Autoinhibitory regulation of SCF-mediated ubiquitination by human cullin 1's C-terminal tail

Abstract

SCF (Skp1·CUL1·F-box protein·ROC1) E3 ubiquitin ligase and Cdc34 E2-conjugating enzyme catalyze polyubiquitination in a precisely regulated fashion. Here, we describe biochemical evidence suggesting an autoinhibitory role played by the human CUL1 ECTD (extreme C-terminal domain; spanning the C-terminal 50 amino acids), a region that is predicted to contact the ROC1 RING finger protein by structural studies. We showed that ECTD did not contribute to CUL1's stable association with ROC1. Remarkably, deletion of ECTD, or missense mutations designed to disrupt the predicted ECTD·ROC1 interaction, markedly increased the ability of SCF^βTrCP2 to promote IκBα polyubiquitination and polyubiquitin chain assembly by Cdc34 in vitro. Thus, disruption of ECTD yields in vitro effects that parallel SCF activation by Nedd8 conjugation to CUL1. We propose that SCF may be subject to autoinhibitory regulation, in which Nedd8 conjugation acts as a molecular switch to drive the E3 into an active state by diminishing the inhibitory ECTD·ROC1 interaction.

The SCF (Skp1·CUL1·F-box protein·ROC1) E3 ubiquitin (Ub) ligase defines a large class of evolutionally conserved four-subunit protein complexes that mediate timely, proteolytic destruction of a wide array of protein substrates, thereby regulating nearly all aspects of cellular and organismal processes (1). SCF is composed of an invariable core that contains Skp1, CUL1, and ROC1 (also named Rbx1 or Hrt1), and one interchangeable subunit belonging to the F-box protein family that consists of >70 members in humans. In SCF, CUL1 functions as a scaffold anchoring simultaneously an Skp1·F-box protein heterodimer and the ROC1 RING finger protein, hence positioning the substrate-bound F-box protein within the proximity of ROC1 that recruits an E2 Ub-conjugating enzyme. By mechanisms yet to be elucidated, the bound substrate, SCF and the E2 engage in concerted interactions that commit catalysis, yielding a covalent linkage connecting a polyubiquitin chain to a substrate lysine residue, which signals for proteolysis.

Initially, human SCF was noted to possess an intrinsic activity that stimulates the Cdc34 E2-conjugating enzyme for the synthesis of unanchored Ub chains (2, 3). Subsequent work has determined that this Ub ligase activity is mediated by ROC1 complexed with the C-terminal half of human CUL1 in a manner that strictly depends on the integrity of the RING finger domain (4, 5). Moreover, conjugation of Nedd8, an Ub-like molecule, to human CUL1 at K720, markedly enhanced the capability of Cdc34 to assemble substrate-free Ub chain (6), and to synthesize a di-Ub species (7, 8). Consistent with human reconstitution studies, Petroski and Deshaies (9) have demonstrated that in budding yeast SCF accelerates the discharge of Ub from Cdc34 and increases the rate of di-Ub formation. Taken together, it appears that the complex interactions between ROC1–CUL1–Nedd8 and an E2 (such as Cdc34) assemble an Ub polymerase activity. All SCF-related Cullin–RING ligases were found to contain Ub polymerization activity (3, 10, 11), suggesting a common mechanism used by these E3s for Ub catalysis.

In this study, we have uncovered an autoinhibitory activity by the CUL1 ECTD (extreme C-terminal domain; spanning amino acids 727–776) and propose that Nedd8 may exert its stimulatory role by reversing the ECTD-mediated inhibition.

Results

Reconstitution of Nedd8-Assisted Ubiquitination of IκBα by SCF^βTrCP2 and Cdc34.

We sought to examine the influence of neddylation on the assembly of polyubiquitin chains onto IκBα by human Cdc34 E2-conjugating enzyme, because its counterpart in budding yeast plays an indispensable role in supporting SCF-mediated proteolysis (12). Our reconstituted in vitro ubiquitination system included GST-IκBα(1–54) phosphorylated by IKKβ^S177E,S181E as the substrate, Ub, E1, human Cdc34, and the four-subunit SCF^βTrCP2 E3 complex. To evaluate the role of neddylation, we used two experimental approaches, with the first procedure involving the initial assembly of the SCF^βTrCP2–IκBα complex, followed by treatment with the neddylation agents and then, the addition of ubiquitination enzymes (Fig. 1A, approach 1). As shown, SCF^βTrCP2, containing CUL1 with ≈10% in neddylated form as revealed by immunoblot analysis (Fig. 1B Lower, lanes 2–6), supported the ubiquitination of ³²P-GST-IκBα(1–54), producing laddered forms of substrate-Ub conjugates (Fig. 1B Upper, lanes 2–5), a majority of which ranged in size from 100 to 250 KDa (Fig. 1B Upper, lane 5). However, incubation of SCF^βTrCP2-IκBα with neddylation agents converted the majority of CUL1 from unmodified to neddylated forms (Fig. 1B Lower, lanes 7–11). The resulting hyperneddylated SCF^βTrCP2 produced a heterogeneous population of ³²P-GST-IκBα(1–54)-Ub conjugates of molecular mass from 150 to far beyond 250 kDa, considerably longer than those formed with the underneddylated E3 (Fig. 1B Upper, compare lanes 4 and 5 with lanes 9 and 10). Notably, the extensive ubiquitination required both neddylation and a high concentration of Cdc34 (Fig. 1B Upper, lanes 9 and 10). Unexpectedly, this neddylation procedure also formed a ³²P-GST-IκBα(1–54) conjugate independently of Cdc34 (Fig. 1B Upper, lanes 7, 8, and 11), which matched the size of mononeddylated substrate. Intriguingly, this mononeddylated species disappeared in the presence of high levels of Cdc34 (Fig. 1B upper, lanes 9 and 10), suggesting that Cdc34 was able to use this modified substrate for Ub chain assembly in vitro. On a cautious note, although our observations suggest enzymatic capability by human Cdc34 in assembling Ub chains on Nedd8, its biological implication is not understood at the present time.

Fig. 1.

Neddylation stimulates SCF^βTrCP2/Cdc34-mediated ubiquitination of IκBα. (A) A scheme of experimental procedures used for the in vitro ubiquitination of IκBα by SCF^βTrCP2. (B and C) Reconstitution of Nedd8-assisted ubiquitination of GST-IκBα(1–54) by SCF^βTrCP2 and Cdc34 using approach 1 or 2, respectively. Ubiquitination was carried out as described in SI Text. Both autoradiogram and immunoblots with indicated antibodies are shown. The numbers in the middle of the autoradiogram denotes the suggested number of Ub moieties conjugated to the substrate. * marks the conjugate formed by the substrate and Nedd8. Note the multiple CUL1-Nedd8 species shown in B. As revealed by recent mass spectrometry studies, Nedd8 K11, K22, K48, and K60 can form chains in vivo, whereas K22 and K48 can be neddylated in vitro (25). The biological significance of the Nedd8 chains remains to be elucidated.

To evaluate IκBα polyubiquitination exclusively, we used approach 2 to eliminate substrate mononeddylation, by treating SCF^βTrCP2 with neddylation agents first, and then washing the complex before the addition of substrate (Fig. 1A, approach 2). Under this condition, mononeddylation of ³²P-GST-IκBα(1–54) did not occur (Fig. 1C Upper, lane 3). Nonetheless, the hyperneddylated SCF^βTrCP2 was found to be more effective than the underneddylated E3 in producing Ub conjugates with molecular mass far >250 kDa (Fig. 1C Upper, compare lanes 2 and 3). Collectively, these findings demonstrate a prominent role for neddylation in promoting extensive ubiquitination of IκBα by human Cdc34.

In Vitro Activation of SCF by Deletion of Human CUL1 ECTD.

We investigated the role of human ECTD (Fig. 2A) in SCF-mediated ubiquitination. ECTD defines a highly conserved region spanning residues 727–776, comprising α-helices H30 and H31 and β-strands S10 and S11 within the CUL1 winged-helix B domain (13). As revealed by crystallographic analysis, the human CUL1 ECTD forms extensive contacts with the ROC1 RING finger protein (13), which involve ROC1 H3 helix and S2 strand, as well as CUL1 H31 helix and S10 strand (Fig. 2B). Specifically, CUL1 Y761 and E758 form hydrogen bonds with ROC1 K89, S85, and T64. Significantly, substitution of yeast ROC1 at the corresponding K89 and R91 positions with alanine failed to rescue ΔRbx1/ROC1 yeast strain (13), underscoring a role for these residues in mediating interactions of critical functional significance.

Fig. 2.

The human CUL1 ECTD. (A) A schematic representation of domains of the human CUL1. ECTD comprises C-terminal amino acids 727–776 (downstream of the neddylation site (K720), with secondary structures as previously determined (13), and with identical residues among CUL1 homologs and paralogs colored in red and conserved amino acids in purple. (B) The interface structure between the CUL1 ECTD H31 helix and S10 β-strand (green), as well as the ROC1/Rbx1 H3 helix (red). This structure was generated based on the crystal structure of the CUL1–Rbx1–Skp1–F-box^Skp2 complex (13), as described in SI Text. Residues involved in direct interface contact are indicated, with white dotted lines denoting hydrogen bonds and salt bridges.

Coexpression of CUL1^324–728, a human CUL1 ECTD deletion mutant, with ROC1 in bacteria formed a complex with efficiency comparable to that of the WT [supporting information (SI) Fig. S1A], demonstrating that ECTD does not contribute to CUL1's stable association with ROC1. However, ECTD deletion rendered CUL1 deficient in neddylation in vitro (Fig. S1B) and in vivo (Fig. S1C), which is in complete agreement with the previous observations with Cdc53 (14) and human CUL1 (15).

To determine the effects of ECTD deletion in the ubiquitination of IκBα, we isolated SCF^βTrCP2 or the mutant E3 containing CUL1^1–728, which were assembled in FCHT293 and FC728HT293 cells, respectively, by using a double-affinity purification approach as described in SI Text. Immunoblot analysis revealed comparable amounts of HA-βTrCP2 and Flag–CUL1 or Flag–CUL1^1–728, associated with SCF^βTrCP2 and SCF^βTrCP2 (CUL1^1–728), respectively (Fig. 3A Lower). Whereas ≈10% of CUL1 was found neddylated (Fig. 3A Lower, lanes 3 and 4), CUL1^1–728 was unmodified exclusively (Fig. 3A Lower, lanes 5–8), as expected.

Fig. 3.

Deletion of the CUL1 ECTD activates the ubiquitination of IκBα by SCF^βTrCP2/Cdc34. (A) Titration. Increasing amounts of SCF^βTrCP2 and SCF^βTrCP2 (CUL1^1–728), prepared as described in SI Text, were compared for their ability to support the ubiquitination of IκBα with Cdc34 (90 pmol). The reaction was analyzed by autoradiogram (Upper), to visualize ³²P-GST–IκBα (1–54)–Ub conjugates, and by immunoblot (Lower), to reveal the levels of HA–βTrCP2 and Flag-tagged CUL1 or mutant in each E3 complex used. (B) Reaction kinetics. The amounts of SCF^βTrCP2 and SCF^βTrCP2 (CUL1^1–728) used were identical to those used in reactions shown in A, lanes 2 and 6, respectively.

Surprisingly, SCF^βTrCP2 (CUL1^1–728) was significantly more active than the underneddylated WT E3 complex in mediating the ubiquitination of ³²P-GST–IκBα(1–54) (Fig. 3A Upper, lanes 6 and 7) in a manner that depends on Cdc34 (Fig. 3A Upper, lane 8). Quantitation analysis using a phosphorimager revealed that SCF^βTrCP2 (CUL1^1–728) yielded GST–IκBα(1–54)–Ub conjugates of >250 kDa in amounts about four times higher than those observed with the WT E3. Furthermore, kinetic analysis showed that SCF^βTrCP2 (CUL1^1–728) accumulated more high-molecular-mass ³²P-GST–IκBα(1–54)-Ub species at 15 min than SCF^βTrCP2 did after 90 min of incubation (Fig. 3B, compare lanes 3 and 4). Thus, SCF^βTrCP2 (CUL1^1–728) resembled the WT E3 complex containing hyperneddylated CUL1, in producing extensively ubiquitinated products (compare Fig. 3A, lanes 6 and 7, and Fig. 3B, lanes 4–6, with Fig. 1B, lanes 9 and 10, and Fig. 1C, lane 3). Taken together, these results strongly suggest that ECTD deletion enhances the ability of SCF^βTrCP2 to promote the ubiquitination of IκBα.

Next, we examined the influence of ECTD truncation in Ub chain assembly by the ROC1–CUL1 subcomplex. Kinetic analysis revealed that ROC1–CUL1^324–728, deleted of the CUL1 ECTD, exhibited much higher Ub polymerization activity than the unmodified WT complex (Fig. 4A, lanes 1–14), stimulating the production of free Ub chains (di-Ub or tri-Ub) as well as polyubiquitinated Cdc34 by 4- to 5-fold (Fig. 4A, graphs). In agreement with previous studies (6, 7), neddylation of CUL1^324–776 markedly enhanced polyubiquitin chain assembly (Fig. 4A, compare lanes 17 and 18). Importantly, the activity by the unmodified ROC1–CUL1^324–728 was similar to that observed with ROC1–CUL1^324–776–Nedd8, as judged by similar production of di-Ub, tri-Ub, and polyubiquitinated Cdc34 by these two complexes (Fig. 4A, compare lanes 18 and 19). As shown, addition of neddylation components did not stimulate ubiquitination by the ECTD-truncated complex (Fig. 4A, compare lanes 19 and 20). In addition, ROC1–CUL1^324–728 was found to be more active than ROC1–CUL1^324–776 to assemble Ub chains in the range of Cdc34 concentrations examined (Fig. S2). Altogether, these data demonstrated that removal of the CUL1 ECTD stimulated human Cdc34 for the synthesis of polyubiquitin chains.

Fig. 4.

Truncation of the CUL1 ECTD promotes polyubiquitin chain assembly catalyzed by human Cdc34. (A) Polyubiquitin chain assembly. The Ub polymerization assay was carried out as described in SI Text. A time course experiment is shown (lanes 1–14), using ROC1-CUL1^324–776 and ROC1-CUL1^324–728 (30 pmol each). Lane 13 omitted the ROC1–CUL1 complex. Lane 14 contained Ub and E1 only. The reaction products were analyzed by autoradiography and quantitated by PhosphoImager, with the synthesis of Ub₂, Ub₃, or Cdc34-Ub_n, presented graphically. The effects of neddylation were examined in reactions shown in lanes 15–21. In this case, ROC1-CUL1^324–776 or ROC1-CUL1^324–728 was treated with neddylation agents as described in SI Text, followed by the addition of E1, Cdc34, and ³²P-Ub to initiate polyubiquitin chain assembly reaction. The incubation time was 90 min. Similar levels of CUL1^324–776 and CUL1^324–728 were used (Fig. S6A). (B) di-Ub synthesis. ROC1-CUL1^324–776 and ROC1-CUL1^324–728 (30 pmol each) were compared for their ability to support Cdc34-catalyzed di-Ub synthesis, measured as described in SI Text. The reaction was incubated for 60 min at 37°C with Cdc34 (30 pmol). Formation of di-Ub was quantitated and graphically presented. Similar levels of CUL1^324–776 and CUL1^324–728 were used (Fig. S6B).

To characterize precisely the effect of ECTD on Ub ligation by Cdc34, we measured the synthesis of di-Ub species formed via the ligation of ³²P-labeled Ub^K48R to bovine Ub, as described (7). Consistent with previous observations (7), in the absence of Nedd8, ROC1–CUL1^324–776 supported the formation of di-Ub poorly (Fig. 4B, compare lane 2 with lanes 3–5). In contrast, ROC1–CUL1^324–728 markedly increased di-Ub synthesis in a concentration dependent manner (Fig. 4B, lanes 6–8, and graph). Taken together, these results concluded that the deletion of ECTD promoted extensive ubiquitination of IκBα by enhancing the ability of human Cdc34 to catalyze the assembly of polyubiquitin chains. Hence, the ECTD truncation appeared to phenocopy neddylation in stimulating ubiquitination.

Disruption of the Human CUL1 ECTD·ROC1 Interface Activates SCF in Vitro.

Given that the human CUL1 ECTD is dispensable for stable association with ROC1 (Fig. S1A) but its removal activated ubiquitination (Figs. 3 and 4), we postulated that the ECTD·ROC1 interface might play an autoinhibitory role. To test this hypothesis, we sought to disrupt the predicted ECTD·ROC1 interface by creating alanine substitution at CUL1 Y761 or ROC1 K89. To avoid neddylation-mediated effects, we generated Flag–CUL1^K720R,Y761A bearing both K720R and Y761A mutations. Using procedures described in SI Text, varying amounts of SCF^βTrCP2, SCF^βTrCP2 (CUL1^K720R), or SCF^βTrCP2 (CUL1^K720R,Y761A) were affinity-purified, with each set containing comparable amounts of HA–βTrCP2 and Flag–CUL1 or mutants, as confirmed by immunoblot analysis (Fig. 5A). As expected, whereas the WT Flag–CUL1 contained ≈10% neddylated form (Fig. 5A Upper, lanes 1 and 2), no detectable neddylation was observed with CUL1^K720R or CUL1^K720R,Y761A (Fig. 5A Upper, lanes 3–6). SCF^βTrCP2 supported the ubiquitination of ³²P-GST–IκBα(1–54) more effectively than SCF^βTrCP2 (CUL1^K720R) (Fig. 5B, compare lanes 1 and 2 with lanes 8–10), in keeping with a stimulating role for neddylation as demonstrated (16–18). Importantly, SCF^βTrCP2 (CUL1^K720R,Y761A) supported the formation of high-molecular-mass ³²P-GST–IκBα(1–54)–Ub conjugates more effectively than SCF^βTrCP2 (CUL1^K720R) (Fig. 5B, compare lanes 4–6 with lanes 8–10), in levels similar to the WT E3, especially when low levels of SCF were used (Fig. 5B, graph). Thus, introduction of a single Y761A substitution into CUL1^K720R was able to reverse the inhibition caused by deficient neddylation.

Fig. 5.

Activation of SCF^βTrCP2-mediated ubiquitination of IκBα by CUL1 Y761A substitution. (A) Immunoblot analysis. SCF^βTrCP2, SCF^βTrCP2 (CUL1^K720R), or SCF^βTrCP2 (CUL1^K720R/Y761A) was prepared as described in SI Text. The levels of HA–βTrCP2 and Flag–CUL1 or mutant (Flag–CUL1^K720R or Flag–CUL1^K720R/Y761A) in isolated E3s were determined by immunoblot analysis and quantitated by the Odyssey infrared imaging system (Licor). As revealed by the numbers displayed, each immuno-purified complexes contained comparable levels of the SCF subunits. In addition, quantitative immunoblot analysis indicated that anti-HA immunoprecipitation copurified predominantly Flag–CUL1 or mutant (with the ratio of recombinant against the endogenous >4:1; data not shown). (B) Autoradiographic analysis of in vitro ubiquitination of IκBα. The amount of purified SCF^βTrCP2 used in lanes 1 and 2, SCF^βTrCP2 (CUL1^K720R/Y761A) in lanes 5 and 6, and SCF^βTrCP2 (CUL1^K720R) in lanes 9 and 10, were identical to those in Fig. 5A, lanes 1–6. The amount of SCF^βTrCP2 (CUL1^K720R/Y761A) used in lane 4, and SCF^βTrCP2 (CUL1^K720R) in lane 8, was 3-fold less than that in lane 5 and lane 9, respectively. The production of GST–IκBα(1–54)–Ub conjugates of >150 kDa by the three SCF^βTrCP2 complexes was quantitated by PhosphorImager, and the ratio between the WT or the K72OR/Y761A mutant against K72OR was presented graphically. (C) Effects of the CUL1 Y761A mutation in polyubiquitin chain assembly. ROC1–CUL1^324–776 and ROC1–CUL1 (Y761A)^324–776 (30 pmol each) were compared for their ability to support Cdc34-catalyzed polyubiquitin chain assembly in the presence or absence of neddylation, as described in Fig. 4A. (D) A model for role of the CUL1 ECTD in autoinhibitory regulation of SCF. In the absence of Nedd8, the CUL1 ECTD makes contacts with ROC1, forming a restrained/inactive conformation. It is proposed that neddylation may disrupt the ECTD–ROC1 interaction to create an active state. The arrow in black denotes the N terminus of ROC1 that establishes primary interface interactions with CUL1 (13).

Next, we examined the effects of the CUL1 Y761A or ROC1 K89A mutation in the assembly of Ub chain by human Cdc34. Supporting our prediction, the purified ROC1–CUL1 (Y761A)^324–776 complex was significantly more active than the WT complex in polyubiquitin chain assembly by Cdc34 (Fig. 5C Upper, compare lanes 3 and 4 with lanes 5 and 6). As shown, CUL1 (Y761A)^324–776 was efficiently neddylated (Fig. 5C Bottom, compare lanes 7 and 8). However, neddylation of CUL1 (Y761A)^324–776 did not further increase the Ub polymerization activity (Fig. 5C Upper, compare lanes 6 and 8). Moreover, the activity by ROC1–CUL1 (Y761A)^324–776 was similar to that observed with ROC1–CUL1^324–776–Nedd8 (Fig. 5C, compare lanes 6 and 8). Similar effects were also observed with ROC1–CUL1 (Y761A)^324–776 and ROC1–CUL1^324–776 at lower concentrations in the presence or absence of neddylation (Fig. S3). In addition, the reaction kinetics of polyubiquitin chain assembly by ROC1–CUL1 (Y761A)^324–776 paralleled that by the neddylated WT complex (data not shown). Similar stimulatory effects were observed with the ROC1 K89A substitution (Fig. S4). Furthermore, both ROC1–CUL1 (Y761A)^324–776 and ROC1 (K89A)–CUL1^324–776 supported the synthesis of di-Ub more effectively than the WT complex (Fig. S5).

Altogether, these results demonstrated that single missense mutations, created to disrupt the predicted ECTD·ROC1 interface, enhanced the ability of SCF^βTrCP2 and ROC1–CUL1 to support substrate ubiquitination and polyubiquitin chain assembly, respectively. These effects resembled those observed with the ECTD deletion mutant (Figs. 3 and 4), thus supporting the hypothesis that the CUL1 ECTD·ROC1 interaction inhibits ubiquitination catalyzed by human Cdc34. Of note, the extent of the activation by single missense mutations was less robust than that seen with the ECTD deletion mutant, most likely caused by incomplete disruption of the CUL1 ECTD·ROC1 interface.

Discussion

Previously, the role of CUL1 has been defined as a scaffold that tethers both Skp1 and ROC1 and provides a site (K720 in human CUL1) for Nedd8 conjugation, which leads to activated ubiquitination. In this work, we have provided compelling biochemical evidence suggesting an autoinhibitory role played by the human CUL1 ECTD (Fig. 2A) in polyubiquitination catalyzed by Cdc34. We showed that: (i) ECTD did not contribute to the stable association between CUL1 and ROC1 (Fig. S1A); (ii) ECTD deletion markedly increased the ability of SCF^βTrCP2 to promote the in vitro ubiquitination of IκBα (Fig. 3) and polyubiquitin chain assembly (Fig. 4); and (iii) single mutations that were designed to disrupt the predicted ECTD·ROC1 interaction, such as CUL1 Y761A and ROC1 K89A, activated SCF^βTrCP2 for IκBα polyubiquitination and the assembly of polyubiquitin chains (Fig. 5 A–C and Figs. S4 and S5).

These findings led us to hypothesize that the human CUL1 ECTD acts as an allosteric module to regulate Cdc34-catalyzed polyubiquitin chain assembly. In this model, the interface between ECTD and ROC1 (Fig. 2B) represents a “restrained” state that suppresses Cdc34 activity. Disruption of the ECTD module through deletion or missense mutation converts this “latent” conformation into an active form. Previous studies have suggested allosteric regulation of an E3 activity, with examples including activation of the Ubr1 E3 by dipeptides (19), and C2 domain-mediated autoinhibition of Smurf2 HECT E3 ligase (20). Allosteric regulation of an enzymatic activity is well documented with studies of c-Src, whereby the kinase domain is kept in an inactive conformation by multiple interactions involving Src homology 3 (SH3), phospho-tyrosine-bound SH2, and the linker that connects the kinase and SH2 domains (21). The c-Src kinase can be activated by displacement of SH2/SH3 domain through competitive polypeptides or tyrosine dephosphorylation.

Moreover, our studies suggest that activation of SCF by disruption of CUL1's ECTD phenocopy Nedd8-mediated effects (Figs. 3–5), raising a possibility for a role of Nedd8 in regulating the ECTD·ROC1 interaction. Molecular modeling analysis of the ROC1–CUL1–Nedd8 docking complex (22) suggests that Nedd8 conjugation causes conformational changes that disrupt the ECTD·ROC1 interface (data not shown). As noted by previous observations (17), substrate-bound SCF contains CUL1 enriched with the neddylated form, suggesting that substrate ubiquitination and neddylation may be coordinated. In light of these findings, we propose that neddylation/deneddylation may switch SCF on/off by regulating the interactions between the human CUL1 ECTD and ROC1. While “on duty” with a substrate, SCF must be in an active neddylated form, presumably with the CUL1 ECTD dissociated from ROC1. However, once exited to an “off-duty” state, SCF may become deneddylated through the action of the COP9 signalosome, re-establishing the ECTD·ROC1 interface and turning the E3 into an inactive/restrained form (Fig. 5D).

SCF activity is regulated at multiple levels. Recent studies have revealed a role for dimerization of F-box proteins in enhancing catalytic efficiency in ubiquitination (23). In addition, it was observed that the Skp1–Skp2 complex promoted the disassociation of the cullin–CAND1 complex (24), suggesting a role for dynamic interactions among Skp1–F-box protein, cullin, and CAND1 in the assembly of active SCF E3 complexes. Autoinhibition and Nedd8-mediated activation of SCF, as suggested by the current study, provides another mechanism for fine-tuning the SCF activity in accordance with its physiological function.

The human CUL1 ECTD·ROC1 interface is established by contacts between the ROC1 H3 helix and CUL1 H31, as well as S10 (Fig. 2B). Aside from binding to the CUL1 ECTD, the ROC1 H3 helix has been predicted to participate in interactions with an E2 as well (13). On this helix, whereas residue K89 interacts with CUL1 Y761 (Fig. 2B), W87 is presumed to bind to E2 (13). Mutagenesis experiments revealed that whereas ROC1 K89A activated ubiquitination (Figs. S4 and S5), ROC1 W87A was unable to support polyubiquitin chain assembly by Cdc34 (data not shown). It can be speculated that the CUL1 ECTD·ROC1 H3 interface may present a steric hinderance that restricts Cdc34 loading or may position the ROC1 H3-bound E2 in a manner that is unfavorable to Ub transfer. Disruption of this interaction, resulted from neddylation as discussed above, would stimulate ubiquitination by promoting Cdc34 recruitment and/or by properly aligning the bound E2 in favor of ubiquitination.

Materials and Methods

This study used the following plasmid vectors: pcDNA3.1-Flag-CUL1 (4), pcDNA3.1-HA-βTrCP2 (HOS; ref. 2), and pGEX-4T3/pET-15b-GST-HA-ROC1-Flag-CUL1^324–776 (6). CUL1 mutants were generated by using a QuikChange Site-Directed Mutagenesis Kit with the mutagenesis primers indicated in SI Text. SCF^βTrCP2, ROC1-CUL1^324–776, or mutant E3 complexes were expressed and affinity-purified by the procedures described in SI Text. Several reconstituted in vitro ubiquitination assays, including ubiquitination of IκBα, neddylation, and substrate-free polyubiquitin chain assembly, were carried out as described (2, 4, 6–8), with concentrations of the components and reaction conditions as specified in SI Text.

Note added in proof.

In a recent study (26), Schulman and colleagues have solved the structure of Nedd8-CUL5 C terminus-ROC1/Rbx1, demonstrating that Nedd8 induces a drastic conformational change that abolishes the interaction between ROC1/Rbx1 and CUL5 WH-B domain (equivalent to CUL1 ECTD). This work provides compelling structural evidence supporting for a role of neddylation in regulating the interaction between the CUL1 ECTD and ROC1.