Maturation promoting factor MPF is a cyclin-CDK complex that catalyzes the phosphorylation

1. Wolfgang Zachariae and
2. Kim Nasmyth

1. Research Institute of Molecular Pathology, A-1030 Vienna, Austria

Cell proliferation depends on the duplication of chromosomes followed by the segregation of duplicates (sister chromatids) to opposite poles of the cell prior to cell division (cytokinesis). How cells ensure that chromosome duplication, chromosome segregation, and cell division occur in the correct order and form an immortal reproductive cycle is one of the most fundamental questions in cell biology. Without such coordination, cells would not maintain a constant chromosome number and sexual reproduction as we know and love it would not be possible.

A halfhearted cell cycle

The discovery of cyclin-dependent kinases (CDKs) went some way to answering this question. Successive waves of S- and M-phase-promoting CDKs first trigger chromosome duplication (S phase)—then the attachment of the replicated chromosomes to a bipolar spindle (M phase). In animal cells, S phase is induced by Cdk2 bound to S-phase cyclins (E- and A-type) whereas M phase is triggered by Cdk1 associated with mitotic cyclins (A- and B-type). In both fission yeast and budding yeast, S and M phase are induced by a single CDK (Cdk1) bound to S-phase- and M-phase-specific B-type cyclins, respectively. We now understand many of the regulatory mechanisms that activate S– and M–CDKs in the correct order. We also have a robust hypothesis for how cells ensure that no genomic sequence is duplicated more than once during the interval between the onset of S and M phases. Initiation of DNA replication requires two distinct steps: first, prereplicative complexes (pre-RCs) are assembled at future origins of replication, a process that can occur only in the absence of CDK activity. The second step, origin unwinding and the recruitment of replication enzymes, is triggered by CDK activation. Because pre-RC assembly is inhibited by CDK activity, chromosome rereplication requires a CDK cycle, a period of low CDK activity followed by a period of high CDK activity. Having activated S–CDKs in late G1, cells maintain high CDK activity until metaphase and this prevents refiring of replication origins.

However, several crucial elements were missing from this CDK-dominated view of the cell cycle. Missing was the impetus that causes sister chromatids to separate at the metaphase-to-anaphase transition; the machinery that destroys mitotic cyclins during anaphase; a mechanism for ensuring that sister chromatid separation normally precedes cytokinesis and chromosome reduplication; and an understanding of how events that trigger sister chromatid separation and exit from mitosis also create the conditions that cause the chromosome cycle to be repeated. Insight into all these questions has recently stemmed from the identification of the machinery responsible for degrading mitotic cyclins, a ubiquitin–protein ligase called the anaphase-promoting complex or cyclosome (APC/C). By destroying anaphase inhibitory proteins, the APC/C triggers the separation of sister chromatids; by destroying mitotic cyclins, it creates the low CDK state necessary for cytokinesis and for reforming the pre-RCs complexes needed for another round of genome replication. Since the discovery of the APC/C 4 years ago, there has been a veritable deluge of results on its roles and regulation, which this article attempts to summarize.

Cyclin degradation in vivo and in vitro

Much of our knowledge about cyclin degradation comes from experiments with eggs from the frog Xenopus leavis and from marine invertebrates such as sea urchins and the clam Spisula solidissima. Upon fertilization, these cells undergo a series of rapid and synchronous cell cycles that consist of alternating S and M phases. Mitotic cyclins A and B steadily accumulate during interphase (the time between two M phases) and are then suddenly degraded during mitosis (Evans et al. 1983). The development of cell-free extracts that reproduce many aspects of the cell cycle opened the way to a biochemical analysis of cyclin destruction (Luca and Ruderman 1989;Murray and Kirschner 1989). The amino terminus of cyclin B was found to be essential for degradation but dispensable for the formation of an active Cdk1–cyclin B kinase (Murray et al. 1989). An amino-terminal cyclin B fragment was sufficient to confer mitotic degradation to heterologous proteins, whereas a version of cyclin B lacking its amino terminus constitutively activated Cdk1, which prevented exit from mitosis. These data suggested that cyclin B degradation is essential for cell cycle progression. Cdk1 inactivation at the end of mitosis has subsequently been found to be essential for cell cycle progression in all eukaryotes. Degradation of mitotic cyclins appears to be a universal mechanism for Cdk1 inactivation even though other mechanisms also exist.

A closer inspection of the amino termini of mitotic cyclins revealed a degenerate nine amino acid motif called the destruction box, whose mutation renders cyclin B resistant to degradation (Glotzer et al. 1991). The observation that the destruction box was also required for the formation of cyclin–ubiquitin conjugates (Glotzer et al. 1991) and that cyclin degradation was sensitive to inhibitors of the ubiquitin system (Hershko et al. 1991) pointed to the protease responsible for cyclin degradation: the 26S proteasome, a multisubunit protease specific for multiubiquitinated substrates (Coux et al. 1996;Baumeister et al. 1998). In contrast to cyclin B ubiquitination, which occurs only in mitotic extracts, the proteasome is active throughout the cell cycle suggesting that cyclin ubiquitination rather than its degradation is cell cycle regulated (Mahaffey et al. 1993). Indeed, cyclin B–ubiquitin conjugates generated in mitotic extracts are efficiently degraded in interphase extracts (J.-M. Peters, pers. comm.).

Cyclin B was one of the first cellular substrates of physiological importance to be identified for the ubiquitin–proteasome pathway (see Fig. ). During this process, ubiquitin is first activated by forming a high-energy thioester with a cysteine in a ubiquitin-activating enzyme known as E1. Ubiquitin is subsequently transferred to one of several ubiquitin-conjugating enzymes (called UBC or E2) to form a second thioester. Finally, an isopeptide bond is formed between ubiquitin’s carboxyl terminus and a lysine residue of the substrate. More ubiquitin molecules are conjugated to those already attached to create a polyubiquitin chain that is recognized by the 26S proteasome, leading to proteolysis of the substrate. The transfer of ubiquitin from E2 enzymes to substrates requires a third activity, called ubiquitin–protein ligase or E3. By interacting with both the substrate and the E2 enzyme, ubiquitin–protein ligases are thought to provide specificity. The E3, and not the E2 enzyme largely determines which proteins are ubiquitinated and subsequently degraded. Cells usually contain a single, conserved E1 enzyme and a family of related E2 enzymes. E3 enzymes appear, however, to be structurally diverse, ranging from single proteins to large multisubunit complexes such as the APC/C.

Degradation of cyclin B by the ubiquitin–proteasome pathway. Ubiquitin (Ub, gray circles) forms thioester intermediates first with the ubiquitin-activating enzyme E1 and then with an ubiquitin-conjugating enzyme E2. Finally, it forms an isopeptide bond with a lysine of the substrate—in this case, cyclin B. Transfer of ubiquitin from the E2 to the substrate is catalyzed by the ubiquitin–protein ligase (E3) activity of the APC/C. Multiubiquitinated cyclin B is recognized by the 26S proteasome (dark gray) and then proteolyzed to yield peptides and free, reusable ubiquitin.

Cyclin A is degraded during metaphase and cyclin B slightly later at the metaphase-to-anaphase transition. The coincidence between cyclin degradation and entry into anaphase gave rise to the idea that cyclin degradation might be the signal that triggers sister chromatid separation. Expression of nondegradable cyclin variants does indeed block Cdk1 inactivation, spindle disassembly, and cytokinesis in many systems. However, sister chromatid separation is not inhibited (Holloway et al. 1993; Surana et al. 1993). Nevertheless, cyclin degradation and sister chromatid separation appeared to be connected. Inhibition of the cyclin degradation system by high concentrations of an amino-terminal cyclin B fragment also blocked sister chromatid separation in Xenopus egg extracts (Holloway et al. 1993). This observation led to the idea that sister chromatid separation might depend on destruction box-dependent degradation of a non-cyclin protein. Such anaphase inhibitors have indeed been identified in yeast and vertebrates (see below). However, recent experiments suggest that depletion of cellular ubiquitin contributes to the inhibitory effect of destruction box peptides (Yamano et al. 1998).

A needle worth the search: identification of the APC/C

The identity of the machinery responsible for cyclin B ubiquitination was revealed by a remarkable convergence of biochemical and genetic studies. Fractionation of extracts from clam oocytes andXenopus eggs suggested that three components might be sufficient for cyclin ubiquitination: an E1 enzyme commonly required for all ubiquitin transfer reactions; a cyclin-specific E2; and a cyclin-specific ligase activity (Hershko et al. 1994; King et al. 1995;Sudakin et al. 1995) (Fig. ). Unlike E1 and E2, the ligase activity was cell cycle regulated, being active when isolated from mitotic extracts but inactive when isolated from interphase extracts. Remarkably, the ligase activity sedimented as a large particle of 20S. Candidates for its components were suggested by a parallel study of cyclin B proteolysis in budding yeast: certain tetratricopeptide repeat (TPR) proteins known to be necessary for nuclear division were found to be essential for cyclin proteolysis in vivo (Irniger et al. 1995). Vertebrate homologs of two of these TPR proteins, Cdc16 and Cdc27, were detected in a 20S complex from human cells (Tugendreich et al. 1995) and were associated with the cyclin–ubiquitin ligase activity fromXenopus (King et al. 1995).

Key to isolating mutants specifically defective in B-type cyclin proteolysis was the recognition that this process is not confined to anaphase cells but continues during the subsequent G1 period until cells enter S phase (Amon et al. 1994). By looking for mutants defective in degrading a cyclin–β-galactosidase fusion protein in G1 cells, it was possible to exclude mutants that failed to degrade mitotic cyclins merely because they arrested in G2 or metaphase, during which mitotic cyclins are stable (Irniger et al. 1995). Temperature-sensitive mutations in five essential genes were identified that cause a defect in cyclin degradation in anaphase and G1. Furthermore, protein extracts from these mutants are defective in cyclin ubiquitination (Irniger et al. 1995; Zachariae and Nasmyth 1996; Zachariae et al. 1996). Whereas overexpression of a nondegradable mitotic cyclin arrests cells in late anaphase, the cyclin degradation mutants arrest earlier, in a metaphase-like state: sister chromatid separation, spindle elongation, Cdk1 inactivation, cytokinesis, and genome rereplication all fail to occur. Thus, all five genes are required for both anaphase onset and cyclin degradation. Furthermore, they all encode subunits of a 36S complex. The clam ubiquitin ligase particle was called the cyclosome, but when it became clear that the equivalent particle was required for the onset of anaphase in budding yeast, the yeast and Xenopus particles were called the anaphase-promoting complex or APC. Similar complexes have been found in fission yeast (Yamashita et al. 1996),Aspergillus nidulans (Lies et al. 1998), and in human cells (Tugendreich et al. 1995; Grossberger et al. 1999), suggesting that the APC/C is a conserved constituent of all eukaryotic cells.

Ubiquitin-conjugating enzymes of the APC/C pathway

Two classes of E2 enzymes are capable of collaborating with the APC/C. E2-C from clam and its homologs fromXenopus (UBCx), humans (UbcH10), and fission yeast (UbcP4) seem to function specifically in ubiquitination mediated by the APC/C (Aristarkhov et al. 1996; Yu et al. 1996; Osaka et al. 1997; Townsley et al. 1997). E2-C is required for the ubiquitination of cyclins but not the bulk of other unstable proteins. Furthermore, expression of a catalytically inactive version of UbcH10 blocks anaphase onset and cyclin degradation in vivo. At least in vitro, cyclin ubiquitination by the APC/C fromXenopus and budding yeast is also supported by Ubc4, which belongs to a different class of E2 enzymes that have been implicated in other degradation pathways (Yu et al. 1996; Charles et al. 1998). However, in budding yeast, the APC/C does not seem to require any particular E2 enzyme. Yeast mutants lacking Ubc4, its close relative Ubc5, or the E2-C homolog Ubc11 do not have any major defect in cyclin ubiquitination/degradation (Zachariae and Nasmyth 1996; Townsley and Ruderman 1998). It has been reported that B-type cyclins are stabilized in budding yeast ubc9 mutants (Seufert et al. 1995). However, the Ubc9 enzyme transfers the ubiquitin-like protein Smt3/Sumo1, not ubiquitin, and this modification does not induce proteolysis (Johnson and Blobel 1997;Schwarz et al. 1998). Ubc9 might support cyclin degradation only indirectly, as a result of its involvement in processes such as nuclear import (Lee et al. 1998).

The APC/C core particle: SCF’s big brother?

The composition of APC/C has been investigated by immunopurification. The Xenopus and human particles contain ≥10 subunits, whereas the yeast particle contains ≥12 subunits (Table ). Most, if not all, yeast subunits have counterparts in vertebrates, which suggests that the APC/C has a similar composition in all eukaryotes (Peters et al. 1996; Yu et al. 1998; Zachariae et al. 1998b; Grossberger et al. 1999). These subunits, the list of which might not yet be complete, remain tightly associated with each other throughout the cell cycle and therefore form the core of the particle (Peters et al. 1996;Grossberger et al. 1999). Additional regulatory subunits, such as the Trp–Asp repeat (WD) activator proteins Cdc20 and Cdh2 (see below), whose association with the APC/C core is both substoichiometric and cell cycle regulated, were initially not detected by biochemical studies.

Subunits of the APC/C in yeast and mammalian cells

It is possibly not surprising, considering the large number of APC/C subunits, to find that many subunits contain motifs thought to mediate protein–protein interactions. The largest subunit, Apc1, shares a motif of unknown function with Rpn1 and Rpn2, which are subunits of the 19S cap complex of the proteasome (Peters et al. 1996;Yamashita et al. 1996; Zachariae et al. 1996; Lupas et al. 1997). Cdc16, Cdc23, Cdc27, and Apc7, a Cdc27-like subunit found only in vertebrates, are all related proteins of 70–100 kD that contain 9–10 copies of the TPR motif (Lamb et al. 1994; Tugendreich et al. 1995; Yu et al. 1998). This degenerate 34-amino-acid motif is not, however, specific for the APC/C, as it is found in many other proteins with unrelated functions (Lamb et al. 1995). Each motif consists of a pair of antiparallel α-helices and multiple motifs are predicted to fold into a superhelical structure with a continuous helical groove suitable for protein–protein interactions (Das et al. 1998). The small APC/C subunit Cdc26 seems to be important for assembly of the complex. Budding yeast Cdc26 is a heat shock protein that stabilizes the interaction of three core subunits (Cdc16, Cdc27, Apc9) with the rest of the particle (Zachariae et al. 1996; Zachariae et al. 1998b). Indeed, overexpression of Cdc26’s fission yeast homolog, Hcn1, suppresses the assembly defect caused by acut9 (cdc16) mutation (Yamada et al. 1997).

Of yet greater interest are those subunits that contain motifs or domains found in other proteins implicated in ubiquitination. Doc1/Apc10, for example, is a 30-kD protein that contains a domain, called the Doc domain, found in several large mammalian proteins (Hwang and Murray 1997; Kominami et al. 1998; Grossberger et al. 1999). It is interesting that these proteins also contain either HECT (see below) or cullin domains, both of which are hallmarks of ubiquitin ligases. Nothing, however, is known about the function of these proteins.

Clearly, the most remarkable finding to emerge from these studies is that two of APC/C’s subunits, Apc2 and Apc11, are related to subunits from SCF (for Skp1–cullin–F-box protein), another ubiquitin ligase complex with an important role in cell cycle regulation (Feldman et al. 1997; Skowyra et al. 1997). This suggests that APC/C and SCF might be derived from the same ancestral ubiquitin protein ligase. Apc2 is a member of the cullin family of proteins, which includes the SCF subunit Cdc53 from budding yeast (Kramer et al. 1998b; Yu et al. 1998; Zachariae et al. 1998b). Cdc53’s carboxy-terminal ‘cullin’ domain recruits the E2 enzyme Cdc34 to the SCF (Patton et al. 1998), whereas its amino-terminal region binds to Skp1, which in turn binds to proteins that associate with substrates (Bai et al. 1996). Cdc53 is therefore thought to have a key role in bringing substrates into contact with their ubiquitin-conjugating enzyme. Apc2 might fulfill a similar role within the APC/C. However, neither Skp1 nor related proteins have been implicated in APC/C function and the mechanism by which substrates are recruited to the APC/C remains obscure. The small subunit Apc11 (Zachariae et al. 1998b) is closely related to Hrt1 (also called Rbx1 or Roc1), which has recently been identified as the fourth essential subunit of the SCF (Kamura et al. 1999; Seol et al. 1999; Skowyra et al. 1999; Tan et al. 1999). Human Hrt1 is also a subunit of the VBC complex (for VHL protein-Elongin B–Elongin C), which contains the von Hippel–Lindau tumor suppressor protein (pVHL) and is implicated in ubiquitin-dependent protein degradation (Kamura et al. 1999; Maxwell et al. 1999). Interestingly, one of VBC’s subunits was identified as the cullin CUL2 (Pause et al. 1997). Apc11 and Hrt1 share a RING–H2 finger domain that coordinates two zinc ions and is suspected to mediate protein–protein interactions (Borden and Freemont 1996). Hrt1 binds both to Cdc53’s carboxyl terminus and to Cdc34 and is thought to tether the E2 enzyme to SCF. Apc11 likewise interacts with the carboxy-terminal region of its cullin, Apc2, and might therefore play a similar role in the APC/C (Ohta et al. 1999). However, a stable interaction between APC/C and its E2 enzyme has so far not been detected. Indeed, such an interaction might be transient and only occur during the transfer of ubiquitin to substrate proteins. Taken together, these data suggest that APC/C, SCF, and VBC are members of a family of complex E3 enzymes that share a cullin and a RING finger subunit and might originate from the same ancestral ubiquitin protein ligase (Fig. A). It is remarkable that RING domains occur in several other proteins involved in ubiquitin-dependent proteolysis such as Ubr1, Hrd1, and Mdm2, which function in the degradation of N-end rule substrates, endoplasmic reticulum (ER) proteins, and p53, respectively. Many RING domain proteins with no apparent role in proteolysis might function in other processes regulated by ubiquitination, such as internalization of cell surface proteins (Hicke 1999).

Comparison of the APC/C with other ubiquitin–protein ligase complexes. (A) APC/C, SCF, and VBC belong to a family of E3 complexes containing a cullin (red) and a RING-finger (orange) subunit. Furthermore, these complexes use specialized subunits (yellow) for substrate recognition and recruitment. SCF collaborates with several different substrate recognition subunits; Cdc4 which binds to the substrate Sic1 is shown as one example. The von Hippel–Lindau protein (pVHL) is thought to play a similar role in the VBC complex by recruiting the hypoxia-inducible factor (HIF). pVHL binds to a Skp1-like protein called Elongin C via an F box-related domain called SOCS box. Arrows with question marks indicate suspected but not yet demonstated interactions. (B) Different reaction mechanisms of HECT domain ligases (top) and cullin–RING finger ligases (bottom). Catalysis might involve neutralization of the oxyanion that is formed as ubiquitin’s carboxyl terminus is transferred to the ε-amino group of a substrate lysine. HECT domain ligases transfer ubiquitin from the E2 to the substrate by forming an E3–ubiquitin thioester intermediate, which requires a conserved cysteine residue located in the HECT domain. Formation of the E3–ubiquitin thioester may occur before or after association with the substrate (Sub). The cullin–RING-finger ligase module is proposed to transfer ubiquitin directly, without forming an E3–ubiquitin thioester.

A key question concerns which APC/C subunits are intimately involved in APC/C’s catalytic activity, ubiquitin–protein ligation. A good bet would be those subunits, such as Apc2 and Apc11, that are found in other ligases. Insight into the role of these two proteins has recently come from studies on their counterparts in SCF. A Cdc53–Hrt1 subcomplex is sufficient to stimulate greatly the enzymatic activity of Cdc34 and therefore fulfills the definition of a minimal ubiquitin–protein ligase (Seol et al. 1999). Thus, APC/C, and SCF, and probably also VBC, might be members of a family of ubiquitin protein ligases whose catalytic centers consist of cullin-related and RING–H2 finger subunits (Fig. A). If so, what then might be the catalytic mechanism of ubiquitin transfer? The only ligases whose mechanism is remotely understood are the HECT (homology to E6-APcarboxyl terminus) domain proteins, which transfer ubiquitin from the E2 enzyme to the substrate via an E3 ubiquitin thioester intermediate (Scheffner et al. 1995) (Fig. B top). Recent experiments argue against such a mechanism for the SCF and, by implication for the APC/C (Seol et al. 1999) (Fig. B, bottom). The ligase activity of the Cdc53–Hrt1 subcomplex is resistant to alkylating agents known to block thioester-forming enzymes such as E1 and E2. Furthermore, the catalytic activity of Cdc34 can be activated by synthetic polycations lacking any sulfhydryl groups. It has been proposed that ubiquitin transfer might be catalyzed by stabilizing the oxyanion formed in the transition state as ubiquitin’s carboxyl terminus is transferred to the ɛ-amino group of a substrate lysine (Seol et al. 1999). The cullin–RING-H2 finger ligase might provide a positive charge neutralizing the oxyanion or it might induce a conformational change in the E2 enzyme. The homology between APC/C and SCF subunits should help to test this hypothesis.

Despite their important and possibly fundamental similarities, the APC/C and SCF differ in several important respects. The APC/C appears to have many more stable core subunits, whereas the SCF appears, at the moment, to have many more ‘activator proteins,’ which are thought to recruit substrates to the core particle. It is interesting that the APC/C has thus far only been implicated in the degradation of proteins with functions in cell division, whereas SCF, possibly through its larger repertoire of substrate-recruiting proteins, mediates the degradation of proteins involved not only in cell division but also in cell signaling. A future challenge will be to explain the differences in APC/C and SCF in terms of the biology with which they are concerned.

WD activators: Cdc20 and Cdh2

Genetic studies in yeast and Drosophila have shown that APC/C-dependent degradation requires activator proteins, called Cdc20/Fizzy and Cdh2/Fizzy-related, which are evolutionarily conserved and play a fundamental role in regulating the activity and substrate specificity of the APC/C. The Drosophila fizzy gene encodes a protein composed of seven WD repeats, which is required for sister chromatid separation and degradation of cyclin A and B during mitosis (Dawson et al. 1995; Sigrist et al. 1995). A related WD protein called Fizzy-related is expressed later in development when cell cycles contain a G1 phase (Sigrist and Lehner 1997). In these somatic cycles, APC/C-dependent degradation during mitosis depends on Fizzy, whereas its maintenance during G1requires Fizzy-related. At least in Drosophila, the two WD proteins seem to function as cell cycle stage-specific activators of cyclin degradation.

A similar picture has also emerged from studies in budding yeast. The Fizzy homolog Cdc20 mediates degradation of the anaphase inhibitor Pds1 (Visintin et al. 1997; Shirayama et al. 1998), the S-phase B-type cyclin Clb5 (M. Shirayama and K. Nasmyth, unpubl.), and the mitotic cyclin Clb3 (Alexandru et al. 1999). Proteolysis of these Cdc20 substrates commences at the metaphase-to-anaphase transition. Degradation of another set of substrates including the mitotic cyclin Clb2, the polo kinase Cdc5, and the spindle protein Ase1, commences during anaphase and requires the Fizzy-related homolog Cdh2/Hct1 (Schwab et al. 1997; Visintin et al. 1997;Charles et al. 1998; Shirayama et al. 1998). Overexpression of Cdc20 and Cdh2 induces APC/C-dependent degradation of their substrates at all stages of the cell cycle. These data suggest, but do not prove, that these WD proteins confer substrate specificity and that their activity limits the rate of proteolysis mediated by the APC/C in vivo. In yeast and human cells, both Cdc20 and Cdh2 bind to the APC/C, but the timing and regulation of their association is regulated differently (Fang et al. 1998b; Kramer et al. 1998a; Zachariae et al. 1998a). The levels of Cdc20 protein rise and fall as cells enter and exit mitosis, with the result that Cdc20 is bound to the APC/C only during M phase (and possibly during late G2). In contrast, the level of Cdh2 remains constant during the cell cycle, but it only binds the APC/C during G1. For the rest of the cell cycle, the association of Cdh2 with the APC/C is inhibited due to its phosphorylation by Cdk1 (Zachariae et al. 1998a).

The APC/C appears to possess other activator proteins besides Cdc20 and Cdh2. Budding yeast, for example, contain a thirdCDH1/CDC20-related gene called AMA1, which is expressed and spliced only during meiosis (K.F. Cooper and R. Strich, pers. comm.). The Ama1 protein binds to the APC/C core and appears to be necessary for Clb1 proteolysis. Fission yeast contain a family of at least five related WD proteins, including a Cdc20 homolog (Slp1) and a Cdh2 homolog (Srw1/Ste9) (Matsumoto 1997; Yamaguchi et al. 1997; Kitamura et al. 1998; Kominami et al. 1998). Whether proteins with motifs other than WD repeats can also activate ubiquitination mediated by the APC/C is not known.

How the WD coactivators promote ubiquitination is currently unknown. One possibility is that they recruit substrates to the APC/C in a manner analogous to SCF’s substrate recognition subunits. These proteins contain two domains: one required for substrate recognition and another one, the F box, which binds to the Skp1 subunit of the SCF complex (Patton et al. 1998). Interestingly, a subset of SCF’s substrate recognition subunits (of which yeast Cdc4 is the founder member) bind their targets via WD-repeat domains. Cdc20 and Cdh2 might recruit specific substrates and present them to an E2 enzyme that resides on the APC/C. A function as substrate recognition factors would be consistent with the substoichiometric binding of these proteins to the core APC/C particle and their dose-dependent stimulation of APC/C activity. However, this model predicts an interaction between the activator proteins and substrates, which has not yet been reported. Neither Cdc20 nor Cdh2 possess F boxes and presumably interact with the APC/C core particle via some other motif.

Cell cycle progression and the APC/C

Just as CDKs both promote the onset of DNA replication and inhibit the formation of pre-RCs needed for a new round, so does the APC/C both promote and hinder cell division progression depending on the state of the cell cycle. By destroying anaphase inhibitors such as Pds1 and Cut2, APC/CCdc20promotes sister chromatid separation at the metaphase-to-anaphase transition. By destroying B-type cyclins and geminin, APC/CCdc20 promotes inactivation of mitotic CDKs (and thereby facilitates cytokinesis) and removes a block to chromosome rereplication at the same time as it promotes sister chromatid separation. However, by maintaining B-type cyclin degradation following exit from mitosis, APC/CCdh2 delays both entry into S phase and preparations for a new round of mitosis. It is interesting that Cdh2 appears to be absent from embryonic cells whose cell cycles do not have a significant gap between M and S phases. The activity of APC/CCdh2 is very possibly crucial for generating the G1 phase of somatic cells. It seems likely that, as new APC/C substrates are discovered, the APC/C and its activators will be found to orchestrate a pattern of proteolysis, whose complexity rivals that of some transcriptional programs.

Control of sister chromatid separation

Sister chromatids are held together during G2 by a multisubunit complex called cohesin. At least in budding yeast, sister chromatid separation at the metaphase-to-anaphase transition appears to be triggered by the sudden disappearance from chromosomes of Scc1, one of cohesin’s subunits (Michaelis et al. 1997). Though the removal of Scc1 from chromosomes depends on APC/CCdc20, it is unlikely to be due to proteolysis mediated by the APC/C itself. Scc1 levels do decline during anaphase and Scc1’s proteolysis during the subsequent G1 period is at least partly dependent on APC/C. However, the key event for the removal of Scc1 from chromosomes appears to be its proteolytic cleavage, which is mediated by the separin protein Esp1 (Uhlmann et al. 1999). Esp1 is inhibited by Pds1, whose destruction is indeed mediated by APC/CCdc20 and is essential for the activation of Esp1 (Cohen-Fix et al. 1996; Ciosk et al. 1998). Thus, APC/CCdc20 mediates sister separation only indirectly, by mediating the proteolysis of a ‘securin’ protein, Pds1, which enables Esp1 to destroy cohesion directly (Fig.). As predicted by this hypothesis, inactivation of Pds1 allows cells to separate sister chromatids and remove Scc1 from chromosomes in the absence of APC/CCdc20activity. Surprisingly, Pds1 is neither essential for viability (Yamamoto et al. 1996a) nor for the correct timing of sister chromatid separation in budding yeast at low temperatures (Alexandru et al. 1999). Additional mechanisms must therefore participate in the control of sister chromatid separation. In an unperturbed cell cycle, Pds1 degradation is an essential precondition rather than the direct trigger for anaphase onset. Although Pds1 is not essential for controlling sister chromatid separation in cycling cells, it is necessary for blocking sister separation in the presence of lagging chromosomes or spindle damage (see below). Interestingly, Pds1 is essential for proliferation at high temperatures and may have some role in preparing Esp1 for its activity once Pds1 has been destroyed (Ciosk et al. 1998).

APC/CCdc20 mediates sister chromatid separation. APC/CCdc20 ubiquitinates Pds1 and thereby marks it for proteolysis by the proteasome. This liberates the separin Esp1 from its inhibitor and allows it to mediate the proteolytic cleavage of the cohesin subunit Scc1 (green circles). Cleavage and disappearance from chromosomes of Scc1 results in the loss of cohesion between sister chromatids. It is not known whether the Esp1–Pds1 complex is dissociated upon ubiquitination or proteolysis of Pds1. It is currently unclear whether Esp1 itself cleaves Scc1 or whether it activates a protease. Lagging chromosomes and spindle damage block the activity of APC/CCdc20, which requires association of Mad2 with Cdc20.

A slightly different picture of sister chromatid separation has emerged from studies of fission yeast, where anaphase onset is also regulated by a separin–securin complex, whose subunits are called Cut1 and Cut2, respectively. Though Cut2 shares many properties with Pds1, such as binding to Cut1 and being degraded by the APC/C shortly before entry into anaphase, it has no obvious sequence homology (Funabiki et al. 1996, 1997). The notion that Cut2 might be an anaphase inhibitor was complicated by the finding that loss of Cut2 function prevents sister chromatid separation, which is the same phenotype as that caused by a nondegradable Cut2 protein lacking its destruction boxes. Because the loading of Cut1 onto spindles depends on Cut2, it has been proposed that sister separation requires first, the loading of the Cut1–Cut2 complex onto spindles and second, the activation of Cut1 by APC/C-dependent degradation of its inhibitor Cut2 (Kumada et al. 1998). It has also been suggested that the Cut1 separin might promote sister separation by increasing the force exerted on sister chromatids by microtubules. Whether Rad21, the fission yeast Scc1 homolog, is a target of Cut1 is not yet known. It is conceivable that separins regulate anaphase spindle function at the same time as they induce dissolution of sister chromatid cohesion by changing the properties of Scc1/Rad21.

APC/CCdc20 has also been implicated in promoting anaphase in animal cells, including humans (Dawson et al. 1995; Sigrist et al. 1995; Tugendreich et al. 1995; Kallio et al. 1998). Furthermore, degradation of a protein with similar properties to those of Pds1 and Cut2 coincides with proteolysis of B-type cyclins and is required for anaphase (Zou et al. 1999). Human Pds1 binds to an Esp1-like protein, suggesting that here too destruction of a securin by APC/CCdc20 liberates a separin protein, which then proceeds to split sister chromatids. Whether securins are the sole targets of APC/CCdc20 with regard to sister separation in animal cells is still unclear. Mutations in theDrosophila pimples gene cause a defect in sister chromatid separation at centromeric regions and the pimples gene product is degraded during mitosis (Stratmann and Lehner 1996). The properties of the Pimples protein therefore resemble those of the securin Cut2. It will be interesting to see whether degradation of Pimples is actually required for sister centromere separation and whether it is mediated by the APC/C.

Control of spindle function

Sister chromatid separation must be coordinated with spindle elongation and spindle disassembly with cytokinesis. It is conceivable that loss of sister chromatid cohesion is essential for proper spindle elongation at the onset of anaphase. Indeed, spindles fail to elongate when sister separation is inhibited by a noncleavable Scc1 variant (Uhlmann et al. 1999). Nevertheless, it would not be surprising were spindle dynamics altered by APC/CCdc20 through mechanisms that were independent of separin activity. Attachment of spindles to kinetochores is thought to be stabilized by tension generated when sister kinetochores attach to spindles emanating from opposite spindle poles. This tension is clearly lost at the onset of anaphase and some other mechanism presumably ensures that chromosomes never disattach from microtubules during anaphase. Thus stabilization of spindle–kinetochore attachments, an increased tendency of kinetochores to move towards the poles, and increased repulsion between polar microtubules (which causes poles to migrate apart) might all directly or indirectly be promoted by APC/C-dependent proteolysis.

The kinesin-related motor proteins Kip1 (D.M. Roof, pers. comm.) and Cin8 (E. Hildebrandt and M.A. Hoyt, pers. comm.), which are required for mitotic spindle function, are both degraded by APC/CCdh2 during anaphase. These motors are thought to cross-link and slide microtubules from opposite poles to create an outward pushing force which counteracts the activity of another motor called Kar3 (Hoyt and Geiser 1996). CENP-E is a mammalian, kinesin-related protein that binds to kinetochores early in mitosis and then localizes to the spindle midzone (Brown et al. 1994). CENP-E is degraded in late mitosis, consistent with its degradation by the APC/C pathway. The budding yeast Ase1 protein localizes to the midzone of mitotic spindles and is important for anaphase spindle elongation (Pellman et al. 1995). Degradation of Ase1 at the end of anaphase and during G1 depends on a destruction box-like sequence and is mediated by APC/CCdh2(Juang et al. 1997). Ipl1/Aurora is a conserved protein kinase that associates with spindles in yeast and animal cells. In yeast, Ipl1 regulates the interaction of kinetochores with microtubules by controlling, together with protein phosphatase 1, the phosphorylation status of the kinetochore component Ndc10 (Biggins et al. 1999; Sassoon et al. 1999). In yeast and animal cells Ipl1/aurora protein levels fluctuate during the cell cycle with kinetics consistent with degradation by the APC/C (Gopalan et al. 1997; Kimura et al. 1997; Roghi et al. 1998). However, it should be noted that degradation of none of these spindle-associated substrates is absolutely essential for spindle function in budding yeast. pds1 apc double mutants clearly manage to segregate chromosomes to opposite spindle poles, implying that Pds1 is the sole protein whose destruction is needed for sister chromatid segregation. Furthermore, cdh2 mutants, which fail to degrade Ase1, are viable. In conclusion, the only proteins whose destruction by the APC/C is known to be crucial for chromosome segregation are securins like Pds1 and Cut2.

Exit from mitosis: inactivation of CDKs

Inactivation of mitotic Cdk1 kinases is essential for several events during exit from mitosis, including disassembly of the mitotic spindle, chromosome decondensation, cytokinesis, reformation of the nuclear envelope, reactivation of transcription, and rebuilding a Golgi apparatus (Murray et al. 1989; Luca et al. 1991; Gallant and Nigg 1992;Holloway et al. 1993; Surana et al. 1993; Sigrist et al. 1995;Gottesfeld and Forbes 1997; Nakamura et al. 1997). Destruction of mitotic cyclins has long been thought to have a key role in Cdk1 inactivation because expression of high levels of nondegradable cyclins blocks Cdk1 inactivation in a wide variety of systems. However, more recent work in budding yeast has questioned the validity of this conclusion. Expression of high levels of a nondegradable variant of the mitotic cyclin Clb2 does indeed block cells in late anaphase with high levels of Cdk1 kinase activity, but this arrest is not observed in cells that express more moderate levels of nondegradable Clb2 (Amon et al. 1994) or in cdh2 mutants, which fail to degrade physiological levels of Clb2 (Schwab et al. 1997). These findings imply that neither degradation of Clb2 nor that of any other Cdh2-dependent substrate is essential for exit from mitosis in budding yeast. Exit from mitosis in the absence of Clb2 proteolysis is possible due to accumulation of the Cdk1 inhibitor Sic1 (Schwab et al. 1997).

These results do not, however, exclude an essential role for cyclin proteolysis in Cdk1 regulation. After all, at least six different B-type cyclins are present during mitosis in budding yeast. Cdk1 inactivation might depend on proteolysis of a B-type cyclin other than Clb2 by APC/CCdc20. Whereas deletion ofPDS1 allows cdc20 or apc mutants to separate their sister chromatids, it does not permit them to inactivate Cdk1 and to exit from mitosis (Lim et al. 1998). The identity of the mysterious protein whose destruction by APC/CCdc20 is necessary for Cdk1 down regulation has recently been revealed by looking for mutations that permit the proliferation of cells lacking both Cdc20 and Pds1. This showed, remarkably, that while APC/CCdc20 mediates sister chromatid separation by destroying Pds1, it promotes Cdk1 inactivation by destroying the S-phase cyclin Clb5 (M. Shirayama and K. Nasmyth, unpubl.). This suggests that in budding yeast, as must be the case in embryonic animal cells, APC/CCdc20 and not APC/CCdh2 has the more crucial role in down regulating Cdk1 during anaphase. Curiously, destruction of Clb5 by APC/CCdc20 is more important than that of other B-type cyclins. Why should this be the case? Cdk1 inactivation in budding yeast occurs through at least three mechanisms: (1) proteolysis mediated by APC/CCdc20, which destroys Clb3 and Clb5; (2) proteolysis mediated by APC/CCdh2, which destroys Clb3 and Clb2; and (3) accumulation of a CDK inhibitor (CKI) Sic1, which inactivates any surviving Cdk1–B-type cyclin complexes. By destroying Clb5, APC/CCdc20 not only directly eliminates a B-type cyclin but also permits the Cdc14 phosphatase (see below) to dephosphorylate Sic1 and Cdh2, which both switches off the proteolysis of Sic1 by SCF and turns on APC/CCdh2 (Fig. ). Clb5 presumably has some unique ability to antagonize Cdc14, and Clb5 must therefore be destroyed by a mechanism that is independent of Cdc14 activity (i.e., by APC/CCdc20-mediated proteolysis). Having activated APC/CCdh2 and the accumulation of CKIs, APC/CCdc20 is no longer required to down regulate Cdk1. Thus, Cdc20, which must not reactivate the APC/C until sister chromatids produced during the next S phase are ready to be separated, is therefore destroyed as cells enter G1. The destuction by APC/CCdc20 of securins that block sister separation, and of B-type cyclins that block cytokinesis and the reformation of pre-RCs ensures that cells cannot attempt cell cleavage or chromosome reduplication until they have already separated sister chromatids during anaphase.

Regulation of cell cycle progression by APC/C-dependent proteolysis. (A) Sister chromatid separation and Cdk1 inactivation both depend on APC/CCdc20. Activation of APC/CCdc20 but not APC/CCdh2 requires phosphorylation of core subunits by Cdk1–cyclin B. APC/CCdc20 mediates sister chromatid separation and Cdk1 inactivation by degrading securins and B-type cyclins, respectively. APC/CCdc20also creates the conditions that keep Cdk1–cyclin B complexes inactive during the ensuing G1 phase. Cdk1 inactivation by APC/CCdc20 permits release from its inhibitor of the Cdc14 phosphatase which, by dephosphorylating Sic1 and Cdh2, causes accumulation of a CKI and activation of APC/CCdh2. Broken lines indicate proteolysis and P in a circle indicates phosphorylation (see text for details). (B) Changes in the abundance and activity of CDKs, the securin Pds1, the CKI Sic1, and both forms of the APC/C during the cell cycle.

It is a reasonable working hypothesis that APC/CCdc20 is responsible for the proteolysis of both cyclin A and cyclins B1 and B2 in animal cells. For example, neither protein is degraded in fizzy (cdc20) mutants in Drosophila (Sigrist et al. 1995). Furthermore, cyclins B1 and B2 disappear at the very onset of anaphase; that is, at the same time as the degradation of securins like Pds1. In animal cells, APC/CCdh2 is probably only required for maintaining mitotic cyclin proteolysis once cells have entered G1. In yeast, however, also APC/CCdh2participates in the destruction of B-type cyclins as cells exit from mitosis. The degradation of Clb2 largely by APC/CCdh2 helps to explain what has long been a curious difference between yeast and animal cells. Endogenous cyclin B molecules are unstable in animal cells arrested in late anaphase due to expression of nondegradable cyclin B variants, whereas Clb2 remains stable under similar circumstances. It turns out that Cdh2 is inactivated by Cdk1 and APC/CCdh2 cannot therefore function (and degrade Clb2) until Cdk1 has been inactivated (see below).

Cytokinesis

In animal and budding yeast cells, the APC/C is essential for cytokinesis. The main role of APC/C in this regard is presumed to be cyclin proteolysis leading to Cdk1 inactivation. It is therefore curious that fission yeast apcmutants eventually arrest with a ‘cut’ phenotype (Yanagida 1998). The mutants fail to separate sister chromatids and to destroy mitotic cyclins but proceed with aspects of cytokinesis such as formation of a septum. Curiously, cytokinesis in fission yeast seems to require inactivation of Cdk1 as much as it does in other organisms (Yamano et al. 1996). One explanation for this paradox is that Cdk1 in fission yeast can be inactivated, albeit slowly, by an APC/C-independent mechanism such as phosphorylation on tyrosine 15.

NIMA

In A. nidulans, entry into mitosis requires Cdk1–cyclin B and a second kinase called NIMA (Osmani et al. 1991; Osmani and Ye 1996). NIMA protein levels fluctuate during the cell cycle with kinetics resembling that of cyclin B and NIMA is stabilized inbimACDC16 and bimEAPC1mutants, suggesting that NIMA is an APC/C target (Ye et al. 1998). In contrast to most APC/C substrates, however, the degradation determinant of NIMA is located in the carboxyl terminus, which does not contain an obvious destruction box. Deletion of this domain does not affect NIMA’s ability to promote entry into mitosis but blocks the mitotic degradation of NIMA and exit of cells from mitosis (Pu and Osmani 1995). At least in A. nidulans, mitotic degradation of NIMA is an essential function of the APC/C. The identification of NIMA as an APC/C substrate sheds new light on the previous proposal that BIMEAPC1 might be a negative regulator of entry into mitosis. This idea was based on the finding thatbimEAPC1 nimA double mutants do not arrest in G2 as nimA single mutants do, but undergo aspects of mitosis such as chromosome condensation (Osmani et al. 1988). Recently, it was shown that cyclin B and the mutant NIMA protein accumulate in the double mutant, which superinduces Cdk1–cyclin B activity and produces sufficient NIMA kinase activity to induce mitosis (Ye et al. 1998). As predicted by this model, bimEAPC1 mutants cannot enter mitosis when the nimA gene is deleted, and the double-mutant cells arrest in G2.

Control of DNA replication

Initiation of DNA replication depends on a period devoid of CDK activity in which pre-RCs are assembled followed by the activation of S-CDKs, which trigger origin unwinding and the emergence of replication forks. Because of its essential role in mitotic Cdk1 inactivation, the APC/C is also thought to be essential for rereplication. Most, if not all, apc mutants arrest with replicated DNA and high Cdk1 activity but then fail to rereplicate their chromosomes. Furthermore, transient inactivation of Cdk1 kinases in yeast by expression of the Cdk1 inhibitor Sic1 causes rereplication inapc mutants shifted to their restrictive temperature (E.A. Noton and J.F.X. Diffley, pers. comm.). This suggests that Cdk1 inactivation is the sole essential role of the yeast APC/C with respect to DNA replication. Recently, certainapc mutants were reported to rereplicate their chromosomes despite arresting with high Cdk1 activity (Heichman and Roberts 1996,1998). It was proposed that the APC/C prevents rereplication by destroying an activator of DNA replication. This interesting but controversial claim fits awkwardly with the notion that cells naturally arrest the cell cycle (and prevent chromosome reduplication) by inhibiting the APC/C (see below). Another study of the same apc mutants detected replication only of mitochondrial DNA (Pichler et al. 1997). In vertebrate cells, DNA re-replication depends not only on degradation by the APC/C of mitotic cyclins but also of geminin (McGarry and Kirschner 1998). In human cells, geminin accumulates during S phase and is degraded during anaphase. Geminin prevents recruitment of MCM proteins into pre-RCs at origins. Initiation of DNA replication requires association of the conserved protein kinase Cdc7 with its regulatory subunit Dbf4 whose abundance fluctuates during the cell cycle. The Dbf4 protein appears as cells enter S phase and is degraded by APC/CCdc20 at the metaphase-to-anaphase transition (M.G. Ferreira and J.F.X. Diffley, pers. comm.; Cheng et al. 1999). At least in budding yeast, Dbf4 proteolysis is not essential for proliferation but it might contribute to the high fidelity of DNA replication. The degradation of Dbf4 might prevent the firing of those origins that have acquired preRCs as Cdk1 activity drops during exit from mitosis.

Polo kinase and Cdc20: first activators, then substrates

Two proteins that regulate APC/C, its Cdc20 activator and the Polo/Cdc5 protein kinase, are themselves regulated by APC/C-mediated proteolysis. Both proteins remain constant during embryonic cell cycles in Xenopus eggs but they fluctuate during the cell cycle in budding yeast and human cells in a manner similar to cyclin B. Ubiquitination and degradation of Cdc5, the polo-like kinase of budding yeast, depend on APC/CCdh2 and on two destruction box-like sequences (Charles et al. 1998; Shirayama et al. 1998). Expression of a nondegradable Cdc5 variant prevents the accumulation of mitotic cyclins and the formation of mitotic spindles suggesting that Cdc5 degradation is important for inactivating APC/C-dependent cyclin degradation as cells enter S phase. Plk1, the humanpolo-like kinase, is also ubiquitinated by the APC/C although it lacks an obvious destruction box (Fang et al. 1998b). Budding yeast Cdc20 contains two destruction boxes that are important for APC/C-dependent degradation during G1 (Shirayama et al. 1998). In contrast to other substrates, Cdc20 is also degraded in an APC/C-dependent manner during other cell cycle stages, but this instability does not require the destruction boxes (Prinz et al. 1998). The fluctuation of Cdc20 results from constitutive degradation of the protein in combination with cell cycle-regulated transcription. Cdc20 degradation does not require Cdh2; however, it remains to be determined whether Cdc20 has a role in promoting its own degradation or whether Cdc20 degradation requires another activator or no activator at all. In either case, the constitutive degradation of Cdc20 demonstrates that APC/C activity is not necessarily cell cycle-regulated and destruction box-dependent. This implies that there might be many more substrates of APC/C that have no role in cell cycle regulation.

Controlled destruction

At least five different mechanisms are known or suspected to contribute to APC/C regulation: (1) abundance and (2) phosphorylation of activator proteins; (3) phosphorylation of core APC/C subunits; (4) binding of inhibitory protein complexes; and (5) substrate phosphorylation. Furthermore, it is clear that different forms of the APC/C, such as APC/CCdc20 and APC/CCdh2, can be regulated differently by the very same regulatory proteins. For example, Cdk1 kinases promote APC/CCdc20 activity at the metaphase-to-anaphase transition whereas they inhibit APC/CCdh2 activity from S phase until exit from mitosis. A major difference between embryonic and somatic cells is the absence of Cdh2 from early embryos (Sigrist and Lehner 1997; Lorca et al. 1998). This has the consequence that cyclins destroyed by APC/CCdc20 at the end of mitosis start to reaccumulate almost as soon as they have been degraded. High levels of cyclin E presumably permits cells to inactivate CDK inhibitory proteins without an extensive transcriptional program. The appearance of Cdh2 later in development introduces a G1 phase, a stable state in which cyclins remain unstable.

Regulation of APC/CCdc20

Activation of APC/CCdc20 during metaphase promotes proteolysis of securins (Pds1/Cut2), which liberates separins (Esp1/Cut2) and thereby induces sister chromatid separation. It also triggers proteolysis of mitotic cyclins and thereby contributes to Cdk1 inactivation. It is through the activity of APC/CCdc20 that eukaryotic cells link preparations for a new round of DNA replication with the prior separation of sister chromatids. Regulation of APC/CCdc20 is therefore central to the eukaryotic cell cycle. In yeast, Cdc20 accumulates during G2/M as a result of the transcriptional activation of the CDC20 gene, which requires mitotic Cdk1 kinases (Prinz et al. 1998). Cdc20 fluctuates in abundance in a similar manner in human cells (Weinstein 1997). In yeast and human cells, Cdc20 binds to the APC/C and the abundance of APC/CCdc20 complexes corresponds to the cellular amount of Cdc20 (Fang et al. 1998b; Kramer et al. 1998a). However, accumulation of Cdc20 alone is insufficient to trigger degradation of APC/CCdc20 substrates like Pds1. Premature expression of CDC20 in yeast causes Cdc20 to appear earlier during the cell cycle but has no dramatic effect on the timing of Pds1 degradation (Prinz et al. 1998).

Recent experiments with extracts from clam oocytes and with purifiedXenopus and human APC/C suggest that activation of APC/CCdc20 involves two steps (Shteinberg et al. 1999; E. Kramer and J.-M. Peters, pers. comm.). First, the APC/C core particle is phosphorylated by Cdk1–cyclin B kinase during mitosis. Subsequently, Cdc20 stimulates the activity of the phosphorylated APC/C core. This conclusion is based on the observation that cyclin B–ubiquitination activity of interphase APC/C is stimulated by Cdc20 only after phosphorylation by Cdk1–cyclin B kinase. Furthermore, phosphatase treatment of the purified mitotic APC/C prevents its activation by Cdc20, which is restored by Cdk1–cyclin B kinase activity. InXenopus extracts, cyclin degradation was found to coincide with the phosphorylation of several APC/C subunits including Apc1 and Cdc27 (King et al. 1995; Peters et al. 1996). Both reactions are defective in extracts induced to enter mitosis in the absence of the Xenopus Cks1 homologue (also called p9 or Suc1), which is a conserved subunit of Cdk–cyclin complexes (Patra and Dunphy 1996, 1998). Cks1/Suc1 stimulates phosphorylation of purified APC/C by Cdk1–cyclin B and a small fraction was found to be associated with the mitotic but not the interphase APC/C in Xenopus extracts. These data suggest that Cks1/Suc1 targets Cdk1–cyclin B to phosphorylate APC/C subunits. Also Cdc20 has been found to be phosphorylated in a cell cycle-regulated manner in Xenopus egg extracts and human cells (Weinstein 1997; Kramer et al. 1998a; Lorca et al. 1998). However, nonphosphorylatable forms of Cdc20 are still able to activate mitotic Xenopus and human APC/C indicating that phosphorylation of Cdc20 by Cdk1–cyclin B is not essential for APC/C activation (E. Kramer, N. Scheuringer, and J.-M. Peters, pers. comm.).

The notion that Cdk1–cyclin B is required for activation of APC/CCdc20 explains how Cdk1 promotes its own destruction in extracts from embryonic cells (Felix et al. 1990). However, there must be a mechanism that delays the onset of cyclin proteolysis until after Cdk1–cyclin B kinases have triggered spindle formation, chromosome condensation, and nuclear envelope breakdown. Such a mechanism is particularly important for embryonic cells that supposedly lack surveillance mechanisms (checkpoints) that inhibit APC/CCdc20 while there still exist lagging chromosomes (see below). Such a time lag could be generated by a signalling cascade which links activation of Cdk1–cyclin B to that of APC/CCdc20. Cyclin degradation upon release ofXenopus egg extracts from a meiotic metaphase arrest (CSF arrest) was shown to depend on the polo kinase Plx (Descombes and Nigg 1998). Activation of Plx by Cdk1–cyclin B appears to be indirect and requires at least one additional kinase called xPlkk, which was shown to directly phosphorylate and thereby activate Plx (Qian et al. 1998). Polo kinases have been implicated in regulating the APC/C core particle. Human Plk1, which is activated during mitosis in vivo, phosphorylates several APC core subunits in vitro and this stimulates the cyclin B ubiquitination activity (Kotani et al. 1998). Transient inhibition of APC/CCdc20 could also account for the time lag between activation and destruction of Cdk1–cyclin B. Several pieces of evidence suggest that APC/C is inhibited by protein kinase A (PKA). Yeast APC/C mutants are suppressed by mutations in the cAMP–PKA pathway that lower the kinase activity of PKA (Yamashita et al. 1996; Yamada et al. 1997). In vitro, PKA phosphorylates purified mammalian APC/C and inhibits its cyclin B ubiquitination activity (Kotani et al. 1998). Consistent with a role in APC/C regulation PKA activity was found to fluctuate during the cell cycle of mammalian cells in culture. PKA activity increased at the beginning of mitosis and fell at metaphase. Taken together, these data provide two mechanisms that might contribute to the timely activation of APC/CCdc20 at the metaphase-to-anaphase transition: phosphorylation of core subunits by Plk1 and dephosphorylation of PKA phosphorylation sites by an as yet unknown phosphatase.

How is APC/CCdc20 inactivated to allow reaccumulation during the next cell cycle of substrates such as B-type cyclins and Pds1? In yeast and human cells, the amount of Cdc20 drops dramatically as cells exit from mitosis. Considering the dose-dependent action of Cdc20, it is likely that this drop causes a considerable reduction in the activity of APC/CCdc20. However, APC/CCdc20 might not be completely inactive in G1 cells. Overexpression of PDS1 was shown to cause accumulation of the Pds1 protein in G1-arrested cdc20 mutants but not in wild-type cells (Visintin et al. 1997).

Regulation of APC/CCdh2

In budding yeast, degradation of several APC/C substrates including Clb2, Cdc5, and Ase1 is solely dependent on Cdh2. Other substrates, like the mitotic cyclin Clb3, are degraded by APC/CCdc20 during anaphase but by APC/CCdh2 during telophase and G1(Alexandru et al. 1999). Many if not most APC/C substrates in animal cells behave like Clb3 in yeast. For example, most B-type cyclins are degraded by APC/CCdc20 at the metaphase-to-anaphase transition and by APC/CCdh2 during G1 (Sigrist et al. 1995; Sigrist and Lehner 1997). Whereas APC/CCdc20 is mainly active during mitosis, APC/CCdh2 is only active from the end of mitosis until shortly before entry into the next S phase (Fig. B). APC/CCdh2’s activity is mirrored by the ability of Cdh2 to bind to the APC/C core. Although Cdh2 is present throughout the cell cycle, it only binds to the APC/C during late anaphase and G1. In yeast, binding is inhibited from S phase until the end of mitosis by phosphorylation of Cdh2, which is mediated by Cdk1 (Zachariae et al. 1998a; Jaspersen et al. 1999). Ectopic inhibition of Cdk1 induces Cdh2 to bind to the APC/C irrespective of the cell cycle stage. Furthermore, a nonphosphorylatable Cdh2 variant binds and activates APC/C throughout the cell cycle, which prevents the accumulation of mitotic cyclins like Clb2 and Clb3. There is evidence that a similar phenomenon occurs in animal cells, where both Cdk1 and Cdk2 presumably conspire together to keep Cdh2 inactive (Knoblich et al. 1994; Lane et al. 1996). Indeed, phosphorylated human Cdh2 is unable to bind to and activate the APC/C, whereas nonphosphorylated Cdh2 can activate both interphase and mitotic APC/C (E. Kramer, N. Scheuringer, and J.-M. Peters, pers. comm.). Cdh2 and mitotic cyclins are therefore mortal enemies, each inhibiting the activity of the other. Supremacy in this battle switches from Cdh2 during G1, where APC/CCdh2prevents the accumulation of mitotic cyclins, to cyclin during S, G2, and M phases, where Cdks inhibit the binding of Cdh2 to the APC/C. The mechanism by which cells switch between these two self-sustaining states lies at the heart of the eukaryotic cell cycle.

Switching on APC/CCdh2

Activation of APC/CCdh2 at the end of mitosis is part of a more general program for inactivating Cdk1 (Fig.). In yeast, this process coincides with the sudden accumulation of the Cdk1 inhibitor Sic1. During S, G2, and M phases, Sic1 is phosphorylated by Cdk1 and then rapidly degraded by SCFCdc4. Meanwhile, Sic1 synthesis at the end of anaphase is activated by a transcription factor, Swi5 (Knapp et al. 1996), whose entry into the nucleus is inhibited by Cdk1 phosphorylation (Moll et al. 1991). Thus, the sudden accumulation of Sic1 at the end of mitosis depends on dephosphorylation of Swi5, which induces SIC1 transcription, and dephosphorylation of Sic1 itself, which abolishes SCFCdc4-mediated proteolysis. Recent work has highlighted the role of a conserved phosphatase called Cdc14 in dephosphorylating simultaneously Swi5, Sic1, and Cdh2 at the end of mitosis (Visintin et al. 1998; Jaspersen et al. 1999). Activation of Cdc14 is therefore a key event in Cdk1 inactivation. During most of the cell cycle, a regulatory subunit called Net1/Cfi1 inhibits the catalytic activity of Cdc14 and sequesters Cdc14 into a large complex which is assembled in the nucleolus (Shou et al. 1999; Visintin et al. 1999). In anaphase, Cdc14 is released from its inhibitor and then spreads throughout the cell, where it dephosphorylates Cdh2, Swi5, and Sic1. This process requires the polo-like kinase Cdc5, the protein kinases Cdc15, Dbf2, and Dbf20, the Ras-like GTPase Tem1, and the GDP/GTP exchange factor Lte1. Of these Cdc14 activators, most is known about Cdc5, whose transcription is activated by Cdk1 during G2. Cdc5 kinase activity appears somewhat later, suggesting that it is regulated by post-translational mechanisms such a phosphorylation (Cheng et al. 1998). Having promoted the activation of Cdc14 and thereby that of APC/CCdh2, Cdc5 is rapidly destroyed by APC/CCdh2. Proteins like Cdc5 and Cdc14 are vital for switching cells from the high Cdk1/low Sic1 state to a low Cdk1/high Sic1 state, but they are not required to maintain the low Cdk1 state during G1.

It is not yet known what triggers the activation of Cdc14. This process does not require sister chromatid separation as the release of Cdc14 from the nucleolus and Cdk1 inactivation both occur in esp1mutants (Surana et al. 1993; Visintin et al. 1999). Instead, Cdc14 activation requires, like sister chromatid separation, proteolysis mediated by APC/CCdc20. Nondegradable Pds1 not only blocks anaphase onset but also Cdk1 inactivation (Cohen-Fix and Koshland 1999; Tinker-Kulberg and Morgan 1999), possibly by inhibiting the release of Cdc14 from the nucleolus. The dependence of Cdk1 inactivation on Pds1 proteolysis helps to ensure that sister chromatid separation precedes cytokinesis and chromosome rereplication. The S phase cyclin Clb5 has been identified as the second Cdc20 substrate whose degradation is essential for Cdk1 inactivation (M. Shirayama and K. Nasmyth, unpubl.). The Cdk1–Clb5 kinase might be especially potent in counteracting the phosphatase activity of Cdc14. In conclusion, APC/CCdc20-dependent activation of Cdc14 tilts the battle between Cdk1–cyclin B versus Cdh2 and Sic1 in favor of the latter and therefore has a key role in inactivating Cdk1 at the end of mitosis (Fig. ). What roles the Cdc14 homologs in animal cells have, remains to be determined (Li et al. 1997). In animal cells, most if not all A- and B-type cyclins appear to be degraded by APC/CCdc20, whose activity does not depend on Cdc14. It is therefore unclear whether Cdc14’s main role in animal cells will be to downregulate Cdk1, for example, by switching off proteolysis of CKIs like p27Kip1, or to regulate other aspects of mitotic exit and cell cleavage.

Maintainance and termination of the low kinase state

How are Cdk1–Clb kinases kept inactive during G1 and how are they reactivated as cells enter S phase? Once APC/CCdh2 and Sic1 have overwhelmed Cdk1 activity with the help of APC/CCdc20 and Cdc14, they promote their own activity by keeping mitotic cyclins unstable and Sic1 stable. The maintainance of the low kinase state and thereby the G1 period depends on the suppression of any Cdk1 activity capable of phosphorylating Cdh2 and Sic1. In yeast, APC/CCdh2 is important for G1 cell cycle arrest in response to mating pheromones and nutrient starvation.apc and cdh2 mutants are unable to halt DNA replication due to precocious accumulation of S phase-promoting B-type cyclins (Kumada et al. 1995; Irniger and Nasmyth 1997; Kominami et al. 1998). APC/CCdh2 activity also persists in G1 in human cells (Brandeis and Hunt 1996). Because differentiation usually requires arrest in G1, it will be interesting to establish whether persistent APC/CCdh2 activity is important in terminally differentiated cells.

Reactivation of Cdk1 activity must await the accumulation of a special class of cyclins whose activity is refractory to both APC/CCdh2 and Sic1. In budding yeast, the G1 cyclins Cln1 and Cln2 have this property. They suddenly appear in late G1 as part of a transcriptional program that is activated as cells reach a critical size (Koch and Nasmyth 1994). The S-phase B-type cyclin Clb5 also accumulates at this stage due to the same transcriptional program. Though sensitive to Sic1, the Cdk1-Clb5 kinase is not destroyed by APC/CCdh2. The accumulation of Cdk1–Cln1 and Cdk1–Cln2 kinases in late G1 leads to the phosphorylation of Sic1 and thereby to its proteolysis. This activates the Cdk1–Clb5 kinase, which by phosphorylating Cdh2 switches cells finally back to a state in which mitotic cyclin B kinases can reaccumulate. The activation of Cdk1–Clb5 irreversibly switches the system to the high kinase state. Once established by Cdk1–Clb5, the high kinase state can be maintained by B-type cyclins such as Clb3 and later Clb2 which are susceptible to APC/CCdh2. In animals, Cdk2–cyclin E may have a role similar to that of Cdk1–Clb5 in yeast. In Drosophila, ectopic expression of cyclin E or a failure to express the Cdk2–cyclin E inhibitor Dacapo results in ectopic accumulation of mitotic cyclins (Knoblich et al. 1994; Lane et al. 1996). S phase-promoting kinases therefore have a more general role in organizing the cell cycle than previously anticipated. They trigger chromosome duplication and simultaneously prepare the stage for the separation of sister chromatids. It is therefore fascinating that their eventual destruction, for example, the proteolysis of Clb5 in yeast by APC/CCdc20, is a crucial trigger for exit from the high Cdk1 state. It is conceivable that destruction of cyclin A has a similar role in animal cells.

Stopping destruction

There are several occasions where cells block cell cycle progression by specifically inhibiting APC/C-dependent proteolysis. One example is the metaphase arrest of vertebrate eggs awaiting fertilization. Other examples are cell cycle delays induced by DNA damage, incomplete DNA replication or defects in the attachment of chromosomes on a bipolar spindle. These so-called checkpoint mechanisms are important for the integrity of chromosomes and the high fidelity with which they are transmitted during mitosis.

CSF arrest

As in most vertebrates, unfertilized Xenopus eggs are arrested in metaphase of the second meiotic division by an activity called cytostatic factor or CSF. During this arrest, Cdk1 activity remains high and both sister chromatid separation and cyclin B degradation are blocked, suggesting that CSF prevents APC/C-dependent degradation. CSF arrest depends on the c-Mos protein kinase (Sagata et al. 1989) and on activation of the mitogen-activated protein (MAP) kinase cascade. How activation of MAP kinase blocks degradation of APC/C substrates is unknown (Haccard et al. 1993; Posada et al. 1993; Kosako et al. 1994). Active MAP kinase does not directly inhibit the APC/C but seems to activate an inhibitor that does not copurify with the APC/C. For example, APC/C purified from CSF extracts is as active in cyclin B ubiquitination as that from mitotic extracts (Vorlaufer and Peters 1998). Fertilization triggers a transient increase in the cytoplasmic calcium concentration that causes activation of APC/C-dependent cyclin A and B degradation and thereby release from the metaphase arrest. Release from the CSF arrest and cyclin degradation can also be induced by treatments that increase intracellular calcium or by adding calcium to CSF extracts. Calcium activates the degradation of c-Mos, but this is not the trigger for cyclin degradation because c-Mos is degraded considerably later than cyclin B (Watanabe et al. 1991). The calcium wave causes a transient activation of calmodulin-dependent protein kinase II (CaMKII), which is both necessary and sufficient for activation of cyclin degradation and release from the CSF arrest (Lorca et al. 1993). Unfortunately, the relevant substrate of CaMKII is unknown.

DNA damage checkpoint

DNA damage causes fission yeast and animal cells to block entry into mitosis by preventing the activation of Cdk1–cyclin B. In budding yeast, cell cycle progression is instead halted by blocking Pds1 degradation. DNA damage induces phosphorylation of Pds1, which depends on MEC1 and RAD9, two genes known to be required for this checkpoint (Cohen-Fix and Koshland 1997). It has been suggested (but not yet proven) that Pds1’s phosphorylation renders it resistant to attack by APC/CCdc20. The persistence of Pds1 in response to DNA damage not only blocks sister chromatid separation but also prevents inactivation of Cdk1 (Tinker-Kulberg and Morgan 1999).

The APC/C as a target of the spindle checkpoint

The segregation of sister chromatids to opposite spindle poles at the metaphase-to-anaphase transition occurs due to loss of the cohesion that holds sisters together. Because loss of cohesion between sister chromatids occurs simultaneously on all chromosomes, it is vital that this process be blocked by ‘lagging’ chromosomes whose sister kinetochores have not yet attached to microtubules emanating from opposite spindle poles. Lagging chromosomes not only block sister separation but also inactivation of Cdk1–cyclin B kinases; they therefore freeze cells in a metaphase-like state. The control mechanism responsible for this cell cycle arrest has been called the mitotic checkpoint. Although this checkpoint is not essential during an unperturbed cell cycle in yeast, it seems to be an integral part of cell cycle control in mammalian cells (Taylor and McKeon 1997; Gorbsky et al. 1998). It is a control mechanism without which mitosis cannot properly function. The same mechanism blocks cell cycle progression upon artificial destruction of the spindle with microtubule depolymerizing drugs. The APC/C has emerged as the key target of the mitotic checkpoint.

The isolation of mutants that die rapidly in the presence of microtubule-depolymerizing drugs has identified several genes involved in the spindle checkpoint: MAD1, MAD2, MAD3,BUB1, BUB2, BUB3, MPS1, andPDS1 (Hoyt et al. 1991; Li and Murray 1991; Weiss and Winey 1996; Yamamoto et al. 1996b). Mutants in these genes are defective in halting various aspects of the cell cycle upon extended exposure to microtubule-depolymerizing drugs. Homologs of Mad1-3, Bub1, Bub3, and Pds1 have been identified in other organisms, including humans (Chen et al. 1996; Li and Benezra 1996; He et al. 1997; Taylor and McKeon 1997;Taylor et al. 1998; Zou et al. 1999).

In yeast and mammalian cells, Mad2 is found associated with Cdc20 and the APC/C (Fang et al. 1998a; Hwang et al. 1998; Kallio et al. 1998; Kim et al. 1998; Wassmann and Benezra 1998). At least in budding yeast, Mad1 and Mad3 also bind to Cdc20. There is compelling evidence that the Mad2–Cdc20 interaction is required to inhibit the ubiquitin ligase activity of APC/CCdc20, which then prevents the destruction of cyclin B and securins like Pds1.CDC20 mutant alleles, which are defective in the binding of Mad2, have a phenotype similar to that of mad mutants (Hwang et al. 1998; Kim et al. 1998; Alexandru et al. 1999). These cells fail to block sister separation and cyclin B destruction when spindles are damaged or the mitotic checkpoint is ectopically activated. How Mad2 inhibits APC/CCdc20 is unknown. It is interesting that tetramers of recombinant human Mad2 block cyclin B degradation in Xenopus egg extracts much more effectively than monomers do (Fang et al. 1998a). It is therefore possible that Mad1 and Mad3 catalyze the formation of active Mad2 tetramers in response to lagging chromosomes or spindle damage.

A key question is how cells sense that spindles are damaged or that kineochores have not formed bivalent attachments to the mitotic spindle. Two lines of evidence suggest that kinetochores themselves have a key role. In vertebrate cells, Mad1, Mad2, Bub1, and Bub3 localize selectively to unattached kinetochores and disappear from them as they attach to the mitotic spindle (Chen et al. 1996, 1998; Li and Benezra 1996; Taylor and McKeon 1997; Gorbsky et al. 1998; Taylor et al. 1998; Waters et al. 1998). Meanwhile in yeast, components of the CBF3 complex that mediate the attachment of centromeres to microtubules are necessary for cell cycle arrest when spindles are damaged (Hyman and Sorger 1995). It has therefore been proposed that unattached kinetochores mediate the assembly or activation of a Mad2-containing complex that when released from the kinetochore inhibits APC/CCdc20 throughout the cell (Chen et al. 1998; Gorbsky et al. 1998). Such complexes must be unstable and are regenerated as long as unattached kinetochores are present. When all kinetochores have been attached, the inhibitory activity declines, allowing APC/CCdc20 to become active.

How do the Mad and Bub proteins block sister chromatid separation and Cdk1 inactivation that require, at least in budding yeast, different forms of the APC/C? Recent analysis of the kinetics of Pds1 degradation and Cdk1 inactivation in mad and bubmutants suggests that the cell cycle arrest induced by microtubule depolymerization involves two distinct pathways (Alexandru et al. 1999;Fesquet et al. 1999; Fraschini et al. 1999). One pathway involving Mad1, Mad2, Mad3, and Bub1 inhibits APC/CCdc20, which blocks degradation of Pds1 and thereby keeps Esp1 inactive. This mechanism blocks sister separation. The persistence of Pds1 also blocks Cdk1 inactivation by inhibiting release of Cdc14 from the nucleolus (M. Shirayama and K. Nasmyth, unpubl.). However, mutants lacking Mad2 or Pds1 or both still block exit from mitosis, suggesting that there is a second pathway blocking Cdk1 inactivation. This pathway depends on Bub2 and Byr4. Only mutants lacking both pathways (e.g., mad2 bub2double mutants) inactivate Cdk1 and enter the next cell cycle as if nothing were amiss in the presence of microtubule depolymerizing drugs. Several pieces of evidence suggest that the Bub2 pathway blocks the release of Cdc14 from the nucleolus. The fission yeast homologs of Bub2 and Byr4 form a two-component GTPase-activating enzyme for a GTPase homologuous to Tem1 (Furge et al. 1998). Bub2 and Byr4 might inactivate Tem1 by converting the GTP-bound form to the presumably inactive, GDP-bound form. Tem1–GTP might be required for activation of the Dbf2 kinase, which was also found to be blocked by the Bub2 pathway (Fesquet et al. 1999). The Bub2 pathway is unlikely to be activated by the kinetochore but by some other as yet unknown defect cause by microtubule depolymerization (Wang and Burke 1995). Unlike Mad proteins, Bub2 is associated with spindle poles (Fraschini et al. 1999) and might coordinate cytokinesis with the arrival of chromosomes at the poles in late anaphase.

Summary

The APC/C was found by searching for the apparatus responsible for destroying mitotic cyclins at the metaphase-to-anaphase transition. This apparatus is a key part of the regulatory network that generates oscillations in the activity of mitotic CDKs. Studies of how the various forms of APC/C are regulated promise to provide a deep understanding of how eukaryotic cells generate CDK waves. However, the APC/C turns out to have a much more fundamental role in the eukaryotic cell cycle than merely being a counterweight to cyclin B synthesis. It mediates the separation of sister chromatids and is the target of regulatory mechanisms whose role is to ensure that daughter cells inherit one (usually two) complete copies of the genome. The control of sister chromatid separation appears to have become more and more important as organisms became more complicated and as their genomes grew in size. Bacteria, for example, barely regulate sister chromatid separation. Indeed, they commence to separate sister chromatids almost as soon as replication has been initiated. At the other extreme are mammalian cells, in which a surveillance mechanism merely needed for high fidelity chromosome transmission in yeast has clearly become an integral part of mitosis.

It was the discovery of the APC/C (and SCF) and the key roles that they have in eukaryotic cell reproduction that established once and for all the importance of ubiquitin mediated proteolysis in eukaryotic cell biology. Once perceived as a system exclusively involved in removing damaged proteins from the cell, ubiquitination is now perceived as a universal regulatory mechanism whose importance approaches that of protein phosphorylation.

The irreversibility of proteolysis is utilized by cells to give the cell cycle directionality. Once CKIs have been destroyed by SCF, it is very difficult for cells to inactivate Cdks. Their activity drives the onset of DNA replication and entry into mitosis. Likewise, the destruction of cyclin B and Pds1 triggers the separation of sister chromatids and the inactivation of Cdk1. By utilizing the same apparatus to degrade mitotic cyclins and anaphase inhibitors, eukaryotic cells ensure that preparations for chromosome rereplication cannot normally precede the separation of sister chromatids generated by a previous round of DNA replication. The recent discovery that destruction of IκB and β-catenin are mediated by SCF suggests that this formula is not restricted to cell division but also has a key role in signal transduction. The persistence of APC/C in fully differentiated quiescent cells suggests that its powerful substrate recognition apparatus is not confined to proteins involved in cell cycle progression.

That proteolysis should have a key role in cell cycle regulation is in retrospect possibly not too surprising. What is still less clear is why so much of the proteolysis that orders the cell cycle should be mediated by large ubiquitin ligase complexes as opposed to more conventional proteases. A key advantage of these ligases is that they can, with great specificity, promote the complete annihilation of a protein without the need for cleavage sites within functional domains of that protein. This then may be one of the key advantages of separating the marking process from proteolysis itself. It is interesting in this regard that the proteolytic program that mediates apoptosis in contrast utilizes proteolytic cleavage. Ubiquitination requires ATP, which is in short supply in dying cells.

The cell cycle ends in an orgy of protein degradation that sets the scene for a new round of duplication. With the discovery of the apparatus responsible for this cleanup, we have embarked on a whole new area of cell biology.

Acknowledgments

We thank Raymond Deshaies, John Diffley, Christian Gieffers, Andrew Hoyt, Marc Kirschner, Edgar Kramer, Jan-Michael Peters, David Roof, Masaki Shirayama, Randy Strich, Kazuo Todokoro, Frank Uhlmann, and Mitsuhiro Yanagida for communicating unpublished results. We are grateful to Jan-Michael Peters for critical comments on the manuscript. WZ thanks the European Molecular Biology Organisation for support through a long-term fellowship.

Footnotes

• ↵Corresponding author. Present address: Max-Planck-Institute of Molecular Cell Biology and Genetics, c/o EMBL D-69117 Heidelberg, Germany.

• E-MAIL Zacharia{at}EMBL-Heidelberg.de; FAX 49-6221-387-306.

• Cold Spring Harbor Laboratory Press

References

1. ↵
2. ↵
3. ↵
4. ↵
5. ↵
6. ↵
7. ↵
8. ↵
9. ↵
10. ↵
11. ↵
12. ↵
13. ↵
14. ↵
15. ↵
16. ↵
18. ↵
19. ↵
20. ↵
21. ↵
22. ↵
23. ↵
24. ↵
25. ↵
26. ↵

    (1998b) Direct binding of CDC20 protein family members activates the anaphase- promoting complex in mitosis and G1. Mol. Cell 2:163–171, .

27. ↵
28. ↵
29. ↵
30. ↵
31. ↵
32. ↵
33. ↵
34. ↵
35. ↵
36. ↵
37. ↵
38. ↵
39. ↵
40. ↵
41. ↵
42. ↵
43. ↵

    (1998) CDC16 controls initiation at chromosome replication origins. Mol. Cell 1:457–463, .

44. ↵
45. ↵
46. ↵
47. ↵
48. ↵
49. ↵
50. ↵
51. ↵
52. ↵
53. ↵
54. ↵
55. ↵
56. ↵
57. ↵
58. ↵
59. ↵
60. ↵
61. ↵
62. ↵
63. ↵
64. ↵
65. ↵
66. ↵
67. ↵
68. ↵
69. ↵
70. ↵
71. ↵
72. ↵
73. ↵
74. ↵
75. ↵
76. ↵
77. ↵
78. ↵
79. ↵
80. ↵
81. ↵
82. ↵
83. ↵
84. ↵
85. ↵
86. ↵
87. ↵
88. ↵
89. ↵
90. ↵
91. ↵
92. ↵
93. ↵
94. ↵
95. ↵
96. ↵
97. ↵
98. ↵
99. ↵
100. ↵
101. ↵
102. ↵
103. ↵

     (1998) Xe-p9, a Xenopus Suc1/Cks protein, is essential for the Cdc2-dependent phosphorylation of the anaphase- promoting complex at mitosis. Genes & Dev. 12:2549–2559, .

104. ↵
105. ↵
106. ↵
107. ↵
108. ↵
109. ↵
110. ↵
111. ↵
112. ↵
113. ↵
114. ↵
115. ↵
116. ↵
117. ↵
118. ↵
119. ↵
120. ↵
121. ↵
122. ↵
123. ↵
124. ↵
125. ↵
126. ↵
127. ↵
129. ↵
130. ↵
131. ↵
132. ↵
133. ↵
134. ↵

     (1998) The human homologue of Bub3 is required for kinetochore localization of Bub1 and a Mad3/Bub1-related protein kinase. J. Cell Biol. 142:1–11, .

135. ↵
136. ↵
137. ↵
138. ↵
139. ↵
140. ↵
141. ↵
142. ↵
143. ↵
144. ↵
145. ↵
146. ↵
147. ↵
148. ↵
149. ↵
150. ↵
151. ↵
152. ↵
153. ↵
154. ↵
155. ↵
156. ↵
157. ↵
158. ↵
159. ↵
160. ↵
161. ↵
162. ↵
163. ↵
164. ↵
165. ↵

Is MPF a cyclin

What are maturation promoting factors MPF )? What do they start?

What is the relationship between cyclin

What does the CDK cyclin complex do?