An essential role for the RNA-binding protein Smaug during the Drosophila maternal-to-zygotic transition

In oviparous organisms that deposit their eggs into the environment,embryogenesis is initiated by a cleavage stage characterized by a rapid,simplified S-phase/M-phase cell cycle and little or no zygotic transcription(Etkin, 1988; O'Farrell et al., 2004). The cleavage stage is driven by mRNAs and proteins that are deposited into the oocyte, and is therefore under maternal genetic control. Transfer of developmental control to the genome of the embryo occurs during the maternal-to-zygotic transition (MZT), when a subset of the maternal mRNAs is eliminated and zygotic transcription is initiated(Schier, 2007; Tadros et al., 2007b; Zurita et al., 2008). While this transfer of genetic control is under way, the cleavage stage divisions slow and cell cycle checkpoints are activated. The MZT ends with the midblastula transition (MBT), which is defined as the first developmental process that depends on zygotic rather than maternal mRNAs.

Typically, the MBT follows a fixed number of cleavage divisions. In Drosophila, for example, the MBT occurs during interphase of nuclear cycle 14. However, haploid Drosophila embryos undergo an additional division before the MBT whereas Drosophila embryos injected with exogenous DNA proceed through fewer divisions before the MBT(Edgar et al., 1986). Similarly, polyspermic Xenopus embryos undergo fewer cleavage divisions before the MBT (Newport and Kirschner, 1982). These studies suggested that titration of a maternally deposited factor by nuclei or DNA – which increase exponentially during the cleavage divisions – might determine the timing of the MBT. However, the timing of maternal Cyclin E destruction in Xenopus and degradation of a pool of maternal mRNAs in Drosophila are independent of the cleavage divisions(Bashirullah et al., 1999; Howe and Newport, 1996; Tadros et al., 2003). In addition, RNAi-mediated knock down of cyclins leads to premature termination of the Drosophila cleavage divisions, but does not alter the timing of cellularization (McCleland and O'Farrell, 2008). These findings suggest that the MBT is under the control of a developmental clock, but the molecular mechanisms that drive this clock remain to be defined.

In Drosophila, the embryonic cleavage divisions slow progressively during the syncytial blastoderm stage, when the majority of nuclei are in a monolayer at the cortex (Foe and Alberts,1983). Following cleavage 13, the cortical nuclei are surrounded by membranes in a process termed cellularization, which is the first morphological transformation that requires zygotic gene expression and therefore represents the MBT (Mazumdar and Mazumdar, 2002; Merrill et al., 1988; Schweisguth et al.,1991; Wieschaus and Sweeton,1988). DNA replication checkpoint mutants fail to slow the syncytial blastoderm stage cell cycle, cellularize or initiate high-level zygotic transcription, which suggested a direct role in the MZT(Sibon et al., 1999; Sibon et al., 1997). However,the cellularization and transcriptional defects associated with checkpoint mutants are dramatically suppressed by mnk (encoding checkpoint kinase 2, Chk2), a mutation that disrupts DNA damage signaling(Takada et al., 2003; Takada et al., 2007). Furthermore, DNA damaging agents can block cellularization and zygotic transcription (Takada et al.,2007). The replication checkpoint thus controls the syncytial divisions and helps maintain DNA integrity, but does not appear to directly control the MZT.

Maternal transcript destruction during the Drosophila MZT requires the conserved RNA-binding protein Smaug(Tadros et al., 2007a). Smaug recruits the CCR4/POP2/NOT deadenylase complex to its target transcripts,triggering poly-A tail removal and transcript destruction(Semotok et al., 2005; Semotok et al., 2008; Zaessinger et al., 2006). Here, we show that smg mutant embryos fail to slow the syncytial blastoderm cell cycle, terminate cleavage divisions, activate the DNA replication checkpoint, cellularize or gastrulate. We also show that Smaug is required for high-level zygotic transcription at the MZT. Significantly,transgenic expression of Smaug in an anterior-to-posterior gradient induces an anterior-to-posterior gradient in the onset of syncytial cell cycle delays,maternal transcript destruction, zygotic gene activation, blastoderm cellularization and even gastrulation. Smaug thus coordinates both the cleavage-dependent and cleavage-independent aspects of the MZT, and may drive a molecular clock that controls the timing of this key developmental transition.

smg and grp mutant embryos were generated by crossing homozygous smg¹/smg^Df(3L)Scf-R6(Dahanukar et al., 1999) or grp^fs(A)4/grp^fs(A)4(Sibon et al., 1997) females to wild-type Oregon R or w¹¹¹⁸ males. The Smaug protein gradient was generated using the UASp-smg-bcd3′UTR(USB) transgenes (Tadros et al.,2007a) and germline driver Nanos-Gal4:VP16 (NGV)(Rorth, 1998). Females with the genotype w¹¹¹⁸; smg¹ USB /smg¹NGV were crossed to wild-type Oregon R males. To assay for phenotypic suppression by mnk, embryos were derived from w¹¹¹⁸;mnk⁶⁰⁰⁶/mnk⁶⁰⁰⁶;smg¹/smg^Df(3L)Scf-R6 females crossed to wild-type males (Abdu et al.,2002). Unfertilized eggs were obtained from Oregon-R females crossed to sterile males of the genotype T(Y;2)bw ^DRev#1, cn bw^DRev#11mr2/SM6a (Reed and Orr-Weaver, 1997). The smg¹, Df(3L)Scf-R6,mnk⁶⁰⁰⁶ and Nanos-Gal4:VP16 stocks were obtained from the Bloomington Drosophila Stock Center.

Cell-cycle progression and replication checkpoint function were assayed by time-lapse confocal microscopy using a Leica TCS-SP inverted laser-scanning microscope, as described previously (Sibon et al., 1997). Time-lapse DIC microscopy was performed using Zeiss Axiovert–S100 inverted microscope and 20× Planapo lens. Images were captured at 20-second intervals, and MetaMorph software was used for image acquisition and processing. The length of cellularization was defined as the time between the end of mitosis 13 and completion of membrane invagination. Basic statistical analyses were performed in Excel(Microsoft).

Fluorescence in situ hybridization using RNA probes was performed as described previously (Takada et al.,2007). For slam antisense RNA probes, the DNA template was amplified by PCR from wild-type genomic DNA, using the following sets of primers: slam 5′-ctgttcagtccgattctcatcc-3′ and 5′-T7-aatcttgtccatgtgctcgctg-3′. The T7 promoter sequence was:5′-cgtaatacgactcactataggg-3′. For cyclin A and Bantisense RNA probes, the DNA templates were amplified by PCR from plasmids containing the appropriate cDNA (Lehner and O'Farrell, 1989; Lehner and O'Farrell, 1990) using T3 and T7 primers. Antisense RNAs were transcribed with the T7 polymerase. Images were acquired using a Leica TCS-SP inverted laser-scanning microscope and processed with Photoshop.

For whole-mount immunostaining, embryos were fixed and immunolabeled as described previously (Theurkauf,1994). Rabbit anti-phospho-Histone H3 (P-H3) (Upstate Innovative Cell Signaling Solution) was used at 1:500; mouse anti-phospho-Tyrosine(P-Tyr) (Cell Signaling Technology) was used at 1:1000; guinea pig anti-Smaug(Tadros et al., 2007a) was used at 1:1000. The following fluorescent secondary antibodies were used: goat anti-rabbit Alexa 488 (Molecular Probes, 1:500); donkey anti-mouse Alexa 488(Molecular Probes, 1:500); donkey anti-guinea-pig FITC (Jackson ImmunoResearch, 1:500). Embryos were labeled for DNA with TOTO3 (Molecular Probes) at 0.1 μM and treated with RNase (Promega) at 10 U/ml during the secondary antibody incubation. Images were acquired using a Leica TCS-SP inverted laser-scanning microscope and processed with Photoshop.

To quantify the Smaug gradient, single plane images of cycle 13 embryos were captured using a Leica TCS-SP inverted laser-scanning microscope. Imaging was carried out under non-saturating conditions, and all genotypes were labeled and imaged under identical conditions. Pixel intensity (from 0 to 250)was extracted from raw images using MetaMorph software (Molecular Probes) and exported to Excel files. Immunolabeling and imaging of the embryo's interior is inefficient, owing to antibody penetration and light scattering. Analysis was therefore restricted to a four-pixel band at the cortex. Average pixel intensity as a function of position along the anterior-posterior axis was then calculated from an analysis of four independent embryos. To derive the profile of the gradient, the average cortical pixel intensity in wild-type embryos was subtracted from the average pixel intensity in wild-type embryos expressing the USB transgene.

The following primary antibodies were used for western blotting:anti-α-Tubulin (Sigma, 1:20,000, mouse), anti-β-Tubulin (E7 DSHB,1:500, mouse), anti-large subunit of RNA polymerase II (ARNA-3a from Research Diagnostics, 1:500, mouse), anti-phospho-ser-2 from the C-terminal domain of RNA polymerase II (H5 from Research Diagnostic, 1:500, mouse) and anti-Smaug(1:5000, guinea-pig). The following HRP-conjugated secondary antibodies were used at a 1:1500 dilution: sheep anti-mouse (Amersham) and donkey anti-guinea-pig (Jackson ImmunoResearch). Embryos were labeled and hand-selected as described previously(Edgar et al., 1994). The equivalent of two embryos or one quarter of an ovary was loaded for each lane. For western analysis of developmentally aged embryos, adult male and female flies were placed in collection cages and fed yeast paste on grape juice agar plates. Cages were kept at 25°C and plates were changed twice before collection for 1 hour. The embryos were then used immediately (0-1 hour time point), or aged 1 to 3 hours. The equivalent of 2.5 embryos of the desired age was loaded for each lane. The ECL Plus detection system as used to detect antibody binding (Amersham). Blots were imaged and quantified using a Kodak Image Station and Excel.

Total RNA was extracted from staged fertilized or unfertilized eggs, as well as from stage 14 oocytes as described previously(Tadros et al., 2007a). To assay mRNA quality, known stable (rpA1) and unstable (Hsp83)transcripts were probed on northern blots. Total RNA was then reverse transcribed with random primers (Tadros et al., 2007a) and labeled as described in the Indirect Labeling of Total RNA for Microarray Hybridization protocol at http://www.flyarrays.com. The fluorescently labeled cDNA probes were hybridized to 12Kv1 and 14Kv1 microarrays obtained from the Canadian Drosophila Microarray Centre(http://www.flyarrays.com;GEO platform accession numbers: GPL1467, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL1467;and GPL3603: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL3603). Hybridization and scanning were performed as previously described(Neal et al., 2003). The 16 bit TIFF image files were quantified using QuantArray Version 3 (PerkinElmer),using the adaptive quantification algorithm and analyzed using GeneTraffic Duo3.2 (Iobion Informatics/Stratagene). The 12Kv1 arrays were normalized using the rank invariant LOWESS extrapolation method(Schadt et al., 2001) and the 14Kv1 arrays were normalized to a set of known stable transcripts: Rpl1,RpL32, Rps5a, Rps3, RpL22, mRpS30, mRpS22, CG6764, bonsai, RpS29, RpL11,mRpL1, CG6764, CG317, RpL37A and RpL40. Additional information on data analysis can be provided on request. To profile expression of miRs, after extraction with Trizol reagent, total RNA was further purified using the RNeasy Mini Kit (50) (Qiagen) as recommended by Exiqon. RNA samples were then sent to Exiqon for miRCURY LNA Array microRNA Profiling. mRNA and miRNA expression profiling data are available from the Gene Expression Omnibus (GEO)server (Accession Numbers GSE13287 and GSE13288, respectively).

miRCURY LNA Detection Probes were purchased from Exiqon. Northern blotting was performed described at http://www.exiqon.com/MicroRNA_northern_blot,except that a 15% polyacrylamide gel was used. For all samples, 12 μg of total RNA was loaded per lane. The hybridization temperature was 40°C for all miRCURY LNA probe-to-membrane hybridizations.

Mutations in smaug (smg) disrupt nuclear organization during the syncytial blastoderm stage(Dahanukar et al., 1999). To gain insight into the genesis of these defects, we directly assayed cycle progression and spindle assembly in smg mutants. For these studies,embryos were injected with rhodamine-Tubulin and the DNA dye OliGreen and monitored by time-lapse confocal microscopy for cell cycle progression. Interphase length normally increases from 6 to 14 minutes between division cycles 11 and 13. In smg mutants, by contrast, cell cycle length did not increase and interphase 13 averaged only 7 minutes(Fig. 1A). The smgmutant embryos also progressed through additional syncytial cycles after mitosis 13 (Fig. 1A) and cellularization appeared to fail. These defects were confirmed by time-lapse differential interference contrast (DIC) microscopy (see Table S1, Movies 1 and 2 in the supplementary material), which reduces photo-damage and reveals membrane invagination, as well as by labeling with an anti-phospho-tyrosine antibody, which reveals cellular membranes in early Drosophilaembryos (see Fig. S1A-C in the supplementary material).

The DNA replication checkpoint is required to slow the cleavage divisions(Sibon et al., 1999; Sibon et al., 1997). To determine whether smg mutations disrupt the replication checkpoint,we injected syncytial blastoderm stage embryos with the DNA replication inhibitor aphidicolin, OliGreen and rhodamine-Tubulin, and assayed progression into mitosis by time-lapse confocal microscopy(Sibon et al., 1997). In wild-type embryos, aphidicolin induced progressively more significant interphase delays during cycles 11 and 13(Fig. 1B)(Crest et al., 2007),indicating that replication checkpoint activity increases progressively during the syncytial blastoderm stage. In smg mutants, by contrast,aphidicolin failed to trigger significant cell cycle delays during any of the syncytial blastoderm divisions (Fig. 1B). Smaug is therefore required for replication checkpoint activation during the Drosophila MZT.

We next assayed activation of zygotic gene expression by microarray-based gene expression profiling. For these studies, we compared transcript abundance in embryos 2-3 hours post-fertilization with transcript abundance in mature oocytes. Zygotic gene expression is normally activated at high levels between 2 and 3 hours after fertilization, whereas mature oocytes contain the full complement of maternally loaded mRNA. The zygotically transcribed genes fall into three distinct classes (Fig. 2A; see Table S2 in the supplementary material). Class I genes are strictly zygotic, as they are not present in oocytes but are expressed in 2-to 3-hour-old embryos. The remaining two classes are present in mature oocytes, and are therefore maternally loaded. Class II mRNAs are stable through the MZT, but increase in abundance as transcription is initiated. Class III transcripts are degraded during this transition, and then re-expressed zygotically.

Smaug is required for upregulation of 85% of the Class I genes and 90% of the Class II genes, including the vast majority of the highly transcribed zygotic genes (Fig. 2A). By contrast, smg mutations blocked expression of only 15% of Class III genes (Fig. 2A). Smaug is required for destruction of the maternal pool of most Class III transcripts(75%) (Tadros et al., 2007a),so continued expression of these genes in smg mutant embryos could reflect stabilization of the maternal pool, Smaug-independent transcription in the embryo or a combination of these two factors. In summary, Smaug is required for normal zygotic accumulation of transcripts from the vast majority of protein-coding genes that are highly expressed during the MZT.

Transcript elongation is linked to phosphorylation of the large subunit of RNA polymerase II (RNAPII) on serine 2 of the C-terminal domain (CTD)(Majello and Napolitano,2001). Accumulation of this phosphorylated form of RNAP II,designated II0, also correlates with the increase in transcription during the Drosophila MZT (Bellier et al.,1997; Dantonel et al.,2000; Leclerc et al.,2000; Palancade et al.,2001). However, II0 is not detectable in smg mutant embryos (Fig. 2B). By contrast,total RNAPII levels are similar in smg mutants and wild-type embryos(Fig. 2B). The smgmutation thus appears to block full activation of the transcription machinery during the MZT.

DNA damage and Chk2 activation can inhibit zygotic transcription and cellularization (Takada et al.,2007), thus the transcription defects in smg mutants could have been secondary to checkpoint failure and DNA damage. We therefore assayed RNAPII CTD II0 phosphorylation in mnk; smg double-mutant embryos (Fig. 2B). The defects in II0 accumulation in smg mutants were only partially suppressed in mnk; smg double mutants. In addition, these double-mutant embryos consistently failed to cellularize or gastrulate (see Fig. S1F in the supplementary material). By contrast, grp checkpoint mutant embryos also showed reduced levels of CTD phosphorylation, but this was restored to wild-type levels in mnk grp double mutants(Fig. 2B). The transcriptional block in smg mutants is therefore independent of checkpoint failure and Chk2 activation.

To determine whether the defects in maternal transcript destruction in smg mutants are secondary to checkpoint failure, we assayed maternal mRNA destruction in grp and mnk grp checkpoint mutants. Fluorescence in situ hybridization (FISH) studies showed that Hsp83(Semotok et al., 2005; Semotok et al., 2008) and cyclin B mRNAs were stabilized in smg mutants(Fig. 3D,F,J,L; and data not shown), but degraded with normal kinetics in grp and mnk grpdouble mutants (Fig. 3K; see Fig. S2F in the supplementary material, data not shown). Cyclin AmRNA was also degraded in grp and mnk grp double mutants,but some transcripts appeared to persist(Fig. 3E, data not shown). These studies indicate that maternal transcript destruction is largely independent of replication checkpoint control and DNA damage signaling.

High-level zygotic expression of miR-309 cluster microRNAs (miRs)directs destruction of a subset of maternal mRNAs at the MZT(Bushati et al., 2008). Hybridization of total RNA purified from staged eggs or embryos to a microarray carrying probes for 68 known Drosophila miRs confirmed that high-level miR-309-cluster transcription occurs only in embryos and not in activated unfertilized eggs (see Fig. S3 in the supplementary material). Northern blot analysis demonstrated that expression of three of the miRs – miR-3, miR-6 and miR-286 – is disrupted in smg mutants (Fig. 2C). Four-hundred and ten maternal mRNAs appear to require zygotic expression of the miR-309 cluster for destabilization at the MZT(Bushati et al., 2008). Consistent with our observations, ∼85% of these maternal transcripts are stabilized in smg mutants (Fig. 2D). Smaug-dependent expression of the miR-309-cluster thus leads to further destabilization of a subset of maternal mRNAs.

smg mRNA is translationally silent in mature oocytes, but is translationally activated on egg activation(Tadros et al., 2007a), and Smaug protein is present at high levels in early embryos(Dahanukar et al., 1999; Smibert et al., 1999). To define the precise kinetics of Smaug protein accumulation during the MZT, we performed western blots on embryos that had been hand sorted according to cell division cycle number (see Materials and methods). These studies showed that Smaug levels progressively increase through the early cleavage divisions, peak during syncytial blastoderm cycles 10 through 13, and then dramatically decline during interphase of nuclear cycle 14(Fig. 4). Thus, initiation of Smaug expression correlates with initiation of maternal mRNA degradation,whereas the peak in Smaug accumulation correlates with the onset of zygotic gene expression and replication checkpoint activation.

Smaug protein levels drop dramatically during early interphase 14, when transcription increases dramatically. In unfertilized eggs, which do no initiate zygotic transcription, Smaug protein accumulates with normal kinetics but then persists at high levels (Fig. 4). Our gene expression profiling data show that maternal smg mRNA is stable through the MZT. Smaug-dependent transcription of early zygotic genes may therefore block smg mRNA translation or decrease Smaug protein stability. This may prevent destruction of zygotic mRNAs carrying Smaug-binding sites, and thus promote the burst of transcription at the end of the MZT.

To determine whether Smaug accumulation functions as a trigger for the MZT,we expressed this protein in an anterior-to-posterior gradient and assayed cell cycle progression, transcription, cellularization and gastrulation. If Smaug functions as a trigger, changing protein levels will change the timing of downstream processes rather than the extent to which these processes are completed. By contrast, if Smaug has a direct role in multiple processes at the MZT, the gradient should produce the equivalent of an allelic series,which would change the extent of phenotypic rescue, but not the timing of rescue. For example, different alleles of scraps produce cellularization defects that differ in severity, but these alleles do not appear to change the timing of this process(Field et al., 2005).

To generate a Smaug gradient, we expressed a hybrid transgene in which the smg open reading frame was fused to the bicoid 3′UTR, in the smg¹ mutant background(Tadros et al., 2007a)(Fig. 5C). We quantified the resulting gradient using immunolabeling for Smaug (see Materials and methods). The smg¹ allele produces a truncated 60 kDa protein that is detected by the anti-Smaug antibody(Fig. 5E). We therefore estimated the shape of the gradient by quantifying immunofluorescence labeling along the anterior-posterior axis in wild-type embryos expressing the USB transgene, and then subtracting the uniform fluorescence observed in control wild-type embryos (Fig. 5D). Immunofluorescence labeling intensity may not be linear with respect to protein concentration, thus these data provide only a rough estimate of the shape of the gradient, not precise information on protein concentration. Nonetheless, these studies indicate that Smaug is expressed above wild-type levels at the anterior pole, drops below average wild-type levels between 75 and 80% egg length (EL), and progressively declines between 75% and 0% egg length (0% and 100% EL represent the posterior and anterior pole, respectively).

To assess the dynamics of the syncytial divisions and cellularization, we analyzed early development by time-lapse differential interference contrast(DIC) microscopy. In wild-type syncytial blastoderm stage embryos, mitosis is synchronous or is initiated at the poles and progresses in converging waves toward the center (Foe and Alberts,1983) (see Movie 1 in the supplementary material). In smg-USB embryos, by contrast, mitosis was delayed at the anterior,leading to posterior-to-anterior waves of nuclear division (see Movies 3 and 4 in the supplementary material). To confirm these observations, we analyzed the distribution of phosphorylated histone H3, a marker for mitosis, along the axis of the embryos. Wild-type embryos showed uniform H3 phosphorylation,labeling at both poles or labeling through the center(Fig. 6C, data not shown). In smg-USB embryos, however, phosphorylated-histone-H3-labeled mitotic nuclei were often observed at the posterior and unlabeled interphase nuclei were present near the anterior pole, indicating a delay in mitosis at the anterior (Fig. 6D). Expressing Smaug in an anterior-to-posterior gradient thus triggers anterior-to-posterior delays in cell cycle progression.

The Smaug gradient also consistently led to a gradient in the timing of cellularization (see Movies 3 and 4 in the supplementary material; Fig. 6F). We found that 100% of wild-type embryos initiated cellularization synchronously along the anteroposterior axis, with a mean mid-cellularization time of 93 minutes after syncytial interphase 10 (see Movie 1 in the supplementary material; Fig. 6E; Fig. S4 in the supplementary material). By contrast, 100% of embryos expressing Smaug in a gradient showed a striking anterior-to-posterior cellularization gradient(n=12, Fig. 6F; see Movies 3 and 4 in the supplementary material). In these embryos,cellularization was initiated over the anterior 20 to 25% of egg length with essentially wild-type timing (mean of 94 minutes after syncytial mitosis 10,see Fig. S4 in the supplementary material). Over the remaining 75% of egg length, where Smaug drops below wild-type levels, cellularization was progressively delayed and mid-cellularization was not observed until 117 minutes after interphase 10 at the posterior pole (see Fig. S4 in the supplementary material). In addition, embryos derived from hemizygous females,which express Smaug at 50% of wild-type levels, proceeded through the normal number of syncytial divisions, but cellularized 10 minutes later than embryos derived from wild-type diploid mothers (data not shown). Reducing Smaug expression thus leads to progressive delays in the MBT.

In 8 of 12 live recordings, the entire embryo completed 13 syncytial divisions before cellularizing in a graded manner (see Movie 3 in the supplementary material). In 4 out of 12 embryos, however, division at the anterior pole terminated after mitosis 12, whereas the rest of the embryo progressed through the normal 13 syncytial divisions (see Movie 4 in the supplementary material; Fig. 6B). However, the timing of cellularization at the anterior pole was the same in both classes of embryos. Smaug overexpression thus triggers premature arrest of the cleavage divisions, but does not advance the MBT.

In wild-type embryos, the gastrulation movements of germband extension and dorsal migration of the pole cells were observed before the cephalic furrow could be clearly identified by DIC (see Movie 1 in the supplementary material). In embryos expressing Smaug in a gradient, by contrast, cephalic furrow formation initiated before cellularization was completed at the posterior pole (see Movies 3 and 4 in the supplementary material). Higher resolution time-lapse DIC imaging confirmed that cellularization and gastrulation were significantly delayed, but these processes appeared to be completed (see Fig. S5, Movies 5 and 6 in the supplementary material), and 19%of embryos expressing the Smaug gradient hatched (n=300). Reducing Smaug thus delays the early developmental program, but does not appear to block execution of this program.

To determine the influence of the Smaug gradient on zygotic gene activation, we used FISH to assay zygotic expression of slam and runt (Fig. 7), which are required for cellularization and segmentation, respectively(Kania et al., 1990; Lecuit et al., 2002). In wild-type interphase 14 embryos proceeding through cellularization, slam transcripts were uniformly expressed and runt was expressed in seven stripes (Fig. 7B,F). slam expression dropped rapidly as cellularization was completed and gastrulation was initiated, while the seven-stripe pattern of runt expression persisted (data not shown). In smg-USB embryos undergoing graded cellularization, slam was first expressed in an anterior cap and runt was expressed in a single anterior stripe (Fig. 7C,G). In later embryos, slam was expressed at peak levels at the posterior and at lower levels at the anterior pole, and the full seven-stripe runt pattern was established(Fig. 7D,H). The Smaug gradient thus triggers a wave of slam and runt transcription that starts at the anterior pole and progresses to the posterior pole. To assay for maternal transcript destruction, we used FISH to monitor maternal cyclin B mRNA levels. This transcript is normally degraded over most of the embryo, but is retained in the pole cells(Fig. 7I-J). In smg-USB embryos, by contrast, cyclin B mRNA was degraded in an anterior-to-posterior gradient (Fig. 7K,L). Graded expression of Smaug thus appears to trigger an anterior-to-posterior gradient in the transition from maternal to zygotic control of development.

Smaug overexpression at the anterior pole did not advance cellularization,suggesting that an additional process or component defines the timing of cellularization under these conditions. To test this hypothesis, we assayed cellularization and maternal transcript destruction in embryos derived from females carrying extra copies of a wild-type smg transgene, which uniformly overexpress Smaug (see Fig. S4 in the supplementary material). These embryos cellularized at the same time as wild type, and FISH analyses indicated that destruction of maternal cyclin B mRNA was completed synchronously over most of the embryo during interphase 14 (data not shown). The timing and pattern of cellularization and maternal transcript destruction was also normal in embryos derived from wild-type females carrying the USB transgene (see Fig. S4 in the supplementary material). These embryos significantly overexpress Smaug at the anterior pole, and levels drop to near wild type at the posterior pole. At or below wild-type levels, Smaug thus determines the kinetics of maternal transcript destruction, zygotic transcription and cellularization. Above wild-type levels, by contrast, an additional process or factor appears to become limiting and determines the timing of the MBT.

Here, we have shown that Smaug is required to slow the cleavage divisions,activate the DNA replication checkpoint and initiate high level expression of the vast majority of genes that are transcribed during the final stages of the MZT. Furthermore, Smaug is essential for blastoderm cellularization, which marks the MBT. Significantly, Smaug protein begins to accumulate during the early cleavage divisions, when maternal transcript destruction is initiated,and Smaug levels peak during the syncytial blastoderm stage, when the replication checkpoint is activated and high-level zygotic transcription begins. In addition, expression of Smaug in an anterior-to-posterior gradient triggers an anterior-to-posterior gradient in the timing of transcript destruction, checkpoint activation, cellularization and gastrulation. Smaug is therefore the first maternal or zygotic factor shown to control the timing of the MZT.

Over-expression of Smaug does not accelerate maternal transcript destruction or cellularization, indicating that an additional factor or process becomes limiting under these conditions. Smaug triggers mRNA destruction by recruiting the CCR4/POP2/NOT complex, which catalyzes poly-A tail removal (Semotok et al.,2005), and an additional factor (`Y') has been proposed to act with Smaug to trigger decay (Tadros et al., 2007a). Factor Y or a component of the CCR4/POP2/NOT complex could become limiting for transcript destruction and the MBT when Smaug is overexpressed. Alternatively, Smaug expression at wild-type levels could saturate the binding sites on target transcripts and the deadenylation machinery could be present in excess. Under these conditions, wild-type levels of Smaug would lead to transcript destruction at maximal rates, which in turn would determine the minimum time for the MZT. Smaug also represses translation of specific targets (e.g. nanos), and this function could also have a role in coordinating the MZT.

Based on the present studies and earlier work, we favor the simple hypothesis that Smaug-dependent maternal transcript destruction triggers the MZT by coordinately downregulating a suite of maternal proteins that suppress zygotic transcription and the replication checkpoint. Consistent with this speculation, Smaug is required for maternal cyclin B transcript destruction, and overexpression of Cyclin B suppresses checkpoint activation and leads to additional rapid syncytial divisions (Crest et al., 2006). Similarly, Smaug triggers destruction of maternal mRNA encoding Tramtrack,which represses transcription of a subset of genes until near the end of the MZT (Pritchard and Schubiger,1996; Tadros et al.,2007a).

An initial wave of Smaug-dependent zygotic gene expression appears to trigger a series of positive and negative feedback loops that drive completion of the MZT (Fig. 8). For example, Smaug is required for zygotic transcription of fruhstart(frs), which promotes destruction of maternal string mRNA(Grosshans et al., 2003). String activates Cyclin-B-Cdk1, and string mRNA destruction in response to Frs expression could cooperate with Smaug-dependent cyclinB mRNA destruction to terminate the rapid cleavage stage divisions (Donzelli and Draetta,2003). Smaug is also required for zygotic expression of transcriptional activators, including cyclin T (see Table S2 in the supplementary material), cdk7 and cdk9 (C.H.H., J.T.W. and H.D.L., unpublished), and CyclinT-Cdk9 complex catalyzes phosphorylation of serine 2 on the RNAPII CTD repeat, which is linked to Smaug-dependent transcription at the MZT (Fig. 3) (Bellier et al.,1997; Palancade et al.,2001; Shim et al.,2002). Transcription of cyclin T and cdk9concomitant with destruction of maternal RNAs encoding transcriptional repressors, could therefore accelerate activation of the zygotic gene expression. Finally, Smaug is required for zygotic expression of the miR-309 cluster, which promotes a second, more rapid phase of maternal transcript destruction (Bushati et al., 2008) that terminates maternal genetic control of embryogenesis (Fig. 8). A recent study has shown that a maternally deposited transcription factor,Zelda, is also required for expression of the miR-309 cluster and number of other early zygotic genes (Liang et al., 2008). There is no evidence that Zelda regulates the timing of the MZT, but Zelda and Smaug could function cooperatively to promote transcription and maternal transcript destruction during this transition.

S-phase length increases during the later cleavage stage divisions,suggesting that maternally deposited replication factors are titrated as the N/C ratio increases. The DNA replication checkpoint is activated as S-phase slows, and thus senses the increasing N/C ratio(Sibon et al., 1999; Sibon et al., 1997). These observations, along with the present studies, suggest a simple model in which Smaug accumulation drives a maternal clock that determines timing of the MZT and cellularization, while the replication checkpoint monitors the N/C ratio and thus controls the number of cleavage divisions that can be completed before the clock runs down. In this model, haploid embryos contain less DNA,and therefore activate the checkpoint later and proceed through an additional division before cellularization. Embryos containing additional DNA, by contrast, trigger the checkpoint earlier and complete fewer divisions within the window provided by the clock. This model thus explains the link between the MBT and the N/C ratio, and control of the timing of this transition by a cleavage-independent clock.