Spatial and temporal control of mitotic cyclins by the Gnu regulator of embryonic mitosis in Drosophila

In eukaryotes the cyclin dependent kinases (cdks), in conjunction with their cyclin partners, are the primary mediators effecting changes in cell cycle phase. During the earliest mitotic cycles of Drosophila embryos, total cyclin B levels do not cycle (Maldonado-Codina and Glover, 1992; Edgar et al., 1994). However, cyclin degradation is clearly important as inhibition of the cyclin degradation machinery with a destruction box peptide or injection of non-destructible cyclin B causes mitotic arrest in the early embryo (Su et al., 1998). Therefore mitotic cyclins may exhibit extreme spatial restriction in their cyclic degradation (Huang and Raff, 1999).

During Drosophila oogenesis, meiosis is initiated but the egg arrests at metaphase I (Jang et al., 1995). Movement of the oocyte through the oviduct is accompanied by its hydration and activation, which involves completion of the meiotic divisions (Fig. 1) (Heifetz et al., 2001). At the transition from meiosis to embryonic mitoses, Drosophila embryos require the maternal genes, gnu plu and pan gu (Freeman et al., 1986; Shamanski and Orr-Weaver, 1991) that encode proteins essential in unfertilized eggs and embryos to establish mitosis and prevent runaway nuclear DNA replication (Fig. 1). All three are small proteins restricted in their expression to oocytes, eggs and embryos. Plu bears ankyrin repeats, a motif for protein-protein interaction (Axton et al., 1994). Pan gu is a serine-threonine protein kinase (Fenger et al., 2000), and biochemical experiments have revealed that Png, Plu and Gnu form a protein kinase complex (Lee et al., 2003).

We have recently cloned the gnu gene and shown that it encodes a novel phosphoprotein that is dephosphorylated by protein phosphatase-1 at egg activation. Gnu is normally expressed in the nurse cells and oocyte of the ovary and is degraded during the embryonic mitoses (Renault et al., 2003). The temporal pattern of Gnu dephosphorylation parallels that of the maternal Ya protein, which is essential for the transition from oocytes to eggs and the early nuclear cycles (Liu et al., 1997). Ya is phosphorylated in ovaries and dephosphorylated at egg activation, and changes in its phosphorylation state correlate with Ya relocalisation to the nuclear lamina. The phosphorylation is mediated by a mitogen-activated protein kinase (MAPK) (Yu et al., 1999). We were interested to determine whether Gnu function or localisation are also regulated through its level of phosphorylation mediated by MAPK. Abolition of three MAPK consensus phosphorylation sites resulted in a Gnu protein with no mobility shift on immunoblots, however, this mutant protein was still functional.

Gnu is normally transported to the oocyte prior to meiosis, and is essential in embryonic mitoses, but is it also needed during meiosis? By imposing the translational and localisation controls of bicoid on gnu, we achieved an anterior-posterior gradient of Gnu that was first translated upon egg activation. Our results indicate that production of Gnu protein after egg activation is triggered is sufficient to rescue embryos laid by gnu mutant mothers.

Several lines of evidence indicate that cyclin B is a critical target of the giant nuclei class of genes. Cyclin B protein levels are decreased in embryos from png, plu, or gnu mutant females (Fenger et al., 2000). Mutation of cycB dominantly enhanced the weak png phenotype, whereas increasing the gene dosage of cycB suppressed the giant nuclei phenotype associated with mutations in png, plu and gnu (Lee et al., 2001). In the present study, embryos containing an anterior-posterior gradient of Gnu activity showed that, above a critical threshold level, Gnu regulates mitotic activity by promoting stability of cyclin B.

Constructs used are summarised in Table 1. The rescuing gnuGFP transformation construct P{w⁺ gnuGFP} in the pCasPer4 vector (Renault et al., 2003) was changed to P{w⁺ gnuGFP S/T>A} via site-directed mutagenesis (QuickChange; Stratagene) with the primers JD1F (5′ GAT CCC GCA gCC CCA CTG gCC CCA CTC gcT CCC CTC TCC 3′) and JD1R (5′ GGA GAG GGG Agc GAG TGG GGc CAG TGG GGc TGC GGG ATC 3′). Constructs were introduced into wild-type flies by P-element-mediated germline transformation (Roberts and Standen, 1998). Transformants were subsequently crossed into the w; gnu/TM3 background. The construct P{w⁺ gnuGFP} (Renault et al., 2003) was altered to P{w⁺ gnuGFP-bcd3′UTR} by the replacement of all gnu sequences 3′ to the 3′ BamHI site of the C-terminal green fluorescent protein (GFP) tag with a BamHI-EcoRI fragment of bicoid 3′ un-translated region (3′UTR) amplified from plasmid pbcdTN3 (Berleth et al., 1988) template with primers B3U EcoRI (5′ CGG AAT TCA GGG ACG GAA ATA TGG GCT A 3′) and B3U BamHI (5′ CGG GAT CCT GGA TGA GAG GCG TGT TAG A 3′). The inserts of both constructs were completely sequenced before transfer from pKS (Stratagene) to the pCasPer4 transformation vector.

Standard protocols for paraformaldehyde fixation of embryos and ovaries were used (Sullivan et al., 2000). Cyclin B immunostaining was achieved with methanol-fixed embryos (Huang and Raff, 1999). Photomicrographs were taken using an Eclipse 800 microscope (Nikon) with a MRC Radiance Plus laser scanning confocal system (BioRad) and LaserSharp software (BioRad). Images represent maximum-brightness projections of 6-10 confocal sections (1 μm) of whole stained egg chambers or embryos. GFP fluorescence was compared with that from ovaries and embryos from wild-type flies prepared in parallel, and imaged with the same confocal settings. All embryos are displayed with the anterior to the left, dorsal uppermost. DNA was stained with 1 μg/ml propidium iodide after 0.5 mg/ml ribonuclease (RNase A) treatment for 2 hours at 37°C. Primary antibodies used were rat anti-alpha tubulin 1 μg/μl (YL1/2, Serotec Ltd, 1:200), rabbit anti-cyclin A Rb270 and anti-cyclin B Rb271 (Whitfield et al., 1990) (1:500), mouse anti-γ-tubulin (Sigma, 1:1000), rabbit anti-phosphorylated histone-H3 (Upstate, 1:200). Secondary antibodies Cy5 anti-rat (Jackson, 1:500), Alexa488 anti-mouse and anti-rabbit (Molecular Probes, 1:500) were preabsorbed against methanol-fixed embryos.

In situ hybridisation was performed according to the method of White-Cooper et al. (White-Cooper et al., 1998). Ovaries and embryos were photographed on an Olympus BX50 microscope with DIC optics, CCD camera and Openlab software.

Proteins were extracted in 50 mM Tris-HCl pH 6.8, 100 mM NaCl, 1 mM benzamidine HCl, 1 mM phenyl methyl sulphonyl fluoride (PMSF), 2 mM dithiothreitol (DTT), 1 mM Na₃VO₄, 50 mM NaF, 10 mM β-glycerophosphate on ice. An equal volume of 2× SDS loading buffer was added and the sample boiled for 5 minutes. Boiling must be immediate because the phosphorylated form of Gnu and Gnu-GFP in ovaries is not stable. The samples were centrifuged for 5 minutes at 10,000 g. Supernatants were separated on 10% acrylamide/methylenebisacrylamide (37.5:1), 0.1% sodium dodecyl sulphate (SDS), pH 8.8 gels (SDS-PAGE) and transferred to polyvinilidene fluoride (PVDF) membrane by semi-dry electrophoresis. Equal amounts of protein were loaded in each lane and checked by Indian ink staining of the blot and by subsequently reprobing with mouse anti-α-actin antibody (ICN) at 1:500,000 dilution as an internal loading control. Mouse anti-GFP monoclonal antibody (Zymed) was used at a 1:500 dilution. Rabbit anti-Gnu antiserum (Renault et al., 2003) was preabsorbed on fixed 5- to 24-hour embryos and used at a dilution of 1:2000 for immunoblotting. Anti-cyclin A and anti-cyclin B antisera were both used at 1:2500 dilution. Detection was achieved with peroxidase-conjugated anti-rabbit and anti-mouse secondary antibodies (Vector) at 1:10,000 dilution, and Supersignal substrate (Pierce).

Three types of transgenes were used in this study (Tables 1 and 2). The first type, P{w⁺ gnuGFP}, expresses wild-type Gnu tagged to GFP; the second type, P{w⁺ gnuGFP S/T>A}, expresses a mutant version of GnuGFP in which MAPK phophorylation sites have been mutated from serine or threonine to alanines, and the third type, P{w⁺ gnuGFP-bcd3′UTR}, expresses wild-type Gnu tagged to GFP under control of the bicoid 3′UTR.

Ovarian Gnu protein migrates more slowly in SDS-PAGE than embryonic Gnu, suggesting that the protein is subject to post-translational regulation. The slower form can be prevented from converting to the faster-moving form in ovarian extracts by specific (Inhibitor-2) inhibition of protein phosphatase-1 (PP1), implicating phosphorylation (Renault et al., 2003). Gnu protein contains five potential MAPK Ser/Thr phosphorylation sites (Renault et al., 2003; Jacobs et al., 1999). Three of these sites are closely clustered together (Ser13, Thr16 and Thr19), suggesting that Gnu could be phosphorylated at more than one residue. Previously we showed that a transgene made from a genomic fragment of gnu with an insertion of GFP at the C-terminus, P{w⁺ gnuGFP}, can fully rescue the gnu mutation (Renault et al., 2003). We mutated the coding sequence of this transgene such that the Ser₁₃, Thr₁₆ and Thr₁₉ were replaced by alanines creating P{w⁺ gnuGFP S/T>A}. The resulting GnuGFP protein cannot be phosphorylated at these sites and no longer exhibited slower mobility in ovarian extracts (Fig. 2A, lanes 1-4), or in embryos (not shown). Two insertions (DF4 and GM3) of the P{w⁺ gnuGFP S/T>A} construct rescued gnu mutant females to fertility (Fig. 2A, lanes 1, 4). Although rescue was not fully penetrant, nuclear divisions were rescued in most embryos (Fig. 3A, Table 2). Therefore, reversible phosphorylation of Gnu on these sites is not essential for its function.

GnuGFP protein from two different insertions (BM1 and EM1) of P{w⁺ gnuGFP S/T>A} was expressed at low levels not detected by immunoblotting (Fig. 2A, lanes 2, 3) but gave partial phenotypic rescue (Table 2). Chromosomal position effect is sufficient to explain the low levels of Gnu expression of these two insertions but does not explain the observation that two insertions (i.e. DF4, GM3) with levels of Gnu expression similar to the wild-type construct P{w⁺ gnuGFP} (Fig. 2A, lanes 1, 4, O) did not rescue all of the embryos.

We compared the temporal and spatial expression of GnuGFP expressed from the P{w⁺ gnuGFP} and P{w⁺ gnuGFP S/T>A} constructs. GnuGFP from the wild-type P{w⁺ gnuGFP} transgene was detectable above w[1118] background in stages 9 and 10 (Fig. 2B) but was only strongly visible at stage 11 and later. GnuGFP from the rescuing P{w⁺ gnuGFP S/T>A} construct was similar except that it was expressed in egg chambers from stage 6 onwards, distributed throughout the nurse cells (Fig. 2C) and was not abundant in oocytes prior to stage 11 dumping. These small differences in expression (Fig. 2C compare with B) are unlikely to account for impaired function since the majority of Gnu is dumped into the oocyte at stage 11 and Gnu only becomes essential after egg activation (see below). We conclude that Gnu encoded by the P{w⁺ gnuGFP S/T>A} transgene that cannot be phosphorylated on the three N-terminal MAPK sites has slightly less than wild-type activity or its accumulation is subtly impaired in the oocyte. This is supported by the fact that the embryos from gnu females partially rescued by the P{w⁺ gnuGFP S/T>A} transgene (Fig. 3B,C) most closely resemble those from weak png alleles (Shamanski and Orr-Weaver, 1991).

Although the phenotype of gnu mutants is first evident after meiosis [references for plu can be found in Axton et al. (Axton et al., 1994)], we wondered whether Gnu has to be present during meiosis. We replaced the 3′UTR of gnu in the P{w+ gnuGFP} construct with that of bicoid (Berleth et al., 1988; Ephrussi and Lehmann, 1992) creating P{w⁺ gnuGFP-bcd3′UTR}. This temporally restricts Gnu expression by coupling its translation to egg activation. In contrast to the endogenous gene, expression from the P{w⁺ gnuGFP-bcd3′UTR} transgene was not detected in ovaries (Fig. 4A), but insertions S2, T1 and T2 rescued the gnu mutation to fertility (Table 2). This indicates that Gnu function is not essential before egg activation. However, although levels of GnuGFP expression in extracts of 0-3 hours embryos (Fig. 4A), in whole fixed embryos (Fig. 5B-E) and unfertilized eggs from females mated to sterile males (Fig. 5F) were comparable to those from the P{w⁺ gnuGFP} transgene, only a third of the embryos from w; P{w⁺ gnuGFP-bcd3′UTR}S2/+; gnu and w; gnu P{w⁺ gnuGFP-bcd3′UTR}T1 mothers exhibited normal syncytial stages (Table 2). The bcd3′UTR also confers spatial control since the mRNA is localized to the anterior of the oocyte. GnuGFP from the P{w⁺ gnuGFP-bcd3′UTR} transgene was therefore translated at the anterior pole of the embryo and diffused posteriorly, creating a concentration gradient of Gnu protein (Fig. 5B-F). In approximately a third of the embryos rescued by the S2, T1 and T2 inserts of P{w⁺ gnuGFP-bcd3′UTR} transgene, a gradient of mitotic activity was apparent (Fig. 5B-F, Table 2). In these embryos, nuclei underwent mitosis normally in the anterior region but failed to migrate into the most posterior region, instead becoming polyploid. A further third of the embryos contained a mixture of mitotic and polyploid nuclei without forming an obvious gradient (Fig. 3B, Table 2).

The gradient of Gnu activity in gnu embryos rescued by P{w⁺ gnuGFP-bcd3′UTR} allowed us, for the first time, to examine the earliest onset of the gnu mutant phenotype in embryonic nuclei. We used this to test two hypotheses: firstly, that Gnu, Plu and Pan gu act to regulate the cell cycle by promoting cyclin B translation or stability (Fenger et al., 2000; Lee et al., 2001) and, secondly, that Gnu is required to tether centrosomes to nuclei (Freeman et al., 1986).

Although cycB mRNA levels and localization in ovaries and embryos from gnu mutant mothers were indistinguishable from wild type (Fig. 6) (Lee et al., 2003), their embryos contain very little cyclin A and B protein (Fig. 4B) (Fenger et al., 2000). Given that Gnu is only essential after egg activation, we were surprised to find that levels of both cyclin A and B protein were also reduced in gnu mutant ovaries (Fig. 4B). Restoring Gnu levels specifically in embryos with the P{w⁺ gnuGFP-bcd3′UTR} transgene (Fig. 4A) restored normal levels of cyclin A and B proteins in these embryos but, as expected, did not restore cyclin A and B levels in ovaries (Fig. 4B).

Comparison within an embryo (Fig. 7A-D) showed that, in the fully rescued anterior region, cytoplasmic cyclin B and metaphase spindle-associated cyclin B showed a staining pattern similar to wild type (Huang and Raff, 1999). In the posterior region where Gnu is limiting, cyclin B was lost from the cytoplasm and eventually from the mitotic spindles. These anterior-posterior gradients of cytoplasmic cyclin B were seen in many embryos, whether in mitosis or in interphase and cytoplasmic gradients were seen even in embryos where all nuclei were synchronized in the cell cycle.

Cyclin B is responsible for the mitosis promoting activity of MPF, leading to phosphorylation of mitotic target proteins, such as histone H3 (Su et al., 1998). In the posterior region of Gnu gradient embryos, below the critical threshold, loss of cyclin B caused a loss of phospho-histone H3, a marker of MPF activity (Fig. 7E). Anaphase chromosomes normally retain phospho-histone H3 staining during segregation (Fig. 7F) (Su et al., 1998). At threshold concentrations of Gnu, phospho-histone H3 staining was progressively lost from the centromere of the chromosomes (centrosome ends of the spindles) before the chromosomes had fully segregated.

Chromosome segregation problems are seen in a number of abnormal cell cycles, for example after experimental inhibition of S-phase with aphidicolin (Raff and Glover, 1988), in embryos defective for the Aurora protein kinase (Glover et al., 1995) and in the case of mutations in DNA damage checkpoints (Sibon et al., 1997; Yu et al., 2000). However, loss of histone H3 phosphorylation is consistent with loss of MPF activity that would result from our observation of reduced levels of cyclin B in the posterior region, even on metaphase spindles (Fig. 7C,D). We infer that the earliest onset gnu phenotype is, therefore, a loss of MPF activity due to low cyclin B levels on the mitotic spindle and the consequent failure of anaphase segregation. Tetraploid mitotic nuclei indicated that the first defect as Gnu levels fall below the functional threshold is a mitotic cycle without anaphase resolution of the chromosomes. The centrosomes attached to these nuclei indicate that the detachment of centrosomes is secondary.

How might Gnu, Plu and Pan gu act to regulate the early embryonic mitotic cycles? We have suggested, on genetic evidence, that the three proteins control a single process (Renault et al., 2003) and the finding that Pan gu is a protein kinase (Fenger et al., 2000) indicates that control might be exerted at the level of selection and phosphorylation of the substrates of Pan gu. We do not believe Gnu to be one of these essential substrates, since the mobility of Gnu in extracts from ovaries and embryos is unaffected by presence or absence of Pan gu (Renault et al., 2003). While it is still possible that Pan gu phosphorylation of Gnu has simply not been detected by us, we have shown here that removal of three MAPK consensus phospho-acceptor sites in the N terminus of the Gnu protein abolished its altered electrophoretic mobility in ovaries. Despite this, the mutant transgene still rescued the gnu mutation. Gnu therefore does not require reversible phosphorylation on these sites to function in the embryo. However, we cannot rule out the possibility that phosphorylation at other sites that do not result in detectable electrophoretic mobility changes could be important for Gnu function.

The gnu mutant phenotype of inappropriate DNA replication is first evident in eggs or embryos. However, we report here that gnu has an earlier function in that the levels of both cyclin A and cyclin B are reduced in gnu mutant ovaries. To test whether this early role is essential, we used the 3′UTR of bicoid to replace nurse cell translation of Gnu with translation in the oocyte at or after the time meiosis is completed at egg activation. Translation of Gnu at egg activation was able to rescue fertility of gnu mutant females demonstrating that Gnu is not essential before egg activation.

The bcd3′UTR also confers localized translation from the anterior pole of the embryo. The P{w⁺ gnuGFP-bcd3′UTR} embryos containing anteroposterior gradients of Gnu fluorescence and activity show that mitosis is supported above a critical threshold level of Gnu. The extreme spatial sensitivity of embryonic nuclei to Gnu activity suggests that the concentration of Gnu protein is critical for mitosis in Drosophila embryos.

Our constructs constitute an allelic series affecting the level of Gnu in whole embryos (P{w⁺ gnuGFP S/T>A} construct), or the local level within an embryo (P{w⁺ gnuGFP-bcd3′UTR} construct). The reduced levels of Gnu reaching the oocyte from P{w⁺ gnuGFP S/T>A} constructs resulted in a range of embryonic phenotypes characterized by progressive loss of the mitotic phase and the onset of excessive nuclear DNA replication. The first threshold defects seen in the mitotic gradients in embryos rescued by the P{w⁺ gnuGFP-bcd3′UTR} construct were loss of anaphase segregation and a mitotic onset of polyploidy rather than the non-mitotic endo-cycle seen in gnu, plu and pan gu loss-of-function mutants. Since centrosomes were seen normally associated with nuclei undergoing abnormal division, physical anchoring of centrosomes to nuclei or synchronising the centrosome cycle to that of nuclear replication (Freeman et al., 1986) is unlikely to be the primary function of Gnu protein in the embryo.

Wild-type mature oocytes and unfertilized eggs are replete with cyclin A and B protein. However, ovaries (Fig. 4B) and embryos (Fig. 4B) (Fenger et al., 2000) from gnu mutant mothers contain reduced levels of cyclin A and B protein. Since cycB mRNA localization and levels in embryos from gnu mutants were indistinguishable from wild type (Fig. 6) (Lee et al., 2003), decreased cyclin B protein levels in gnu mutant embryos are not a consequence of decreased cycB transcript. Therefore gnu could act to either activate translation or regulate degradation. Several lines of evidence suggest the later is more likely. Firstly, different phosphorylation forms of cyclin A are differentially reduced in gnu, plu and png mutants, which suggests that cyclin A is unstable in these mutants (Fenger et al., 2000). Secondly, blocking translation, but not cyclin degradation, with cycloheximide does not lead to inappropriate DNA replication, although nuclei do become enlarged (Zalokar and Erk, 1976).

We cannot currently exclude that Gnu, Plu and Pan gu act both to stimulate cyclin A and B translation and to inhibit their global cytoplasmic degradation. However, we prefer the simpler model in which Gnu (with Plu and Pan gu) acts to maintain the stability of global cytoplasmic cyclin B (Fenger et al., 2000). Gnu stabilisation of cytoplasmic cyclin B stockpiled in oogenesis is key to a developmental strategy that enables extremely rapid embryonic mitoses to take place via localized spindle-limited anaphase degradation, obviating a need for extensive cyclin B translation in the early embryo. In support of this idea, Gnu becomes undetectable 1-2 hours after fertilization (Renault et al., 2003). At the same time, cycles of cyclin B synthesis and degradation first become detectable in embryo extracts (Edgar et al., 1994). Since cyclin B can inhibit DNA replication and chromosome condensation in mitosis (Hayles et al., 1994; Donaldson and Blow, 1999), the reduction in cyclin B protein in the gnu mutant might account for the over-replication phenotype seen in the gnu mutant. The mechanism by which gnu regulates cyclin B protein levels remains to be determined.

Plu, Png and Gnu have functions in addition to their effects on the S-M cycles via cyclin B and Cdk1 activity. gnu, plu and png are needed for the destabilization of certain maternal transcripts that occurs at egg activation (Tadros et al., 2003). The mechanism through which they play this role is also unclear although it appears to be Cdk1 independent (Tadros et al., 2003).

The P{w⁺ gnuGFP-bcd3′UTR} transgene includes the gnu promoter, to provide low level expression in the female germline, and bcd3′UTR, which confers anterior localisation and translation upon egg activation. The transgene is a useful tool for developmental cell cycle research in two important respects. Firstly, the vector can be used to rescue maternal effect mutations using an appropriate ORF and a graded response to the rescuing protein can be assessed within a single embryo. In early development, nuclei proliferate in an initially anterior position, then axial expansion pushes some of them into a more posterior region. These nuclei are now in a region of low protein abundance where, as in the Gnu example, they would develop mutant phenotypes below the functional threshold. Antibody staining for cell cycle proteins can be examined in single embryos using the rescued anterior region as a control for the abnormal posterior region. Secondly, the vector also provides a novel experimental strategy to permit translation of candidate cell cycle activators – including potential sperm supplied factors – in unfertilized eggs.

We conclude that Gnu protein is transported from nurse cell to oocyte, but is first required after egg activation is triggered, to promote zygotic mitoses. Following fertilisation, Drosophila embryos progressively degrade a small fraction of their maternal cyclin B protein stockpile on mitotic spindles every anaphase. This developmental strategy is possible because Gnu acts to stabilize the bulk cytoplasmic pool of cyclin B protein during the embryonic mitoses.

We gratefully acknowledge the Cancer Research Campaign (Cancer Research UK) for Project Grant support and the MRC for Co-operative Group and Project Grants. L.L. was supported by a Nuffield Bursary. We thank Christiane Nüsslein-Volhard for the pbcdTN3 clone, David Glover for antibodies, Karen Clifton and Attila Tasnádi for maintaining fly stocks, Jianqiao Jiang, James Wakefield and Luke Alphey for helpful advice. The authors declare they have no competing financial interests.