Calcium and cell cycle control

There are two sides to cell division cycle control. One is a cell’s decision from the outside about whether to enter the cell cycle. The other, on the inside, is the decision about when key cell cycle events, such as mitosis, should take place. In deciding to enter the cell cycle, the cell responds to external signals such as growth factors. Decisions on the inside are taken on the basis of inside information like cell size or the completion of genome replication. We shall discuss cell division cycle control from this internal perspective.

The idea that there are particular points during the cell cycle when a cell must make up its mind and commit itself to one particular outcome or another has arisen from work on yeast. A large number of mutant strains of yeast with anomalous cell cycle control has been selected. Three cell cycle control points have been identified by analysis of mutant phenotypes: START, at which a decision between mitosis, meiosis, arrest or conjugation is taken; ENTRY into mitosis after DNA synthesis is completed and EXIT from mitosis once chromosome segregation has occurred. The yeast cell cycle mutants have also provided the opportunity to isolate cell cycle control proteins, using molecular genetics techniques to identify the mutant genes. It is now evident that all eukaryotic cells possess close homologues of the key yeast cell cycle control proteins.

It’s very satisfying to come across so telling an example of what used to be called the Uniformity of Nature. It makes it so much easier to justify studying cell cycle control in frog, starfish, clam and sea urchin oocytes, eggs and embryos. Early embryos have a very rapid and straightforward cell cycle. Immature oocytes are arrested at meiosis ENTRY and resume the cell cycle in response to hormones. In many cases, mature oocytes arrest at mitosis EXIT until they are fertilized. Cell cycle control points can be defined easily in eggs and oocytes because these cells have evolved natural cell cycle pauses from which they are released by an external signal. The natural breaks in the cell cycle in eggs and oocytes are specific adaptations of cell cycle control points that exist in all dividing cells. Our experiments with early sea urchin embryos show that control points analogous to START, mitosis ENTRY and mitosis EXIT are found in the sea urchin. We shall discuss the idea that the most basic internal cell cycle control mechanisms involve the generation of internal signals that govern the transition from one phase of the cell cycle to the next. The virtue of the basic cell cycle in yeast is that it is amenable to analysis by genetic techniques. The virtue of the basic cell cycle in eggs and embryos is that it can be analysed at the level of the cell’s physiology, in terms of the cell’s internal signalling systems.

A useful way of thinking about cytoplasmic signalling is in terms of messengers that are generated within the cell to hit targets. The targets in cell cycle signalling are, by definition, a subset of the cell cycle control proteins. The messengers in cell cycle signalling are all or a subset of the second messengers that have been identified by studying how cells respond to hormones. Calcium is a cell messenger that seems to play a central part in cell cycle control. Much of the evidence that points to calcium as a signal that triggers cell cycle START, mitosis ENTRY and mitosis EXIT comes from work on sea urchin eggs and embryos.

A cell must generally accomplish four things during the cell cycle: grow, replicate its DNA, duplicate its centrosomes and segregate its chromosomes. In general, the decision to begin DNA synthesis (S phase) is made by monitoring cell growth (Killander and Zetterberg, 1965; Brookes and Shields, 1985); passage through mitosis ENTRY is triggered only when DNA synthesis is completed (Pardee, 1974; Pardee et al. 1978) and requires that centrosome duplication has occurred (Rattner and Phillips, 1973; Kuriyama and Borisy, 1981). Mitosis EXIT awaits the condensation and assembly of chromosomes on the mitotic spindle (Sluder, 1979). The G, S and M phases of the cell cycle (Fig. 1) can be thought of as different cell states. The cell cycle control points are situated just prior to state transitions. START is in G₁, just before the S-phase transition, mitosis ENTRY in G₂, just prior to mitosis onset and mitosis EXIT is in M, probably at the metaphase-anaphase transition. The pattern is most obvious in yeast (Fig. 1), where cell cycle mutants have been found that enter S phase at the wrong cell size (Nurse, 1975; Nurse and Thuriaux, 1980) (I), or fail to enter (Nurse and Bissett, 1981; Russell and Nurse, 1986; Booher and Beach, 1988) (II) or exit (Moreno et al. 1988) (III) mitosis.

This pattern is also found in eggs and oocytes. The cell cycle pauses after S phase in immature oocytes (Kanatani, 1973; Masui and Clarke, 1979; Meijer and Guerrier, 1985) while waiting for the hormonal signal to resume. There is a second hiatus after oocyte maturation in frog and mammalian eggs at mitosis EXIT as the cell cycle waits for the egg to be fertilized (Whitaker, 1989). The sea urchin egg is arrested in interphase and STARTs the cell cycle at fertilization. Eggs and oocytes differ from somatic cells in two important respects. They do not grow: the early embryo is the same size as the egg. Nor do they think of leaving the cell cycle: there is nowhere yet for them to go.

The similarities between cell cycle control points in yeast and in oocytes are striking. It is also evident that the internal control points can be governed by either internal (size control in yeast) or external signals (the hormones and sperm in eggs and oocytes). The evidence for similar cell cycle control points in mammalian somatic cells is on the whole more indirect, but the general scheme does not need too much Procrustean stretching to apply to their cell cycle too. There is good evidence for a cell cycle commitment point analogous to yeast START in late G₁ of the somatic cell cycle (Pardee, 1974). To find cell cycle arrest at other points is rare or abnormal. The best argument for START, mitosis ENTRY and mitosis EXIT as control points in the mammalian cell cycle is that the cell cycle control proteins isolated from yeast and oocytes have their exact equivalents in mammalian cells (Lee and Nurse, 1987; Hunt, 1989).

The unfertilized sea urchin egg waits in interphase for the sperm. It is haploid and metabolically quiescent. Cell cycle progression (START) is triggered by fertilization. 30 min after fertilization S phase begins and is completed at 60 min. Mitosis onset is marked by dissolution of the nuclear envelope (nuclear envelope break-down. NEB) and chromatin condensation, 70 min after fertilization. Mitotic anaphase occurs at 90min and the nucleus reforms at 110 min. Only the first cell cycle shows a lengthy G₁ phase. In subsequent cycles, G₁ is almost elided, START follows quickly on mitosis EXIT and S phase of the next cell cycle begins very soon after mitotic telophase. S phase and M phase follow one another helter skelter until the mid-blastula stage (12 h after fertilization), when the cell cycle lengthens, regains its longer G₁ and G₂ phases and paternal genes are first transcribed. It is perhaps worth emphasizing that the early embryonic cell cycles take place without transcription: proteins are made from maternal message laid down as the oocyte matures (Davidson et al. 1982).

The cell cycle resumes abruptly at fertilization: within 10 min metabolism and protein synthesis rates have increased 20-fold. The cell signal responsible for cell cycle progression is a transient increase in intracellular free calcium concentration, [Ca]_i.

[Ca]_i increases rapidly at fertilization from its resting value of 0.1 μM to 1-5 μM and declines towards the resting value over a period of 5-10 min (Steinhardt et al. 1977; Fig. 2). The source of calcium is intracellular, since the Ca_i transient is not affected by removing calcium from the extracellular medium (Chambers, 1980; Chambers and Angeloni, 1981; Schmidt et al. 1982; Crossley el al. 1988). Some experiments done by Steinhardt and Epel in the 1970s (Steinhardt and Epel, 1974; Steinhardt et al. 1976) indicate that the increase in [Ca]_i is a necessary and sufficient signal for resumption of the cell cycle (Whitaker and Steinhardt, 1982). Calcium chelators that prevent the Ca_i increase block cell cycle progression (Zucker and Steinhardt, 1978). Causing an artificial Cai increase with the calcium ionophore A23187 leads to a metabolic activation and resumption of the cell cycle identical to activation at fertilization (Steinhardt and Epel, 1974; Whitaker and Steinhardt, 1982). Using a trick or two in addition, parthenogenetic activation with A23187 leads to normal (though haploid) embryonic development (Brandriff et al. 1975). These experiments demonstrate that the Ca_i transient is the sole causal agent for cell cycle resumption at fertilization.

One major and long-lasting consequence of the transient Ca_i increase at fertilization is a sustained increase in intracellular pH (Shen and Steinhardt, 1978; Johnson et al. 1976; Johnson and Epel, 1981 and Fig. 2). The unfertilized egg cytoplasm is unusually acidic (pH 6.7-6.8). A normal cytoplasmic pH (pH7.2-7.3) is restored at fertilization by the activation of a Na-H antiporter in the egg plasma membrane (Johnson et al. 1976; Payan et al. 1983). The pH_i signal is an essential component of cell cycle onset. Artificially maintaining the cytoplasm at an acidic pH (Grainger et al. 1979) or preventing the extrusion of H⁺ by removing external Na⁺ or inhibiting the antiporter (Chambers, 1976; Johnson et al. 1976; Shen and Steinhardt, 1979) blocks cell cycle progression. The intracellular pH controls the rate and extent of protein synthesis (Grainger et al. 1979; Winkler et al. 1980; Dube et al. 1985). The increase in the rate of protein synthesis at fertilization is largely due to the removal of the metabolic brake represented by the unusually acidic pH of the unfertilized egg (Whitaker and Steinhardt, 1982).

Microinjection of the phosphoinositide messenger InsP₃ stimulates resumption of the cell cycle in sea urchin eggs (Whitaker and Irvine, 1984; Fig. 4). This was, incidentally, the first report of the consequences of InsP₃ injection in a living cell. InsP₃ releases calcium from an internal store within the egg (Clapper and Lee, 1985; Crossley et al. 1988) and causes a Ca_i transient that is very similar to the Ca_i transient at fertilization in magnitude and duration (Swann and Whitaker, 1986; Swann et al. 1987). Blocking InsP₃ production with neomycin (a not-altogether specific inhibitor of PtdInsP₂ hydrolysis) prevents cell cycle onset (Swann and Whitaker, 1986).

The other arm of the phosphoinositide signalling system is DAG (diacylglycerol). Synthetic DAG activates the Na-H antiporter, causing cytoplasmic alkalinization (Shen and Burgart, 1986; Lau et al. 1986), as does PM A (Swann and Whitaker, 1985; Lau et al. 1986), a tumour promoter that stimulates protein kinase C (Nishizuka, 1984). Metabolism of polyphosphoinositide (PPI) lipids increases dramatically at fertilization (Turner et al. 1984) and both InsP₃ and DAG rise and fall in concert with the Ca_i transient (Ciapa and Whitaker, 1986). This seems to be the result of a positive feedback mechanism in which Ca_i increases stimulate further InsP₃ production (Whitaker and Aitchison, 1985; Swann and Whitaker, 1986). The messenger pathways are illustrated in Fig. 3.

There are very clear parallels between InsP₃ and DAG production at fertilization and the response of cells to hormones and growth factors (Berridge, 1987a; 1987b). In both circumstances, an external messenger (the growth factor or the sperm) stimulate changes in the phosphoinositide cell messengers (InsP₃, DAG, Ca_i). The PPI messengers must then search out the specific protein targets that are responsible for cell cycle onset. This analogy appears to be imperfect, though, in at least one respect. Signal transduction of hormone and growth factor signals across the plasma membrane involves the interaction of receptors with GTP-binding proteins at the inner surface of the plasma membrane (Gilman, 1987; Cockcroft and Stutchfield, 1988). It has been suggested that signal transduction at fertilization operates through a GTP-binding protein (Turner et al. 1986; Turner et al. 1987), but it seems rather that the sperm triggers the concerted increase in InsP₃, Ca_i and DAG by another route (Whitaker et al. 1989).

Whether the Ca_i transient in the unfertilized egg is elicited by natural means, or artificially by InsP₃ microinjection or calcium ionophore treatment, the consequence is that the cell cycle resumes and mitosis onset occurs 80-100min later (Fig. 4). These experiments also illustrate the point that the sperm’s main contribution to cell cycle control is to trigger the calcium transient, since eliciting the calcium transient with InsP₃ or the calcium ionophore A21387 also cause resumption of the cell cycle. Nor do markedly polyspermie eggs show an altered cell cycle timing. The absence or excess of the sperm’s DNA does not appear to affect the timing of mitosis onset. Any cell cycle anomalies in parthenogenetically activated sea urchin eggs are due to the lack of the sperm’s centriole without which a mitotic spindle cannot form (Brandriff et al. 1987). Indeed, the nuclear cycle of S phase and chromatin condensation does not require a functional mitotic spindle (Mazia, 1974; Mazia and Ruby, 1974; Patel et al. 1989a,b).

Transient calcium signals occur throughout the first cell cycle in sea urchin embryos (Poenie et al. 1985 and Fig. 5). In a way, the calcium signals later in cell cycle are more interesting than the START signal because they are an example of antonomous internal cell signalling. The evidence that points to the existence of a calcium signal at mitosis entry is shown schematically in Fig. 6. We can measure an increase in [Ca]_i in eggs 1-2 min before normal mitosis ENTRY (Figs 5 and 6). In polyspermie eggs, the transient is large and easily detected using whole cell fura2 recording techniques. Causing a premature increase in Ca_i by microinjecting InsP₃ or calcium leads to a rapid and precocious mitosis ENTRY (Twigg et al. 1988). Mitosis ENTRY is blocked by microinjection of the calcium chelators EGTA or BAPTA and the block can be relieved by microinjecting calcium (Twigg et al. 1988; Steinhardt and Aiderton, 1988). These experiments show that the Ca_i transient at mitosis ENTRY is a necessary signal for mitosis onset. It is also a sufficient signal, provided the egg has passed START and the protein synthesis requirement is met (Twigg et al. 1988; Patel et al. 1989 and Fig. 10).

The immediate target of the mitosis ENTRY Ca_i transient in sea urchin embryos is calmodulin and a type II calmodulin-regulated kinase (CAM-II kinase). Microinjection of antibodies or inhibitory peptides directed against the CAM-II kinase arrests the cell cycle at mitosis ENTRY (Baitinger et al. 1989; Baitinger er al. 1990).

The timing of mitosis ENTRY is independent of the degree of polyspermy at fertilization (Fig. 4). In contrast, the mitosis entry Ca_i transient is much larger in polyspermie than in monospermic eggs; it is often not easily detected in whole cell fura2 experiments after monospermic fertilization (Fig. 5) and is sometimes undetectable. Strictly speaking, something that cannot be detected doesn’t exist. For obvious reasons, we prefer the idea that, in many normal, monospermic embryos, the Ca_i transient at mitosis entry is localized to the area of the nucleus. This would make it difficult to pick out from the noise in fura2 measurements. It seems that the mitosis ENTRY Ca_i transient is amplified when several sperm nuclei are present in the cytoplasm, though the underlying cell cycle timing mechanisms that lead to the generation of the calcium transient are unaffected.

Surf clam (Spisula) oocytes ENTER meiosis immediately after fertilization in response to an internal signal triggered by sperm (Fig. 1). Cell cycle onset through meiosis ENTRY is also triggered in these eggs by microinjection of InsP₃ (Bloom et al. 1988), presumably because InsP₃ causes a Ca_i transient. An increase in PPI turnover has been measured at fertilization (Eckberg and Szuts, 1987) and it has been suggested that DAG production leads to cytoplasmic alkalinization via protein kinase C and a Na-H antiporter, as in sea urchin eggs (Dube, 1988). Calmodulin antagonists prevent meiosis onset (Carroll and Eckberg, 1986). There are strong similarities, therefore, with the cell signals that operate at START in sea urchin eggs, though in Spisula the source of the activating calcium is extracellular (Dube, 1988).

Ca_i transients occur in sea urchin embryos at mitosis EXIT (Poenie et al. 1985). Figure 5 shows the timing of the transient. Embryos micro-injected during mitosis with the calcium chelator BAPTA arrest in mitosis, usually at mitotic metaphase (Fig. 11). These observations locate the mitosis EXIT control point at the metaphase-anaphase transition.

EXIT from meiosis is also controlled by an internal Ca_i signal in mammalian and frog eggs. The pre-fertilization arrest is at second meiotic metaphase (Fig. 1). Anaphase resumes and the chromosomes move apart and decondense in response to the fertilization Ca_i transient (Busa and Nuccitelli, 1985; Miyazaki et al. 1986; Kubota et al. 1987). The calcium chelator EGTA prevents the cell cycle from resuming when injected prior to fertilization (Kline, 1988; Miyazaki and Igusa, 1982). The cell messengers responsible may well be InsP₃ and DAG, as in sea urchin eggs, since microinjection of InsP₃ triggers meiosis EXIT in both cases (Busa et al. 1985; Miyazaki, 1988; Whitaker, 1989).

Cell cycle Ca_i transients occur in somatic mammalian cells. The Ca_i transient at mitosis EXIT has been studied the most. Full-blown Ca_i transients have been recorded in mammalian (Poenie et al. 1985 ; Poenie et al. 1986) and plant cells (Hepler and Callahan, 1987; Keith et al. 1985). Injection of calcium chelators retards mitosis EXIT (Izant, 1983). Blocking calcium entry prevents mitosis EXIT in plant cells (Hepler, 1985). Calcium gradients centred on the mitotic pole have been seen (Ratan and Shelanski, 1986). These observations are consistent with a Ca_i signal at mitosis EXIT in somatic cells. On the other hand, it has been argued that calcium is merely a permissive requirement for mitosis exit, on the grounds that Ca, transients are not always observed during passage through mitosis EXIT (Toombes and Borisy, 1989) or, again, that the Ca_i signal at mitosis EXIT is a slow and small increase in Ca_i in the region of the mitotic apparatus, rather than a transient (Hepler, 1989). A comparative study of mammalian cell lines has revealed that the mitosis EXIT Ca_i transient is strong and reproducible in some cell lines and sometimes undetectable in others (Aiderton et al. 1988; Kao et al. 1990). In Swiss 3T3 cells, passage through mitosis ENTRY was blocked by calcium chelators and induced by artificial Ca) transients. These data are reminiscent of our observations on the mitosis ENTRY Ca_i transient in sea urchin embryos, where we have seen clear effects with calcium injection and calcium chelators, but where the Ca_i transient is often difficult to measure (Figs 5, 6). It may be that the cell cycle Ca, transients in mammalian cells are sometimes small, local and difficult to detect.

It has been known for some time that a critical external calcium concentration is necessary for mammalian cells to pass START (Whitfield et al. 1976; Whitfield et al. 1980; Hazelton et al. 1979; Tupper et al. 1980), but attention has largely been directed at the parts played by Ca_i and PPI messenger signals earlier in the cell cycle (Berridge, 19876). The cell signals generated by hormones and growth factors represent external cell cycle control, exerted in early G_i (G_o). The growth factor signals act at an external cell cycle control point distinct from START, mitosis ENTRY and mitosis EXIT. We shall not discuss them.

A more indirect case for calcium regulation of START, mitosis ENTRY and mitosis EXIT can be made by looking at the calcium target, calmodulin. Inhibitors of calmodulin cause cell cycle arrest at both START, mitosis ENTRY and mitosis EXIT (Chafouleas et al. 1982; Sasaki and Hidaka, 1982; Eilam and Chernichovsky, 1988; Keith et al. 1983; Border et al. 1983), though the caveat that these inhibitors may have other less specific effects in living cells must always be mentioned. Overexpressing calmodulin causes cells to pass START precociously and reducing expression with antisense mRNA leads to arrest at mitosis EXIT (Rasmussen and Means, 1989a; 19896). Calmodulin localizes to the centrosome at mitosis (Welsh et al. 1979; Wolniak et al. 1980). These data are good circumstantial evidence for regulation of the cell cycle control points by Ca,, but would not convict without corroboration. Some small amount is available from experiments in yeast where in S. cerevisiae it has been shown that a patent calmodulin gene is necessary for passage through mitosis EXIT, though sadly it is not required for START progression (Ohya and Anraku, 1989).

Sea urchin embryos divide normally in sea water that lacks calcium (Chambers, 1980; Schmidt et al. 1982). This straightforward observation demonstrates that the source of calcium for cell cycle Ca_i transients is intracellular, as it is at fertilization. The unfertilized egg has a membrane-bound store that sequesters calcium via an ATP-driven calcium pump (Clapper and Lee, 1985; Payan et al. 1986). The early embryo has a store with very similar properties (Suprynowicz and Mazia, 1985) that doubtless represents the same endoplasmic reticulum, though there may be changes in subcellular distribution and activity during the cell cycle. The isolated mitotic apparatus contains a calcium store of the same genre (Silver et al. 1980) and the sequestering activity of the post-fertilization store varies during the cell cycle, being most active at mitosis (Suprynowicz and Mazia, 1985). An antibody to the Ca-ATPase from the skeletal muscle calcium store arrests the cell cycle when microinjected into early embryos (Silver, 1986) and microinjection of a monoclonal antibody raised against a 46x10³M_r component of the reticulum calcium pump arrests the cell cycle at mitosis entry by inhibiting calcium sequestration and causing deleteriously high Ca, levels (Hafner and Petzelt, 1987). Calcium stores associated with the mitotic apparatus have been found in both mammalian and plant cells (Hepler and Wolniak, 1984).

The calcium store of the early embryo also resembles the store of the unfertilized egg in its sensitivity to InsP₃. Microinjecting InsP₃ stimulates a Ca, transient at all times during the cell cycle (Twigg et al. 1988). This leads us to the presumption that the PPI messenger system controls the cell cycle Ca, transients, a presumption firmly based on the knowledge that InsP₃ is a ubiquitous trigger for calcium release from intracellular sources (Berridge and Irvine, 1989). The supposition finds some confirmation in two observations. The first is that inositol rescues lithium-arrested sea urchin embryos (Sillers and Forer, 1985). Lithium treatment depletes PPI lipids. The messenger system is repleted and restored by exogenous inositol (Berridge and Irvine, 1984). It seems that a patent PPI messenger system is essential for cell cycle progression. The second observation is that a phosphatase that degrades and inactivates InsP₃ is found concentrated on vesicles in the centrosome (Petzelt et al. 1989): the production of InsP₃ is likely to occur at a site close to or co-incident with its site of degradation.

From what we’ve said so far, it emerges that cell cycle Ca_i signals control progression through START, mitosis ENTRY and mitosis EXIT in the early embryonic cell cycles of sea urchins. We know little about how the Ca_i signals are generated. We know more about their targets. One obvious inference is that the same cell cycle signal has different effects, indicating that it cannot be affecting the same final target at each of the three control points. This might arise in a number of ways. One possibility is that each of the Ca_i transients is localized in a different part of the cell, so modifying targets that are themselves selectively localized. We suggested earlier that the mitosis ENTRY Ca_i transient may be local in extent. There is only tenuous support for this idea from calcium imaging experiments. Detectable cell cycle Ca_i transients (except, perhaps the mitosis ENTRY transient) are global (Poenie and Steinhardt, 1987; our unpublished observations), though it must be admitted that detecting very local Ca, changes in the large, spherical sea urchin embryo is difficult with present imaging techniques. Another likely explanation for the multiple effects of a single signal is that the Ca_i transient encounters different or modified targets at each of the cell cycle control points. Protein synthesis or degradation is necessary to produce different targets. We have shown that a protein or proteins synthesized after START sensitizes the nuclear membrane to the mitosis entry Ca_i transient (Twigg et al. 1988). Modified targets must be the result of posttranslational mechanisms, of which the most common are protein phosphorylation and dephosphorylation. This is indeed what seems to happen.

The targets of cell messengers are protein kinases and protein phosphatases, with very few exceptions. Where the links between cell messengers, kinases and phosphatases have been analysed in detail, the chains are branched and have not been completely untangled (Krebs, 1986; Cohen, 1988). Protein kinases and phosphatases are themselves nearly always substrates for other kinases and phosphatases. It is clear, though, that the patterns of cross-phosphorylation and cross-dephosphorylation are determined by kinases and phosphatases directly stimulated by cell messengers. CAM-kinases are regulated by calcium-calmodulin, A-Kinase by cAMP and C-kinase by calcium and acyl lipids (for example, DAG) (Schulman and Lou, 1989; Nishizuka, 1988; Beebe and Corbin, 1986). The type 2B-phosphatase calcineurin is calcium-calmodulin regulated and phosphatase 1 is regulated by cAMP through Inhibitor 1 (Cohen, 1988; Ballou and Fischer, 1986).

Interfering with protein phosphorylation or déphosphorylation arrests the cell cycle at cell cycle control points (Pondaven and Meijer, 1986; Meijer et al. 1986; Néant and Guerrier, 1988; Néant et al. 1989; Rime et al. 1989; Huchon et al. 1981; Foulkes and Mailer, 1982). Many of the cell cycle control proteins whose existence has been proven or inferred from experiments on cell cycle mutants of yeast and Aspergillus are kinases and phosphatases (Cyert and Thorner, 1989). No homologies have been reported with kinases or phosphatases that are known to be regulated by calcium, though the bimG mutant of Aspergillus and the dis2 mutant of S. pombe show homology with protein phosphatase 1. It may be for this reason, as much as for the lack of cell physiology data in yeast, that little emphasis has been given to cell messenger regulation of the yeast cell cycle, despite reports that specifically interfering with the PPI messenger system causes dose-dependent cell cycle retardation in yeast (Uno et al. 1988). The known protein kinases and phosphatases identified as exerting cell cycle control in yeast appear to be members of the class that are the indirect targets of cell messengers.

The absence of cell cycle mutants showing anomalies in genes that code for direct cell messenger targets can be taken to argue against cell cycle regulation by cell messengers in yeast. It is equally likely that mutations in such genes have such widespread consequences that they produce only banal lethal phenotypes. These two opinions illustrate two distinct views of cell cycle regulation. On one view, cell cycle control can be described by the interactions between phosphoproteins, kinases and phosphatases without the intervention of cell messenger signals. This view is illustrated by the models of Murray and Kirschner (Murray and Kirschner, 1989; Murray et al. 1989) where the emphasis is on the yeast control proteins and the reconstitution of the cell cycle in cell-free systems. The other view, taken here, is that cell messengers are central to cell cycle regulation. This view (Poenie and Steinhardt, 1987; Rasmussen and Means, 1989a; Hepler, 1989) originates from experiments on the physiology and biochemistry of the cell cycle in intact cells, as the above discussion of the calcium signal at START, mitosis ENTRY and mitosis EXIT illustrates. The two views can easily be reconciled by demonstrating the precise links between second messengers and their cell cycle control protein targets.

pp34 is the protein product of a yeast gene whose counterpart has been found to regulate cell cycle transitions in frog and starfish oocytes (Dunphy et al. 1988; Gautier et al. 1988; Labbe et al. 1989a; Labbe et al. 1989b’, Arion et al. 1988), surf clam oocytes (Draetta et al. 1989), early sea urchin embryos (Meijereta/. 1989; Patel et al. 1989b) and mammalian cells (Lee and Nurse, 1987; Draetta et al. 1987; Draetta et al. 1988; Draetta and Beach, 1988; Lee et al. 1988; Krek and Nigg, 1989). In yeast, it is required at START and mitosis ENTRY (Nurse and Thuriaux, 1980; Nurse and Bissett, 1981; Hartwell, 1974). In somatic cells it is phosphorylated on both threonine and tyrosine residues in interphase (Draetta and Beach, 1988), suggesting, by analogy with yeast, that pp34 phosphorylation controls the onset of S phase (Draetta et al. 1988). Fig. 7 shows that pp34 is phosphorylated in sea urchin eggs at fertilization (Patel et al. 1989b).

We have analysed the control of pp34 phosphorylation at fertilization by comparing the effects of various parthenogenetic agents (Fig. 7). Parthenogenetic treatments (A23187, the calcium ionophore or 15 HIM NH4CI) that lead ultimately to DNA synthesis and cell cycle progression cause rapid pp34 phosphorylation. Treatments that stimulate only the pH_i increase (2 HIM NH4CI or PMA) do not stimulate DNA synthesis nor cell cycle onset and do not lead to phosphorylation of pp34. Of the two ionic signals at fertilization, it is the Ca_i signal that is responsible, since inducing the Ca_i signal without the subsequent increase in pH_i is sufficient (Fig. 7, legend; Patel et al. 1989b). These experiments illustrate the similarities between START in the sea urchin egg and START in the yeast and mammalian cell cycles and demonstrate a link between the Ca_i signal and a pivotal cell cycle control protein.

The striking behaviour of the cyclins was first noticed in clam and sea urchin embryos (Rosenthal et al. 1980; Evans et al. 1983). They are synthesized continuously throughout the cell cycle but disappear abruptly during mitosis, suggesting that they are involved in cell cycle regulation. Cyclin homologues have been found in frog oocytes (Pines and Hunt, 1987; Minshull et al. 1989), starfish oocytes (Swenson et al. 1986; Standart et al. 1987), Drosophila embryos (Lehner and O’Farrell, 1989; Whitfield et al. 1989) and yeast (Solomon et al. 1988; Goebl and Byers, 1988). The yeast cyclin is a product of the cdc 13 cell cycle control gene that was identified by analysis of cell cycle mutants.

Unfertilized sea urchin eggs contain undetectable levels of cyclin (Tim Hunt, unpublished). The Ca_i START signal causes cyclin levels to rise by markedly stimulating the rate of protein synthesis. This is in large part due to the increase in cytoplasmic pH brought about by activation of the Na-H antiporter (Johnson et al. 1974; Payan et al. 1983). Cyclin synthesis can be triggered independently of Ca_i by treating eggs with 2DIM NH4CI or with PMA, a direct activator of the Na-H-antiporter (Fig. 8).

Mitosis promoting factor (MPF) is a cytoplasmic activity found during mitosis or meiosis in frog and starfish oocytes (Kishimoto et al. 1982; Picard et al. 1981a,b;Gerhart et al. 1984; Wasserman and Smith, 1978), yeast and mammalian cells (Tachibana et al. 1987; Weintraub et al. 1982; Kishimoto et al. 1982; Nelkin et al. 1980; Sunkara et al. 1979) that itself stimulates mitosis and meiosis ENTRY (Kishimoto et al. 1984; Sunkara et al. 1979; Miake-Lye et al. 1983). MPF consists of at least two proteins, pp34 (Gautier et al. 1988; Dunphy eta/. 1988; Labbe et al. 1989a; Draetta and Beach, 1988; Arion el al. 1988) and cyclin (Swenson et al. 1986; Draetta et al. 1989; Labbe eta/. 19896). MPF co-purifies with a kinase (histone Hl kinase) whose activity marks mitosis entry (Labbe et al. 1989a; Arion et al. 1988). Cyclin is required for MPF activation (Draetta et al. 1989; Murray and Kirschner, 1989). It may direct pp34 to the nucleus (Booher et al. 1989). The interactions between cyclin and pp34 and thus the formation of active MPF are regulated by kinase and phosphatase activity. Histone Hl kinase activation and the association of cyclin with pp34 correlate temporally with cyclin phosphorylation (Meijer et al. 1989). Déphosphorylation of pp34 occurs prior to mitosis (Gautier et al. 1989; Labbe et al. 1989a; Gould and Nurse, 1989; Moria et al. 1989) and this dephosphorylation is required for passage through mitosis ENTRY. MPF probably consists of phosphorylated cyclin (Pcyclin) and pp34 whose threonine and tyrosine residues (phosphorylated at START) have been dephosphorylated.

The order in which this conjugate phosphorylation/dephosphorylation occurs is not known, though it has been shown that artificially dephosphorylating pp34 stimulates histone Hl kinase activity and cyclin phosphorylation (Pondaven et al. 1990).

We can monitor cyclin phosphorylation in the first cell cycle in sea urchin embryos by monitoring the phosphorylation-related apparent molecular weight change of cyclin on polyacrylamide gels (Patel et al. 1989b; Meijer et al. 1989). Cyclin is constitutively phosphorylated as it is synthesized: cyclin and Pcyclin are found in one-to-one proportions. In addition, there is a burst of cyclin phosphorylation that just precedes mitosis onset, detected as an increase in the Pcyclin/cyclin ratio (Fig. 9). If we microinject the calcium chelator BAPTA into sea urchin embryos during the first cell cycle interphase, we block both mitosis entry and the burst of cyclin phosphorylation (Fig. 9). These data suggest that the Caj signal at mitosis ENTRY causes onset of mitosis by stimulating cyclin phosphorylation. They support the idea that cyclin phosphorylation is a pre-requisite for mitosis entry. Cyclin is an indirect target of the Ca_i transient. It is possible that cyclin is phosphorylated directly by the Ca.-regulated CAM kinase that is active in sea urchin embryos at mitosis entry (Baitinger et al. 1989; 1990). It is equally likely that the action is very indirect. pp34 has itself been shown to possess cyclin kinase activity (Gautier et al. 1988; Meijer et al. 1989; Pondaven et al. 1990). One of a large number of possible mechanisms would have it that the CAM kinase targets a phosphatase that dephosphorylates pp34, activating its cyclin kinase activity, for example.

Injecting cyclin mRNA into frog oocytes at meiosis ENTRY triggers cell cycle onset and leads to mitosis onset (Swenson et al. 1986; Pines and Hunt, 1987), as does adding cyclin mRNA to a frog oocyte cell-free system (Murray and Kirschner, 1989; Murray et al. 1989). At first sight, these observations suggest that cyclin synthesis is a sufficient stimulus for mitosis onset. Yet we have identified two regulatory changes caused by the Ca_i transient at START in sea urchin eggs: phosphorylation of pp34 and the stimulation of cyclin synthesis. We can manipulate each of these changes separately to test whether stimulating cyclin synthesis at START is a sufficient signal for mitosis entry in sea urchin eggs. We can induce an increase in cyclin synthesis without phosphorylation of pp34 by treating unfertilized eggs with 2mM NH₄C1 or PMA (Patel et al. 1989b’, and Figs 7, 8). Eggs treated with PM A do not progress to mitosis ENTRY even after many hours, despite cyclin synthesis to levels greater than observed in normal cycling embryos (Table 1 in Fig. 8: legend). Spontaneous Ca_i transients can occur in PMA-treated eggs without triggering mitosis onset and artificial Ca_i transients triggered by microinjection of InsP₃ are also without effect (Patel et al. 1989b). It appears that the Ca_i transient at START is essential for passage through mitosis ENTRY 70 min later. The simplest hypothesis is that pp34 phosphorylation at START, as well as cyclin synthesis, is necessary to prime cells for mitosis ENTRY, though our experiments do not prove this. Mitosis ENTRY is not merely blocked by an inadequate S phase, because the DNA-synthesis inhibitor aphidicolin does not prevent successful mitosis onset (Twigg et al. 1988).

On the other hand, partial inhibition of protein synthesis delays mitosis onset. Increasing concentrations of the protein synthesis inhibitor emetine lead to increasingly late mitosis onset (Fig. 10). A critical concentration of some protein (probably cyclin) is essential.

There are clear differences, then, between the resumption of the cell cycle at START in sea urchin eggs and cell cycle resumption at mitosis ENTRY in Xenopus oocytes. For the latter, cyclin synthesis induced by exogenous cyclin message is a sufficient stimulus for mitosis onset, while for the former, cyclin synthesis alone is inadequate. Admittedly, cyclin message does not produce a full maturation response in Xenopus oocytes: they arrest at an abnormal meiosis, suggesting overexpression of the protein (Swenson et al. 1986). Yet cyclin message will also stimulate . mitosis entry in Xenopus egg homogenates and here both mitosis onset and mitosis exit occur cyclically (Murray and Kirschner, 1989; Minshull et al. 1989). We shall discuss the possible reasons for this discrepancy later.

The Ca_i transient associated with mitosis ENTRY occurs with normal timing in eggs in which cyclin synthesis is prevented by emetine treatment (Patel et al. 1989b). Our conclusion is that the Ca_i signal and cyclin are independently required for mitosis onset.

Xenopus egg extracts driven into mitosis by a modified cyclin mRNA that escapes destruction cannot exit mitosis (Murray et al. 1989). Sea urchin embryos are blocked in mitosis by calcium chelators. This suggests the idea that the Ca_i spike at mitosis EXIT may stimulate cyclin degradation. Fig. 11 shows that introducing a cytoplasmic calcium chelator to block the cell cycle at mitosis EXIT prevents both cyclin déphosphorylation and cyclin degradation, implying a connection between the two events. The result supports the view that the mitosis EXIT Ca_i spike is responsible for cyclin degradation in sea urchin embryos as it is at meiosis EXIT in frog eggs (Murray et al. 1989). It also strongly suggests that only phosphorylated cyclin is degraded (Luca and Ruderman, 1989; Meijer et al. 1989).

The natural cell cycle arrest in eggs and oocytes (Fig. 1) is released by an external stimulus. Immature clam, frog and starfish oocytes are arrested at meiosis ENTRY. Mature mammalian and frog oocytes (now called eggs: Hagiwara and Jaffe, 1979) arrest again at meiosis EXIT. Sea urchin eggs are stopped at START. The block at all three control points in eggs and oocytes can be relieved by a Ca, transient caused by the sperm at fertilization (Fig. 12). There are close analogies with the sea urchin embryo cell cycle (Fig. 13). At meiosis ENTRY in clam oocytes, the Ca_i transient liberates sequestered cyclin (Westendorf et al. 1989). The analogy suggests that it will also cause déphosphorylation of pp34. At meiosis exit in frog eggs, the Ca_i transient leads to cyclin destruction (Murray et al. 1989).

The odd man out is the meiosis entry stimulated by hormones in immature frog and starfish oocytes. It does not appear in Fig. 13 because the balance of evidence argues against control by a Ca_i transient (Whitaker, 1989; Whitaker et al. 1989; Doree et al. 1990; Smith, 1989). Other cell messengers are involved, predominantly cyclic AMP (Mailer and Krebs, 1980). Meiosis entry in frog oocytes is different in a second important respect. Whereas in clam oocytes or in sea urchin embryos, the Ca_i transient that triggers passage through meiosis or mitosis ENTRY occurs only a few minutes before meiosis or mitosis onset, the cell messenger signals that trigger cell cycle resumption in frog oocytes precede meiosis onset by 5 h. In other words, a different control point is involved. We should look for the meiosis ENTRY Ca_i transient in frog later, immediately prior to mitosis onset.

Each cell cycle Cai transient meets a different constellation of the cell cycle control proteins (Fig. 13). This is probably enough to explain why the same cell signal can lead to different outcomes. Each transient signals a transition from one cell state to another. At START, the transient stimulates phosphorylation of pp34 (Fig. 7) and initiates cyclin synthesis via its effect on internal pH. At mitosis ENTRY, we suggest that the transient leads to dephosphorylation of pp34, followed by phosphorylation of cyclin (Fig. 9). At mitosis EXIT, Pcyclin is destroyed (Fig. 11), leaving pp34 to be phosphorylated again by the next transient at START. This scheme has much in common with others (Hunt, 1989; Murray, 1989; Luca and Ruderman, 1989), with the important difference that the cell cycle state changes are brought about by calcium signals.

The weakest aspect of the argument linking calcium and cell cycle control is our vagueness about the mechanisms that govern the timing of the Ca, transients. There must be feedback between the state changes and the transients, but we are ignorant of it. A simple flip-flop model in which each pp34 transition primes the next transient and Pcyclin primes the transient that leads to its own destruction is adequate and can be tested. A model of this sort is consistent with the observation that the pp34 PSTAIR sequence will elicit a Ca_i transient when injected into starfish eggs (Picard et al. 1990). It is equally likely that the feedback mechanisms exist, not at the level of the final targets, pp34 and cyclin, but at the intermediate level of the kinases and phosphatases that are responsible for modification of the final targets. The feedback mechanisms here may be rather more difficult to unravel. In trying to determine what generates the cell cycle Ca_i transients, it may be worth paying more attention to the cell cycle organelle, the centrosome.

Centrosome duplication is independent of DNA synthesis and mitosis in sea urchin eggs (Sluder and Lewis, 1987), suggesting that it may be the cell cycle pacemaker. Micro-injecting heterologous centrosomes triggers cell cycle onset in starfish eggs (Picard el al. 1987a). We have already mentioned that one of the enzymes of the PPI messenger pathway (Petzelt et al. 1989) and the calcium target calmodulin (Welsh et al. 1979; Wolniak et al. 1980) are found concentrated at the centrosome. pp34 also localizes to the centrosome during mitosis (Riabowol et al. 1989). If the centrosome were the source of cell cycle Ca_i transients, it would explain why the mitosis ENTRY transient is large in polyspermie eggs (Fig. 5). Polyspermie eggs differ from monospermic eggs in having, not one, but several centrosomes.

We have demonstrated a link between the cell cycle Ca_i transients in early sea urchin embryos and their targets, the cell cycle control proteins pp34 and cyclin. We think it likely that Ca_i transients control cell cycle progression in other cells, too. There are two main objections to the idea that calcium may be a universal cell cycle control signal. The first is that Ca_i signals cannot always be detected as cells pass through START, mitosis ENTRY and mitosis EXIT. We have suggested that this may be because cell cycle Ca, transients can be, on occasion, so small or local as to escape detection using the fluorescent dye techniques that are presently available. The other objection to the idea of calcium as a cell cycle signal arises from experiments on cell-free cell cycle systems.

Cytoplasmic extracts from activated Xenopus eggs and Spisula embryos can support nuclear cycles of chromatin condensation/decondensation and nuclear envelope breakdown/re-formation that are driven by addition of mRNA coding for the cell cycle control protein, cyclin (Murray and Kirschner, 1989; Luca and Ruderman, 1989). The implication is that it is cyclin synthesis that drives the cell cycle. Moreover, these cell-free systems undergo repeated chromatin condensation cycles (Andrew Murray, unpublished) and cyclin destruction (Luca and Ruderman, 1990) in the presence of 5 HIM EGTA, a concentration of this calcium chelator that might be expected to prevent any transient increase in Ca_i (Steinhardt and Aiderton, 1988; Twigg et al. 1988). The experiments with cell-free systems seem to argue that calcium is not an essential signal. On the other hand, as we have seen, there is little doubt that calcium chelators introduced into living cells prevent progression through the cell cycle control points.

We can resolve the contradiction by supposing that the cell cycle Ca, signals are pacemakers that govern the precise timing of cell cycle transitions. The corollary is that, given time, the cell cycle transition will occur in the absence of a calcium signal. The pacemaker idea would also explain why the cycle times of cell-free cell cycles (with normal levels of cyclin expression) are longer than their cellular counterparts. It is also possible, however, that the cycles in cell-free systems are simulacra of their cellular originals, rather than precise copies. The overexpression of cyclin in these cell-free systems may produce an in vitro nuclear cycle that lacks some of the control mechanisms that operate in an intact cell. In intact dividing cells it is very probable that it is the cell cycle Ca_i transients that provide cell cycle control.

We thank Chris Ford, Marcel Doree, Robert Brookes, Roy Golsteyn and Tim Hunt for their advice and Melanie Lee, Chris Norbury and Paul Nurse for supplying the anti-PSTAIR antibody and PSTAIR peptide. We particularly wish to thank Michael Aitchison for preparing the figures. The work was supported by grants from the Wellcome Trust, the Science and Engineering Research Council and the Royal Society.