Ubiquitin ligation to F-box protein targets by SCF-RBR E3-E3 super-assembly

Abstract

E3 ligases are typically classified by hallmark domains such as RING and RBR, which are thought to specify unique catalytic mechanisms of ubiquitin transfer to recruited substrates^1,2. However, rather than functioning individually, many neddylated cullin-RING E3 ligases (CRLs) and RBR-type E3 ligases in the ARIH family-which together account for nearly half of all ubiquitin ligases in humans-form E3-E3 super-assemblies^3-7. Here, by studying CRLs in the SKP1-CUL1-F-box (SCF) family, we show how neddylated SCF ligases and ARIH1 (an RBR-type E3 ligase) co-evolved to ubiquitylate diverse substrates presented on various F-box proteins. We developed activity-based chemical probes that enabled cryo-electron microscopy visualization of steps in E3-E3 ubiquitylation, initiating with ubiquitin linked to the E2 enzyme UBE2L3, then transferred to the catalytic cysteine of ARIH1, and culminating in ubiquitin linkage to a substrate bound to the SCF E3 ligase. The E3-E3 mechanism places the ubiquitin-linked active site of ARIH1 adjacent to substrates bound to F-box proteins (for example, substrates with folded structures or limited length) that are incompatible with previously described conventional RING E3-only mechanisms. The versatile E3-E3 super-assembly may therefore underlie widespread ubiquitylation.

Fig. 1. Snapshots that represent intermediates during ARIH1-catalysed ubiquitylation of the phosphorylated cyclin E substrate of SCF^FBXW7.

Coordinates were refined against high-resolution cryo-EM data (TS1 and TS2), or modelled by fitting high-resolution structures into moderate-resolution maps (pre- and post-TS1). The Rcat domain of ARIH1 traverses more than 120 Å to relocate from its autoinhibited position, collect ubiquitin from UBE2L3 and then deliver ubiquitin to a substrate bound to the F-box protein. Sub, substrate. a, Pre-TS1. Neddylated SCF-activated ARIH1 binds thioester-bonded UBE2L3~ubiquitin (Ub), mimicked by ubiquitin isopeptide bonded to a mutant UBE2L3 in which the catalytic cysteine is substituted with lysine. b, TS1. Transfer of the C terminus of ubiquitin from UBE2L3 to the catalytic cysteine (Cys^cat) of ARIH1 (in the Rcat domain) is mimicked by simultaneously linking ubiquitin and the catalytic cysteines of UBE2L3 and neddylated SCF-activated ARIH1. c, Post-TS1. The C terminus of ubiquitin thioester-bonded to the catalytic cysteine of ARIH1 was mimicked by neddylated SCF-dependent reaction of ARIH1 with ubiquitin–VME. In the major class from cryo-EM data, the only ARIH1 domain visible in the map is the Ariadne domain bound to CUL1–RBX1 in the E3–E3act super-domain. Much of ARIH1, its Rcat-linked ubiquitin, CUL1 WHB and NEDD8 are presumably dynamic. d, TS2. Transfer of ubiquitin from the catalytic cysteine of ARIH1 to the substrate bound to the F-box protein is mimicked by simultaneously linking ubiquitin, the catalytic cysteine of ARIH1 and the acceptor site on the cyclin E phosphopeptide. e, Guide to colouring of domains and proteins that participate in E3–E3-mediated ubiquitylation of substrates bound to an F-box protein. C/R, cullin–RBX; N, N terminus.

Fig. 2. Amalgamated intermolecular E3–E3act super-domain and E3–E3 platform activate neddylated SCF–ARIH1 ubiquitylation.

a, Autoinhibited ARIH1 (Protein Data Bank (PDB) code 4KBL), showing switch helix residues of the Ariadne domain securing the Rcat domain. Yellow star labels the catalytic cysteine. RBX1- and CUL1-binding regions are indicated. b, In the E3–E3act super-domain, the Ariadne domain of ARIH1 (purple) binds CUL1 (green)–RBX1 (blue), accompanied by an activating bend-to-kink remodelling of the switch helix and ARIH1 side-chain relocation (inset). c, Side-by-side comparison of UBE2L3~ubiquitin-binding elements of ARIH1 in autoinhibited (left) (grey) (PDB 5UDH) or TS1 (right) conformation (coloured as in Fig. 1e). Structures were aligned over the IBR domain. Upon binding to neddylated SCF E3 ligase, the RTI helix of ARIH1 is remodelled and other elements are rearranged. d, Intricate catalytic assembly for TS1. The E3–E3 platform, which includes CUL1-linked NEDD8, cradles UBE2L3~ubiquitin. Inset, the Ub-guided helix of ARIH1 binds the Ile44 patch of ubiquitin and guides the Rcat domain, which captures the C terminus of ubiquitin from UBE2L3. Colouring as in Fig. 1e.

Fig. 3. ARIH1 ubiquitylation of a range of substrates recruited to diverse F-box proteins.

a, Structure representing TS2: ubiquitin transfer from ARIH1 to SCF substrate. The Ub-guided helix of ARIH1 and ensuing Rcat domain bound to ubiquitin form a ubiquitin transferase module barricaded by the switch helix of ARIH1. b, Approximately 100° reorientation of ubiquitin transferase module (shown in surface) between TS1 (light pink and melon) and TS2 (dark pink and orange), shown by aligning CUL1–RBX1 and the E3–E3act super-domain of the two transition state structures. Yellow star denotes the catalytic cysteine of ARIH1. c, Cartoon representations of various F-box proteins (grey) and their substrates (red) relative to zones accessible to ubiquitin-linked active sites of ARIH1 (pink) and UBE2D (cyan), based on structural modelling. K, substrate lysine. d, Graphs showing the mean value of catalytic efficiencies (k_obs/K_m) for ARIH1 (pink)- and UBE2D3 (cyan)-mediated ubiquitylation with indicated substrate and F-box protein, as depicted in c. n = 3 independent experiments.

Fig. 4. Proposed transient amalgamation–ubiquitylation–disassembly cycle of SCF–ARIH1 RBR-type E3–E3 ligase.

Model for neddylated SCF–ARIH1 E3–E3 ubiquitylation. Portions of proteins that contribute to the transition states are coloured as in Fig. 1e. (1) ARIH1 is autoinhibited, and SCF is unassembled and unneddylated. (2) NEDD8 linkage persists on substrate-bound SCF E3 ligase, enabling amalgamation with ARIH1 via the E3–E3act super-domain and E3–E3 platform. The switch helix of the Ariadne domain of ARIH1 is twisted, and the Rcat domain of ARIH1 is freed from autoinhibition. (3) In TS1, the Ariadne domain of ARIH1 binds CUL1–RBX1 in the E3–E3act super-domain, and NEDD8-bound ARIH1 elements are remodelled in the E3–E3 platform. The E3–E3 platform displays the linked ubiquitin of UBE2L3, which lures the Ub-guided helix of ARIH1 to promote the capture of UBE2L3-linked ubiquitin by the catalytic cysteine of ARIH1. The Ub-guided helix and Rcat domains of ARIH1, and bound ubiquitin, form a ubiquitin transferase module. (4) In TS2, the ubiquitin transferase module has undergone an approximately 100° translocation to deliver ubiquitin to the substrate bound to the F-box protein. (5) Without bound ubiquitin, the Ub-guided helix of ARIH1 could resume an autoinhibitory conformation. E3–E3 disassembles.

Extended Data Fig. 1. Stable mimics for fleeting E3–E3 ubiquitylation intermediates and transition states (TS1 and TS2).

a, Neddylated SCF–ARIH1 RBR E3–E3-catalysed ubiquitylation of an F-box-protein-bound substrate proceeds through two transition states. First, the C terminus of ubiquitin is transferred from the catalytic cysteine of the E2 UBE2L3 to the catalytic cysteine of ARIH1 (TS1), and then from ARIH1 to substrate (TS2). b, SDS–PAGE gels (left, nonreducing; right, reducing) of rapid-quench flow experiments resolving intermediates in fluorescent ubiquitin transfer from UBE2L3 to ARIH1 to cyclin E phosphopeptide substrate of neddylated SCF^FBXW7 (CUL1–SKP1–FBXW7(ΔD)) (millisecond time scale). Gel images are representative of independent biological replicates (n = 2). c, Quantification of results obtained from rapid-quench flow and monitoring fluorescent ubiquitin transfer, with or without cyclin E substrate included in the reaction. d, Chemical structures of native ubiquitylation intermediates and stable mimics used in this study. In the pre-TS1 intermediate, the C terminus of ubiquitin is linked to the catalytic cysteine of UBE2L3 by a thioester bond; in the stable mimic, the C terminus of ubiquitin is linked by an isopeptide bond to a lysine replacement for the catalytic cysteine of UBE2L3. The TS1 intermediate is mimicked by using an ABP with an electrophilic moiety installed between the C terminus of ubiquitin and the catalytic cysteine of UBE2L3 to trap the catalytic cysteine of ARIH1 via a stable three-way cross-link. Following the native TS1 reaction, in the post-TS1 intermediate, ubiquitin is thioester-bonded to the catalytic cysteine of ARIH1. The stable mimic used ubiquitin–VME to stably couple ubiquitin to the catalytic cysteine of ARIH1. In the TS2 reaction, ubiquitin is transferred from ARIH1 to the substrate. Our electrophilic TS2 ABP forms a stable mimic by simultaneously linking the catalytic cysteine of ARIH1, the C terminus of ubiquitin and the acceptor site on a peptide substrate. e, SDS–PAGE gel monitoring the formation of the stable TS1 mimic, in which ARIH1, ubiquitin and UBE2L3 are linked via a single atom. For comparison, reactions were carried out at the same concentrations of ARIH1 and neddylated CUL1–RBX1, and similar buffer and temperature conditions as in b. Gel image is representative of independent technical replicates (n = 2). Higher protein concentrations and extended times were used to generate samples for cryo EM. f, SDS–PAGE gel tracking neddylated SCF-dependent generation of the stable TS2 mimic, in which the catalytic cysteine of ARIH1, the C terminus of ubiquitin and the peptide substrate are linked. Gel image is representative of independent technical replicates (n = 2).

Extended Data Fig. 2. Synthesis of ABPs to capture TS1 and TS2.

a, Strategy to generate an ABP to visualize TS1 (TS1 ABP). The goal was to generate an ABP with a warhead between the catalytic cysteine of UBE2L3 and the C terminus of ubiquitin that would react with ARIH1 only when assembled with a neddylated CRL. An intein-based semisynthesis route was used to couple Ub(1–75)–MESNa and (E)-3-[2-(bromomethyl)-1,3-dioxolan-2-yl]prop-2-en-1-amine (BmDPA) to yield a cyclic ketal-protected ubiquitin species. Acidic deprotection of the cyclic ketal yields a reactive ubiquitin species, which when conjugated to a single-cysteine-containing version of UBE2L3 produces an ABP with a Michael acceptor between the C terminus of ubiquitin and the active site of UBE2L3. b, Quality controls comparing predicted masses for Ub–MESNa, ABP precursor and TS1 ABP entities with measurements obtained by electrospray ionization–time-of-flight mass spectrometry. c, SDS–PAGE gel confirming TS1 ABP reaction depends on the catalytic cysteine of ARIH1 (C>S refers to serine replacement). Gel image is representative of independent technical replicates (n = 2). d, SDS–PAGE gel demonstrating TS1 ABP reaction with ARIH1 depends on neddylated CUL1–RBX1 and ARIH1 residues required for ubiquitylating client substrates bound to an F-box protein. Gel image is representative of independent technical replicates (n = 2). e, Because structural biology is an empirical endeavour, various TS2 ABP approaches were tested to identify a strategy yielding high-quality electron microscopy data visualizing TS2. The concept was to place a warhead between a substrate and the C terminus of ubiquitin, to generate an ABP that would react only with ARIH1 super-assembled with the SCF containing the cognate F-box protein of the substrate. The fully synthetic TS2 ABP alternative (left) displayed reactivity and specificity matching the native reaction when assembled with cyclin E or p27 phosphopeptide substrate mimics, as shown by SDS–PAGE gel (right). f, In parallel, we tested a semisynthetic strategy, which led to high-resolution cryo-EM structures, and thus complexes generated with this strategy are referred to as TS2 throughout the article. Ub–MESNa and a substrate phosphopeptide with an N-terminal cysteine placed to mimic the acceptor site were fused via native chemical ligation and the free cysteine was converted to dehydroalanine. g, SDS–PAGE showing TS2 p27 ABP reaction with ARIH1 requires all elements needed for native TS2, or use of a mutant version of ARIH1 (ARIH1(F430A/E431A/E503A)) bypassing the need for NEDD8 for this reaction. Gel image is representative of independent technical replicates (n = 2). h, SDS–PAGE testing specificity of TS2 ABPs for cognate F-box proteins. Phosphorylated cyclin E and p27 are substrates of SCF^FBXW7 and SCF^SKP2, respectively. All experiments with SCF^SKP2 also contained the essential protein partner CKSHS1 unless otherwise indicated. Gel image is representative of independent technical replicates (n = 2).

Extended Data Fig. 3. Cryo-EM map quality analysis.

a, Cryo-EM map representing TS1, coloured by local resolution in Å as estimated by ResMap. The map was generated using our TS1 ABP for the complex representing UBE2L3~ubiquitin~ARIH1 bound to neddylated CUL1–RBX1–SKP1–SKP2–CKSHS1–p27–cyclin A–CDK2. b, Fourier shell correlation curve (FSC) displaying an overall resolution of 3.83 Å with the FSC = 0.143 criterion. c, Structure shown in electron microscopy density. Left, close-up of E3–E3act domain, showing side-chain density for interactions between the RING domain of RBX1 and the Ariadne domain of ARIH1. Middle, close-up of E3–E3 platform showing interactions with UBE2L3-linked ubiquitin. Right, ubiquitin transferase module. The map quality permitted modelling of side chains either visible in the density or by wholesale docking of previous crystal structures. d, Cryo-EM map representing TS2, coloured by local resolution in Å as estimated by ResMap. The map was generated using our TS2 p27 ABP for the complex representing ARIH1~ubiquitin~p27 bound to CUL1–RBX1–SKP1–SKP2-CKSHS1. This particular map is from a complex using ARIH1(F430A/E431A/E503A), which was previously shown to relieve autoinhibition and bypass the need for neddylation. e, FSC curve displaying an overall resolution of 3.91 Å with the FSC = 0.143 criterion. f, Structure shown in electron microscopy density, representing highest, medium and lowest resolution areas of the map. Left, close-up of E3–E3act domain, showing side-chain density for interactions between the RING domain of RBX1 and the Ariadne domain of ARIH1. Middle, portion of CUL1. Right, ubiquitin transferase module. After the ubiquitin transferase module was modelled and refined for the structure representing TS1, it was wholesale-docked into the lower-resolution density for this region in the cryo-EM maps representing TS2.

Extended Data Fig. 4. Cryo-EM maps representing intermediates in neddylated SCF-dependent ubiquitin transfer from UBE2L3 to ARIH1 and from ARIH1 to substrates recruited by structurally diverse F-box proteins.

a, Low-pass-filtered cryo-EM map representing intermediate with ubiquitin linked to E2 UBE2L3, before transfer to ARIH1 bound to neddylated SCF (pre-TS1). Model was generated by docking the following structures into the higher-resolution (4.5 Å) map: SKP1–FBXW7–cyclin E (PDB 2OVQ), and UBE2L3~ubiquitin bound to ARIH1-neddylated CUL1–RBX1 refined from the structure representing TS1. The Ub-guided helix and Rcat domain of ARIH1 are not visible in this intermediate and were removed before fitting. b, Low-pass-filtered cryo-EM maps representing TS1 intermediate, neddylated SCF-dependent ubiquitin transfer from UBE2L3 to ARIH1, for SCF ligases with two structurally divergent F-box proteins. Left, ribbon diagram for the final refined model for one of these SCF^SKP2.The concave leucine-rich repeat domain of the F-box protein SKP2 enwraps its partner CKSHS1 to corecruit the intrinsically disordered C-terminal domain of phosphorylated p27. This assembly is sufficient for efficient phospho-p27 ubiquitylation in vitro, but the complex can be further augmented by additional protein–protein interactions. CKSHS1 can also bind cyclin A–CDK2, which also binds the N-terminal domain of p27^,,. This TS1 structure comprises a p27 N-terminal domain-bound cyclin A–CDK2, phospho-p27 C-terminal-domain-bound CKSHS1–SKP2–SKP1 bound to a neddylated CUL1–RBX1-activated ARIH1~ubiquitin~UBE2L3 assembly generated with the TS1 ABP. The right panel shows a comparable assembly with the monomeric (ΔD) version of the F-box protein FBXW7, in which the top side of the eight-bladed WD40 β-propeller domain of FBXW7 recruits a phosphopeptide substrate derived from cyclin E. The model was made by docking SKP1–FBXW7–cyclin E (PDB 2OVQ) and catalytic portions of the final refined model for neddylated CUL1–RBX1-activated ARIH1~ubiquitin~UBE2L3 TS1 complex with SKP2 on the left. c, Cryo-EM data for the post-TS1 intermediate, with ubiquitin linked to the catalytic cysteine of ARIH1, yielded three classes. (1) In the most abundant ‘E3–E3act super-domain only’ class (middle), only the E3–E3act super-domain and SCF scaffold are visible. (2) A ‘pre-TS1 class’ largely resembles the pre-TS1 assembly, as the E3–E3act super-domain and E3–E3 platform as well as the E2 UBE2L3 are visible. The ubiquitin transferase module is not visible. (3) A low-resolution ‘TS2 class’ resembles the structural architecture of TS2, with density presumably corresponding to the ubiquitin transferase module (the ubiquitin-guided helix and Rcat domain of ARIH1 bound to ubiquitin) poised adjacent to the F-box-protein-bound substrate. d, Low-pass-filtered cryo-EM maps for some TS2 complexes, representing ubiquitin transfer from ARIH1 to an SCF-bound substrate. Top left, ribbon diagram for the final refined model for one of these, with SKP1–CUL1–SKP2. This structure comprises ARIH1(F430A/E431A/E503A)~ubiquitin~phospho p27 peptide (generated with a TS2 ABP) bound to CKSHS1–SKP2–SKP1–CUL1–RBX1. ARIH1(F430A/E431A/E503A) has previously been shown to relieve autoinhibition and bypass the need for neddylation. Top middle and top right, cryo-EM maps from the complex representing neddylated SCF^FBXW7-dependent ubiquitin transfer from ARIH1 to cyclin E phosphopeptide substrate (captured with our cyclin-E-based TS2 cyclin E ABP). The ribbon diagram corresponds to SKP1–FBXW7(ΔD)–cyclin E from a previous crystal structure, CUL1–RBX1–ARIH1~ubiquitin from the refined structure with SKP2 (top left), and for the unrefined class, the neddylated CUL1–ARIH1 platform (the UBAL, RING1, RTI helix and IBR of ARIH1 and NEDD8 isopeptide-linked to the WHB domain of CUL1) from the refined structure of TS1 (top right). The latter portion of the complex is not visible upon refinement to higher resolution (top right). Bottom left, low-pass-filtered cryo-EM map and docked structures for SCF^FBXW7 and ARIH1(F430A/E431A/E503A) captured with our cyclin-E-based TS2 ABP, which superimposes with the refined class from neddylated SCF^FBXW7 and wild-type ARIH1, thus validating the use of ARIH1(F430A/E431A/E503A) to improve cryo-EM map quality for TS2. Middle, right, two conformations of TS2 obtained by major classification of the cryo-EM data are shown. Although the conformations differ in their relative positions of the substrate-bound F-box protein FBXW7(ΔD) and ubiquitin transferase module, in both the ubiquitin-guided helix and Rcat domain of ARIH1 direct the C terminus of ubiquitin towards the F-box-protein-bound substrate. e, The catalytic architecture for TS2 is conserved irrespective of the TS2 ABP probe strategy. Cryo-EM map using TS2 ABP alternative (Extended Data Fig. 2e) is shown in grey, with docked model from corresponding SCF^FBXW7–ARIH1(F430A/E431A/E503A)~ubiquitin~cyclin E complex captured with the TS2 cyclin E probe (d bottom left) (Fig.1d). The low-pass-filtered cryo-EM map obtained with the cyclin-E-based TS2 is shown in violet, superimposed on the right.

Extended Data Fig. 5. Structural rearrangements and implications of super-assembly for neddylated SCF and RBR E3 ligases.

a, Close-ups showing a common RBX1 RING surface binding to ARIH1 Ariadne domain in the E3–E3act super-domain and binding to the E2 UBE2D (PDB 6TTU). Structures on left align CUL1–RBX1–ARIH1 and CUL1–RBX1–UBE2D over RBX1, and on right over CUL1. b, Unique RBX1 RING (blue) and CUL1 (green) arrangement in E3–E3act domain. For reference, the relative RBX1 RING orientations are shown for neddylation in teal (RBX1 N, PDB 4P5O), and in an inhibited complex in sky blue (RBX1 I, PDB 4F52). c, Close-ups comparing relative orientations of NEDD8 and the isopeptide-linked WHB domain of CUL1 in TS1 (yellow–green) with the ‘activation module’ activating conventional UBE2D-dependent ubiquitylation of an SCF^β-TRCP substrate (PDB 6TTU) (grey) (left), or during neddylation (PDB 4P5O) (grey) (middle), or captured by crystal packing in a structure of neddylated CUL5–RBX1 that revealed orientational flexibility of neddylated CUL WHB and RBX RING domains (PDB 3DQV) (grey) (right). NEDD8 is superimposed across the different structures. d, Relative to CUL1 scaffold, NEDD8 (yellow)-linked CUL WHB domain (dark green) position in TS1 compared to positions of these domains from structures shown in c. After superimposing the CUL–RBX C/R domains, positions of NEDD8 are shown in light yellow and of its linked CUL1 WHB domain in dark green from structure representing UBE2D-dependent ubiquitylation of an SCF^β-TRCP substrate (PDB 6TTU) (left), or during neddylation (PDB 4P5O) (middle) or captured by crystal packing in a structure of neddylated CUL5–RBX1 that revealed orientational flexibility of neddylated CUL WHB and RBX RING domains (PDB 3DQV) (right). Dotted lines show regions of structures connected but not modelled owing to lack of density. e, Comparison of structure and location of ARIH1 element corresponding to Ub-guided helix in TS1 (light pink) and TS2 (dark pink) relative to that linking the Ariadne and Rcat domains in the autoinhibited configuration (grey). For perspective, the helix preceding the Ub-guided helix, which is similar in all of the structures, is also coloured. f, In the context of transient E3–E3 assembly, neddylated SCF-activated UBE2L3~ubiquitin~ARIH1 TS1 adopts the canonical activated RBR configuration. g, Corresponding region of structure with RBR E3 HOIP (PDB 5EDV). h, Corresponding region of structure with PARKIN (PDB 6N13). The structure of HOIP is represented as a monomer, although it is a domain-swapped dimer in the crystal. A HOIP ubiquitin-guided helix was previously noted to promote ubiquitin transfer from an E2 to this RBR E3 ligase. This concept could possibly apply in PARKIN as well.

Extended Data Fig. 6. Mutational validation of structural mechanism of neddylated SCF–ARIH1-catalysed ubiquitylation.

a, Close-up of E3–E3act domain with spheres showing locations of strongly defective (red), marginally defective (orange) and hyperactive (green) mutations identified by previous ARIH1 Ala scanning mutagenesis^,. Defective mutants map to key CUL1- and RBX1-binding residues, whereas hyperactive mutants map to the site of activating bend-to-kink conformation change within the switch helix. b, Close-up of CUL1-WHB-domain-linked NEDD8 interactions with ARIH1 UBAL domain in TS1, showing locations of strongly defective mutants as red spheres. Mutants in ARIH1 UBAL domain (V123D or F150A) at interface with NEDD8 were previously described. A representative SDS–PAGE gel of experiments testing effects of mutating NEDD8 residues at interface with ARIH1 or with CUL1 in the context of neddylated CUL1–RBX1-activated ubiquitin transfer from UBE2L3 to ARIH1 is shown below. Gel image is representative of independent technical replicates (n = 2). c, Close-up of catalytic elements for TS1 (ubiquitin transfer from E2 UBE2L3 to ARIH1 catalytic cysteine (yellow star)). Red spheres show sites of previously identified strongly defective mutants, or tested on the basis of the structure representing TS1. SDS–PAGE gel (left) shows neddylated CUL1–RBX1-activated ubiquitin transfer from UBE2L3 to ARIH1 or (right) from UBE2L3 via ARIH1 to phosphopeptide substrate derived from cyclin E, testing effects of ARIH1 mutants in the ubiquitin-guided helix preceding the Rcat domain. These residues are markedly remodelled for TS1 and were not tested in the previous ARIH1 Ala scanning mutagenesis study. Gel images are representative of independent technical replicates (n = 2). d, Close-up of catalytic elements for TS2, ubiquitin transfer from ARIH1 to SCF-bound substrate. Spheres indicate sites of mutation defective in achieving substrate targeting configuration. Sites of blue mutations map to region in switch helix contributing to ARIH1 autoinhibition and substrate targeting, and accordingly lead to accumulation of ARIH1~ubiquitin in assays monitoring fluorescent ubiquitin transfer from UBE2L3 to ARIH1 to a neddylated SCF substrate (non-reducing conditions, −DTT (botttom left); reducing conditions, +DTT (bottom right)). Gel images are representative of independent technical replicates (n = 2). e, SDS–PAGE gels monitoring fluorescent ubiquitin transfer to cyclin E phosphopeptide, for ARIH1 triple-glycine mutants probing roles of the N-terminal end of the switch helix of the Ariadne domain. The mutations are upstream of the triple-glycine mutants shown as blue spheres in d. Gel images are representative of independent technical replicates (n = 3). f, SDS–PAGE gels of assays testing whether mutants adopt configuration for ubiquitin transfer from ARIH1 to substrate, probed by reaction with cyclin-E-based TS2 ABP. Gel image is representative of independent technical replicates (n = 2).

Extended Data Fig. 7. Comparison of structural mechanisms of neddylated SCF substrate ubiquitylation by ARIH1 (E3–E3) or conventional ubiquitylation with UBE2D (E2–E3).

a, Side-by-side comparison of E3–E3-catalysed substrate ubiquitylation, via the two transition states (left), versus conventional E2–E3 mechanism (right). Structures on left represent neddylated SCF^FBXW7-dependent ubiquitin transfer from UBE2L3 to ARIH1 (TS1) and from ARIH1 to cyclin E substrate (TS2) on the basis of electron microscopy data shown in Extended Data Fig. 4b, d. These structures are shown in two different relative orientations to highlight key features of TS1 or TS2 reactions. Structure on right shows neddylated SCF^β-TRCP-dependent ubiquitin transfer from the E2 UBE2D to IκBα (PDB 6TTU). The structures are aligned over CUL1. In the E3–E3 mechanism, ubiquitin linked to the Rcat domain of ARIH1 is projected towards the substrate irrespective of F-box protein identity. However, in the E2–E3 mechanism, optimally positioned UBE2D specifically contacts β-TRCP, contributing to the notable catalytic efficiency of the conventional E2–E3 mechanism for neddylated SCF^β-TRCP ubiquitylation of several of its substrates. b, ARIH1~ubiquitin active site faces F-box-protein-bound substrate within a confined zone. Structures from two electron microscopy classes are shown (Extended Data Fig. 4d). After superimposing CUL1–RBX1–ARIH1~ubiquitin from both classes, the substrate (red)-bound F-box protein in one is shown in purple (conformation 1) and the other in grey (conformation 2). c, E3–E3 catalytic configuration is generalizable for substrates recruited to structurally diverse F-box proteins: p27 recruited to SKP1–SKP2–CKSHS1 (left) and cyclin E recruited to SKP1–FBXW7 (right). d, Structural modelling and comparison of E3–E3 versus E2–E3-mediated ubiquitylation with cyclin E as a substrate. The structure on the left corresponds to SKP1–FBXW7–cyclin E (PDB 2OVQ), fitted into map corresponding to conformation 1, with neddylated CUL1–RBX1-activated ARIH1~ubiquitin~substrate from the refined structure representing TS2 for SCF^SKP2. Proximity of the ubiquitin transferase domain to the substrate phospho-degron explains how ARIH1 efficiently ubiquitylates a ‘short’ cyclin E substrate, with only four residues between the phospho-degron and acceptor lysine (Fig. 3d, Extended Data Fig. 8). On the right is a model of E2–E3-mediated ubiquitylation of cyclin E by neddylated SCF^FBXW7. The model was generated by aligning the SKP1–F-box portion of SKP1–FBXW7–cyclin E (PDB 2OVQ) in place of the corresponding region showing UBE2D-mediated ubiquitylation of a substrate of neddylated SCF^β-TRCP (PDB 6TTU), which shows the distance separating the catalytic cysteine of UBE2D and the cyclin E substrate acceptor. This rationalizes the inefficient ubiquitylation of the short cyclin E peptide substrate by the conventional E2–E3 mechanism (Fig. 3d, Extended Data Fig. 8). e, As in d, but modelled with SKP1–FBXL3–CRY2 (PDB 4I6J) based on the F-box proteins. f, Competition assay testing whether a neddylated SCF ligase can mediate conventional UBE2D-catalysed ubiquitylation if occupied by ARIH1. SDS–PAGE gels monitor neddylated SCF^FBXW7-dependent transfer of fluorescent ubiquitin from UBE2D3 to cyclin E phosphopeptide substrate. Ubiquitylation is severely hindered upon addition of the stable proxy for the E3–E3 post-TS1 intermediate (ARIH1~ubiquitin, generated by ARIH1 reaction with ubiquitin–VME) to the reaction, but not when ARIH1 is added on its own (right). The results are rationalized by the same portions of RBX1 and NEDD8 binding ARIH1 and UBE2D (Extended Data Fig. 5a, c). A catalytically inactive mutant version of ARIH1 in which the catalytic cysteine is substituted with Ser was used to prevent any potential spurious activity from ARIH1. Gel image is representative of independent technical replicates (n = 2). g, SDS–PAGE gels monitor neddylated SCF^FBXW7-dependent transfer of fluorescent ubiquitin from UBE2R2 to ubiquitin-linked cyclin E phosphopeptide substrate, testing competition upon adding the stable proxy for the E3–E3 post-TS1 intermediate (ARIH1~ubiquitin) to the reaction, or inactive ARIH1 on its own (right). Gel image is representative of independent technical replicates (n = 2).

Extended Data Fig. 8. Quantitative kinetics show ARIH1 efficiently ubiquitylates diverse SCF-bound substrates, whereas UBE2D3 activity is relatively specialized.

a, Plots of the fraction of substrate that had been converted to ubiquitylated products versus concentration of ARIH1 (magenta) or UBE2D3 (cyan). Various neddylated SCF complexes were assayed that contained the F-box proteins FBXL3, β-TRCP2 (referred to throughout the Article as β-TRCP) or FBXW7 (ΔD version) and substrates cryptochrome 1 (CRY1) (relative position of substrate fixed by its folded structure), phosphopeptides derived from β-catenin (long (containing a lysine in a known substrate position) or short (separated from the degron by only a four-residue linker)) or phosphopeptides derived from cyclin E (long (containing a lysine in a known substrate position), short (separated from the degron by only a four-residue linker) or ubiquitylated (contains a single ubiquitin fused by sortase-mediated transpeptidation)). Triplicate data points from independent experiments are shown and were fit to the Michaelis–Menten model to estimate the K_m of ARIH1 or UBE2D3 for the respective neddylated SCF–substrate complexes using nonlinear curve fitting (GraphPad Prism). b, Plots comparing various neddylated SCF ligases and substrates described in a, showing appearance of ubiquitylated substrate (CRY1) or disappearance of unmodified substrate in rapid-quench flow reactions all performed under single-encounter conditions. Triplicate data points from independent experiments are shown. The data were fit to closed form equations (Mathematica) as previously described to obtain rates of ubiquitylation (k_obs) as well as their associated standard error (Extended Data Table 2).