NEDD8 nucleates a multivalent cullin–RING–UBE2D ubiquitin ligation assembly

Abstract

Eukaryotic cell biology depends on cullin–RING E3 ligase (CRL)-catalysed protein ubiquitylation¹, which is tightly controlled by the modification of cullin with the ubiquitin-like protein NEDD8^2,3,4,5,6. However, how CRLs catalyse ubiquitylation, and the basis of NEDD8 activation, remain unknown. Here we report the cryo-electron microscopy structure of a chemically trapped complex that represents the ubiquitylation intermediate, in which the neddylated CRL1^β-TRCP promotes the transfer of ubiquitin from the E2 ubiquitin-conjugating enzyme UBE2D to its recruited substrate, phosphorylated IκBα. NEDD8 acts as a nexus that binds disparate cullin elements and the RING-activated ubiquitin-linked UBE2D. Local structural remodelling of NEDD8 and large-scale movements of CRL domains converge to juxtapose the substrate and the ubiquitylation active site. These findings explain how a distinctive ubiquitin-like protein alters the functions of its targets, and show how numerous NEDD8-dependent interprotein interactions and conformational changes synergistically configure a catalytic CRL architecture that is both robust, to enable rapid ubiquitylation of the substrate, and fragile, to enable the subsequent functions of cullin–RING proteins.

Main

CRLs orchestrate numerous eukaryotic processes—including transcription, signalling, cell division and differentiation—and CRL dysregulation underlies many pathologies. The activities of these enzymes depend on coordinated but dynamic interactions between dedicated cullin–RING complexes and several regulatory partner proteins¹. Cullins (CULs) 1–5 bind cognate RING-containing partners (RBX1 or RBX2) through a conserved intermolecular cullin/RBX (hereafter C/R) domain, in which a CUL β-sheet stably embeds an RBX strand⁷. On one side of the C/R domain, the CUL N-terminal domain can associate interchangeably with numerous substrate-recruiting receptors. As examples, human CUL1–RBX1 binds around 70 SKP1-F-box protein complexes and CUL4–RBX1 binds around 30 DDB1–DCAF complexes, forming the E3 enzymes CRL1^{F-box protein} or CRL4^DCAF, respectively (in which F-box protein and DCAF represent substrate receptors for a given CRL)^8,9,10,11. The C-terminal WHB domain of the CUL protein and the RING domain of the RBX protein emanate from the other side of the C/R domain. To achieve E3 ligase activity, the RING domain recruits one of several ubiquitin-carrying enzymes, which presumably use distinct mechanisms to transfer ubiquitin to receptor-bound substrates.

CRLs are regulated by reversible NEDD8 modification of a specific lysine residue within the WHB domain of CULs. Although it has approximately 60% sequence identity to ubiquitin, NEDD8 uniquely activates CRL-dependent ubiquitylation². NEDD8 has been suggested to have multiple roles in catalysis—including assisting in the recruitment of ubiquitin-carrying enzymes, facilitating juxtaposition of the substrate and ubiquitylation active site, and promoting conformational changes—although the structural mechanisms of these effects remain unknown^3,4,5,12. NEDD8 also stabilizes cellular CRLs by blocking the exchange factor CAND1 from ejecting substrate receptors from unneddylated CUL–RBX complexes¹³. Neddylation controls around 20% of ubiquitin-mediated proteolysis and presumably many nondegradative functions of ubiquitin, and an inhibitor of NEDD8 (MLN4924, also known as Pevonedistat) blocks HIV infectivity and is in clinical trials as an anticancer agent^6,14.

Here we determine structural mechanisms that underlie ubiquitylation by human neddylated CRL1^β-TRCP and E2 enzymes from the UBE2D family, in which the F-box protein β-TRCP recruits a specific phosphodegron motif in substrates including β-catenin and IκBα^{15,16,17,18,19,20}. UBE2D knockdown stabilizes the CRL1^β-TRCP substrate IκBα²¹, whereas mutations that impair the ubiquitylation of β-catenin by CRL1^β-TRCP promote tumorigenesis²². Furthermore, hijacking of CRL1^β-TRCP enables HIV to evade host immunity²³, and deamidation of Gln40 of NEDD8 by an enteropathogenic and enterohaemorrhagic Escherichia coli effector results in the accumulation of the CRL1^β-TRCP substrate IκBα as well as substrates of other CRLs^24,25,26.

NEDD8 activation of ubiquitylation

We used rapid quench-flow methods to obtain kinetic parameters for CRL1^β-TRCP and UBE2D catalysed ubiquitylation of model substrates (phosphopeptides from β-catenin and IκBα, containing single acceptor lysines). We found that NEDD8 substantially stimulates the reaction, by nearly 2,000-fold (Fig. 1a, Extended Data Fig. 1, Extended Data Table 1). Performing experiments under conditions that allow for multiple UBE2D turnover events enabled us to quantify the effects of neddylation on substrate ‘priming’, in which ubiquitin is ligated directly to the substrate, compared with ‘chain elongation’, in which it is linked to a substrate-linked ubiquitin. The individual rates for the linkage of successive ubiquitins during polyubiquitylation showed that NEDD8 activates both substrate-priming and chain-elongation reactions. However, ubiquitin ligation to a substrate is tenfold faster than ubiquitin ligation to a substrate-linked ubiquitin, suggesting that neddylated CRL1^β-TRCP—together with UBE2D—optimally catalyses substrate-priming reactions (Extended Data Table 1).

Fig. 1: Role of NEDD8 and strategy for the visualization of dynamic ubiquitin transfer from UBE2D to substrate by neddylated CRL1^β-TRCP.

a, Effect of the neddylation of CUL1 on CRL1^β-TRCP-catalysed ubiquitin transfer from UBE2D to a radiolabelled IκBα-derived peptide substrate. The plots show the proportion of substrate remaining during pre-steady-state rapid quench-flow ubiquitylation reactions with saturating UBE2D3 and either unneddylated or neddylated CRL1^β-TRCP. The symbols show the data from independent experiments (n = 2 technical replicates). b, Schematic representing substrate priming by neddylated CRL1^β-TRCP and UBE2D~Ub. The inset shows the transition state during ubiquitylation. c, Chemical mimic of the ubiquitylation intermediate, in which surrogates for the active site of UBE2D, the C terminus of ubiquitin and the ubiquitin acceptor site on the IκBα-derived substrate peptide are simultaneously linked.

Cryo-EM reveals cullin–RING dynamics

Ubiquitin transfer from a RING-docked UBE2D~Ub intermediate (in which ~ indicates a thioester bond or thioester-bond mimic, Ub indicates ubiquitin) to a substrate that is bound to an F-box protein was difficult to rationalize from previous structural models^3,7,20,27 and from our cryo-electron microscopy (cryo-EM) reconstructions of unneddylated and neddylated substrate-bound CRL1^β-TRCP (Extended Data Fig. 2, Extended Data Table 2). We observed a well-resolved ‘substrate-scaffolding module’ that resembles models based on crystal structures of substrate-bound SKP1–β-TRCP and the portion of an F-box–SKP1–CUL1–RBX1 complex that includes the N-terminal domain of CUL1 and the intermolecular C/R domain^7,20. However, for the RING domain of RBX1 and the WHB domain of CUL1—with or without covalently linked NEDD8—density is either lacking or visualized only at low contour, in varying positions in different classes. These domains apparently sample multiple orientations, and it is therefore difficult to conceptualize rapid ubiquitylation of a flexible substrate by uncoordinated nanometre-scale motions of RBX1-activated UBE2D~Ub (Extended Data Fig. 2).

Capturing CRL–substrate ubiquitylation

Substrate priming involves fleeting simultaneous linkage of the active site of UBE2D, the C terminus of ubiquitin and the substrate (Fig. 1b). We chemically linked surrogates for these entities to form a stable mimic of the transition state, which avidly binds neddylated CRL1^β-TRCP (Fig. 1c, Extended Data Fig. 3a–d). After screening several complexes by cryo-EM, we obtained a reconstruction at 3.7 Å resolution that showed our proxy for a UBE2D~Ub–IκBα substrate intermediate bound to a hyperactive version of neddylated CRL1^β-TRCP (Fig. 2a, Extended Data Figs. 3, 4, Extended Data Table 2).

Fig. 2: Cryo-EM structure representing neddylated CRL1^β-TRCP-mediated ubiquitin transfer from UBE2D to the substrate IκBα.

a, Cryo-EM density representing the neddylated CRL1^β-TRCP–UBE2D~Ub–IκBα substrate intermediate, in which UBE2D~Ub is activated and juxtaposed with the substrate. b, The substrate-scaffolding module connects β-TRCP-bound substrate to the intermolecular cullin–RBX (C/R) domain. c, The catalytic module consists of RING–UBE2D~Ub of RBX1 in the canonical closed activated conformation, and additional density corresponding to the chemical surrogate for the substrate undergoing ubiquitylation. d, NEDD8 and the covalently linked WHB domain of CUL1 form the activation module.

This complex, representing the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate, explains rapid ubiquitylation through unprecedented neddylated cullin–RING arrangements (Extended Data Fig. 5). NEDD8 is nearly encircled by interactions; it binds the WHB domain of CUL1—to which it is linked—in an ‘activation module’, and positions a ‘catalytic module’ relative to the substrate-scaffolding module to juxtapose the active site and the substrate (Fig. 2b–d).

In the catalytic module, RBX1 binds UBE2D~Ub in the canonical RING-activated ‘closed’ conformation, in which noncovalent interactions between UBE2D and ubiquitin allosterically activate the thioester bond between them^28,29,30. Compared with previously isolated RING–UBE2D~Ub structures^28,29,30, the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate shows additional density corresponding to the substrate proxy along a trajectory to the ubiquitylation active site (Fig. 2c, Extended Data Figs. 6a, 7b). A UBE2D groove seems to engage the substrate polypeptide in a manner poised to assist in projecting adjacent lysines into the active site. This engagement of substrate polypeptide may contribute to the ability of UBE2D to ubiquitylate a broad range of proteins³¹.

The structure revealed that the distance between the β-TRCP-bound phosphodegron of IκBα and the UBE2D~Ub active site is around 22 Å, which is compatible with the spacing between this motif and potential acceptor lysines in many substrates³² (Extended Data Fig. 6). We could therefore make the following predictions: firstly, peptide substrates with sufficient residues (for example, 13 and 9) between the phosphodegron and the acceptor lysine to span a distance of 22 Å should be rapidly primed in a NEDD8-dependent manner; secondly, a peptide substrate with too few spacer residues (for example, 4) to span this gap should be severely impaired for priming by neddylated CRL1^β-TRCP and UBE2D, but that the addition of a ubiquitin could satisfy geometric constraints and enable further polyubiquitylation; and thirdly, the substrates should show little difference in UBE2D-mediated priming with unneddylated CRL1^β-TRCP. Comparing kinetic parameters for peptide substrates with 13-, 9-, and 4-residue spacers confirmed these predictions, with the priming rate of the latter peptide with neddylated CRL1^β-TRCP reduced below the limit of our quantification (Extended Data Fig. 6, Extended Data Table 1).

NEDD8 coordinates ubiquitin ligation assembly

NEDD8 is covalently linked to the WHB domain from CUL1, and together they form a globular activation module. A NEDD8 groove, comprising the Ile36/Leu71/Leu73 hydrophobic patch and the C-terminal tail, embraces the hydrophobic face of the isopeptide-bound CUL1 helix (Figs. 2d, 3a, b). At the centre, Gln40 of NEDD8 contacts CUL1, the isopeptide bond, and the C-terminal tail of NEDD8 in a buried polar interaction that is typical of such organizing apolar interfaces³³. This rationalizes how pathogenic bacterial effectors that catalyse Gln40 deamidation impair CRL1-dependent ubiquitylation^12,24,25,26: the resultant negative charge would destroy the CUL1–NEDD8 interface.

Fig. 3: Intra- and inter-module interfaces specifying the catalytic architecture for ubiquitin priming of the substrate by neddylated CRL1^β-TRCP with UBE2D.

a, Cryo-EM density highlighting noncovalent interfaces that contribute to the catalytic architecture for neddylated CRL1^β-TRCP-mediated ubiquitin transfer from UBE2D to a substrate. Circled regions correspond to interfaces within the activation module, and between activation and catalytic, activation and substrate-scaffolding, and catalytic and substrate-scaffolding modules shown in b–f. b, Close-up view of the intra-activation module interface, showing the buried polar residue Gln40 of NEDD8 and the Ile36/Leu71/Leu73 hydrophobic patch making noncovalent interactions with the WHB domain of CUL1, adjacent to the isopeptide bond linking NEDD8 and CUL1. c, Close-up view of the interface between the activation and catalytic modules, showing key residues at the interface between NEDD8 and the backside of UBE2D. d, Close-up view highlighting (in orange) the residues of NEDD8 that differ in ubiquitin, and that are at the interface with the substrate-scaffolding module. e, Close-up view highlighting His32 of UBE2D at the interface with the substrate-scaffolding module. f, Close-up view indicating the role of the loop-out conformation of NEDD8, which is required for the binding of UBE2D.

The activation module binds the catalytic module, which explains how neddylation helps CRL1^β-TRCP to recruit UBE2D in cells³⁴.The Ile44 hydrophobic patch of NEDD8 engages the ‘backside’ of UBE2D, which is opposite the active site for ubiquitylation (Fig. 3c, Extended Data Fig. 7a). Contacts resemble those described for free NEDD8 or ubiquitin binding to the backside of UBE2D and allosterically stimulating the intrinsic reactivity of an isolated RING–UBE2D~Ub subcomplex^35,36,37,38. We examined intrinsic reactivity by monitoring ubiquitin discharge to free lysine using a previously described hyperactive neddylated CRL1^β-TRCP mutant³⁹ and high concentrations of enzyme and lysine. Substrate-independent ubiquitin transferase activity was impaired by a NEDD8 mutant that disrupts the integrity of the activation module, and by mutations of UBE2D that hinder its interactions with the covalently linked ubiquitin, the RING domain of RBX1 or NEDD8 (Extended Data Fig. 7c–f). The architecture observed in the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate complex structure may therefore both stimulate the intrinsic reactivity of the UBE2D~Ub intermediate and place the catalytic centre in proximity to the β-TRCP-bound substrate.

The activation module is itself positioned by the binding of NEDD8 to the substrate-scaffolding module. Leu2, Lys4, Glu14, Asp16, Arg25, Arg29, Glu32, Gly63 and Gly64 of NEDD8—which nestle in a concave CUL1 surface—differ in ubiquitin, and these amino acids account for nearly one-third of the differences in sequence between the two proteins (Fig. 3d). The counterparts of these amino acids in ubiquitin would be expected to repel CUL1, which rationalizes the need for NEDD8 as a distinctive ubiquitin-like protein.

In addition, the catalytic module contacts both CUL1–RBX1 and the substrate receptor sides of the substrate-scaffolding module (Fig. 3e, Extended Data Fig. 5). On one side, the RING domain of RBX1 stacks on the Trp35 side chain of its C/R domain, which is consistent with previously reported effects of introducing a W35A mutation into RBX1³⁹. On the other side, the curved β-sheet of UBE2D complements the propeller of β-TRCP (Fig. 3e).

NEDD8 conformation coincidence coupling

Different conformations of ubiquitin and ubiquitin-like proteins have long been known to influence their interactions, with ubiquitin-binding domains selecting between ‘loop-in’ or ‘loop-out’ orientations of the Leu8-containing β1/β2-loop. However, it remains largely unknown how these conformations might simultaneously affect binding to multiple partners⁴⁰. Our data show that NEDD8 must adopt the loop-out conformation to both form the activation module and to engage UBE2D in the catalytic module (Fig. 3f, Extended Data Fig. 7g–i). The conformation of NEDD8 apparently serves as a coincidence detector, coupling noncovalent binding to the linked WHB domain of CUL1 and to the catalytic module.

Synergistic catalytic assembly

We next assessed the importance of the structurally observed interfaces by testing the effects of mutations in relevant regions of the proteins. We used an attenuated pulse-chase assay format that qualitatively, but exclusively, monitors NEDD8-activated substrate priming by CRL1^β-TRCP and UBE2D (Extended Data Fig. 8). Mutations that were designed to destroy the activation module (NEDD8(Q40E)) or to hinder interactions between the activation and catalytic modules (NEDD8(I44A) or UBE2D(S22R)) substantially impaired substrate priming, as did swapping key NEDD8 residues at the interface with the substrate-scaffolding module with those found in ubiquitin. Moreover, although the structural basis for ubiquitylation by other CRLs requires further investigation, these mutations also impair the UBE2D-mediated priming of a cyclin E phosphopeptide substrate with neddylated CRL1^FBW7, and of IKZF zinc finger 2 by neddylated CRL4^{CRBN/Pomalidomide} (Extended Data Fig. 8g–o).

Given that the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate depends on several conformational changes and large interfaces within and between modules (Fig. 2, Extended Data Figs. 5, 7), we hypothesized that pairing mutations that affect various interfaces would have synergistic effects. We used two types of experiment to define kinetic parameters for peptide substrate ubiquitylation—the Michaelis constant, K_m, was measured by titrating UBE2D, and the rate constant, k_obs, for ubiquitin transfer was determined by rapid quench-flow at saturating UBE2D concentrations to isolate catalytic defects. We drew three main conclusions from the results (Fig. 4a, Extended Data Fig. 1, Extended Data Table 1). First, although no individual mutation is as detrimental as eliminating neddylation, the mutation of each interface on its own shows a decrease in k_obs/K_m of approximately tenfold or more compared with the wild-type. The most detrimental ‘single’ mutation is the substitution of NEDD8 with Ub(R72A)—this confirms the importance of NEDD8 and of interactions between the activation- and substrate-scaffolding modules. The next most detrimental mutation is Q40E in NEDD8—this underscores the importance of the structure of the activation module and its role in establishing the loop-out conformation of NEDD8. Second, combining mutations at several sites has a devastating effect on activity, even if the effect of the individual mutations alone is mild. For example, when using NEDD8(I44A) the activity is reduced by tenfold compared with the wild-type; however, when this mutant is combined with UBE2D(H32A) at the catalytic module–substrate receptor interface—which is mildly defective in an attenuated assay with subsaturating UBE2D (Extended Data Fig. 8f)—a near 200-fold reduction in activity is observed. Third, the mutations have comparatively little effect on chain elongation, which is consistent with the structure of the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate defining ubiquitin transfer from UBE2D directly to the unmodified substrate.

Fig. 4: Multifarious interactions configuring rapid substrate priming.

a, Effects of the indicated mutants—within the activation module, between activation and catalytic, activation and substrate-scaffolding and substrate-scaffolding and catalytic modules, alone or in combination—on the catalytic efficiency of substrate priming, as quantified by overall fold difference in k_obs/K_m compared with wild-type neddylated CRL1^β-TRCP and UBE2D-catalysed ubiquitylation of a peptide substrate. Reactions with unneddylated CRL1^β-TRCP serve as a reference, and used CUL1(K720R) to prevent obscuring the interpretation of results by artefactual ubiquitin transfer to CUL1 and the resultant artefactual activation of substrate priming. Graphs show the average value from two different experiments (technical replicates), for which curve fits and values are provided in Extended Data Fig. 1 and Extended Data Table 1. b, On their own, neddylated CRL1^β-TRCP and UBE2D~Ub are dynamic, and at an extreme their constituent proteins and/or domains may be substantially mobile. Mobile entities are harnessed in the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate. There could be multiple routes to the catalytic architecture. It seems equally plausible that the UBE2D~Ub intermediate would first encounter either the RING domain of RBX1, or NEDD8—either of which would increase the effective concentration for the other interaction. Likewise, noncovalent-binding between NEDD8 and its linked WHB domain, or with the backside of UBE2D, would stabilize the loop-out conformation of NEDD8—this would also favour the other interaction. Ultimately, NEDD8, the cullin, and the RBX1-bound UBE2D~Ub intermediate make numerous interactions that synergistically establish a distinctive catalytic architecture that places UBE2D adjacent to β-TRCP.

Discussion

Our cryo-EM structure representing the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate suggests a model for substrate priming that addresses many longstanding questions. First, rapid substrate ubiquitylation can be explained by NEDD8, the cullin and RBX1-bound UBE2D~Ub making numerous interactions that activate UBE2D and synergistically place the catalytic centre adjacent to β-TRCP. Second, biochemical features of neddylated CRLs that were incompatible with previous structures are now rationalized, including NEDD8-stimulated crosslinking between a CRL1^β-TRCP-bound phosphopeptide and UBE2D⁴; simultaneous NEDD8 linkage to a cullin and binding to the backside of RBX1-bound UBE2D^3,7,37; and the detrimental effects of bacterial-effector-catalysed NEDD8 Gln40 deamidation^24,25,26 (Figs. 2, 3). Third, residues in ubiquitin that differ from those in NEDD8 would clash in the catalytic architecture (Fig. 3d, Extended Data Table 1), thus rationalizing the existence of NEDD8 as a distinct ubiquitin-like protein.

In the absence of other factors, the scaffolding module of neddylated CRL1^β-TRCP robustly bridges the substrate with the C/R domain, whereas NEDD8, its linked CUL1 WHB domain and the RING domain of RBX1 are relatively dynamic and apparently mobile (Extended Data Fig. 2). These mobile entities are harnessed in the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate (Fig. 2). The numerous requisite protein–protein interactions and conformational changes suggest that there could be several routes to the catalytic architecture (Fig. 4b), in which the formation of interfaces successively narrows the range of options—akin to progression down a free-energy funnel. Because ubiquitylation does occur with mutant substrates or enzymes, albeit at substantially lower rates (Extended Data Table 1), we cannot exclude that ubiquitin could be transferred from RING- and NEDD8-bound UBE2D in various orientations relative to the substrate-scaffolding module. However, if the thioester bond is both in the RING-activated configuration and adjacent to the substrate—as in the structure—this would increase the rate at which the presumably random exploration of three-dimensional space by a substrate lysine would lead to productive collision with the active site. Accordingly, reducing any single contribution to the structurally observed catalytic architecture increases the relative importance of other contacts—even lesser ones (Fig. 4a, Extended Data Table 1).

Whereas neddylated CRL1^β-TRCP and UBE2D seem to be optimal for ubiquitin priming of peptide-like substrates, the limited effect of mutations on the linkage of subsequent ubiquitins (Extended Data Table 1) raises the possibility that different forms of ubiquitylation involve alternative—currently unknown—catalytic architectures. Although not overtly observed for our substrates with a single acceptor lysine, an outstanding question is whether there are other circumstances in which a substrate-linked ubiquitin could mimic NEDD8 and activate further ubiquitylation. Moreover, in addition to UBE2D, neddylated CRLs recruit a range of other ubiquitin-carrying enzymes—from ARIH-family RBR E3s for substrate priming to other E2s for polyubiquitylation^21,41,42,43—and UBXD7, which in turn recruits the AAA-ATPase p97 to process some ubiquitylated substrates⁴⁴. We speculate that these CRL partners uniquely harness the dynamic NEDD8, its linked CUL WHB domain and/or the RING domain of RBX1 to specify distinct catalytic activities, much like the thioester-linked UBE2D~Ub intermediate captures neddylated CRL1^β-TRCP through multiple surfaces to specify substrate priming. The malleability of neddylated CRLs—coupled with numerous ubiquitin carrying enzyme partners—may underlie successful molecular glue or PROTAC-targeted protein degradation, whereas potential limitations in their ability to achieve the optimal multivalent catalytic architectures may explain failures in such chemically directed ubiquitylation of heterologous substrates⁴⁵. Additionally, it seems likely that the conformational dynamics of unneddylated CRLs (Extended Data Fig. 2) would enable transitioning between the different conformations coordinating cycles of neddylation–deneddylation with CAND1-driven substrate–receptor exchange^{39,46,47,48,49}. Thus, the multifarious nature of interactions and conformations that determine robust and rapid substrate priming—as revealed by the structure of the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate—also provides a mechanism by which common elements can be transformed by different protein partners to interconvert between distinct CRL assemblies to meet the cellular demand for ubiquitylation.

Methods

Cloning, protein expression and purification

All proteins are of human origin. All variants of UBE2D2, UBE2D3, UBE2M, RBX1, CUL1, CUL4, NEDD8 and ubiquitin were generated using PCR, Quikchange (Agilent), or were synthesized by Twist Biosciences.

UBE2D2 was purified as previously described⁵⁰, and UBE2D3 was purified in a similar manner. Ubiquitin was expressed in BL21(DE3) RIL as previously described⁵¹. Wild-type CUL1, RBX1(5-C), SKP1, β-TRCP2, CUL4A (from residue 38 to C terminus, hereafter referred to as CUL4), CRBN, DDB1 and UBA1 were cloned into pLIB vectors⁵². GST–TEV–RBX1 and CUL1, GST–TEV–RBX1 and CUL4A, His-TEV–β-TRCP2 and SKP1 or His-TEV–DDB1 and GST-TEV-CRBN were co-expressed by co-infecting with two baculoviruses. UBA1 was cloned with an N-terminal GST-tag with a TEV cleavage site. These proteins were expressed in Trichoplusia ni High-Five insect cells, purified by either GST or nickel-affinity chromatography, overnight TEV cleavage, followed by ion-exchange and size-exclusion chromatography. All variants of CUL1–RBX1 were purified similarly. Purification of NEDD8, UBE2M, APPBP1–UBA3, SKP1–FBW7 (from residue 263 to C terminus⁵³), neddylation of CUL1–RBX1, and fluorescent labelling of ubiquitin used for biochemical assays were performed as previously described³⁹. β-TRCP1 (monomeric form, from residue 175 to C terminus²⁰, hereafter referred to as β-TRCP1) with an N-terminal His-MBP followed by a TEV cleavage site was cloned into a pRSFDuet vector with SKP1ΔΔ (SKP1 with two internal deletions, of residues 38–43 and 71–82)⁵⁴. SKP1∆∆–β-TRCP1 was expressed in BL21(DE3) Gold E. coli at 18 °C, purified with nickel affinity chromatography, followed by TEV cleavage, anion exchange and size-exclusion chromatography. Modification of RBX1–CUL1 and RBX1–CUL4A by ubiquitin instead of NEDD8 was performed with the Ub(R72A) mutant that allows its activation and conjugation by neddylating enzymes APPBP1–UBA3 and UBE2M^55,56. The reaction for ubiquitylating CUL4A–RBX1 was performed at pH 8.8 to drive the reaction to completion. The previously described Y130L mutant of UBE2M was used to modify CUL1–RBX1 with the I44A mutant of NEDD8³⁹. IKZF1 ZF2 (residues 141–169, with two point mutations (K157R/K165R) and with a lysine added at position 140 to create a single target lysine at the N terminus⁵⁷) was cloned with an N-terminal GST with a 3C-Prescission cleavage site and a noncleavable C-terminal Strep-tag. IKZF1 ZF2 was purified by GST affinity chromatography, 3C-Prescission cleavage overnight, and size-exclusion chromatography. UBE4B RING-like U-box domain (residues 1200–C terminus) containing D1268T and N1271T point mutations that enhance activity⁵⁸ (hereafter referred to as UBE4B) was cloned with an N-terminal GST with TEV cleavage site. UBE4B was purified by GST affinity chromatography, TEV cleavage overnight, followed by ion exchange and size-exclusion chromatography.

Peptides

All peptides were stated to be of >95% purity by HPLC and were used as received.

Peptides used to quantify enzyme kinetics had the following sequences: IκBα, KERLLDDRHD(pS)GLD(pS)MRDEERRASY (obtained from New England Peptide); β-catenin short, KSYLD(pS)GIH(pS)GATTAPRRASY (obtained from Max Planck Institute of Biochemistry Core Facility); β-catenin medium, KAWQQQSYLD(pS)GIH(pS)GATTTAPRRASY (obtained from New England Peptide); β-catenin long, KAAVSHWQQQSYLD(pS)GIH(pS)GATTAPRRASY (obtained from Max Planck Institute of Biochemistry Core Facility); β-catenin for sortase-mediated transpeptidation to ubiquitin to generate a homogeneously ubiquitin-linked substrate, GGGGYLD(pS)GIH(pS)GATTAPRRASY (obtained from Max Planck Institute of Biochemistry Core Facility).

Peptides used for the qualitative assays monitoring substrate priming—that is, fluorescent ubiquitin transfer from UBE2D~Ub to substrate—had the following sequences: IκBα, KKERLLDDRHD(pS)GLD(pS)MKDEE (as previously described⁴¹); CyE, KAMLSEQNRASPLPSGLL(pT)PPQ(pS)GRRASY (as previously described⁴¹).

The nonmodifiable substrate analogue used in the competition experiment in Extended Data Fig. 3d had the following sequence: IκBα, RRERLLDDRHD(pS)GLD(pS)MRDEE (obtained from Max Planck Institute of Biochemistry Core Facility).

Peptides used in cryo-EM experiments are as follows: for the structure representing neddylated CRL1^β-TRCP–UBE2D~Ub–IκBα substrate described in detail and cryo-EM experiments shown in Extended Data Fig. 3e, f, h, i): IκBα, CKKERLLDDRHD(pS)GLD(pS)MKDEEDYKDDDDK (obtained from Max Planck Institute of Biochemistry Core Facility); for cryo-EM reconstructions of unnneddylated and neddylated CRL1^β-TRCP–IκBα substrate shown in Extended Data Fig. 2a, b: IκBα, KKERLLDDRHD(pS)GLD(pS)MKDEE (as previously described⁴¹).

Enzyme kinetics

UBE2D3 titrations under substrate single-encounter conditions for estimation of the K _m for E2 used by neddylated CRL1

These experiments used full-length CRL1^β-TRCP2 and UBE2D3, referred to here as CRL1^β-TRCP and UBE2D. Fifty micromolar peptide substrate (for a list of peptides that were used in the assay, see ‘Peptides’) was radiolabelled with 5 kU of cAMP-dependent protein kinase (New England Biolabs) in the presence of [γ³²P]ATP for 1 h at 30 °C. Two mixtures were prepared before initiation of the reaction: a UBA1/Ub mix containing unlabelled substrate competitor peptide that was identical in sequence to the labelled one (the one exception being the Ub-β-catenin substrate, in which the unlabelled β-catenin for sortase peptide was used); and a neddylated or unneddylated (with the CUL1(K720R) mutant, to prevent low-level ubiquitylation by UBE2D³) CRL1^β-TRCP/labelled peptide substrate mix. The UBA1/Ub mix contained reaction buffer composed of 30 mM Tris-HCl, 100 mM NaCl, 5 mM MgCl₂, 2 mM ATP and 2 mM DTT pH 7.5. The concentration of ubiquitin was 80 μM, with 1 μM UBA1 and 100 μM unlabelled peptide. UBE2D was first prepared as a twofold dilution series from a variable stock concentration, then introduced individually into tubes containing equal amounts of the UBA1/Ub mix. The CRL1^β-TRCP/labelled peptide substrate mix contained the same reaction buffer as the UBA1/Ub/UBE2D mix, 0.5 μM CRL1^β-TRCP, and 0.2 μM labelled peptide substrate. The reactions were initiated at 22 °C by combining equal volumes of both mixes, rapidly vortexed and quenched after 10 s in 2× SDS–PAGE buffer containing 100 mM Tris-HCl, 20% glycerol, 30 mM EDTA, 4% SDS and 4% β-mercaptoethanol pH 6.8. Each titration series was performed in duplicate and resolved on hand-cast, reducing 18% SDS–PAGE gels. The gels were imaged on a Typhoon 9410 Imager and quantification of substrate and products was performed using Image Quant (GE Healthcare). The product of each lane was measured as the fraction of the ubiquitylated products divided by the total signal, plotted against the UBE2D concentration, and fit to the Michaelis–Menten equation to estimate K_m (GraphPad Prism software). The standard error was calculated using Prism and has been provided in Extended Data Table 1 for all estimates of K_m.

Estimating the rates of ubiquitin transfer to CRL1-bound substrate using pre-steady-state kinetics

These experiments used full-length CRL1^β-TRCP2 and UBE2D3, referred to here as CRL1^β-TRCP and UBE2D, respectively. Separate UBA1/UBE2D/Ub and CRL1^β-TRCP/labelled peptide substrate mixes were prepared to assemble single-encounter ubiquitylation reactions. For most reactions, the UBA1/UBE2D3/Ub mix contained reaction buffer, 80 μM Ub, 1 μM UBA1, 40 μM UBE2D, and 200 μM unlabelled competitor substrate peptide. For all reactions containing either CUL1(K720R) or UBE2D(H32A) assayed with wild-type neddylated CUL1, CUL1 modified either with the mutant NEDD8(I44A) or with Ub(R72A) permitting ligation to CUL1, 120 μM Ub and 70 μM UBE2D were used. The CRL1^β-TRCP/labelled peptide substrate mixes contained reaction buffer, 0.5 μM CRL1^β-TRCP, and 0.2 μM labelled peptide (for a list of peptides that were used in the assay, see ‘Peptides’). Each mix was separately loaded into the left or right sample loops on a KinTek RQF-3 quench flow instrument, and successive time points were taken at 22 °C by combining the mixtures with drive buffer composed of 30 mM Tris-HCl and 100 mM NaCl pH 7.5. Reactions were quenched at various time points in 2× SDS–PAGE buffer to generate the time courses. Substrate and products from each time point were resolved on hand-cast, reducing 18% SDS–PAGE gels. The gels were imaged on a Typhoon 9410 Imager, and substrate and product bands were individually quantified as a percentage of the total signal for each time point using ImageQuant (GE Healthcare). Reactions were performed in duplicate, and the average of each substrate or product band was used for the analysis. The data for substrate (S0) or mono-ubiquitylated product (S1) bands were fit to their respective closed-form solutions as previously described⁵⁹ using Mathematica to obtain the values k_obs^S0–S1 and k_obs^S1–S2 (Extended Data Table 1). The standard errors were calculated in Mathematica and have been provided in Extended Data Table 1 for all estimates of k_obs.

Multiturnover assays with short β-catenin

The multiturnover assay showing the ubiquitin transfer to short β-catenin (Extended Data Fig. 6c, d) was performed as described in ‘Estimating the rates of ubiquitin transfer to CRL1-bound substrate using pre-steady-state kinetics’, but without the excess unlabelled β-catenin peptide substrate. Time points were collected by quenching in 2× SDS–PAGE loading buffer. Substrate and product were separated by SDS–PAGE followed by autoradiography. The fraction of unmodified substrate (S0) was quantified and fit to either a one-phase decay (unneddylated) or linear (neddylated) model (Prism 8). Similarly, products containing five or more ubiquitins were quantified and fit to either an exponential growth (neddylated) or linear (unneddylated) model. Experiments were performed in duplicate.

Generation of ubiquitylated β-catenin fusion via sortase reaction

A mimic of ubiquitylated β-catenin was generated by fusing a ubiquitin with a C-terminal LPETGG with a GGGG-β-catenin peptide. The reaction was incubated with concentrations of 50 μM UB^LPETGG, 300 μM GGGG-β-catenin peptide, and 10 μM 6×His-Sortase A for 10 min in 50 mM Tris, 150 mM NaCl, 10 mM CaCl₂, pH 8.0. Sortase A was removed by retention on nickel resin, and the product was further purified by size-exclusion chromatography in 25 mM HEPES, 150 mM NaCl, 1 mM DTT at pH 7.5.

The sequence of ubiquitin with sortase motif was as follows: MQIFVKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGSGSGSLPETGG.

Other biochemical assays

For experiments comparing the activity of neddylated with unneddylated CRLs or CUL–RBX1 complexes, in the unneddylated versions the NEDD8 modification sites of CUL1 and CUL4 were mutated to Arg (CUL1(K720R) and CUL4A(K705R)) to prevent obscuring interpretation of the results by low-level ubiquitylation of the NEDD8 consensus Lys during the ubiquitylation reactions³.

Substrate priming assays

Experiments in Extended Data Fig. 8a–f used full-length CRL1^β-TRCP2 and UBE2D3. Ubiquitylation of IκBα by CRL1^β-TRCP via UBE2D3 was monitored using a pulse-chase format that specifically detects CRL^β-TRCP-dependent ubiquitin modification from UBE2D to IκBα independently of effects on UBA1-dependent formation of the UBE2D3~Ub intermediate. The pulse reaction generated a thioester-linked UBE2D~Ub intermediate and contained 10 μM UBE2D, 15 μM fluorescent ubiquitin, 0.2 μM UBA1 in 50 mM Tris, 50 mM NaCl, 2.5 mM MgCl₂, 1.5 mM ATP pH 7.5 incubated at room temperature for 10 min. The pulse reaction was quenched with 25 mM EDTA on ice for 5 min, then further diluted to 100 nM UBE2D in 25 mM MES, 150 mM NaCl pH 6.5 for subsequent mixture with components of the reaction for neddylated CRL1^β-TRCP-dependent ubiquitin transfer to the substrate in the chase reaction. The chase reaction mix consisted of 400 nM CRL (NEDD8–CUL1–RBX1–SKP1–β-TRCP), and 1 μM substrate (phosphorylated peptide derived from IκBα) in 25 mM MES, 150 mM NaCl pH 6.5 incubated on ice. After the quench, the pulse reaction mix was combined with the chase reaction mix at a 1:1 ratio on ice. The final reaction concentrations were 50 nM UBE2D (in thioester-linked UBE2D~Ub complex) and 200 nM neddylated CRL1^β-TRCP to catalyse substrate ubiquitylation. Samples were taken at each time point, quenched with 2× SDS–PAGE sample buffer, protein components were separated on nonreducing SDS–PAGE, and the gel was scanned on an Amersham Typhoon imager (GE Healthcare).

Substrate priming reactions assaying the effects of variations in UBE2D shown in Extended Data Fig. 8g–i on CRL1^FBW7-dependent ubiquitylation—a phosphopeptide derived from CyE—were performed similarly to those for CRL1^β-TRCP as described above with 100 nM UBE2D~Ub (based on concentration of UBE2D from the pulse reaction), 500 nM neddylated CUL1–RBX1–SKP1–FBW7 (residues 263 to the C terminus), and 2.5 μM CyE phosphopeptide in 25 mM HEPES, 150 mM NaCl pH 7.5 at room temperature. Experiments testing the effects of variations in NEDD8 (or its substitution with Ub(R72A)) were performed in the same manner, except with 250 nM NEDD8 (or variant)-modified CUL1–RBX1–SKP1–FBW7 (residues 263 to the C terminus).

Our assay for CRL4^CRBN ubiquitylation of IKZF was established on the basis of findings that ZF2 mediates tight immunomodulatory-drug-dependent interactions sufficient to target degradation, and UBE2D3 contributes to the stability of CRL4^CRBN neomorphic substrates in cells^57,60,61. Substrate priming reactions showing controls and assaying effects of variations in NEDD8 and UBE2D3 shown in Extended Data Fig. 8j–o monitored ubiquitylation of IKZF ZF2 with 400 nM UBE2D~Ub (concentration determined by that of UBE2D in the chase reaction), 500 nM NEDD8–CUL4–RBX1–DDB1–CRBN, 5 μM pomalidomide and 2.5 μM IKZF ZF2 in 25 mM HEPES, 150 mM NaCl pH 7.5 at room temperature. Effects of swapping NEDD8 for Ub(R72A) on the CRL4^CRBN-mediated ubiquitylation of IKZF ZF2 are shown in Extended Data Fig. 8m and were performed similarly but with 100 nM UBE2D~Ub, 250 nM NEDD8 or Ub-modified CUL4–RBX1–DDB1–CRBN, 2.5 μM pomalidomide and 1.25 μM IKZF ZF2.

Assays for intrinsic activation of UBE2D~Ub intermediate

Assays shown in Extended Data Figs. 3b, g, 7d–f—monitoring neddylated CUL1–RBX1 activation of the thioester-linked UBE2D~Ub intermediate (that is, in the absence of substrate)—used the RBX1(N98R) variant that is hyperactive towards UBE2D~Ub³⁹. Experiments in Extended Data Figs. 3b, g, 7f were performed in pulse-chase format similar to substrate-priming assays, but with 9 μM UBE2D~Ub (loading reaction with 20 μM UBE2D, 30 μM Ub and 0.5 μM UBA1), 500 nM E3 and 5 mM free lysine. For unneddylated CUL1–RBX1- or UBE4B-dependent discharge, 50 mM free lysine was used instead. Discharge assays shown in Extended Data Fig. 7d, e used 5 μM UBE2D~Ub (loading reaction with 20 μM UBE2D, 20 μM Ub and 0.5 μM UBA1), 500 nM E3 and 10 mM free lysine for neddylated CUL1–RBX1-dependent discharge, and 50 mM free lysine for unneddylated CUL1–RBX1-dependent discharge. All assays were visualized by Coomassie-stained SDS–PAGE.

Generation of a stable proxy for the UBE2D~Ub–substrate intermediate

Preparation of His-TEV-Ub(1–75)-MESNa

His-TEV-Ub(1–75) was cloned using a previously described method⁶² into pTXB1 (New England Biolabs) and transformed into BL21(DE3) RIL. Cells were grown in terrific broth at 37 °C to an optical density at 600 nm (OD₆₀₀) of 0.8 and then induced with IPTG (0.5 mM), shaking overnight at 16 °C. The collected cells were resuspended (20 mM HEPES, 50 mM NaOAc, 100 mM NaCl, 2.5 mM PMSF pH 6.8), sonicated and then centrifuged (50,000g, 4 °C, 30 min). Ni-NTA resin (1 millilitre resin per litre of broth, Sigma Aldrich) was equilibrated with the resuspension buffer and incubated with the cleared lysate at 4 °C on a roller (30 rpm) for 1 h. The resin was then transferred to a gravity column and washed (5 × 1 column volume with 20 mM HEPES, 50 mM NaOAc, 100 mM NaCl pH 6.8). Protein was then eluted (5 × 1 column volume with 20 mM HEPES, 50 mM NaOAc, 100 mM NaCl 300 mM imidazole pH 6.8). Ubiquitin was then cleaved from the chitin-binding domain by diluting the eluted protein 10:1 (v/v) with 20 mM HEPES, 50 mM NaOAc, 100 mM NaCl, 100 mM sodium 2-mercaptoethanesulfonate (Sigma Aldrich) pH 6.8. This solution was incubated at room temperature overnight on a roller (30 rpm). Ub-MESNa was finally purified by size-exclusion chromatography (SD75 HiLoad, GE Healthcare) equilibrated with 12.5 mM HEPES, 25 mM NaCl pH 6.5.

Sequence of His-TEV–Ub(1–75)-chitin-binding domain:

MGSSHHHHHHENLYFQGSGGMQIFVKTLTGKTITLEVEPSDTIENVKAKIQD

KEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGCFAKGTNVL

MADGSIECIENIEVGNKVMGKDGRPREVIKLPRGRETMYSVVQKSQHRAH

KSDSSREVPELLKFTCNATHELVVRTPRSVRRLSRTIKGVEYFEVITFEMGQ

KKAPDG

Native chemical ligation to make Ub(1–75)-Cys–IκBα

His-Ub(1–75)–MESNa (200 μM final concentration) and freshly dissolved IκBα peptide (H-CKKERLLDDRHDpSGLDpSMKDEEDYKDDDDK-OH) (1,000 μM final concentration) were combined in a 1.5-ml tube in 50 mM NaPO₄, 50 mM NaCl pH 6.5. This was incubated with rocking at 30 rpm for 1 h at room temperature before TCEP was added to 1 mM. After rocking for an additional hour at room temperature, the reaction was quenched by adding 500 mM NaPO₄ pH 8.0 to 45 mM. The entire solution was then incubated with Ni-NTA resin (300 μl for a 1 ml reaction) at 30 rpm for 1 h at 4 °C. In a gravity column, the resin was then washed with 6 × 300 μl 50 mM NaPO₄, 50 mM NaCl, 1 mM β-mercaptoethanol pH 8.0. Protein was eluted with 50 mM NaPO₄, 50 mM NaCl, 1 mM β-mercaptoethanol, 300 mM imidazole pH 8.0. Fractions were analysed by SDS–PAGE and nanodrop.

Formation of disulfide linkage between UBE2D Cys85 and Ub(1–75)-Cys-IκBα

The same approach was used to generate complexes for UBE2D2 and UBE2D3, referred to collectively as UBE2D. UBE2D(C21I/C107A/C111D) was purified from size-exclusion chromatography (see ‘Cloning, protein expression and purification’) and then immediately used without freezing. After size-exclusion chromatography, the protein was concentrated (Amicon, EMD Millipore) to 600 μM. Protein (2 × 100 μl) was separately desalted (2 × Zeba, 0.5 ml column, 7,000 molecular weight cut-off filter, Thermo Fisher) to 20 mM HEPES, 250 mM NaCl, 5 mM EDTA pH 7.0. Elutions were combined and immediately added together to 34 μl 10 mM 5,5′-dithiobis-(2-nitrobenzoic acid) (Sigma Aldrich, dissolved in 50 mM NaPO₄ pH 7.5) and mixed by pipetting before incubating at room temperature for 30 min. The solution was then desalted (2 × Zeba, 0.5 ml column, 7,000 molecular weight cut-off filter, Thermo Fisher) to 20 mM HEPES, 250 mM NaCl, 5 mM EDTA pH 7.0 at the same time that Ub(1-75)–Cys–IκBα (500 μl at 100 μM) was desalted (1 × Zeba, 2 ml column, 7,000 molecular weight cut-off filter, Thermo Fisher) to the same buffer. The UBE2D and ubiquitin components were then immediately combined and incubated at room temperature for 30 min, at which point the sample was loaded to a Superdex 75 Increase column (GE Healthcare) equilibrated with 20 mM HEPES, 250 mM NaCl, 5 mM EDTA pH 7.0.

Comparing the ability of stable proxy for the UBE2D~Ub–substrate intermediate and subcomplexes to compete with ubiquitylation

Assays comparing ubiquitylation in the presence of competitors (stable proxy for the UBE2D~Ub–substrate intermediate, stable isopeptide-linked mimic of UBE2D~Ub, and nonmodifiable substrate peptide) were carried out similarly to that described for our substrate priming assay described in ‘Substrate priming assays’ with the following modifications. The assay was performed in pulse-chase format to exclude the potential for competitors to affect generation of the UBE2D~Ub intermediate. In the pulse reaction, a thioester-linked UBE2D~Ub intermediate was generated by incubating 10 μM UBE2D, 15 μM fluorescent Ub, and 0.2 μM UBA1 in a buffer that contained 50 mM Tris, 50 mM NaCl, 2.5 mM MgCl₂, and 1.5 mM ATP pH 7.6 at room temperature for 10 min. The pulse reaction was next quenched by the addition of an equal volume of 50 mM Tris, 50 mM NaCl, 50 mM EDTA pH 7.6 and placed on ice for 5 min, then further diluted to 100 nM in a buffer containing 25 mM MES pH 6.5 and 150 mM NaCl. The E3-substrate mix consisted of 400 nM NEDD8–CUL1–RBX1, 400 nM SKP1–β-TRCP, 1 μM IκBα peptide, with or without 1 μM competitor in a buffer consisting of 25 mM MES and 150 mM NaCl pH 6.5, and was incubated at 4 °C for 10 min to achieve equilibrium. Reactions were initiated on ice by the addition of an equal volume of pulse reaction to the E3-substrate mix, resulting in final reaction conditions of 50 nM UBE2D~Ub, 200 nM E3 neddylated CRL1^β-TRCP, and 500 nM substrate with or without 500 nM competitor. Samples were taken at the indicated time points and quenched with 2× SDS–PAGE sample buffer. Substrate and products were then separated by SDS–PAGE, and subsequently visualized using an Amersham Typhoon imager (GE Healthcare).

Early attempt to visualize ubiquitin transfer by neddylated CRL1^β-TRCP and UBE2D and rationale for approaches to improve electron microscopy samples

In our initial attempt to determine a structure visualizing ubiquitin transfer by neddylated CRL1^β-TRCP and UBE2D, we used a full-length β-TRCP (β-TRCP2), which is a homodimer²⁷, and a proxy for a UBE2D~Ub–substrate intermediate based on a method used to capture a Sumoylation intermediate⁶³. In the previous study, SUMO was installed via an isopeptide bond on a residue adjacent to the E2 catalytic Cys, and substrate was crosslinked to the E2 Cys via an ethanedithiol linker. Here, we introduced the corresponding lysine substitution in the background of an optimized E2 (UBE2D2(L119K/C21I/C107A/C111D)). Using high concentrations of UBA1 and high pH, we generated an isopeptide-bonded complex between ubiquitin and this UBE2D2 variant in a manner dependent on the L119K mutation, and crosslinked the Cys of the IκBα substrate mimic peptide to the UBE2D~Ub complex using EDT as previously described⁶³. The resultant cryo-EM data map, shown in Extended Data Fig. 3e, presented two major challenges. First, the dimer exacerbated structural heterogeneity, with minor differences presumably based on natural motions between the two protomers. Second, the donor ubiquitin was poorly visible, presumably owing to it not being linked to the catalytic Cys. Thus, we generated many samples in parallel to overcome these challenges by (1) using a monomeric version of β-TRCP1 that had previously been crystallized²⁰; (2) removing two loops in SKP1 that are known to be flexible and not required for ubiquitylation activity (although they are required for CAND1-mediated substrate–receptor exchange)^13,54; (3) devising a chemical approach to synthesize a proxy for the UBE2D~Ub–substrate intermediate in which all three entities are simultaneously linked to the E2 catalytic Cys; (4) using a point mutant version of RBX1 that is hyperactive for substrate priming with UBE2D but defective for ubiquitin chain elongation with UBE2R-family E2s³⁹. Notably, with a K_m for UBE2D3 of 350 nM and rates of ubiquitylating the medium β-catenin peptide substrate of 8.9 s⁻¹ (S0–S1) and 0.25 s⁻¹ (S1–S2), we confirmed that the monomeric version of neddylated CRL1^β-TRCP1 is kinetically indistinguishable from full-length, homodimeric neddylated CRL1^β-TRCP2.

Cryo-EM

Sample preparation

For neddylated or unneddylated CUL1–RBX1–SKP1–β-TRCP–IκBα samples, subcomplexes were mixed in an equimolar ratio with 1.5-fold excess substrate peptide, incubated for 30 min on ice and purified by size-exclusion chromatography in 25 mM HEPES, 150 mM NaCl, 1 mM DTT pH 7.5. The complex was further concentrated and crosslinked by GraFix⁶⁴. The sample was next liberated of glycerol using Zeba Desalt Spin Columns (Thermo Fisher), concentrated to 0.3 mg ml⁻¹, and 3 μl of sample was applied to R1.2/1.3 holey carbon grids (Quantifoil) and was plunge-frozen by Vitrobot Mark IV in liquid ethane. Structure determination of the neddylated CUL1–RBX1–SKP1–β-TRCP–Ub~UBE2D–IκBα complex used a similar method as above, with 1.5-fold excess of the stable proxy for the UBE2D~Ub–IκBα intermediate, but with no DTT in buffer. After SEC, GraFix, and desalting, 3 μl of 0.08 mg ml⁻¹ sample was applied to graphene oxide-coated Quantifoil R2/1 holey carbon grids (Quantifoil)⁶⁵ and was plunge-frozen by Vitrobot Mark IV in liquid ethane.

Electron microscopy

Datasets were collected on a Glacios cryo transmission electron microscope at 200 kV using a K2 Summit direct detector in counting mode. For the CUL1–RBX1–SKP1–β-TRCP1∆D–IκBα dataset, 6,433 images were recorded at 1.181 Å per pixel with a nominal magnification of 36,000×. A total dose of 60 e⁻ Å⁻² was fractionated over 50 frames, with a defocus range of −1.2 μm to −3.3 μm. For NEDD8–CUL1–RBX1–SKP1–β-TRCP1∆D–IκBα, 2,061 images were recorded at 1.885 Å per pixel with a nominal magnification of 22,000×. A total dose of 59 e⁻ Å⁻² was fractionated over 38 frames, with a defocus range of −1.2 μm to −3.3 μm.

Datasets were also collected on a Talos Arctica at 200 kV using a Falcon II direct detector in linear mode. For each sample, around 800 images were recorded at 1.997 Å per pixel with a nominal magnification of 73,000×. A total dose of approximately 60 e⁻ Å⁻² was fractionated over 40 frames, with a defocus range of −1.5 μm to −3.5 μm.

High-resolution cryo-EM data were collected on a Titan Krios electron microscope at 300 kV with a Quantum-LS energy filter, using a K2 Summit direct detector in counting mode. 9,112 images were recorded at 1.06 Å per pixel with a nominal magnification of 130,000×. A total dose of 70.2 e⁻ Å⁻² was fractionated over 60 frames, with a defocus range of −1.2 μm to −3.6 μm.

Data processing

Frames were motion-corrected using RELION-3.0⁶⁶ with dose weighting. Contrast transfer function was estimated using CTFFIND⁶⁷. Particles were picked with Gautomatch (K. Zhang, MRC Laboratory of Molecular Biology). Two-dimensional classification was performed in RELION-3.0, followed by 3D ab initio model building by sxviper.py from SPARX⁶⁸. The initial model from sxviper.py was imported to RELION-3.0 for further 3D classification, refinement, post-processing and particle polishing using frames 2–25.

Protein identification and model building

The final reconstructions displayed clear main chain and side chain densities, which enabled us to model and refine the atomic coordinates. Known components (CUL1–RBX1, RCSB Protein Data Bank codes (PDB) 1LDJ and 4P5O; SKP1∆∆–β-TRCP1, PDB 1P22 and 6M90; UBE2D~Ub with a backside-bound ubiquitin to be replaced by NEDD8 sequences, PDB 4V3L) were manually placed as a whole or in parts and fit with rigid-body refinement using UCSF Chimera⁶⁹. The resultant complete structure underwent rigid-body refinements in which each protein or domain was allowed to move independently. Further iterative manual model building and real space refinements were carried out until good geometry and map-to-model correlation was reached. Manual model building and rebuilding were performed using COOT⁷⁰, and Phenix.refine⁷¹ was used for real space refinement.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Extended data figures and tables

Extended Data Fig. 1 Quantitative pre-steady-state enzyme kinetics of neddylated CRL1^β-TRCP- and UBE2D-dependent ubiquitylation.

Gel images are representative of independent technical replicates (n = 2); the symbols on the graphs show the data from independent experiments (n = 2). a, Autoradiogram of SDS–PAGE gel showing products of ubiquitylation reactions under single-encounter conditions for the interaction of radiolabelled substrate (medium β-catenin substrate peptide derived from β-catenin) with neddylated CRL1^β-TRCP, titrating UBE2D3 (hereafter denoted UBE2D). Each lane represents a single ubiquitylation reaction that was used to estimate the fraction of peptide that had been converted into ubiquitylated products as a function of UBE2D concentration. b, Plots of the fraction of substrate that had been converted to ubiquitylated products against UBE2D concentration for ubiquitylation reactions containing either wild-type UBE2D (as shown in a), or the mutants UBE2D(S22R) or UBE2D(H32A). Various CRL1^β-TRCP complexes were assayed that contained either wild-type neddylated CRL1^β-TRCP (red), CRL1^β-TRCP complexes modified by NEDD8 variants containing I44A (orange), Q40E (green) or ‘ubiquitylizing’ (L2Q/K4F/E14T/D16E/G63K/G64E) substitutions (blue), or CRL1^β-TRCP modified by Ub(R72A) that is competent for ligation to CUL1 (purple) or unmodified CRL1^β-TRCP (CUL1 with the neddylation-site mutation K720R, black). Duplicate data points from independent experiments performed with identical samples are shown and were fit to the Michaelis–Menten model to estimate the K_m of UBE2D for CRL1^β-TRCP using nonlinear curve fitting (GraphPad Prism). c, Plots of the fraction of substrate that had been converted to ubiquitylated products against UBE2D concentration for ubiquitylation reactions with various substrate peptides: derived from IκBα (but with a single acceptor Lys); derived from β-catenin; derived from β-catenin with different spacing between the phosphodegron motif and a potential acceptor Lys (a medium β-catenin substrate peptide with a nine-residue spacer between the β-catenin phosphodegron and acceptor, matching the relative position of these moieties in IκBα; and a short β-catenin substrate in which the four residues between these moieties are too few to bridge the structurally observed gap between the substrate receptor and UBE2D~Ub active site); and a homogeneous ubiquitin linked-β-catenin generated by sortase-mediated transpeptidation wherein the only lysines are from ubiquitin. d, Autoradiogram of SDS–PAGE gel showing results from rapid quench-flow reactions under pre-steady-state single encounter conditions for the interaction of radiolabelled substrate (a medium β-catenin phosphopeptide) with CRL1^β-TRCP. The representative raw data are from a reaction using wild-type UBE2D and wild-type neddylated CRL1^β-TRCP, and show time-resolved conjugation of increasing numbers of individual ubiquitin molecules. S0, substrate with 0 ubiquitins; S1, substrate with one ubiquitin; S2, substrate with two ubiquitins; and so on. e, Plots comparing various substrate peptides described in c, showing disappearance of unmodified substrate (S0) with black circles, and the appearance of mono-ubiquitylated substrate (S1) with grey triangles, in rapid quench-flow reactions all performed as in d and under single-encounter conditions as in a. Duplicate data points from independent experiments performed with identical samples are shown. The data were fit to closed form equations (Mathematica) as previously described⁵⁹ to obtain both the rates for the transfer of the first ubiquitin to substrate (k_obs^S0–S1) and of the second ubiquitin to the singly ubiquitin-modified substrate (k_obs^S1–S2) as well as their associated standard error (Extended Data Table 1). f, Plots from experiments performed and analysed as described in e, except with radiolabelled medium β-catenin peptide substrate, CRL1^β-TRCP variants containing the indicated versions of CUL1–RBX1, and with either wild-type or indicated mutant versions of UBE2D.

Extended Data Fig. 2 In CRL1^β-TRCP, the RING domain of RBX1 and the WHB domain of CUL1—without or with a covalently linked NEDD8— are dynamic in the absence of other factors and are harnessed in the catalytic architecture for substrate ubiquitylation with UBE2D.

a, Cryo-EM density corresponding to substrate-scaffolding regions of unneddylated CRL1^β-TRCP is shown in surface representation, with the density encompassing the RING domain of RBX1 and the WHB domain of CUL1 outlined for different 3D classes in different colours corresponding to the percentage of particles in that 3D class. b, As in a but with neddylated CRL1^β-TRCP, with its surfaces outlined in different classes encompassing the RING domain of RBX1, the WHB domain of CUL1, and covalently modified NEDD8. c, Refined cryo-EM density from CRL1^β-TRCP reveals the substrate-scaffolding module bridging the substrate recruited to substrate receptor β-TRCP with the intermolecular C/R domain, readily fitted with crystal structures of SKP1–β-TRCP (PDB: 1P22)²⁰ and the N-terminal domain of CUL1 and the C/R domain of CUL1–RBX1 (PDB: 1LDK)⁷. d, Schematic showing dynamics of NEDD8, its linked CUL1 WHB domain and the RING domain of RBX1 based on cryo-EM data in b for substrate-bound neddylated CRL1^β-TRCP, and model for varying locations of the RBX1 RING-bound UBE2D~Ub relative to the substrate awaiting ubiquitylation. e, Left, schematic of the catalytic architecture based on the cryo-EM data shown in Fig. 2, representing neddylated CRL1^β-TRCP-catalysed ubiquitin transfer from E2 UBE2D to an IκBα-derived substrate peptide. Right, semi-transparent version of the schematic, highlighting the three modules (substrate-scaffolding, catalytic and activation modules), their constituents and locations establishing the catalytic architecture for substrate priming by neddylated CRL1^β-TRCP and UBE2D.

Extended Data Fig. 3 Generation of a stable proxy for the UBE2D~Ub–substrate intermediate, and characterization in complexes with neddylated CRL1^β-TRCP by cryo-EM and biochemistry.

Gel panels in this figure are representative of two independent experiments; n = 2. a, Our strategy for trapping a mimic of the transient neddylated CRL E2~Ub–substrate complex requires that the E2 UBE2D contain only a single cysteine at the active site. However, UBE2D contains three additional cysteines (Cys21, Cys107 and Cys111). Standard replacements of cysteine by serine or alanine severely compromised activity. On the basis of the structural locations of these cysteines, we presumed that their mutation hindered formation of the RING-activated, closed, active UBE2D~Ub conformation^28,29,30. We thus devised a systematic structure- and random-based approach to identify suitable replacements that qualitatively maintain wild-type levels of activity with neddylated CRLs. Structural analysis showed that Cys21 and Cys107 are in close proximity, such that mutation of both residues to alanine may generate a destabilizing cavity at this site. Combining UBE2D2(C107A) with Cys21 mutated to isoleucine, leucine or valine to compensate for the reduced hydrophobic volume led to the identification of C21I(C107A) as a suitable version for testing all other possible replacements for Cys111. A similar approach was taken for UBE2D3. A total of 48 different versions of UBE2D were tested to identify the UBE2D(C21I/C107A/C111D) mutant for chemical trapping at the remaining active site cysteine. b, Top, schematic of pulse-chase assay testing intrinsic activation of thioester-linked UBE2D~Ub intermediates. Although this is often tested by monitoring RING-dependent discharge of ubiquitin from UBE2D to free lysine, RBX1 RING-dependent activity is limited in this assay owing to sequence constraints imposed by the requirements for binding to partners other than UBE2D³⁹. Nonetheless, substrate-independent activation of UBE2D~Ub can be readily visualized using CUL1 complexed with a previously described hyperactive mutant RBX1(N98R)³⁹, and high enzyme and lysine concentrations. UBE2D~Ub generated in a pulse reaction was mixed with NEDD8-modified CUL1–RBX1 (shown here with the N98R mutant) and free lysine, and ubiquitin discharge was monitored over time by Coomassie-stained SDS–PAGE (as shown by the representative gel at the bottom) demonstrating that standard serine or alanine mutations of noncatalytic cysteines compromised activity (shown for the mutant C21A/C107A/C111S), whereas the optimized mutant (C21I/C107A/C111D) retains activity similar to that of the wild type. c, Overview of the generation of our stable proxy for the phosphorylated IκBα substrate intermediate linked at a single atom, and comparison to the previous method used to visualize noncanonical Lys sumoylation⁶³. d, Experiment validating our stable proxy for the UBE2D~Ub-phosphorylated IκBα substrate intermediate linked at a single atom, based on the hypothesis that its simultaneous occupation of the binding sites for the UBE2D~Ub intermediate and substrate should result in more potent inhibition of a neddylated CRL1^β-TRCP-dependent substrate priming reaction compared to the individual constituents of the complex. e, Cryo-EM reconstruction of neddylated CRL1^β-TRCP2 (with full-length, dimeric β-TRCP2) bound to a mimic of UBE2D2~Ub–IκBα generated by adapting the method used previously to visualize noncanonical lysine sumoylation⁶³. Ubiquitin is isopeptide-bonded to the substituted residue of a UBE2D(L119K) mutant, and a cysteine residue that replaces the acceptor in the substrate is disulfide-bonded to the catalytic cysteine of UBE2D2. This electron microscopy map visualizes the catalytic architecture of dimeric CRL1^β-TRCP2 in which the dimerization domain agrees well with the previous crystal structure²⁷, and its linked NEDD8 (circled in yellow) is bound to the backside of UBE2D, but the donor ubiquitin (absent from the region circled in orange) was not visible—presumably owing to inadequacies of the method used to generate this mimic of the catalytic intermediate, in which the ubiquitin and substrate are not both simultaneously linked to the UBE2D catalytic cysteine. Variations between the two protomers of the dimer also exacerbated sample heterogeneity. f, Cryo-EM reconstruction of neddylated CRL1^β-TRCP1∆D (with monomeric version of β-TRCP1, from residue 175 to the C terminus²⁰) bound to our newly developed proxy for the UBE2D3~Ub–IκBα intermediate. The phospho-IκBα peptide-substrate-bound β-TRCP–SKP1–CUL1–RBX1–NEDD8–UBE2D portion of this map superimposes with the map for the dimeric complex shown in e, but here the entire complex is visible—including both the NEDD8 (circled in yellow) and donor ubiquitin (circled in orange). g, To further increase cryo-EM sample homogeneity, we considered that the RBX1 RING sequence represents a compromise to meet requirements for its many different catalytic activities achieved with neddylation E2s, various ubiquitin carrying enzymes, and regulators including the inhibitor GLMN⁷². Therefore, we introduced a second RBX1 linchpin residue via mutation (N98R), which has previously been shown to improve neddylated CRL and UBE2D-dependent substrate priming at the expense of other RBX1-dependent functions (for example, with UBE2M and UBE2R2)³⁹. A Coomassie-stained SDS–PAGE gel from an assay for the intrinsic activity of UBE2D~Ub is shown, showing enhanced neddylated CRL-dependent activation of discharge to free lysine with the RBX1 N98R mutation. h, i, Cryo-EM reconstructions of neddylated CRL1^β-TRCP1∆D with RBX1(N98R) bound to our newly developed proxies for the UBE2D3~Ub–IκBα and UBE2D2~Ub–IκBα intermediates, the latter of which was pursued for high-resolution electron microscopy (final reconstruction refined to 3.7 Å resolution, shown on right).

Extended Data Fig. 4 Flow chart showing the stages of cryo-EM image processing.

a, Cryo-EM image-processing flow chart. Ultimately, reconstruction of the data yielded a focused refinement at 3.46 Å resolution and a global refinement at 3.7 Å resolution that superimposes well with lower-resolution maps that were obtained during attempts to visualize substrate priming with neddylated wild-type dimeric CRL1^β-TRCP. b, Two-dimensional classes representing particles used for final reconstructions. c, Angular distribution of final reconstruction. d, Gold-standard Fourier shell correlation (FSC) curve showing overall resolution at 3.72 Å at an FSC of 0.143. e, Electron microscopy density map coloured by local resolution. NEDD8, circled in yellow, is the entity displaying the highest local resolution in the map.

Extended Data Fig. 5 Extraordinary cullin–RING conformational changes in catalytic architecture juxtaposing the substrate and the active site of ubiquitylation.

a, Side-by-side comparison of relative RING-domain locations in different CRL complexes after superposition of the C/R domains from the original CUL1–RBX1 structure (PDB: 1LDJ, ‘pre-neddylation’—which data herein show is dynamic—although the crystal structure probably captured the conformation that enabled CAND1 binding and substrate receptor exchange)⁷, the structure representing the neddylation reaction (PDB: 4P5O)³⁹, and a structure of a neddylated CUL5–RBX1 domain (PDB: 3DQV, labelled ‘post-neddylation’, which revealed the potential for conformational changes in the neddylated CUL WHB- and RBX1 RING-domains³, and data herein shows is dynamic), and the structure presented here showing how the neddylated CUL1 WHB domain and RBX1 RING domain are harnessed in a catalytic architecture for ‘active ubiquitylation’. Trp35 of RBX1 is highlighted to show how it serves as a multifunctional platform for either the RING domain in different orientations, or for the E2-linked NEDD8 during neddylation³⁹. b, Superposition of the structures shown in a, highlighting different relative positions of the RING domain. c, Comparison of the relative locations of the CUL WHB domain in different structures after superimposing their C/R domains (not shown). d, Cryo-EM density from the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate intermediate complex, showing patchiness of the region corresponding to CUL1 ‘helix-29’⁷. This CUL1 region connecting the C/R and WHB domains is visible only as patchy density, whereas in previous cullin crystals this forms the rod-like helix-29 continuing into the WHB domain⁷. It seems that helix-29 of CUL1 dissolves into a flexible tether, which rationalizes the previously observed proteolytic sensitivity of this region in a neddylated CUL1–RBX1 complex³, and enables the displacement and rotation required for placing the ensuing WHB domain and its linked NEDD8 at the centre of the ubiquitylation complex.

Extended Data Fig. 6 Geometry between phosphodegron and acceptor in structure, substrates and ubiquitylation.

a, Cryo-EM density highlighting the relative placement of the substrate degron and the UBE2D~Ub active site. The approximately 22 Å distance between the UBE2D~Ub active site and the phosphodegron of β-TRCP-bound substrate requires at least 6 intervening residues in a substrate. b, Alignments for several reported β-TRCP substrates³², highlighting the degron sequence (yellow) and nearby lysines (red). Also shown are sequences of peptide substrates with a single acceptor Lys that were used in kinetics analyses. The peptide sequences were derived from IκBα, and from β-catenin with varying spacers between phosphodegron and acceptor Lys: wild-type β-catenin peptide, medium β-catenin peptide with lysine corresponding to IκBα, and short β-catenin peptide with a lysine five residues upstream of the N-terminal phosphoserine in the degron, which would be too short to bridge the structurally observed distance between the phosphodegron binding site on β-TRCP and the UBE2D catalytic cysteine in the ubiquitylation active site. c, Representative autoradiogram (n = 2) of SDS–PAGE gel showing products from indicated time points of ubiquitylation reactions under multiturnover conditions with either neddylated or unneddylated CRL1^β-TRCP and radiolabelled short β-catenin peptide substrate. The amount of short β-catenin peptide modified by neddylated CRL1^β-TRCP and UBE2D is too low in the single-encounter ubiquitylation reaction to enable quantification of kinetic parameters; however, product formation is apparent under multi-turnover conditions and shows that most products are heavily ubiquitylated. d, Plots fitting consumption of unmodified short β-catenin peptide substrate (S0) compared to formation of polyubiquitin chains with five or more ubiquitins (S5+) from reactions as in c. The symbols show the data from independent experiments (n = 2 technical replicates).

Extended Data Fig. 7 Interactions shaping the catalytic architecture of neddylated CRL1^β-TRCP–UBE2D~Ub–IκBα substrate intermediate.

a, NEDD8 and the catalytic module from the structure representing the neddylated CRL1^β-TRCP–UBE2D~Ub–IκBα intermediate, highlighting distinctive interactions between NEDD8 (yellow) and donor ubiquitin (orange) with UBE2D. b, Catalytic module from the neddylated CRL1^β-TRCP–UBE2D~Ub–IκBα intermediate, highlighting the covalently linked proxy for the IκBα substrate’s acceptor in the active site relative to a superimposed representative previous crystal structure of an isolated RING–UBE2D~Ub complex (grey, PDB: 4AP4)^28,29. In the inset, the density for the covalently linked proxy for IκBα substrate’s acceptor is shown in red in the active site. The chemical trap superimposes with consensus acceptors visualized in active sites of sumoylation⁶³ and neddylation³⁹ intermediates, where aromatic side chains guide the lysine targets (blue and green, respectively)^39,73. However, the myriad substrates of UBE2D neither conform to a specific motif nor do they or UBE2D display specific side chains that guide lysine acceptors into the catalytic centre. Instead, in the neddylated CRL1^β-TRCP–UBE2D~Ub–substrate complex, density from backbone atoms preceding the chemical proxy for the acceptor lysine corresponds to the aromatic guides in sumoylation and neddylation intermediates. c, Overview of assays for activation of intrinsic reactivity of the UBE2D~Ub intermediate. Top, schematic of pulse-chase assay for testing the effects of UBE2D mutations on activation, monitoring UBE2D~Ub discharge to free lysine activated by neddylated CUL–RBX1 compared to unneddylated or RING-like UBE4B controls. Bottom, sites of mutations shown as spheres on the structure of UBE2D from the cryo-EM structure of neddylated CRL1^β-TRCP–UBE2D~Ub–substrate complex. The colours of spheres reflect both the locations and the effects on UBE2D~Ub discharge to free lysine. Sites of mutations with marginal or no effect are shown in cyan, whereas those with major effects are otherwise coloured. Mutations that cause major defects map to the RBX1 RING-binding site (blue), the interaction surface with the donor ubiquitin (orange), and the interaction surface with NEDD8 (yellow). d, Representative Coomassie-stained SDS–PAGE gels (of two independent experiments) shown for reactions monitoring substrate-independent discharge of UBE2D~Ub to free lysine, in the presence of CUL1–RBX1(N98R) that was either neddylated or unneddylated (K720R), with either wild-type or the indicated UBE2D3 mutants at binding sites for backside-bound NEDD8(S22R), the RBX1 RING(F62A), and the covalently linked donor ubiquitin in the closed conformation (S108L). e, As in d, except testing the effect of NEDD8(Q40E), which would disrupt the activation module. f, Reactions performed as in d, except with indicated variants of UBE2D2, in reactions with CUL1–RBX1(N98R) that was either neddylated or unneddylated (K720R), or with the optimized RING-like U-box domain from UBE4B as a reference⁵⁸. For mutations reporting on the catalytic conformation (G24K, T36K, M38K, A96K and D112K), representative gels are shown for two experiments. All other experiments were performed once. g, Comparison of β1/β2-loop conformations after superimposing the indicated structures of NEDD8 and ubiquitin. The comparison suggests that whereas NEDD8 and ubiquitin can adopt both loop-in and loop-out conformations, donors linked to E2 active sites in RING-activated complexes adopt the loop-in conformation, and those bound to the UBE2D backside adopt loop-out conformations. h, Only a loop-out conformation is compatible with the neddylated CRL activation module structure, because loop-in conformation from the structures shown in g would prevent noncovalent interactions with the WHB domain of CUL1 (green). i, Only a loop-out conformation is compatible for the CUL1-linked NEDD8 to bind the catalytic module, because loop-in conformations in the structures shown in g would prevent noncovalent interactions with the backside of UBE2D (cyan).

Extended Data Fig. 8 Qualitative validation of the mechanistic principles that underlie substrate priming by neddylated CRLs and UBE2D.

Gel panels in this figure are representative of two independent experiments; n = 2 technical replicates. a, Schematic of a qualitative substrate priming assay for testing effects of mutations in neddylated CRL1^β-TRCP or UBE2D on substrate priming, monitoring fluorescent ubiquitin transfer from UBE2D3 to the phosphorylated IκBα substrate. b, Scan of SDS–PAGE detecting fluorescent ubiquitin transferred to the IκBα-derived substrate in a qualitative assay for NEDD8 activation of substrate priming. c, As in b, showing the effect on substrate priming of disrupting the activation module with the Q40E mutation of NEDD8. d, As in b, showing the effect on substrate priming of disrupting interactions between the activation and catalytic modules with the NEDD8(I44A) or UBE2D(S22R) mutants. e, As in b, showing the effect on substrate priming of disrupting interactions between the activation and substrate-scaffolding modules, though CUL1 modification by a ‘ubiquitylized’ NEDD8 mutant with six residues swapped for their ubiquitin counterparts (L2Q/K4F/E14T/D16E/G63K/G64E). f, As in b, showing the effect of the H32A in UBE2D at the interface between the catalytic and substrate-scaffolding modules. g, Scheme of pulse-chase assay for testing the effects of mutations in neddylated CRL1^FBW7 or UBE2D on substrate priming. The assay monitors the transfer of fluorescent ubiquitin from UBE2D to peptide substrate derived from phosphorylated cyclin E (pCyE). h, Fluorescent scan detecting ubiquitin transferred to the pCyE substrate by neddylated CRL1^FBW7 and the indicated mutants of UBE2D. i, Fluorescent scan detecting ubiquitin transferred to the pCyE substrate by UBE2D and indicated variants of neddylated (or ubiquitylated) CRL1^FBW7. Experiment with unneddylated CRL1^FBW7 used the K720R variant of CUL1 to prevent artefactual ubiquitylation. j, Scheme of pulse-chase assay for testing effects of mutations in neddylated CRL4^CRBN or UBE2D on substrate priming, monitoring fluorescent ubiquitin transfer from UBE2D to the IKZF1/3 ZF2 substrate in the presence of the immunomodulatory drug pomalidomide. k, Fluorescent scan of assay validation, showing dependence on pomalidomide. l, Fluorescent scan detecting ubiquitin transferred to the IKZF substrate by CRL4^CRBN, pomalidomide and the indicated variants of UBE2D. m–o, Fluorescent scan detecting ubiquitin transferred to the IKZF substrate by UBE2D and the indicated variants of neddylated (or ubiquitylated) CRL4^CRBN with pomalidomide. Experiments with unneddylated CRL4^CRBN used the K705R variant of CUL4A to prevent artefactual ubiquitylation.

Extended Data Table 1 Estimates of K_m and k_obs for substrate ubiquitylation

Extended Data Table 2 Cryo-EM data collection, refinement and validation statistics