Structures of APC/C(Cdh1) with substrates identify Cdh1 and Apc10 as the D-box co-receptor.

The ubiquitylation of cell-cycle regulatory proteins by the large multimeric anaphase-promoting complex (APC/C) controls sister chromatid segregation and the exit from mitosis. Selection of APC/C targets is achieved through recognition of destruction motifs, predominantly the destruction (D)-box and KEN (Lys-Glu-Asn)-box. Although this process is known to involve a co-activator protein (either Cdc20 or Cdh1) together with core APC/C subunits, the structural basis for substrate recognition and ubiquitylation is not understood. Here we investigate budding yeast APC/C using single-particle electron microscopy and determine a cryo-electron microscopy map of APC/C in complex with the Cdh1 co-activator protein (APC/C(Cdh1)) bound to a D-box peptide at ∼10 Å resolution. We find that a combined catalytic and substrate-recognition module is located within the central cavity of the APC/C assembled from Cdh1, Apc10--a core APC/C subunit previously implicated in substrate recognition--and the cullin domain of Apc2. Cdh1 and Apc10, identified from difference maps, create a co-receptor for the D-box following repositioning of Cdh1 towards Apc10. Using NMR spectroscopy we demonstrate specific D-box-Apc10 interactions, consistent with a role for Apc10 in directly contributing towards D-box recognition by the APC/C(Cdh1) complex. Our results rationalize the contribution of both co-activator and core APC/C subunits to D-box recognition and provide a structural framework for understanding mechanisms of substrate recognition and catalysis by the APC/C.

The APC/C is a multimeric E3 ubiquitin ligase assembled from 13 individual subunits 1,2. Many of the APC/C’s core proteins are comprised of multiple repeat motifs whose principle function is to provide a molecular scaffold, but whose exact biological role is not well understood. The best characterised APC/C subunits are the cullin and RING proteins Apc2 and Apc11 responsible for catalytic activity, and the TPR subunit Apc3/Cdc27 that interacts with co-activator (either Cdc20 or Cdh1) 10,11,12 and the APC/C subunit Apc10 (also known as Doc1) 13. Both co-activator 9,11,14,15,16,17, and core APC/C subunits 5,6,7,8,9 have been implicated in substrate recognition, but the structural basis for this process is unknown. To address this question, we used single particle electron microscopy to determine structures of budding yeast APC/C in complex with Cdh1 (APC/CCdh1) and substrates. The resultant EM maps are of exceptional quality and detail. Our new maps show the characteristic triangular shape of the APC/C 18,19,20,21 (Supplementary Fig. 1), but at higher resolution we visualise a lattice-like scaffold assembled from individual APC/C subunits defining a central cavity.

The APC/C co-activator Cdh1 was identified in negative stain EM reconstructions as a prominent and discrete density feature present within the central cavity of APC/CCdh1 and absent from APC/C (Fig. 1a,b). Its disc-shaped density, characteristic of an exposed WD40 β-propeller domain, is connected to the APC/C via an edge-on interface. Overall, with the exception of the Cdh1 density, APC/C and APC/CCdh1 are similar, and the large conformational changes that accompany co-activator binding to vertebrate APC/C 19,21 are not evident. An ellipsoid-shaped density feature, resembling the β-sandwich of Apc10 13,22, situated adjacent to, but not in contact with Cdh1, is more prominent in the presence of Cdh1 (Fig. 1a,b). Its close proximity to Cdh1 was intriguing in view of the role of Apc10 in contributing towards substrate recognition 6, and the D-box-dependent processivity of the ubiquitylation reaction 5,7. To unequivocally identify Apc10, we generated APC/CΔApc10 in complex with Cdh1 (APC/CΔApc10-Cdh1). The resultant APC/CΔApc10-Cdh1 map showed complete loss of this ellipsoid density (Fig. 1c), confirming its identity as Apc10. Deletion of Apc10 also resulted in a depletion of Cdh1 density around the circumference of the β-propeller most distant from its contact to APC/C (Fig. 1c). Since deletion of Apc10 does not affect the APC/C subunit composition 6 or abrogate Cdh1 binding (Supplementary Fig. 2), the partial loss of Cdh1 density is indicative of an increased flexibility of Cdh1’s WD40 domain. This finding and the reduced density for Apc10 in APC/C imply conformational interdependence of Apc10 and Cdh1.

Figure 1

Negative stain electron microscopy reconstructions of budding yeast APC/C show that substrate binding to APC/CCdh1 involves Cdh1 and Apc10. (a) APC/CCdh1, (b) APC/C, (c) APC/CΔApc10·Cdh1. Density assigned to Cdh1 and Apc10 is shown in magenta and blue, respectively. The resolution of the APC/CCdh1 binary complex is ~18-20 Å (Supplementary Fig. 10d). Negative stain EM reconstructions of (d) APC/CCdh1·Hsl1 complex, (e) APC/CCdh1·D-box, (f) APC/CCdh1·KEN-box. Lower panels in (d), (e) and (f) show details of the structural changes associated with Cdh1 and Apc10 in the presence of substrate compared with the superimposed binary APC/CCdh1 map represented in mesh. Hsl1 and D-box and KEN-box peptides were used at saturating concentrations to promote stoichiometric APC/CCdh1-substrate ternary complexes.

To identify substrate-binding sites on APC/CCdh1, we utilised a fragment of Hsl1, a D-box (RxxLxxI/VxN) 3 and KEN-box (KEN) 4 containing substrate with high affinity for APC/CCdh1 (refs 14,23). The ternary APC/CCdh1·Hsl1 complex was catalytically competent, as judged by its ability to ubiquitylate Hsl1 (Supplementary Fig. 3a). Engagement of Hs1l with APC/CCdh1 is accompanied by a pronounced structural change involving Cdh1 and Apc10 (Fig. 1d). Specifically, the β-propeller domain of Cdh1 is bulkier, shifts ~7 Å towards Apc10, and new, well-defined density bridges Cdh1 to Apc10. Thus, Hsl1 promotes the formation of new connections between Cdh1 and Apc10, a result consistent with direct co-activator - substrate interactions 9,11,14,15,16,17 and a role for Apc10 in mediating optimal substrate binding 5,6,7,23.

To define the specific roles of the D- and KEN-boxes in contributing to these conformational changes, we determined structures of APC/CCdh1 in complex with synthetic peptides containing either a D-box or a KEN-box. Similar to previous results with D-box peptides 11,24, an 18-residue D-box peptide modelled on cyclin B (S. pombe Cdc13) completely inhibited APC/CCdh1 activity towards Clb2 (a mitotic cyclin with D- and KEN-boxes) at 0.1 mM (Supplementary Fig. 4a). Fig. 1e shows that D-box peptide generated similar structural changes to Hsl1, specifically the WD40 domain of Cdh1 is shifted, and new density connects it with Apc10 (Supplementary mov1). However, in contrast to the APC/CCdh1·Hsl1 map, the extent of new density associated with Cdh1 is markedly reduced, suggesting that the additional density in APC/CCdh1·Hsl1 represents the larger Hsl1 substrate. Control experiments show that a mutant D-box peptide, which fails to bind APC/CCdh1 (Supplementary Fig. 4c), induces no conformational changes (Supplementary Fig. 5). Binding of the KEN-box peptide to APC/CCdh1 also promotes a repositioning of Cdh1 towards Apc10, but notably without the connecting density (Fig. 1f). This indicates that only D-box substrates promote a physical interconnection between Cdh1 and Apc10.

To explore the structure of APC/CCdh1·D-box in more detail, we collected cryo-EM images of the complex and determined its structure at ~10 Å resolution. The cryo-EM map reproduces the overall features of the APC/CCdh1·D-box map generated from negatively stained particles, but with greatly enhanced detail and resolution (Fig. 2, Supplementary Fig. 6,7). Similar to the APC/CCdh1·D-box ternary complex obtained from negative stain EM, the cryo-EM reconstruction shows density connecting Cdh1 and Apc10 (Figs. 2,4). Docking the crystal structure of Apc10 (refs 13,22) and the modelled Cdh1 WD40 domain into their respective densities, indicates additional unassigned density linking Cdh1 to Apc10 (Fig. 4a,c). Strikingly, the best fit of Apc10 into the cryo-EM map positions a highly conserved loop, required for D-box recognition 7, adjacent to the density linking Apc10 with Cdh1. In contrast, residues on Apc10’s opposite surface that contribute to APC/C interactions 7, are oriented towards Apc2 (Fig. 4c).

Figure 2

Cryo-electron microscopy reconstruction of budding yeast APC/CCdh1·D-box reveals the lattice-like architecture of the complex. Three views of the complex with (b) similar to views shown in figure 1. Resolution is ~10 Å (Supplementary Fig. 12c).

Figure 4

Cdh1, Apc10, Apc2 and Apc11 form a substrate recognition-catalytic module. (a) and (b). Two views of the cryo-EM APC/CCdh1·D-box complex. Protein density is represented by a mesh with fitted atomic coordinates of Cdh1 β-propeller (modelled), Apc10 (ref. 22), Apc2-Apc11 (modelled on Cul4a-Rbx1 of SCF) and Cdc27 (ref. 26). Only the N-terminal β-strand of Apc11 bound to the Apc2 CTD is modelled (orange). The two subunits of Cdc27 are shown in light and dark green. View in (a) shows the 2-fold symmetry axis of Cdc27. Density connecting Cdh1 to a TPR-super-helix of the Cdc27 dimer is indicated by an arrow. TPR motifs 8 to 10 of Cdc27, implicated in IR-tail recognition 23, are shown in lighter colourer. In (b) the final residue of Apc10 observed in the crystal structure (Ser 256), 25 residues N-terminal to the IR motif, is indicated by red spheres. (c) Details of the Cdh1 and Apc10 co-receptor for D-box. Both Cdh1 and Apc10 connect to Apc2. The N-terminus of Cdh1, including the C-box linking the WD40 domain to Apc2, is not modelled. Red arrow i denotes the conserved loop (residues His239 to Asp244) of Apc10 implicated in D-box recognition 7, red arrow ii denotes Lys162 and Arg163 of Apc10 responsible for APC/C affinity 7. Two models for a possible fit of D-box to the density interconnecting Cdh1 and Apc10 are shown in Supplementary Fig. 8. (d) Schematic of combined catalytic and substrate recognition module responsible for D-box binding and substrate ubiquitylation. D-box is represented as binding to an interface between Cdh1 and Apc10.

These structural data revealing that Cdh1 and Apc10 become interconnected by bridging density in the presence of D-box substrates rationalises biochemical studies demonstrating that both co-activator and core APC/C subunits 8,9,11,14,15,16,17, specifically Apc10 5,6,7,23, contribute to D-box dependent recognition and processive ubiquitylation. The unassigned density bridging Apc10 and Cdh1 in the APC/CCdh1·D-box complex can be modelled as a D-box peptide, indicating that the binding site for D-box is shared between the WD40 domain of Cdh1 and the β-sandwich of Apc10. Cdh1 and Apc10 therefore generate a D-box co-receptor (Supplementary Fig. 8). Although biochemical data show that the D-box interacts with the conserved surface of co-activator’s WD40 domain 11,15, direct interactions between D-box and Apc10 alone have not previously been demonstrated (our unpublished data and 7), possibly due to the weak affinity of isolated Apc10 for D-box.

We used 15N-HSQC NMR, a technique suitable for detecting weak protein-ligand interactions, to investigate potential Apc10 - D-box interactions. The 15N-HSQC NMR spectrum of S. cerevisiae Apc10, shown in Fig. 3, has a substantial number of well-dispersed peaks consistent with the Apc10 β-sandwich architecture 22. However, the number of visible peaks is ~ half that expected for a 221 residue protein, and the visible peaks have a wide range of intensities. Reduced peak number and intensity variation are characteristic of proteins undergoing exchange between different conformational or oligomeric states. Spectra recorded with a two-fold difference in protein concentration showed no change in position or shape of any dispersed peak suggesting there is no sensitivity to any possible oligomerisation equilibrium. Consequently, the features of the 15N-HSQC spectrum are best explained as a result of Apc10 adopting multiple conformations in intermediate to slow exchange (sub-millisecond to second timescales) in solution. Addition of a stoichiometric excess (~40-fold) of the D-box peptide used to generate the APC/CCdh1·D-box ternary complex resulted in more than 20 changes in amide peak position or relative intensity (Fig. 3). NMR-based measurement of the translational diffusion coefficient showed that the NMR-observed species is an Apc10 monomer of ~26 kDa. Thus, the changes in specific peaks on addition of peptide demonstrate that the D-box peptide interacts with monomeric Apc10, altering the chemical environment and/or the conformational equilibrium of a subset of its residues. However, the low intensity and proportion of visible amide peaks made sequential assignment and full characterisation of the D-box binding site on Apc10 impracticable.

Figure 3

1H-15N HSQC spectra of Apc10. Overlaid are spectra of the apo-protein and protein in the presence of stoichiometric excess of each of four peptides. The complete amide region (a) and for clarity expanded views of two boxed sub-regions (b, c) are shown. Spectra in the presence of either of the two D-box containing peptides show common changes with respect to the apo-protein spectrum, namely absence of the peaks seen in the apo-protein (black arrows) and new or shifted peaks not seen in the apo-spectrum (blue arrows). In contrast, spectra in the presence of either the Cdc13-derived peptide in which four residues of the D-box motif are mutated to alanine or a peptide containing a KEN-box motif are very similar to the apo-spectrum, retaining all of the peaks marked by black arrows. The spectrum with the mutant Cdc13 peptide does in some cases show low intensity peaks at the positions indicated by blue arrows (see expanded views b, c) indicating a very weak residual interaction. These spectra are consistent with a D-box dependent interaction with Apc10. (Peaks arising from natural abundance 15N amides in the unbound peptide that are protected from solvent exchange are indicated by an asterisk).

To establish whether the peptide-induced changes of the Apc10 NMR spectrum are specifically D-box-dependent, we performed a series of control experiments. First, a different D-box peptide (a 19-residue peptide modelled on S. cerevisiae Clb2 whose sequence identity with Cdc13 is confined to the D-box) produced very similar NMR spectral changes as the Cdc13 D-box (Fig. 3). Second, a mutant D-box Cdc13 peptide resulted in only minor changes in the Apc10 NMR spectrum, consistent with greatly reduced binding. Finally, the Hsl1 KEN-box peptide, which from the APC/CCdh1·KEN-box EM analysis does not bridge Cdh1 and Apc10, resulted in an essentially identical spectrum to that of the apo-protein, with none of the changes seen for the two D-box containing peptides. These NMR data therefore provide strong evidence for a direct interaction between Apc10 and D-box, supporting the notion that Apc10 participates in D-box recognition.

To gain further insight into the mechanisms of substrate recognition and ubiquitylation, we modelled atomic structures of Apc2 and Cdc27 into the molecular envelope of the APC/CCdh1·D-box map. We fitted a homology model of Apc2, based on Cul4a-Rbx1, allowing for small adjustments of the C-terminal domain (CTD) relative to the cullin repeats (Fig. 4, Supplementary Fig. 7,9). Continuous density attaches the globular CTD to that of the cullin repeats, which are seen as a long stalk-like density that transverses one side of the complex (Fig. 4, Supplementary Fig. 7). The APC/CCdh1·D-box cryo-map reveals that Cdh1 and Apc10 are both connected to the Apc2 CTD (Fig. 4c, Supplementary Fig. 7). Notably, the interaction of the CTD of Apc2 with substrate adaptor subunits contrasts with the SCF complex in which the N-terminal cullin repeat of Cul1 interacts with substrate adaptors 25.

Cdc27 is a dimer and we docked its N-terminal dimerisation domain 26 into the globular structure at the head of the TPR sub-complex, and independently positioned the modelled C-terminal TPR super-helices of the Cdc27 subunits into the curved tubular densities extending from the globular domain (Fig. 4a,b, Supplementary Fig. 7), consistent with the mapping of Cdc27 (ref. 27). Although not imposed in the fitting, these docked TPR super-helices are related by the same dyad symmetry as the Cdc27 dimerisation domain, therefore preserving the overall two-fold symmetry of Cdc27 (Fig. 4a). The organisation of Apc2 and Cdc27 in close proximity to Cdh1 and Apc10 visualised in our APC/CCdh1·D-box structure unifies previous models of APC/C subunit topologies 10,12,13,23 (Fig. 4d). Cdh1 is known to interact through its C-terminal Ile-Arg (IR) tail with Cdc27 10,11,12, and in S. cerevisiae, Cdh1 also requires Apc2 for optimal binding 12. The structures fitted to the EM map show that with the C-terminus of Cdh1 in contact with Cdc27, its N-terminal C-box is positioned to contact Apc2 (Fig. 4, Supplementary Fig. 7) 12,17. Pulldown experiments on recombinant human proteins have shown that Apc10 interacts with Cdc27 through its C-terminal region, which also contains an IR-motif 13, whereas in S. cerevisiae, Apc10 associates preferentially with a sub-complex of Apc1, Apc2 and Apc11 (ref. 12). Our EM data position Apc10 close to the second Cdc27 subunit. Consequently, the human and yeast biochemical data are explained by the extensive interface between Apc10 and Apc2, and the flexible C-terminal IR-tail of Apc10 binding to the Cdc27 TPR super-helix.

This study identifies Cdh1 and Apc10 as a co-receptor for D-box. Individually, co-activator and APC/C possess low affinity and specificity for substrate 9 and therefore cooperatively enhances substrate affinity through multivalency. Definition of the subunit organisation and generation of a pseudo-atomic structure of the APC/C 27, together with characterisation of the D-box co-receptor presented here, provide the conceptual framework for a mechanistic understanding of the APC/C.

Methods Summary

Generation of APC/C and complexes with Cdh1 and substrates

APC/C and APC/CΔApc10 were isolated from S. cerevisiae and ubiquitylation assays were performed essentially as described 6. S. cerevisiae His6-Cdh1 was expressed in Sf9 cells and purified using Ni-NTA. APC/CCdh1 was prepared by loading excess Cdh1 to APC/C immobilised on calmodulin resin, and eluted as for APC/C. APC/CCdh1-substrate complexes were generated as described in Methods.

Electron microscopy and image analysis

Purified APC/C (~0.2 mg/ml) from peak elution fractions was applied to Quantifoil 1.2 or 2 μm aperture grids coated with continuous thin carbon and either negatively stained for electron microscopy at room temperature or flash frozen using a Vitrobot for cryo-EM. Images were recorded in an FEI TF20 electron microscope under low dose conditions using a Tietz F415 CCD camera. Three-dimensional maps were calculated from molecular images using programs from Imagic 28, Spider 29 and EMAN 30.

NMR analysis

1H-15N HSQC spectra were recorded at 25°C over 5.5 or 11 hours for samples of Apc10 alone and in the presence of four peptides samples using a (1H,15N,13C) triple resonance cryoprobe on a 700MHz Bruker Avance III spectrometer. Spectra were processed identically and displayed to compensate for concentration and/or recording time differences.