Cdc28 Activates Exit from Mitosis in Budding Yeast

To exit mitosis, cells must accomplish two tasks: chromosome segregation and inactivation of complexes between mitotic cyclins and cyclin-dependent kinase 1 (Cdk1), known as Cdc28 in budding yeast and Cdc2 in other eukaryotes, which allows the cell cycle to progress into G1. Both steps require the activity of the anaphase-promoting complex (APC) or cyclosome, a multiprotein complex that is required for the ubiquitination of cyclin and other unstable substrates (King et al. 1995; Sudakin et al. 1995; Zachariae and Nasmyth 1996, Zachariae and Nasmyth 1999; Yanagida et al. 1999). The activity of the APC depends on its interaction with two WD-40 proteins, Cdc20 (Sethi et al. 1991; Sigrist et al. 1995; Visintin et al. 1997; Lorca et al. 1998) and Hct1 (Schwab et al. 1997; Sigrist and Lehner 1997; Visintin et al. 1997; Yamaguchi et al. 1997; Kitamura et al. 1998). Cdc20 initiates the metaphase to anaphase transition by inducing ubiquitination of the anaphase inhibitor, Pds1 (Cohen-Fix et al. 1996; Funabiki et al. 1997). This reaction causes Pds1 degradation and sister chromatid separation (Funabiki et al. 1996; Ciosk et al. 1998). Cdc20-dependent APC activity is inhibited by the spindle checkpoint, which senses defects in chromosome attachment to the spindle and delays the onset of anaphase until these defects are corrected (Rieder et al. 1995; Fang et al. 1998a; Hardwick 1998; Hwang et al. 1998; Kim et al. 1998).

The second WD-40 protein, Hct1, also activates the APC (Schwab et al. 1997; Sigrist and Lehner 1997; Visintin et al. 1997; Kramer et al. 1998). In budding yeast, Hct1-dependent APC activity is necessary for the ubiquitination and degradation of Clb2, the major mitotic cyclin, which causes the sudden drop in the protein kinase activity of Cdc28 at the end of mitosis and keeps the APC active throughout G1 (Zachariae et al. 1998; Jaspersen et al. 1999). The activation of the Hct1-dependent APC depends on the prior activation of the Cdc20/APC, and this dependency helps ensure that the events of mitosis occur in the proper sequence (Lim et al. 1998; Visintin et al. 1998; Shirayama et al. 1999; Yeong et al. 2000).

In budding yeast, active Cdc28 inhibits the transition from anaphase to G1 (Amon 1997; Li and Cai 1997). One of the primary targets that Cdc28/Clb2 complexes inhibit is Hct1. Phosphorylation of Hct1 by Cdc28/Clb complexes prevents it from binding to and activating the APC (Zachariae et al. 1998; Jaspersen et al. 1999). This inhibition is opposed by the phosphatase Cdc14, which de-phosphorylates Hct1, initiating a positive feedback loop that drives the cell into G1 (Visintin et al. 1998; Jaspersen et al. 1999). As Hct1 activity rises, the rate of Clb destruction increases, reducing the kinase activity of Cdc28 and further activating Hct1. The CDK inhibitor, Sic1 (Mendenhall 1993; Donovan et al. 1994), which inhibits Cdc28/Clb complexes, also participates in this feedback loop, since both its transcription and stability are inhibited by Cdc28/Clb activity (Amon 1997; Toyn et al. 1997; Visintin et al. 1998).

Does Cdc28 also play a role in inducing anaphase? In frog and clam egg extracts, activation of the APC depends on active Cdc2/Cyclin B complexes (Felix et al. 1990; Hershko et al. 1994; Minshull et al. 1994; Shteinberg and Hershko 1999). This activating role might be explained by the finding that Cdc2/Cyclin B can phosphorylate subunits of the APC (Patra and Dunphy 1998; Kotani et al. 1999), and this phosphorylation is correlated with activating the APC and binding of Fizzy, the Cdc20 homologue in frogs (Peters et al. 1996; Fang et al. 1998b; Kotani et al. 1998; Shteinberg et al. 1999). In embryonic cell cycles, no Hct1 homologue is present (Sigrist and Lehner 1997; Lorca et al. 1998), suggesting that Cdc20-dependent APC activity targets both Pds1 homologues and mitotic cyclins for destruction. In budding yeast, the inability of Cdc20 to catalyze the complete destruction of mitotic cyclins means that Hct1 or Sic1 is required for cells to exit from mitosis (Schwab et al. 1997; Visintin et al. 1997). Since both Hct1 and Sic1 are inhibited by Cdc28, it has been difficult to ask whether a previous step in the exit from mitosis requires Cdc28.

Cdc28/Clb activity is also regulated by inhibitory phosphorylation on residues tyrosine 19, and possibly threonine 18 (Amon et al. 1992; Sorger and Murray 1992; Booher et al. 1993). Tyrosine 19 is phosphorylated by Swe1 (the homologue of Wee1 in fission yeast) and dephosphorylated by Mih1 (the homologue of Cdc25 in fission yeast; Russell et al. 1989; Booher et al. 1993). Phosphorylation on the homologous sites of fission yeast and vertebrate Cdc2 controls the timing of entry into mitosis and also can be induced by the checkpoints that detect unreplicated or damaged DNA (Gould and Nurse 1989; Enoch and Nurse 1990; Norbury et al. 1991; Jin et al. 1996; Rhind et al. 1997). In budding yeast, tyrosine 19 is phosphorylated during S-phase, but CDC28-T18A, Y19F (CDC28-AF) cells respond normally to the DNA damage checkpoint (Amon et al. 1992; Sorger and Murray 1992). Inhibitory phosphorylation of Cdc28 is required for the bud emergence or morphogenesis checkpoint, which delays mitosis in cells that have not budded (Lew and Reed 1993).

Our previous work suggested that inhibitory phosphorylation of Cdc28 might aid in mitotic exit. cdc55Δ cells, which lack a B subunit of protein phosphatase 2A (PP2A; Healy et al. 1991), are spindle checkpoint defective and have increased phosphorylation on tyrosine 19 of Cdc28 (Minshull et al. 1996; Wang and Burke 1997). The premature exit from mitosis in cdc55Δ cells with damaged spindles is suppressed by CDC28-T18V, Y19F (CDC28-VF), suggesting that inhibitory phosphorylation of Cdc28 might work in concert with Hct1 and Sic1 to reduce Cdc28-associated activity at the end of mitosis.

Here, we show that this hypothesis is incorrect. Although CDC28-VF has defects in leaving mitosis, these are not due to a lack of inhibitory phosphorylation, but reflect a second defect of CDC28-VF. This defect slows the normal activation of the Cdc20-dependent APC and reveals that Cdc28-associated activity is essential for the activation of the APC in mitosis.

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 were confirmed by PCR or by mutant phenotype. The sequences of all oligonucleotide primers used in this study are available upon request. The strains TG1 and DH5α were used for all bacterial manipulations.

The strains used for the crosses in Table were: JM434 (CDC28-VF), JM469 (CDC28-V), JM467 (CDC28-F), ADR1541 (CDC28-AF), ADR2035 (CDC28-F88G), KH208 (swe1Δ), ADR484 (cdc28-1N), ADR840 (cdc28-4), ADR314 (clb2Δ), ADR719 (cdc23-1), ADR1147 (cdc16-1), LH226 (doc1-1), LH125 (cdc20-1), ADR1435 (hct1Δ), JC126 (cdc5-1), ADR1298 (cdc13-1), and K1993 (cdc15-2). For all crosses, at least 22 tetrads were analyzed.

BAR1 was deleted using pJGsst1 (a gift of Jeremy Thorner, University of California, Berkeley, CA). PDS1-myc18 strains were made using a PDS1-myc18 replacement plasmid (Shirayama et al. 1998). pCUP-GFP12-lacI12, lacO:TRP1 and lacO:LEU2 were integrated using pSB116 (Biggins et al. 1999), pAFS52, and pAFS59 (Straight et al. 1996), respectively. GAL-MPS1 strains were made with pAFS120 (Hardwick et al. 1996). MAD3 was deleted using pKH181 (Hardwick et al. 2000). swe1Δ strains were made by crossing JM449 (a gift of Jeremy Minshull, Maxygen, Redwood City, CA) to the appropriate strains. MIH1 was deleted using pIP33 (a gift of Peter Sorger, Massachusetts Institute of Technology, Cambridge, MA). CDC55 was deleted using pJM6 (Minshull et al. 1996). clb2Δ strains were made by crossing K1890 (Surana et al. 1991) to the appropriate strains. CDC28-F88G strains were made by crossing JAU02 (a gift of Jeff Ubersax, University of California, San Francisco, CA) to the appropriate strains. The 2μ-CDC28 plasmid is EF190 (a gift of Eric Foss, Fred Hutchinson Cancer Research Institute, Seattle, WA) and is the CDC28 gene cloned into pRS425 (Sikorski and Hieter 1989). pHIS3-GFP-TUB1 was integrated using pAFS71 (a gift of Aaron Straight, Harvard Medical School, Boston, MA). cdc20-3 strains were made by crossing K8029 (Shirayama et al. 1998) to the appropriate strains. pGAL-PDS1-HA strains were made by crossing RTK43 (Jaspersen et al. 1998) to the appropriate strains.

Mutants in CDC28-VF-HA, CDC28-F-HA, CDC28-V-HA, and CDC28-HA were made as described previously (Booher et al. 1993). CDC28-AF-HA was made by cloning the HindIII/XhoI fragment from pSF35 (Peter Sorger, MIT, Cambridge, MA) into pRD96 (a gift of Ray Deshaies, California Institute of Technology, Pasadena, CA) cut with HindIII/XhoI to create pAR155, which was used to create a CDC28-AF-HA strain as previously described for pRD96 (Booher et al. 1993). The pGAL-CDC28 plasmids were made as follows: full-length CDC28-HA (pAR106) was made by cutting out the 1.1-kb XhoI-HindIII fragment of CDC28-VF-HA from pRD96 and replacing it with the same fragment from pRD47. The BstBI-BamHI fragment of pRD96 or pAR106 was then cloned into pDK20 (a gift of Doug Kellogg, University of California, Santa Cruz, CA) cut with SmaI and BamHI, to create pAR109 (yIP-pGAL-CDC28:URA3) and pAR108 (yIP-pGAL-CDC28-VF:URA3). The plasmids were cut with Stu1 and integrated at the URA3 locus.

pGAL-CLB2-Δ176 strains were made with pAR39. An EcoRI/BamHI fragment of CLB2 that lacks the first 176 amino acids was amplified by PCR and cloned into pDK20 cut with EcoRI and BamHI to create pAR39. The plasmid was cut with StuI for integration at the URA3 locus.

cks1-38 strains are cks1Δ::KAN^R covered by cks1-38 integrated at TRP1 (pAR183). CKS1 was deleted by the PCR-targeting method. Diploid cells were transformed with a cassette containing the bacterial KAN^Rgene, which confers G418 resistance in W303. The cassette was amplified by PCR from pFA6a-kanMX6 (Longtine et al. 1998) with primers that contain the sequences that flank the CKS1 open reading frame. pAR183 (cks1-38:TRP) was constructed by cloning an BamHI/SphI fragment from pSE271 (Tang and Reed 1993) into YIplac204 (Gietz and Sugino 1988), the resulting plasmid was cut with EcoRV and integrated into the CKS1/cks1Δ::KAN^R heterozygote at the TRP1 locus. The diploid was sporulated, and a resulting spore, ADR1767, which is temperature sensitive, TRP⁺, and KAN^R was used to create the strains used in this study.

HCT1 was deleted using pAR127. An EcoRI/HindIII fragment of the HCT1 locus was amplified by PCR and cloned into pSK(−) (Stratagene) to create pAR125. An XbaI/SmaI fragment of the HIS3 gene was then cloned into pAR125 cut with SpeI/XmnI, to create pAR127, which replaces the entire HCT1 open reading frame with HIS3. The EcoRI/NotI fragment of pAR127 was used to transform yeast.

CDC20 was tagged at the NH₂ terminus. A cassette containing 12 myc tags (pLH71) was inserted into pCM4 (Hwang et al. 1998) cut with BsteII, to create pLH83. The resultant plasmid was cut with EcoNI and BglII, removing the CEN-ARS sequences, blunted, and religated to create pLH92. This plasmid was cut with BspEI for integration at the CDC20 locus, creating a duplication of CDC20 and CDC20-12myc marked with the URA3 gene. This strain was grown on 5-fluoroorotic acid to select against the URA3 gene, and the resulting colonies were screened by Western blot for the presence of CDC20-12myc to create LHH371. This strain was crossed to the appropriate strains to create those used in this study.

Stock solutions of inhibitors were: 10 mg/ml alpha factor (Biosynthesis); 10 mg/ml nocodazole (Sigma-Aldrich). All stocks were stored at −20°C in DMSO.

For microcolony assays, cells were grown to mid-log phase in yeast extract and peptone (YEP) + 2% raffinose, spotted onto a YEP + 2% galactose plate, and unbudded cells were picked out into a grid with a dissection needle. The number of cells in each microcolony was counted at different times after incubation at 30°C. Each bud is counted as a cell, and the original cells that did not bud are not included in the analysis.

For serial dilution and spotting, cells were prepared in a multiwell dish or a microtiter plate, and using a multiprong applicator, ∼10 μl of each strain and its dilutions were spotted onto various plates and incubated at either 23 or 30°C.

Sister chromatid separation was visualized using a fusion of green fluorescent protein (GFP) and lacI bound to repeats of lacO, which had been integrated at specific locations on yeast chromosomes (Straight et al. 1996). Small samples of cells were harvested at the indicated times and either scored live, or fixed for 10 min in 3.7% paraformaldehyde at 4°C, and then washed twice in 0.1 M KPO₄ pH 7.4. The cells were then mounted on slides and viewed by fluorescence microscopy (Nikon). In all experiments, a minimum of 200 cells were counted per time point.

The anaphase arrest of pGAL-CLB2-Δ176 was visualized using pHIS3-GFP-TUB1. GFP-Tub1 was induced by transferring cells to complete synthetic medium (CSM)-His + 2% galactose + 10 mM 3-aminotriazole for the first 30 min after alpha factor removal. Cells were then transferred into YEP + 2% galactose. Spindle length was scored in living cells by fluorescence microscopy. The anaphase arrest of cdc15-2 was scored by fixing cells for 5 min in 70% ethanol, washing in 50 mM Tris-Cl, pH 7.5, and then resuspending in 50 mM Tris-Cl, pH 7.4, + 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes). Cells were then mounted on slides and viewed by fluorescence microscopy.

In experiments where only Western blots are shown, yeast extracts were prepared by bead beating (multitube bead beater; Biospec) frozen cell pellets in 1× SDS sample buffer (2% SDS, 80 mM Tris-Cl, pH 6.8, 10% glycerol, 10 mM EDTA, 0.02% bromophenol blue, 1 mM Na₃VO₄, 1 mM PMSF, and 5 mM NaF) and an equal volume of acid washed glass beads (Biospec) for 1 pulse of 90 s. Samples were then normalized based on OD₆₀₀ readings of the original yeast samples taken during the time course. We have found that this method allows even loading of the samples and works as well as other techniques.

Yeast extracts for immunoprecipitation and Western blots were made by bead beating frozen cell pellets 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, 1 mM Na₃VO₄, 1 mM PMSF, and leupeptin, pepstatin and chymostatin all at 1 μg/ml) and an excess of acid washed glass beads for 2 pulses of 90 s, incubating on ice for 5 min between each pulse. The resulting lysate was separated from the glass beads and centrifuged in a microfuge at 14,000 rpm for 5 min to remove insoluble material. Protein concentration of each lysate was determined using Bradford reagent (0.04% Coomassie blue G-250 dissolved in 4.75% ethanol, then mixed with 8.5% o-phosphoric acid and H₂O) and samples were normalized based on these measurements. A portion of the lysate was mixed with an equal volume of 2× SDS sample buffer. Standard methods were used for PAGE and protein transfer to nitrocellulose (Schleicher and Schuell; Minshull et al. 1996). Blots were stained with Ponceau S to confirm transfer and equal loading of samples, and then blocked for 30 min in antibody-specific blocking buffer (see below). All antibodies were incubated overnight at 4°C or 2 h at 25°C. After washing in PBST (PBS+ 0.1% Tween-20; Maniatis et al. 1982), the blots were then incubated in HRP-conjugated anti-rabbit or anti-mouse antibodies (Amersham Pharmacia Biotech) at a 1:5,000 dilution in PBST for 30 min at 25°C, washed again, incubated in ECL detection reagents (Amersham Pharmacia Biotech) or Renaissance reagents (NEN Life Science Products), using the manufacturer's instructions, and then exposed to X-OMAT film (Kodak).

The following antibodies were used in Western blots: 9E10 ascites (BabCO) was used at a dilution of 1:1,000 in PBST + 0.02% NaN₃ after blocking in 4% nonfat dried milk in PBST. Affinity-purified rabbit polyclonal anti-Clb2 and anti-Clb3 antibodies (Kellogg and Murray 1995) were used at a dilution of 1:1,200 in blocking buffer (2% BSA, PBST, 0.5M NaCl, 0.02% NaN₃). Rabbit polyclonal anti-Sic1 serum (a gift of Mike Mendenhall, University of Kentucky, Lexington, KY) was used at 1:1,000 in blocking buffer (4% nonfat dried milk, 2% BSA, PBST, 0.02% NaN₃) + 10 μg/ml cell lysate made from sic1Δ cells (JM408, made with MDMp203; a gift of Mike Mendenhall). 12CA5 ascites (BabCO) was used at 1:1,000, rabbit polyclonal anti-Cdc16 and anti-Cdc23 (Lamb et al. 1994) were used at 1:2,000, and anti-Cdc27 (Lamb et al. 1994) was used at 1:2,500. These four antibodies were all diluted in blocking buffer (4% nonfat dried milk, PBST, 0.02% NaN₃).

The remaining lysate was used for immunoprecipitation. 2–20 mg of lysate per sample was used depending on the experiment. 0.33–3 μg of antibody was added to the lysate and incubated on ice for 20 min. Samples were then centrifuged in a microfuge at 14,000 rpm for 5 min at 4°C and transferred to 10–15 μl of protein A CL-4B Sepharose beads (Sigma-Aldrich) that had been equilibrated in lysis buffer. The beads were rotated at 4°C for 1–2 h. The beads were manipulated as described below.

For histone H1 kinase reactions, 0.33 μg of anti-Clb2 or anti-Clb3 antibody was used for immunoprecipitation in 1–5 mg of cell lysate. After immunoprecipitation, the beads were 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 twice in kinase buffer (80 mM Na-β-glycerophosphate, pH 7.4, 15 mM MgCl₂, 20 mM EGTA). All washes were performed on ice. Kinase reactions were performed in 15 μl of kinase buffer containing 1 mM DTT, 25 μM ATP, 2.5 μg histone H1 (Upstate Biotechnology) and 1 μCi of γ[³²P]ATP (Amersham Pharmacia Biotech) and were incubated for 15 min at 25°C. Reactions were stopped by adding 15 μl of 2× SDS sample buffer and heating samples to 99°C for 5 min. Samples were run on a 15% polyacrylamide gel, stained, and dried. Kinase gels were quantified using a Molecular Dynamics PhosphoImager and ImageQuant software.

For antiphosphotyrosine blots, 2 μg of 12CA5 antibody was used to immunoprecipitate Cdc28-HA or Cdc28-VF-HA from 15–20 mg of cell lysate. The beads were washed three times in kinase bead buffer and twice in PBS. Standard methods were used for PAGE and protein transfer to nitrocellulose. The blot was then blocked in P-Tyr blocking buffer (1% BSA, 10 mM Tris-Cl, pH 7.5, 100 mM NaCl, 0.1% Tween 20) and incubated overnight at 4°C in blocking buffer containing an anti-P-Tyr Fab fragment conjugated to HRP, RC20H (Transduction Laboratories), diluted to 1:2,500. The blot was washed five times in TBST (10 mM Tris-Cl, pH 7.5, 100 mM NaCl, 0.1% Tween 20), incubated in SuperSignal chemiluminescent substrate (Pierce Chemical Co.), using the manufacturer's instructions, and then exposed to X-ray film. After exposure, the blot was stripped in an SDS-containing buffer, and reprobed with the 12CA5 antibody to confirm that the same amount of Cdc28 had been immunoprecipitated.

For APC assay, 1 μg of rabbit polyclonal anti-Cdc26 antibody (Hwang and Murray 1997) was used to immunoprecipitate the APC from 5 mg of cell lysate. The APC assay was conducted as previously described (Charles et al. 1998), except that cells were lysed in the lysis buffer described above, the beads were washed in APC bead buffer (250 mM NaCl, 50 mM Tris-Cl, pH 7.4, 50 mM NaF, 5 mM EGTA, 5 mM EDTA, 0.1% NP-40, 1 mM DTT) three times, and then twice in QA+NaCl buffer (20 mM Tris-Cl, pH 7.6, 100 mM NaCl, 1 mM MgCl₂). 25 nM of purified Hct1 (Jaspersen et al. 1999) was added to some samples during the ubiquitination reaction.

For Cdc20 binding experiments, 3 μg of anti-Cdc26 antibody was used to immunoprecipitate 10–20 mg of cell lysate. After immunoprecipitation, the beads were washed three times in Cdc20 bead buffer (200 mM NaCl, 50 mM Tris-Cl, pH 7.4, 50 mM NaF, 5 mM EGTA, 5 mM EDTA, 0.1% NP-40, 1 mM DTT), and then twice in low salt kinase buffer (10 mM NaCl, 20 mM Hepes-KOH, pH 7.4, 5 mM MgCl₂). The immunoprecipitates were then processed for Western blots as described above.

Since CDC28-VF suppresses the checkpoint defect of cdc55Δ (Minshull et al. 1996), we wondered if the CDC28-VF mutant alone might have difficulty leaving mitosis. Fig. 1 A shows that progression through mitosis is delayed in CDC28-VF. Wild-type and CDC28-VF cells were arrested in G1 by the mating pheromone alpha factor and released into the cell cycle. Once cells had budded, alpha factor was readded to arrest cells that had completed the cycle. This regimen allows us to look clearly at one synchronous cell cycle. CDC28-VF cells show a 30-min delay in the degradation of the anaphase inhibitor Pds1 and sister chromatid separation. Clb2 proteolysis and the fall in Clb2-associated kinase activity are delayed by >30 min. These delays can be partially attributed to a 15-min delay in Clb2 accumulation, but the persistence of peak levels of Clb2 for at least 60 min clearly reflects an additional defect in CDC28-VF. In addition, mitotic entry is not delayed because short spindles, a marker for mitotic entry, appear at the same time in CDC28-VF and wild-type cells (data not shown). CDC28-VF does not delay exit from G1, since the CDK inhibitor, Sic1 (Schwob et al. 1994), disappears at the same time (30 min) in wild-type and CDC28-VF.

Although CDC28-VF delays passage through mitosis, the doubling time of CDC28-VF cells is nearly identical to that of wild-type cells (data not shown). This apparent paradox can be explained by the fact that CDC28-VF cells exit mitosis at a larger cell size than wild-type cells, and therefore have to grow less in G1 to reach the critical cell size needed to pass Start, the cell cycle transition that commits them to replicating their DNA (Johnston et al. 1977). Thus, the increase in time spent in mitosis is made up by a decrease in time spent in G1.

To see if the mitotic delay in CDC28-VF is due to difficulty exiting mitosis, we examined cells that were recovering from activation of the spindle checkpoint. Wild-type and CDC28-VF cells were arrested in mitosis by treating them with nocodazole (an inhibitor of microtubule polymerization) for three hours and then released from this arrest into fresh medium containing alpha factor to arrest them in G1 as they left mitosis. Wild-type cells degrade Clb2, inactivate Clb2-associated kinase, and separate their sisters within 90 min of removing nocodazole (Fig. 1 B). CDC28-VF cells, however, take 150–180 min to fully escape from the nocodazole arrest.

Since CDC28-VF cells are delayed in exiting mitosis, we investigated how they responded to a prolonged mitotic arrest caused by the spindle checkpoint. We examined the response to overexpressing the protein kinase Mps1, which arrests cells in mitosis by activating the checkpoint, but does not damage the spindle (Hardwick et al. 1996). After about eight hours, wild-type cells overcome the arrest, divide, and resume proliferating. In contrast, CDC28-VF cells overexpressing Mps1 cannot proliferate (Fig. 2 A); when individual cells are followed microscopically, many never divide, and the remainder go through only one or two divisions (Fig. 2 B and data not shown). The cell cycle arrest and eventual lethality are completely suppressed by the mad3Δ mutation, which inactivates the spindle checkpoint (Li and Murray 1991; Hardwick et al. 2000). mad3Δ and mad3Δ CDC28-VF cells divide as if there were no activation of the checkpoint. These results show that CDC28-VF cells, unlike wild-type, cannot escape from mitosis in the presence of constant stimulation of the spindle checkpoint. CDC28-VF cells are also sensitive to other perturbations that activate the spindle checkpoint, including the presence of short linear chromosomes (Wells and Murray 1996) and mutations that damage the spindle (Hardwick et al. 1999). Like Mps1 overexpression in CDC28-VF, these treatments are lethal and cause long delays in mitosis (data not shown).

We initially observed that CDC28-VF and cdc55Δ have opposite effects on the exit from mitosis, consistent with the idea that inhibitory phosphorylation of Cdc28 aids exit from mitosis. More careful examination reveals that the mitotic exit defect in CDC28-VF is not due to effects on Cdc28 phosphorylation. Fig. 3 A shows serial dilutions of a panel of mutants, spotted onto galactose-containing plates, which induce Mps1 overexpression, or glucose-containing plates, which do not. swe1Δ cells lack the tyrosine kinase that phosphorylates Cdc28 (Booher et al. 1993). Like CDC28-VF cells, they have no phosphotyrosine present on Cdc28 (Fig. 3 B), but unlike CDC28-VF or CDC28-F, they are no more sensitive to Mps1 overexpression than wild-type cells. Since another kinase might phosphorylate threonine 18 in swe1Δ cells, we investigated the behavior of swe1Δ CDC28-V cells, which should lack all inhibitory phosphorylation. swe1Δ CDC28-V resemble CDC28-V cells, both being only slightly more sensitive to Mps1 overexpression than wild-type cells. mih1Δ cells, which like cdc55Δ cells have increased inhibitory phosphorylation on Cdc28 (Fig. 3 B), do not have a spindle checkpoint defect as judged by their sensitivity to Mps1 overexpression or to microtubule depolymerizing agents (Fig. 3 A and data not shown).

In principle, the relative insensitivity of swe1Δ to Mps1 overexpression could be explained by the existence of other kinases that phosphorylate tyrosine 19 of Cdc28. We do not believe such a kinase exists. We have never detected phosphotyrosine on Cdc28 in swe1Δ or swe1Δ mih1Δ cells (Fig. 3 B and data not shown), and mih1Δ cells show a 15-min delay in entering mitosis that is completely suppressed by deleting SWE1, suggesting that Swe1 is the only kinase responsible for inhibiting Cdc28 (data not shown). In addition, if such a kinase existed, we would expect CDC28-AF (Amon et al. 1992; Sorger and Murray 1992), which substitutes alanine at position 18 of Cdc28 rather than valine, to behave identically to CDC28-VF. However, CDC28-AF behaves like CDC28-F, both of which are less sensitive to Mps1 overexpression than CDC28-VF (Fig. 3 A).

We favor the idea that the T18V, Y19F substitution causes a phosphorylation-independent defect in Cdc28. Based on the crystal structure of human Cdk2, tyrosine 19 of Cdc28 is adjacent to the gamma phosphate of bound ATP (De Bondt et al. 1993). Thus, CDC28-VF, and to a lesser extent the CDC28-F and CDC28-AF mutations, might affect ATP binding, substrate binding, catalytic activity, or substrate specificity of Cdc28.

Because the CDC28-VF defect is not due to a lack of inhibitory phosphorylation, we asked if other mutations that affect mitotic Cdc28 activity might share phenotypes with CDC28-VF. cdc28-1N, clb2Δ, and cks1-38 (a mutant in Cks1, a Cdc28-binding protein required for passage through Start and mitosis) are all more sensitive to Mps1 overexpression than wild-type (Fig. 4 A; Piggott et al. 1982; Hadwiger et al. 1989; Surana et al. 1991; Tang and Reed 1993). This phenotype is not seen in cdc28-4, a temperature-sensitive allele of Cdc28 that prevents passage through Start and is primarily defective in the G1 function of Cdc28 (Reed 1980).

Does a reduction in mitotic Cdc28 activity cause the CDC28-VF phenotype? We have observed that both Clb2- and Clb3-associated kinase activity and total Cdc28-associated kinase activity of CDC28-VF cells is lower than wild-type (Fig. 4 B and data not shown). This is seen both in synchronously cycling cells (Fig. 1, Clb2-associated kinase activity) and in cells arrested by the spindle checkpoint (Fig. 4 B). We estimate the specific activity of a Cdc28-VF/Clb complex is roughly half that of a wild-type Cdc28/Clb complex. Because the difference between wild-type and CDC28-VF is small, it is difficult to know if it is biologically significant. However, an independent substitution in the ATP binding site of Cdc28, CDC28-F88G, has a reduced specific activity in vitro (Bishop et al. 2000) and is as sensitive to overexpression of Mps1 as CDC28-VF (Fig. 4 C).

The mitotic defect of CDC28-VF is semidominant. The heterozygote CDC28/CDC28-VF has an intermediate sensitivity to Mps1 overexpression (data not shown) and overexpression of Cdc28-VF in otherwise wild-type cells creates cells that are fully sensitive to overexpressed Mps1 (data not shown and Fig. 5). These results suggest that Cdc28-VF is a dominant negative mutant, which competes with the wild-type kinase for substrates, mitotic cyclins, and Cks1. In support of this idea, multiple copies of the CDC28 gene on a 2μ plasmid suppress the lethality of overexpressing Mps1 in a CDC28-VF strain (Fig. 4 D). The semidominant phenotype of CDC28-VF does not reflect haploinsufficiency of Cdc28, since CDC28/cdc28Δ diploids, which contain half as much Cdc28 as CDC28/CDC28 diploids, do not have a phenotype like CDC28-VF (data not shown).

Because CDC28-VF impairs the exit from mitosis, crippling other pathways involved in this process might kill CDC28-VF cells. Our inability to make double mutants between CDC28-VF and mutants in the APC supports this idea (Table). CDC28-VF is synthetically lethal in combination with mutations in components of the APC (cdc23-1, cdc16-1, apc10-1, formerly doc1-1; Lamb et al. 1994; Irniger et al. 1995; Hwang and Murray 1997) and positive regulators of the APC (cdc5-1, hct1Δ and cdc20-1; Schwab et al. 1997; Visintin et al. 1997; Charles et al. 1998; Shirayama et al. 1998). These double mutants are inviable at all temperatures, and although the double mutant spores are able to germinate, they die in microcolonies of large-budded cells (data not shown), indicating a terminal arrest in mitosis. These genetic interactions are specific to the APC and regulators of the APC, because two other mutants that arrest in mitosis, cdc13-1 (which activates the DNA damage checkpoint) and cdc15-2 (which arrests cells in anaphase) are both viable in combination with CDC28-VF.

The interactions of CDC28 alleles and clb2Δ with Mps1 overexpression and the cdc23-1 mutant are correlated with each other: cdc28-1N, CDC28-F88G, and clb2Δ are synthetically lethal in combination with cdc23-1 and cannot proliferate when overexpressing Mps1, whereas cdc28-4 is viable in combination with cdc23-1 or Mps1 overexpression (Table II; also, see Irniger et al. 1995). CDC28-F and CDC28-AF, which have milder phenotypes than CDC28-VF, are synthetically lethal with cdc23-1, but viable in combination with cdc16-1, a weaker APC mutant (Table). Lastly, swe1Δ is viable in combination with cdc23-1 and cdc16-1 (Table), demonstrating that Swe1 is unlikely to have a role in promoting the exit from mitosis.

The genetic interactions between CDC28-VF and the APC and its regulators are a mixed blessing. Although they suggest that Cdc28 may help activate the APC, they prevent us from examining the phenotype of a CDC28-VF apc⁻ double mutant, since it is impossible to create such a mutant. We overcame this difficulty by exploiting the fact that overexpressing Cdc28-VF creates cells that behave phenotypically like CDC28-VF. This overexpression in cdc23-1 cells at the permissive temperature of 23°C is toxic and no cells survive even a brief 1-h pulse of CDC28-VF expression driven by the GAL1 promoter (data not shown).

We overexpressed CDC28 and CDC28-VF in cdc23-1 cells that had been arrested by alpha factor and then released them from the arrest into fresh medium at 23°C, adding alpha factor after budding so that cells will rearrest when they reach the next G1. Overexpression of CDC28-VF causes a permanent large-budded arrest with high levels of Clb2 and Clb3 (Fig. 5 A). Sic1 levels never rise and Clb2-associated kinase activity never falls. Sister separation and spindle elongation are delayed by 2 h in cells expressing CDC28-VF (Fig. 5 B and data not shown). These observations show that the induction of anaphase is delayed and the transition from anaphase to G1 is completely blocked in cdc23-1 CDC28-VF cells. These phenotypes are not due to activating the spindle checkpoint because neither mad2Δ nor mad2Δ bub2Δ bypass the arrest (data not shown; Hoyt et al. 1991; Li and Murray 1991; Alexandru et al. 1999). Therefore, we conclude that CDC28-VF is defective in activating both the Cdc20-dependent APC, which triggers anaphase, and the Hct1-dependent APC, which completes cyclin proteolysis.

The failure to induce mitotic cyclin proteolysis in cdc23-1 mutants overexpressing Cdc28-VF could reflect the requirement for Cdc20 activity to activate the Hct1-dependent APC (Visintin et al. 1998; Shirayama et al. 1999; Yeong et al. 2000), or it could reflect a Cdc20-independent defect in the activity of Hct1. To distinguish between these possibilities, we asked if CDC28-VF has normal APC activity in G1, a time when all APC activity is Hct1-dependent and Cdc20 is absent (Charles et al. 1998; Prinz et al. 1998; Shirayama et al. 1998; and see below). We immunoprecipitated the APC from alpha factor-arrested cells with antibodies raised against Cdc26, a nonessential component of the APC (Hwang and Murray 1997), and measured its ability to ubiquitinate an iodinated fragment of sea urchin Cyclin B in a reconstituted ubiquitination assay (Charles et al. 1998). We detected no differences in APC activity in wild-type and CDC28-VF, and the activity of immunoprecipitates from both wild-type and CDC28-VF could be increased by adding recombinant Hct1 (Fig. 6; Jaspersen et al. 1999). In addition, we have shown that Hct1-dependent proteolysis of Clb2 and Pds1 in G1 is normal in CDC28-VF (data not shown and Fig. 7 C). These experiments are consistent with the idea that the CDC28-VF mutant has no direct effect on the activity of the Hct1-dependent APC.

Cdc20 is required for sister chromatid separation (Shirayama et al. 1998). CDC28-VF shows delays in sister separation (Fig. 1 and Fig. 5), suggesting that Cdc20-dependent APC activity is defective in these cells. We have approached this issue in more detail by examining the half life of Pds1, a substrate of the Cdc20-dependent APC (Visintin et al. 1997; Shirayama et al. 1999; Tinker-Kulberg and Morgan 1999) during anaphase, a time when the Cdc20-dependent APC is thought to be active (Jaspersen et al. 1998). First, we needed to confirm that degradation of Pds1 in anaphase is due to Cdc20 and not Hct1. We arrested wild-type, cdc20-3, and hct1Δ strains in anaphase by overexpressing a nondegradable Clb2 (pGAL-CLB2-Δ176; Surana et al. 1993). These strains also contained an epitope-tagged form of Pds1 replacing the endogenous gene. When all cells had reached anaphase (Fig. 7 A, t = 0), the cultures were shifted to 37°C to see if Pds1 levels would rise in either the cdc20-3 or hct1Δ strain, as an indication that Pds1 had become more stable. Pds1 levels rose in the cdc20-3 strain, but not in the hct1Δ strain, showing that the stability of Pds1 in anaphase is controlled by the Cdc20- rather than the Hct1-dependent APC (Fig. 7 A).

We next examined the half life of Pds1 during an anaphase arrest caused by the cdc15-2 mutation. Pds1 is unstable during this arrest and Cdc20 is required to exit from the arrest (Jaspersen et al. 1998; Tinker-Kulberg and Morgan 1999). cdc15-2, cdc15-2 CDC28-VF, and cdc15-2 clb2Δ cells were arrested in anaphase, an epitope-tagged PDS1 gene driven by the GAL1 promoter was induced by adding galactose for 1 h, and its expression was terminated by adding glucose. The half life of Pds1 in cdc15-2 cells in this experiment was <15 min, but was >1.5 h in CDC28-VF cdc15-2 and clb2Δ cdc15-2 cells (Fig. 7 B), showing that the CDC28-VF and clb2Δ mutations compromise Cdc20-dependent APC activity.

Earlier studies argued that Hct1 and Cdc20 were specificity factors for the APC, with Cdc20 directing the ubiquitination of Pds1 and Hct1 directing that of mitotic cyclins (Visintin et al. 1997). The instability of Pds1 in G1 cells, which lack detectable Cdc20 (Prinz et al. 1998), prompted us to reexamine this issue. Wild-type, CDC28-VF, and cdc20-3 cells were arrested in G1 by alpha factor, or in the case of hct1Δ, which is resistant to alpha factor (Schwab et al. 1997), by the cdc28-13 mutation (Reed 1980). Once arrested, wild-type, CDC28-VF, and cdc20-3 cells were shifted to 37°C (for cdc28-13 and cdc28-13 hct1Δ, asynchronous cultures were transferred to 37°C), pGAL-PDS1 was induced by adding galactose for two hours and then expression was shut off by adding glucose. Pds1 was equally unstable in wild-type, CDC28-VF, cdc20-3, and cdc28-13 cells, but was completely stable in the hct1Δ cdc28-13 cells. This control shows that the defect in Pds1 stability in CDC28-VF is specific to anaphase and, together with Fig. 7 A, shows that Pds1 is targeted for destruction by the Cdc20-dependent APC in mitosis and by the Hct1-dependent APC in G1. In addition, this experiment reinforces the conclusion derived from in vitro experiments in Fig. 6 that CDC28-VF has no defects in Hct1-dependent APC activity in G1. Our results differ from those of Visintin et al. 1997, who found that stability of Pds1 in G1 is regulated by Cdc20. Their results may have been due to incomplete alpha factor arrest of the cdc20-1 allele used in their study. Our results agree with the recent observation that Clb2 is targeted for destruction by both the Cdc20- and Hct1-dependent forms of the APC (Yeong et al. 2000).

The stabilization of Pds1 in mitotic CDC28-VF cdc15-2 cells suggests that CDC28-VF is defective in the Cdc20-dependent APC. As a first step in investigating the biochemical defect of CDC28-VF, we examined the interaction between Cdc20 and the APC in anaphase, a time when the Cdc20-dependent APC is active. We arrested cdc15-2 and cdc15-2 CDC28-VF cells in anaphase, immunoprecipitated the APC with anti-Cdc26 antibodies, and examined the amount of associated Cdc20. Equal amounts of APC was immunoprecipitated from all three strains, but in CDC28-VF cells there was less associated Cdc20 (Fig. 8), even though the total level of Cdc20 was similar in wild-type and CDC28-VF.

We have shown that the CDC28-VF mutant is defective activating the Cdc20-dependent APC. CDC28-VF cells show a short delay in exiting mitosis, but this delay becomes more severe when the APC is compromised by spindle checkpoint activation or defects in the APC or its regulators. These phenotypes allowed us to show a requirement for Cdc28 in exiting mitosis. The defect in the exit from mitosis in CDC28-VF is correlated with reduced binding of Cdc20 to the APC and lower Cdc20-dependent APC activity. CDC28-VF cells have normal Hct1-dependent APC activity in G1, suggesting that their failure to exit mitosis is the result of interrupting the chain of events that normally leads from mitosis to G1: reduced Cdc20-dependent APC activation prevents the activation of Cdc14, which normally leads to activation of Hct1 and increased levels of Sic1, the key steps in inactivating the mitotic activity of Cdc28 (Lim et al. 1998; Visintin et al. 1998; Jaspersen et al. 1999; Shirayama et al. 1999).

Previously, it has been shown that Cdc28 inhibits exit from mitosis. High levels of Cdc28-associated kinase activity cause arrest in anaphase (Holloway et al. 1993; Surana et al. 1993), and inhibiting Cdc28 in mitotically arrested cells activates the APC, driving them through cytokinesis and into G1 (Amon 1997; Li and Cai 1997). These observations led to the conclusion that the major role for mitotic Cdc28 in regulating the APC was to inhibit the activity that drove cells out of mitosis. Our results show that Cdc28 has different effects on APC activity at different stages of mitosis. Cdc28/Clb activity activates the Cdc20-dependent APC to induce the metaphase to anaphase transition, but inhibits the Hct1-dependent APC, thus inhibiting cytokinesis and the transition from anaphase to G1. This pattern of regulation helps ensure chromosome segregation precedes cell division and allows budding yeast to regulate these events separately, a useful feature in an organism that specifies the site of cytokinesis long before spindle assembly.

Proteins that were identified as members of the spindle checkpoint have two roles in keeping cells with depolymerized microtubules from leaving mitosis. Six of them, Mps1, Mad1–Mad3, Bub1, and Bub3, detect kinetochores that are not attached to the spindle (Hoyt et al. 1991; Li and Murray 1991; Weiss and Winey 1996) and prevent sister chromatid separation by inhibiting the Cdc20-dependent APC (Hwang et al. 1998; Kim et al. 1998). In contrast, Bub2, also identified as part of the spindle checkpoint (Hoyt et al. 1991), detects an unknown lesion and arrests cells in anaphase, probably by preventing the activation of Cdc14 (Alexandru et al. 1999; Fesquet et al. 1999; Fraschini et al. 1999; Li 1999). The opposition between the inhibition of the Cdc20-dependent APC by the spindle checkpoint and activation by Cdc28 explains why CDC28-VF cells, which are defective in activating the Cdc20-dependent APC, have difficulty overcoming a checkpoint-dependent arrest.

We do not know how wild-type cells eventually escape from mitosis despite continued Mps1 overexpression, a process that could be described as adaptation to the spindle checkpoint (Rieder and Palazzo 1992; Minn et al. 1996; Rudner and Murray 1996). In particular, we cannot distinguish between constitutive mechanisms, such as stabilization of mitotic cyclins leading to a slow rise in Cdc28/Clb activity that eventually leads to APC activation, and induced mechanisms, such as a slow process initiated by components of the checkpoint that leads to reduced inhibition of Cdc20. Although we initially thought of CDC28-VF as an adaptation mutant (Minshull et al. 1996), we now believe that it achieves its effect by reducing the ability of Cdc28 to activate the Cdc20-dependent APC, whether or not the checkpoint is active. It is only when the checkpoint is active, however, that this reduction is sufficient to keep cells from leaving mitosis. This discussion highlights a general difficulty in studying adaptation mutants: the ability to overcome prolonged checkpoint-dependent arrests appears to be sensitive to small, checkpoint-independent defects in basic cell cycle machinery (Toczyski et al. 1997).

Genetic analysis of CDC28 has revealed two types of mutants, those that primarily affect G1 (cdc28-4, cdc28-13; Reed 1980) and those that primarily affect exit from mitosis (CDC28-VF, cdc28-1N; Piggott et al. 1982; Surana et al. 1991). The mitotic mutants share phenotypes with clb2Δ and cks1-38. Do these four mutants, CDC28-VF, cdc28-1N, clb2Δ and cks1-38, have a common biochemical defect? clb2Δ cells contain no Clb2, and therefore have reduced levels of mitotic Cdc28 activity (Grandin and Reed 1993). The defects of cdc28-1N and cks1-38, which arrest cells in mitosis, may have more to do with altering the substrate specificity of Cdc28. When assayed by immunoprecipitation of Clb2-associated kinase, cdc28-1N strains have similar kinase activity to wild-type, but they have no kinase activity associated with Cks1-coupled beads (Surana et al. 1991; Kaiser et al. 1999). These results suggest that the primary defect in cdc28-1N is its failure to bind Cks1, which has been shown in frog and clam extracts to be essential for exit from mitosis and APC phosphorylation by Cdc2/Cyclin B (Patra and Dunphy 1996, Patra and Dunphy 1998; Shteinberg and Hershko 1999).

Our only clue to the biochemical defect of Cdc28-VF is that it appears to have a small reduction in its specific activity (Fig. 4 D). We think this defect may be important because CDC28-F88G, a mutant in the ATP binding site of Cdc28, also has reduced specific activity (Bishop et al. 2000), and shares phenotypes with CDC28-VF (Fig. 4 C). The cdc28-1N and cks1-38 mutations may produce their effects by reducing the level of the Cdc28/Clb/Cks1 complexes, whereas in clb2Δ cells, the complexes of Cdc28 with the remaining Clb proteins may be less capable of activating the Cdc20-dependent APC.

Although the Cdc28 activity is lower in CDC28-VF cells, we do not think that the defect in CDC28-VF is simply due to lower total Cdc28 activity per cell. Overexpression of Clb2 and Clb3, which raises Cdc28 activity in cells (Stueland et al. 1993), does not suppress the mitotic defect of CDC28-VF (data not shown). This result is consistent with the idea that the specific activity per Cdc28 molecule, not the total Cdc28 activity per cell, is critical for activating the exit from mitosis. An alternative explanation is that the different mutants change the substrate specificity of Cdc28, preventing phosphorylation of important mitotic substrates.

Protein phosphorylation is a common way of regulating protein activity. Mutating putative phosphorylation sites to nonphosphorylatable residues is a widely used technique for assessing the biological function of phosphorylation of specific proteins (Li et al. 1995; Zachariae et al. 1998; Jaspersen et al. 1999). Our analysis of CDC28-VF shows that such mutations can have unanticipated effects that are independent of phosphorylation.

It is difficult to tell how common such effects are. Our observations of CDC28-VF suggest that the studies conducted with the CDC28-AF, CDC28-VF, and CDC28-F mutants in budding yeast (as well as experiments with the corresponding mutants in Cdc2) should be reexamined to exclude the possibility that the observed effects of these mutants were due to phosphorylation-independent defects. A simple control is to ask whether the phenotype of the CDC28 mutants is exactly mimicked by deletion of SWE1. If so, the conclusions of the original experiments are secure. If not, phosphorylation-independent effects due to mutating the inhibitory residues may contribute to the observed phenotypes. In the general case, the ideal control is to show that inactivating the kinase that phosphorylates a particular protein produces a similar effect on the substrate's activity as do the phosphorylation site mutants.

How does Cdc28 promote anaphase? Experiments in frogs, clams, and mammalian cell culture have all suggested that phosphorylation activates the APC by modifying four of its subunits: Cdc16, Cdc23, Cdc27, and Apc1 (BimE; Hershko et al. 1994; Peters et al. 1996; Kotani et al. 1998, Kotani et al. 1999; Patra and Dunphy 1998). These proteins are phosphorylated during mitosis and the phosphorylated APC has greater Cdc20-dependent activity in vitro, whereas dephosphorylation of purified APC causes a loss of activity (Lahav-Baratz et al. 1995; Shteinberg et al. 1999). Studies in clams have suggested that this phosphorylation is required for proper Cdc20 binding (Shteinberg et al. 1999). In the accompanying paper (Rudner and Murray 2000, this issue), we show that the APC is phosphorylated by Cdc28 in budding yeast, and that a defect in this phosphorylation causes reduced Cdc20-dependent APC activity and contributes to the CDC28-VF phenotype.

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

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