The importance of the ubiquitin system largely depends on ubiquitin ligases, E3s, as they determine the specificity of the system. Rbx1/ROC1/Hrt1, a RING finger protein, functions as an important component of the cullin-containing SCF and VBC-Cul2 ligases. Modification of cullins by NEDD8 (NEDDylation), has been shown to be essential for the E3 activity of both SCF and VBC-Cul2, and it was suggested that Rbx1 acts as the E3 for cullin NEDDylation. RING finger is composed of eight cysteine and histidine residues that bind to zinc ions. Rbx1 is a highly evolutionarily conserved protein; however, the eighth coordination residue in its RING finger is aspartate (D97) rather than cysteine. Substitution of D97 with each of the other 19 amino acids demonstrates that aspartate is superior to cysteine in cullin NEDDylation. Interestingly, however, different D97 mutants demonstrate different activities towards 6 cullins tested. Importantly, we were able to discriminate between the NEDDylating activity of Rbx1 and its involvement in the ubiquitylation reaction within the context of VBC-Cul2. Moreover, while Rbx1 is not involved in governing the stability of SCF, Rbx1 mutants destabilize VBC-Cul2. Taken together, these results indicate that various mechanisms regulate both the activities and the stability of cullin-based ligases.
The ubiquitin system plays important roles in the regulation of a broad array of cellular functions by controlling the fate of numerous regulatory proteins. Degradation of a substrate by the ubiquitin system requires its prior ubiquitylation before it is recognized by the downstream 26S proteasome. Ubiquitin is conjugated to target proteins by a cascade of reactions catalyzed by three enzymes, E1, E2 and E3 (Hershko & Ciechanover 1998; Weissman 2001). In this process, ubiquitin is first activated by ATP to form a high-energy thiol-ester intermediate. The activated ubiquitin is then transferred from E1 to one of several E2s. In the presence of ubiquitin ligase E3, E2 transfers the activated ubiquitin to the target protein that is bound specifically to a particular E3. Therefore, E3s play an important role in the system, as they determine its substrate specificity (Hershko & Ciechanover 1998; Weissman 2001). Several distinct E3 ligase families have been identified based on commonly shared structural motifs/subunits/mechanism of action. RING finger-containing E3s comprise the largest ligase family, and contain both monomeric and multimeric ubiquitin ligases (Weissman 2001). Of the multimeric RING finger ligases, the Cul1-based SCF (Skp1-Cul1-F-box) ligases form the largest subfamily (Deshaies 1999).
von Hippel-Lindau (VHL) disease is a dominant inherited syndrome characterized by the predispostion of benign and malignant tumors. VHL disease is caused by germline mutations in one allele of the VHL tumor suppressor gene and inactivation or loss of the other allele (George & Kaelin 2003). pVHL, the product of the VHL tumor suppressor gene, has been shown to associate with Elongin B (Elo B), Elongin C (Elo C) and Cul2 to generate a complex termed VBC-Cul2 (Pause et al. 1997), which is an E3 (Lisztwan et al. 1999; Iwai et al. 1999). Recent studies indicate that VBC-Cul2 targets the α subunits of hypoxia-inducible factors, including HIF-1 and HIF-2, for ubiquitylation which leads to subsequent degradation by the proteasome (Huang et al. 1998). The ubiquitylation of HIF-1α and HIF-2α is triggered by oxygen-dependent hydroxylation of specific proline residues in the target proteins (Ivan et al. 2001; Jaakkola et al. 2001), a reaction that is catalyzed by several HIF prolyl hydroxylases termed EGLN1–3 (Epstein et al. 2001; Bruick & McKnight 2001).
Both SCF and VBC-Cul2 share several common characteristics. First, both contain the Rbx1/ROC1/Hrt1 RING finger protein subunit. One important function of the RING finger component is recruitment of the E2 enzyme to the complex (Weissman 2001). Second, the cullin components in both complexes (Cul1 in SCF and Cul2 in VBC-Cul2) are modified by NEDD8, a ubiquitin-like protein (Hori et al. 1999). NEDDylation has been shown to be essential for the E3 activity of both SCF and VBC-Cul2 (Osaka et al. 2000; Gong & Yeh 1999; Kamura et al. 1999). NEDDylation is catalyzed by several enzymes that are similar, though are not identical, to those involved in ubiquitylation. NEDD8 is activated by a heterodimeric E1-like enzyme, the APP-BP1 and Uba3 complex, and is then transferred to an E2-like enzyme, Ubc12 (Gong & Yeh 1999). Rbx1 has been shown to function not only as a component of the ubiquitylating E3 complex, but also a NEDD8 ligase for Rub1, the yeast homolog of NEDD8 (Kamura et al. 1999). Rbx1 is a RING finger protein that has been conserved along evolution from yeast to higher eukaryotes, including humans. Typically, the RING finger, which is found in many other proteins as well, is composed of eight cysteine and histidine residues that bind to zinc ions to form a unique three dimensional structure. However, in Rbx1, the eighth coordination residue of the RING finger is an aspartate (D97; Fig. 1). To investigate the role of the D97 in cullin-based E3s, we substituted it with each of the other 19 amino acids. The function of the different Rbx1 proteins was examined either following its purification with the different components of the VBC-Cul2, or directly in cells. We demonstrate that different mutants have different effects on NEDDylation of distinct cullins, and that the NEDDylating activity of Rbx1 is distinct from its involvement in ubiquitylation within the VBC-Cul2 complex. Also, we show that different Rbx1 mutants have different effects on VBC stability, unlike their lack of such a role in governing the stability of SCF complexes. We propose that the fine structure of Rbx1 RING is crucial for both regulating cullin-based ligase activity as well as determining VBC-Cul2 ligase stability.
Rbx1 D97 is more efficient than its C97 consensus counterpart in cullin NEDDylation
To determine why the D97 residue in the RING finger of Rbx1 is conserved throughout evolution, in contrast to a conserved cysteine residue in this position in all other RING finger proteins, we mutated D97 by substituting it with each of the other 19 amino acids. We then assessed the effect of these mutations on the NEDDylation of various cullins. We generated recombinant baculoviruses expressing each of the mutant Rbx1 proteins and infected Hi Five insect cells with these baculoviruses along with viruses expressing GST-NEDD8 and myc-tagged cullin proteins. Since six cullins—Cul1, Cul2, Cul3, Cul4A, Cul4B and Cul5—have been identified in humans and have all been shown to bind to Rbx1 and to be modified by NEDD8 (Hori et al. 1999), we tested the NEDDylation of each of these proteins by the mutant Rbx1 proteins (Fig. 2). As a control, we tested the effect of the C75A/H77A mutant Rbx1, which is known to lack E3 NEDDylating activity (Chen et al. 2000). Interestingly, expression of any of the Rbx1 proteins, the WT and the different mutants alike, resulted in higher expression of Cul2, Cul3, Cul4A and Cul4B (Fig. 2, compare lane 1 to all other lanes for the respective cullins). The level of Cul1 and Cul5 was not affected by Rbx1 expression. It is possible that complex formation between Rbx1 and the different cullins stabilize the latter. Cells that are not expressing Rbx1 did not demonstrate, as expected, any cullin-NEDDylating activity (Fig. 2, lane 1). In contrast, all cullins were NEDDylated when expressed along with WT Rbx1 (lane 2). This finding indicates that the NEDDylation of cullins depends on the activity of the expressed Rbx1. Substitution of D97 by bulky hydrophobic amino acids (I, L, M, P, V: lanes 9, 11, 12, 14, 19, respectively), aromatic amino acids (F, Y, W: lanes 6, 21, 20, respectively), or basic amino acids (K, R: lanes 10, 16, respectively) abolished the NEDDylation of all the cullins tested. In contrast, Rbx1 substituted with an acidic residue (D (WT), E: lanes 2, 5, respectively), threonine (T: lane 18), or alanine (A: lane 3) supported NEDDylation of all the cullins. The remaining mutations exhibited non-uniform effects on cullin NEDDylation. Of these, D97C (lane 4), an Rbx1 mutant carrying a consensus cysteine in the eighth zinc coordination site, supported the NEDDylation of Cul1, Cul2, Cul4A and Cul5, but only poorly of Cul3 and Cul4B. These results explain why cysteine, the conserved residue in this position in all other RING finger proteins, cannot function in Rbx1.
VBC-Cul2 containing defined mutated Rbx1 species display distinct E3 activities in ubiquitylation and NEDDylation
Rbx1 was shown to function as an important component in cullin-containing E3s. Therefore, it was important to assess the effect of mutations in Rbx1 D97 on the E3 activity of VBC-Cul2. As a substrate, we used the proline-hydroxylated ODD domain of HIF-2α. As can be seen in Fig. 3, HIF-2α ODD was not ubiquitylated in the absence of the EGLN3 proline hydroxylase (lane 1), but strongly ubiquitylated in the presence of EGLN3 by VBC-Cul2 that contains WT-Rbx1 (lane 3), indicating that ubiquitylation of HIF-2α ODD depends on the hydroxylation of the specific proline residue (proline 531 in HIF-2α) in the domain. Moreover, the ubiquitylation of the ODD domain depended on the presence of VBC-Cul2 ligase (lanes 2, 3). NEDD8 conjugation to Cul2 had little effect on ODD ubiquitylation by VBC-Cul2 ligase in this in vitro assay (lanes 3, 5). We suspect that it is due to the large amount of E2 (200 ng of UbcH5c) in our reaction mixture, because it has been shown that NEDD8 conjugation to Cul1 enhances recruitment of E2 for ubiquitin to Rbx1 and substrates are ubiquitylated efficiently in the presence of large amount of E2 (Kawakami et al. 2001). We thus decided to examine the effect of mutations in Rbx1 D97 on ODD ubiquitylation in the absence of the NEDD8 conjugation system to neglect the effect of the Rbx1 mutants on the NEDDylation. Interestingly, certain Rbx1 mutants revealed distinctive ubiquitylating and NEDDylating activities. Thus, Rbx1 D97S demonstrated efficient ODD ubiquitylation (Fig. 3, lane 10) while it failed to NEDDylate most cullins (Fig. 2, lane 17). Conversely, VBC-Cul2 that contains Rbx1 D97A demonstrated a low ubiquitylating activity (Fig. 3, lane 11), though it was efficient in NEDDylating all cullins (Fig. 2, lane 3). In other Rbx1 mutants, the two activities correlated. Collectively, these results suggest that Rbx1 possesses distinct E3 activities for ubiquitylation and NEDDylation.
Rbx1 mutants defective in Cul2 NEDDylation down-regulate VBC-Cul2
It has been shown that Rbx1 is essential for the E3 activity of both SCF and VBC-Cul2, probably by recruiting the E2 component (Osaka et al. 2000; Ohh et al. 2002). To shed light on other potential roles of Rbx1 in the activity/fate of these ligases, we examined whether the expression of NEDDylation-defective Rbx1 mutants affect the protein levels of the VBC-Cul2 ligase components. Initially, 293T cells were co-transfected with expression plasmids encoding Rbx1 WT or D97R along with those encoding Elo B, Elo C, Cul2 and pVHL as indicated in Fig. 4A. Co-expression of Cul2 and Rbx1 WT enhanced the amount of Rbx1 (lane 1, 3) as well as Cul2 (lane 2, 3), indicating that the expression of Cul2 and Rbx1 WT are interdependent. The amounts of Elo B and Elo C were also increased by the presence of Cul2 and Rbx1 WT (lanes 1–3), suggesting that the presence of Cul2 and Rbx1 WT enhances the amounts of Elo B and Elo C. However, the amounts of Elo B and Elo C were decreased in cells expressing Cul2 and Rbx1 D97R in spite that the amount of Cul2 was not affected by the presence of Rbx1 D97R (lane 3, 4). The amount of pVHL, however, was barely affected by the presence of the mutant. To probe the mechanism involved in the decrease of Elo B and Elo C in Rbx1 D97R-expressing cells, we examined the half-life of Elo B as a representative of these two proteins in the presence of Rbx1 WT or D97R. 293T cells co-transfected with expression vectors encoding Rbx1 WT or D97R along with those encoding Elo B, Elo C, Cul2 and pVHL were treated with CHX 36 h after transfection (‘chase’). The amounts of Elo B and Cul2 were assessed by immunoblotting. The amount of Elo B decreased more rapidly in cells expressing Rbx1 D97R than in cells expressing Rbx1 WT, whereas the amount of Cul2 was not affected by the co-expressing Rbx1 (Fig. 4B–D). These results suggest that Rbx1 D97R destabilizes Elo B and Elo C.
To confirm the effect of the Rbx1 mutant on the destabilization of Elo B and Elo C observed in mammalian cells, Hi Five cells were infected with recombinant baculoviruses expressing different mutants of Rbx1 along with NEDD8, and Elo B, Elo C, Cul2 and pVHL. As can be seen in Fig. 5A, less Elo C were observed in the lysates when Rbx1 mutants that do not support Cul2 NEDDylation (Rbx1 D97P, D97R or C75A/H77A: lanes 4–6) were present compared to when NEDDylation-supporting Rbx1 proteins (WT, D97C or D97E: lanes 1–3) were present. To our surprise, the amount of pVHL is lower in cells expressing Rbx1 D97P, D97R or C75A/H77A than in those expressing Rbx1 WT, D97C or D97E. Similar results were obtained in a pull-down experiment comparing WT to D97R Rbx1s (Fig. 5B). Interestingly, the levels of Cul2 and of IRP2 (a nonspecific control) are not affected, regardless of the Rbx1 protein present (Fig. 5A, lanes 1–6). To probe the mechanism involved in the decrease of the VBC level in Rbx1 D97R-expressing cells, we examined the half-life of Elo C as a representative of the VBC components of the ligase complex in the presence of Rbx1 WT or D97R. CHX treatment of Hi Five cells infected with baculovirus expressing Rbx1 WT or D97R along with those expressing NEDD8, Elo B, Elo C, Cul2 and pVHL revealed that the amount of Elo C decreased more rapidly in cells expressing Rbx1 D97R than in cells expressing Rbx1 WT (Fig. 5C,D) although the amount of Cul2 was not affected by the co-expressed Rbx1 (Fig. 5C,E). Thus, expression of Rbx1 mutants that are defective in Cul2 NEDDylation selectively decreases the cellular level of the VBC complex, probably by destabilizing its components. Collectively, these results suggest that Rbx1 mutants defective for cullin NEDDylation down-regulate the VBC complex in insect cells as well as in mammalian cells and the effect of the Rbx1 mutant is stronger in insect cells than in mammlian cells.
It is possible that Cul2 binding of a Rbx1 mutant that is NEDDylation defective can reduce the amount of VBC-Cul2 by destabilizing the VBC complex. To test this hypothesis, we expressed increasing amounts of Rbx1 WT in the presence of a constant amount of Rbx1 D97R, and monitored the level of Elo C as a representative of the VBC complex components in insect cells. As can be seen in Fig. 5F, Elo C expression is down-regulated by the co-infection, even of a small amount of Rbx1 D97R (lane 2). Co-infection of Rbx1 WT along with the D97R mutant restores the amount of Elo C (as well as of Cul2 NEDDylation), which indicates that it is the presence of the Rbx1 WT in the VBC-Cul2 complex that reverses the suppressive effect of Rbx1 D97R on the maintenance of VBC-Cul2 and Cul2 NEDDylation.
Rbx1 mutants defective in cullin NEDDylation have no effect on the amount of SCF complex
We next examined whether the expression of NEDDylation-defective Rbx1 mutants affect the protein levels of the SCF ligase components. Hi Five cells were infected with baculoviruses expressing the components of SCFβTrCP1 (the F-box protein β-TrCP1, Skp1, and Cul1) together with viruses expressing NEDD8 and WT or mutant Rbx1. As can be seen in Fig. 2, Cul1 NEDDylation was suppressed in the presence of Rbx1 D97P, D97R and C75A/H77A (Fig. 6A, lanes 4–6). In striking contrast, however, the levels of Skp1 and β-TrCP1 in the lysates were almost identical in the presence of the WT and mutant Rbx1, as was the case with the amount of Skp1 and β-TrCP1 that were co-immunoprecipitated with Cul1 (Fig. 6A,B). The half-life of Skp1 in Rbx1 D97R-expressing cells was not different from that measured in WT Rbx1-expressing cells (Fig. 6C–E). Taken together, these results suggest that the presence of Rbx1 mutants defective in NEDDylation down-regulate in a specific manner, VBC-Cul2.
Cul2 NEDDylation has no effect on the amount of VBC-Cul2 in cells
We found that the levels of the VBC subunits were decreased when NEDDylation-defective mutants of Rbx1 were present (Figs 4 and 5). To determine whether the lack of Cul2 NEDDylation is responsible for the lower levels of VBC, we used a Cul2 mutant in which lysine 689, the residue that is NEDDylated (Gong & Yeh 1999), was substituted with arginine (K689R). As expected, no NEDDylation of Cul2 K689R was observed when the Cul2 K689R virus was co-infected along with WT Rbx1 (Fig. 7, lane 7). However, the levels of pVHL, Elo B and Elo C were not down-regulated in the presence of NEDDylation-supporting Rbx1 proteins regardless of whether Cul2 K689R (lanes 7, 8) or WT Cul2 (lanes 2, 3) were present. These results indicate that Cul2 NEDDylation is dispensable for the maintenance of VBC-Cul2.
Rbx1 is a small RING finger protein that is essential for the E3 activity of SCF and VBC-Cul2 (Osaka et al. 2000; Ohh et al. 2002). The protein is highly conserved and its RING finger differs from other RING fingers in that its eighth zinc coordination residue is an aspartic acid (D97) rather than a cysteine.
Rbx1 plays distinct roles in NEDDylation and ubiquitylation
We show here that Rbx1 D97C, an Rbx1 mutant possessing a cysteine residue in the eighth zinc coordination site, could NEDDylate Cul1, Cul2, Cul4A and Cul5, but not Cul3 and Cul4B. In contrast, Rbx1 WT could NEDDylate all the cullins (Fig. 2). Thus, while the role of NEDDylation of Cul3 and Cul4B is still not known, it is clear that a consensus, cysteine-containing Rbx1 could not NEDDylate these species.
RING finger domains have been shown to be the binding sites for E2s (Lorick et al. 1999). Therefore, Rbx1 should bind to both Ubc12, the NEDD8 E2, and the E2s involved in ubiquitylation. Our series of mutations of Rbx1 have unexpectedly revealed that Rbx1 RING has different requirements in its function as an E3 for ubiquitin or NEDD8. While Rbx1 D97A can efficiently NEDDylate all cullins (Fig. 2, lane 3), its ubiquitylating activity is relatively weak (Fig. 3, lane 11). Conversely, while Rbx1 D97S shows a weak NEDDylating acticity towards Cul2, Cul3, Cul4A and Cul4B (Fig. 2, lane 17), its ubiquitylating activity towards the ODD domain of HIF-2α is only slightly affected (Fig. 3, lane 10). Kamura et al. (1999) reported that Rbx1 mutants capable of supporting NEDDylation can also ubiquitylate model substrates. While their data appear to differ from ours, this may not be the case. Both Rbx1 D97A and D97S maintain the E3 activity for ubiquitin as well as for NEDD8 even though the ability to conjugate the two modifiers is different among these mutants. Therefore, our analyses of Rbx1 mutants suggest that some Rbx1 mutations preferably affect ubiquitylation, whereas others affect more NEDDylation. Mechanistically, it is possible that the differences in activities are due to differential binding of the different E2s, Ubc12 and the ubiquitylating E2s, to the different Rbx1 mutants. An interesting question is how does Rbx1 function as the active center for the two conjugation systems that target distinct substrates using different E2 enzymes? A recent report by Deffenbaugh et al. (2003) has shown that in the case of the yeast SCFCdc4 complex, the ubiquitin-charged E2 binds to the E3 only temporarily and transfers ubiquitin to the substrate when it is released from the E3 complex, indicating that E2s do not necessarily have to be bound to E3s when they transfer the modifiers to the target molecules. In analogy, Rbx1, which is a small protein, may induce NEDDylation and ubiquitylation at the same time by allowing alternate binding of the different E2s. If the E2s bind to different sites, then one can explain how a certain mutation affects one process more than the other. The recently solved crystal structures of VBC (Stebbins et al. 1999) and SCF (Zheng et al. 2002b), which have revealed that the cullin NEDDylation site is spatially opposite to the substrate ubiquitylation site, also support the notion that Ubc12 and E2s for ubiquitin bind to different sites of Rbx1. However, it is also possible that the two systems may act sequentially with the mono-NEDDylation preceding the more dynamic multiubiquitylation process. According to an alternative mechanism, NEDDylation can result in a reduced affinity of Rbx1 for the Ubc12 compared to the affinity the Rbx1 has to the ubiquitin E2.
Rbx1 and regulation of VBC-Cul2 ligase complex
Notably, we found that the co-expression of Rbx1 WT and all of the D97 mutants increases the levels of Cul2, Cul3 and Cul4B (Fig. 2). Since this phenomenon is independent of the NEDDylation of these cullins, this suggests that Rbx1 may play additional roles in cullin-based complexes.
We show that Rbx1 plays a role in stabilizing the cellular VBC-Cul2 complex (Figs 4 and 5). Our results show that in contrast to Rbx1 WT, Rbx1 D97R that does not support Cul2 NEDDylation appear to lower the amount of Elo B and Elo C components of the VBC complex by destabilizing them in mammalian cells (Fig. 4). Moreover, pVHL as well as Elo B and Elo C were destabilized by the presence of Rbx1 mutants defective for Cul2 NEDDylation in insect cells (Fig. 5). It is well established that the amount of proteins expressed in baculovirus-infected insect cells is much more than that in transfected 293T cells, which, we suspect, may result in the stronger effect on destabilization of pVHL by Rbx1 D97R in insect cells than in 293T cells. Alternatively, it is because most of pVHL expressed in 293T cells seems to be a monomer (data not shown) and Rbx1 D97R expression has no influence on the amount of pVHL. The amount of the Rbx1 mutants defective for NEDDylation in the lysates is lower than that of Rbx1 supporting NEDDylation (Figs 4A, 5A, 6A and 7). Therefore, it is possible that Rbx1 mutants defective for Cul2 NEDDylation could not stabilize VBC-Cul2 because of the low expression. However, we suspect that it is not the case. The amount of Rbx1 D97R, which is defective to support Cul2 NEDDylation, bound to Cul2 is almost identical to that of Rbx1 WT because the amount of Rbx1 D97R co-immunoprecipitated with Cul2 is almost equal to that of Rbx1 WT although the amount of the former in cells is lower than that of the latter (Fig. 5B). We suspect that Rbx1 mutations rendered the protein unstable when Rbx1 did not bind to Cul2, which result in the low level of Rbx1 mutants defective for Cul2 NEDDylation. It should be noted that the amounts of pVHL, Elo B and Elo C are smaller in cells co-infected with Rbx1 than in those without Rbx1 (Figs 5B and 7). The precise mechanism underlying the decrease of VBC complex by the presence of Rbx1 is currently unknown. However, it might result in low expression of VBC complex in cells co-infected with Rbx1 since the titer of the infected baculoviruses is higher in cells co-infected with Rbx1 than in those without Rbx1. Alternatively, Rbx1 may have some roles in down-regulating the VBC complex not to bind to Cul2. Further analyses will be needed to clarify it.
Collectively, the present results strongly suggest that Rbx1 D97R destabilizes the VBC complex, which results in the decrease of the amount of VBC-Cul2 ligase in cells. Moreover, although the destabilization of VBC-Cul2 is associated with NEDDylation-deficient Rbx1 mutants, the lack of Cul2 NEDDylation does not appear to play a role in this disruptive effect (Fig. 7). As shown in Fig. 6, we also found that the Rbx1 mutants that destabilize VBC-Cul2 have no effect on the stability of the SCF complex. Therefore, Rbx1 plays a different role in the regulation of SCF complexes compared to that it plays in regulating VBC-Cul2. We speculate that the binding of the Rbx1 mutants to Cul2 destabilizes the binding between VBC and Cul2, and that this results in the degradation of the VBC complex. Then, how does Rbx1 control the binding between VBC and Cul2? Rbx1 may play several potential roles in the stabilization of VBC-Cul2. First, the binding of Rbx1 mutants to Cul2 may directly affect the VBC binding to Cul2, which results in the destabilization of the VBC complex. It has shown that, however, Rbx1 binds to C-terminal domain of cullins, but substrate recognition complexes, namely, VBC and Skp1-F box complexes, bind to N-terminal domain of the proteins, and that cullins are rigid proteins (Stebbins et al. 1999; Zheng et al. 2002b). Therefore, it is unlikely that Rbx1 directly affects the binding between VBC and Cul2. The recent identification of Cul1 binding protein, CAND1 (Liu et al. 2002; Zheng et al. 2002a) may have shed light on the mechanism. CAND1 inhibits SCF complex formation by competing with the Skp1-F-box complex for binding to Cul1. Rbx1 can bind to CAND1-bound Cul1. However, Cul1 is not NEDDylated in the Cul1-Rbx1-CAND1 complex, as this requires the recruitment of the Skp1-F-box protein complex to Cul1. It is the binding of the Skp1-F-box protein complex that induces the NEDDylation of Cul1, which in turn releases CAND1 from the complex. Thus, CAND1 plays a central role in the regulation of SCF formation and its E3 activity. Since CAND1 cannot bind to Cul2, or its binding to Cul2 is very weak (Liu et al. 2002; our unpublished observation), it is unlikely that CAND1 plays a role in the regulation of VBC-Cul2. Instead, Rbx1 binding to Cul2 might regulate the binding of an unidentified regulatory protein to Cul2, which is crucial for the maintenance of VBC-Cul2. Rbx1 mutants defective for NEDDylation may not be able to regulate the binding of the unidentified protein to Cul2, possibly because the Rbx1 mutants cannot NEDDylate the protein, then destabilizes the VBC-Cul2 complex. TIP120B/CAND2, a homolog of CAND1, might be a candidate of the unidentified regulatory protein although its expression is rather restricted in muscle and heart (Aoki et al. 1999). Alternatively, structural changes in Rbx1 induced by D97 mutations inhibit the binding of VBC to Cul2 by unknown mechanisms. Further analyses will clarify the precise mechanism underlying destabilization of VBC-Cul2 by these Rbx1 mutants.
As mentioned above, we speculate that the binding of the Rbx1 mutants to Cul2 destabilizes the binding between VBC and Cul2, and that this results in the degradation of the VBC complex. Supporting this hypothesis is the observation that pVHL is degraded rapidly when it does not complex with Elo B/C (Schoenfeld et al. 2000), and that pVHL is ubiquitylated by Rbx1 (Kamura et al. 2002). We have extended the observation and demonstrated that co-expression of Cul2 along with the destabilizing Rbx1 mutants not only results in accelerated degradation of pVHL but also of Elo B/C. However, the mechanism that underlies the destabilization of VBC demonstrated here may differ from that involved in the degradation of pVHL. That is because we have found that the destabilizing Rbx1 D97R mutant does not exhibit the E3 activity (Fig. 3).
Rbx1 regulates VBC-Cul2 in multiple ways. First, it acts as an important component of the E3 complex by recruiting the E2 component. Second, it functions as a center of NEDDylation. Here we show that a third role of Rbx1 can be its involvement in the maintenance of VBC-Cul2. Since proteins targeted for ubiquitylation are recognized by the pVHL component of the VBC-Cul2 complex (Lisztwan et al. 1999; Iwai et al. 1999), this function of Rbx1 may be important in ubiquitylation in that it maintains an intact and complexed VBC-Cul2 ligase that can target the protein substrates. Thus, considering the different roles Rbx1 plays in SCF and VBC-Cul2, it is possible that it regulates the E3 activity of the two complexes via different mechanisms.
Generation of plasmids and recombinant baculoviruses
Human Cul3, Cul4A, Cul4B, Cul5 cDNAs and Rbx1 C75A/H77A mutant cDNA were described (Hori et al. 1999; Kawakami et al. 2001). Human Cul1, Cul2 and GST-NEDD8 cDNAs were provided by Dr Arnim Pause (McGill University, Canada). Rbx1 mutants carrying mutations at D97 and the Cul2 K689R mutant were generated by two-step PCR as described (Iwai et al. 1995). The sequences of all the mutants were verified by using the ABI 3100 autosequencer (Applied Biosystems). cDNAs encoding myc-Cul2, HA-VHL, HSV-Elo B, FLAG-Elo C and T7-Rbx1were subcloned into pcDNA 3.1 vector (Invitrogen). cDNAs encoding Rbx1 mutants containing N-terminal T7 tag, GST-NEDD8 and myc-Cul1, Cul2, Cul2K689R, Cul3, Cul4A, Cul4B and Cul5 were subcloned into pVL1393 vector (Invitrogen). Recombinant baculoviruses encoding those cDNAs were generated by using the Bac-PAK6 baculovirus expression system (Clontech). HPC4-Elo B, HSV-Elo C and FLAG-pVHL baculoviruses were provided by Dr Joan Conaway (Stowers Institute for Medical Research, Kansas City, USA) (Kamura et al. 1998). Baculoviruses encoding T7-Rbx1, (His)6-βTrCP1, FLAG-Skp1, (His)6-E1 (His)6-APP-BP1 and T7-Uba3 were described (Iwai et al. 1999; Kawakami et al. 2001).
Anti-HSV and anti-T7 were purchased from Novagen. Anti-Flag M2, anti-RGS-(His)6, anti-MBP and anti-myc antibodies were purchased from Sigma, Qiagen, Santa Cruz, CA, USA and Convance, respectively. Anti-IRP2 antibody was described (Ashizuka et al. 2002). Anti-HPC4 antibody was provided by Dr Arnim Pause.
Expression of recombinant proteins in Hi Five insect cells or bacterial cells
Hi Five cells cultured in Grace's medium with 10% fetal calf serum (FCS) at 27 °C were infected with the appropriate baculoviruses as indicated in the figure legends and as described (Iwai et al. 1999). The cells were collected 60 h after infection and lyzed in ice cold buffer containing 50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1 mm DTT, 1% Triton X-100, and 2 mm PMSF. The lysates were centrifuged at 15 000 g for 20 min at 4 °C, and the supernatants were used for immunoblotting, immunoprecipitation or purification of recombinant proteins. VBC-Cul2 containing WT or Rbx1 mutants were purified as described (Iwai et al. 1999). E1 and APP-BP1/Uba3 were purified by Ni-NTA affinity chromatography.
pMAL-c2X-HIF2α-ODD was generated by subcloning the DNA fragment encoding the oxygen-dependent degradation (ODD) domain of HIF-2α (aa 404–572) into pMAL-c2X (New England Biolab). pT7-7-(His)6-EGLN3 was generated by subcloning human EGLN3 cDNA amplified by RT-PCR from HeLa cells into pT7-7-(His)6. pET28b-NEDD8, pT7-7-(His)6-Ubc12, pT7-7-(His)6-UbcH5c were described (Iwai et al. 1999; Kawakami et al. 2001), and pGEX-ubiquitin was kindly provided by Dr Peter Howley (Harvard Medical School, Boston, MA, USA).
BL21(DE3) transformed with the appropriate expression plasmids were cultured in 2xYT medium. IPTG (400 µm) was added at A600 of 0.6, and cells were harvested following 2 h of induction. UbcH5c and Ubc12, GST-ubiquitin, and MBP-ODD proteins were purified by Ni-NTA, glutathione, or maltose resin affinity chromatography, respectively.
Cell culture and transfection to cells
293T cells were cultured in DMEM supplemented with 10% FCS, 100 IU/mL penicillin G and 100 µg/mL streptomycin. Transient transfection to 293T cells was performed by using Lipofectamine 2000 according to the manufacturer's (Invitrogen) instructions. The cells were harvested 36 h after transfection and lyzed in ice cold buffer containing 50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1 mm DTT, 1% Triton X-100, and 2 mm PMSF. The lysates were centrifuged at 15 000 g for 20 min at 4 °C, and the supernatants were used for immunoblotting.
Lysates from Hi Five cells infected with the indicated baculoviruses were incubated with anti-myc antibody for 2 h on ice, followed by 1 h incubation with Protein A Sepharose beads (Amersham Bioscience) at 4 °C. The beads were washed 6 times with a buffer containing 50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1% Triton X-100, followed by two washes with 20 mm Tris-HCl pH 7.5.
Lysates and immunoprecipitates from Hi Five cells were resolved via SDS-PAGE and transferred to PVDF membranes (Pall). After blocking in a buffer containing 5% skim milk, PBS, and 0.1% Tween 20, the membranes were incubated with the indicated antibodies followed by incubation with the appropriate horseradish peroxidase-conjugated second antibodies (Amersham Bioscience). The proteins were visualized following treatment of the membranes with the SuperSignal West Pico chemiluminescent detection system (Pierce) according to the manufacturer's instructions.
Half-lives of the proteins were determined as follows: cycloheximide (CHX; 100 µg/mL) was added to either cultured Hi Five cells infected with the appropriate baculoviruses 64 h following infection or 293T cells 36 h after transfection. Cells were collected at the indicated time points and the lysates were resolved via SDS-PAGE. The membranes were probed with appropriate antibodies to detect Skp1, Elo B, Elo C or cullins. Signals obtained by immunoblotting were quantified by the LAS3000 luminescent image analyzer (Fuji Film).
In vitro ubiquitin conjugation assay
The conjugation reaction mixture contained (in a final volume of 20 µL) 200 ng of MBP-ODD, 100 ng of E1, 200 ng of E2 (UbcH5c), 20 mm Tris-HCl pH 7.5, 5 mm MgCl2, 2 mm DTT, 2 mm 2-oxoglutarate, 2 mm ascorbic acid, 5 µg of GST-ubiquitin, and 800 ng of VBC-Cul2 containing either WT or mutant Rbx1, in the presence or absence of 1 µg of EGLN3 as indicated. 50 ng of APP-BP1/Uba3, 160 ng of Ubc12 and 1 µg of NEDD8 were added to the reactions indicated to contain the NEDD8 system. The reaction mixtures were incubated in the presence of ATP and ATP-regenerating system (0.5 mm ATP, 10 mm creatine phosphate, 10 µg of creatine phosphokinase) for 120 min at 37 °C. Reactions were terminated by adding SDS-sample buffer and resolved on a 6% SDS-PAGE followed by immunoblotting with anti-MBP antibody.
We thank P. Howley, J. Conaway and A. Pause for reagents. This work was partly supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan to K.I.