Regulation of p34^cdc2 protein kinase: new insights into protein phosphorylation and the cell cycle

The reversible phosphorylation of specific residues on proteins is a ubiquitous mechanism for the regulation of cellular processes. Recently there has been a tremendous expansion of interest in the role of protein phosphorylation in the eukaryotic cell cycle. This has been a consequence of the discovery that the cdc2 gene in the fission yeast Schizosaccharomyces pombe, which is required at the onset of mitosis and also at the ‘start’ control point at the G₁/S transition, encodes a 34kDa protein-serine/ threonine kinase catalytic subunit (p34^cdc2). Homologous genes have been identified in Saccharomyces cerevisiae (CDC28) and in all other eukaryotes examined. p34^cdc2 is a component of MPF (maturation- or M-phase promoting factor), and it is now apparent that p34^cdc2 protein kinase is central to the control mechanism of the cell cycle in all eukaryotic cells (reviewed by Nurse, 1990).

The mitotic function of p34^cdc2 protein kinase has been somewhat better characterized than its role at ‘start’. At the G₂/M transition, p34^cdc2 protein kinase is specifically and dramatically activated, and catalyses the phosphorylation of key proteins (reviewed by Moreno and Nurse, 1990; Pines and Hunter, 19906) to bring about the cellular changes that occur in mitosis. The level of p34^cdc2 protein is constant throughout the cell cycle, and its protein kinase activity (usually assayed in vitro using histone Hl as substrate) requires the association of p34^cdc2 with cyclins, proteins that are synthesised and degraded in a cell cycle-dependent manner (Evans et al. 1983). Activation also involves a complex series of phosphorylation/ dephosphorylation events catalysed by a network of protein kinases and protein phosphatases. Although the picture is still incomplete, important developments have occurred recently in our understanding of these molecular events. Some of the protein kinases and protein phosphatases involved in controlling the phosphorylation of p34^cdc2 itself have been tentatively identified and found to recognise both serine/threonine and tyrosine residues. This unexpected finding calls for a revision of our previous assumptions about the substrate specificity of such enzymes. In fact, the study of the cell cycle is now yielding new insights into the intricate mechanisms by which protein phosphorylation provides molecular switches and precise temporal control of cellular processes.

p34^cdc2 is itself phosphorylated at multiple sites in a cell cycle-dependent manner. Although the phosphorylation sites have not been directly sequenced, four have been identified by a combination of genetic methods and phosphopeptide mapping. Putative phosphorylated residues were mutated to non-phosphorylatable residues and the resulting phenotype examined. In parallel, phosphopeptides derived from p34^cdc2 were compared with synthetic peptides phosphorylated in vitro. The first phosphorylation site to be identified in this way was Tyrl5 in S. pombe p34^cdc2 (Fig. 1). Mutation of Tyrl5 to phenylalanine causes the cells to enter mitosis prematurely, suggesting that this mutation causes premature activation of p34^cdc2 protein-kinase (Gould and Nurse, 1989). Furthermore, a purified human protein-tyrosine phosphatase can remove the tyrosine phosphate from wild-type p34^cdc2 isolated from cells in Gr, and activate its histone Hl kinase activity (Gould et al. 1990). These results demonstrate that dephosphorylation on Tyrl5 alone is sufficient to activate the Gr form of S. pombe p34^cdc2 protein kinase.

In vertebrates, the regulation of p34^cdc2 protein kinase activity by phosphorylation on Tyrl5 is conserved. However, there is an additional inhibitory phosphorylation site on the adjacent residue, Thrl4 (Fig. 1). p34^cdc2 phosphorylated on both Thrl4 and Tyrl5 can be isolated from chicken (Krek and Nigg, 1991a) and mouse (Norbury et al. 1991) cultured cells in S and Gr phases. Norbury et al. (1991) have examined the activation of wild-type and mutant human p34^cdc2 expressed in Xenopus extracts in the presence of cyclin B. The activity of the wild-type human p34^cdc2 protein kinase is suppressed like the endogenous p34^cdc2 protein kinase, only activating after a long lag. Only when both Thrl4 and Tyrl5 are mutated to non-phosphorylatable residues (Phe and Ala, respectively) is the transient suppression of kinase activity relieved (Norbury et al. 1991). Similar results have been obtained by Krek and Nigg (1991b), who have expressed wild-type and mutant chicken cdc2 cDNAs in HeLa cells. They find that mutation of Thrl4 and Tyrl5 together to non-phosphorylatable residues induces premature mitotic events. However, the two sites are not equivalent: mutation at Thrl4 alone did not produce a phenotype, while mutation of Tyrl5 alone partially induces premature mitosis.

Thrl4 and Tyrl5 lie in the putative ATP-binding site of p34cdc2 p34^cdc2 protein kinase, and the presence of a phosphate group probably hinders the binding of ATP, inactivating the kinase. Inhibitory phosphorylation within the ATP-binding site constitutes a novel mechanism for the regulation of a protein kinase, although isocitrate dehydrogenase of Escherichia coli is regulated in a similar manner (Hurley et al. 1990).

Another putative phosphorylation site in p34^cdc2 is a threonine residue, Thrl67 in fission yeast, Thrl6l in vertebrates (Krek and Nigg, l99la) (Fig. 1). Recently, two groups have mutated Thrl61/167 in p34^cdc2 to various other residues, and examined the effect of expression in S. pombe (Ducommun et al. 1991; Gould et al. 1991). Mutation to a neutral non-phosphorylatable residue results in failure to rescue temperature-sensitive cdc2 mutants (ts-cdc2) and overexpression is lethal or leads to elongated cells containing a single nucleus. However, mutation to serine allows ts-cdc2 rescue. Replacement with glutamic acid, which might mimic a phosphorylated residue, permits some growth when expressed in the ts-cdc2 strain. Overexpression of this mutant p34^cdc2 produces cells that are multiply septated and with multiple mitotic-like nuclei. These results indicate that phosphorylation of Thrl6l/167 is required for p34^cdc2 activity, and indeed this site appears to be phosphorylated when the kinase is active during M-phase (Ducommun et al. 1991; Gould et al. 1991; Krek and Nigg, l99la).

Phosphorylation at a further site, Ser277 (chicken cdc2) (Fig. 1) occurs during G?, and decreases on entry into S phase (Krek and Nigg, l99la). Although the effect of this phosphorylation is, at present, unknown, it might be involved in the proposed function of cdc2 at the G₁/S transition.

The form of p34^cdc2 protein kinase that is active in M-phase consists of a complex of p34^cdc2 and cyclin B (Draetta et al. 1989; Labbé et al. 1989). The cyclin subunit is also phosphorylated, although the possible function of this phosphorylation is still unclear. Xenopus cyclin B2 can be phosphorylated in. vitro by the product of the mos proto-oncogene, a serine/threonine protein kinase (Roy et al. 1990). It has been proposed that this may stabilize the active p34^cdc2/cyclin B2 complex and account for the effect of mos as the ‘cytostatic factor’ responsible for the arrest of unfertilized eggs in second meiotic metaphase (Mailer, 1991). Conversely, it is possible that phosphorylation of cyclin by p34^cdc2 protein kinase itself acts as a signal for cyclin degradation, p34^cdc2 protein kinase inactivation and exit from M-phase (Felix et al. 1990b).

Until very recently, known protein kinases could be divided into two classes on the basis of their substrate specificity, having a mutually exclusive preference for either tyrosine or serine/threonine residues. This specificity was reflected in the conservation of certain residues in the primary sequences of kinases belonging to each group (Hanks et al. 1988). However, several protein kinases have now been described that phosphorylate both classes of residues, although their sequences resemble serine/threonine-specific kinases (Ben-David et al. 1991; Dailey et al. 1991; Howell et al. 1991; Stern et al. 1991). One of these, the 107 kDa product of the weel⁺ gene in S. pombe (p107^weel), is a negative regulator of entry into mitosis (Russell and Nurse, 1987b). p107^weel can autophosphorylate on tyrosine and serine residues, and phosphorylates an exogenous peptide substrate on tyrosine (Featherstone and Russell, 1991). P107^weel has not been shown to phosphorylate p34^cdc2 directly, but co-expression of p34^cdc2 with p107’^u∞7 in insect cells using a baculovirus vector results in phosphorylation of p34^cdc2 on tyrosine. Moreover, this phosphorylation is greatly stimulated by coexpressed cyclin (Parker et al. 1991). Although non-physiological phosphorylations could occur when kinases are vastly over-produced in this system, these results strongly suggest that p107^weelprotein kinase catalyses the tyrosine phosphorylation of p34^cdc2, and that the p34^cdc2/ cyclin complex is a better substrate than p34^cdc2” alone. Although p107^weelfrom S. pombe phosphorylates serine/ threonine and tyrosine residues in vitro, Thrl4 has not been found phosphorylated in S. pombe p34^cdc2. Tyrl5 phosphorylation involves the cooperation of p107^weelwith another putative protein kinase, the product of the mik1⁺ gene (Lundgren et al. 1991). In vertebrates both Thrl4 and Tyrl5 are phosphorylated, but it remains to be seen whether phosphorylation of both sites is catalysed by a homologue of p107^weelor by two distinct kinases. Perhaps distinct inhibitory pathways could act on each site.

There are certain parallels between the phosphorylation of p34^cdc2 and some other protein-serine/threonine kinases. Mitogen-activated protein (MAP) kinases p42^mapk and p44^mapk (Sturgill and Wu, 1991) are members of a family of protein-serine/threonine kinases that includes the extracellular signal-regulated kinases (ERK) (Boulton et al. 1991), yeast kinases KSS1 (Courchesne et al. 1989) and FUSS (Elion et al. 1990), and probably several other vertebrate kinases usually assayed using microtubule-associated protein-2 (MAP-2) or myelin-basic protein (MBP) as exogenous substrates, e.g. (Ahn et al. 1990; Gotoh et al. 1991; Sanghera et al. 1990). These kinases, which are related to the cdc2 family (26–41 % identity for ERKs; Boulton et al. 1991), play an important role in the cellular response to a variety of extracellular signals, forming part of the signal transduction pathway from receptor tyrosine kinases to serine/threonine phosphorylation of target proteins. They may also be involved in regulation of the cell cycle, being activated during M-phase (Sturgill and Wu, 1991). Like p34^cdc2, the activity of MAP kinases is controlled by both tyrosine and threonine phosphorylation, although in this case the phosphorylations are not inhibitory, but required for activation. Recently, the regulatory phosphorylation sites on p42^mapk have been identified by peptide sequencing using mass spectrometry (Payne et al. 1991). Two phosphorylated residues, one tyrosine and one threonine, are separated by a single glutamic acid residue and situated near to the conserved kinase subdomain VIII (Hanks et al. 1988) (Fig. 1). The corresponding region in p34^cdc2 contains the Thr161/167 phosphorylation site (Fig. 1). In fact, the double activating phosphorylations in p42^mapk may be considered homologous to the single activating (Thrl61/167) phosphorylation in p34^cdc2. In some other protein kinases auto-phosphorylation sites are found in this region. Dephosphorylation of p42^mapk on either threonine or tyrosine, using protein serine/ threonine phosphatase 2A or a tyrosine-specific phosphatase, respectively, inactivates the kinase in vitro (Anderson et al. 1990). However, the phosphatase(s) and kinase(s) that act in vivo on these sites in p42^mapk are unknown. Growth factor-stimulated protein kinases related to p42^mapk may be phosphorylated on both tyrosine and threonine residues and activated in response to treatment with a single factor (Ahn et al. 1991; Gomez and Cohen, 1991). It has been suggested that this factor may not itself be a kinase, but rather stimulates autophosphorylation of MAP kinases (Seger et al. 1991).

Tight control of the activity of p34^cdc2 and MAP kinases is essential to prevent the cell from unregulated growth and division. In the case of p42’^mapk, phosphorylation at both threonine and tyrosine sites is required for activity, so that unregulated phosphorylation by kinases specific for tyrosine or threonine/serine residues alone will not result in activation. For vertebrate p34^cdc2, it is dephosphorylation on both threonine and tyrosine residues that is required for activation, providing a double check against unwarrented activation. S. pombe p34^cdc2 does not seem to have the dual tyrosine/threonine inhibitory phosphorylations, however, and is activated by dephosphorylation on tyrosine alone (Gould and Nurse, 1989). Tyrosine phosphorylation may be very rare in S. pombe, and it is possible that the extra safety catch of inhibitory threonine phosphorylation on p34^cdc2 is not required.

MAP kinases, like p34^cdc2, may recognise certain serine/threonine residues followed by a proline in their substrates, although a second proline, two residues N-terminal to the phosphorylated residue, may also be required (Sturgill and Wu, 1991). It has been suggested that ‘proline-directed phosphorylations may constitute a sub-set of phosphorylation sites on proteins involved in growth control and mitotic induction (Hall and Vulliet, 1991). There are some clear differences in the ability of p34^cdc2 and MAP kinases to phosphorylate certain proteins in vitro’, however, conclusions about the identity of the kinase that phosphorylates a particular serine/threo-nine-proline site in vivo must be interpreted with caution until we know the precise specificity determinants.

Protein phosphatases have been divided, like protein kinases, into two major classes: those that dephosphorylate tyrosine residues and those with a preference for serine/threonine residues (Cohen, 1989; Tonks and Charbonneau, 1989). Unlike protein kinases, however, the two classes of phosphatase do not show sequence homology (Cohen et al. 1990; Tonks, 1990). Recently, the situation has changed somewhat with the discovery that vaccinia virus VH1 gene encodes a protein phosphatase that shows some sequence homology with tyrosine phosphatases, although it can dephosphorylate well both phosphotyrosine and phosphoserine residues (Guan et al. 1991). So, as for protein kinases, a third class of protein phosphatase may be emerging.

The product of the cdc25 gene in S. pombe is a positive regulator of cdc2. Related genes have been identified in Saccharomyces cerevisiae, Drosophila, Xenopus and humans. Genetic evidence has suggested that cdc25 could function in the dephosphorylation and activation of p34^cdc2 at the G₂-M transition. Recently, this has been substantiated by the finding that a Drosophila homologue of cdc25 activated p34^cdc2 when added to Xenopus egg extracts. Activation was concomitant with dephosphorylation of p34^cdc2 on tyrosine (Kumagai and Dunphy, 1991). In fact, purified p34^cdc2/cyclin B complex isolated in an inactive form can be activated and dephosphorylated on tyrosine by incubation with a purified human cdc25 homologue (Stausfeld et al. 1991). These results strongly suggest that cdc25 can catalyse the dephosphorylation of p34^cdc2 directly.

Initially, cdc25 was not thought to be a p34^cdc2-tyrosine phosphatase, because no sequence similarity with known tyrosine phosphatases could be discerned (Nurse, 1990). However, the vaccinia VH1 gene product contains only a few residues in common with other tyrosine phosphatases. A closer inspection of the amino acid sequence of cdc25 does reveal some sequence similarity with VH1 phosphatase and other tyrosine phosphatases (Moreno and Nurse, 1991; Fig. 2). In particular, a motif His-Cys-(5 residues)-Arg is present in all of them. In VH1 phosphatase (Guan et al. 1991) and other tyrosine phosphatases (Sreuli et al. 1990) the Cys and Arg residues in this motif are essential for activity. Mutation of the corresponding residues in cdc25 prevents its activation of p34^cdc2 (M. W. Kirschner, personal communication), indicating that cdc25 also has phosphatase activity, cdc25 has not been shown to produce the efficient dephosphorylation of any substrate other than p34^cdc2 (on tyrosine) when it is complexed to cyclin, and so may be highly substrate-specific. Whether cdc25 is able to dephosphorylate the inhibitory phosphorylated threonine residue in vertebrate p34^cdc2 is not certain, although there is some evidence to support this notion (Stausfeld et al. 1991).

Okadaic acid, which specifically inhibits type 1 and 2A serine/threonine phosphatases, causes the premature activation of cdc2 in extracts of Xenopus eggs. The type 1 phosphatase-specific protein inhibitors 1 and 2 do not, indicating that the effect is due to the inhibition of phosphatase 2A (Felix et al. 1990a; reviewed by Karsenti et al. 1991). In extracts without endogenous cyclins, but to which bacterially expressed cyclin B has been added, okadaic acid abolishes the normal requirement for a threshold of cyclin concentration and reduces the time lag between the addition of cyclin and kinase activation (Solomon et al. 1990). Recently, a factor that inhibits the activation of p34^cdc2 (termed INH) has been purified and found to contain the catalytic subunit of phosphatase 2A as well as other components (Lee etal. 1991). However, the target of phosphatase 2A is not certain. Because inhibition of phosphatase 2A stimulates the dephosphorylation of p34^cdc2 on tyrosine, it has been proposed that the activity of a tyrosine phosphatase (probably a cdc25 homologue) is reduced by dephosphorylation catalysed by phosphatase 2A (Fig. 3; reviewed by Karsenti et al. 1991). Although it is not clear whether cdc25 is regulated by phosphorylation, it is a phosphoprotein (Moreno et al. 1990). In vitro, phosphatase 2A can also dephosphorylate and inactivate p34^cdc2 directly (Lee et al. 1991; Gould et al. 1991). Since Thrl61/167 phosphorylation seems to be required for tight association between p34^cdc2 and cyclin (Ducommun et al. 1991), dephosphorylation of that site by phosphatase 2A could cause dissociation of the complex. However, it is not at all certain that phosphatase 2A acts in this way in vivo. Phosphatase 2A might also activate a tyrosine kinase, probably p107^weel, to maintain p34^cdc2 in an inactive state (Fig. 3). Indeed, genetic evidence suggests that weel is repressed by phosphorylation due to another kinase, the product of the niml gene (Russell and Nurse, 1987a). Further investigation is required to determine the actual component(s) of the system on which phosphatase 2A acts in the cell. This is an important question, because phosphatase 2A seems to be part of the Gr checkpoint pathway that monitors when p34^cdc2/cyclin B can be activated.

The mechanism of activation of p34^cdc2 complexed to cyclin B is intricate and not yet completely understood. The Thrl6l/167 phosphorylation site in p34^cdc2 is required for tight cyclin binding (Ducommun et al. 1991) and activation of the kinase (Gould et al. 1991). It is not yet clear whether phosphorylation of this site precedes cyclin B association or is a consequence of it, but the two events may be cooperative. For instance, cyclin binding to p34^cdc2 could increase the rate of Thrl61/167 phosphorylation, which may then stabilize the complex. Cyclin B binding also induces inhibitory tyrosine phosphorylation (Tyrl5) (Solomon et al. 1990; Meijer et al. 1991), but the timing of Thrl4 phosphorylation and its possible dependence on cyclin B binding is uncertain. As cyclin B is synthesised it associates with p34^cdc2 in an inactive complex that accumulates before the final activation leading to metaphase. Why should it be necessary to have inhibitory phosphorylation of p34^cdc2 after cyclin binding? One answer is that each regulatory phosphorylation on the complex can be used as a checkpoint, sensing the accomplishment of S-prophase processes before permitting activation of the mitotic form of the kinase (Enoch and Nurse, 1991). Recent genetic (Enoch and Nurse, 1990) and biochemical (Dasso and Newport, 1990; Kumagai and Dunphy, 1991) evidence indicates that the tyrosine phosphorylation of p34^cdc2 is affected by the state of replication of DNA; if DNA replication is blocked then tyrosine phosphorylation occurs normally but dephosphorylation does not take place and the p34^cdc2 protein kinase remains inactive, cdc25 is somehow involved in this check point (Enoch and Nurse, 1990). Unreplicated DNA could act by inhibiting the activity of cdc25, or by activating the opposing tyrosine kinase (Fig. 3). In-triguingly, in Xenopus egg extracts to which a large number of nuclei have been added, when traffic through nuclear pores is blocked, p34^cdc2 is tyrosine phosphorylated normally but not dephosphorylated (Kumagai and Dunphy, 1991). This may suggest that a regulator of cdc25 or cdc25 itself shuttles between nucleus and cytoplasm, coupling the tyrosine dephosphorylation of p34^cdc2 to the completion of DNA replication.

However, tyrosine phosphorylation may not be an obligatory step in the pathway of p34^cdc2 protein kinase activation. In the cleaving Xenopus embryo, after the 1st mitotic cycle, the cell divides rapidly from the 2nd to 12th mitotic cell cycles without G₁ or G₂ phases. During this period, p34^cdc2 protein kinase oscillates between inactive and active forms, although tyrosine phosphorylation is not detected (Ferrell et al. 1991). The lack of inhibitory tyrosine phosphorylation could explain why the activation of p34^cdc2protein kinase is not dependent on the completion of DNA replication until the midblastula transition.

In this commentary, we have discussed the regulation of the activity of p34^cdc2 protein kinase by its association with regulatory subunits, in particular B-type cyclins, and by phosphorylation/dephosphorylation of the p34^cdc2 catalytic subunit. However, catalytic subunits of other putative cyclin-dependent protein kinases, closely related to p34^cdc2, are being identified (Lehner and O’Farrell, 1990; Paris et al. 1991; Pines and Hunter, 1990a). In addition to the specific regulation of p34^cdc2/cyclin B at the G₂-M transition, cell cycle control probably involves other protein kinases consisting of combinations of p34^cdc2-like catalytic subunits with different cyclins. Cyclin-like proteins specific of the G1phase have been identified in different species (Forsburg and Nurse, 1991; Richardson et al. 1989; Wittenberg et al. 1990), and during the G₂-M period, there are both A- and B-type cyclins (Draetta, 1990). These subunits may confer specific functions to a protein kinase and target it to particular subcellular compartments. Each kinase may be recognised differently by regulatory kinases and phosphatases, and hence be differentially controlled. Each kinase may prime the activation of the next kinase in the sequence (Minshull et al. 1990), so that a successsion of phases is assured. In fact, we are only just beginning to uncover the complexity of protein phosphorylation in the cell cycle.

We thank those authors who communicated their results prior to publication. Work in this laboratory is supported by a grant from the Human Frontiers Science Programme. P.R.C. is the recipient of a research fellowship of The Wellcome Trust.