Phosphorylation by Cdc28 Activates the Cdc20-Dependent Activity of the Anaphase-Promoting Complex

Proteolysis plays a critical role in the eukaryotic cell cycle. During the exit from mitosis, ubiquitin mediated proteolysis destroys an inhibitor of sister chromatid separation (Pds1 in budding yeast and Cut2 in fission yeast; Holloway et al. 1993; Funabiki et al. 1996; Yamamoto et al. 1996) and the mitotic cyclins (Clb1–Clb4 in budding yeast; Ghiara et al. 1991; Glotzer et al. 1991; Yamano et al. 1996). These proteins are targeted for degradation by the anaphase-promoting complex (APC) or cyclosome, which is the E3 ubiquitin ligase for cyclins (King et al. 1995; Sudakin et al. 1995; Zachariae et al. 1996), Pds1 (Cohen-Fix et al. 1996; Funabiki et al. 1997), and other substrates (Juang et al. 1997; Prinz et al. 1998; Shirayama et al. 1998), marking them for destruction by the 26S proteasome. The APC is regulated by the binding of two conserved activators, Cdc20 and Hct1 (also known as Cdh1; Schwab et al. 1997; Visintin et al. 1997; Fang et al. 1998b; Kitamura et al. 1998; Lorca et al. 1998). In budding yeast, Cdc20-dependent APC activity initiates the metaphase to anaphase transition and the series of events that activate the Hct1-dependent APC, which induces complete mitotic cyclin destruction (Lim and Surana 1996; Visintin et al. 1997; Shirayama et al. 1999). Hct1 acts in conjunction with the cyclin-dependent kinase (Cdk) inhibitor Sic1 to induce the rapid drop in Cdc28-associated kinase activity that drives cells out of mitosis and into the next G1 (Mendenhall 1993; Donovan et al. 1994; Amon 1997; Li and Cai 1997). The Hct1- and the Cdc20-dependent APC can both target Pds1 for destruction (Visintin et al. 1997; Rudner et al. 2000), suggesting that the main difference between them is the time during the cell cycle when each is active (Prinz et al. 1998).

Phosphorylation of Hct1 by Cdc28/Clb complexes keeps it from binding or activating the APC (Zachariae et al. 1998; Jaspersen et al. 1999). This phosphorylation is removed by Cdc14, a phosphatase that is activated after Cdc20-dependent destruction of Pds1, Clb2, and the S-phase cyclin, Clb5 (Visintin et al. 1998; Jaspersen et al. 1999; Shirayama et al. 1999; Yeong et al. 2000). The late activation of Cdc14 ensures that cells do not inactivate Cdc28 and exit mitosis until well after they have initiated sister chromatid segregation.

Cdc20 is regulated in at least three ways: the gene is transcribed only in mitosis, the protein is targeted for destruction by the APC, and Cdc20 activity is inhibited by the spindle checkpoint, which monitors whether chromosomes have attached to the spindle properly (Weinstein 1997; Fang et al. 1998a; Hwang et al. 1998; Kallio et al. 1998; Kim et al. 1998; Kramer et al. 1998; Prinz et al. 1998; Shirayama et al. 1998).

The Cdc20-dependent APC is regulated by phosphorylation. APC subunits are phosphorylated in fission yeast, frogs, clams, and mammalian tissue culture cells (Hershko et al. 1994; Peters et al. 1996; Yamada et al. 1997; Kotani et al. 1998). Phosphorylation correlates with APC activity in vivo, and experiments in vitro have suggested that phosphorylating the APC regulates Cdc20 binding and APC activity (Kotani et al. 1998, Kotani et al. 1999; Shteinberg et al. 1999). Studies in frog egg extracts and mammalian tissue culture cells have shown that the protein kinase Plk (known as Cdc5 in budding yeast and Plx1 in frogs) and the complex of Cdc2, Cyclin B, and Cks1, a small Cdk binding protein, can phosphorylate the APC in vitro. Depletion of either Cks1 or Plx1 from frog extracts blocks cyclin destruction, suggesting that both Cdc2 and Plx1 may activate the APC (Patra and Dunphy 1996; Descombes and Nigg 1998; Kotani et al. 1998, Kotani et al. 1999), but the relative importance of these two kinases in vivo is unclear. Phosphorylation of the APC by cAMP-dependent protein kinase A inhibits the APC both in vivo and in vitro (Yamashita et al. 1996; Kotani et al. 1998). Lastly, protein phosphatase 2A (PP2A) inhibits the APC (Lahav-Baratz et al. 1995; Shteinberg et al. 1999), whereas PP1 activates the APC (Yamada et al. 1997).

In the accompanying paper (Rudner et al. 2000), we show that CDC28-T18V, Y19F (CDC28-VF), and other mutants with altered mitotic Cdc28 activity are compromised in activating the Cdc20-dependent APC, revealing a requirement for Cdc28 in APC activation. Here, we show that CDC28-VF is defective in the mitotic phosphorylation of the APC and that this phosphorylation depends on Cdc28 activity both in vivo and in vitro. Mutating potential phosphorylation sites in the APC components Cdc16, Cdc23, and Cdc27 reduces Cdc20 binding to the APC and Cdc20-dependent APC activity in vivo.

Table lists the strains used in this work. All strains are derivatives of the W303 strain background (W303-1a; Rodney Rothstein, Columbia University, NY). Standard genetic techniques were used to manipulate yeast strains (Sherman et al. 1974) and standard protocols were used for DNA manipulation (Maniatis et al. 1982). All deletions and replacements were confirmed by PCR or by mutant phenotype. The sequences of all primers used in this study are available upon request. The bacterial strains TG1 and DH5α were used for amplification of DNA.

BAR1 was deleted using pJGsst1 (a gift of Jeremy Thorner, University of California, Berkeley, CA). CDC27-MBP strains were made by crossing JC35 (a gift of Julia Charles, University of California, San Francisco, CA) to the appropriate strains. cdc26Δ strains were described previously (Hwang and Murray 1997). clb2Δ strains were made by crossing K1890 (a gift of Kim Nasmyth, Institute of Molecular Pathology, Vienna, Austria) to the appropriate strains. pCUP-GFP12-lacI12 and lacO:LEU2 were integrated using pSB116 (Biggins et al. 1999) and pAFS59 (Straight et al. 1996), respectively. pGAL-MPS1 strains were made with pAFS120 (Hardwick et al. 1996). pGAL-PDS1-HA strains were made by crossing RTK43 (a gift of Rachel Tinker-Kulberg, Johns Hopkins University, MD) to the appropriate strains. APC9 was tagged by the PCR-targeting method. Cells were transformed with a cassette containing the bacterial KAN^Rgene that confers G418 resistance in W303. The cassette was amplified by PCR from pFA6a-3HA-kanMX6 (Longtine et al. 1998) with primers containing the sequences that flank the stop codon of APC9. The construction of CDC20-myc12 and cks1-38 is described in Rudner et al. 2000(this issue).

Alanine-substituted mutants in CDC16, CDC23, and CDC27 were made using site-directed mutagenesis (Kunkel 1985). Mutations were confirmed by the introduction of new restriction enzyme sites and by sequencing (ABI). For CDC16, the EcoR1/Xho1 fragment of pWAM10 (Lamb et al. 1994) was cloned into KS (−) (Stratagene) to create pAR290. pAR290 was mutagenized to create pAR293, which contains all six serine/threonine to alanine substitutions. pAR294 was cut with EcoRI and NotI, and ligated to a EcoRI/PstI PCR fragment that contains the 3′ end of CDC16, a PstI/SpeI PCR fragment that contains the TRP1 gene, and a SpeI/NotI PCR fragment that contains the 3′ untranslated region of the CDC16 gene. The resultant plasmid, pAR303, was cut with XhoI and NotI, and integrated at the CDC16 locus. The TRP⁺ transformants were screened by PCR for the presence of all mutations. For CDC23, the BamHI/NotI fragment of pRS239 (Lamb et al. 1994) was cloned into KS (−) to create pAR228. pAR228 was mutagenized to create pAR240, which contains the single serine to alanine substitution in CDC23. pAR228 was cut with BamHI and NotI, transformed into cdc23-1 cells (ADR1285), and selected for growth at 37°C. Transformants were screened by Western blot for the HA tag present at the 3′ end of the gene, and by PCR for the presence of the alanine substitution. For CDC27, the PstI/NotI fragment of pJL25 (Lamb et al. 1994) was cloned into KS (−) to create pAR201. pAR201 was mutagenized to create pAR203, which contains all five serine/threonine to alanine substitutions in CDC27. pAR203 was cut with NdeI and NotI, and ligated to a NdeI/XbaI PCR fragment that contains the KAN^R gene and a XbaI/NotI PCR fragment containing the 3′ untranslated region of CDC27. The resultant plasmid, pAR271, was cut with KpnI and NotI, and integrated at the CDC27 locus. Transformants were screened by PCR for the presence of all mutations.

Physiological experiments were performed as described in the accompanying paper (Rudner et al. 2000, this issue). Hydroxyurea (HU; Sigma-Aldrich) was added directly to media at a final concentration of 200 mM.

Cells were fixed for indirect immunofluorescence in 3.7% formaldehyde for 1 h. The spindles were visualized by antialpha-tubulin (Harlan Sera-Lab) immunofluorescence as described previously (Hardwick and Murray 1995), except that the blocking reagent used was 2% BSA, PBS. Short spindles are bipolar spindles <2 μm long.

Immunoprecipitation, Western blots, APC assay, and Cdc20 binding to the APC were performed as described in the accompanying paper (Rudner et al. 2000). Modifications of the basic protocol are detailed below.

To resolve the phosphorylated forms of Cdc27 by Western blot, samples were electrophoresed on a 12.5% polyacrylamide gel containing 0.025% bisacrylamide. The phosphorylated forms of Cdc16 were resolved by Western blot on a 10% polyacrylamide gel containing 0.13% bisacrylamide.

The following antibodies were used in this study: 9E10 ascites (BabCO); affinity-purified rabbit polyclonal anti-Clb2 and anti-Clb3 antibodies (Kellogg and Murray 1995); rabbit polyclonal anti-Sic1 serum (a gift of Mike Mendenhall, University of Kentucky, Lexington, KY); 12CA5 ascites (BabCO); rabbit polyclonal anti-Cdc16, anti-Cdc23, and anti-Cdc27 (Lamb et al. 1994); and rabbit polyclonal anti-Cdc26 antibody (Hwang and Murray 1997). Details on the use of these antibodies can be found in the accompanying paper (Rudner et al. 2000).

Yeast cells were arrested in G1 with alpha factor, in S-phase with HU, and in mitosis by spindle checkpoint activation and temperature shift. Once the cells were arrested at the indicated stage of the cell cycle, 50 ml of OD₆₀₀ 0.8 cells were harvested by centrifugation, washed twice in H₂O, and resuspended in 1 ml phosphate-free complete synthetic medium (Rothblatt and Schekman 1989) containing 0.5–1 mCi ³²PO₄ (Amersham Pharmacia Biotech). Cells were labeled for 1 h, harvested by centrifugation, washed once in H₂O, and were then frozen in screw-cap tubes (Sarstedt). These tubes were used throughout the procedure to prevent radioactive contamination. The frozen yeast pellets were processed for immunoprecipitation as described in the accompanying paper (Rudner et al. 2000) with the following modifications. 2–3 μg anti-Cdc26 antibody was bound to 20 μl protein A beads for 20 min on ice. These beads were then incubated with 10–20 mg of unlabeled cell lysate made from cdc26Δ cells for 1–2 h. After incubation, the beads were washed twice in lysis buffer. At the same time, the labeled cell lysate (typically 10 mg) was precleared in 75 μl protein A CL-4B Sepharose beads (Sigma-Aldrich) for 1 h, and then centrifuged at 14,000 rpm for 5 min at 4°C. The labeled lysate was then added to the antibody-bound protein A beads and incubated with rotation for 1–2 h. The beads were washed four times with kinase bead buffer (500 mM NaCl, 50 mM Tris-Cl, pH 7.4, 50 mM NaF, 5 mM EGTA, 5 mM EDTA, 0.1% Triton X-100; transferring the beads to fresh tubes after the fourth wash), and then twice with 50 mM Tris-Cl, pH 7.5. The beads were then rotated in 50 mM Tris-Cl, pH 7.5, containing 0.5 mg/ml RNAse A for 30 min at 4°C, washed an additional two times in kinase bead buffer (transferring the beads to fresh tubes after the second wash), and then a final wash in 50 mM Tris-Cl, pH 7.5.

Cells were arrested in G1 by alpha factor, were harvested by centrifugation, frozen, and processed for immunoprecipitation. 10–15 mg of cell lysate was precleared in 50 μl protein A beads, and then the APC was immunoprecipitated with 2 μg anti-Cdc26 antibodies that were prebound to protein A beads as described above. After immunoprecipitation, the beads were washed three times in kinase bead buffer (transferring the beads to fresh tubes after the second wash), and then twice in low salt kinase buffer (10 mM NaCl, 20 mM Hepes-KOH, pH 7.4, 5 mM MgCl₂, 1 mM DTT). 5 ng of purified Cdc28-His₆, 50 ng purified Clb2-MBP (gifts of Jeff Ubersax, University of California, San Francisco, CA), and 100 ng purified Cks1 (see below) in 2 μl of kinase dilution buffer (300 mM NaCl, 25 mM Hepes-KOH, pH 7.4, 10% glycerol, 0.1 mg/ml BSA) were added to a 13 μl of low salt kinase buffer containing 10 μm ATP, 2 μCi γ[³²P]ATP (Amersham Pharmacia Biotech), and 10 μm okadaic acid (Calbiochem-Novabiochem). This reaction mix was added to the immunoprecipitated APC and incubated at 25°C for 20 min. The beads were then washed three times in kinase bead buffer containing 1 μm okadaic acid (transferring the beads to fresh tubes after the second wash), and then twice in low salt kinase buffer containing 1 μm okadaic acid. These washes remove Clb2-MBP and proteolytic fragments of Clb2-MBP, which are well phosphorylated and obscure APC phosphorylation. Cdc5 phosphorylation was performed by adding the following to immunoprecipitated APC: purified His₆-HA-Cdc5 (a gift of Julia Charles, University of California, San Francisco, CA) in 5 μl of Cdc5 storage buffer (250 mM KCl, 20 mM Hepes-KOH, pH 7.4, 10% glycerol, 5 mM NaF, 0.1 mg/ml BSA) added to 15 μl of Cdc5 kinase reaction buffer (20 mM KCl, 20 mM Hepes-KOH, pH 7.4, 2 mM MgCl₂, 2 mM MnCl₂; final concentrations in 20 μl reaction) containing 10 μm ATP, 2 μCi γ[³²P]ATP, and 10 μm okadaic acid.

Cks1 protein was made as described previously (Booher et al. 1993) using pCKS1-1. After the ammonium sulfate precipitation, the pellet was resuspended in lysis buffer (50 mM Tris-Cl, pH 8.0, 2 mM EDTA, 10% glycerol) and then desalted on a PD-10 column (Amersham Pharmacia Biotech) that had been equilibrated in CnBr coupling buffer (500 mM NaCl, 100 mM Na₂CO₃, pH 8.3). Cks1 was then coupled to CnBr-activated Sepharose 6MB or 4B (Amersham Pharmacia Biotech) according to the manufacturer's instructions. Beads were washed and stored in lysis buffer (100 mM NaCl, 50 mM Tris-Cl, pH 7.5, 50 mM NaF, 50 mM Na-β-glycerophosphate, pH 7.4, 2 mM EGTA, 2 mM EDTA, 0.1% Triton X-100, 0.02% NaN₃). 3–5 mg of cell lysate was incubated with 10 μl Cks1-coupled beads for 1–2 h, washed three times in kinase bead buffer (500 mM NaCl, 50 mM Tris-Cl, pH 7.4, 50 mM NaF, 5 mM EGTA, 5 mM EDTA, 0.1% Triton X-100; transferring the beads to fresh tubes after the second wash), and then twice in low salt kinase buffer. Phosphatase treatment of Cks1 bead pulldowns was performed as previously described (Hardwick and Murray 1995) using lambda phosphatase (New England Biolabs, Inc.).

Mutants that reduce mitotic Cdc28 activity have difficulty activating the Cdc20-dependent APC, suggesting that Cdc28 might phosphorylate the APC or Cdc20 (Rudner et al. 2000). Therefore, we asked if the budding yeast APC is phosphorylated in vitro. We used APC that was isolated by immunoprecipitating cell lysates with antibodies against Cdc26, a nonessential component of the APC (Hwang and Murray 1997), and used these immunoprecipitates as a substrate for purified recombinant Cdc28/Clb2/Cks1 (a gift of Jeff Ubersax and David Morgan, University of California, San Francisco, CA) in the presence of γ[³²P]ATP. In APC isolated from wild-type cells, three major bands and a single minor band were phosphorylated (Fig. 1, top). We determined the identity of these four bands by phosphorylating the APC isolated from cells containing epitope-tagged subunits that change their molecular weight. If the band shifted up in the epitope-tagged APC, we concluded that the phosphorylated protein is the APC subunit. By this criterion, the protein at 97 kD is Cdc16, the protein at 85 kD is Cdc27, and the minor species at 65 kD is Cdc23 (Fig. 1, top). Similar experiments have shown the band at 42 kD is Apc9 (data not shown).

We do not think the phosphorylation of the APC in these reactions is due to kinases that coimmunoprecipitate with the APC; no labeling is seen in immunoprecipitates lacking added Cdc28/Clb2/Cks1. However, a kinase bound to the APC might need to be activated by Cdc28, as has been reported for Plk phosphorylation of the mammalian APC (Kotani et al. 1998). Therefore, we tested whether Cdc5, the Plk homologue in budding yeast, was required for in vitro APC phosphorylation (Kitada et al. 1993). We isolated the APC from a cdc5-1 mutant that had been arrested in G1 by alpha factor at 25°C and then shifted to the restrictive temperature of 37°C for one hour. This APC is fully phosphorylated in vitro by Cdc28 (Fig. 1), showing that Cdc5 is not required for APC phosphorylation in this in vitro assay. In addition, Cdc5 is not detectable in alpha factor-arrested cells (Hardy and Pautz 1996; Charles et al. 1998; Shirayama et al. 1998), or in anti-Cdc26 immunoprecipitates of the APC, isolated from mitotic cells that contain Cdc5 (David Morgan, personal communication; and data not shown).

Is the APC phosphorylated in vivo? Wild-type cells were arrested at three points in the cell cycle: during G1 by adding alpha factor (a mating pheromone), during S-phase by adding HU (a DNA synthesis inhibitor), and in mitosis with nocodazole (a microtubule polymerization inhibitor). The arrested cells were labeled with ³²PO₄ and the APC was isolated by immunoprecipitating cell lysates with antibodies against Cdc26. Three major proteins of 97, 85, and 65 kD were strongly labeled in nocodazole-arrested cells, and to a lesser extent in HU- and alpha factor-arrested cells (Fig. 2 A). These three proteins do not precipitate from cdc26Δ cells. The molecular weights of these proteins suggest that they are the APC subunits Cdc16, Cdc27, and Cdc23, and mutating phosphorylation sites in these proteins abolishes in vivo phosphorylation of the APC (see below).

Since Cdc28/Clb complexes are inactive in G1, the differences in APC phosphorylation during different cell cycle stages suggests this reaction depends on Cdc28/Clb complexes. We tested this hypothesis directly by comparing the phosphorylation of the APC in CDC28-VF, clb2Δ, and cdc28-1N cells, three mutants that affect the mitotic activity of Cdc28 (Piggott et al. 1982; Surana et al. 1991; Grandin and Reed 1993; Rudner et al. 2000). The cells were arrested in metaphase at 25°C (Fig. 2 B) or at 35°C (Fig. 2 C) by overexpressing Mps1 from the galactose inducible GAL1 promoter, which activates the spindle checkpoint. All three mutants reduce the phosphorylation of the APC by a factor of 2–4 compared with wild-type.

Previous studies have suggested that in mammalian tissue culture cells, the protein kinase Plk is primarily responsible for phosphorylating the APC (Kotani et al. 1998). A mutant in CDC5, the yeast homologue of Plk, cannot activate the Hct1-dependent APC (Charles et al. 1998). To determine whether APC phosphorylation is dependent on Cdc5, we examined APC phosphorylation in a cdc5-1 mutant, arrested in metaphase by overexpressing Mps1 at a semirestrictive temperature of 35°C. We observed a similar reduction in APC phosphorylation as in CDC28-VF and cdc28-1N (Fig. 1 C), suggesting that Cdc5 contributes to APC phosphorylation in vivo.

To confirm the identities of the phosphorylated APC subunits and to determine if the APC is phosphorylated by Cdc28 in vivo, we mutated all the potential Cdc28 sites in Cdc16, Cdc23, and Cdc27. Using the weakest possible consensus phosphorylation site (serine or threonine, followed by proline; S/TP) as our criterion, we mutated six sites in Cdc16, one in Cdc23, and five in Cdc27. We refer to the resulting genes as CDC16-6A CDC23-A and CDC27-5A. As Fig. 3 A shows, most of the mutated sites fit only the minimal S/TP motif and lack a nearby basic residue found in many biochemically determined Cdk phosphorylation sites (Brown et al. 1999).

We directly assessed the ability of the mutant subunits to be phosphorylated in vivo and in vitro. In vivo, each alanine-substituted subunit is resistant to phosphorylation (Fig. 3 B). This result confirms our conclusion that Cdc16, Cdc27, and Cdc23 are the three major phosphorylated proteins in the APC (Fig. 2) and shows that among the phosphorylation sites we mutated are the relevant in vivo sites. In addition, the phosphorylation of the different subunits are largely independent of each other. For example, the CDC23-A mutant eliminates the phosphorylation of Cdc23, but not that of Cdc16 and Cdc27. In vitro, Cdc23-A and Cdc27-5A are resistant to phosphorylation in vitro by Cdc28 (Fig. 3 C). Cdc16-6A is still weakly phosphorylated, though much less than the wild-type protein.

During the course of this work we discovered that the budding yeast APC, like the animal APC, can bind to Cks1-coupled beads (Sudakin et al. 1997). This interaction is thought to be critical for APC phosphorylation and reflects the ability of Cks1 to bring Cdc2/Cyclin B complexes into proximity with the APC by interacting with both complexes simultaneously (Patra and Dunphy 1998; Shteinberg and Hershko 1999). Fig. 4 A shows that the APC from mitotically arrested yeast cells binds to Cks1-coupled beads. Comparing Western blots of the material recovered from wild-type and CDC28-VF cells reveals that less APC from CDC28-VF cells binds to Cks1-coupled beads. Reduced recovery of the APC does not reflect decreased binding of Cdc28-VF to Cks1 beads, since equal amounts of Cdc28-VF/Clb2 and Cdc28/Clb2 are recovered with the beads.

Fig. 4 A also shows that Cdc27 runs as a doublet on Western blots, with the upper band predominating in wild-type and the lower band predominating in CDC28-VF. The slower mobility form of Cdc27 is a phosphorylated form because it can be converted to the faster one by treating the Cks1-bound material with lambda phosphatase (Fig. 4 B). We are also able to detect phosphorylation-dependent mobility shifts in Cdc16 and Cdc23 (data not shown). These phosphorylation-dependent shifts confirm our in vivo labeling data (Fig. 2) that show the APC is phosphorylated in vivo.

To investigate the relationship between reduced Cks1 binding and reduced Cdc27 phosphorylation in CDC28-VF cells, we followed both through the cell cycle. Fig. 4 C shows that through most of the cell cycle Cdc27 is partially phosphorylated, but as cells go through mitosis and Clb2 levels peak, Cdc27 and Cdc16 phosphorylation increases (Fig. 4 C, arrow). The amount of phosphorylation on both subunits increases further when nocodazole treatment arrests cells in mitosis by activating the spindle checkpoint. In CDC28-VF cells, as Clb2 levels increase, Cdc27 phosphorylation decreases before eventually returning to its partially phosphorylated state (Fig. 4 C, bracket).

These changes in APC phosphorylation correlate with its ability to bind Cks1-coupled beads (Fig. 4 C, bottom). In wild-type, no APC binds Cks1-coupled beads in an alpha factor arrest, and its binding increases as levels of Clb2 rise. In CDC28-VF cells, little binding of the APC to Cks1-coupled beads is seen at any stage of the cell cycle, even though the peak levels of Clb2 are similar in wild-type and CDC28-VF cells. These differences suggest that mitotic phosphorylation by Cdc28 is required for APC binding to Cks1-coupled beads (Sudakin et al. 1997). Although the difference in APC phosphorylation between wild-type and CDC28-VF cells in a synchronous cell cycle is transient and subtle, it is reproducible, and it correlates with a large difference in the ability of the APC to bind Cks1-coupled beads.

We next tested whether APC phosphorylation is required for the APC to bind Cks1-coupled beads. The beads do not bind an APC containing Cdc27-5A (Fig. 4 D), but do bind an APC containing either Cdc16-6A and Cdc23-A. This result suggests that phosphorylation of Cdc27 is critical for Cks1 binding to the APC.

We have shown that mutations that alter the mitotic activity of Cdc28 have reduced APC phosphorylation (Fig. 2B and Fig. C). Our ability to detect APC phosphorylation on Western blots allowed us to examine additional mutants that affect Cdc28 activity. Mutants that reduce the mitotic activity of Cdc28 (CDC28-VF, cdc28-1N, clb2Δ, and cks1-38) are hypersensitive to checkpoint arrest caused by overexpression of Mps1, whereas a mutant that primarily affects G1 activity (cdc28-4) is not (Reed 1980; Surana et al. 1991; Tang and Reed 1993; Rudner et al. 2000). To test if this correlation extended to the phosphorylation state of the APC, we arrested these strains in mitosis with nocodazole and immunoblotted for Cdc27 and Cdc16. This analysis correlates perfectly with our earlier findings: cdc28-4 have normal levels of Cdc16 and Cdc27 phosphorylation, whereas clb2Δ, cdc28-1N, and cks1-38 all have reduced levels and resemble CDC28-VF (Fig. 4 E).

We wanted to rule out the possibility that the phosphorylation site mutants had general effects on the activity of the APC, as opposed to a specific effect on its mitotic, Cdc20-dependent form. Since the Hct1-dependent APC is maximally active when Cdc28 is inactive, loss of Cdc28-dependent phosphorylations should not affect Hct1-dependent APC activity in G1-arrested cells that lack active Cdc28 (Zachariae et al. 1998; Jaspersen et al. 1999). Therefore, we examined the activity of APC containing the alanine-substituted subunits that had been isolated from G1-arrested cells. APC activity was measured in an in vitro ubiquitination assay that uses an iodinated fragment of sea urchin cyclin B as a substrate and the APC provided from anti-Cdc26 immunoprecipitates (Charles et al. 1998). Fig. 5 shows that there is no difference in G1-specific APC activity between wild-type cells and those carrying alanine mutations in APC subunits (CDC16-6A, CDC23-A, or CDC27-5A). Thus, the mutations in putative Cdc28-dependent phosphorylation sites have not disrupted the ability of these subunits to associate with other APC components or produce normal levels of Hct1-dependent APC activity. In addition, cells carrying alanine-substituted APC subunits show no obvious growth defects at any temperature.

We asked if the alanine substitutions in the APC, like CDC28-VF, have difficulty leaving mitosis (Rudner et al. 2000). Wild-type and CDC16-6A CDC23-A CDC27-5A cells were arrested in G1 by the mating pheromone alpha factor and then released into the cell cycle. Once cells had budded, alpha factor was readded to arrest cells that had completed the cycle. CDC16-6A CDC23-A CDC27-5A cells show a 20-min delay in sister chromatid separation (Fig. 6 A). Clb2 and Clb3 proteolysis are delayed by >40 min. This defect is not due to slower mitotic entry, because wild-type and CDC16-6A CDC23-A CDC27-5A cells initiate budding, degrade Sic1, and form a short mitotic spindle at the same time (Fig. 6 A and data not shown).

Mutating APC phosphorylation sites also causes an increased sensitivity to spindle checkpoint arrest caused by Mps1 overexpression (Hardwick et al. 1996; Rudner et al. 2000). Serial dilutions of wild-type, CDC28-VF, mutants in single APC subunits, double mutants, and the triple mutant were spotted on plates where Mps1 is induced to high levels (Fig. 6 B). Both CDC16-6A and CDC27-5A are sensitive to Mps1 overexpression and combining the two mutants creates a phenotype similar to that of CDC28-VF. The CDC23-A mutation alone has little phenotype, but exacerbates the effect of both the CDC16-6A and CDC27-5A mutations. These data suggest that phosphorylation of the APC subunits contribute to the ability to overcome the spindle checkpoint and suggest that the alanine-substituted APC, like CDC28-VF, may be defective in the Cdc20-dependent APC.

To test Cdc20-dependent APC function more directly, we examined the ability of these nonphosphorylatable APC mutants to support Pds1 degradation in vivo. Pds1 is normally unstable in anaphase with a half life of about ten minutes (Jaspersen et al. 1998). We arrested wild-type and nonphosphorylatable APC mutants in anaphase (using the cdc15-2 mutant), induced Pds1 expression from the GAL1 promoter by adding galactose for one hour, and then shut the promoter off by adding glucose and examined the rate of Pds1 degradation. Previously, we have shown that in this anaphase arrest, CDC28-VF and clb2Δ stabilize Pds1 (Rudner et al. 2000). We see a similar effect when the CDC27-5A and CDC16-6A mutants are combined with cdc15-2. The half life of Pds1 is increased to 30 min in anaphase-arrested CDC27-5A, and to >90 min in CDC16-6A cells (Fig. 7 A).

We also have examined the association of Cdc20 with the APC in the alanine-substituted mutants at the cdc15-2 block, a time when the Cdc20-dependent APC is active. This association is impaired in CDC28-VF (Rudner et al. 2000). We arrested cdc15-2, cdc15-2 CDC16-6A, and cdc15-2 CDC27-5A cells in anaphase, immunoprecipitated the APC with anti-Cdc26 antibodies, and examined the amount of associated Cdc20. The association of Cdc20 to the APC is reduced in CDC27-5A cells and severely impaired in CDC16-6A cells (Fig. 7 B).

Different groups debate whether Cdc2 (Cdk1) or Plk is the major APC kinase in vivo (Lahav-Baratz et al. 1995; Kotani et al. 1998; Patra and Dunphy 1998; Shteinberg and Hershko 1999; Shteinberg et al. 1999). Since we find that APC phosphorylation is reduced in a cdc5-1 mutant (Fig. 2 C), we tested whether purified recombinant Cdc5 could phosphorylate the APC in vitro. Like Cdc28, Cdc5 phosphorylates Cdc16, Cdc27, and Apc9 (Fig. 8 A), but unlike Cdc28, appears to not phosphorylate Cdc23. Cdc5 also phosphorylates proteins that have the molecular weights of several other APC subunits (Apc1, -2, -4, and -5; Zachariae et al. 1996). We have not confirmed the identity of these proteins because there is little evidence that these proteins are major targets of phosphorylation in vivo (Fig. 2). The ability of human Plk to activate the APC in vitro depends on pretreatment with Cdc2/Cyclin B complexes (Kotani et al. 1998), a result that has been interpreted to suggest that Cdc2 activates Plk's kinase activity against the APC. In our hands, however, Cdc5's ability to phosphorylate the APC does not increase when the kinase is pretreated with purified Cdc28/Clb2/Cks1 complexes (data not shown).

We next tested whether purified Cdc5 can phosphorylate the alanine-substituted APC. The Cdk sites we mutated on Cdc16, Cdc23, and Cdc27 are also potential sites of phosphorylation by Cdc5. Substrates of the frog homologue of Cdc5, Plx1, become epitopes for the MPM-2 antibody after phosphorylation by Plx1 (Kumagai and Dunphy 1996) and MPM-2 recognizes phosphorylation at SP or TP sites (Westendorf et al. 1994). In contrast to their effect on phosphorylation by Cdc28, the APC phosphorylation site mutants had no effect on in vitro phosphorylation of the APC by recombinant Cdc5 (Fig. 8 B). This observation makes it likely that the reduced in vivo APC phosphorylation seen in cdc5-1 cells is an indirect effect of reduced Cdc5 activity, rather than a direct in vivo phosphorylation of these APC subunits by Cdc5.

We have shown that the budding yeast APC subunits, Cdc16, Cdc23, and Cdc27, are phosphorylated in vivo and in vitro. Phosphorylation in vivo depends on Cdc28, and in vitro it is catalyzed by pure Cdc28/Clb2/Cks1 complexes. Mutating potential Cdc28 phosphorylation sites in Cdc16, Cdc23, and Cdc27 abolishes their in vivo phosphorylation and compromises the mitotic, but not the G1 functions of the APC. We have also shown that Cdc5 affects APC phosphorylation in vivo and can catalyze APC phosphorylation in vitro. Our analysis of APC phosphorylation site mutants in vivo and in vitro, however, argues that in vivo Cdc5 indirectly induces the phosphorylation of Cdc16, Cdc23, or Cdc27, rather than directly modifying these subunits.

Our results agree with studies on other organisms that show mitosis-specific APC phosphorylation. Cdc16, Cdc23, Cdc27, and Apc1 are phosphorylated in frogs; Apc1, Cdc16, and Cdc27 are phosphorylated in mammalian tissue culture cells; and Cdc16 (Cut 9) is phosphorylated in fission yeast (Peters et al. 1996; Yamada et al. 1997; Patra and Dunphy 1998; Kotani et al. 1999). Although APC phosphorylation has been shown to activate the Cdc20-dependent APC in mammalian tissue culture and clam egg extracts (Kotani et al. 1998; Shteinberg et al. 1999), and Cks1 depletions prevent mitotic exit in frog extracts (Patra and Dunphy 1996), this is the first report to examine the in vivo function of APC phosphorylation. Although phosphorylation of Cdc16, Cdc23, and Cdc27 is not essential for viability in budding yeast, our studies suggest that it stimulates Cdc20-dependent APC activity and Cdc20 binding to the APC in vivo.

Cdc27 remains partially phosphorylated in G1 cells (Fig. 4 C). The presence of slower migrating Cdc27 in G1 cells could arise two ways: during the exit from mitosis, if Cdc28-catalyzed phosphorylation declines after phosphatases have been inactivated; or in G1, by phosphorylation catalyzed by another kinase. Because Cdc27-5A runs as a single band on Western blots in G1 (Fig. 5), we favor the possibility that G1 phosphorylation on Cdc27 remains from the previous mitosis. This finding would suggest that the phosphatase that removes phosphorylation from the APC is only active in mitosis. PP2A has been proposed to play such a role in clams and frogs (Lahav-Baratz et al. 1995; Vorlaufer and Peters 1998).

In one report, Plk has been identified as the major kinase of the mammalian APC (Kotani et al. 1998), although others have argued that this role is played by Cdc2 (Lahav-Baratz et al. 1995; Patra and Dunphy 1998; Shteinberg and Hershko 1999; Shteinberg et al. 1999). We asked if its budding yeast homologue, Cdc5, plays a similar role. In vivo, APC phosphorylation is reduced in the cdc5-1 mutant and purified Cdc5 phosphorylates the APC in vitro (Fig. 2 C and 8 A). Three observations argue that in living cells Cdc5 does not directly phosphorylate the APC subunits we have studied: phosphorylation site mutations that completely block phosphorylation of Cdc16, Cdc23, and Cdc27 in vivo, do not block in vitro phosphorylation of these subunits by Cdc5 (Fig. 8 B); purified Cdc28/Clb2/Cks1, that lacks detectable Cdc5, efficiently phosphorylates immunoprecipitated APC (Fig. 1); and the same mutations that block Cdc28-catalyzed phosphorylation in vitro also block in vivo APC phosphorylation (Fig. 3).

If Cdc5 does not phosphorylate Cdc16, Cdc23, and Cdc27 directly, how does it regulate the phosphorylation of these subunits? Cdc5 may be responsible for phosphorylating other APC subunits (Apc1, -2, -4, and -5 are potential substrates; Fig. 8) in vivo, and the phosphorylation of these subunits may affect the phosphorylation of the Cdc28 targets Cdc16, Cdc23, and Cdc27. Alternatively, Cdc5 may modulate Cdc28/Clb/Cks1 activity or localization.

We have shown that phosphorylation site mutants in the APC reduce activation of the Cdc20-dependent APC. The half-life of Pds1 is increased in CDC27-5A and CDC16-6A cells, and this defect in Cdc20 function could be explained by the observed inability of Cdc20 to bind an APC containing Cdc16-6A. This data supports genetic experiments showing that reduced mitotic Cdc28 activity compromises the Cdc20-dependent APC (Rudner et al. 2000).

If Cdc20 binding and activity depend on a phosphorylated APC, why is the triple mutant CDC16-6A CDC23-A CDC27-5A viable? Even in the triple mutant there is some residual Cdc20 binding to the APC (data not shown), which is presumably sufficient to drive the metaphase to anaphase transition. The residual binding of Cdc20 to the APC could depend on the phosphorylation of the other subunits. In support of this idea, we see weak phosphorylation of proteins we believe to be Apc1, -4, -5, and -9 in some in vivo labelings (data not shown), and a protein we believe to be Apc9 is phosphorylated in vitro by Cdc28/Clb2/Cks1 complexes (data not shown and Fig. 1; Zachariae et al. 1996). In addition, cdc28-1N, a mutation in Cdc28 that cannot bind Cks1 (Kaiser et al. 1999; and data not shown) and cks1-38, have reduced APC phosphorylation (Fig. 1 C and 4 E). These two mutants are temperature-sensitive for growth and arrest in mitosis (Piggott et al. 1982; Tang and Reed 1993), suggesting that APC phosphorylation may be essential. Alternatively, it has been proposed that the primary defect in cdc28-1N and cks1-38 is in proteasome function (Kaiser et al. 1999), though proteasome activity was examined in G1, not in mitosis, leaving the relevance of this finding to the exit from mitosis uncertain.

Our data suggests that activation of the APC by phosphorylation opposes its inhibition by the spindle checkpoint. Although CDC16-6A CDC23-A CDC27-5A is viable, its delay in mitosis (Fig. 6 A) becomes lethal when the spindle checkpoint is activated (Fig. 6 B). Both APC phosphorylation and the spindle checkpoint affect the ability of Cdc20 to activate the APC, but have no effects on the G1, Hct1-dependent activity of the APC.

Phosphorylation of the APC in frogs and clams in vitro depends on homologues of the small Cdk binding protein, Cks1, and in clams, Cks1 stimulates Cdc20-dependent APC activity (Patra and Dunphy 1998; Shteinberg and Hershko 1999; Shteinberg et al. 1999). In budding yeast, the role of Cks1 remains uncertain. Although we add purified Cks1 to our in vitro kinase reactions, Cks1 is not required for APC phosphorylation in vitro (Fig. 1 and data not shown). However, APC phosphorylation in vivo clearly depends on Cks1 (Fig. 4 E) and phosphorylation of Cdc27 is required for APC binding to Cks1-coupled beads (Fig. 4 D). We do not think that the binding of the APC to Cks1-coupled beads correlates with the ability of Cdc28 to phosphorylate the APC in vivo: although an APC containing Cdc27-5A does not bind to Cks1-coupled beads, Cdc16 and Cdc23 are fully phosphorylated in a CDC27-5A mutant. Despite this in vivo finding, we do see reduced in vitro phosphorylation of Cdc16 and Cdc23 in an APC containing Cdc27-5A (Fig. 3 C).

Mutants that affect the mitotic form of Cdc28 have reduced levels of phosphorylation of the APC, whereas cdc28-4 cells, which are defective in the G1 form of Cdc28 (Reed 1980), show normal phosphorylation of Cdc27 and Cdc16. We were surprised to discover that APC phosphorylation in cdc28-4 appears to be normal (Fig. 4 E), because this mutant has ∼20% the amount of Cdc28 protein as wild-type cells at the permissive temperature of 23°C, and very little detectable Cdc28-associated kinase activity when immunoprecipitated from cell lysates (Surana et al. 1991; and data not shown). A possible explanation of the absence of mitotic defects in cdc28-4 cells is that the specific activity of each Cdc28-4 molecule is equal to that of wild-type Cdc28, although the total number of active kinases is drastically reduced. The specific activity of individual Cdc28 molecules may be critical for APC phosphorylation because one Cdc28/Clb2/Cks1 complex may bind persistently to the APC. Once bound to the APC, this single complex might be responsible for multiple phosphorylations. If the steady state phosphorylation of the APC is determined by the balance between phosphorylation by Cdc28 and dephosphorylation by protein phosphatases, and Cdc28 remains bound to the APC, a drop in specific activity of Cdc28 would reduce the phosphorylation and activity of the APC.

How do cells escape from mitosis? If activating Cdc28/Clb/Cks1 complexes activates the Cdc20-dependent APC, which in turn triggers chromosome segregation, how do cells ensure that the lag between activating Cdc28 and activating the Cdc20-dependent APC is long enough to assemble a spindle and align chromosomes on it? Although one answer is that the spindle checkpoint inhibits Cdc20 in cells with misaligned chromosomes (Hwang et al. 1998; Kim et al. 1998), this explanation is not enough. Inactivating the spindle checkpoint does not kill yeast cells, implying other mechanisms exist to block premature activation of the Cdc20-dependent APC. Another possible mechanism is regulating the abundance of Cdc20. High levels of CDC20 transcripts are restricted to mitotic cells and APC-dependent proteolysis restricts the abundance of Cdc20 (Weinstein 1997; Kramer et al. 1998; Prinz et al. 1998; Shirayama et al. 1998). None of these forms of regulation exist in early frog embryos, where Cdc20 (Fizzy) levels are constant through the cell cycle and spindle depolymerization does not induce mitotic arrest (Minshull et al. 1994; Lorca et al. 1998). In addition, overexpressing Cdc20 in budding yeast raises the level of Cdc20 mRNA and protein, but does not advance the exit from mitosis, suggesting that other mechanisms must exist to regulate Cdc20-dependent APC activity (Prinz et al. 1998). Regulating the rate of Cdc28-catalyzed APC phosphorylation provides an additional mechanism. If this phosphorylation were slow relative to spindle assembly, most cells would manage to align their chromosomes on the spindle before activating the Cdc20-dependent APC, which in turn induces Pds1 destruction and anaphase.

Note Added in Proof. Similar results showing that Cdc20 only binds to a phosphorylated APC have been published recently (Kramer, E.R., N. Scheuringer, V. Podrelejnikov, M. Mann, and J.M. Peters. 2000. Mol. Biol. Cell. 11:1555–1569).

We would like to thank Julia Charles, Jeff Ubersax, and David Morgan for sharing unpublished results and reagents; Doug Kellogg, Lena Hwang, Sue Jaspersen, Rachel Tinker-Kulberg, Dave Morgan, Mike Mendenhall, Ray Deshaies, Bob Booher, Andrew Page, Phil Hieter, Wolfgang Zacharaie, Kim Nasmyth, Jeremy Thorner, and Steve Reed for yeast strains, plasmids, and antibodies; Jeff Ubersax, Sue Jaspersen, Dave Morgan, and the Murray Lab for critical reading of the manuscript; Bodo Stern, Alex Szidon, Julia Charles, Rachel Tinker-Kulberg, Hironori Funabiki, Sue Biggins, and Dara Spatz Friedman for invaluable discussions and extraordinary support.

This work was supported by grants from the National Institutes of Health and Human Frontiers in Science Program to A.W. Murray. A.D. Rudner was a pre-doctoral fellow of the Howard Hughes Medical Institute during this work.