Identification of Elongin C and Skp1 Sequences That Determine Cullin Selection*

The multiprotein von Hippel-Lindau (VHL) tumor suppressor and Skp1-Cul1-F-box protein (SCF) complexes belong to families of structurally related E3 ubiquitin ligases. In the VHL ubiquitin ligase, the VHL protein serves as the substrate recognition subunit, which is linked by the adaptor protein Elongin C to a heterodimeric Cul2/Rbx1 module that activates ubiquitylation of target proteins by the E2 ubiquitin-conjugating enzyme Ubc5. In SCF ubiquitin ligases, F-box proteins serve as substrate recognition subunits, which are linked by the Elongin C-like adaptor protein Skp1 to a Cul1/Rbx1 module that activates ubiquitylation of target proteins, in most cases by the E2 Cdc34. In this report, we investigate the functions of the Elongin C and Skp1 proteins in reconstitution of VHL and SCF ubiquitin ligases. We identify Elongin C and Skp1 structural elements responsible for selective interaction with their cognate Cullin/Rbx1 modules. In addition, using altered specificity Elongin C and F-box protein mutants, we investigate models for the mechanism underlying E2 selection by VHL and SCF ubiquitin ligases. Our findings provide evidence that E2 selection by VHL and SCF ubiquitin ligases is determined not solely by the Cullin/Rbx1 module, the target protein, or the integrity of the substrate recognition subunit but by yet to be elucidated features of these macromolecular complexes.

acid degenerate motif of sequence (A,P,S,T)LXXXCXXX(A,-I,L,V) (5)(6)(7)(8)(9). In the context of the VHL ubiquitin ligase, the VHL protein serves as the substrate recognition subunit and is linked by the adaptor protein Elongin C to a heterodimeric Cul2/Rbx1 module that functions as a potent activator of ubiquitylation of target proteins by an E2 ubiquitin-conjugating enzyme. Elongin B, a ubiquitin-like protein, associates with the complex through interactions with Elongin C and appears to stabilize the binding of Elongin C to VHL.
The best characterized substrates of the VHL ubiquitin ligase are the HIF␣ family of DNA binding transcription factors (1-3, 10, 11), which function together with the constitutively expressed aryl hydrocarbon nuclear translocator to regulate expression of hypoxia-inducible genes when cells are starved for oxygen (12). Cellular HIF␣ protein levels are tightly controlled by the ubiquitin-dependent proteolytic system (13,14). In cells grown in a plentiful supply of oxygen, HIF␣ family members are hydroxylated at critical proline residues within their oxygen-dependent degradation domains (15)(16)(17)(18)(19)(20). Upon VHL binding to their hydroxylated oxygen-dependent degradation domains, HIF␣ family members are rapidly ubiquitylated by the VHL ubiquitin ligase complex and degraded by the proteasome. In cells grown in hypoxic conditions, HIF␣ family members are not hydroxylated. Accordingly, under hypoxic conditions, the VHL ubiquitin ligase complex cannot bind HIF␣s and promote their ubiquitylation and degradation, so cellular levels of these transcription factors rise, leading to activation of hypoxically regulated genes.
Many substrates of SCF ubiquitin ligases have been characterized, and ubiquitylation of most, but not all, depends upon the E2 ubiquitin-conjugating enzyme Cdc34 (30 -34). In contrast, at least in vitro, VHL ubiquitin ligase-dependent ubiquitylation of HIF␣s depends on members of the Ubc5 family of E2s (2,3).
Our laboratory is currently engaged in biochemical studies to define the functions of subunits of VHL and SCF ubiquitin ligases. In this report, we investigate the functions of the Elon-gin C and Skp1 adaptor proteins in reconstitution of active VHL and SCF ubiquitin ligases. We identify Elongin C and Skp1 structural elements responsible for selective interaction with their cognate Cullin/Rbx1 modules. In addition, by exploiting altered specificity Elongin C-Skp1 and Cdc4-VHL chimeric proteins, we provide insights into the determinants of E2 selectivity of VHL and SCF ubiquitin ligases.

EXPERIMENTAL PROCEDURES
Antibodies-Anti-VHL monoclonal antibody (Ig32) was purchased from BD PharMingen. Anti-Myc (9E10) and anti-HA (12CA5) monoclonal antibodies were obtained from Roche Applied Science. Anti-HSV monoclonal antibody was from Novagen. Anti-FLAG monoclonal antibody (M2) was purchased from Sigma. Anti-Cul2 and anti-Elongin C monoclonal antibodies were obtained from BD Transduction Laboratories. Anti-Cul2 (CT2) rabbit polyclonal antibodies were from Zymed Laboratories Inc.. Anti-HIF1␣ monoclonal antibody (H1␣67) was purchased from Novus Biologicals. Anti-protein C monoclonal antibody (HPC4) (35) was provided by C. Esmon (Oklahoma Medical Research Foundation) and is commercially available as anti-protein C from Roche Applied Science. Anti-Elongin B polyclonal antibody was prepared as described (36).
Expression of Recombinant Proteins in Sf21 Insect Cells-Wild type human Elongin C and Elongin C mutants containing N-terminal epitope tags recognized by the HPC4 monoclonal antibody were subcloned into pBacPAK8. Recombinant baculoviruses were generated with the BacPAK baculovirus expression system (Clontech). Baculoviruses encoding human Cul1 and Cul2 containing N-terminal HA epitope tags were described previously (7). Baculoviruses encoding human Cul2, human VHL, human VHL containing an N-terminal His 6 tag, human Elongin B and mouse Rbx1 containing N-terminal Myc epitope tags, and human HIF1␣ containing N-terminal His 6 and HPC4 tags were described previously (3). Baculoviruses encoding S. cerevisiae Cdc53 and mouse Rbx1 containing N-terminal His 6 and Myc tags were described previously (4).
Immunoprecipitations and Western Blotting-Approximately 100 g of Sf21 cell lysates was incubated at 4°C for 2 h with either 10 l of protein G-Sepharose and 2 g of HPC4 antibody or 10 l of protein A-Sepharose and 2 g of the antibodies indicated in the figures. Sepharose beads were washed three times in buffer containing 40 mM Hepes-NaOH (pH 7.9), 150 mM NaCl, 1 mM dithiothreitol, and 0.5% (v/v) Triton X-100. Immunoprecipitated proteins were fractionated by SDS-polyacrylamide gel electrophoresis and transferred to Hybond P membranes (Amersham Biosciences). Membranes were incubated at 4°C overnight with the indicated antibodies in buffer containing 40 mM Tris-HCl (pH 7.6), 100 mM NaCl, and 3% (w/v) nonfat dry milk and then with peroxidase-conjugated secondary antibodies (Sigma). Proteins were visualized on film following treatment of membranes with Supersignal West Pico, Dura, or Femto chemiluminescence reagent (Pierce). In some experiments, proteins were visualized with a Storm gel and blot imaging system following treatment of membranes with ECL plus Western blotting reagents (Amersham Biosciences). Alternatively, membranes were incubated at 4°C overnight with the indicated antibodies in Odyssey blocking solution (LI-COR Biosciences) and then with the appropriate Alexa Fluor 680 (Molecular Probes) or IRDye 800 (Rockland) secondary antibodies. In this case, proteins were visualized and quantitated with the Odyssey infrared imaging system (LI-COR Biosciences).

Purification of Recombinant VHL and Cullin/Rbx1
Complexes from Sf21 Cell Lysates-Plates containing 1 ϫ 10 7 Sf21 cells were coinfected with the recombinant baculoviruses indicated in the text and figures. Sixty hours after infections, cells were collected by centrifugation and resuspended in 1 ml/plate of ice-cold buffer containing 40 mM Hepes-NaOH (pH 7.9), 150 mM NaCl, 5 g/ml leupeptin, 5 g/ml antipain, 5 g/ml pepstatin A, 5 g/ml aprotinin, and 40 mM imidazole (pH 7.9). Cells were lysed by French press (1-inch piston; 16,000 p.s.i. of pressure; American Instrument Company). Following centrifugation at 10,000 ϫ g for 20 min at 4°C, the resulting supernatant was mixed with 1 ml of Ni ϩ2 -agarose pre-equilibrated in buffer containing 40 mM Hepes-NaOH (pH 7.9), 150 mM NaCl, and 40 mM imidazole (pH 7.9). The slurry was incubated at 4°C for 2 h with gentle mixing, washed three times with the same buffer, and packed into a 0.8 cm-diameter column. The column was eluted stepwise with buffer containing 40 mM Hepes-NaOH (pH 7.9), 50 mM NaCl, 10% (v/v) glycerol, and 300 mM imidazole (pH 7.9). VBC-Cullin/Rbx1 complexes were reconstituted by mixing Ni ϩ2agarose-purified VHL-Elongin BC complex (VBC) and Cullin/Rbx1 together and incubating them for 1 h at 4°C. For the experiment in Fig.  5 (see below), VBC-Cul2/Rbx1 complex was further purified by TSK DEAE-NPR HPLC. Ni ϩ2 -agarose fractions were dialyzed against buffer containing 40 mM Tris-HCl (pH 7.9), 100 mM KCl, 1 mM dithiothreitol, 0.5 mM EDTA, and 10% (v/v) glycerol. Following centrifugation at 10,000 ϫ g for 20 min at 4°C, the resulting supernatant was applied to a TSK DEAE-NPR HPLC column (4.6 ϫ 35 mm; Toso-Haas) preequilibrated in the same buffer. The column was eluted at 0.2 ml/min with a 3-ml linear gradient from 100 to 500 mM KCl in the same buffer, and 0.1-ml fractions were collected. For the experiment of Fig. 7 (see below), complexes containing Cdc4-VHL, Elongins B and C, and Cul2/ Rbx1 were purified by the same procedure.
Preparation of Hydroxylated HIF1␣-Sf21 cells were infected with baculoviruses encoding human HIF1␣ containing N-terminal His 6 and HPC4 tags. HIF1␣ was purified from Sf21 cell lysates by Ni ϩ2 -agarose chromatography as described previously (3). Approximately 5 g of purified HIF1␣ was incubated at room temperature for 2 h with 200 l of TNT wheat germ extract translation system (Promega) programmed with a pcDNA3 expression vector encoding the EGLN1 prolyl hydroxylase (20) and containing 100 M FeCl 2 , 2 mM sodium ascorbate, and 5 mM ␣-ketoglutarate. Prolyl hydroxylated HIF1␣ was then purified from the reaction mixture by Ni ϩ2 -agarose chromatography (3).
Computational Methods-Multiple sequence alignments were made using the MACAW program (37). Homology modeling and loop rebuilding were done on the Swiss-Model server (38). Automated docking was done using the GRAMM program (39). Fig. 8 (see below) was generated with the PyMOL package (DeLano Scientific, LLC; www.pymol.org).

RESULTS
As illustrated in Fig. 1, the Elongin C and Skp1 proteins share significant amino acid sequence similarity and belong to a protein superfamily with shared three-dimensional structure. Elongin C and Skp1 function as adaptors that recruit heterodimeric Cul2/Rbx1 and Cul1/Rbx1 modules, respectively, into VHL and SCF ubiquitin ligase complexes. In the case of the VHL ubiquitin ligase, Elongin C recruits a Cul2/Rbx1 module to the complex by interacting specifically with a BC box in the VHL protein (Ref. 40 and references therein) and with an N-terminal Cul2 region (41). In the case of SCF ubiquitin ligase complexes, Skp1 functions by an apparently similar mechanism as Elongin C to recruit a Cul1/Rbx1 module to the com-plex by interacting specifically with an F-box in F-box proteins and with an N-terminal Cul1 region (42,43).
To begin to study the relationship between the roles of Elongin C and Skp1 in reconstitution of their respective ubiquitin ligases, we sought to identify Elongin C sequences required for its interaction with Cul2 and for its recruitment of the Cul2/ Rbx1 module into the VHL complex. To accomplish this, we constructed baculoviruses encoding a systematic series of Elongin C deletion mutants and analyzed the abilities of these Elongin C mutants to support the assembly of VBC and of the complete VBC-Cul2/Rbx1 ubiquitin ligase in Sf21 insect cells. Sf21 cells were coinfected with baculoviruses encoding VHL, Elongin B, Cul2, Rbx1, and wild type or mutant Elongin C with an N-terminal epitope tag recognized by the HPC4 monoclonal antibody. Elongin C-containing complexes were immunoprecipitated with HPC4 monoclonal antibodies and analyzed by Western blotting for the presence of individual subunits of the VHL complex. As shown in Fig. 2A, recruitment of Cul2 to the VHL complex was strongly dependent on both Elongins B and C. Elongin C was quite sensitive to deletions. Only three of the mutants could assemble similarly to wild type Elongin C into the VBC complex; two of these, ⌬41-50 and ⌬51-60, were defective in recruitment of Cul2 into the VHL ubiquitin ligase complex (Fig. 2B). These results suggested that Elongin C sequences between residues 41 and 60 are not crucial for VBC assembly but are important for Elongin C interaction with Cul2 and recruitment of Cul2 into the VHL ubiquitin ligase complex. Notably, some residues within the corresponding region of Skp1 appear to contribute to formation of the Skp1-Cul1 interface (43).
To investigate further the role of Elongin C sequences 41-60 in recruitment of Cul2 into the VHL ubiquitin ligase, we carried out domain swaps between Elongin C and Skp1 in an effort to identify an Elongin C-Skp1 chimeric protein capable of recruiting a Cul1/Rbx1 module into the VHL complex. In one mutant, designated M2, Elongin C residues 40 -47 were replaced with Skp1 residues 25-32. These stretches of sequence in Elongin C and Skp1 fall largely within conserved helix 2 ( Fig. 1) (40,44). A second Elongin C-Skp1 chimeric protein, designated M1, was created by replacing Elongin C residues 47-57 with Skp1 residues 32-42. Elongin C residues 47-57 form a loop (loop 3) between helix 2 and strand 3 in human Elongin C; loop 3 was poorly resolved in the Elongin C x-ray structures (40,45). The corresponding sequences from Skp1 fall within a protease-sensitive region that is also proposed to form an unstructured loop between helix 2 and strand 3 and that was partially deleted from the recombinant Skp1 molecules used to crystallize SCF F-boxSKP2 and Skp1-Skp2 for structural studies (43,44). The detailed structure of this region in both Skp1 and Elongin C remains elusive, and the sequence of loop 3 is not well conserved.
As shown in Fig. 3, VBC complexes containing M1 and M2 Elongin C mutants were both capable of assembling into VHL complexes with Cul2/Rbx1, although less efficiently than VBC Cell lysates and HPC4 immunoprecipitates were subjected to 9 or 13% SDS-polyacrylamide gel electrophoresis and analyzed by Western blotting as described under "Experimental Procedures." Proteins detected in Western blots are indicated to the left of the figure. Myc-tagged Rbx1 was detected with anti-Myc antibody, and HPC4-tagged Elongin C was detected with HPC4 antibody. Visualization of Western blots was accomplished with a Storm gel and blot imaging system as described under "Experimental Procedures." wt, wild type.
FIG. 1. Alignment of Elongin C (EloC) and Skp1 amino acid sequences. Human Elongin C and selected representatives of the Skp1 family were downloaded from the NCBI Conserved Domain Data base (www.ncbi.nlm.nih.gov/Structure/cdd/cddsrv.cgi?uidϭsmart00512) and realigned using Gibbs sampling. Unique identifiers of each sequence in GenBank TM or PDB databases and distances from the N termini of the proteins are shown to the left of each sequence. Segments missing from the three-dimensional structures (see "Results") are underlined. Sequences from Skp1 that were used to replace Elongin C sequences in C M1 and C M2 mutants are shown in magenta. The green font with red shading indicates the residues that are predicted to be within 3 Å of a Cullin mollecule and are likely to make direct contacts. Yellow shading indicates conserved bulky hydrophobic residues (Ile, Leu, Val, Met, Phe, Tyr, and Trp), blue type indicates conserved residues with small side chains (Ala, Gly, and Ser), and green shading indicates other conserved residues (including a group of positively charge side chains Lys, Arg, and His and acidic/amine side chains Asp, Glu, Asn, and Gln). In the secondary structure lines, s stands for a strand, and h stands for a helix. complexes containing wild type Elongin C. In contrast, only complexes containing the Elongin C M1 mutant, in which the Elongin C flexible loop residues were replaced with the corresponding residues from Skp1, were capable of assembling into VHL complexes with Cul1/Rbx1. Taken together, these results argue that Elongin C sequences between amino acids 47 and 57 and Skp1 sequences between amino acids 32 and 42 are important for determining the specificity with which these adaptor proteins interact with their respective Cullins and for recruitment of Cullin/Rbx1 modules into their respective E3 ubiquitin ligase complexes.
Evidence from previous studies indicated that HIF1␣ ubiquitylation by the wild type VHL complex is carried out specifically by the E2 ubiquitin-conjugating enzyme Ubc5 (2, 3), whereas ubiquitylation of SCF target proteins is in most cases carried out by the E2 Cdc34 (30 -34). The E2 ubiquitinconjugating enzyme is believed to be recruited to these ubiquitin ligases by interactions with the RING finger protein Rbx1. In the SCF F-boxSKP2 complex, Rbx1 interacts with other subunits of the complex only through the C-terminal domain of the long, rigid Cul1 molecule, whereas Skp1 and its associated substrate-binding F-box protein subunit are bound to the Cul1 N-terminal domain nearly 100 Å away (43). For these reasons, we considered the possibilities that (i) the Cullin protein might play a significant role in determining E2 selectivity and (ii) HIF1␣ ubiquitylation by the VHL ubiquitin ligase complex containing the altered specificity Elongin C M1 mutant and the Cul1/Rbx1 module might depend on Cdc34 rather than Ubc5. To address these possibilities and to confirm the functionality of the altered specificity Elongin C M1 mutant, we tested the ability of the VHL ubiquitin ligase containing either wild type Elongin C and Cul2/Rbx1 or the Elongin C M1 mutant and Cul1/Rbx1 to activate HIF1␣ ubiquitylation in reactions carried out with the E2 ubiquitinconjugating enzymes human Ubc5a (hUbc5a), human Cdc34 (hCdc34), or yeast Cdc34 (yCdc34).
The recombinant E2 ubiquitin-conjugating enzymes hUbc5a, hCdc34, and yCdc34 used in these experiments were purified to FIG. 4. Characterization of E2 ubiquitin-conjugating enzymes used in this study. A, yCdc34, hCdc34, and hUbc5a, expressed in and purified from E. coli, were analyzed by SDS-polyacrylamide gel electrophoresis and visualized with Coomassie Brilliant Blue. B, ability of yCdc34, hCdc34, and hUbc5a to form thioester bonds. The indicated amounts of each E2 were incubated for 1 h at 25°C with 50 ng of yeast Uba1 and 100 ng of His-T7-Xpress-ubiquitin in 10-l reactions containing 40 mM Hepes-NaOH (pH 7.9), 60 mM potassium acetate, 2 mM dithiothreitol, 6 mM MgCl 2 , 0.5 mM EDTA (pH 8.0), 10% (v/v) glycerol, and 1.5 mM ATP. Reactions were stopped with 15 l of buffer containing either 30 mM dithiothreitol (reducing) or 6 M urea (nonreducing). Reaction products were subjected to SDS-polyacrylamide gel electrophoresis and analyzed by Western blotting (WB) with anti-T7 antibodies as described under "Experimental Procedures." Reaction products were visualized using the LI-COR Odyssey infrared imager as described under "Experimental Procedures." C, hUbc5a supports VBC-Cul2/Rbx1dependent ubiquitylation of hydroxylated HIF1␣. As described under "Experimental Procedures," purified His-HPC4-HIF1␣ was treated with wheat germ extract programmed with or without HA-EGLN1 prolyl hydroxylase, purified by Ni ϩ2 -agarose chromatography, and assayed in ubiquitylation reactions containing 500 ng of hUbc5a in the presence or absence of the purified VBC-Cul2/Rbx1 complex. Reaction mixtures were subjected to 8% SDS-polyacrylamide gel electrophoresis and analyzed as described under "Experimental Procedures" by Western blotting with HPC4 monoclonal antibody. D, yCdc34 and hCdc34 support SCF Cdc4 -dependent ubiquitylation of phosphorylated Sic1 (circled P indicates phosphorylation). Lane 1, phosphorylated HPC4-Sic1. Lanes 2-9, yCdc34 (60 ng), hCdc34 (110 ng), and hUbc5a (50 ng) were assayed for their abilities to support ubiquitylation of phosphorylated HPC4-Sic1 in the absence (lanes 2-5) or presence (lanes 6 -9) of SCF Cdc4 , which had been immunopurified with anti-FLAG antibody from lysates of baculovirus-infected Sf21 cells expressing Cdc53, HSV-Cdc4, Myc-Rbx1, and FLAG-Skp1 as described (4). Reaction products were fractionated by SDS-polyacrylamide gel electrophoresis and analyzed as described under "Experimental Procedures" by Western blotting with HPC4 antibody. The band marked with an asterisk is a nonspecific band derived from FLAG immunoprecipitation. near homogeneity from Escherichia coli (Fig. 4A) and assayed for their abilities to form E2-ubiquitin thioester conjugates in the presence of an E1 ubiquitin-activating enzyme, GST⅐ubiquitin, and ATP. As shown in the experiment in Fig.  4B, all three enzyme preparations were active in this assay, although the maximum level of E2-ubiquitin thioester conjugate detected with hCdc34 was lower than with yCdc34 or hUbc5a. In addition, hUbc5a supported VHL ubiquitin ligasedependent ubiquitylation of prolyl-hydroxylated HIF1␣, and both yCdc34 and hCdc34 supported SCF Cdc4 -dependent ubiquitylation of phosphorylated Sic1 (Fig. 4, C and D). We note that hUbc5a could also support SCF Cdc4 -dependent ubiquitylation of phosphorylated Sic1, albeit at a much reduced level when compared with yCdc34 and hCdc34.
In preliminary experiments, we found that it was difficult to reproducibly and in good yield prepare recombinant VHL complexes containing stoichiometric amounts of Cul2/Rbx1 or Cul1/Rbx1 by coexpression of VHL, Elongins B and C, Rbx1, and Cul2 or Cul1 in insect cells (data not shown). To reconstitute VHL ubiquitin ligase complexes, we therefore mixed recombinant heterotrimeric VBC with recombinant Cullin/Rbx1 complexes. Confirming the validity of this approach, hydroxylated HIF1␣ could be ubiquitylated with comparable efficiency by recombinant VHL complexes that had been expressed in and purified intact from insect cells or by VHL complexes reconstituted with a mixture of VBC and Cul2/Rbx1 complexes, which had been expressed and purified separately (Fig. 5).
As shown in Fig. 6, both wild type VHL complex and VHL complex containing the altered specificity Elongin C M1 mutant were capable of supporting HIF1␣ ubiquitylation. As expected, the VHL complex containing wild type Elongin C was much more active in HIF1␣ ubiquitylation in the presence of Cul2/Rbx1 than Cul1/Rbx1. VHL complex containing the Elongin C M1 mutant was active in HIF1␣ ubiquitylation in the presence of Cul1/Rbx1 but exhibited little activity in the presence of Cul2/Rbx1, although it could bind Cul2/Rbx1. Remarkably, HIF1␣ ubiquitylation was dependent on the E2 Ubc5a in all cases, despite previous studies indicating that ubiquitylation of target proteins by most SCF ubiquitin ligases containing the Cul1/Rbx1 module is carried out by the E2 Cdc34. Taken together, these findings demonstrated that VBC-dependent HIF1␣ ubiquitylation can be activated by a Cul1/Rbx1 module. In addition, they support the idea that the Cullin/Rbx1 module is not solely responsible for determining the E2 selectivity of Cullin-containing ubiquitin ligases, although the E2 is believed to dock on the Cullin/Rbx1 module at the C terminus of Cullin proteins, a relatively large distance away from other subunits of the ubiquitin ligase complex.
We next investigated the possibility that the identity of the target protein might influence E2 selectivity in our assays. In these experiments, we took advantage of a Cdc4-VHL chimeric substrate recognition subunit (46). The S. cerevisiae F-box protein Cdc4 interacts through its WD repeat domain with the S. cerevisiae Cdk inhibitor Sic1 and recruits it for ubiquitylation by the SCF Cdc4 ubiquitin ligase. The Cdc4-VHL chimeric protein is composed of the entire 213-amino-acid human VHL protein fused to the C terminus of a Cdc4 deletion mutant containing residues 341-779, which includes the Cdc4 WD repeat domain but lacks the Cdc4 F-box (Fig. 7A). The Cdc4-VHL chimera binds stably to the Elongin BC complex through the VHL BC box but does not interact with Skp1 since it lacks an intact F-box (46).
The SCF Cdc4 ubiquitin ligase targets phosphorylated Sic1 for ubiquitylation by the E2 ubiquitin-conjugating enzyme Cdc34 (21)(22)(23)47). Cdc4 binds phosphorylated Sic1 via its WD repeat domain and recruits it to the SCF Cdc4 complex for ubiquitylation by Cdc34. To determine whether the Cdc4-VHL chimeric protein can assemble into an E3 ubiquitin ligase containing Elongins B and C and a Cul2/Rbx1 module, Sf21 insect cells FIG. 5. Reconstitution of VHL ubiquitin ligase with isolated VBC and Cul2/Rbx1 complexes. VBC-Cul2/Rbx1, VBC, and Cullin/ Rbx1 complexes were prepared as described under "Experimental Procedures." EGLN1-treated HIF1␣ was assayed as described under "Experimental Procedures" in ubiquitylation reactions containing ϳ800 ng of hUbc5a in the absence (lane 1) or presence (lane 2) of purified VBC-Cul2/Rbx1 complex or a mixture of purified VBC and Cul2/Rbx1 that had been preincubated for 60 min at 4°C (lane 3). Reaction mixtures were subjected to 8% SDS-polyacrylamide gel electrophoresis. Proteins in gels were transferred to Hybond P and visualized as described under "Experimental Procedures" by Western blotting with anti-HIF1␣ antibody.
FIG. 6. Cul1-dependent ubiquitylation of HIF1␣ by hUbc5a. A, VBC complexes containing either wild type (WT) or M1 mutant Elongin C were preincubated for 60 min at 4°C alone or in the presence of Cul1/Rbx1 or Cul2/Rbx1. The resulting mixtures were assayed for their abilities to support ubiquitylation of prolyl-hydroxylated HIF1␣ in the presence of 0.5 g of hUbc5a or 5.4 g of hCdc34. Reaction products were analyzed by Western blotting with anti-HIF1␣ antibody and visualized on film. B, VBC complexes containing either wild type or M1 mutant Elongin C were preincubated for 60 min at 4°C in the presence of Cul1/Rbx1 or Cul2/Rbx1. The resulting mixtures were assayed for their abilities to support ubiquitylation of prolyl-hydroxylated HIF1␣ in the presence of the indicated amounts of yCdc34, hCdc34, or hUbc5a. Reaction products were analyzed by Western blotting with anti-HIF1␣ antibody and visualized using an Odyssey infrared imaging system (LI-COR Biosciences).
were coinfected with baculoviruses encoding the Cdc4-VHL chimera containing an N-terminal His 6 tag, Elongins B and C, Cul2, and Rbx1. The Cdc4-VHL complex was purified from cell lysates by Ni 2ϩ -agarose chromatography followed by TSK DEAE 5-PW HPLC. As shown in the Coomassie-stained SDSpolyacrylamide gel of Fig. 7B, the Cdc4-VHL chimera could be purified as part of a multiprotein complex with roughly stoichiometric amounts of Elongins B and C, Cul2, and Rbx1. As shown in Fig. 7C, Sic1 ubiquitylation by the Cdc4-VHL ubiquitin ligase complex was strongly dependent on the E2 ubiquitin-conjugating enzyme Ubc5 rather than Cdc34. Sic1 ubiquitylation was dependent on the Cdc4 portion of the Cdc4-VHL chimeric protein since the purified VHL complex did not support Sic1 ubiquitylation in the presence of either Ubc5 (Fig.  7D) or Cdc34 (data not shown).

DISCUSSION
In this report, we investigate the functions of the Elongin C and Skp1 proteins in reconstitution of VHL and SCF ubiquitin ligases. We identify Elongin C and Skp1 structural elements responsible for selective interaction with their cognate Cullin/Rbx1 modules. In addition, using altered specificity Elongin C and F-box protein mutants, we investigate models for the mechanism underlying E2 selection by VHL and SCF ubiquitin ligases.
Our findings suggest that the ϳ11 amino acid loop 3 between helix 2 and strand 3 in the conserved three-dimensional structure of Elongin C and Skp1 (Fig. 1) plays an important role in Cullin selection by the VBC complex. Wild type Elongin C preferentially binds to Cul2. However, when loop 3 is replaced by its counterpart from Cul1-binding Skp1, the chimeric Elongin C M1 mutant can bind both Cul1 and Cul2, and it requires Cul1 for efficient HIF1␣ ubiquitylation.
Little is known about loop 3 structure and function. In the available structures of Elongin C (Protein Data Bank (PDB) accession numbers 1VCB and 1LM8), this region is disordered (40,45), and in Skp1 (PDB accession number 1LDK), it appeared to interfere with crystallization and had to be deleted (43,44). In an effort to gain better understanding of the function of loop 3, we rebuilt both the wild type loop and its replacement in Elongin C M1, using the corresponding sequences from either Elongin C or Skp1 with the isolated C chain of 1LM8 as the template. The resulting model was docked onto Cul1 (the A chain of 1LDK) by global energy minimization or was homology-modeled onto the Skp1 structure (the C chain of 1LDK). The results obtained using both approaches are very similar and suggest that loop 3 might not be involved in a direct interaction with the Cullin protein ( Fig. 8 and data not shown). In contrast with numerous direct, mainly hydrophobic interactions between helices that are conserved between Skp1 and Elongin C and between Cul1 and Cul2, respectively, all atoms in the predicted loop 3 structure in Elongin C M1 are separated by at FIG. 8. Model of EloC-Cul1 interface. The pink ribbon indicates the fragment of Cul1 (residues 16 -55 and 112-149 of PDB entry 1LDK, chain A) that contributes to the interface. Elongin C is shown with rebuilt loop 3 from wild type Elongin C (blue) or Skp1 (magenta). Elongin C ␣-helices are shown in red, ␤-strands are shown in yellow, and the rebuilt loop 3 from wild type Elongin C is shown in blue whereas that from Skp1 (corresponding to the M1 mutant) is shown in magenta. Side chains predicted to be involved in EloC-Cul1 interactions are shown in green.

FIG. 7. Ubiquitylation of phosphorylated Sic1 by an Elongin BC-and
Cul2-containing ubiquitin ligase by hUbc5a. A, the Cdc4-VHL chimera contains residues 341-779 of Cdc4 fused to the complete VHL open reading frame. The Cdc4 F-box is indicated by the black box, the Cdc4 WD repeats are indicated by the dark gray boxes, and the VHL BC box is indicated by the white box. WT, wild type. B, ubiquitin ligase complexes containing either Cdc4 or Cdc4-VHL as F-box protein were expressed in insect cells and purified as described. ElobB, Elongin B; ElobC, Elongin C. C, purified Cdc4-VHL-BC-Cul2/Rbx1 complexes were assayed for their abilities to support ubiquitylation of phosphorylated HPC4-Sic1 in the presence of ϳ100 ng of hUbc5a or yCdc34. Reaction products were analyzed by Western blotting (WB) with HPC4 monoclonal antibody and visualized on film. D, purified Cdc4-VHL-BC-Cul2/ Rbx1 or VBC-Cul2/Rbx1 complexes were assayed as in panel C for their abilities to support ubiquitylation of phosphorylated HPC4-Sic1. least 6 Å from any atoms in Cul1, and this distance is even larger for the predicted loop 3 structure in wild type Elongin C.
What might be the molecular mechanism by which loop 3 affects the specificity with which Elongin C interacts with Cullin proteins? There are several possibilities that could not be addressed by our approximate structural model. First, loop 3 might adopt a folded or bent conformation in the process of Elongin C-Cullin interaction, which perhaps could produce additional contact(s) between Cul1 and the Elongin C M1 mutant but not between Cul1 and wild type Elongin C. Second, the ϳ6-Å distance between Elongin C M1 loop 3 and Cul1 could be bridged by a water molecule, thus producing additional Elongin C M1-specific interactions. Finally, we note that our model predicts that loop 3 of Elongin C M1 would contribute an extra hydrogen bond to interact with a lysine residue in the middle of strand 1 (data not shown), perhaps inducing a change in the packing of the structural core of the Elongin C M1 mutant. Experimental structure-function studies are needed to determine whether any of these possibilities contribute to the specificity of protein-protein interactions that we observed; what is clear is that the mechanism of loop 3-dependent Cullin selection is subtle.
By exploiting the altered specificity Elongin C M1 mutant, we have investigated the importance of the Cullin/Rbx1 module in E2 selection by the VHL ubiquitin ligase complex. Our observation that hUbc5a and not Cdc34 is the preferred E2 for ubiquitylation of prolyl-hydroxylated HIF1␣ directed by a Cul1-containing VHL ubiquitin ligase suggests that the Cullin/ Rbx1 module is not solely responsible for determining the E2 selectivity of substrate-specific ubiquitylation. In addition, our observation that hUbc5a is the preferred E2 for ubiquitylation of phosphorylated Sic1 directed by an Elongin BC-based ubiquitin ligase containing a chimeric Cdc4-VHL substrate recognition subunit suggests that the target protein is not solely responsible for determining the E2 selectivity of the reaction. It could be argued that Sic1 presented to the E2 by the Cdc4-VHL ubiquitin ligase is not the same substrate as Sic1 presented to the E2 by SCF Cdc4 since the position and orientation of Sic1 relative to the E2 could be different in the two cases. For several reasons, however, we do not favor this possibility. Comparison of the VHL-Elongin BC (1VBC) (40) and Skp1-Cdc4 (1NEX) (49) structures suggests that Sic1 could be positioned similarly in the two cases. The space occupied by the Cdc4 F-box and the F-box binding helices of Skp1 in the Skp1-Cdc4 structure is approximately the same as that occupied by VHL in the VHL-Elongin BC structure. Although the Cdc4 F-box is located N-terminal to the Sic1-binding WD repeat domain, and VHL is located C-terminal to the WD repeat in the Cdc4-VHL chimera, the N and C termini of the WD repeat domain emerge from the same face of the ␤-propeller, suggesting that the Sic1 binding site is likely to be oriented similarly in SCF Cdc4 and in the Cdc4-VHL ubiquitin ligase. Furthermore, high affinity binding of Sic1 to Cdc4 is due to cooperative interactions between any of several spatially separated, low affinity phosphobinding sites on Sic1 to a single site on Cdc4, suggesting that Sic1 can bind productively to Cdc4 in any of several orientations (48,49).