Vol. 16, No. 17, pp. 2179-2206, September 1, 2002

REVIEW
The anaphase-promoting complex: it's not just for mitosis any more

¹ Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas 77030, USA; ² Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520-8024, USA

	Introduction

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

The ability of cells to make exact replicas of themselves is central to the life and development of complex organisms. Initial insights into the question of how cells divide came during the latter half of the 19th century when Walther Flemming visualized structures he called threads (which we now call chromosomes) and described how these threads change during cell multiplication, a process he called mitosis. Now, more than a century later, we have a molecular understanding of many of the cellular processes that Flemming observed. Indeed, major cytological events occurring during mitosis are known to constitute cell cycle transitions and are regulated by complex signal transduction pathways whose major components have been identified during the past decade. In this review, we describe recent efforts to understand how central components of this regulatory apparatuscyclin-dependent kinases and the anaphase-promoting complex/cyclosome (APC/C)control progression through the cell division cycle and how regulatory mechanisms impinge on the APC/C. The APC/C is the multisubunit ubiquitin ligase whose activity is precisely regulated to ensure the timely degradation of cyclins and other key cell cycle regulators in unperturbed cells and to respond to mitotic checkpoints that prevent their degradation. We pay particular attention to recent developments as excellent reviews are available from a few years ago (Morgan 1999; Zachariae and Nasmyth 1999).

	Cell cycle transitions: interplay between cyclin-dependent kinases and ubiquitin-mediated proteolysis

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

The primary task of the cell division cycle is to duplicate genetic information precisely through the process of DNA replication (S phase) and then to allocate this information equally to two daughter cells through mitosis. Inaccuracies in this process can be problematic. For example, cells that attempt to separate chromosomes that are incorrectly or incompletely duplicated are much more likely to incur fatal or irreparable damage as a result of either loss or gain of genetic information. Thus, a large number of signaling pathways collaborate to enforce order on cell division events.

Cell cycle transitions are the primary mechanism used by the cell to establish the order and timing of cell cycle events. Such transitions occur when there is a change in the biochemical status of the cell division machinery. Early cell-fusion experiments showed that major cell cycle phases can be incompatible with one another. For example, when a G₂ cell is fused with an S-phase cell, the G₂ nucleus waits until the S-phase nucleus has completed replication before both nuclei enter mitosis synchronously (Rao and Johnson 1970). Thus, progression through G₂ into mitosis is incompatible with ongoing DNA synthesis. Through subsequent genetic and biochemical analysis, we now understand in general terms how these cell cycle dependencies are generated and controlled. Moreover, molecules that play key roles in defining particular cell cycle stages have been uncovered. One frequently used paradigm involves an inhibitor-activator module; a protein complex in which an activator of a particular transition is held in an inactive form by an inhibitory factor (Fig. 1A). The definition of a specific component as an inhibitor or activator is frequently complicated by the fact that a given protein may perform both roles; an activator of a particular transition may become an inhibitor of a subsequent transition. A case in point is the Cdk inhibitor Sic1 in budding yeast. Sic1 promotes mitotic exit by inhibiting the activity of mitotic cyclin/Cdk complexes (Clb2/Cdc28) but blocks S-phase entry by inhibiting a closely related pair of S-phase Cdk complexes (Clb5-6/Cdc28). As described in more detail below, the mitotic inhibitor Pds1 and B-type cyclins display similar dual functions. Thus, molecules that function as both inhibitors and activators of cell proliferation are frequently nodes of regulation and are used as focal points for the integration of multiple signaling pathways that monitor ongoing and completed cellular events and that link these with transitions.

View larger version (22K): [in this window] [in a new window]   Figure 1.   (A) Schematic representation of a common mechanism used to control cell cycle transitions. When an activating component A is associated with an inhibitory component I, the cell cycle transition is blocked. In response to appropriate signals, association of A with I is abolished (in some cases through proteolytic destruction of I) and A can activate the transition. In some cases, I has served a positive role in a previous step in the cell cycle. (B) Degradation of protein substrates through the ubiquitin-proteasome pathway involves three components, E1, E2, and E3. E1 uses ATP to form a high-energy thiol ester with the C-terminal glycine of ubiquitin. This ubiquitin is then transferred to a cysteine residue in one of several E2s. Specific E2s assemble with appropriate E3s to initiate transfer of ubiquitin to an associated substrate. Multiple rounds of ubiquitination of the initial ubiquitin conjugate lead to the formation of polyubiquitin chains, which are then recognized by the proteasome, which degrades the ubiquitinated protein. (C,D) Regulation of ubiquitination can occur through multiple mechanisms. Some ubiquitination reactions (C), such as those involving known SCF-mediated pathways, require that the substrate be phosphorylated to be recognized by the E3. In other cases such as the APC (D), the E3 is the target of regulation by phosphorylation, and substrate recognition does not require that the substrate be phosphorylated. With the APC (D), phosphorylation can either act to positively (pathway a) or negatively (pathway b) regulate activity, depending on the context.

Ubiquitin-mediated degradation of inhibitory factors is a primary mechanism by which a change in cell cycle state is achieved. This process involves the covalent attachment of ubiquitin chains to lysine residues in a target protein, leading to its recognition and degradation by the proteasome. The specificity of ubiquitin-mediated proteolysis is exquisite, allowing a single protein within a larger complex to be modified and destroyed. Moreover, the fact that ubiquitin-mediated proteolysis is both rapid and irreversible means that the change in cell cycle state can occur in a unidirectional switch-like fashion. Ubiquitination involves three major steps (Fig. 1B; Hershko and Ciechanover 1998). First, the ubiquitin-activating enzyme E1 uses ATP to generate a thiol ester between its active-site cysteine and the C-terminal glycine residue of ubiquitin. Second, the ubiquitin is transferred to the active-site cysteine of an E2 (ubiquitin-conjugating enzyme). Third, the E2 assembles with an E3 (ubiquitin ligase) and then transfers ubiquitin to one or more lysine residues of a substrate protein to generate a stable isopeptide linkage. Multiple rounds of ubiquitin transfer to the initial ubiquitin conjugate lead to the formation of polyubiquitin chains, which are then recognized by the regulatory cap of the proteasome. E3s are key components of ubiquitination pathways because they determine the substrate specificity of ubiquitination reactions, they recruit the appropriate E2, and they may also contribute to E2 activity.

Because the degradation of key cell cycle regulators typically occurs in response to the fulfillment of multiple criteria, it is not surprising that complex signaling systems impinge on key ubiquitination reactions. Many of these signaling systems involve input from a protein kinase pathway and can act either positively or negatively. In some cases, the substrate itself is the recipient of regulatory information, and its phosphorylation sets up the timing of its ubiquitination. For example, many substrates of the SCF (Skp1/Cul1/F-box protein) ubiquitin ligase need to be phosphorylated prior to binding to the SCF, which can be constitutively active (for review, see Deshaies 1999; Koepp et al. 1999). In other cases, such as the APC/C, protein kinases affect primarily the activity status of the E3 in either positive or negative ways (for review, see Morgan 1999; Zachariae and Nasmyth 1999). Of course, variations on these themes exist, including the use of coactivators and inhibitory subunits.

	Cycling into destruction: discovery of the APC/C

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

The APC/C and SCF ubiquitin ligases have emerged from studies aimed at understanding how key components of the cell cyclecyclins and cyclin-dependent kinase inhibitorsare regulated. An unexpected surprise was that the discovery of these ligases provided not only the key to understanding how Cdk activity is regulated but also insights into how other basic processes such as sister-chromatid separation are controlled, and how numerous phosphorylation-driven signaling pathways are regulated.

Initial insight into the problem of regulated proteolysis of cell cycle molecules came from an attempt to understand how mitotic cyclins are degraded as cells pass through mitosis (for review, see Zachariae and Nasmyth 1999). Multiple systems contributed to developments in this area, but experiments in clam, sea urchin, and Xenopus egg extracts were crucial in framing the problem. Cyclins A and B were identified in marine invertebrates as proteins that are translated in response to fertilization but rapidly degraded at each cleavage of the early embryo, suggesting that this disappearance was cell cycle regulated (Evans et al. 1983). Efforts to link A- and B-type cyclins with maturation promoting factor, an activity required for the interphase-to-metaphase transition, and with the Cdc2 protein kinase simultaneously led to the finding that cyclin B alone is sufficient to induce mitosis in interphase egg extracts and that cyclin B degradation occurs precisely as cells exit mitosis (Luca and Ruderman 1989; Murray and Kirschner 1989; Murray et al. 1989). These early experiments also revealed that cyclin B degradation could be triggered in interphase by addition of active Cdc2 (foreshadowing the role of Cdk-mediated APC/C phosphorylation in its activation) and that cyclin B degradation required elements in its N terminus (foreshadowing the elucidation of a motif, the "destruction box," that serves as a recognition element for the APC/C; Murray et al. 1989; Félix et al. 1990). In fact, the first suggestion that cyclin B degradation was required for exit from mitosis came from the finding that sea urchin cyclin B90 (lacking its N-terminal 90 amino acids) could activate Cdc2 and induce mitotic entry but that it was stable and prevented mitotic exit (Murray et al. 1989).

Efforts to elucidate the mechanism underlying cyclin B degradation led to the finding that cyclin B undergoes covalent modification to form a ladder of proteins with reduced electrophoretic mobility in mitotic extracts but not in interphase extracts (Glotzer et al. 1991). The spacing between these modified forms (~7 kD) suggested that ubiquitin might be involved. At the time, E1 and E2 components had begun to be defined, but a major gap in the ubiquitination field concerned the nature and identity of E3s. Moreover, actual in vivo substrates of the ubiquitin-proteasome system were poorly defined. Thus, cyclin B became an important model system for defining E3s and for understanding how ubiquitin-mediated proteolysis controlled cell cycle transitions, ultimately revealing that the E3 activity required for cyclin B degradation is also used to destroy other important mitotic regulators.

Initial insight into the cyclin B ubiquitin ligase came from its partial purification from marine invertebrates and Xenopus. Reconstitution of cyclin B ubiquitination activity from fractionated extracts required E1, an E2 activity, and an E3 activity that had the properties of a large complex, with an estimated size of 1500 kD (20S; Hershko et al. 1994; King et al. 1995; Sudakin et al. 1995). The identification of components of the E3 was advanced by the observation that B-type cyclins are not only unstable during mitotic exit but also for an extended period in G₁ in budding yeast (Amon et al. 1994). This instability during G₁ facilitated the development of a genetic screen that allowed the identification of several genes, including CDC16 and CDC23, which, when mutated, blocked mitotic cyclin degradation during G₁ (Irniger et al. 1995). Cdc16 and Cdc23 contain a repetitive protein-protein interaction domain, the TPR motif, but this provided few clues as to how these proteins promoted B-type cyclin degradation. Furthermore, extracts from cdc16 and cdc23 cells (as well as two additional mitotic-arrest mutants, cdc26 and cdc27) were defective in cyclin ubiquitination, as expected if these proteins were part of an E3 (Zachariae and Nasmyth 1996). Importantly, cdc16 and cdc23 mutants are not only defective in exit from mitosis but are also defective for separation of sister chromatids at the metaphase-to-anaphase transition. These findings provided genetic evidence that the machinery used for cyclin degradation is also involved in other aspects of mitosis (Irniger et al. 1995; for review, see Nasmyth 1999), an idea that was previously proposed based on the finding that the proteasome, but not cyclin B destruction, is required for anaphase in Xenopus extracts (Holloway et al. 1993). The availability of antibodies to human Cdc16 and Cdc27 made it possible to show directly that these proteins are present in purified APC/C complexes from human cells and Xenopus egg extracts, and also made it possible to show that anti-Cdc27 immune complexes contain cyclin-B E3 activity (King et al. 1995). The identification of essential APC/C components not only provided the building blocks with which to define the complex in greater detail (see below) but also, together with the identification of APC/C-targeting sequences, the Destruction and KEN boxes, facilitated the identification of APC/C substrates other than mitotic cyclins.

	APC/C composition and structure: relationship with the SCF ubiquitin ligase

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

The large size of the purified cyclin-B E3 suggested that components in addition to Cdc16 and Cdc23 would be present. These components were identified biochemically from both yeast and Xenopus, and the corresponding human cDNAs were identified (Table 1; Peters et al. 1996; Zachariae et al. 1996, 1998b; Yu et al. 1998). Several APC/C components were also identified in a collection of cut mutants in Schizosaccharomyces pombe (Yamashita et al. 1999; Tatebe and Yanagida 2000). Budding yeast APC/C contains 11 core subunits that remain tightly associated throughout purification and are all required for timely mitosis. Like the TPR-repeat-containing subunits, temperature-sensitive mutations in APC1, APC2, and APC11 arrest in metaphase at the nonpermissive temperature because of an inability to induce loss of sister-chromatid cohesion (Zachariae et al. 1996, 1998b; K.M. Kramer et al. 1998). These mutants are also defective in B-type cyclin degradation in vivo and cyclin B ubiquitination in extracts (Zachariae et al. 1998b). Apc9, although not essential for viability, is nevertheless required for efficient entry into anaphase. Doc1 was originally identified in budding yeast (Hwang and Murray 1997) and later found as a component of the human and S. pombe APC/C, referred to as APC10 (Kominami et al. 1998; Grossberger et al. 1999; Kurasawa and Todokoro 1999). Doc1 is required for Clb2 degradation during mitotic exit in budding yeast. Although a role for Doc1 in early mitotic events has not been reported, a radiation-induced mouse mutant called oligosyndactylism displays defects in the metaphase-to-anaphase transition. This mutation appears to inactivate the APC10/Doc1 gene (Pravtcheva and Wise 2001). Detailed analyses of Apc4 and Apc5 mutants have not been reported in budding yeast. However, APC5/ida mutants in Drosophila and APC4/emb-30 mutants in Caenorhabditis elegans produce metaphase arrests (Furuta et al. 2000; Bentley et al. 2002). At present, Apc9 appears to be unique to the budding yeast APC/C (Zachariae et al. 1998b), whereas Apc7 has only been found in metazoans (Table 1; Yu et al. 1998). It is possible that additional APC/C subunits exist, as a small number of proteins in APC/C immune complexes from yeast have not yet been identified (Zachariae et al. 1998b).

The SCF ubiquitin ligase and its role in Cdk regulation

The majority of APC/C subunits have, as yet, unknown functions and in many cases lack conserved domains that would suggest a function. Major exceptions to this generalization are the Apc2 and Apc11 subunits, which display sequence identity with core components of the SCF ubiquitin ligase (Fig. 2A; Yu et al. 1998; Zachariae et al. 1998b). SCF complexes are modular E3s that contain a core ubiquitin ligase composed of Cul1/Cdc53, Skp1, the ring finger protein Rbx1/Roc1/Hrt1, and a member of the F-box family of proteins, which serves as the substrate receptor (Fig. 2A; Bai et al. 1996; Feldman et al. 1997; Skowyra et al. 1997). The SCF ubiquitin ligase was discovered during efforts to identify the mechanism of degradation of the Sic1 Cdk inhibitor and G₁ cyclins in budding yeast, but it is now recognized as one of the largest families of ubiquitin ligases in eukaryotes (for review, see Patton et al. 1998; Deshaies 1999; Koepp et al. 1999). Initial insight into Sic1 turnover came from an analysis of temperature-sensitive yeast mutants that blocked DNA replication but not budding or spindle-pole-body duplication. At the nonpermissive temperature, cdc53, cdc4, skp1, and cdc34 strains arrest in G₁ with elevated levels of Sic1 and reduced Clb5/Cdc28 activity, whereas analogous mutant cells also lacking Sic1 proceed through S phase and arrest in mitosis (Schwob et al. 1994; Bai et al. 1996). This suggests a role for SCF^Cdc4 in mitosis, but the nature of this role and the substrates of this E3 during mitosis have yet to be identified.

View larger version (35K): [in this window] [in a new window]   Figure 2.   (A) Domain structures of SCF, CBC, and APC/C complexes. All three types of E3s contain a core ubiquitin ligase composed of a cullin-homology domain (CHD) containing protein (yellow), a RING-H2 protein (red/orange), and an adapter such as Skp1 (purple; unknown in the case of the APC/C) to facilitate interaction with substrate receptors (F-box proteins, BC-box proteins, and Cdc20/Cdh1 proteins). The RING-H2 finger assembles with the E2. The contacts made between various APC/C subunits are not known. (B) Schematic representation of the CHD of the SCF complex emphasizing the subdomains that form the cavity where the RING-H2 protein Rbx1 binds (red; see text for details). (C) Conservation between Apc11 and Rbx1. The structure of an Rbx1/Cul1(CHD)/Ubc7 complex was used to display the conservation between Apc11 and Rbx1. The CHD is shown in yellow, Rbx1 in red, and Ubc7 in magenta. Residues conserved in APC11 and Rbx1 are shown in red spacefill. Residues in spacefill blue correspond to those that are conserved in Apc11 and Rbx1 and that are thought to interact with the E2 (Trp 87, Pro 95, Leu 96 in Rbx1). Residues in Ubc7 that are thought to contact the ring domain of Rbx1 are shown in magenta spacefill (Pro 62, Phe 63, Pro 97). The active-site cysteine of Ubc7 is shown in cyan.

Although the fact that CDC34 encodes a ubiquitin-conjugating enzyme suggested a role for ubiquitin-mediated proteolysis in Sic1 regulation (Schwob et al. 1994), it was unclear how these genetically defined components cooperated to regulate Sic1 turnover. The finding that CDC34 and CDC53 were also genetically required for degradation of G₁ cyclins (Tyers et al. 1992; Deshaies et al. 1995; Lanker et al. 1996; Mathias et al. 1996; Willems et al. 1996) suggested that these components functioned in multiple proteolysis pathways, but it was not clear how different substrates would be targeted or recognized. The identification of Skp1 as a high-copy suppressor of cdc4 mutants (Bai et al. 1996) provided an answer to this question. Like cdc53 and cdc34 mutants, skp1 mutants display defects in both Sic1 and G₁ cyclin turnover. Moreover, Skp1 was found to interact with a 40-amino-acid domain in Cdc4 called the F-box, which was also found in a number of other proteins (Bai et al. 1996). One of these proteins, Grr1, had been implicated in turnover of G₁ cyclins (Barral et al. 1995). This, together with genetic interactions among SKP1, CDC53, and CDC34, led to the idea that these proteins function together with different F-box proteins to direct the ubiquitination of different targets (Bai et al. 1996). This prediction proved correct as reconstituted SCF^Cdc4 complexes can direct the ubiquitination of Sic1, which interacts with the WD40 repeats in the C terminus of Cdc4 in a phosphorylation-dependent manner (Feldman et al. 1997; Skowyra et al. 1997), whereas SCF^Grr1 complexes can promote the ubiquitination of G₁ cyclins (Seol et al. 1999; Skowyra et al. 1999), which bind to leucine-rich repeats in Grr1 in a phosphorylation-dependent manner (Skowyra et al. 1997; Hsiung et al. 2001).

SCF-mediated ubiquitination is used to control the flux through multiple signaling pathways, including transcriptional control pathways, developmental pathways, and hormone signaling pathways in plants. In all the cases analyzed thus far, protein turnover through the SCF pathways depends on phosphorylation of the substrate, allowing it to interact with the appropriate F-box protein (Deshaies 1999; Koepp et al. 1999). Recent data indicate that turnover of Sic1 is ultrasensitive to Cln/Cdc28 concentrations (Nash et al. 2001), providing a mechanism to link the timing of Clb/Cdc28 activation with G₁ cyclin levels and nutrients. An ultrasensitive response occurs when a system responds to a graded input with an output that is switch-like or "all-or-none." Ultrasensitivity in the case of Sic1 destruction arises because multiple phosphorylation events are required to allow it to engage SCF^Cdc4. Sic1 contains nine Cdk phosphorylation sites, and at least six of these need to be phosphorylated for efficient recognition by SCF^Cdc4 (Verma et al. 1997; Nash et al. 2001). Assuming that Cdc28-mediated phosphorylation of Sic1 occurs via a dissociative mechanism, this is expected to generate an ultrasensitive response with a Hill coefficient >6 (Nash et al. 2001). The Hill coefficient could actually be much larger owing to positive feedback via Clb5/Cdc28 activity upon Sic1 destruction (for review, see Deshaies and Ferrell 2001; Harper 2002). Interestingly, none of the individual phosphorylation sites in Sic1 conform to the optimal phospho-peptide recognition motif for Cdc4, but replacing a single phosphorylation site with an optimal Cdc4 recognition sequence in an otherwise nonphosphorylatable Sic1 protein makes the modified Sic1 more unstable than wild-type Sic1 (Nash et al. 2001). This change leads to defects in G₁/S control in vivo and suggests that the use of suboptimal sites and the resulting ultrasensitivity provide a mechanism to tightly link Sic1 turnover to the level of Cln/Cdc28 activity. Other phosphorylation-driven ubiquitination reactions may also be ultrasensitive.

Core components of the SCF and APC/C are structurally related

There are notable structural similarities between SCF and APC/C subunits (Fig. 2A). The identification of Apc2 revealed that it is related to Cdc53 in that they both share a 180-residue domain referred to as the cullin homology domain (CHD; Zachariae et al. 1998b; Wirbelauer et al. 2000). This domain is found in six closely related cullin family members in multicellular eukaryotes, in Apc2, and in several other more distantly related proteins. Because Cul1 had been implicated in binding to the E2 Cdc34 (Feldman et al. 1997; Skowyra et al. 1997), this suggested that Apc2 might also recruit an E2.

A more complete understanding of the relationship between SCF complexes and the APC/C came with the identification of the RING-H2 protein Rbx1/Roc1/Hrt1 as a subunit of the SCFs in both yeast and human cells (Kamura et al. 1999; Ohta et al. 1999; Seol et al. 1999; Skowyra et al. 1999). When the RING-H2 protein Apc11 was initially identified as a component of the APC/C, it was shown to be required for APC/C activity (Zachariae et al. 1998b), but its role in ubiquitination was unknown. The identification of a highly related protein in the SCF immediately suggested that the APC/C and SCF shared mechanistic similarities. Like Apc11, Rbx1 is an essential gene in budding yeast, and extracts from cells containing temperature-sensitive mutations in Rbx1 display defects in ubiquitination of Cln1 and Sic1. Importantly, Rbx1 interacts directly with Cdc53, and this interaction strongly stimulates the association of Cdc34 with Cdc53. Moreover, Rbx1 interacts with Cdc34 by itself, although this binding is stimulated by Cdc53 (Seol et al. 1999; Skowyra et al. 1999). The consequence of this interaction is the activation of ubiquitin-conjugating enzyme activity. Cdc34 displays weak autoubiquitination activity in vitro, but addition of Cdc53/Rbx1 complexes stimulates this activity (Seol et al. 1999; Skowyra et al. 1999). Similarly, human Rbx1/Roc1 can greatly stimulate formation of ubiquitin chains in vitro (Ohta et al. 1999). These findings suggested that RING-H2 domains such as those contained in Rbx1 function to both recruit and activate E2s. Subsequent studies have shown that Apc2 and Apc11 form a ubiquitin ligase core analogous to the Cul1/Rbx1 module (Fig. 2A) and that this complex has the ability both to bind E2s and to stimulate nonspecific ubiquitination activity (Gmachl et al. 2000; Leverson et al. 2000; Tang et al. 2001a). Mutation of the conserved ring-finger domain within Apc11 abolishes this activity.

The cullin/RING-H2 ubiquitin ligase core

Structural studies of SCF complexes and another RING-H2 ubiquitin ligase, Cbl, have helped define how the cullin/RING-H2 modules function in ubiquitination reactions (Zheng et al. 2000, 2002). Cul1 is a highly elongated protein composed of an N-terminal domain of three repetitive -helical bundles. The extreme N-terminal bundle interacts with Skp1. The C-terminal region contains the CHD and additional sequences that interact with Rbx1 to form a globular structure composed of a 5-strand -sheet and three independently folded helical bundles (Fig. 2B; Zheng et al. 2002). The previously defined cullin homology domain is composed of a 4-helix bundle linked to an /-domain in which the second strand of the -sheet is provided by an extended N terminus of Rbx1 (residues 20-35). This -sheet interaction constitutes all of the interactions between Rbx1 and the previously defined CHD and explains the sequence conservation in this domain. One unexpected outcome of this structural analysis is the expansion of the CHD. The C-terminal 180 residues of the CHD contain two winged-helix (WH) domains. The second WH domain (WH-B) is highly conserved in cullin family members; it forms a cradle that interacts with the RING-H2 domain of Rbx1 and also contains the conserved lysine residue that is the site of modification by Nedd8 (Fig. 2B), a small 76-residue protein with sequence similarity to ubiquitin. Nedd8's molecular function is unknown, and members of the cullin family of proteins are the only known recipients of Nedd8 conjugation (Tanaka et al. 1998), although more are likely to be found in the future. Linkage of Nedd8 to Cul1 is required for full catalytic activity of SCF complexes (for review, see Deshaies 1999), and the juxtaposition of the Nedd8 modification site with the site of Rbx1 binding is presumably of functional importance. Apc2 is not neddylated and displays little sequence identity with cullins in this C-terminal region. Nevertheless, crystallographic analysis has revealed that the C terminus of Apc2 also forms a WH domain very similar to that found in Cul1 (Zheng et al. 2002). Overall, these results indicate that the boundaries previously used to define the cullin homology domain should be expanded to include the extreme C termini of cullin and Apc2 family members. Based on the Rbx1/Cul1 structure, it is expected that Apc11 will interact with the C-terminal WH domain of Apc2, and one could speculate that other APC/C subunits might play a role analogous to that played by Nedd8 in cullin complexes.

A major question concerns how E2s are recruited to and recognized by the APC/C. Previous studies have shown that the APC/C can use multiple E2 family members, including human UbcH10 and its close homologs UbcX and E2-C (Aristarkhov et al. 1996; Yu et al. 1996; Osaka et al. 1997; Townsley et al. 1997), as well as Ubc4 from yeast (Charles et al. 1998). However, the APC/C is apparently unable to use Ubc3/Cdc34, a major SCF E2. Structural information relevant to E2 selection by RING-H2-based E3s has come from the analysis of Cbl, a ring-finger protein involved in receptor tyrosine kinase turnover, in complexes with UbcH7 (Zheng et al. 2000). Highly conserved Phe and Pro residues in UbcH7 interact with a hydrophobic surface in the Cbl ring-finger domain, which is formed primarily by the canonical RING-H2 domain. Key residues in Cbl involved in interaction with UbcH7 (Trp 87, Pro 95, Leu 96) are all conserved in Rbx1, suggesting that Rbx1 will interact in a similar way with Ubc3/Cdc34. Consistent with the involvement of these residues, mutation of Trp 87 in Rbx1 leads to a nonfunctional protein (Zheng et al. 2002). Interestingly, Apc11 and Rbx1 are quite similar in this region (Fig. 2C). Although structural studies are necessary to determine precisely how E2s interact with Rbx1/Cul1 and Apc11/Apc2 complexes, the available data suggest a generally conserved mechanism for interactions between E2s and distantly related RING-H2 domains. However, it is unclear at present what dictates the specificity of interaction of E2s with various RING-H2 proteins. In addition, Ubc4 interacts with Apc11 and most likely the Apc11/Apc2 complex, whereas UbcH10 does not require coexpressed Apc11 to interact with Apc2 (Tang et al. 2001a). Precisely how the Apc2/UbcH10 interaction occurs and whether it contributes to ubiquitination is unclear.

Structural analysis of the APC/C

Although we are now beginning to understand how the cullin-RING-H2 modules of the APC/C function in ubiquitination, we have little mechanistic information concerning other APC/C subunits. In principle, these subunits could be involved in binding to the APC/C activators Cdc20 and Cdh1 or in localizing the APC/C to particular structures in the cell. Alternatively, some of these subunits, such as Cdc16, Cdc23, and Cdc27, which are phosphorylated (see below), could be used as recipients of regulatory information. A complete understanding of APC/C function will require structural analysis, but this goal will be a challenge because of the APC/C's complex composition. Recent work has led to a low-resolution (24-Å) structure of human APC/C determined by cryo-electron microscopy and image reconstruction (Gieffers et al. 2001). The structure is dominated by an outer protein wall containing an inner channel that has been hypothesized to be the reaction chamber. It is unclear where critical subunits are in this structure, and additional work is needed to establish the quaternary organization of the individual subunits. One approach will involve structure determination of APC/C subcomplexes and individual subunits, which can then be modeled on low-resolution structures. It has been known for some time that loss of particular subunits such as the nonessential subunits Cdc26 and Apc9 leads to rearrangement of the APC/C (Zachariae et al. 1998b). For example, Apc2 immune complexes from cells lacking Cdc26 contain greatly reduced levels of Cdc16, Cdc27, and Apc9, suggesting that these subunits form a subcomplex.

The only APC/C subunit to be examined crystallographically thus far is Doc1/Apc10, which is the founding member of a family of proteins that contain a Doc1 domain. Doc1 domains are not limited to the APC/C. In fact, several Doc1-domain-containing proteins have been identified that also contain other domains implicated in ubiquitination, including the CHD, suggesting that Doc1 domains play a general role in ubiquitination reactions (Grossberger et al. 1999). The Doc1 domain from both yeast (Au et al. 2002) and humans (Wendt et al. 2001) forms a twisted -sandwich jelly-roll structure. This structure is quite similar to the fold found in several proteins involved in biomolecular interactions, including galactose oxidase, sialidase, and XRCC1, although this similarity is not evident at the level of primary sequence. Interestingly, conserved residues in Doc1 domains cluster on a single surface of the protein that is analogous to the surface used by sialidase to bind its ligand. Moreover, this patch also contains Ser 148, which is the site of a temperature-sensitive mutation in yeast Doc1 (Hwang and Murray 1997). Taken together, these data suggest that Doc1 may interact with APC/C components or substrates of the reaction through this conserved domain. Finding proteins that interact with Doc1 may help uncover the function of this subunit and may also aid in understanding the role of Doc1 domains in other classes of E3.

APC/C activators: Cdc20 and Cdh1

Most of the substrate selectivity of the APC/C resides in the so-called activator proteins Cdc20 (also called p55^CDC and fizzy, fzy) and Cdh1 (also called Hct1 and fizzy-related, fzr). The initial characterization of cell cycle mutants in budding yeast showed that cells mutant for CDC20 had a similar arrest phenotype as cells mutant for what subsequently became known as genes for APC/C subunits such as CDC23, CDC26, and CDC27. The first direct connection to cell cycle proteolysis came from Drosophila, where fizzy mutants failed to degrade mitotic cyclins (Dawson et al. 1993; Sigrist et al. 1995). Cdc20 is required for APC/C function in mitosis (see below). Cdh1/Hct1 (Cdc20 homolog/Homolog of Cdc twenty) was first identified in budding yeast and found to be necessary for the activity of the APC/C in late mitosis and in G₁ (Schwab et al. 1997; Visintin et al. 1997). Similarly, Drosophila fzr was found to be required for maintaining low mitotic cyclin levels in G₁ (Sigrist and Lehner 1997). Cdc20 and Cdh1 are members of a multigene family. For example, a meiosis-specific form, called Ama1, has been found in Saccharomyces cerevisiae (Cooper et al. 2000), and multiple Cdh1 homologs have been found in chickens (Wan and Kirschner 2001; see below). As described below, the use of multiple APC/C-activating proteins allows flexibility in APC/C function.

Complementing the genetic evidence that Cdc20 and Cdh1 are required for APC/C function in vivo, biochemical experiments showed that they were necessary for APC/C activity in vitro (Fang et al. 1998a; E.R. Kramer et al. 1998; Jaspersen et al. 1999). Added Cdh1 was necessary for the activity of the APC/C immunoprecipitated from yeast cell extracts (Jaspersen et al. 1999). Similarly, added Cdh1 and Cdc20 were each able to activate immunopurified Xenopus and human APC/C (Fang et al. 1998a; E.R. Kramer et al. 1998). Interestingly, addition of Cdc20 and Cdh1 conferred distinct substrate selectivities on the APC/C. This specificity is reflected in the general tendency of Cdc20 to target the APC/C to substrates containing a degradation signal called the destruction box and for Cdh1 to target the APC/C to substrates containing a distinct signal termed the KEN box (see below). Although the simplest explanation for the distinct substrate specificities of APC^Cdc20 and APC^Cdh1 would be that Cdc20 and Cdh1 recruited substrates to the APC/C, there was no evidence for this until recently. Hence, for some years Cdc20 and Cdh1 were known simply as "APC/C activators" to indicate simultaneously their requirement for APC/C function and our ignorance about their biochemical roles. Because they bind to the APC/C and confer substrate specificity upon it, Cdc20 and Cdh1 play the role that F-box proteins play in SCF complexes. However, Cdc20 and Cdh1 only associate with the APC/C transiently, essentially making them alternative substoichiometric APC/C subunits. How these two proteins coordinate distinct cell cycle transitions is discussed below.

	Integration of APC/C activity with cell cycle transitions

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

Research during the last decade has revealed that the chromosome cycle is inextricably linked with the Cdk cycle. There are two major states of Cdk activity during the cell cycle; a state where Cdk activity is high and a state where Cdk activity is low (Fig. 3A; Amon 1997; Irniger and Nasmyth 1997). This periodicity underlies temporal control of DNA replication and links it to the process of mitosis. Initiation of DNA replication requires the integration of two central processes: (1) formation of a prereplication complex and (2) activation of DNA-unwinding and polymerase functions. The former can occur only when Cdk activity is low, whereas the latter is promoted by high Cdk activity (Dahmann et al. 1995; Diffley 1996, 2001; Piatti et al. 1996; Noton and Diffley 2000). Thus, the switch from low Cdk activity to high Cdk activity is critical to proper control of DNA replication. Moreover, because Cdks are inhibitory to the formation of prereplication complexes, reinitiation cannot occur until cells reduce their Cdk activity by passing through mitosis, thereby ensuring the temporal order of S- and M-phases.

View larger version (34K): [in this window] [in a new window]   Figure 3.   Temporal control and substrates of the APC/C. (A) Periodic activities of APCCdc20, cyclin B/Cdc2, and APC/CCdh1. Cdk kinase activity is high in S phase and persists to the metaphase-anaphase transition. APCCdc20 is activated as cells approach metaphase, and this activity is enforced by high Cdc28 activity. APCCdh1 is activated during exit from mitosis and persists until the G1/S transition. (B) Cell cycle targets of the APC/C and the SCF in budding yeast and human cells.

Efforts to understand how this periodicity in Cdk activity is achieved have led to a greater understanding of both the positive and negative roles played by the APC/C in cell cycle control. The APC/C is required to reduce B-type cyclin levels as cells pass through anaphase and telophase, but the APC/C also restrains the accumulation of B-type cyclins during G₁, and its inactivation is required for timely S-phase entry (Irniger and Nasmyth 1997). Moreover, the APC/C is required to coordinate chromosome separation at the metaphase-to-anaphase transition in a process that requires Cdk activity.

The necessity that the APC/C be active under conditions of both high and low Cdk activity is reflected in the use of the two activators, Cdc20 and Cdh1, which display differential sensitivity to Cdk activity. APC^Cdc20 functions in the presence of high Cdk activity, and, indeed, this form of the APC/C appears to require Cdk function for activation (but see below for caveats to this generalization). In contrast, Cdh1 is directly inhibited by Cdks (Zachariae et al. 1998a; Jaspersen et al. 1999; Sorensen et al. 2000, 2001). Thus, the ability of the APC/C to shuffle through states of activation by Cdc20 and Cdh1 is central to its ability to control Cdk activity through cyclin degradation and the degradation of other substrates. Layered on this control is the activity of G₁ cyclin/Cdk complexes. Unlike B-type cyclins, G₁ cyclins are immune to the action of the APC/C and therefore can accumulate when APC^Cdh1 activity is high. This property is important because Cln/Cdc28 activity in budding yeast is required to activate degradation of the B-type cyclin/Cdk inhibitor Sic1 and initiate DNA synthesis (Schwob et al. 1994; Tyers 1996).

Switching off Clb5-dependent Cdk activity and destroying the mitotic inhibitor Pds1 are the sole essential functions of Cdc20 in budding yeast

APC^Cdc20 plays two essential roles in the early stages of mitosis: (1) degradation of a regulator of sister-chromatid cohesion and (2) degradation of Clb5, an S-phase cyclin that potently antagonizes APC^Cdh1 activity. APC^Cdh1 is, in turn, required to destroy the bulk of B-type cyclins to allow exit from mitosis (Fig. 3B). The defining event in mitosis is separation of sister chromatids (for review, see Nasmyth 1999; Nasmyth et al. 2000). The bonds that hold sister chromatids together are set up during DNA synthesis through the process of cohesion and are maintained until cells undergo the metaphase-anaphase transition (Fig. 4; Uhlmann and Nasmyth 1998; Tomonaga et al. 2000). The temporal linkage of cohesion with DNA replication provides a means by which to ensure that sister chromosomes are tethered until metaphase and explains why cohesion cannot be established during G₂. Sister-chromatid cohesion occurs via a multiprotein complex containing Scc1/Mcd1/Rad21, Scc3, Smc1, and Smc3 (Guacci et al. 1997; Michaelis et al. 1997; Furuya et al. 1998; Ciosk et al. 2000). In yeast, cohesion is maintained along the length of the chromosome until metaphase (Tanaka et al. 1999), but in mammalian cells, loss of cohesion is more complex. In an initial step, most cohesin complexes are released from chromosome arms, with a small fraction remaining bound in centromeric regions (Fig. 4; Waizenegger et al. 2000). This initial loss of arm cohesion depends on the Polo protein kinase (Sumara et al. 2002). Complete loss of cohesion occurs upon cleavage of Scc1 by an endoprotease called separase (Esp1 in budding yeast and Cut1 in fission yeast; Uhlmann et al. 1999, 2000; Yanagida 2000; Hauf et al. 2001). Regulation of separase is critical. Inappropriate separase activity can lead to precocious dissociation of sister chromatids. The timing of separase activity is controlled through the activity of a protein generically referred to as securin (Pds1 in budding yeast, Cut2 in fission yeast, and PTTG in mammals). Securin plays a negative role in mitosis by binding to separase and inhibiting its protease activity. Thus, activation of anaphase depends on proper degradation of securin by APC^Cdc20 (Cohen-Fix et al. 1996; Funabiki et al. 1996; Yamamoto et al. 1996; Visintin et al. 1997; Lim et al. 1998; Schott and Hoyt 1998; Tinker-Kulberg and Morgan 1999; Zou et al. 1999; Salah and Nasmyth 2000). Degradation of Pds1 liberates active separase, which then cleaves Scc1 to promote loss of cohesion (Fig. 4; Ciosk et al. 1998; Uhlmann et al. 1999, 2000).

View larger version (30K): [in this window] [in a new window]   Figure 4.   The role of the APC/C in loss of sister-chromatid cohesion at the metaphase-to-anaphase transition. Cohesion is established during S phase and is lost at the metaphase-anaphase transition through the action of a protease, separase, that cuts Scc1, thereby breaking the bonds that hold sister chromatids together. Separase is initially cytoplasmic and, in budding yeast, is localized to the nucleus in a pathway that is dependent on Cdc28-mediated phosphorylation of Pds1. When conditions for anaphase are met, securin is ubiquitinated by APCCdc20, and separase is released in an active form to cleave Scc1. In budding yeast, Scc1 cleavage is facilitated by Cdc5-mediated phosphorylation (Alexandru et al. 2001). Separase is also required to activate transient release of Cdc14 from the nucleolus through the Cdc-FEAR (fourteen early anaphase release) pathway. In human cells, loss of cohesion occurs through a polo-kinase (Plk1)-mediated step without cleavage of Scc1 in prophase, followed by a separase-dependent step at the metaphase-anaphase transition.

Cdks play a major role in regulating sister-chromatid separation in yeast (as well as other organisms). First, Cdks phosphorylate core subunits of the APC/C (Cdc16, Cdc23, and Cdc27), which promotes APC^Cdc20 activation (see below). Second, Cdks play a critical role in a positive function of securin (Fig. 4; Agarwal and Cohen-Fix 2002). In budding yeast, accumulation of Esp1 in the nucleus depends on the presence of securin/Pds1 (Jensen et al. 2001). Similarly, in S. pombe, localization of separase/Cut1 on spindle poles requires securin/Cut2 (Kumada et al. 1998). Recent studies (Agarwal and Cohen-Fix 2002) have revealed that Cdc28 directly phosphorylates Pds1, thereby enhancing its ability to bind to Esp1 and promoting Esp1 nuclear localization. However, Cdk-mediated Pds1 phosphorylation does not appear to be required for Pds1 to function as a mitotic inhibitor as cells expressing Pds1 that lack critical Cdk phosphorylation sites are resistant to the microtubule-depolymerizing drug benomyl, unlike pds1 cells. Third, in vertebrate cells, separase is phosphorylated, and this appears to negatively regulate its ability to induce sister-chromatid separation (Stemmann et al. 2001). The kinase responsible has not yet been identified, but Cdc2 is capable of inducing a blockade to sister-chromatid separation even though securin is destroyed. Thus, phosphorylation of separase may provide an additional mechanism for controlling the timing of loss of cohesion.

Once cells have initiated anaphase, they begin to switch off mitotic cyclin activity (Fig. 3A). This process occurs in two phases. First, APC^Cdc20 eliminates the bulk of cyclins present during mitosis, including Clb3 and Clb5 (Shirayama et al. 1999; Baumer et al. 2000; Yeong et al. 2000). Clb5 degradation is particularly important, and, in fact, Pds1 and Clb5 are the only essential substrates of APC^Cdc20 in budding yeast as deletion of CLB5 and PDS1 suppresses the lethality of cdc20 cells (Shirayama et al. 1999). As described in detail below, activation of APC^Cdh1 requires collaboration between a protein phosphatase Cdc14 and the Cdk inhibitor Sic1, which extinguish Cdh1 phosphorylation and promote assembly of an active APC^Cdh1 complex (Visintin et al. 1998; Jaspersen et al. 1999). One possible interpretation of the data is that, among B-type cyclin complexes, Clb5/Cdk is most active toward Sic1 and Cdh1 during this phase of the cell cycle, and this activity cannot normally be overcome by Cdc14. In the absence of Clb5, normal levels of Cdc14 are sufficient to maintain Sic1 and Cdh1 in their unphosphorylated forms despite the presence of other B-type cyclin/Cdk complexes. The second phase of the switch from high to low Bcyclin/Cdk activity comes with the full activation of APC^Cdh1, which is capable of efficiently destroying Clb2 (Schwab et al. 1997; Visintin et al. 1997). Complete degradation of Clb2 is required for exit from mitosis. Integrating degradation of Pds1 and Clb5 with activation of APC^Cdh1, together with the fact that degradation of Pds1 is required for release of Cdc14 from the nucleolus (Shirayama et al. 1999; Tinker-Kulberg and Morgan 1999), provide an effective mechanism for ensuring that cells exit mitosis only after they have separated their chromosomes.

Spatial control of cyclin degradation

A major question in the ubiquitination field concerns the spatial and temporal control of ubiquitination, which is largely an unexplored area. In many cases, substrates are thought to be destroyed essentially in an all-or-none fashion. However, analysis of B-type cyclin degradation indicates that temporal control of degradation may go hand-in-hand with spatial control. In budding yeast, Clb2 degradation occurs in two waves, a Cdc20-dependent pathway that destroys a substantial fraction of Clb2 during anaphase and a second wave of degradation that occurs via APC^Cdh1 during mitotic exit (Lim et al. 1998; Baumer et al. 2000; Yeong et al. 2000). The basis of this differential selectivity is unknown but could potentially reflect localized activation of Cdh1 during mitosis. Perhaps the best example of spatial control in ubiquitination by the APC/C comes from an analysis of human cyclin B-GFP degradation in real time in vivo (Clute and Pines 1999). Cyclin B is localized to chromosomes and spindle poles during prophase and is also diffusely localized in the nucleus. Once the last chromosome is aligned at the metaphase plate, cyclin B-GFP is immediately eliminated from the spindle poles and chromosomes, but most of the cyclin B-GFP remains, only to be eliminated as cells proceed into anaphase (Clute and Pines 1999). Using immunofluorescence, it has also been found that Drosophila cyclin B is lost from spindle poles more rapidly than from other nuclear structures (Huang and Raff 1999). What is not clear is to what extent spatial degradation is controlled by the activity of the APC/C or by some event that renders the target susceptible to ubiquitination. In mammalian cells, some components of the APC/C, including APC1, CDC27, and DOC1, are located on centrosomes (King et al. 1995; Jorgensen et al. 1998), but this localization is insufficient to provide a mechanism for this level of control. It will be important to determine to what extent spatial control of cyclin degradation is linked with specific biological activities of cyclins. On a general note, analysis of spatial distribution of proteolysis is complicated by the fact that proteins can also undergo non-ubiquitin-dependent alterations in localization. The development of methods that identify substrate-E3 interactions in living cells may be required to explore these important questions.

Linking the APC/C with degradation of diverse proteins

Although securin and mitotic cyclins are perhaps the best understood APC/C substrates, a large number of other proteins display cell cycle-regulated degradation via the APC/C (Fig. 3B). Many of these targets have been identified through the use of temperature-sensitive mutations in budding yeast APC/C subunits. These include multiple proteins linked with spindle function, such as the anaphase spindle-elongation-control protein Ase1, the kinesin-related motor protein Cin8, and Cdc20 itself. In budding yeast, both Ase1 and Cin8 degradation are controlled by APC^Cdh1 (Juang et al. 1997; Hildebrandt and Hoyt 2001). In contrast, degradation of Kip1, a motor protein that appears to function redundantly with Cin8, is Cdc20-dependent (Gordon and Roof 2001). Although these proteins are clearly APC/C substrates and are degraded in a cell cycle-dependent manner, cells expressing nondegradable forms tend to have mild phenotypes, indicating that their degradation is not essential for cell proliferation. For example, although expression of endogenous levels of nondegradable Cin8 causes an increase in the fraction of cells without spindles, indicating a defect in assembly or maintenance of the mitotic spindle, the cells are viable. A balance between spindle motor forces is important for ensuring proper spindle formation. Because overexpression of Cin8 can disturb spindle function and cause premature elongation of spindles, it seems likely that degradation of Cin8 during G₁ provides a mechanism for resetting the motor activity to low levels prior to assembly of the mitotic spindle later in the cell cycle (Hildebrandt and Hoyt 2001). This would predict that Kar3 (Hoyt and Geiser 1996), the motor protein that functions in opposition to Cin8, might also display cell cycle-regulated abundance.

Aurora-A kinase family members are localized to spindle-pole bodies and have been implicated in centrosome duplication and separation and in spindle assembly. In Xenopus and mammalian tissue culture cells, aurora-A accumulates in G₂/M and is absent in G₁. Recent work has shown that aurora-A is ubiquitinated by APC^Cdh1 (Castro et al. 2002; Taguchi et al. 2002). The budding yeast homolog of aurora is Ipl1. Although the levels of Ipl1 also vary in the cell cycle, it has not yet been shown that Ipl1 is subject to regulation by the APC/C. Aurora family members have been linked with cancer in humans, and cancers frequently display elevated aurora levels, in some cases owing to increased gene dosage. High levels of aurora do not alter centrosome duplication directly but induce cells to undergo an aberrant mitosis without cytokinesis, producing tetraploid cells in G₁ (Meraldi et al. 2002). In the absence of p53, these cells pass through S phase and enter into an aberrant mitosis because of the presence of extra centrosomes, giving rise to aneuploid cells. Thus, the ability of APC^Cdh1 to keep aurora levels in check would appear to be important in maintaining ploidy.

The APC/C has also been linked to the control of DNA synthesis. In metazoans, Cdc6 and Cdt1 collaborate to establish prereplication complexes at the G₁/S transition. Cdc6 and Cdt1 assemble at sites of replication initiation and recruit MCM proteins onto chromatin to establish active replication complexes (for review, see Lygerou and Nurse 2000; Diffley and Labib 2002). Both Cdc6 and Cdt1 are regulated through control of their localization, and Cdc6 is additionally regulated by its abundance. Human Cdc6 protein levels are low in G₁, in part because of the action of APC^Cdh1 (Petersen et al. 2000). Cdc6 is ubiquitinated by APC^Cdh1 in vitro, and Cdh1 is limiting for its accumulation in tissue culture cells. The ability of Cdc6 to accumulate and assemble into replication complexes will likely depend on inactivation of Cdh1 at the G₁/S transition (see below). The critical role played by Cdt1 in MCM loading dictates that its activity be tightly controlled after initiation. This is accomplished through the action of geminin, a small protein that binds to Cdt1 and blocks its activity (Wohlschlegel et al. 2000). Geminin is cell cycle-regulated; it is present during S and G₂ phases, but is destroyed as cells proceed through the metaphase-anaphase transition (McGarry and Kirschner 1998; Nishitani et al. 2001). Maintaining low geminin levels during G₁ allows for loading of Cdt1 onto origins. Geminin was initially identified in a screen for substrates of the Xenopus APC/C (McGarry and Kirschner 1998). Although the form or forms of the APC/C that are capable of ubiquitinating geminin have not been determined, the pattern of geminin expression through the cell cycle suggests that Cdh1 is responsible for its turnover.

The APC/C has also been implicated in the turnover of additional components of the budding yeast replication pathway (Fig. 3B). The protein kinase Cdc7 is required for initiation of DNA synthesis and is regulated by Dbf4, whose protein levels are cell cycle-regulated. Once initiation is complete, cells inactivate Cdc7 by destroying Dbf4 via APC^Cdc20-catalyzed ubiquitination (Cheng et al. 1999; Oshiro et al. 1999; Ferreira et al. 2000). Mutations in Dbf4 that block efficient ubiquitination do not induce cell cycle arrest; Dbf4 degradation is therefore not essential for cell division. Mammalian Dbf4 is also cell cycle-regulated and may use the APC/C to control its stability. In contrast to Dbf4, the levels of Cdc6 during S and G₂ phases in budding yeast are controlled by SCF^Cdc4 in conjunction with Cdk-mediated phosphorylation (Drury et al. 1997, 2000; Perkins et al. 2001).

Vertebrate Cdh1 and G₁/S control

Although the role of Cdh1 in G₁ control is well characterized in yeast, significantly less is known about Cdh1 function in animal cells. However, recent work has shown that Cdh1 is not required for cell viability and proliferation in DT40 chicken cells (Sudo et al. 2001). As might have been expected, these cells displayed increased levels of mitotic cyclins in G₁. Thus, these cells are refractory to inhibition by rapamycin, which induces the Cdk inhibitor p27. A major complication with interpretation of this work is that the chicken genome has at least three other Cdh1 homologs (Wan and Kirschner 2001; see below), and it is possible that one or more of these is partially redundant with Cdh1 for important mitotic functions in these cells.

	Substrate recognition

Top Introduction Cell cycle transitions:... Cycling into destruction:... APC/C composition and... Integration of APC/C activity... Substrate recognition APC/C regulation Spindle assembly checkpoint Roles for the APC/C... Conclusion References

Destruction boxes and KEN boxes

The first motif found to be necessary for the degradation of an APC/C substrate was the destruction box. When Glotzer et al. (1991) first showed that mitotic cyclins were degraded via the ubiquitin system, they also showed that an N-terminal 91-amino-acid fragment was sufficient for ubiquitination and degradation, and that it could cause the degradation of an unrelated protein when the two were fused. Inspection of the N termini of the known cyclins revealed a conserved short motif that might be important either for substrate recognition or for ubiquitination. This motif was termed the destruction box. Mutagenic analysis has identified the degenerate core destruction-box motif to be RxxLxxxxN (Glotzer et al. 1991; King et al. 1996), although there are some clear differences in the destruction boxes of A- and B-type cyclins (Glotzer et al. 1991; King et al. 1996; Klotzbucher et al. 1996; Geley et al. 2001) and some flexibility at the last position. Of course, such a small motif occurs frequently and does not, by itself, contain sufficient information to confer APC/C-dependent ubiquitination on a protein. Little is understood about other elements located near the destruction box that are required for its proper recognition. Clearly, such elements may be highly degenerate and only recognizable at the structural level. Even when a small piece of an APC/C substrate (such as a 27-amino-acid fragment of a mitotic cyclin; King et al. 1996) can confer degradation on an unrelated protein, one must always be cautious in assuming that all of the recognition elements are contained in the transferred sequence. Degenerate elements, not just including the ubiquitinated lysines themselves, may be present in the fusion partner. Destruction boxes are widespread and have been found in the majority of APC/C substrates.

A second degradation motif, the KEN box, was only identified in 2000 (Pfleger and Kirschner 2000). It was suspected that human Cdc20 might contain a novel motif because it lacked an identifiable destruction box yet was a substrate for APC^Cdh1 in Xenopus egg extracts. Deletion analysis narrowed the motif to the first ~100 amino acids of the protein, and alanine-scanning mutagenesis identified four amino acids as essential: KENxxxN. Given that other Cdc20 homologs contain aspartic acid in place of the second asparagine, it was proposed that the minimal KEN box consists of KENxxx(N/D), although it is plausible that glutamic acid and possibly other amino acids may also be allowed in the terminal position. Indeed, the budding yeast Hsl1 protein contains a functional KEN box consisting of KENxxxE (Burton and Solomon 2001), the KEN box of the yeast Clb2 protein is KENxxxS (Hendrickson et al. 2001), and the KEN box of human securin is KENxxxG (Zur and Brandeis 2001). Few KEN boxes have been examined in any detail, so this consensus may evolve, particularly if suboptimal KEN boxes retain significant function. Like the destruction box, the KEN box is portable, and a 28-amino-acid section of human Cdc20 containing the KEN box could destabilize an unrelated protein, though not as well as larger pieces of Cdc20, suggesting that there may be additional determinants for recognition. So far, only a modest number of APC/C substrates have been found to have KEN boxes, although this number will grow as more substrates are examined for their presence and as we fine-tune our understanding of what sequences constitute functional motifs. Some substrates require both a destruction box and a KEN box (Petersen et al. 2000; Burton and Solomon 2001; Hendrickson et al. 2001; Jacobs et al. 2001; Zur and Brandeis 2001). In such cases, initial identification of a necessary destruction box may delay recognition of an essential KEN box.

The existence of well-characterized Destruction boxes and KEN boxes does not preclude the possibility of additional degradation motifs. For instance, a new motif, the A box, has just been found to be essential for the APC^Cdh1-mediated degradation of Xenopus Aurora-A (Littlepage and Ruderman 2002).

Direct binding to Cdc20 and Cdh1

When Cdc20 and Cdh1 were found to confer substrate specificity on the APC/C, it was generally assumed that they would bind substrates directly and recruit them to the APC/C. This was certainly the simplest explanation. However, through years of study, there was no evidence that Cdc20 or Cdh1 bound substrates directly, necessitating that other models be considered. For instance, the distinct but overlapping specificities that Cdc20 and Cdh1 conferred on the APC/C suggested that each activator might cause a different substrate-binding site on the APC/C to open. Given the inelegance of this model, it was a relief when four reports appeared within a two-week span last year showing various aspects of the direct recognition of substrates by Cdc20 and Cdh1 in diverse systems.

Schwab et al. (2001) showed by coimmunoprecipitation that budding yeast Cdc20 and Cdh1 were associated with substrates. The specificity of the associations paralleled the degradation requirements. Thus, Cdc20, but not Cdh1, bound to Pds1, which requires Cdc20 for its degradation, and Cdh1, but not Cdc20, bound to Clb2, Clb3, and Cdc5, all of which require Cdh1 for their degradation. Mutation of the destruction boxes in Clb2 or Cdc5 did not affect the association with Cdh1, suggesting either that there is another degradation motif, or that degradation motifs are not involved in this interaction. A KEN box has since been identified in Clb2 (Hendrickson et al. 2001), although whether it mediates the interaction with Cdh1 has not been tested. Hilioti et al. (2001) also examined the association of Pds1 with Cdc20. They found that the destruction box in Pds1 was necessary for this association and that Pds1 produced in Escherichia coli could bind to Cdc20 produced by in vitro translation, strongly suggesting that the interaction was direct. Importantly, they also showed that the spindle assembly checkpoint (see below) had no effect on substrate<