Structural insights into anaphase-promoting complex function and mechanism

David Barford

Abstract

The anaphase-promoting complex or cyclosome (APC/C) controls sister chromatid segregation and the exit from mitosis by catalysing the ubiquitylation of cyclins and other cell cycle regulatory proteins. This unusually large E3 RING-cullin ubiquitin ligase is assembled from 13 different proteins. Selection of APC/C targets is controlled through recognition of short destruction motifs, predominantly the D box and KEN box. APC/C-mediated coordination of cell cycle progression is achieved through the temporal regulation of APC/C activity and substrate specificity, exerted through a combination of co-activator subunits, reversible phosphorylation and inhibitory proteins and complexes. Recent structural and biochemical studies of the APC/C are beginning to reveal an understanding of the roles of individual APC/C subunits and co-activators and how they mutually interact to mediate APC/C functions. This review focuses on the findings showing how information on the structural organization of the APC/C provides insights into the role of co-activators and core APC/C subunits in mediating substrate recognition. Mechanisms of regulating and modulating substrate recognition are discussed in the context of controlling the binding of the co-activator to the APC/C, and the accessibility and conformation of the co-activator when bound to the APC/C.

1. Introduction

The anaphase-promoting complex or cyclosome (APC/C) is an unusually large multi-subunit cullin-RING E3 ubiquitin ligase that functions to regulate progression through, and exit from, the mitotic phase of the cell cycle and controls entry into the S phase ([14], reviewed by [59]). The APC/C also plays a role to regulate meiosis, and has been implicated in post-mitotic functions including dendrite formation in neurons, metabolic, and learning and memory processes [1015]. APC/C-mediated coordination of cell cycle progression is achieved through the temporal regulation of APC/C activity and substrate specificity. The APC/C becomes activated at the onset of mitosis, and at prometaphase ubiquitylates Nek2A (table 1) and cyclin A. At metaphase, the APC/C ubiquitylates two inhibitors of the transition to anaphase: securin and cyclin B [16,17]. Securin is a protein inhibitor of separase, a protease that cleaves the cohesin subunit SCC1 [18]. Cleavage of SCC1 causes disassembly of cohesin, necessary to allow sister chromatid segregation [1922]. Reduced cyclin levels are also required for entry into anaphase, since cyclin-dependent kinase (Cdk) activity inhibits separase [2325]. After anaphase, cyclin destruction continues to maintain negligible Cdk activity, necessary for the cell to disassemble the mitotic spindle and exit mitosis [2628]. During G1, the main role of the APC/C is to sustain low levels of mitotic Cdk activity to allow for resetting of replication origins as a prelude to a new round of DNA replication in the S phase [29,30].

View this table:
Table 1.

APC/C substrates with non-canonical destruction motifs.

The temporal regulation of APC/C activity is achieved through a combination of two structurally related co-activator subunits, Cdc20 and Cdh1 [3141], coupled to protein phosphorylation, APC/C inhibitors and differential affinity for APC/C substrates. The two APC/C co-activators have opposing activity profiles. Cdc20 activates the APC/C during early mitosis, when the APC/C is phosphorylated and Cdh1 activity is low owing to its Cdk-dependent phosphorylation, whereas APC/CCdc20-mediated reduction of Cdk activity stimulates Cdh1. In turn, APC/CCdh1 ubiquitylates Cdc20, leading to APC/CCdc20 inactivation. Thus, Cdc20 activates Cdh1, which in turn antagonizes Cdc20 activity. The switching between APC/CCdc20 and APC/CCdh1 fulfils two main functions. First, APC/CCdc20 and APC/CCdh1 have overlapping but nevertheless distinct substrate specificities. Therefore, specific cell cycle regulators are degraded during the separate phases of APC/CCdc20 and APC/CCdh1 activity, allowing for ordered progression through the cell cycle. Second, Cdc20 and Cdh1 are subject to control by different regulatory mechanisms. Cdc20 activates the APC/C phosphorylated by Cdk and Plk1 protein kinases during early mitosis, whereas Cdh1 is inhibited by its Cdk-mediated phosphorylation. Importantly, APC/CCdc20 activity towards securin and cyclin B is inhibited by the mitotic checkpoint complex (MCC), a multi-protein complex generated in response to the spindle assembly checkpoint (SAC) [42].

2. Anaphase-promoting complex/cyclosome: substrate recognition

(a) Canonical anaphase-promoting complex/cyclosome recognition motifs: D box and KEN box

The capacity of the APC/C to ubiquitylate a large number of distinct substrates at specific phases of the cell cycle is determined by complex mechanisms for substrate recognition. A universal mode of APC/C-substrate recognition does not exist. The classical APC/C degron is the destruction box or D box, a nine-residue motif (RxxLxxI/VxN), first characterized in B-type cyclins as being sufficient for APC/C-mediated ubiquitylation [4345], common to most, but not all APC/C substrates. Another APC/C degron, the KEN motif (KENxxxN/D), is often present in APC/C substrates usually in addition to the D box [46]. Efficient ubiquitylation by either APC/CCdc20 or APC/CCdh1 of substrates harbouring both D and KEN boxes is dependent on both degrons [47,48]. However, some substrates contain only either a D box or a KEN box, in one or more copies. The use of two distinct degrons imposes a degree of substrate specificity on the APC/C, with APC/CCdc20 being more dependent on the D box and APC/CCdh1 more dependent on the KEN box [49,50]. Importantly, APC/CCdh1 is capable of ubiquitylating KEN box-only-containing substrates, for example, Tome-1 [51] and Sororin [52]. Additionally, as discussed below, more rarely used APC/C degrons have also been characterized.

In their study of signals responsible for cyclin destruction, King et al. [44] found that Lys residues in close proximity to the D box serve as targets of ubiquitylation, consistent with findings that a Lys residue immediately C-terminal to the D box can function as a ubiquitin acceptor [53].

(b) Mechanism of D box and KEN box recognition

Without the co-activator, the core APC/C is inactive as an E3 ubiquitin ligase and is unable to bind most APC/C substrates (exceptions include recognition of the kinase Nek2A, cyclin A and a tandem D box in vertebrates). It has therefore been proposed that the role of the co-activator is to act as a substrate-recognition subunit, analogous to the F box protein receptors of the SCF, and recruit substrates to the APC/C. This notion is consistent with observations that co-activators promote the interaction of the substrate to the APC/C to form a stoichiometric ternary complex of APC/C–co-activator–substrate, dependent on the D box and KEN box degrons [48,54,55]. In support of the adaptor model, numerous studies demonstrate direct and D box- and KEN box-dependent binding of APC/C substrate proteins to co-activators [47,5660]. Chemical cross-linking of substrates to the WD40 domains of Cdc20 and Cdh1 in a D box- and KEN box-dependent manner confirms a role of the co-activators to contribute towards the recruitment of the substrate to the core APC/C [53,56]. However, given the immeasurably low affinity of co-activator–substrate interactions, it is unlikely that the co-activators alone are sufficient to confer high-affinity substrate binding to the APC/C–co-activator binary complex. Rather, core APC/C subunits contribute towards substrate association, especially D box-containing substrates. The APC/C alone binds tandem D boxes [61], and the Apc10 (also known as Doc1) subunit of the APC/C is necessary for optimal co-activator-dependent substrate recognition [48], and for contributing towards substrate processivity [62]. In the instance of D box recognition, recent single-particle electron microscopy studies have shown that the co-activator and Apc10 function as a D box co-receptor [63]. Earlier studies also support the notion of a composite substrate-binding site shared between a co-activator and a core APC/C subunit(s), dependent on Apc10. First, substrates enhance the binding of Cdh1 to the APC/C in a D box-dependent manner [60], a finding supported by a study showing that the substrate compensates for loss of affinity of Cdh1 to APC/C owing to mutations in the putative Cdh1-binding region of Cdc27 [64].

(c) Non-canonical anaphase-promoting complex/cyclosome recognition motifs

Ubiquitylation of some APC/C substrates is not dependent on either the D box or the KEN box. Thus, a class of non-canonical APC/C recognition motifs can contribute towards APC/C-dependent substrate ubiquitylation.

Cyclin A and Nek2A evade the SAC mediated by the MCC, even though the MCC inhibits ubiquitylation of the APC/CCdc20-dependent targets securin and cyclin B [38,40,6568]. Both cyclin A and Nek2A possess modified D boxes (cyclin A-type D boxes), located close to the N-terminus in cyclin A and towards the C-terminus in Nek2A. However, for both cyclin A and Nek2A, mutation of the D box (and the KEN box in Nek2A) does not stabilize the respective protein at prometaphase, suggesting a different mode of substrate recognition from that occurring at other phases of the cell cycle [65,6769]. In the instance of cyclin A, both a region immediately C-terminus of the D box and interactions with Cdk1 and Cks (a Cdk-regulatory protein) [65,70,71] are required for APC/C recognition, whereas a C-terminal Met-Arg motif is responsible for associating Nek2A to the APC/C [67]. The Nek2A KEN box is a required degradation signal after the onset of metaphase, consistent with altered modes of substrate recognition at prometaphase [67]. Similarly, recent work from Izawa & Pines [72] revealed that degradation of cyclin A at prometaphase does not require either a D box or Apc10, whereas both are required at metaphase.

Other examples of substrates that incorporate non-canonical destruction motifs (table 1) include the A box of Aurora A kinase [73,74], an N-terminal 34-residue segment of Cik1L (a splice variant of Cik1) [75], a C-terminal element of Cin8p [76], a 43-residue segment at the C-terminus of the kinesin Kip1p [77], a GxEN motif in Xenopus chromokinesin XKid [78], an LxExxxxN motif in Saccharomyces cerevisiae Spo13 [79] and a Cry box in Cdc20 [80].

3. Anaphase-promoting complex/cyclosome subunit organization and structure

(a) Overview of anaphase-promoting complex/cyclosome subunits

In most eukaryotes, APC/C comprises 13 different proteins, most of which are highly conserved, essential for activity, with some present in two copies per complex (table 2) [48,8290]. Molecular weights and stoichiometry of the subunits have been established from recombinant assemblies by nano-electrospray mass spectrometry, analysis of X-ray structures and multi-angle laser light scattering [81,91,92]. The catalytic core of the APC/C is composed of the cullin subunit Apc2 and RING H2 domain subunit Apc11, analogous to the cullin and Rbx1 subunits, respectively, of cullin-RING ligases of the SCF superfamily [93] (table 2 and figure 1). The C-terminal domain (CTD) of Apc2 forms a tight complex with Apc11 and together are competent to catalyse ubiquitylation, although are devoid of substrate selectivity [9496]. Many of the APC/C's core proteins are composed of multiple repeat motifs whose principal function is to provide a molecular scaffold. The largest APC/C subunit, Apc1, is approximately 200 kDa in size. The C-terminal region of Apc1 is composed of 11 tandem repeats of a 35–40 amino acid motif shared with the Rpn1 and Rpn2 subunits of the proteasome 19S regulatory particle [97] (figure 1). Such motifs, termed proteasome/cyclosome repeats (PC repeats), are intriguing given the close functional relationship between the APC/C and proteasome. The Apc2 N-terminus comprises three cullin repeats of approximately 130 residues. The most abundant structural motif observed in APC/C subunits is the tetratricopeptide repeat (TPR) with four yeast and five vertebrate APC/C subunits composed almost exclusively of 13–14 copies of this 34-amino acid motif arranged in tandem (table 2 and figure 1). The TPR motif, originally discovered within proteins subsequently identified as components of the APC/C [98101], is one of the most abundant protein motifs present in eukaryotic genomes, and functions in scaffolding proteins to mediate protein–protein interactions, and the assembly of multi-protein complexes. As discussed later, the super-helical architecture generated by consecutive TPR motifs facilitates protein–protein interactions. Finally, the functionally and structurally related co-activators Cdc20 and Cdh1 contain WD40 β-propeller domains that contribute to substrate recognition (table 2 and figure 1).

View this table:
Table 2.

Core and regulatory subunits of the APC/C. Molecular mass corresponds to S. cerevisiae subunits. Stoichiometry based on [81].

Figure 1.

Primary structure of selective APC/C subunits. (a) Conserved TPR subunits of the APC/C: Cdc16, Cdc23, Cdc27 and Apc5. Schematic of Cdc27 and Cdc16 based on Zhang et al. [91,92], respectively. (b) Apc1 showing the PC (proteasome–cyclosome repeats). (c) Apc2 showing N-terminal cullin repeats and the C-terminal domain (CTD) responsible for binding Apc11. (d) Cdh1. ‘P’ indicates consensus Cdk-phosphorylation sites that are responsible for blocking APC/C interaction with Cdh1.

(b) Anaphase-promoting complex/cyclosome subunit structures

Recently, the crystal structures of numerous TPR subunits [91,92] have been determined to augment the only other APC/C subunit structure (Apc10) determined a decade ago [102,103]. Apc10 is composed of a Doc homology domain also present in putative E3 ubiquitin ligases characterized by cullin and HECT domains [104], including a variant SCF composed of CUL7 [105]. The Doc homology domain forms a β-sandwich structure that is related in architecture to the galactose-binding domain of galactose oxidase, the coagulation factor C2 domain and an XRCC1 domain (figure 2a). Residues that are invariant among Apc10 map to a β-sheet region of the molecule, whose counterpart in galactose oxidase, the coagulation factor C2 domains and XRCC1 mediate biomolecular interactions. Interestingly, mutations of this region abrogate APC/C E3 ubiquitin ligase activity in a D box-dependent manner, indicating a role for this region in mediating D box recognition [106].

Figure 2.

Crystal structures of selective APC/C subunits. (a) Apc10/Doc1 (S. cerevisiae) [102]. (b) Cdc16–Cdc26 heterodimer [91]. Cdc16 and Cdc27 share a related architecture and mode of homodimerization. (c) Cdc27 homodimer [91,92]. Both Cdc16 and Cdc27 homodimerize through an N-terminal domain composed of seven TPR units forming a TPR superhelix.

The recent crystal structure determinations of a Cdc16–Cdc26 complex from Schizosaccharomyces pombe (Cut9–Hcn26) [91] and the N-terminal domain of Cdc27 (Encephalitozoon cuniculi) [92] reveal insights into the molecular organization and architecture of APC/C TPR-containing subunits (figure 2b,c). Both Cdc16 and Cdc27 form homodimers, with the dimer interface mediated by the self-association of their respective N-terminal TPR domains. A TPR superhelix created by seven successive TPR motifs (each TPR motif folds into a pair of anti-parallel α-helices) forms an interlocking dimer interface whereby the two subunits of the dimer associate in an N-terminal-to-C-terminal configuration (figure 2b,c) [91,92]. An intimate homotypic interface is generated as the inner concave surface of each subunit's TPR superhelix encircles its dimer counterpart in a clasp-like arrangement. Cdc16 associates with Cdc26, and in the Cdc16–Cdc26 heterotetramer the C-terminal seven TPR motifs of Cdc16 are arranged into a TPR superhelix contiguous with the TPR superhelix of the N-terminal dimerization domain [91]. Thus, each Cdc16 subunit adopts a rod-like structure of a contiguous TPR superhelix of 14 TPR motifs with two complete TPR turns. Dimerization of Cdc16 via its N-terminal domain generates a shallow ‘V’-shaped molecule measuring some 155 Å in length. Cdc26 stabilizes Cdc16 through an unusual TPR–protein interaction. The N-terminal 12 residues of Cdc26, which are unstructured in isolation, adopt an extended conformation that interacts with the inner concave groove of the Cdc16 superhelix [91,92,107] (figure 2b). Residues 14–21 of Cdc26 on the other hand form an α-helix that packs against the C-terminal non-consensus TPR α-helix of the superhelix, mimicking a TPR motif, although with both helices arranged in parallel. Interactions between Cdc16 and Cdc26 explain the stabilizing influence exerted by Cdc26 on Cdc16 [107].

The C-terminal seven TPR units of Cdc27, missing from the crystallized N-terminal domain of Cdc27 (Cdc27Nterm), were predicted to form a continuous TPR superhelix contiguous with the Cdc27Nterm TPR superhelix [92]. Thus, similar to Cdc16, the Cdc27 homodimer will adopt a ‘V’-shaped molecule, with the C-terminal TPR superhelix projecting away from the apex of the ‘V’ formed by the Cdc27Nterm domain interface. Structural similarity of Cdc16 and Cdc27 architecture and mode of dimerization suggest evolutionary divergence from a common ancestor. Cdh1 interactions with Cdc27, mediated via its C-terminal Ile-Arg (IR) motif, involve the highly conserved TPR8 and TPR9 motifs of the CTD of Cdc27 [64]. Residues implicated in engaging the Cdh1-IR motif line the inner groove of the extended TPR superhelix, distal from the Cdc27 dimerization domain.

In the S. pombe Cdc16–Cdc26 (Cut9–Hcn1) structure, the N-terminal acetylated Met residue of Cdc26 is located within an enclosed chamber formed from the inner groove of the Cdc16 TPR superhelix [91] (figure 3). The N-terminus of Cdc26 is completely surrounded by its interacting Cdc16 subunit. The upper extent of the chamber is capped by the N-terminus of the opposing Cdc16 subunit (figure 3), although Cdc26 does not participate at the dimer interface. Thus, the acetyl group of the N-terminal Met of Cdc26 is completely inaccessible to bulk solvent. Recently, N-terminal acetylation was identified as a secondary degradation signal (degron) for the Doa10 E3 ubiquitin ligase [108]. One suggested role for N-terminal acetylation, which is present in 80 per cent of proteins, was to control the relative stoichiometry of subunits constituting a multi-subunit complex. Burial of the N-acetyl Met of Cdc26 in complex with Cdc16 would prevent its recognition and degradation via the Doa10–proteasome pathway, whereas excess levels of Cdc26 not in complex with Cdc16, and therefore with an exposed N-acetyl Met residue, would be available for recognition by the Doa10 E3 ubiquitin ligase and therefore subject to degradation via the ubiquitin proteasome system.

Figure 3.

The N-terminal Met of Cdc26 is acetylated and enclosed within a chamber formed from the Cdc16 TPR superhelix and the homodimer interface. The N-terminal region of Cdc26 is shown in cyan with Cdc16 shown as a surface representation, with secondary structure indicated. Reproduced with permission from [91].

(c) Reconstituting recombinant anaphase-promoting complex/cyclosome

Prior to the work of Schreiber et al. [81], research on the APC/C had been restricted to the use of native systems. Because most APC/C subunits are essential, genetic manipulations are intrinsically difficult, and the low natural abundance of the APC/C had limited structural, biochemical and biophysical studies. To overcome these restrictions, and to allow defined manipulation of APC/C complexes, a recombinant expression system was recently generated that allowed the reconstitution of holo APC/C and its subcomplexes in milligram quantities [81]. Initially, two S. cerevisiae APC/C subcomplexes were generated, SC8 and TPR5, the selection of component subunits guided by the subunit topology of S. cerevisiae APC/C [84,109]. SC8 comprised the core APC/C subunits associated with catalysis and substrate recognition (Apc2, Apc11 and Apc10) together with subunits that are thought to play a structural role (Apc1, Apc4, Apc5 and Cdc23), and Mnd2. A second complex, TPR5, contained the TPR subunits Cdc16 and Cdc27, together with the smaller accessory subunits Cdc26, Apc9 and Apc13. Holo APC/C was generated by combining the co-expression of SC8 and TPR5 in insect cells.

(d) The anaphase-promoting complex/cyclosome comprises 17 or 18 subunits with a total mass of 1127–1158 kDa

A quantitative estimate of the mass and subunit stoichiometry of the APC/C was obtained by applying nano-electrospray mass spectrometry to SC8. SC8 was assigned a mass of 698.8 kDa, in good agreement with that predicted for a complex containing all SC8 subunits in unit stoichiometry plus an additional copy of Cdc23. Combining the stoichiometry data for Cdc27 and Cdc16–Cdc26 derived from crystallographic analysis [91,92] with mass spectrometry data for SC8 and Apc13 EGFP labelling provided a quantitative estimate of APC/C subunit stoichiometry and molecular mass, with only the absolute stoichiometry of the small budding yeast-specific Apc9 still uncertain. Its association with Cdc27 [48] implies the possibility of two copies per complex. Saccharomyces cerevisiae APC/C therefore comprises 17–18 subunits with a molecular mass in the range of 1127–1158 kDa (table 2).

(e) A pseudo-atomic structure for the anaphase-promoting complex/cyclosome

Information on the APC/C structure is based on electron microscopy single-particle three-dimensional reconstruction studies of budding and fission yeast, human and Xenopus APC/C. Yeast APC/C structures were determined using cryo-electron microscopy (cryo-EM), whereas the human and Xenopus structures were determined using a novel approach of cryo-negative stain [50,110112]. All studies reveal that the basic overall shape of the APC/C resembles a triangular structure of approximately 200–230 Å in dimensions, featuring a central cavity surrounded by a protein wall (figure 4). At higher resolution (10 Å cryo-EM map), S. cerevisiae APC/C revealed a lattice-like scaffold generated from the multiple repeat motifs of most APC/C subunits defining a central cavity that contains the catalytic and substrate-recognition module (figure 5).

Figure 4.

Comparison of the EM structure of the APC/C from S. cerevisiae, S. pombe and human APC/C shows similar overall structures. (a) Negative-stain EM map of S. cerevisiae APC/C [63] and (b) the cryo-EM map of S. cerevisiae APC/CCdh1–D box [63] are compared with (c) the cryo-EM map of S. pombe APC/C at 27 Å resolution [112] and (d) the cryo-negative-stain EM map of human APC/C at 25–19 Å resolution [111]. Stars denote positions of subunits identified by antibody labelling. Reproduced with permission from da Fonseca et al. [63].

Figure 5.

Cryo-EM reconstruction of budding yeast APC/CCdh1–D box reveals the lattice-like architecture of the complex. Two views of the complex. Resolution is 10 Å.

Structurally the APC/C is divided into two overlapping subcomplexes, termed the TPR subcomplex and the catalytic subcomplex (SC8: containing the catalytic and substrate-recognition module and platform). Electron microscopy analysis of the S. cerevisiae TPR subcomplex (TPR6, comprising Cdc16, Cdc23 and Cdc27 with accessory subunits Cd26, Apc9 and Apc13) yielded a quasi-twofold symmetrical structure (figure 6a) [81]. TPR6 adopts an oval bowl-like architecture measuring 180 Å by 120 Å by 80 Å, characterized by well-defined tubular-like densities organized into a lattice arrangement. An electron microscopy reconstruction of SC8 revealed an asymmetric architecture, also comprising rod-like and globular features (figure 6b). Both TPR6 and SC8 closely match their corresponding regions in the intact APC/C, suggesting that these subcomplexes adopt stable autonomous conformations (figures 6) [81].

Figure 6.

Three-dimensional electron microscopy structure comparisons of recombinant APC/C and APC/C subcomplexes. (a) Superimposition of TPR6 (red) onto endogenous APC/CCdh1 (blue mesh) and (b) SC8 (yellow) onto recombinant APC/C (purple mesh). Reproduced with permission from Schreiber et al. [81].

Comparing the structures of the intact APC/C with SC8 and TPR6 subcomplexes and APC/C lacking defined subunits allowed individual APC/C subunits to be mapped within the overall APC/C molecular envelope [81]. For example, by comparing the apo-APC/C with APC/C lacking Cdc27–Apc9 (APC/CΔCdc27–ΔApc9), the difference density between these two structures was assigned to Cdc27 (figure 7a). The Cdc27 density showed an elongated shape at the top of the APC/C with twofold symmetry, consistent with Cdc27 being a homodimer. The globular portion of the assigned density corresponds to the Cdc27 dimerization domain, whereas the tubular densities extending out from either side of the dimerization domains correspond to the C-terminal TPR super-helices. A similar procedure was applied to fit Cdc16, through a superimposition of SC8 onto APC/CΔCdc27–ΔApc9 (figure 7b). The assigned Cdc16 density also displayed twofold symmetry, matching very closely the Cdc16–Cdc26 heterotetramer structure [91]. A small difference density feature, not accounted for by Cdc16–Cdc26, facing the central TPR cavity, was assigned as Apc13. SC8 and TPR6 share Cdc23 in common, thus the overlapping density between these two structures after superimposition onto the APC/C reference map was assigned to Cdc23. The density is dominated by a central globular domain from which two curved tubular features project in opposite directions with a pronounced structural resemblance to the assigned Cdc27 density, related by a dyad symmetry operation centred on the Cdc16 dimer axis. Cdc27 as a homology model for Cdc23 was an almost perfect fit to the assigned Cdc23 density, indicating that Cdc23 is a dimer structurally related to Cdc27.

Figure 7.

Three-dimensional localization of TPR subunits and atomic coordinate docking. (a) Three-dimensional localization of Cdc27–Apc9 by subtracting the APC/CΔCdc27ΔApc9 EM map from the recombinant APC/C map. The difference density is drawn as a grey mesh and used as restraints for Cdc27 docking. The two subunits within the homodimer are coloured in different shades of green. The symmetry axis of the Cdc27 homodimer is indicated in the right panel. (b) Three-dimensional localization of Cdc16–Cdc26–Apc13. The difference density (grey mesh) was calculated by subtracting the SC8 from the APC/CΔCdc27ΔApc9 EM map. The atomic coordinates of the S. pombe Cdc16–Cdc26 heterotetramer were used for rigid body docking. The two Cdc16 subunits within the heterotetramer are shown in red and light red and the Cdc26 N-terminus is shown in cyan. The molecular envelope corresponds to the APC/CΔCdc27ΔApc9 EM structure with density assigned to SC8 in yellow surface representation. The symmetry axis of the Cdc16–Cdc26 heterotetramer is indicated in the right panel. Reproduced with permission from Schreiber et al. [81].

Apc10 and Cdh1 were assigned to opposite sides of the inner cavity of the APC/C [63,113] (figure 8a). The disc-shaped density of Cdh1, 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, S. cerevisiae APC/C and APC/CCdh1 are similar, in contrast to the large conformational changes that accompany co-activator binding to vertebrate APC/C [110,111]. An ellipsoid-shaped density feature, resembling the β-sandwich of Apc10 [102,103], situated adjacent to but not in contact with Cdh1, was confirmed by analysis of reconstructions of APC/C lacking Apc10 [63,113] (figure 8b). 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 (figure 8c), indicative of an increased flexibility of the WD40 domain of Cdh1.

Figure 8.

Negative-stain EM reconstructions of S. cerevisiae APC/C show positions of Cdh1 and Apc10. Molecular envelopes of (a) APC/CCdh1, (b) apo APC/C, and (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. Reproduced with permission from da Fonseca et al. [63].

Guided by the density assignment of APC/C subunits and subunit stoichiometry, APC/C subunit coordinates were docked into the cryo-EM map of the APC/CCdh1–D box complex determined at 10 Å resolution [63]. The resultant pseudo-atomic model (figure 9) incorporated Cdc16, Cdc26, Cdc23 and Cdc27 of the TPR subcomplex and Apc2, Apc11 (N-terminal β-strand) and Apc10, as well as the co-activator Cdh1 [81]. The atomic fitting of the TPR subunits to the cryo-EM map accounts for the major density of the TPR lobe, and rationalized its repetitive layered architecture.

Figure 9.

Subunit organization and the pseudo-atomic model of APC/C. Atomic coordinates of Cdc16–Cdc26, Cdc23, Cdc27, Apc2, Apc10 and Cdh1 were docked in the 10 Å cryo-EM map of the APC/CCdh1–D box ternary complex represented in the grey mesh. The surface molecular boundaries of Apc1 (salmon) and Apc4–Apc5 (green) are indicated. Symmetry-related monomers of the Cdc16, Cdc23 and Cdc27 homodimers are represented in light and dark red, orange and green, respectively. Local twofold symmetry axes of Cdc27 and Cdc23 are indicated by diamonds. (a) View onto the central cavity orthogonal to the dyad axis of the Cdc27 homodimer. (b,c) Views related to (a) by rotations shown. (d) View approximately coincident with the Cdc16–Cdc26 dyad axis. Red spheres indicate the C-termini of Cdc16 and Cdc23, whereas red and blue spheres in Cdc27 denote the N- and C-termini of the inter-TPR insert. PC repeats of Apc1 are indicated. Reproduced with permission from Schreiber et al. [81].

Subunit assignments located Apc1, Apc4 and Apc5 to the platform of the APC/C (figure 9). Apc5 was assigned to the curved hook-like density of the Apc4–Apc5 molecular boundary based on its resemblance to a TPR superhelix. Combining this definition of Apc4–Apc5 with the localization of Apc2 and Apc10 [63], the remaining unassigned density within the platform region of the APC/C was assigned to Apc1. In the APC/C molecular envelope, this density assumed an ‘L’-shape comprising a rod-shaped feature connected to a globular disc-like density that links Apc2 to Cdc23 and incorporates the Apc1 PC repeats. Cdc23 is therefore connected to Apc5 and Apc1, with Apc4 interconnecting Apc1 and Apc5 at the opposite end of the platform. Cdh1 and Apc10 are positioned in close proximity and just adjacent to the TPR superhelices of the Cdc27 homodimer. Both Cdh1 and Apc10 bind to the C-terminal TPR superhelices of Cdc27 through their C-terminal IR tails [53,103,114] and one subunit of Cdc27 binds to Cdh1 whereas the other subunit binds to Apc10. In the pseudo-atomic model, Apc10 contacts the CTD of Apc2 [81], consistent with biochemical data that Apc10 associates preferentially with a subcomplex of Apc1, Apc2 and Apc11 [109]. The C-box at the N-terminus of the co-activator, an APC/C-interaction motif [57], previously proposed to bind the catalytic module through an undefined subunit [109], is positioned in close proximity to a conserved region of Apc2. Thus, the TPR subcomplex, together with Apc1, Apc4 and Apc5, coordinates the juxtaposition of the catalytic and substrate-recognition module subunits Apc2, Apc11 and Apc10 relative to co-activators.

Unambiguous density corresponding to the RING domain of Apc11 was not visible in the cryo-EM map. One reason for this could be the small size of the RING domain (approx. 7.5 kDa), potentially making it difficult to distinguish from other components. An alternative possibility is that in the ternary complex of APC/C, co-activator and substrate, the RING domain is flexible, similar to Rbx1 in activated neddylated SCF [115].

(f) Structures of APC/CCdh1 with substrates identified Cdh1 and Apc10 as the D box co-receptor

Substrate-binding sites on APC/CCdh1 were identified using a fragment of Hsl1, a D-box (RxxLxxI/VxN) [43) and KEN-box [46]-containing substrate with a high affinity for APC/CCdh1 [47,64], in two independent studies [63,113]. In the study of S. cerevisiae APC/C, engagement of Hs1l with APC/CCdh1 was accompanied by a pronounced structural change involving Cdh1 and Apc10 (figure 10). Specifically, the β-propeller domain of Cdh1 became bulkier, shifted approximately 7 Å towards Apc10 and new, well-defined density bridged Cdh1 to Apc10. Thus, Hsl1 promoted the formation of new connections between Cdh1 and Apc10, a result consistent with direct co-activator–substrate interactions [47,53,5558] and a role for Apc10 in mediating optimal substrate binding [48,62,64]. A synthetic peptide modelled on the D box of S. pombe Cdc13 generated similar structural changes to Hsl1; specifically, the WD40 domain of Cdh1 was shifted and new density connected it with Apc10. However, in contrast to the APC/CCdh1–Hsl1 map, the extent of new density associated with Cdh1 is markedly reduced, indicating that the additional density in APC/CCdh1–Hsl1 represented the larger Hsl1 substrate.

Figure 10.

Negative-stain EM reconstructions of the APC/C show that substrate binding to APC/CCdh1 involves Cdh1 and Apc10. Negative-stain EM reconstructions of (a) the S. cerevisiae APC/CCdh1–Hsl1 complex, (b) S. cerevisiae APC/CCdh1–D box, (c) S. cerevisiae APC/CCdh1–KEN box, (d) human APC/CCdh1 (Cdh1 in red), and (e) human APC/CCdh1–Hsl1 (Hsl1 in purple). Lower panels in (ac) 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 the mesh. Panels (ac) reproduced with permission from da Fonseca et al. [63] and (d,e) reproduced with permission from Buschhorn et al. [113].

The cryo-EM reconstruction of the APC/CCdh1–D box ternary complex of APC/CCdh1 in complex with a D box peptide showed density connecting Cdh1 and Apc10 (figure 11). Docking the crystal structure of Apc10 [102,103] and the modelled Cdh1 WD40 domain into their respective densities indicated additional unassigned density linking Cdh1 to Apc10 (figure 11). The best fit of Apc10 into the cryo-EM map positioned a highly conserved loop, required for D box recognition [106], adjacent to the density linking Apc10 with Cdh1. In contrast, residues on Apc10's opposite surface that contribute to APC/C interactions [106] are oriented towards Apc2 (figure 11). 15N-HSQC NMR, a technique suitable for detecting weak protein–ligand interactions, confirmed direct Apc10–D box interactions [63]. Binding of the KEN box peptide to APC/CCdh1 also promoted a repositioning of Cdh1 towards Apc10, but notably without the connecting density (figure 10). This indicated that only D box substrates promote a physical interconnection between Cdh1 and Apc10, consistent with the lack of direct KEN box–Apc10 interactions revealed through NMR studies [63].

Figure 11.

Cdh1, Apc10, Apc2 and Apc11 form a substrate-recognition catalytic module. (a) View of the cryo-EM APC/CCdh1–D box complex. Protein density is represented by a mesh with fitted atomic coordinates of the Cdh1 β-propeller (modelled), Apc10, Apc2–Apc11 and Cdc27. The two subunits of Cdc27 are shown in light and dark green. The view shows the twofold symmetry axis of Cdc27. Density connecting Cdh1 to a TPR superhelix of the Cdc27 dimer is indicated by an arrow. TPR motifs 8–10 of Cdc27, implicated in IR tail recognition [64], are shown in lighter colours. (b,c) Modelling of the D box peptide into density bridging Cdh1 and Apc10. (b) As an 8-residue α-helix. (c) As an extended chain. In both instances, the binding site for the D box is shared between Cdh1 and Apc10. (d) Schematic of the 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. Reproduced with permission from da Fonseca et al. [63].

Structural data revealing that Cdh1 and Apc10 became interconnected by bridging density in the presence of D box substrates rationalized biochemical studies demonstrating that both the co-activator and core APC/C subunits [47,53,5558,61], specifically Apc10 [48,62,64,106], contribute to D box-dependent recognition and processive ubiquitylation. Cdh1 and Apc10 therefore generate a D box co-receptor. Individually, the co-activator and APC/C possess low affinity and specificity for the substrate [55] and therefore cooperatively enhance substrate affinity through multi-valency.

4. Processivity requires two E2 activities

Proteasome-dependent degradation of ubiquitylated protein substrates is determined by a ubiquitin polymer composed of at least four ubiquitin moieties [116,117]. In addition to Lys48 linkages, others including Lys11, Lys29 and single ubiquitin moieties are recognized for degradation by the proteasome [118122]. Human APC/C has been shown to catalyse the formation of Lys11-linked polyubiquitin [118], whereas S. cerevisiae APC/C generates Lys48 linkages [123].

(a) E2s define polyubiquitin linkage

For humans and S. cerevisiae, two E2s are responsible for generating a polyubiquitin chain through a sequential mechanism. For S. cerevisiae APC/C, Ubc4 catalyses the rapid monoubiquitylation of multiple lysines on APC/C substrates, whereas Ubc1 catalyses Lys48-linked polyubiquitin chain assembly onto a pre-attached ubiquitin moiety [123]. The capacity of Ubc1 to elongate an existing ubiquitin modification is conferred by its unusual C-terminal UBA domain. Deletion of the UBA domain results in shorter reaction products than those produced with wild-type Ubc1, indicating that the UBA domain confers processivity on the APC/C–E2 complex. However, specificity for Lys48 linkages is retained in the absence of the UBA domain. In humans, the E2 Ube2S is the E2 responsible for extending APC/C substrates primed by UbcH10 or UbcH5 [124126]. Insights into the mechanism of specific polyubiquitin chain linkage have been provided by biochemical and structural analysis of the elongating E2s, Ubc1 of S. cerevisiae [127] and Ube2S of humans [128]. In the case of Ube2S, the Ubc domain binds both donor (ubiquitin covalently attached via its C-terminal Gly residue through a thioester bond to the catalytic cysteine residue) and acceptor ubiquitin, orientating the Lys11 of the acceptor ubiquitin optimally for nucleophilic attack onto the donor ubiquitin. Acceptor ubiquitin also promotes a catalytically competent conformation of Ube2S to facilitate its reaction with the donor ubiquitin [128]. The identification of a cluster of polar residues in the E2 Ubc1 and within ubiquitin itself, necessary to mediate specific Lys48-linked chain elongation, suggests a related mechanism for chain elongation by S. cerevisiae APC/C-Ubc1 [127].

5. The mitotic checkpoint complex disrupts anaphase-promoting complex/cyclosome: co-activator substrate-recognition sites

(a) The spindle assembly checkpoint complex is exerted by the mitotic checkpoint complex

The SAC ensures correct segregation of sister chromosomes, delaying anaphase onset until all chromosomes have achieved bipolar kinetochore–microtubule attachment. The assembly checkpoint is mediated by the MCC comprising the checkpoint proteins Mad2, Mad3/BubR1, Bub3 and Cdc20 [129131]. Mad2 and Mad3/BubR1 bind directly to Cdc20 and can inhibit APC/CCdc20 activity in vitro [130,132136]. Mad3/BubR1 and Bub3 interact directly [129] via a WD40 domain in Bub3 and the GLEBS motif in Mad3/BubR1 [137,138]. Functional MCCs assemble in vitro from recombinant Mad2, Mad3/BubR1, Bub3 and Cdc20 proteins in the absence of kinetochores, although at a much slower rate [139]. A single unattached kinetochore is sufficient to inhibit all APC/CCdc20 activity, owing to a catalytic amplification effect of the kinetochore-induced APC/CCdc20 inhibitor. Unattached kinetochores catalytically generate a diffusible Cdc20 inhibitor. Immobilized Mad1–Mad2 at kinetochores provides a template for initial assembly of Mad2 bound to Cdc20 that is then converted to a final mitotic checkpoint inhibitor with Cdc20 bound to Mad3/BubR1 [139144]. Additionally, the checkpoint may act via promoting Cdc20 degradation since checkpoint proteins including Mad3/BubR1 are required for APC/C-mediated degradation of Cdc20 [145,146].

The molecular mechanisms underlying MCC-mediated inhibition of APC/C activity towards securin and cyclin B are related to the pseudo-substrate-based inhibition of APC/C by Acm1 (D box- and KEN box-dependent) [147149] and Emi1 (D box-dependent) [150], notions consistent with findings that the presence of MCC components on the APC/C leads to strongly reduced substrate binding [111]. The N-terminus of Mad3/BubR1 functions as a pseudo-substrate inhibitor binding to Cdc20 and blocking substrate interactions [151154]. Two KEN boxes of Mad3/BubR1 are required for optimal SAC function, whereas the D box (present in S. cerevisiae) and the N-terminal KEN box are required for Cdc20 binding. Supporting the notion that these motifs mediate binding of Mad3/BubR1 to Cdc20 via a mechanism that is reminiscent of Cdc20–substrate interactions, binding to common overlapping sites, Mad3/BubR1 binding to Cdc20 is mutually competitive with that of the substrate Hsl1, and peptides corresponding to the D box and the KEN box of Hsl1 compete with Mad3/BubR1 for binding to Cdc20 [151]. Recent studies extend these findings by showing that the Cdc20-binding domain within the N-terminal region of Mad3/BubR1, which adopts a TPR-like structure [155], acts as a soluble pseudo-substrate inhibitor of Cdc20 during interphase to inhibit cyclin B ubiquitylation, allowing cyclin B accumulation prior to mitosis onset [153].

A key question for understanding regulation by the SAC is the mechanism of its inactivation. In one scenario, it is possible that appropriate attachment of all kinetochores to the mitotic spindle terminates the conformational activation of Mad2. However, other models have recently been proposed based on observations that Cdc20 undergoes auto-ubiquitylation catalysed by the APC/C in conjunction with UbcH10, which is reversed by the DUB, Usp44 [156,157]. These studies showed that Cdc20 ubiquitylation releases APC/CCdc20 from the SAC by promoting the disassembly of Cdc20 and Mad2, in a process counteracted by Usp44. The study of Pines and colleagues [145] showing that Cdc20 ubiquitylation leads to its degradation, thereby imposing the SAC, clearly contradicts that of Reddy et al. [156] and Stegmeier et al. [157]. Another factor contributing to SAC inactivation is p31comet, which promotes disassembly of the MCC in an ATP-dependent process [158]. One proposed mechanism of p31comet action is to block Mad2 activation through structural mimicry [159].

(b) Structure of the anaphase-promoting complex/cyclosome–mitotic checkpoint complex

The position of the MCC bound to human APC/C isolated from cells arrested by the SAC has been visualized using electron microscopy [111] (figure 12). The MCC was localized to the front-end of the platform domain. Cdc20 alone was located to a site that partially overlaps with the MCC, suggesting that the precise location of Cdc20 changes on MCC binding to the APC/C. The co-localization of the MCC to the site of co-activator binding is consistent with the notion that the MCC inhibits APC/C activity by blocking Cdc20–substrate interactions. Moreover, the MCC promotes a more closed rigid conformation of the APC/C, concomitant with rearrangement of Apc4, which might prevent the conformational changes potentially required for the catalytic cycle. Superimposing the pseudo-atomic coordinates based on S. cerevisiae APC/C [81] into the human APC/C–MCC molecular envelope [111] reveals an exceptionally good fit, underlying the similarity between the human and S. cerevisiae structures (figure 12). Unassigned density in the human APC/C–MCC map represents the MCC situated in the lower region of the APC/C cavity in contact with Cdc23 and Apc4, whereas Apc7, unique to vertebrate APC/C, is situated at the head of the TPR lobe.

Figure 12.

Docking of the pseudo-atomic model of S. cerevisiae APC/C into the electron microscopy-derived molecular envelope of the human APC/C–MCC complex reveals the position of the MCC and Apc7. Two views of the complex with the molecular envelope of human APC/C–MCC [111] are shown in mesh. Coordinates from Schreiber et al. [81]. Note the downward shift of the co-activator (here shown as Cdh1 in purple) when bound to the APC/C compared with the MCC.

6. Switching Cdc20-binding sites on the anaphase-promoting complex/cyclosome on release of the spindle assembly checkpoint may be correlated with a switch to recognition of D box-dependent substrates

The quasi-symmetry of the TPR subcomplex may have functional and evolutionary implications. Cdc23 and Cdc27 share a similar structure and are symmetrically disposed about Cdc16. The TPR motifs of Cdc27 implicated in binding the IR tails of the co-activator and Apc10 are highly conserved in Cdc23, and mutations of equivalent conserved residues in either Cdc23 or Cdc27 diminished co-activator binding [64,72]. This might suggest multiple IR tail-binding sites on the APC/C. One possible prediction might be that the co-activator switches binding between Cdc23 and Cdc27 during different phases of the cell cycle, specifically during the transition from prometaphase to metaphase, which coincides with a switch of substrate specificity from substrates that bind the APC/C independently of the co-activator, Apc10 and D box to a binding mode that is dependent on the co-activator, Apc10 and D box [72]. Specifically, during metaphase, when APC/CCdc20 ubiquitylates D box substrates in an Apc10-dependent manner, also dependent on Cdc27 [72], a functional D box co-receptor would be required. Since the D box co-receptor is generated from both co-activator and Apc10 [63], this would require the co-activator to move towards Cdc27, hence explain the requirement for Cdc27 to ubiquitylate D box substrates at metaphase [72]. In the APC/C–MCC complex, when Cdc20 is displaced towards Cdc23 and Apc5, away from Cdc27 [111], the D box co-receptor cannot be generated. This may explain how the MCC blocks recognition of D box-containing substrates. In addition, the KEN box pseudo-substrate of MAD3 by binding to Cdc20 blocks KEN box substrate interactions with Cdc20. These interactions may also sterically interfere with D box binding to Cdc20.

Another consequence of positioning the co-activator in close proximity to Cdc23 and Apc5 at prometaphase is that the IR tail-binding site on Cdc27 is available for potential interactions with prometaphase substrates. Because the C-terminal Met-Arg sequence of Nek2A, responsible for mediating APC/C interactions [67], is structurally related to the IR tails of co-activators and Apc10, the Nek2A MR tail may engage the IR tail-binding site of Cdc27. In contrast, cyclin A is recruited to the APC/C through its binding partner Cks [71] that recognizes the phosphorylated Cdc27 subunit of the TPR lobe [72].

7. Conclusions and future perspectives

Critical to the function of the APC/C is the modulation of its substrate specificity according to different phases of the cell cycle. Switching of Cdc20 and Cdh1 alters substrate specificity on progression from metaphase to late anaphase. However, the APC/C recognizes different substrates when in complex with Cdc20 at prometaphase and metaphase, and this is due to both repositioning of Cdc20 within the APC/C central cavity and direct steric hindrance of the KEN box-binding site on Cdc20 mediated by MCC subunits. Dynamic changes in co-activator position relative to core APC/C subunits may provide combinations of substrate-recognition interfaces. For example, the interface between the co-activator and Apc10 creates the D box co-receptor, only possible with the co-activator in contact with Cdc27 following release of the SAC.

Whether dynamic structural changes in core APC/C subunits are fundamental to APC/C activity is not clear. In one scenario, it is possible to envision dynamic movements of the Apc11 RING domain, similar to the liberation of the Rbx1 RING domain of the SCF promoted by Cul1 neddylation [115]. One mechanistic question that remains unexplained concerns the role of Cdc20 at prometaphase for the ubiquitylation of cyclin A and Nek2A, but not to mediate their binding to the APC/C [67,72]. The N-terminus of Cdc20, without its substrate-binding WD40 domain, activates the E3 ligase catalytic activity of the APC/C [160]. This activation of catalytic activity may result from conformational changes of Apc2 and Apc11 exerted by the co-activator, possibly involving release of Apc11's RING domain.

Further structural studies at higher resolution, either by cryo-electron microscopy or by crystallography, will resolve these and other questions. The availability of recombinant functional APC/C will facilitate such studies by making available sufficient quantities of the APC/C and allowing the manipulation of APC/C subunits. The latter will enable biochemical and biophysical experiments to probe the roles and mechanisms of individual domains, subunits and subcomplexes, and allow the modification of APC/C subunits for single-molecule analysis that will be particularly relevant to addressing questions of APC/C dynamics.

Acknowledgements

Work in the author's laboratory is funded by Cancer Research UK. We thank Jan-Michel Peters and Holger Stark for figure 10d,e.

Footnotes

References

View Abstract