564 hydroxyl group. Another contributing factor for the specificity at the hydroxyprolinebinding site of pVHL is the restriction of the conformational flexibility of the HIF-1␣ backbone in the vicinity of HyP 564 , which is fixed through an extensive network of hydrogen bonds involving both backbone and side chain groups from pVHL (23). Coordination of atomic motions within pVHL and organization of its dynamic modules may be required to establish and/or maintain the recognition of hydroxylated HIF-␣ through dynamics-mediated entropic effects. Disruption of these coordinated motions and destabilization of pVHL dynamic modules by tumorigenic mutations such as S65L and Q145H leads to loss of efficient and/or specific binding of HIF-␣. It is hopeful that future development of simulation techniques will allow an assessment on whether coordinated atomic motions are correlated with proper positioning of pVHL residues involved in the complex network of interactions that support the formation of the hydroxyproline binding pocket and/or the structural rigidity of the HIF-␣ backbone in the vicinity of hydroxyproline.
Taken together, combined with both computational and biochemical studies, our data points to a previously unrecognized FIG. 7. pVHL Q145H exhibits gross disorganization of molecular motions that couple the HIF-1␣ binding site to distant regions of pVHL and the ElonginC-interacting ␣-domain. A, normalized covariance matrix of backbone C␣ atom positions for pVHL Q145H ⅐HIF-1␣. The red to blue color gradient indicates regions that are positively (breathing-type, squares) and negatively coupled (hingelike, circles), respectively. B, schematic mapping of positive (red circles) and negative (blue diamonds) coupling, as derived from the analysis of C␣ covariance patterns for pVHL Q145H ⅐HIF-1␣. role played by both the L1 and L7 linker regions to coordinate atomic motions that are likely critical for the ␤-domain to mediate its interaction with HIF-␣ in a hydroxylation-dependent manner. We have therefore suggested a new mechanism explaining how tumor-derived mutations within these regions, which are distant to the HIF-␣ binding pocket, profoundly perturb the function of pVHL in targeting HIF-␣ for degradation.