Mos limits the number of meiotic divisions in urochordate eggs

Entry into and exit from mitosis is regulated by cyclin accumulation and destruction (Evans et al., 1983; for a review, see Hunt, 2004). A central tenet of somatic cell cycles is that directionality is provided by cyclin B destruction rendering exit from M phase irreversible (Potapova et al., 2006). The corollary of this is that only one M phase occurs per cycle. However, during meiosis, two M phases occur one after the other. Moreover, akin to mitotic exit, exit from meoisis I depends on cyclin B destruction in oocytes from diverse species: C. elegans (Furuta et al., 2000), ascidian (Levasseur and McDougall, 2000) and mouse (Herbert et al., 2003). This lack of irreversibility creates the potential problem of how the oocyte limits the number of M phases to just two during maternal meiosis. Interestingly, it has been known for many years that the premature activation of mouse oocytes causes them to enter meiosis III rather than interphase (Kubiak, 1989). The Mos/MAPK pathway is known to regulate both the cell cycle and asymmetric cell division in eggs (Kishimoto, 2003; Verlhac et al., 2000; Amiel et al., 2009). However, although it is well established in mammals and urochordates that a fall in MAPK activity precedes entry into interphase, it is not known whether it is the duration of Mos/MAPK activity that limits the number of meioses to just two.

Mos and its associated MAPK activity are elevated during maternal meiosis I and II throughout the Eumetazoa: Cnidaria (Amiel et al., 2009), Mollusc (Shibuya et al., 1992), Echiura (Gould and Stephano, 1999), Echinoderms (Tachibana et al., 1997), Urochordates (Russo et al., 1996; McDougall and Levasseur 1998), Vertebrates (Hashimoto et al., 1994; Verlhac et al., 1996). Although the role of Mos/MAPK pathway in CSF arrest is widespread (Sagata, 1989; Sagata, 1997; Masui, 2000; Yamamoto et al., 2008; Wu and Kornbluth, 2008), its role in other oocyte processes, such as asymmetric cell division (ACD), S phase inhibition and meiosis I to II transition, has either not been as well characterised [ACD (Verlhac et al., 2000; Amiel et al., 2009; Azoury et al., 2009)] or is less conserved [MI-to-MII transition (Furuno et al., 1994; Picard et al., 1996; Walter et al., 1997; Bitangcol et al., 1998; Tachibana et al., 2000; Verlhac et al., 2000; Dupre et al., 2002). Even less well studied is whether Mos/MAPK activity has to be switched off to prevent entry into meiosis III. Some evidence supporting this hypothesis has been found in mammalian oocytes. For example, maintaining MAPK active in mouse oocytes leads to the re-establishment MPF activity and prevents pronuclei formation after exit from meiosis II (Moos et al., 1995; Moos et al., 1996). However, it has not been tested whether keeping MAPK active in these eggs generates a third meiotic division, leading to extrusion of a third polar body (PB). As mouse oocytes can enter meiosis III (Kubiak, 1989) with elevated MAPK activity during the meiosis II to III transition (Verlhac et al., 1996), this suggests that the level of Mos/MAPK activity might be the determining factor limiting the number of meiosis to just two in eggs but this hypothesis has not been formally tested.

Ascidian and mammalian eggs share several common features. Ascidians are marine chordates (termed urochordates) positioned as a sister group of vertebrates (Delsuc et al., 2006). Ascidian eggs are arrested at metaphase I by CSF (Russo et al., 2009) in a similar manner to mammalian eggs that arrest at metaphase II (Shoji et al., 2006). As in mammals, fertilization in urochordates triggers a series of periodic Ca²⁺ waves that stimulate metaphase exit driven by cyclin B destruction and the loss of CDK1 activity (Levasseur and McDougall, 2000; Nixon et al., 2002). In both ascidian and mammalian eggs, sperm-triggered Ca²⁺ oscillations terminate minutes prior to meiotic exit (Stricker, 1999). Moreover, these sperm-triggered Ca²⁺ oscillations can be maintained by preventing meiotic exit. This positive-negative feedback cycle ensures that Ca²⁺ oscillations cease upon exit from the meiotic cell cycle (Dumollard et al., 2002). In order to dissect the role of Mos in CSF arrest from its role in the transition from meiosis II to interphase, we exploited the ascidian model as their eggs are arrested in metaphase I rather than metaphase II (as in mammals). Indeed, CSF is inactivated at fertilization during meiosis I in ascidians making it possible to study the meiosis II to interphase transition independent of CSF.

Remarkably, for a protein that has been studied for more than 20 years we would like to attribute one additional and central role to Mos in urochordates. We find that maintaining Mos and the downstream MAPK activity elevated after fertilization induces meiosis III, meiosis IV and even meiosis V, and also that each additional phase of meiosis is separated by a morphologically normal ACD. Mos kinase from three different taxa and mediating three different CSF arrests were tested [from the jellyfish Clytia hemisphaerica (G1 CSF), the ascidian Ciona intestinalis (Metaphase I CSF) and mouse Mos (Metaphase II CSF)] and all types of Mos had the same ability to produce supernumerary PBs in ascidian eggs. These observations prompt us to conclude that it is the duration of Mos/MAPK activity during fertilization of urochordate eggs that limits the number of meiotic M phases to precisely two.

Eggs from the ascidians Phallusia mammillata and Ciona intestinalis were harvested from animals obtained in Sète and kept in the laboratory in a tank of natural sea water at 16°C. Eggs from Ascidiella aspersa were harvested from Blyth, UK and kept in a tank of natural sea water at 16°C. Egg preparation and microinjection have been described previously (Dumollard and Sardet, 2001) [see detailed protocols by Sardet et al. (Sardet et al., 2010)]. All imaging experiments were performed at 19°C and the kinase assays at 21°C.

Construction of Δ90-cyclin B::GFP has been described previously (Levasseur and McDougall, 2000; Levasseur and McDougall, 2003). Cyclin B Y170 was created by site-directed mutagenesis from full-length cyclin B [characterized by Levasseur and McDougall (Levasseur and McDougall, 2000)]. Clytia hemispherica mos1 has been cloned and described recently (Amiel et al., 2009), whereas Mus musculus Mos was a kind gift from Marie-Helene Verlhac (Université Pierre et Marie Curie, Paris, France). Ci-Mos (gene Id: KH.C2.835.v1.A.ND1-1), Ci-cyclin A (gene Id: KH.C2.821.v1.A.ND1-1), Ci-cyclin E (KH.S498.1.v2.A.SL1-1) and Ci-shugoshin (KH.C12.362.v1.A.SL1-1) were amplified from a Ciona intestinalis Gateway-compatible cDNA library using PCR. All constructs were made using pSPE3 and the Gateway cloning system (Invitrogen) unless otherwise stated [for a detailed protocol, see Sardet et al. (Sardet et al., 2010)]. For all fluorescent fusion protein constructs, we indicate the N or C terminal position of the fluorescent protein (FP) by the following convention: FP-protein X or protein X::FP.

Imaging of microtubules (to visualize spindles) was achieved by expression of microinjected mRNA coding for the microtubule binding domain of Map7 (also known as ensconsin) fused to Venus fluorescent protein while H2B::Rfp1 was used to follow DNA (see Prodon et al., 2010).

Synthetic mRNAs were prepared as concentrated solutions (>2 μg/μl) in distilled water and small aliquots were frozen at −80°C. Solutions were centrifuged for 15 minutes at 15,000 g prior to loading the needle in order to sediment particles that might block the injection needle. Fluorescence from fluorescent protein constructs was observed in an unfertilized egg a few hours after injection of concentrated mRNAs and less than one hour after injection in an embryo (two-cell stage).

Treatments with U0126 and nocodazole were done by bathing in sea water. UO126 was dissolved in DMSO and added to a final concentration of 1 μM for UO126 and 2 μM for nocodazole. Ionomycin, made up in ethanol, was added to eggs at 2 μM concentration for 4-5 minutes and then thoroughly washed out to allow for undisturbed PB extrusion.

S-phase was determined by 5-Bromo-2-deoxyUridine (BrdU) incorporation as modified from Denker et al. (Denker et al., 2008). In brief, eggs or embryos were incubated in 1.5 mM BrdU in ASW for 2-4 hours followed by rapid rinsing in ASW without BrdU before fixation in 3.7% formaldehyde for 1 hour, rinsing, then treatment in 4 M HCl for at least 1 hour. After several rinsing steps, BrdU was detected by anti-BrdU antibodies and the appropriate fluorescent secondary antibodies, and observed by confocal microscopy on a Leica TCS SP2.

Time-lapse imaging of Venus and Rfp1 constructs was performed on an Olympus IX70 inverted microscope set up for epifluorescence imaging. Sequential bright-field and fluorescence images were captured using a cooled CCD camera (Micromax, Sony Interline chip, Princeton Instruments, Trenton, NJ, USA) and data collected was analysed using MetaMorph software (Molecular Devices, Sunnyvale, CA, USA).

Calcium imaging was performed by measuring the fluorescence of injected calcium green 1 dextran (10,000 MW, Invitrogen) as previously described (Dumollard and Sardet, 2001) [for detailed protocols, see Sardet et al. (Sardet et al., 2010)].

Imaging of eggs and embryos were analysed by MetaMorph software (Molecular Devices) and Image J (NIH, USA).

The MBP (Myelin Basic Protein) kinase activity and histone H1 kinase activity were assayed as described previously (McDougall and Levasseur, 1998). Samples of five oocytes were collected by first washing the oocytes through 1 M glycine (pH 8) three times to remove the sea water (this does not alter meiotic progression). The oocytes were then removed in a volume of 1 μl and transferred to 4 μl reaction buffer [25 mM HEPES, 80 mM β-glycerophosphate, 5 mM EGTA, 10 mM MgCl₂, 1 mM DTT, 10 mg/ml leupeptin/pepstatin/aprotinin, 0.2 mM AEBSF, 1 mM benzamidine, 100 mM NaVO₄, 5 mM NaF (pH 7.2)]. At this point, the oocytes were snap-frozen in liquid nitrogen. After defrosting, samples were treated exactly as described previously (McDougall and Levasseur, 1998).

Mature ascidian eggs are arrested at metaphase I and display elevated MPF and MAPK activity (Russo et al., 2009; McDougall and Levasseur, 1998). We estimated MAPK and MPF activity in Phallusia mammillata eggs by measuring MBP and histone H1 kinase activity in fertilized and Ca²⁺-activated eggs (Fig. 1). H1 kinase measurements show two peaks of MPF activity corresponding to the two meiotic divisions following fertilization (Fig. 1A). Activation of the egg with a Ca²⁺ pulse (4 minutes application of 2 μM ionomycin) also resulted in two peaks of MPF activity (see Fig. S1A in the supplementary material) as well as extrusion of two polar bodies (PBs) followed by meiotic exit and pronucleus (PN) formation (see Fig. S4 in the supplementary material) with similar kinetic to natural fertilization. MBP kinase assay indicates that MAPK activity is elevated before fertilization, that it increases slightly during extrusion of the first PB (meiosis I to meiosis II transition) and then decreases 20 minutes after Ca²⁺ activation before exit from meiosis II (Fig. 1B, natural fertilization) (Russo et al., 1996; McDougall and Levasseur, 1998). Expression of Ch-Mos::Rfp1 (from jellyfish) in ascidian eggs is able to maintain MAPK activity indefinitely after egg activation (Fig. 1C). Expression of Mos from the ascidian Ciona intestinalis (Ci-Mos::Cherry) in unfertilized Phallusia mammillata eggs increased MAPK activity by 46% (Fig. 4D) and was also able to maintain MAPK activity after egg activation (data not shown). Incubation of ascidian eggs in the MEK inhibitor UO126 decreased MBP kinase activity in 10 minutes (Fig. 1D, Fig. 2B) confirming that MEK-induced MAPK activity is measured in our MBP kinase assay. These data demonstrate that UO126 rapidly suppresses MAPK activity and that, conversely, maintaining Mos levels in the egg is able to suppress MAPK downregulation usually observed after Ca²⁺ induced egg activation.

In order to assess the involvement of MAPK in metaphase I CSF we inactivated MAPK (using UO126) in unfertilized eggs and monitored egg activation in UO126-treated eggs (Fig. 2). Egg activation was monitored by measuring H1 kinase activity, cyclin B destruction as well as PB formation, PN formation and entry into S phase (Fig. 2A-C). Fig. 2A and Fig. S1C in the supplementary material show that, at around 32′ after incubation in 1 μM UO126, unfertilized Phallusia mammillata eggs extruded a PB that was sometimes reabsorbed (n=25 eggs). Unfertilized Ciona intestinalis eggs also extruded a PB after UO126 treatment (see Fig. S2A in the supplementary material). However, these PBs did not contain DNA (see Fig. S2A in the supplementary material) suggesting that proper chromosomes segregation did not occur. Nevertheless, such pseudo-PB formation was followed ~7′ 30″ later by formation of a PN (n=13, Fig. 2A and see Fig. S1C in the supplementary material). H1 kinase measurements showed that MPF declined at around 40 minutes after UO126 treatment, confirming exit from metaphase arrest (Fig. 2B). Furthermore, we consistently found that cyclin B was destroyed following inhibition of MAPK by UO126 treatment (Fig. 2C, n=7 eggs). Although PB formation was substantially delayed after U0126 treatment when compared with Ca²⁺ activated control eggs, PN formation occurred 7′ 30″ after PB extrusion in U0126-activated eggs, similar to the delay between PB2 and PN in controls (Fig. 2A; see Fig. S1C in the supplementary material). It should be noted that meiosis II was completely absent and that U0126-activated eggs entered interphase directly following extrusion of one PB.

Inhibition of the Mos/MAPK/p90rsk cascade in starfish oocytes can induce embryonic cleavages and parthenogenesis (Tachibana et al., 2000; Mori et al., 2006), but such parthenogenesis was never observed in ascidian eggs treated with UO126. However, we did notice cyclical deformations during the first hours of UO126 treatment with some deformations resembling pseudo-cleavages (see Movie 1 in the supplementary material). To assess for mitotic cell cycles in UO126-treated eggs, we monitored the cycles of accumulation of cyclin A during interphase and destruction during M phase. Ascidians possess one isoform of cyclin A. Cyclin A cannot accumulate during meiosis, which possesses only M phases, and can accumulate only during interphase. We thus expressed and imaged Ci-cyclin A::Venus in UO126-treated eggs (Fig. 2D). We first confirmed that Ci-cyclin A::Venus was destroyed during mitosis and synthesized during interphase in living ascidian embryos (see Fig. S3 in the supplementary material). Ci-cyclin A::Venus was found to be nuclear during interphase and accumulated until NEB, before being destroyed during the whole of mitosis (see Fig. S3 in the supplementary material), thus reproducing the known behaviour of cyclin A in somatic cells (Geley et al., 2001; Den Elzen and Pines, 2001).

Imaging of Ci-cyclin A::Venus consistently showed cycles of synthesis and destruction of cyclin A::Venus in UO126-treated eggs (Fig. 2D and see Movie 1 in the supplementary material; n=10 eggs). To confirm that DNA replication occurred repetitively in these eggs, we assessed BrdU incorporation during the first 2 hours of UO126 treatment (Fig. 2A) and between 2 and 4 hours after UO126 treatment (Fig. 2D) and found that these eggs indeed replicated DNA during both time windows. These observations demonstrate that MAPK inactivation alone is able to release CSF arrest, leading to exit from meiosis I immediately followed by entry into first interphase and subsequent mitotic cell cycles.

Both cyclin E/CDK2 and the SAC have also been implicated in mediating CSF arrest in eggs (Tunquist et al., 2002; Tunquist and Maller, 2003; Grimison et al., 2006). For example, studies in Xenopus egg extracts revealed that keeping cyclin E/CDK2 active can maintain M-phase after Ca²⁺ activation of CSF extracts (Tunquist et al., 2002). We thus tested for a potential role for cyclin E/CDK2 in CSF arrest in ascidians (see Fig. S4 in the supplementary material). Overexpression of cyclin E did not prevent PB formation but did inhibit PN formation (see Fig. S4B,D in the supplementary material). We reasoned that if cyclin E alone constitutes CSF then it should block egg activation induced by inactivation of MAPK. However, we found that inactivating MAPK in cyclin E-expressing eggs resulted in PB extrusion with the same kinetics as control eggs activated with U0126 (n=5, data not shown). This suggests that cyclin E/CDK2 may not be involved in metaphase I arrest in ascidians but confirmation of this will await knockdown of cyclin E in unfertilized eggs.

In order to assess whether the SAC is part of CSF-mediated arrest in ascidian eggs, we fertilized eggs in the presence of microtubule-depolymerising drugs (see Fig. S4 in the supplementary material). We found that although spindle disruption by nocodazole treatment did prevent PB extrusion it did not alter the timing of PN formation, suggesting that the SAC might not be functioning in ascidian eggs (see Fig. S4C,D in the supplementary material; n=18 eggs). However, these data do not rule out a role for SAC proteins in Meta-I CSF, which could be determined by a knockdown approach of candidate proteins. However, these data highlight a lack of feedback control monitoring progression of PB extrusion.

Together these data suggest that the Mos/MAPK cascade is the major regulator of CSF activity in urochordate eggs. Furthermore, the direct entry into interphase after one meiotic division in UO126-treated eggs indicates that MAPK is necessary for entry into meiosis II.

Treatment of unfertilized egg with UO126 stimulates completion of one meiotic division before entry into first interphase, suggesting that the Mos/MAPK cascade promotes entry into meiosis II. However, when eggs are activated with UO126, meiotic completion requires around 30 minutes, whereas meiosis I completion takes only 7-9 minutes in Ca²⁺ activated (and fertilized) eggs (see Fig. S1C in the supplementary material). In order to confirm that Mos/MAPK promotes entry into meiosis II after fertilization, we treated eggs with UO126 and then activated the eggs with Ca²⁺ (see Fig. S1B,C in the supplementary material). Such eggs extruded a first PB at the normal time (~8 minutes after Ca²⁺ activation, see Fig. S1C in the supplementary material) but the second PB was not consistently observed (data not shown) and the metaphase II spindle could not form (see Fig. S1B in the supplementary material). PN formation occurred at the normal time (~31 minutes after Ca²⁺ activation, see Fig. S1C in the supplementary material). The late occurrence of PN formation indicates that meiosis II did occur. Measurements of H1 kinase activity showed Ca²⁺ activation in an egg devoid of MAPK activity (treated with UO126 10 minutes before ionomycin treatment, Fig. 1D) also resulted in two peaks of MPF activity (see Fig. S1B in the supplementary material), suggesting that MPF could reactivate after PB1 extrusion without MAPK activity (20′ after Ca²⁺ activation, see Fig. S1B in the supplementary material). These observations indicate that MAPK activity is not necessary to drive entry into meiosis II during natural fertilization, even though MAPK activity is necessary for the formation of the metaphase II spindle (and hence extrusion of PB2).

In order to confirm the involvement of the Mos/MAPK pathway in CSF activity, we expressed Mos in embryonic blastomeres by injecting capped RNAs coding for Mos kinase in one blastomere of two-cell stage embryos (see Fig. S2B in the supplementary material; Fig. 3). Injection of either jellyfish Ch-mos, mouse Mm-mos or ascidian Ci-mos cRNAs into one blastomere of a two-cell embryo all resulted in early cleavage arrest giving rise to large round blastomeres (see Fig. S2B in the supplementary material). Strikingly, we noticed that before reaching cleavage arrest (and sufficient accumulation of Mos protein expressed from injected capped RNA), Mos-expressing blastomeres underwent rounds of ACD reflected by the formation of PB-like cells in the blastomeres expressing Mos (arrowheads in Fig. 3A). Mos-expressing blastomeres eventually ceased cleaving and closer examination of blastomeres showed that they arrested in mitotic metaphase as seen by the sustained presence of a metaphase spindle (Mm-mos, n=5 embryos, Fig. 3B; Ch-mos, n=3 embryos, Fig. 3C). Remarkably, we noticed that the mitotic spindle in Mos-expressing blastomeres always had one pole subjacent to the plasma membrane suggestive of spindle pole anchoring to the cell cortex (Fig. 3B,C).

Together, these observations further suggest that the Mos/MAPK pathway mediates metaphase I CSF arrest in ascidians. They also indicate that, although a sufficient level of Mos is necessary to impose metaphase arrest in the embryo, a sub-threshold level of Mos promotes meiotic-like ACD over mitotic cell division.

After establishing the role of the Mos/MAPK pathway in metaphase I CSF in urochordates, we sought to characterize other roles of Mos/MAPK during ascidian meiosis. We observed that inactivating MAPK in an unfertilized egg triggers meiotic exit after one PB (Fig. 2). Furthermore inactivating MAPK in a fertilized egg impairs formation of meiosis II spindle (see Fig. S1 in the supplementary material). These observations indicate that an active Mos/MAPK pathway may promote the meiotic phenotype of the mature egg (i.e. high MPF, small peripheral meiotic spindle). We thus tested whether maintaining the Mos/MAPK pathway active in fertilized eggs was able to inhibit Ca²⁺ triggered meiotic exit.

We used several criteria to define the meiotic state of the egg: (1) inhibition of pronuclear envelope formation after extrusion of PB2; (2) persistence of sperm-triggered Ca²⁺ oscillations after PB2 formation; (3) formation of a small meiotic spindle subjacent to the cortex after PB2 formation; (4) inhibition of cyclin A accumulation during interphase of the mitotic cell cycles observed after Ca²⁺ activation of the egg. We found that Mos-expressing fertilized eggs do not form pronuclei after completion of meiosis II. Although Mos expression did not affect the timing of PB1 and PB2 (Fig. 4A, t=32 min; see Fig. S5 in the supplementary material), no pronuclei could be observed in Mos-expressing eggs (boxed in red) at the time control eggs completed pronuclear migration (52 min post-fertilization, boxed in blue in Fig. 4A; see also Movie 2 in the supplementary material). Even though Ch-Mos blocks jellyfish eggs at G1 phase and Mm-Mos and Ci-Mos block eggs of their respective species at metaphase, the three different Mos proteins all inhibited PN formation in ascidian eggs (Fig. 4A; see Fig. S5 in the supplementary material).

The second criterion is based on the observation that sperm-triggered Ca²⁺ oscillations are associated with the meiotic cell cycle and that they terminate at the time the second PB is extruded ~7 minutes before the pronuclei form (McDougall and Levasseur, 1998). Moreover, blocking meiotic exit with a non-degradable form of cyclin B maintains the sperm-triggered Ca²⁺ oscillations indefinitely (Levasseur and McDougall, 2000). Strikingly, Mos overexpression also maintained sperm-triggered Ca²⁺ oscillations after extrusion of two PBs (Fig. 4B, n=7 eggs). Remarkably, we could observe in these eggs a small peripheral meiotic spindle after PB2 extrusion (inset in Fig. 4B), suggesting that maintaining Mos levels after fertilization induced entry into meiosis III.

Finally, we monitored cyclin A levels to further determine whether Mos maintained a meiotic-like phenotype. Activation of cyclin A::Venus-expressing ascidian eggs with ionomycin provoked extrusion of two PBs followed by PN formation (see Fig. S4 in the supplementary material) and cycles of accumulation and destruction of cyclin A::Venus indicative of mitotic cell cycles (Fig. 4C, n=5 eggs). Even though these Ca²⁺ activated eggs never cleaved, they proceeded through S phase, as monitored by BrdU incorporation into the egg DNA (see Fig. S4A in the supplementary material). Thus, Ca²⁺ activation of the egg is sufficient to stimulate exit from meiosis and to trigger a mitotic cell cycle oscillator driving successive S and M phases. We assessed the impact of Mos on this cell cycle oscillator by activating eggs expressing both cyclin A::Venus and Ch-Mos::Rfp1 with ionomycin (Fig. 4D). High Mos expression (overnight expression in the unfertilized egg) increased MAPK activity by only 46% (graph in Fig. 4D). In such eggs we could not measure any significant increase in cyclin A::Venus levels (Fig. 4D, n=7 eggs), even though they deformed, extruded two PBs and stimulated protein synthesis (seen as an increase in Ch-Mos::Rfp1 levels, red graph in Fig. 4D). The lack of cyclin A accumulation in the presence of Mos is indicative of a lack of mitotic cycles comprising interphases. Overall, these data show that Mos prevents the egg-to-embryo transition and maintains the fertilized egg in a meiosis-like state by inhibiting the triggering of a mitotic cell cycle oscillator. Strikingly Mos persistence after fertilization seems to promote entry into metaphase III.

We have shown so far that maintaining MAPK active in fertilized eggs can induce entry into metaphase III, as found previously in mouse eggs (Moos et al., 1995; Moos et al., 1996). However, we were surprised to find that longer time-lapse imaging of such eggs revealed that they extruded more than two PBs (Fig. 5 and see Movie 3 in the supplementary material). Fig. 5A-C shows fertilized ascidian eggs expressing either Mm-mos or Ch-mos, which were not blocked at metaphase III but extruded three PBs (Fig. 5A) while other eggs extruded four PBs (Fig. 5B,C) and sometimes even five PBs (Fig. 5F). This indicates that Mos expressing eggs are undergoing several additional rounds of meiosis (which we call meiosis III, IV and V) following completion of meiosis II. Exogenous Mos from either mouse or jellyfish or ascidian (Ci-Mos) all had this effect in Phallusia mammillata eggs (Fig. 5A-D). This striking effect was not limited to Phallusia eggs, as Ci-Mos could also induce more than two PBs in Ascidiella aspersa eggs (n=5 eggs, four PBs in the example shown in Fig. S6 in the supplementary material). We did not try this experiment in Ciona intestinalis eggs, as they have a low translation efficiency before fertilization (Prodon et al., 2010). None of the Mos constructs affected the timing of meiotic resumption as Mos-expressing eggs extruded the first two PBs with normal kinetics (Mm-mos, Fig. 5A, n=4 eggs; Ch-mos, Fig. 5B,C, n=14 eggs; Ci-mos, Fig. 5D, n=11 eggs, see Fig. S5 in the supplementary material). As the generation of supernumerary PBs requires a peripheral spindle to be present, we imaged microtubules in eggs overexpressing Mos. During meiosis III, IV and V, a peripheral meiotic spindle reformed after each PB extrusion and these spindles became thinner across their midsection with each round of meiosis (Fig. 5C, n=4 eggs). On one occasion, we observed that the meiotic spindle was eventually captured by the sperm aster, thereby preventing further formation of PBs (see Movie 4 in the supplementary material). Together, these observations demonstrate that maintaining Mos/MAPK cascade active in the fertilized egg is sufficient to stimulate entry into meiosis III, meiosis IV and even meiosis V. We performed the rescue experiment by treating Ci-Mos-expressing eggs that had been fertilized and had entered meiosis III with U0126 (i.e. 70 minutes after fertilization, Fig. 5E). Strikingly, we found that nuclei could be observed in such eggs around 30 minutes after perfusion of 5 μM UO126 (Fig. 5E, n=3 eggs) indicating that the ability of Mos to maintain a meiotic cell cycle is mediated entirely by MAPK

Overall these data show that an active Mos/MAPK pathway promotes a cortically located metaphase spindle in the fertilized egg, thereby ensuring spatial aspects of ACD during PB extrusion. Furthermore, the Mos/MAPK pathway must be downregulated after fertilization to limit the number of meioses to just two.

By examining the role played by the Mos/MAPK pathway in metaphase I CSF arrest in urochordate eggs, we exposed the mechanism that limits the number of meioses in these eggs to two. Remarkably, by keeping the Mos/MAPK pathway active following fertilization at near physiological levels, the zygote continued to cycle in and out of meiosis. Specifically, these fertilized eggs entered meiosis III, formed a PB then entered meiosis IV and formed another PB and even occasionally entered meiosis V and formed a fifth PB (Fig. 6). We define these extra M phases as extra rounds of meiosis rather than mitosis by the following criteria: (1) a small anastral meiotic spindle forms subjacent to the cortex that drives rounds of PB formation during asymmetric cell division rather than a central mitotic spindle with large asters that drives symmetric cell division; (2) interphase-dependant cyclin A accumulation is abolished; (3) sperm-triggered calcium oscillations are maintained; and (4) nuclear envelope formation does not occur after PB formation. Therefore the Mos/MAPK pathway supports all aspects of meiotic asymmetric cell division in the egg and downregulation of the Mos/MAPK pathway is necessary to limit the number of meiotic M phases to two in fertilized urochordate eggs.

A central tenet of mitosis is the irreversibility of mitotic exit. For example, in somatic cells a dramatic mechanism based on cyclin B destruction ensures that mitotic exit is irreversible and is followed by interphase (Potapova et al., 2006). This is not the case for meiosis. During maternal meiosis even though cyclin B destruction ensures exit from meiosis I in both urochordates and vertebrates oocytes, MPF activity comes back to drive entry into meiosis II (Levasseur and McDougall, 2000; Dupre et al., 2002; Herbert et al., 2003; Madgwick et al., 2006; Tang et al., 2008). This is probably due to the precocious inhibition of APC/C after anaphase I thereby allowing for early accumulation of cyclin B during interkinesis (Iwabuchi et al., 2000; Madgwick et al., 2006; Tang et al., 2008). The fact that meiosis I can be followed immediately by meiosis II makes it unclear how eggs avoid entering meiosis III. The Mos/MAPK pathway is a good candidate for regulating the number of meioses as it is a major regulator of the meiotic cell cycle (Kishimoto, 2003). However, this is not a universal mechanism because in both jellyfish and starfish the Mos/MAPK cascade is active during meiosis I and II and crucially remains active as the eggs exit meiosis II and enter G1 [jellyfish (Amiel et al., 2009), G1 or G2 in starfish (Fisher et al., 1998; Mori et al., 2006; Hara et al., 2009)]. Moreover, in Xenopus overexpression of Mos does not prevent meiotic exit and instead increases the duration of G2 in fertilized eggs (Murakami et al., 1999) or blocks activated CSF extracts in G2 phase (Bitangcol et al., 1998; Walter et al., 1997; Murakami and Vande Woode, 1998; Abrieu et al., 1997). However, in sharp contrast to Xenopus, maintaining MAPK activity in fertilized mouse eggs does inhibit pronucleus formation (Moos et al., 1995) but it is not know whether supernumerary meioses are induced.

Here, we find that maintaining Mos/MAPK activity in urochordate eggs not only inhibits entry into interphase but is also sufficient to generate supernumerary meioses. This is consistent with Mos-induced stabilization of MPF (Tachibana et al., 2000; Dupre et al., 2002; Kishimoto, 2003; Tang et al., 2008) and causes re-entry into additional rounds of meiotic M phase in urochordates (Fig. 6). The loss of the Mos/MAPK activity is necessary for preventing entry into further rounds of meiosis. However, this does not explain exit from these additional rounds of meiosis. As Mos also maintains sperm-triggered Ca²⁺ oscillations, we postulate that these Ca²⁺ increases drive exit from meiosis III, IV and V as they do at fertilization (McDougall and Sardet, 1995). Importantly, both low (2-3 hours expression) and high (15-20 hours expression) levels of Mos had the same effect in that all Mos-expressing eggs that were subsequently fertilized entered then exited meiosis III (data not shown).

Intriguingly, although MAPK activity drives entry into supernumerary meiosis (Fig. 5), Mos/MAPK activity was not found to be necessary for MPF reactivation after PB1 extrusion (see Fig. S1 in the supplementary material). This discrepancy might be due to a time delay between MAPK inactivation and MPF inactivation (Fig. 2). In fact, we observed that even though MAPK is inactivated within 10 minutes of UO126 treatment (Fig. 1), MPF declines and cyclin B is destabilized only 30 minutes after UO126 treatment (Fig. 2). This suggests that MAPK stabilizes cyclin B (and MPF) indirectly and that the MAPK target inhibiting the APC/C remains active for up to 20 minutes after MAPK inactivation. For these reasons, we cannot exclude a role for a MAPK-dependent mechanism for entry into meiosis II in urochordates. By contrast, we demonstrate here that MAPK is required for formation of the metaphase II spindle.

Finally, the ability to form multiple PBs suggests that bipolar meiotic spindles in Phallusia (and Ascidiella) eggs can form independently of centrioles, as is the case in mouse oocytes. In mouse oocytes, meiotic spindles will form around chromosomes and several meiotic spindles can form in the same metaphase II-arrested oocyte (Maro et al., 1986). Even DNA-covered beads cause acentriolar bipolar meiotic spindles to form in mouse oocytes (Deng et al., 2006). However, in starfish oocytes, the metaphase I meiotic spindle contains two centrioles at each pole whereas the meiosis II spindle has one centriole at each pole (Tamura and Nemoto, 2001). Starfish oocytes lose the ability to form a bipolar meiotic spindle following extrusion of the second PB as only one centriole (which is non reproductive) remains in the egg (Tamura and Nemoto, 2001). We therefore surmise that a bipolar meiotic spindle can form in the ascidian without the need of centrioles similar to mouse oocytes. However, this should be confirmed by direct visualization of centrioles in ascidian eggs.

CSF was originally described as a catalytic activity that is able to block amphibian eggs at metaphase II and was demonstrated by the cell cycle arrest caused by injecting egg cytoplasm into blastomeres at the two cell stage (Masui and Markert, 1971; for a review, see Masui, 2000; Kishimoto, 2003). Vertebrate CSF is mediated by an inhibitor of the APC/C called Emi2/Erp1 (for a review, see Wu and Kornbluth, 2008). Erp1 levels normally increase during meiosis II in Xenopus (Rauh et al., 2005; Tang et al., 2008) and in the mouse (Shoji et al., 2006), and although Erp1 levels decline following egg activation, Erp1 protein is still present up to the four-cell stage in mouse (Shoji et al., 2006) and up to the blastula stage in Xenopus (Nishiyama et al., 2007). Mos induces cleavage arrest in Xenopus embryo at the two-cell stage (Sagata, 1989) and this is probably due to residual Erp1. In amphibians Mos/MAPK/p90rsk phosphorylates Erp1 to maintain its inhibitory activity towards the APC/C (Nishiyama et al., 2007; Inoue et al., 2007).

Mos/MAPK activity mediates metaphase I CSF arrest in urochordates (Fig. 6). Our observations suggest that metaphase I CSF in urochordates is also an APC/C inhibitor activated by Mos/MAPK-dependant phosphorylation. Therefore, a possible conservation between metaphase I CSF in urochordates and metaphase II CSF in vertebrates might exist. However, although the identity of the ascidian APC/C inhibitor is not known, it is distinct from Emi2/Erp1, which is not present in invertebrate genomes, including urochordates (Russo et al., 2009; Fernandez-Guerra et al., 2006; Yamamoto et al., 2008). Further studies are thus needed to identify metaphase I CSF in invertebrate eggs. Once such a protein is found in urochordates, it will be possible to test the hypothesis that entry into meiosis II and determination of the number of meioses is due to re-establishment of CSF after meiotic exit and that re-establishment of CSF must be circumvented to enter the first mitotic cell cycle

Another aspect of our findings is that Mos controls meiotic events occurring after fertilization and so meiotic processes distinct from CSF-induced cell cycle arrest such as asymmetric cell division (ACD). This additional role of Mos in ACD is probably ancestral and may have originated with anisogamy. As egg size increased relative to the sperm size in the Eumetazoa, this created the need to conserve oocyte cytoplasm during both meiotic divisions. Indeed, most species display oocytes that form small PBs, which minimizes loss of oocyte cytoplasm during both meiotic divisions. Some insects have even dispensed with cytokinesis altogether during meiosis and retain all four sets of haploid chromatids in their cytoplasm (Drosophila and Sawfly). Interestingly, Mos is not required for meiotic progression in Drosophila (Ivanovska et al., 2004). Nevertheless, for the majority of Eumetazoan species, Mos/MAPK activity correlates with PB extrusion during the two phases of ACD. An interesting example of how the Mos/MAPK pathway correlates with meiotic cell divisions rather than with CSF cell cycle arrest is found in some species of lophotrochozoa (e.g. some molluscs and Echiuran worms). For example, oocytes from some mollusc and Echiuran species are arrested at the GV stage before fertilization when MAPK activity is low (Shibuya et al., 1992; Gould and Stephano, 1999). It is only after fertilization that the MAPK activity increases correlating with ACD during meiosis I and II (Shibuya et al., 1992; Gould and Stephano, 1999). This correlation suggests that MAPK is required for processes associated with meiosis I and II rather than CSF in these species. Indeed, in ascidians (this study) and in many other species (jellyfish, echinoderm, amphibian and mammals) disrupting Mos function inhibits PB extrusion (Amiel et al., 2009; Tachibana et al., 2000; Bodart et al., 2005; Choi et al., 1996). For example, mouse oocytes lacking Mos display large PBs and a number of other defects (Verlhac et al., 2000). Premature downregulation of the Mos/MAPK pathway in ascidian eggs prevents metaphase II spindle formation, further confirming the major role of this pathway not only in cell cycle arrest but also in meiotic spindle function.

Thus, in addition to CSF, the Mos/MAPK cascade imposes the meiosis-like phenotype partly characterized by a small peripheral spindle necessary for PB formation during ACD.

In conclusion, we propose that the Mos/MAPK pathway maintains an egg phenotype typified by a small peripheral spindle, successive rounds of M phase without intervening interphases or cyclin A accumulation and ongoing Ca²⁺ oscillations. Because urochrodates are at the crossroad between the invertebrates and the vertebrates (Delsuc et al., 2006) it will be interesting to elucidate which processes are shared with either invertebrates or vertebrates.

We thank the ANR (08-BLAN-0136-02), ARC, AFM and ATIP for financial support. R.D. is supported by CNRS. We also w thank Evelyn Houliston and Sandra Chevallier for providing RNAs coding for Ch-Mos::rfp1, Marie-Helene Verlhac for providing the plasmid pRN3-Mm-Mos, and Christian Rouvière for technical support with image acquisition and image analysis.