Time-resolved cryo-EM (TR-EM) analysis of substrate polyubiquitination by the RING E3 anaphase-promoting complex/cyclosome (APC/C)

Abstract

Substrate polyubiquitination drives a myriad of cellular processes, including the cell cycle, apoptosis and immune responses. Polyubiquitination is highly dynamic, and obtaining mechanistic insight has thus far required artificially trapped structures to stabilize specific steps along the enzymatic process. So far, how any ubiquitin ligase builds a proteasomal degradation signal, which is canonically regarded as four or more ubiquitins, remains unclear. Here we present time-resolved cryogenic electron microscopy studies of the 1.2 MDa E3 ubiquitin ligase, known as the anaphase-promoting complex/cyclosome (APC/C), and its E2 co-enzymes (UBE2C/UBCH10 and UBE2S) during substrate polyubiquitination. Using cryoDRGN (Deep Reconstructing Generative Networks), a neural network-based approach, we reconstruct the conformational changes undergone by the human APC/C during polyubiquitination, directly visualize an active E3-E2 pair modifying its substrate, and identify unexpected interactions between multiple ubiquitins with parts of the APC/C machinery, including its coactivator CDH1. Together, we demonstrate how modification of substrates with nascent ubiquitin chains helps to potentiate processive substrate polyubiquitination, allowing us to model how a ubiquitin ligase builds a proteasomal degradation signal.

Fig. 1. TR-EM reveals structure of APC/C–CDH1–UBE2C–UBE2S^CTP during substrate polyubiquitination.

a, Cartoon representation of reaction cycle carried out by the APC/C as it ubiquitinates substrates. The transition from the ‘CRL (cullin–RING) down’ inactive conformation to the ‘CRL up’ active conformation is facilitated by coactivator binding. Initiation of ubiquitination on target substrates is carried out by the recruitment of UBE2C~Ub by the APC2 WHB (winged-helix B) and APC11 RING domains. UBE2S elongates chains initiated by UBE2C. b, Overview of TR-EM approach. Two mixtures containing the reaction components are incubated at room temperature (1). The mixtures are combined to start the reaction (2). Samples are taken from the reaction at timepoints indicated and applied to a grid (3) that is then plunge frozen (4) and imaged using cryo-EM. Representative samples were taken at the time the grids were plunge frozen for SDS–PAGE (5), shown in c. c, Fluorescent monitoring of an SDS–PAGE gel showing APC/C^CDH1-E2 (UBE2C or both UBE2C and UBE2S)-dependent substrate modification at each timepoint for which a grid was frozen. At least three experimental replicates were repeated to optimize conditions. Uncropped gel image available in source data. d, 3D reconstruction of particles from the filtered APC/C–CDH1–UBE2C–UBE2S dataset at 3.5 Å showing the active catalytic architecture assembled. Atomic models of key subunits and ubiquitination components fitted into the cryo-EM density, including CDH1 bound to the APC/C scaffold (i and ii), UBE2C clamped by the APC2 WHB and APC11 RING domains (iii), and the UBE2S CTP bound to the groove formed by APC2 (green) and APC4 (light pink) (iv). Source data

Fig. 2. Reconstruction of a functional conformational landscape of APC/C–CDH1-mediated polyubiquitination using cryoDRGN.

a, PCA representation of the particle distribution in latent space with explained variance (EV) noted in parentheses. The first component captures the continuous conformational changes the APC/C undergoes in the ‘CRL down’ to ‘CRL up’ transition. Density for APC2, APC11 and CDH1 are shown at indicated points as green, blue and purple, respectively. b, 3D models generated using homogeneous 3D refinement of each set of major state particles. Density maps are colored by proximity to the subunit indicated as modeled using rigid body fitting. c,d, Left: UMAP representation of the particle distribution in latent space clustered using k-means (k = 500) clustering and then classified into the four major APC/C states present within the APC/C^CDH1–UBE2C (c) and APC/C^CDH1–UBE2C–UBE2S (d) datasets. Points denote cluster centers where volumes were generated. Right: subplots of particle distributions separated by major states. Particle distributions for each class are colored by their local density overlayed onto the particle distribution for the entire dataset (gray). e,f, Top: charts show distribution of particles across major states for the APC/C^CDH1–UBE2C (e) and APC/C^CDH1–UBE2C–UBE2S (f) datasets. Bottom: line graphs show change in relative fraction of each major state over time in the two datasets.

Fig. 3. UBE2S CTP enhances the APC/C ‘CRL up’ state, increasing the association of UBE2C~Ub with APC/C–CDH1.

a,b, Quantification of substates within the APC/C ‘CRL up’ and ‘CRL down’ major states assigned from 500 sampled cryoDRGN volumes. ‘CRL down’ clusters partition into a majority of particles where CDH1 is absent (~80%). The ‘CRL up’ particles contain a minor subset of particles where the CRL is only partially in the ‘up’ position (a). Addition of UBE2S increases the relative number of particles with density attributed to UBE2C present in the clusters compared to the absence of UBE2S (b). c, Schematic of TIRF microscopy substrate immobilization and the localization of ubiquitination reaction components. Alexa488-labeled CycB^N is immobilized to a neutravidin-functionalized slide through an N-terminal, biotinylated Avitag (captured using a 488 nm laser; substrate channel). APC/C–CDH1 will localize to the substrate and subsequently recruit fluorescent UBE2C–JF549 (captured using a 561 nm laser; UBE2C channel) and a UBE2S^CTP peptide that binds the APC2/APC4 groove. Thus, the number of binding events in the UBE2C channel correlates to UBE2C localizing to APC/C–CDH1–substrate. d, Titrating the UBE2S^CTP increases the recruitment of UBE2C to APC/C–CDH1, localized to substrate at the slide surface. Quantification of the total number of UBE2C binding events (Extended Data Fig. 5f) from three independent experiments is shown (see Supplementary Video 2 for representative movies and Extended Data Fig. 5g,h for single molecule binding event montage and representative traces). Error bars: standard error of the mean. Data are normalized to the control where APC/C-CDH1 is absent. Experiments were compared using one-way analysis of variance (*P = 0.0176). Source data

Fig. 4. Analysis of active APC/C-dependent ubiquitination architecture reveals unexpected Ub binding modes.

a, Focused 3D classification on CDH1, UBE2C and APC2/APC11 using a mask to generate structures of the active APC/C. Discrete states were found showing a Ub interacting with the known APC11 RING exosite (left) and forming contacts with CDH1 (center left) and UBE2C (center right). A class of particles also contained a structure with two Ubs making simultaneous contact with the coactivator and UBE2C simultaneously (right). b, Phage display selected a UbV (UbV^CDH1) that binds to APC/C coactivators. c, Sequence alignment of Ub and UbV^CDH1 with differences highlighted. d, Coomassie-stained SDS–PAGE gel showing binding of GST–UbV^CDH1 towards APC/C coactivators (CDH1 and CDC20), but not GST–Ub, after GST pulldown. n = 3 independent experiments. e, Degradation of APC/C substrates Cyclin A, Cyclin B, Geminin and Securin in mitotic HeLa cell extracts was inhibited by the addition of UbV^CDH1, but not Ub, as monitored by immunoblotting. n = 3 independent experiments. f, Attachment of the Hsl1 D-box, but not the Hsl1 KEN-box, to UbV^CDH1 potentiates its inhibition of APC/C–CDH1–UBE2C-mediated substrate ubiquitination, monitored in three independent experiments by fluorescent scanning of an SDS–PAGE gel. g, Left: the cryo-EM structure of APC/C-CDH1 bound to the UbV^CDH1(orange)–Hsl1 D-box (red) shows the localization of UbV^CDH1 near the KEN-box-binding site on the CDH1 β-propeller (purple). Right: previously published cryo-EM map of APC/C bound to CDH1 and Hsl1 (EMD-2651) for comparison. Uncropped gels representative of n = 3 independent experiments for d–f available in source data. Source data

Fig. 5. Multiple Ub-binding sites promote processive ubiquitination by APC/C^CDH1.

a, Cartoon model showing binding sites of CDH1-binding (UbV^CDH1) and APC11-binding (UbV^RING) UbVs. b, Workflow for single-encounter experiments. APC/C^CDH1 and *CycB^N containing multiple lysines or a single lysine (1K) are preincubated. The UbVs were added to this mixture. A second mix is prepared containing E1, UBE2C, MgATP, an excess of unlabeled Hsl1, and Ub or methylated Ub (meUb). The mixtures are combined, resulting in fluorescent substrate ubiquitination during a single binding event. c, Fluorescent scanning of an SDS–PAGE gel (top) and quantification (bottom) from single-encounter APC/C–CDH1–UBE2C-dependent polyubiquitination reactions with CycB^N* and Ub. Dashed box indicates the region of CycB^N* with >4 Ubs quantified in Extended Data Fig. 7a. n = 3 independent experiments. Error bars: standard error of the mean. d, Fluorescent scanning of an SDS–PAGE gel (top) and quantification (bottom) from single encounter of APC/C–CDH1–UBE2C-dependent substrate priming reactions with CycB^N*(1K) and meUb. n = 3 independent experiments. Error bars: standard error of the mean. e, APC/C–CDH1-dependent ubiquitination reactions of Ub–Securin with its KEN- and D-box degrons mutated (Ub–Securin KD^mut*) using either UBE2C (left) or UBE2S (right), monitored by fluorescent scanning. n = 3 independent experiments. f, Ubiquitination of fluorescently labeled *UbV^CDH1 by APC/C–UBE2C is dependent on CDH1 and the APC2 WHB, monitored by fluorescent scanning. n = 3 independent experiments. g, METRIS setup to monitor binding of APC/C to a biotinylated substrate, Ub–substrate or Ub on a streptavidin-coated surface. Magnetic beads with biotinylated APC/C are subjected to a magnetic field. The movement of individual beads is quantitated to determine a rolling parameter (RP) that correlates with the strength of the interaction. h,i, Scatter plots quantitating the RP of the APC/C on surfaces containing CycB^N or Ub-CycB^N (h) or Ub (i) in the presence of UbV^RING, UbV^CDH1 or both. Experiments were compared using one-way analysis of variance (****P ≤ 0.0001, ***P ≤ 0.001, **P ≤ 0.005, *P ≤ 0.05). NS, not significant. j, Schematic of APC/C–CDH1–UBE2C–UBE2S-dependent substrate polyubiquitination. The APC/C catalytic architecture and recruitment of UBE2C~Ub (dUb) is influenced by CDH1 and UBE2S CTP binding to drive ubiquitination efficiency. Allosterically driven processivity of ubiquitination is further mediated by stabilizing contacts between multiple substrate-linked Ubs and CDH1 and APC11. Uncropped gels representative of n > 3 independent experiments for c–f available in source data. Source data

Extended Data Fig. 1. Processing and fitting of the APC/C from the TR-EM datasets.

a – Diagram depicting domain map of Ub-CycB^N*. b, c – Cryo-EM data processing workflow showing representative micrographs from the APC/C-CDH1-UBE2C and APC/C-CDH1-UBE2C-UBE2S datasets (B). Particle detection was optimized using Topaz and filtered using 3 rounds of 2D classification. Representative 2D projections from the final 2D classification step show the APC/C in multiple orientations (C). d – Local resolution map of the consensus refinements from the datasets. e – FSC curves showing resolution of the 3D density maps from the consensus refinement from the APC/C-CDH1-UBE2C and APC/C-CDH1-UBE2C-UBE2S datasets. Lines show the 0.143 gold-standard and 0.5 cut-off with values denoting resolution at 0.143 cut-off. f – Similar to Fig. 1d, fitted model of APC/C-CDH1-UBE2C. g – Overlay of APC/C-CDH1-UBE2C electron density map from this study with published maps from previous studies. Values shown indicate map-to-map correlation of maps from previous studies (EMD-2925 and EMD-2929) with APC/C-CDH1-UBE2C from this data after gaussian filtering^,. h – Overlay of electron density maps from the APC/C-CDH1-UBE2C and APC/C-CDH1-UBE2C-UBE2S consensus refinements, value shown indicates map-to-map correlation. i – Models showing UBE2S-CTP bound to the APC2/APC4 groove in previously described datasets (EMD-2925; EMD-3433)^, and here in the APC/C-CDH1-UBE2C-UBE2S dataset.

Extended Data Fig. 2. Heterogeneity analysis and classification of particles from the TR-EM datasets using cryoDRGN and PCA.

a – Diagram depicting the cryoDRGN data processing workflow. Particle images are mapped into a continuous low-dimensional vector space, producing a latent embedding, z. A volume decoder can then reconstruct a unique 3D volume given any value of z, for example a particle’s mapped location in latent space. Principle Component Analysis (PCA) can be applied to either the set of particle latent embeddings for a dataset or a set of reconstructed 3D volumes to map the heterogeneity of the particles or volumes in a continuous manner. b, c – UMAP (left) and PCA (right) depiction of the latent space with representative volumes shown at indicated locations, for APC/C-CDH1-UBE2C (B) and APC/C-CDH1-UBE2C-UBE2S (C), explained variance (EV) noted in parentheses. d – PC trajectory along the second component showing the transition between a ‘CRL Partial’ to ‘CRL Up, E2 Bound’ state. Values in parentheses denote explained variance (EV). Densities CDH1, APC2, APC11, and UBE2C are colored as purple, green, blue, and cyan, respectively.

Extended Data Fig. 3. Movement of the APC/C during catalysis, shown in cryoDRGN volume PCA trajectories.

a – Volume PCA trajectories showing conformational changes along first four components to visualize displacement of CDH1 and the APC2 CRL along continuous conformational trajectories. Density changes localized near CDH1 and at the APC/C actives site shown in black stripes. b – Expected variance plotted for the first four components shown in A. c – Histograms showing distributions of volumes along the PCs shown in A.

Extended Data Fig. 4. FSC curves and UMAP particle distributions over TR-EM datasets.

a – FSC curves depicting the resolution of the consensus 3D refinements shown in Fig. 2b. Lines show the 0.143 gold-standard and 0.5 cut-off with values denoting resolution at 0.143 cut-off. b – UMAP visualization of the particle distributions in the latent space plotted separately by timepoint.

Extended Data Fig. 5. Development of TIRF-based APC/C-dependent substrate ubiquitination assay following fluorescently-labeled UBE2C.

a, b – UMAP representation showing the particle distribution of the sub-states in the ‘CRL Down’ (A) and ‘CRL Up’ (B) major states. c – TIRF microscopy movies capture substrate-E2 interactions prior to complete substrate depletion. Experimental mixtures for the given conditions were split evenly between microscopy flow cells for data capture (Fig. 3d) and a ubiquitination reaction with 50 nM Biotin-CycB^N*-A488 spiked in, shown above, that was ran on an SDS-PAGE gel and fluorescently scanned. The banding of the ubiquitinated substrate is distorted due to the high BSA concentration needed for microscopy. The experiments were performed independently three times. d – Duration of detected binding events in seconds across multiple experiments and with increasing concentration of UBE2S CTP. e – There is minimal background signal from slide functionalization (left) and robust signal from immobilized substrate (center left) in the Substrate channel, as well as minimal background fluorescence from the substrate in the UBE2C channel (center right). The addition of UBE2C to immobilized substrate without APC/C-CDH1 shows minimal non-specific UBE2C binding (right). f – Example images showing UBE2C binding events upon addition of APC/C-CDH1 and increasing levels of UBE2S CTP. g – Montage showing representative time series of a UBE2C binding event in the UBE2C channel at 200 ms intervals. h – Montage (top) depicting duration of individual binding events in experiments described in (C) indicated by red or blue asterisk. Trace of raw intensity values (bottom) showing individual UBE2C binding events; red or blue asterisks denote corresponding binding events shown in montage (top). Source data

Extended Data Fig. 6. Previous artificially-trapped APC/C structures mimicking substrate ubiquitination.

a – Fitted cryoEM models from previous studies of the APC/C showing monoubiquitination (left), multi-ubiquitination (center), and polyubiquitination (right)^,. An Hsl1 truncation harboring its KEN- and D-box was used as a substrate where a lysine was replaced with cysteine and crosslinked to the active site of UBE2C to mimic substrate priming (left). A similar Hsl1 truncation was genetically fused to the UbV^RING, purified, and crosslinked to UBE2C and Ub to mimic multi-monoubiquitination (center). For chain elongation, a UbV^RING-Hsl1 D-box chimera harboring a K11C substitution was crosslinked to the active site of UBE2S (right). b – Similar to Fig. 4e, immunoblotting of APC/C-dependent substrate degradation in mitotic HeLa cell extracts reveals that this process is not inhibited by the addition of excess Ub, in contrast to the UbV^CDH1. The experiments were performed independently three times. c – Similar to Fig. 4f, Securin ubiquitination is inhibited more strongly by the UbVCDH1-Hsl1 D-box fusion, as monitored by fluorescent scanning of APC/C-CDH1-UBE2C-dependent ubiquitination reactions. The experiments were performed independently three times. d –Cryo-EM map of APC/C-CDH1-UbV^CDH1-Hsl1 D-box rigid body fitted with a surface representation model of UBE2C bound to APC2 (green)/APC11(blue). Left, UbV^CDH1 (orange) is in position to receive Ub from UBE2C (cyan). Right, cryo-EM map fitted with atomic models of CDH1 (purple) and UbV^CDH1 (orange)-Hsl1 D-box (red). Yellow rectangle indicates the KEN-box binding site of CDH1, similar to where UbV^CDH1 binds. Uncropped gels representative of n = 3 independent experiments for C available in source data. Source data

Extended Data Fig. 7. Ub binding to coactivator serves multiple purposes during APC/C-dependent polyubiquitination.

a – Quantitation of CycB^N* modified with >4 Ubs from single encounter assays in Fig. 5c. Error bars: SEM. The experiments were performed independently three times. b – Substrate-independent, APC/C-dependent hydrolysis of an oxyester-linked Ub from UBE2C C114S is not impeded by the addition of UbV^CDH1, monitored by Coomassie-stained SDS-PAGE gel. The experiments were performed independently three times. c – Synthesis of di-Ub by APC/C-UBE2S is not inhibited by UbV^CDH1 as revealed by scanning of an SDS-PAGE gel where the acceptor Ub* is fluorescently labeled. The experiments were performed independently three times. d – UbV^CDH1 inhibits APC/C-CDH1-dependent ubiquitination activity of all substrate and E2 combinations tested, monitored by fluorescent scanning. The experiments were performed independently three times. e – Ubiquitination of Ub-Securin and Ub-Securin D^Mut as shown in Fig. 5c. The experiments were performed independently three times. f – Similar to Fig. 5f, ubiquitination of *UbV^CDH1 requires UBE2C and does not occur with UBE2S alone, as monitored by fluorescent scanning. Uncropped gels representative of n≥3 independent experiments for B-F available in source data. Source data