A role for the Cdc14-family phosphatase Flp1p at the end of the cell cycle in controlling the rapid degradation of the mitotic inducer Cdc25p in fission yeast

Exit from mitosis in eukaryotic cells requires the inactivation of mitotic cyclin-dependent kinases (CDKs) (Morgan, 1999; Bardin and Amon, 2001). In budding yeast, Cdc14p inactivates CDKs by promoting B-type mitotic cyclin degradation and accumulation of the CDK inhibitor Sic1p (Visintin et al., 1998; Jaspersen et al., 1999). The CDC14 gene is essential for cell viability, and thermosensitive mutants arrest cell cycle progression in telophase with high levels of CDK1/Clb-associated kinase activity (Visintin et al., 1998; Wan et al., 1992; Taylor et al., 1997). Cdc14p is not only required for mitotic exit (Jaspersen et al., 1999; Visintin et al., 1998) but also plays a role in cytokinesis (Mensen et al., 2001; Luca et al., 2001; Yoshida et al., 2001). Cdc14p is regulated in part by changes in its subcellular localisation. It is sequestered in the nucleolus from late telophase to the metaphase-anaphase transition of the following cell cycle by the Net1p/Cfi1p protein of the RENT complex (Regulator of Nucleolar Silencing and Telophase) (Visintin et al., 1999; Straight et al., 1999; Traverso et al., 2001), from where it is released during anaphase, dependent upon the FEAR (Cdc Fourteen Early Anaphase Release) and MEN (Mitotic Exit Network) group of proteins (Shou et al., 1999; Straight et al., 1999; Visintin et al., 1999; Stegmeier et al., 2002). Budding yeast Cdc14 is regarded as the effector of the MEN signal network that inactivates mitotic cyclin-dependent kinases (mitotic CDKs) at the end of anaphase so that cells can exit mitosis (Traverso et al., 2001).

Cdc14 orthologues have been identified in other species including Schizosaccharomyces pombe, Caenorhabditis elegans and Homo sapiens on the basis of sequence similarity to the budding yeast Saccharomyces cerevisiae protein (Li et al., 1997; Cueille et al., 2001; Trautmann et al., 2001; Gruneberg et al., 2002). (All) Cdc14 proteins characterised to date are dual-specificity protein phosphatases that share a core of approximately 300 amino acids towards the amino terminus that includes a strictly conserved protein-phosphatase domain motif (Gray et al., 2003).

In fission yeast, the Cdc14-related protein phosphatase Flp1p (also named Clp1) seems to performs a different role in cell cycle progression to that of its S. cerevisiae counterpart. A null mutant of the S. pombe flp1 is viable but divides at a reduced size and has defects in septation (Cueille et al., 2001; Trautmann et al., 2001). In contrast to Cdc14p, Flp1p is not essential for mitotic exit. Consistent with this, Flp1p is not required for the dephosphorylation and activation of APC activator Ste9p, nor for the accumulation of the cyclin-dependent kinase inhibitor Rum1p, nor for the destruction of the mitotic B-type cyclin Cdc13p (Cueille et al., 2001). Although both Flp1p and Cdc14p proteins are nucleolar-located in interphase, from where they are released to the nucleus at mitosis, Flp1p it is also present on the spindle pole body during interphase as well as on the spindle and the contractile ring during mitosis, although its substrates at these locations are unknown (Cueille et al., 2001; Trautmann et al., 2001). The regulation of Flp1p subcellular localisation is complex and differs in details from the mechanisms regulating Cdc14p localisation in S. cerevisiae. Budding yeast Cdc14p nucleolar release requires MEN function. In fission yeast the orthologues of the MEN genes are collectively termed the SIN (Septation Initiation Network) (Le Goff et al., 1999; Balasubramanian et al., 2000). Recently, we have shown that SIN function is not required to release Flp1p from the nucleolus in S. pombe, although inactivation or attenuation of the SIN is needed for the relocalisation of Flp1p proper to the nucleolus at the end of mitosis (Cueille et al., 2001).

The fact that flp1Δ mutant cells are advanced in mitosis suggests an interaction with the proteins controlling mitosis. This is supported by the finding that overexpression of flp1⁺ arrests cells in G2 with dephosphorylated Cdc25p, and that this arrest depends upon active wee1 (Cueille et al., 2001). These studies suggested that elements of the mitotic control system (Cdc25p and Wee1p) or their upstream regulators might be of the targets of Flp1p (Cueille et al., 2001; Trautmann et al., 2001). We have explored this further, and examined the role of Flp1p in regulating the phosphoprotein phosphatase Cdc25p.

Standard molecular biology and genetic methods were used as described previously (Moreno et al., 1991; Norbury and Moreno, 1997). Yeast transformation was carried out using the lithium acetate transformation protocol (Norbury and Moreno, 1997). Fission yeast were grown and manipulated according to standard protocols in either yeast extract (YE) or minimal (M) medium, containing appropriate supplements [YES media: yeast extract plus supplements (225 mg/l adenine, histidine, leucine, uracil and lysine hydrochloride] (Moreno et al., 1991). S. pombe strains used in this study were from Moreno, Simanis and Bueno's laboratory collections, in particular, all strains involving the flp1Δ deletion carried the construction described previously (Cueille et al., 2001). The induction of synchrony by nda3-KM311 was performed by blocking cells at metaphase by incubating them for 6 hours at 19°C or after elutriation of G2 cells by incubating them for 4 hours at 20°C (cold sensitive arrest or cs arrest). Under these conditions more than 80% of cells blocked at metaphase. Alternatively, to obtain a better degree of synchrony and prior to the cs arrest, nda3-KM311 cells were block for 3 hours with hydroxyurea (HU). For induction of the nmt1 promoter, a culture growing exponentially in medium containing thiamine was washed twice and resuspended in medium without thiamine as described previously (Cueille et al., 2001).

To construct the GST derivatives used in this work different PCRs were performed to amplify the full ORF of flp1+ and for creation of the flp1CS (C286S) mutant with standard cloning and site-direct mutagenesis methods. After cloning, all plasmids generated were sequenced to check that not other changes than the desired one was introduced in the ORF by PCR. Oligonucleotides used for cloning of GST-flp1+ or GST-flp1CS were:

• 5′-TACTTGTCATATGGATTACCAAGA-3′ where the underlined sequence is an NdeI site for cloning.

• 5′-AAAAACTGCGGCCGCCATTTAAACCAG-3′ where the underlined sequence is a NotI site for cloning.

Oligonucleotides used for mutagenesis (to obtain flp1CS) were:

5′-GCTGTTCATAGCAAAGCAGG-3′ and 5′-CCTGCTTTGC-TATGAACAGC-3′.

nmt:GST-flp1-containing plasmids were constructed as described previously (Shiozaki and Russell, 1997).

Antisera and tagged strains allowing detection of Cdc25p, Cdc13p, Cdc2p, Wee1p and tubulin have been described previously (Bähler and Pringle, 1998; Blanco et al., 2000; Cueille et al., 2001).

Cell extracts for western blotting and immunoprecipitation were obtained as described previously (Cueille et al., 2001). For western blots, 75 mg of total protein extract was run on a 12% standard SDS-PAGE gel, transferred to nitrocellulose and probed with rabbit affinity purified anti-Cdc25 (1:250), SP4 anti-cdc13 (1:250) and anti-Cdc2 (1:500) polyclonal antibodies or anti-haemagglutinin (HA) (12CA5, Roche) and anti-myc monoclonal antibodies. Secondary antibodies used were goat anti-rabbit or goat anti-mouse conjugated to horseradish peroxidase (Amersham) (1:3500). Mouse TAT1 anti-tubulin monoclonal antibodies (1:500) and goat anti-mouse conjugated to horseradish peroxidase (1:2000) as secondary antibody was used to detect tubulin as a loading control. Immunoblots were developed using the ECL kit (Amersham) or Super Signal (Pierce). GST purification has been described previously (Shiozaki and Russell, 1997). Phosphatase assays were performed with para-nitrophenylphosphate (pNPP), immunoprecipitated Cdc25p or GST-Cdc25p phosphorylated in vitro as substrates by incubation in phosphatase buffer (imidazole-HCl 50 mM, pH 6.6, EDTA 1 mM, DTT 1 mM) with purified GST-Flp1p (and appropriate controls) for 40 minutes at 30°C (or as indicated). Kinase assays were performed with H1 histone (Calbiochem) or bacterially produced GST-Cdc25p as substrates on immunoprecipitates at 32°C as described previously (Moreno et al., 1989; Calzada et al., 2000), and quantified with a phosphorimager. H1 histone signals were normalised to the nda3-KM311 arrest level.

Approximately 10⁷ cells were collected by centrifugation, washed once with water, fixed in 70% ethanol and processed for flow cytometry or DAPI staining, as described previously (Moreno et al., 1991).

The mitotic index was determined from the percentage of binucleate cells (cells with two nuclei and without a septum) after DAPI staining. The septation index was determined from the percentage of cells with a septum after calcofluor staining. Also, to evaluate the completion of the cell cycle, the percentage of cells in anaphase B was estimated by counting cells with two nuclei and an elongated spindle. Staining of microtubules was performed as described previously (Bähler and Pringle, 1998; Balasubramanian et al., 1997) using anti-tubulin antibody TAT-1 (Bähler and Pringle, 1998) or using a GFP-atb2⁺ strain (Yamamoto et al., 1999). Staining of myc₁₂-Cdc25p protein was performed as described previously (López-Girona et al., 1999). Fluorescence images were collected using a Zeiss Axioplan 2 microscope with a 63× or a 100× objective and a digital camera (Hamamatsu ORCA-ER) and processed with Improvision software. More than 1000 cells were examined for each time point, and each experiment was repeated at least three times, to gain an estimate of error.

In S. pombe cells, as well as in most eukaryotes, entry into mitosis is controlled by the balance between the activities of the inhibitory kinase Wee1p and the opposing effect of the phosphatase Cdc25p in regulating the activity of Cdc2p (Russell and Nurse, 1986; Russell and Nurse, 1987a; Gould and Nurse, 1989; Norbury et al., 1991; Norbury and Nurse, 1992; McGowan et al., 1993), the catalytic subunit of CDK-complexes. We therefore examined the consequences of mutating the flp1 gene upon the stability and phosphorylation state of Cdc25p and Wee1p. Deletion of the flp1 gene resulted in the accumulation of hyperphosphorylated Cdc25p (Fig. 1). This pattern was independent of the stage of the cell cycle, given that cells blocked in G1, S, G2 phase or mitosis accumulated hyperphosphorylated Cdc25p when flp1 was mutated (Fig. 1), which is consistent with the idea that Flp1 may influence the phosphorylation state of Cdc25p. However, Wee1p was partially dephosphorylated in flp1 mutant strains (Fig. 1), indicating that Flp1p does not play an essential role in Wee1p dephosphorylation.

Though Flp1p is not essential for completion of mitosis (Cueille et al., 2001; Trautmann et al., 2001), we examined whether flp1 null cells showed any defects in mitotic exit. We therefore analysed the M to G1 transition in cells synchronised by nda3-KM311 arrest release. Exponentially growing cultures of nda3-KM311 and isogenic nda3-KM311 flp1Δ were blocked at 19°C for 6 hours and released by transfer to 32°C. Samples were taken at intervals and processed for cytological observation, western blot analysis and assayed for Cdc13-associated kinase activity (Fig. 2). In nda3-KM311 flp1Δ mutants the rate of disappearance of binucleate cells (mitotic index) (Fig. 2A, upper plot) was delayed compared with the control. Consistent with this, the appearance of septated cells was also delayed in flp1 mutant cells (Fig. 2A, middle plot). However, analysis of the exit from mitosis by measuring the rate of disappearance of binucleated cells with elongated spindles (anaphase B cells) showed no major differences between flp1 mutants and controls (data not shown). Therefore, we conclude that the delay of disappearance of binucleate cells in strains deleted for flp1 is the result of the delay in cytokinesis. Though the levels of the Cdc13-cyclin B protein declined similarly in the control and the flp1 deletion mutant as cells exit from mitosis, following release from the metaphase arrest, the activity of Cdc13-associated H1 kinase declined faster in the nda3-KM311 control than in nda3-KM311 flp1Δ double mutant (Fig. 2A, lower plot, and 2B). Thus, Flp1p may contribute to the prompt inactivation of CDK as cells initiate anaphase.

In wild-type cells, Cdc25p is rapidly dephosphorylated, and then degraded during mitotic exit (Fig. 2C). The protein then reaccumulates during S and G2 phases, reaching a peak as cells enter mitosis (Moreno et al., 1990; Ducommun et al., 1990). In contrast, upon release of a nda3-KM311 flp1Δ mutant, Cdc25p remained phosphorylated and its steady state level did not change significantly up to 30 minutes after the release from the metaphase block (Fig. 2C).

To obtain better synchrony, this result was confirmed using a culture of nda3-KM311 generated by centrifugal elutriation, which allowed us to obtain over 90% of cells in metaphase after 4 hours at 20°C. This experiment confirmed that p80 ^cdc25 remains phosphorylated longer and is more stable in the absence of flp1 function (Fig. 2D). Together, these results strongly suggest that rapid degradation of the CDK activator Cdc25p at the end of mitosis is regulated by the Flp1p phosphoprotein phosphatase.

Next, we analysed the localisation of the Cdc25p protein in flp1Δ deletion mutant cells expressing Myc-tagged Cdc25p by indirect immunofluorescence microscopy. As described previously, in the myc₁₂-cdc25⁺ control the amount of nuclear Cdc25p increased as cells progress throughout G2, to be mostly nuclear in late G2 and mitosis and then decreased in the transition of mitosis to G1 (López-Girona et al., 1999) (Fig. 3A, upper panels). We observed that in myc₁₂-cdc25⁺ flp1Δ mutant cells the signal of Cdc25p was clearly more intense throughout the cell, in both cytoplasm and nuclei, in every cell cycle stage, indicating that deletion of flp1 causes a higher stability and inappropriate localisation of the Tyr15 phosphatase from late G2 to the G1/S transition in fission yeast cells (Fig. 3A, lower panels). To ensure that flp1Δ mutants accumulated more Cdc25p protein than wild-type, both types of cells were mixed and co-prepared for immunofluorescence (Fig. 3B). In these experiments the myc₁₂-cdc25⁺ control was also labelled with GFP-atb2⁺ (Yamamoto et al., 1999) to distinguish them from myc₁₂-cdc25⁺ flp1Δ mutant cells. This approach confirmed that Cdc25p is more stable in S. pombe cells lacking the Cdc14p-related phosphatase.

Consistent with a higher stability of active (hyperphosphorylated) Cdc25p in the absence of Flp1p, deletion of the flp1 gene partially rescued the ts defect of cdc2-33 cells, a cdc2^ts allele that causes a tight G2 arrest with low CDK-associated kinase activity (Carr et al., 1989; Castellanos et al., 1996), allowing them to form colonies up to 33°C (Fig. 3C). Conversely, we found that flp1Δ is synthetically lethal at semi-permissive temperatures when combined with cdc25-22 ts mutants. A flp1Δ cdc25-22 strain is inviable at 33°C (Fig. 3C), a temperature at which both flp1Δ and cdc25-22 single mutants grew to form colonies. cdc25-22 flp1Δ double mutants arrested at 33°C with a similar cdc arrest phenotype to that of cdc25-22 at 35.5°C (Russell and Nurse, 1986). These observations are consistent with data indicating that cells lacking Flp1 are defective in regulating the levels and localisation of Cdc25p (Fig. 2 and Fig. 3A), and favour the hypothesis that Flp1p controls the steady state levels of Cdc25p.

Since Cdc25p was not promptly dephosphorylated and degraded during mitotic exit in the flp1::kanMX6 mutant, we next performed a biochemical analysis to determine whether p60^flp1 could dephosphorylate p80^cdc25 in vitro. We constructed a vector to produce full-length Flp1p (or a mutant form in which the cysteine 286 was changed to serine: Flp1pC286S) as a fusion protein with glutathione S-transferase under the control of the thiamine repressible nmt1 promoter. We expressed this nmt1:GST-flp1⁺ fusion and found that it induced a G2 cell cycle arrest similar to nmt1:flp1⁺-expressing cells, indicating that GST-Flp1p is functional in vivo. In contrast to our previous report (Cueille et al., 2001), we found that expression of nmt1:GST-flp1CS also caused cells to arrest in G2, but the block was leaky and delayed by approximately 4 hours as compared to a nmt1:flp1⁺ control. The difference may reflect the time of expression used.

GSTp (control), GST-Flp1p or GST-Flp1CSp were purified virtually to homogeneity from S. pombe cells after 20 hours of induction (Fig. 4A). Three high molecular mass GST-Flp1p species were consistently observed. How these differ is presently unclear, but they are likely to be either differentially phosphorylated forms of the fusion protein or degradation products.

To address whether Cdc25p and Flp1p proteins interact in vivo we tested for the presence of Cdc25p in GST-Flp1p protein samples after the purification with glutathione-Sepharose (GSH) beads. By western blot analysis, we found that GST-Flp1p and GST-Flp1CSp, but not GSTp, associate in vivo with Cdc25p (Fig. 4B). Significantly, the Cdc25p protein that associates with GST-Flp1CSp mutant form lacking phosphatase activity is hyperphosphorylated.

With the purified proteins from cdc25Δ cdc2-3w (cdc25 deleted strains) we next tested the ability of GST-fused proteins to function in an in vitro phosphatase assay using a standard substrate. We found that GST-Flp1p dephosphorylated para-nitrophenylphosphate (pNPP) (Fig. 4C). The dephosphorylation reaction was linear during the time course. Under similar conditions of protein or substrate concentration and time, we did not detect any phosphatase activity either with the GST protein alone or with GST-Flp1CSp mutant control, indicating that Flp1p has intrinsic phosphatase activity. To perform a phosphatase assay with Cdc25p as substrate we isolated phosphorylated p80^cdc25 from mitotic cells. A cold-sensitive nda3 strain was arrested in metaphase at the restrictive temperature, cell extracts were prepared and protein samples immunoprecipitated with a polyclonal antibody raised against Cdc25p. Washed immunoprecipitates were incubated in the presence of either GSTp, GST-Flp1p or GST-Flp1CSp and subjected to western analysis to detect Cdc25p after denaturing SDS gel electrophoresis. As shown (Fig. 4D), GST-Flp1p, but not GST-Flp1CSp, partially dephosphorylates Cdc25p protein. The incomplete dephosphorylation of Cdc25p could be due to either low efficiency of GST-Flp1p or, more likely, to dephosphorylation of specific phosphorylated residues in Cdc25p, since the phosphorylation pattern did not change with time.

Using a different approach to assess whether Cdc25p is a substrate of the Flp1p phosphatase, we developed an assay in which purified GST-Cdc25p from bacteria was first phosphorylated by immunoprecipitated Cdc2/Cdc13 (CDK) complexes and then used as a substrate for purified GST-Flp1p from fission yeast with appropriate controls (Fig. 5). In this assay even the lowest quantities of GST-Flp1p used were able to partially dephosphorylate CDK-phosphorylated GST-Cdc25p (plot in Fig. 5). These data are consistent with those shown in Fig. 4, and together, they support the hypothesis that Flp1p may regulate Cdc25p in vivo directly by dephosphorylation.

Having shown that Flp1p controls the level of the CDK-activator Cdc25p, we then performed a metaphase-block and -release analysis with nda3-KM311 wee1_3HA and isogenic nda3-KM311 flp1Δ wee1_3HA in order to test whether the Wee1p protein kinase was also regulated by the Flp1p phosphatase at the end of mitosis (Fig. 6A). We did not detect any changes either in the steady state level of Wee1p, or in its gel mobility compared to the nda3-KM311 wee1_3HA control, suggesting that Flp1p does not dephosphorylate Wee1p at the M-G1 transition.

Finally, to examine further why over-expression of flp1⁺ causes cells to arrest in G2 with the accumulation of a hypophosphorylated, and most likely, inactive form of Cdc25p (Fig. 6), we analysed the localisation of the Cdc25p protein in these cells. Interestingly, most of the Cdc25p was sequestered in the nucleolus (Fig. 6B), perhaps providing an explanation for the cell cycle arrest. Thus, we believe that the G2 arrest may not reflect the true physiological point of action of Flp1p in the cell cycle, though confirmation of this awaits generation of a conditional allele of flp1.

An important feature of the Cdc25p mitotic inducer is its oscillatory pattern of appearance at each fission yeast cell division cycle. The cdc25⁺-encoded tyrosine phosphatase slowly accumulates during interphase to reach a critical level at the end of G2 (Moreno et al., 1990; Ducommun et al., 1990), to be first activated by polo, and then by Cdc2/Cdc13 kinases, to allow full activation of the latter kinase that induces entry into mitosis (Hoffmann et al., 1993; Kovelman and Russell, 1996; Glover et al., 1998; Nigg, 1998). During mitosis, Cdc25p is abruptly degraded (in a cell-cycle-dependent manner) (Moreno et al., 1990; Ducommun et al., 1990; Nefsky et al., 1996). Because of this pattern of synthesis and degradation it was suggested that mitosis is regulated by the cyclic accumulation (and activation) of the Cdc25p protein (Moreno et al., 1990). Fission yeast Cdc25p is regarded as the main effector that activates mitotic cyclin-dependent kinases at the G2/M transition.

Several lines of evidence shown in this work strongly suggest that the flp1⁺-encoded phosphatase induces the rapid degradation of the Cdc25p mitotic inducer at the end of mitosis. In this study we provide evidence that Flp1p can act as a phosphatase in vitro (Fig. 4A), and, that it interacts in vivo with, and can dephosphorylate Cdc25p in vitro (Fig. 4B,D, Fig. 5). Since Cdc25p is not dephosphorylated during mitotic exit in an flp1 null mutant (see Figs 2, 3) we propose that dephosphorylation of specific residues in Cdc25p by Flp1p may be part of the mechanism that triggers its destruction. However, it is also possible that Flp1p activates another phosphatase, which is in turn responsible for the direct dephosphorylation of Cdc25p. Future experiments to determine which CDK-phosphorylated residues of Cdc25p are substrates of Flp1p will resolve this. Alternatively, Flp1p might regulate the activity of the degradation machinery of Cdc25p (see below).

We have suggested earlier that Flp1p does not play any role in the inactivation of mitotic cyclin-dependent kinases through Cdc13p-cyclin degradation or Rum1p-CDK inhibitor stabilisation at the end of mitosis in fission yeast (Cueille et al., 2001). In the present work we have extended the analysis of flp1 null mutants as they exit from mitosis and observed that a major defect in flp1 null mutant cells is the timing of Cdc25p proteolysis, which correlates with a delay in Cdc2/Cdc13 inactivation and septation (see Figs 2, 3). Although, it was known the Cdc25p is unstable (Moreno et al., 1990; Ducommun et al., 1990; Nefsky and Beach, 1996), our work suggests that proteolysis of the mitotic inducer contributes to CDK inactivation as cells exit mitosis for the proper coordination of nuclear division with cytokinesis. Additionally, the degradation of Cdc25 at the M to G1 transition may be part of the mechanism that ensures a minimum cell mass before entering mitosis in the next cell cycle (Daga and Jimenez, 1999), and since it would also [in a redundant manner with Wee1p and Rum1p (Russell and Nurse, 1987a; Moreno and Nurse, 1994)] prevent activation of any Cdc2/Cdc13 complexes that form during G1, we propose that it may also be important for proper coordination between the end of mitosis and cytokinesis in fission yeast, as it has been suggested that cytokinesis occurs only when mitotic B-type cyclin CDK activity is reduced to low levels (reviewed by Irniger, 2002; Simanis, 2003). This model does not exclude the existence of others substrates for the Flp1p phosphatase, but our results strongly suggest that an important one is the CDK activator Cdc25p.

We and others suggested earlier that Flp1p may regulate the activation of Cdc2p at the onset of mitosis by controlling Cdc25p and Wee1p (Cueille et al., 2001; Trautmann et al., 2001). This hypothesis was founded upon two observations: first the flp1 null mutant is advanced into mitosis, and second, the G2 arrest imposed by over-expression of flp1⁺. The experiments that we present here show that in the absence of flp1, S. pombe cells initiate the cell cycle with more Cdc25p protein (Figs 2, 3). This raises the possibility that they are wee because they reach the threshold level of Cdc25p required for Cdc2/Cdc13 activation earlier, potentially explaining their advanced-into-mitosis (wee) phenotype.

Our analysis shows that Flp1p phosphatase does not play an essential role in Wee1p dephosphorylation (see Figs 1, 6). Nevertheless, we propose that one of the regulatory mechanisms that trigger the inactivation of mitotic CDKs before S. pombe cells undergo septation, is the shift in the balance between dephosphorylation and phosphorylation of the tyrosine 15 of Cdc2. Although our study only proves that Flp1p down regulates Cdc25p, a tyrosine kinase has to be active at this transition of the cell cycle. As fission yeast cells enter mitosis, Nim1 phosphorylates Wee1p (Russell and Nurse, 1987b) at the C-terminus to inactivate it (Mosser and Russell, 2000). Wee1p remains phosphorylated, and hence probably inactive, as S. pombe cells exit from mitosis (Fig. 6A). It is tempting to speculate that the Cdc2-tyrosine 15 kinase Mik1p (Lundgren et al., 1991) may play a role in inactivation of cdc2 during G1, in addition to its role during S phase (Christensen et al., 2000; Baber-Furnari et al., 2000). Whether Flp1p regulates Wee1p localisation and/or the activation of Mik1 will be addressed in future studies.

The Cdc25p protein is ubiquitinated for proteolysis (Nefsky and Beach, 1996). Although the E3 ligase Pub1p is required for the predominant mechanism of Cdc25p polyubiquitination to be active, there is evidence for the existence of an alternative Cdc25p protein ubiquitination pathway (Nefsky and Beach, 1996). Dephosphorylation of Cdc25p may make it a target for Pub1p, or Flp1p may also control the activity of Pub1p. Alternatively, Flp1p may activate a Pub1p-independent degradation pathway of Cdc25p. Distinguishing between these alternatives will be the subject of future studies.

In many cases CDK-mediated phosphorylation of proteins implicated in cell cycle control targets them for proteolysis after polyubiquitinylation (Deshaies, 1995; Deshaies et al., 1995). In the case of the Cdc25p mitotic inducer, CDK-mediated phosphorylation is the signal to bring it to full activity (Hoffmann et al., 1993; Kovelman and Russell, 1996). Given that our in vivo biochemical analysis strongly suggests that once Flp1p (a Ser/Thr phosphatase) dephosphorylates Cdc25p (a Tyr phosphatase) the mitotic inducer becomes unstable, we speculate that the signal that targets Cdc25p for destruction will be given by Flp1p-mediated dephosphorylation. Future studies will test this hypothesis.

We are grateful to J. Correa, A. Calzada, R. Daga, K. Labib and M. Sacristán for helpful discussions and comments on the manuscript. A.B. received financial support for this research work through PGC and JCyL grants from the Spanish Science and Technology Ministry and Junta de Castilla y León. This research was also supported by The Swiss National Science Foundation, ISREC and PGC of the Spanish Science and Technology Ministry grants to V.S. and S.M. This work was also supported by FPU/MEC and EMBO short-term fellowships to V.E.