twin, a CCR4 homolog, regulates cyclin poly(A) tail length to permit Drosophila oogenesis

Cell cycle regulation is central to the morphogenesis of organs and tissues and to terminal differentiation of cells. In eukaryotes, cell cycle transitions are driven by the successive activation and inactivation of cyclin-dependent kinases (cdks). In Drosophila, Cyclin E associates with Cdk2 to promote S phase. CycA, CycB and CycB3 sequentially associate with Cdk1 to promote initiation and progression of M phase(Lee and Orr-Weaver., 2003). Just as in yeast and mammals, Drosophila cyclins are regulated at the protein level via proteasome-mediated degradation(Doronkin et al., 2003; Echard and O'Farrell, 2003; Ohlmeyer and Schupbach, 2003; Sigrist et al., 1995). At the RNA level, Drosophila and other eukaryotes use E2F and other factors to regulate cyclin transcription (Duronio et al., 1996). Vertebrates also regulate cyclins A1, A2 and B1 via polyadenylation of mRNA, which enhances the stability and translational efficiency of RNA (Fuchimoto et al.,2001; Groisman et al.,2002).

During Drosophila oogenesis, the cell cycle is tightly regulated to coordinate egg chamber production. At the beginning of oogenesis, in the germarium, cystoblasts undergo four rounds of synchronous, incomplete cell divisions to generate cysts of 16 germ cells interconnected by ring canals. Two cells, the pro-oocytes, share ring canal connections with four neighbors. One of these cells, the oocyte, arrests in Prophase I of meiosis and condenses its chromosomes into a spherical karyosome. The other cells all become polyploid nurse cells and produce mRNA and proteins, such as Orb, that are trafficked to the oocyte (Spradling,1993). Elevation of CycA and of CycE can induce the cyst to undergo a fifth division (Doronkin et al.,2003; Lilly et al.,2000; Ohlmeyer and Schupbach,2003). CycE/Cdk2 kinase activity also directs germ cell fate. Increasing CycE/Cdk2 kinase activity by removing the Cdk-inhibitor dacapo directs oocytes to take on nurse cell characteristics. A mutation that reduces CycE/Cdk2 kinase activity levels results in more cells with oocyte character (Hong et al.,2003; Lilly and Spradling,1996). Study of bag of marbles (bam) further illustrates the close relationship between cell cycle control and differentiation in the female germ line. Consistent with its role in germ cell differentiation, loss of bam results in uncontrolled stem cell proliferation without differentiation(McKearin and Ohlstein, 1995),and ectopic bam causes stem cell loss(Ohlstein and McKearin, 1997). A role for Bam in cell cycle control is revealed by the finding that reduction of bam gene copy number suppresses the extra cyst division induced by excess cyclin activity (Lilly et al.,2000).

The fusome, a vesicular organelle that contains cytoskeletal proteins such as Hu Li Tai Shao (Hts), coordinates cell cycle and cell fate regulation in the cysts. With each cyst division, the fusome branches and extends through the ring canals connecting new daughter cells to the cyst(de Cuevas et al., 1996; Lin et al., 1994). One of the daughters always inherits more fusome material than the others and this cell has been posited to differentiate as the oocyte(de Cuevas and Spradling, 1998; Deng and Lin, 2001; Lin and Spradling, 1995; McKearin, 1997). During the cyst divisions, the fusome associates with one pole of each mitotic spindle to regulate the division plane and the number and synchrony of divisions(de Cuevas et al., 1997; McKearin, 1997). Loss of Hts eliminates the fusome and disrupts synchrony and number of the cyst divisions as well as oocyte specification (Lin et al., 1994; Yue and Spradling,1992; Zaccai and Lipshitz,1996). CycA colocalizes with the fusome during the M-phase of cystoblast divisions (Lilly et al.,2000) and may help the fusome to coordinate the number and timing of divisions. Components of the proteasome are also localized to the fusome,where they may play a role in CycE turnover(Doronkin et al., 2003; Ohlmeyer and Schupbach,2003).

In this study, we describe the molecular identification and phenotypic characterization of twin, a gene required for synchronous cyst divisions, oocyte fate specification and cyst maturation. We show that twin encodes the Drosophila homolog of the yeast ccr4 gene. ccr4 (carbon-catabolite-repression) was first identified in S. cerevisiae as a regulator of RNA levels of the alcohol-dehydrogenase II gene (Denis,1984). Although CCR4 protein was previously shown to associate with basal transcription machinery (Liu et al., 1998), recent data demonstrate that CCR4 catalyzes the degradation of poly(A) tails in yeast and flies(Chen et al., 2002; Tucker et al., 2002; Temme et al., 2004). We show that in twin germaria, CycA is misexpressed and cycA mRNA has a longer poly(A) tail. Furthermore, decreasing the dose of cycAsuppresses twin egg chamber degradation. We also find that twin germaria accumulate less cytoplasmic Bam protein. Genetically increasing the dose of bam suppresses twin cyst division and oocyte fate specification defects. We conclude that Twin/Ccr4 is required for the synchrony and number of cyst divisions and the oocyte/nurse cell fate decision. We propose that Twin/Ccr4 exerts control via regulation of poly(A)tail length of mitotic cyclins and by indirectly affecting the expression of others factors required for germline differentiation, such as Bam.

Ovaries were dissected, fixed and stained as previously described(Morris et al., 2003) with the following modifications: all fixations for BamC and anti-cyclin antisera in 4%formaldehyde. No serum was used with the primary antibody in these samples:Anti-Cyclin A, 1:600; anti-Cyclin B, 1:2000; anti-ORB, 1:1:1 of block, 6H4,4H8 antibodies; anti-Bam, 1:500; 1B1, 1:20. Phalloidin (1 μl) was added to primary antibody incubation for ring canal labeling. Ovary RNA in situ hybridizations performed as described previously(Ephrussi et al., 1991). Embryo RNA in situ hybridizations was performed as described by Lehmann and Tautz (Lehmann and Tautz,1994).

Counts are from stages 3-10. We did not count chambers within the germaria because, especially in twin germaria, it is difficult to determine number and fate of cells in immature chambers. All crosses were kept at 25°C except for the initial quantitation of the twin phenotypes(Fig. 1) which were at room temperature to reveal the full range of phenotypes.

The twin phenotypes are age and temperature sensitive. Whenever comparing flies of different genotypes, we used sisters of the same cross and bottle, of similar age, dissected on the same day. The only exceptions were when we could not use sisters of the same cross (i.e. Oregon R BAMC staining and comparisons of cycA twin^ry3/Df to twin^ry3/Df). In these cases, we dissected and stained flies grown in identical conditions on the same day.

Genomic DNA was digested with BamHI and BglII. The DNA was then ligated and PCR was performed with the Expand Hi Fidelity kit using primers extending outward from the P element ends. Sequencing ABI sequencing of the fragments and analysis with Editview, DNAstar (Lasergene) and BLAST(Altschul et al., 1997)followed.

RNA was isolated from twin^ry3 homozygous and wild-type ovaries using Trizol. The RNA was then poly(A) purified using the PolyATract mRNA Isolation System III (Promega), and was run out on an agarose gel,blotted and probed using standard methods.

We mutagenized ru h th st sr e ca males with EMS. Mutants were tested for failure to complement the ry5 allele for female sterility. ry5 is viable in trans to deficiency and fertile in males, so null alleles can be isolated in such a screen and recovered via male progeny. ESTs were ordered from the Berkeley Drosophila Genome Project and were ABI-sequenced from ends. New primers were generated from each round of sequencing to generate the next round.

As described by Salles and Strickland(Salles and Strickland, 1999). RNA was prepared from ovary extracts of young (1- and 2-day-old) adults using TRIZOL. Very young twin and wild-type adults had comparably little development of mid- and late-stage egg chambers. Total RNA (100 ng) used for first-strand synthesis. PCR reactions performed with expand Hi-Fidelity kit. ³²P dCTP was added to 10 mM cold dNTP stock before setting up PCR reactions, and a fraction of the PCR product was tested on a gel to permit approximately equal loading of hot PCR product for the experiment. Samples were run on 6% acrylamide gels, dried down and exposed as autoradiographs.

twin mutant females are nearly sterile, and their ovaries show defects in cyst division, cyst survival and germ cell fate. We quantitated frequencies of egg chamber degradation, cyst division defects and oocyte specification defects in twin alleles in trans to deficiency,allowing us to place the alleles in a series: S1>ry3∼ry5>S3(Table 1). The stronger alleles in the series also caused more egg chamber death within the germarium (not shown). At room temperature, 27% of twin^S1/Df egg chambers degrade between stages 3 and 10(Fig. 1B, Table 1), 41% of egg chambers contain fewer than 16 germ cells, (Fig. 1C,D,G, Table 1),and 41% of egg chambers fail to specify an oocyte(Fig. 1D,G, Table 1). Twenty-nine percent of the egg chambers were defective in both germ cell number and oocyte specification (Fig. 1D, Table 1). We also observed chambers containing an oocyte at the center rather than the posterior(Fig. 1E,G), chambers with two oocytes (Fig. 1F) and chambers containing cells with more than four ring canals, indicating that these cells divided more than four times (Fig. 1G, fifth ring canal in a different plane). Each allele exhibited stronger mutant phenotypes at higher temperatures (not shown). twincysts contain any number of germ cells from one to more than 25. All alleles displayed the full range of phenotypes, indicating either that none of the alleles are nulls or that twin activity is not absolutely required for egg chamber development.

The cell division defects in twin egg chambers are very striking. In contrast to other mutants defective in counting the number of cyst divisions, such as orb, encore and tribbles (which maintain synchrony of division) (Deng and Lin,2001; Hawkins et al.,1996; Johnston,2000; Lantz et al.,1994; Mata et al.,2000; McKearin and Christerson, 1994; Seher and Leptin, 2000; Van Buskirk et al., 2000), the number of germ cells in twin egg chambers did not generally correspond to a power of two(Fig. 1C). Germ cell numbers in adjacent chambers did not generally add up to 8, 16 or 32 cells, arguing that encapsulation defects cannot explain this phenotype. Finally, every cell in each cyst was connected to at least one neighbor by a single ring canal,arguing against germ cell death as the cause of unusual germ cell number(Fig. 1G). We conclude that twin affects the synchrony and number of cyst divisions.

When egg chambers contained oocytes, these oocytes appeared normal and contained the most ring canals. We observed chambers containing three germ cells, including an oocyte (not shown) as well as rare chambers specifying two oocytes (Fig. 1F). In all cases, the oocytes condensed their chromosomes into a normal karyosome and accumulated oocyte specific factors such as Orb protein(Fig. 1C) and osk mRNA(not shown). We observed two oocytes in chambers containing too many, too few or exactly 16 germ cells. In these cases, both oocytes appeared normal, even to the extent that they attracted migrating follicle cells in stage 9 egg chambers. When there were two oocytes they were often both at the posterior of the chamber (Fig. 1F), although they were also found at opposite ends of the chamber (not shown). Because cyst division and oocyte specification defects occur independently in twinmutants, Twin may couple these processes in the wild type to ensure exactly one oocyte forms in each 16 cell egg chamber.

In spite of the striking defects observed in twin ovaries, most other tissues in twin animals are apparently unaffected. twin/Df flies are viable and males are fertile. Within the ovary, the defect is restricted to the germline. Follicle cells form normal epithelia and migrate properly. In germline clones, twin mutant germ cells exhibit mutant phenotypes in an otherwise heterozygous background (data not shown),indicating that the functioning of twin is required in the germline.

To determine how twin affects germline development, we molecularly identified the gene using two P-element insertion alleles, twin^ry3 and twin^ry5(Spradling, 1993). We reverted the mutant phenotype by precise excision of the P-elements, confirming that the insertions caused the twin defects. twin maps to chromosomal region 95F and the twin alleles failed to complement Df(3R)CRB87-4. Inverse PCR and sequencing of the distal and proximal genomic DNA flanking the P-element insertions showed that the twin^ry3 and twin^ry5 transposons were only offset by 1 bp from each other and in opposite orientations. Using BLAST analysis we found ESTs that matched the distal flanking region(Fig. 2A,B). These ESTs included several splice isoforms of a Drosophila ccr4 homolog that differ in their 5′UTRs but not in their coding regions(Fig. 2B). Sequence analysis showed that the P-insertions as well as the P-element proximal DNA fell in a 15 kb intron of the ccr4 gene.

We hybridized labeled distal and proximal flanking DNA to northern blots containing mRNA from ovaries isolated from OR flies and twin^ry3 homozygous flies(Fig. 2C). A 3 kb doublet detected by the distal probe and by ccr4 ESTs was strongly reduced in twin^ry3 homozygous ovaries. Overexposing the blot showed very weak expression. A 1.5 kb band detected by the proximal probe lies within the 15 kb intron of ccr4. The genome project and Genefinder do not predict this transcript. The transcript was not altered in size or reduced in levels in the twin^ry3 ovaries(Fig. 2C).

We next wanted to see if ccr4 is expressed in a pattern consistent with the role of twin in oogenesis. Consistent with FlyBase reports we find that ccr4 RNA is expressed throughout embryogenesis. Newly deposited embryos show very strong maternal deposition of ccr4 RNA(Fig. 2D). The maternal transcript is rapidly degraded, but the germline and soma exhibit low-level uniform expression in the syncytial blastoderm(Fig. 2E) and throughout embryogenesis (Fig. 2F,G). ccr4 is expressed throughout oogenesis, most strongly in the germline but also in the soma (Fig. 2I,J,K) (see Temme et al.,2004).

S. cerevisiae CCR4 has three functional domains(Fig. 3A)(Malvar et al., 1992). The N-terminal region is rich in glutamine and asparagine. The central region contains five leucine-rich repeats (LRR), which are crucial for dimerization with CAF1/POP1, an adapter protein that facilitates the association of CCR4 with other proteins and is required for deadenylation of targets(Clark et al., 2004). The Twin LRR domain contains three and a half repeats(Fig. 3A,B) which are most similar to the repeats shown to be essential to yeast CCR4 function(Clark et al., 2004). The C-terminal domain of S. cerevisiae CCR4 has 3′-5′exoribonuclease and ssDNA exonuclease activity, and it catalyses the poly(A)-deadenylation reaction (Tucker et al., 2002; Tucker et al.,2001; Chen et al.,2002; Draper et al.,1994; Malvar et al.,1992). The catalytic domain of Twin shares extensive homology with S. cerevisiae CCR4. Blast analysis of the Drosophila genome showed two other predicted ccr4 homologs in Drosophila,angel and Dnocturnin (Fig. 3A). Angel (Kurzik-Dumke and Zengerle, 1996) and Dnocturnin share extensive homology with the CCR4 catalytic domain, although Dnocturnin is missing an asparagine shown to be essential for catalysis (Chen et al.,2002). Neither protein possesses an LRR repeat region(Fig. 3A); Twin is therefore the best candidate to encode the Drosophila CCR4 homolog. Consistent with a prominent role of Twin in deadenylation, Temme et al.(Temme et al., 2004) showed that Drosophila CCR4 catalyzes deadenylation when purified from extracts. Partial depletion of CCR4 either from tissue culture or flies led to an overall increase in bulk poly(A)tail length and more specifically affected the poly(A)tail length and stability of heat-shock, hsp 70, RNA.

In order to conclusively determine that the twin phenotypes are due to disruption of Drosophila ccr4, we carried out an EMS mutagenesis screen for mutants that failed to complement the ry5allele. We screened 2749 independent mutants and isolated two lines, twin^S1 and twin^S3, that failed to complement twin^ry5, twin^ry3, Df(3R)CRB87-4 and each other. Sequencing of ccr4 showed that the S1 mutation results in the substitution of an isoleucine for a highly conserved asparagine in the last LRR repeat (Fig. 3B). This mutation could therefore prevent Twin from deadenylating mRNAs. The S3 mutation, the weakest twin allele, changes an isoleucine to a serine within a conserved motif of the deadenylase catalytic domain (Fig. 3C). This amino acid is conserved but has not been shown to be essential for catalysis in yeast. Because all twin alleles affect the ccr4 transcript,we conclude that twin encodes Drosophila ccr4.

twin mutants exhibit defects in the divisions of the cystoblast. Such defects have been observed in mutants directly affecting cell cycle regulators (Mata et al., 2000)and in mutants affecting the integrity of the fusome, which co-localizes with CycA in a cell cycle-dependent manner (de Cuevas et al., 1996; Lilly et al., 2000; Lin et al.,1994). Since CCR4 in S. cerevisiae regulates gene expression, we examined twin mutants for changes in cyclin regulation in the germarium (Fig. 4A-F). Staining of wild-type or twin^S1/+ germaria showed that, as reported previously, CycA is expressed in the dividing cysts of germarium region 1 and is downregulated in the post-mitotic cysts from region 2 on(Fig. 4A,C). In twin^S1/Df ovaries, CycA was ectopically expressed throughout the germarium, including what should be the postmitotic regions(Fig. 4B). Similar, but less strong misexpression was also observed in twin^ry3/Df(Fig. 4D). Somatic expression appeared unaffected. We also analyzed the expression of CycB, which acts together with CycA in promoting entry into mitosis(Knoblich and Lehner, 1993; Knoblich et al., 1994). CycB is misexpressed in a pattern similar to that of CycA in twin germaria(Fig. 4E,F), although in contrast to the very consistent CycA misexpression phenotype, not all germaria misexpressed CycB. CycB misexpression may be an indirect result of reduced Twin activity or CycB may have other important regulators in addition to Twin. We stained ovaries with antibodies raised against the S-phase cyclin CycE, and did not detect a difference between twin^S1/Df and twin/+ germaria (not shown). Our findings suggest that Twin is required for downregulating CycA expression in the germarium and, to a lesser extent, expression of CycB.

Since twin mutants exhibit cyst division defects similar to those observed in hts mutants, we used 1B1 antiserum to follow fusome morphology during cyst division. As in wild type, 1B1 labeled the fusome and cell cortices in twin mutants(Fig. 4A-F). The fusome seemed normal in size and extended throughout the cysts(Fig. 4B,D). We obtained the same results with anti-Spectrin antiserum (not shown). In addition, we observed CycA and CycB co-localizing with the fusome in a subset of cysts throughout the germarium (Fig. 4B,F), which suggests that the fusome retains some normal function and that cysts may be delayed in progressing through mitosis in twin. In many twin chambers, the fusome had fewer branches(Fig. 4B,F), which we attribute to the fact that twin mutant cysts frequently contain fewer cells. Taken together, these data suggest that the twin defects in cyst division synchrony and oocyte specification occur in parallel or downstream of fusome integrity, possibly by acting with the fusome to regulate cyclin levels within the cyst.

Our findings show that CycA levels are elevated in twin. We analyzed whether reduction in the levels of cyclins affected the twinphenotypes. For these experiments we used the twin^ry3/Dfallelic combination, whose intermediate phenotype seemed suitable for modification studies. Removing one copy of cycA markedly suppressed the number of stage 3-10 chambers that were degrading(Table 2, 51% compared with 70%, a relative suppression of 30%, P<0.001), suggesting that excess CycA induces chamber degradation in twin (see Discussion, Table 2). We did not observe a significant change in the twin^ry3/Df phenotype when we removed one copy of cycE or cycB(Table 2). It remains unclear whether a specific stage of the cell cycle is particularly sensitive to loss of Twin function. Staining germaria with anti-phospho-Histone 3 antiserum did not reveal an enrichment of cysts with the condensed chromosomes characteristic of early to mid-M phase (not shown), but twin cysts could be delayed elsewhere in the cell cycle. On the basis of the CycA expression and genetic interaction experiments, we conclude that misexpression of CYCA contributes to the twin egg chamber phenotype.

We assayed expression of Bam in twin germaria because Bam is an important determinant of cystoblast division and differentiation. In bam mutant ovaries, stem cells divide symmetrically so that the germaria are filled with germline tumors rather than egg chambers(McKearin and Ohlstein, 1995). Precocious expression of Bam in larvae induces all stem cells to differentiate and form cysts (Gilboa and Lehmann,2004; Kai and Spradling,2004; Ohlstein and McKearin,1997). Removing Bam during cyst division can revert differentiating cysts into single stem cell-like cells(Kai and Spradling, 2004). Removing one copy of bam suppresses extra cell divisions caused by a pulse of ectopic CycA or by loss of function of encore, a gene that has been proposed to promote cyclin protein degradation(Hawkins et al., 1996; Lilly et al., 2000; Ohlmeyer and Schupbach,2003).

We stained twin and wild-type ovaries with BamC antiserum(Fig. 5). In wild-type ovaries,BamC stains the cystoblasts and the dividing cysts but not later stages(McKearin and Ohlstein, 1995)(Fig. 5A,A′). The BamC staining in the dividing cysts of twin^S1/Df ovaries was much fainter (Fig. 5B,B′). Most germaria showed quite weak or undetectable levels of BamC, although some germaria exhibited strong BamC expression. We conclude that Bam is downregulated or not localized properly to the cytoplasm(and thus not detectable by the BamC antiserum) in twin mutants.

In order to test the functional significance of BamC reduction in twin ovaries, we induced expression of bam and assayed for suppression of the twin phenotype. To avoid alteration of the twin phenotype caused by heat shock, we compared twin^ry3/Dfwith hs-bam; twin^ry3/Df ovaries grown at 25°C to induce some bam expression from the hs-promoter. We found that increasing Bam levels suppresses twin(Table 3). Fifty percent more chambers contained an oocyte in hs-bam; twin^ry3/Df ovaries than in twin^ry3/Df ovaries (61% compared with 40%, P<0.001, Table 3). Comparing the percentage of chambers with no mutant phenotypes to those with oocyte specification or cyst cell number defects showed an even more striking contrast. hs-bam; twin^ry3/Df egg chambers were nearly twice as likely to contain 15 nurse cells and one oocyte than twin^ry3/Df ovaries egg chambers (43% of non-degrading chambers versus 26%, not shown). Taken together, these results suggest that the cell cycle and oocyte specification defects in twin are due, in part, to reduced Bam expression.

twin encodes the Drosophila CCR4 homolog. In S. cerevisiae, CCR4 is found associated with a large, multi-protein complex and affects RNA levels (Chen et al.,2001; Denis and Malvar,1990; Verdone et al.,1997). Although CCR4 has been proposed to regulate transcription,recent biochemical experiments strongly support a direct role of CCR4 as the major cytoplasmic deadenylase (Chen et al.,2002; Tucker et al.,2002; Tucker et al.,2001; Temme et al.,2004). In this role, CCR4 regulates mRNA target genes by shortening their poly(A) tails. Short poly(A) tails destabilize RNA and decrease the translational efficiency of mRNAs, leading to lower protein levels. We performed poly(A) tail length (PAT) assays(Salles and Strickland, 1999)to compare the poly(A) tail length of cycA, cycB, cycE and bam in total ovary extracts from dissected ovaries of 1- to 2-day-old twin^S1/Df and twin/+ adult females These qualitative assays use PCR to amplify the whole range of poly(A) tail lengths associated with a particular transcript present in an extract. cycApoly(A) tails were significantly longer in twin^S1/Dfovaries than in twin/+ ovaries(Fig. 6A). Mutant extracts were highly enriched with cycA message with poly(A) tails of 40-80 As,while cycA poly(A) tails in wild-type extracts predominantly contained 0-20 As. We also observed an upward shift in poly(A) tail length when monitoring cycB and cycE poly(A) tails in twinmutant extracts, although the shift for cycB was not as dramatic as the cycA shift (Fig. 6B-C). We did not detect an obvious increase in bampoly(A) tail length in twin extracts(Fig. 6D). The fact that we do not observe elevated CycE protein levels in spite of the shift in poly(A) tail length in twin ovaries suggests that other mechanisms of regulating CycE levels (e.g. proteolysis) are sufficient to compensate for longer cycE poly(A) tail length in twin mutants.

Taken together, our data strongly suggest that wild-type Twin/Ccr4 deadenylates a number of downstream target mRNAs, including several cyclins. Because CycA levels are elevated in twin mutant ovaries and reducing CycA dose suppresses the twin egg chamber degradation phenotype, our data support the model that Twin/Ccr4 deadenylation regulates CycA levels,which in turn affects egg chamber growth and maturation.

In S. cerevisiae and more recently in Drosophila, CCR4 was shown to possess deadenylase activity(Chen et al., 2002; Tucker et al., 2002; Tucker et al., 2001; Temme et al., 2004). It remained unclear whether mutations in CCR4 had specific developmental defects and whether these defects might reveal specific targets sensitive to CCR4 function. Here, we show that twin encodes the Drosophilahomolog of CCR4 and that mutations in twin cause defects during early stages of oogenesis. twin mutant cysts divide asynchronously and less than four times, oocyte specification is defective and many egg chambers die and degrade. We show that the mitotic cyclins, CycA and CycB, are misexpressed in twin, and reducing the gene copy number of cycA partially suppresses the twin egg chamber degradation phenotype. Furthermore,we show that the poly(A) tails of cycA, cycE and, to a lesser extent, cycB are longer in twin extracts, suggesting that Twin/Ccr4 deadenylase activity directly controls the RNA levels of these cell cycle regulators. By contrast, BamC staining is reduced in twin. Induction of extra bam expression suppresses the cyst division and oocyte fate specification defects in twin mutants, implicating low Bam levels as one of the causes of these twin phenotypes.

Our twin alleles are viable and specifically affect the female germline. In S. cerevesiae, ccr4 mutations are not lethal, although CCR4 is thought to be the main cytoplasmic deadenylase(Denis, 1984). It is possible that angel and Dnocturnin (CG4796), two other genes with extensive homology to the ccr4 catalytic domain but lacking the crucial LRR repeats, can partially compensate for loss of Twin function. Alternatively, as our mutations are probably not complete nulls, oogenesis may be more sensitive than the soma to decreased Twin function. Like the ovary,the early embryo relies on precise post-transcriptional gene regulation(Tadros et al., 2003). The mature egg contains high levels of maternally loaded twin, consistent with a role for Twin in deadenylation, and probably explaining why twin mutants carry out embryogenesis normally.

Mitotic cells regulate cyclin levels in order to progress through the cell cycle. At the protein level, Drosophila regulates CycA, CycB and CycE, via proteasome-mediated degradation(Doronkin et al., 2003; Echard and O'Farrell, 2003; Sigrist et al., 1995). In the Drosophila ovary, the novel protein Encore has been proposed to localize components of the proteasome complex to the fusome to regulate CycE(Doronkin et al., 2003; Ohlmeyer and Schupbach, 2003). encore mutant cysts undergo an extra cell division and contain 32 cells (Hawkins et al., 1996),probably as a consequence of misexpressing not only CycE, but also CycA(Ohlmeyer and Schupbach,2003). Other experiments have shown that cyst divisions are sensitive to CycA levels. Adding a brief pulse of CycA by inducing a heat-shock construct can lead to an extra round of cyst division, suggesting that downregulation of CycA is crucial for cell cycle progression(Lilly et al., 2000). Only a small number of cysts respond to such a CycA pulse, suggesting that in the wild type not all germ cells are in a susceptible phase of the cell cycle (G2)during which they can respond to CycA.

Cyclin RNA levels are regulated by control of poly(A) tail length. In Xenopus and mouse oocytes, cycB RNA is not translated in the absence of CPEB-mediated poly(A) tail lengthening. Longer poly(A) tails also enhance cyclin translation in Drosophila embryos. In the Drosophila ovary, Orb, the CPEB homolog, regulates poly(A) tail length and expression of its own RNA and oskar RNA(Castagnetti and Ephrussi,2003; Tan et al.,2001). Consistent with a role for Orb in cyclin regulation and cyst division, orb mutant cysts frequently contain eight germ cells(Lantz et al., 1994).

Our data suggest that Twin-mediated deadenylation of cyclin RNA regulates cyst divisions. Cyclin polyadenylation has been well studied, but much less is known about cyclin RNA deadenylation. In Drosophila, Nanos and Pumilio have been shown to control deadenylation of cycB mRNA in primordial germ cells (Asaoka-Taguchi et al., 1999). Furthermore, Xenopus Pumilio interacts with CPEB (Richter and Theurkauf,2001), and Nanos, Pumilio and Orb/CPEB are all expressed early in Drosophila oogenesis (Forbes and Lehmann, 1998). It is intriguing to speculate that Twin may regulate the poly(A) tail lengths in the dividing cyst in conjunction with Nanos, Pumilio and/or Orb.

BamC expression is reduced in twin germaria; a phenotype we would not predict if Twin directly regulated Bam expression via deadenylation. Indeed, we did not observe a substantial change in bam poly(A) tail length in twin ovaries. We therefore propose that bam is an indirect target of Twin/Ccr4.

Although bam is known to control the differentiation of the cystoblast and to promote cyst division, the biochemical role of Bam is unknown. Removing one copy of bam suppresses the extra division in cysts lacking encore or overexpressing CycA. Our results further implicate Bam in the events of early oogenesis. Increased bamexpression suppresses not only the cyst division defects observed in twin mutants, but also the twin oocyte specification defects. Because Twin regulates cycA directly and may regulate Bam indirectly, the simplest model would posit that high levels of CycA are sufficient to suppress Bam expression. Two pieces of evidence argue against this model: Bam and CycA are both present at high levels in the dividing cyst;and Bam is required for the fifth cyst division induced by high levels of CycA. In addition, hs-bam induces stem cells to develop into normal cysts, indicating that high Bam levels do not disrupt CycA expression. We favor a model by which Bam and CycA act in parallel to each other, downstream of Twin.

Although several models could explain our data, we propose that increased mitotic cyclin levels together with low Bam expression cause many of the twin phenotypes. If Bam expression were normal, overexpressing cyclins could lead to extra cyst divisions. The low level of Bam in twin germaria does not permit continued cell division, yet cyclin levels remain high, delaying cell cycle progression and probably causing the egg chamber degradation we observe in twin. This model is consistent with the fact that reducing the copy number of bam suppresses the extra cyst division phenotype of encore and of hs-cycA(Hawkins et al., 1996; Lilly et al., 2000). Corroborating evidence comes from the observation that reducing the gene dose of cycA or increasing the dose of bam can partially suppress the degradation phenotype. However, there are likely to be other, unidentified targets of twin that also contribute to the twinphenotype.

twin and hts mutants disrupt the number and synchrony of cyst divisions and oocyte specification. This array of defects is not shared by the cell cycle mutants discussed above or by other mutants such as orb (Lantz et al.,1994), the M-phase inhibitor tribbles or the M-phase activator string (Mata et al.,2000), which affect the number but not the synchrony of cyst divisions. Comparison of twin and hts may therefore be instructive. hts cysts have no fusome, and are thought consequently not to coordinate the cyst divisions (Lin et al., 1994; Yue and Spradling, 1992). By contrast, cysts in twin mutants contain branched fusomes that are capable of colocalizing with CycA,suggesting the possibility that Twin/Ccr4 gene regulation may mediate the coordination of the cyst divisions with oocyte specification downstream of the fusome.

We thank L. Gilboa, C. Navarro, R. Martinho, D. Siekhaus, A. Renault, P. Kunwar and Lehmann laboratory members past and present for helpful discussion. We thank C. Berg for advice and encouragement at the initial stages of the twin project. We thank C. F. Lehner and D. McKearin for kindly providing antibodies. We thank the Bloomington Drosophila Stock Center for providing us with the deficiency kit and many mutant stocks. R.L. is a Howard Hughes Medical Institute investigator. J.M. was supported in part by the American Cancer Society and by the Howard Hughes Medical Institute.