PAP- and GLD-2-type poly(A) polymerases are required sequentially in cytoplasmic polyadenylation and oogenesis in Drosophila

In many species, the oocyte and early embryo develop in the absence of transcription. Therefore, the first steps of development depend on maternal mRNAs and on their regulation at the level of translation, stability and localization. Regulation of mRNA poly(A) tail length is a common mechanism of translational control. Deadenylation or poly(A) tail shortening results in mRNA decay or translational repression. Conversely, poly(A) tail elongation by cytoplasmic polyadenylation results in translational activation(Richter, 2000; Wickens et al., 2000). How the poly(A) tail length of a particular mRNA and, consequently, its level of translation are determined has been a matter of investigation for many years. It is becoming clear that poly(A) tail length results from a balance between concomitant deadenylation and polyadenylation(Kim and Richter, 2006).

The molecular mechanisms of cytoplasmic polyadenylation have been investigated in Xenopus oocytes. The specific RNA-binding protein in the reaction is CPEB (Cytoplasmic polyadenylation element binding protein),which binds the CPE in the 3′-UTR of regulated mRNAs. Two other factors,CPSF (Cleavage and polyadenylation specificity factor) and Symplekin, are required in addition to a poly(A) polymerase(Barnard et al., 2004; Richter, 2007). Before meiotic maturation, the polyadenylation complex also contains PARN, a deadenylase whose activity counteracts poly(A) tail elongation(Kim and Richter, 2006). At meiotic maturation, CPEB phosphorylation results in the release of PARN from the complex, thus leading to polyadenylation and translational activation.

CPSF and Symplekin are also required for nuclear polyadenylation, a cotranscriptional reaction that leads to the synthesis of a poly(A) tail at the 3′ end of all mRNAs (Edmonds,2002). A canonical poly(A) polymerase (PAP) is responsible for poly(A) tail synthesis during nuclear polyadenylation. Particular isoforms of PAP were first thought to be required for cytoplasmic polyadenylation(Ballantyne et al., 1995). Moreover, TPAP (Papolb - Mouse Genome Informatics), a testis-specific PAP in mouse, is cytoplasmic in spermatogenic cells and has been shown, using a Tpap knockout, to be required for cytoplasmic polyadenylation of specific mRNAs and for spermiogenesis(Kashiwabara et al., 2002; Zhuang et al., 2004). More recently, a new family of atypical poly(A) polymerases, the GLD-2 family, has been characterized, with a first member identified in C. elegans(Wang et al., 2002). GLD-2-type proteins exist in all eukaryotes, where they have different functions (Buhler et al., 2007; Kwak and Wickens, 2007; Rissland et al., 2007).

In C. elegans, GLD-2 is required for entry into meiosis from the mitotic cycle in the gonad, and for meiosis I progression(Kadyk and Kimble, 1998). C. elegans GLD-2 has a poly(A) polymerase activity in vitro(Wang et al., 2002) and in vivo (Suh et al., 2006). In Xenopus oocytes, GLD-2 is found in the cytoplasmic polyadenylation complex, within which it directly interacts with CPEB and CPSF, and it has a poly(A) polymerase activity in vitro in the presence of the other factors of the complex (Barnard et al.,2004). GLD-2 is in complexes with mRNAs, such as cycB1and mos, that are regulated by cytoplasmic polyadenylation(Rouhana et al., 2005). It is thus very likely that GLD-2 plays a role in cytoplasmic polyadenylation during Xenopus meiotic maturation. However, although cytoplasmic polyadenylation of mos and cycB1 mRNAs is required for meiotic maturation (Sheets et al.,1995; Stebbins-Boaz et al.,1996), the functional role of Xenopus GLD-2 in meiotic maturation has not been addressed. Unexpectedly, although mouse GLD-2 (Papd4 -Mouse Genome Informatics) is found in oocytes at metaphases I and II, a recent study shows that oocyte maturation in GLD-2 knockout mice is not altered,demonstrating that if mouse GLD-2 acts as a poly(A) polymerase at this stage,another protein acts redundantly(Nakanishi et al., 2006; Nakanishi et al., 2007).

In Drosophila, poly(A) tail regulation by deadenylation and cytoplasmic polyadenylation is essential for controlling mRNAs involved in axis patterning and other aspects in early development(Benoit et al., 2005; Castagnetti and Ephrussi,2003; Kadyrova et al.,2007; Morris et al.,2005; Semotok et al.,2005; Vardy and Orr-Weaver,2007; Zaessinger et al.,2006). In ovaries, cytoplasmic polyadenylation regulates the translation of oskar (osk), the posterior determinant, and of CycB mRNAs, and this polyadenylation depends on Orb, the Drosophila homolog of CPEB (Benoit et al., 2005; Castagnetti and Ephrussi, 2003; Chang et al.,1999; Juge et al.,2002). Orb is required at the earliest steps of oogenesis for the regulation of the synchronous divisions of a cystoblast that lead to the production of sixteen germ cells per cyst, and for the restriction of meiosis to one oocyte (Huynh and St Johnston,2000). A single gene, hiiragi (hrg), which encodes one isoform of canonical PAP, exists in the Drosophila genome(Juge et al., 2002; Murata et al., 2001). Genetic interactions have implicated orb and hrg in the cytoplasmic polyadenylation of osk mRNA and accumulation of Osk protein at the posterior pole of the oocyte during mid-oogenesis. This led to the conclusion that canonical PAP has a role in cytoplasmic polyadenylation at this stage(Juge et al., 2002).

Cytoplasmic poly(A) tail elongation is also crucial in early embryos to activate the translation of mRNAs, including that of bicoid(bcd), which encodes the anterior morphogen(Salles et al., 1994). Polyadenylation and translation occur upon egg activation, a process that also induces the resumption of meiosis from the metaphase I arrest in mature oocytes, and which is triggered by egg laying, the passage of the egg through the oviduct (Heifetz et al.,2001). A link has been established between cytoplasmic polyadenylation and meiotic progression at egg activation because mutants defective for meiotic progression are also defective for poly(A) tail elongation (Horner et al.,2006; Lieberfarb et al.,1996; Page and Orr-Weaver,1996).

Here, we analyze the function of Drosophila GLD-2 in the female germline. We show that this protein is encoded by wispy(wisp), a gene previously identified genetically(Brent et al., 2000), and we therefore refer to this protein as Wisp. We find that Wisp has a poly(A)polymerase activity in vitro and in vivo, and that it is required for poly(A)tail elongation of maternal mRNAs during late oogenesis and early embryogenesis. Wisp is required for meiotic progression in mature oocytes. A key target of Wisp during this process is cortex (cort)mRNA, which encodes a meiosis-specific activator of the anaphase-promoting complex (APC). This demonstrates the role of polyadenylation and translational activation in meiotic progression. In addition, we investigate the respective roles of conventional PAP and of Wisp in oogenesis and show that PAP and Orb are involved earlier than Wisp and Orb. Our results establish the requirement of two poly(A) polymerases for cytoplasmic polyadenylation at different steps of oogenesis.

The w¹¹¹⁸ stock was used as control. The wisp^KG5287 (previously called CG15737^KG5287) mutant was from the Berkeley Drosophila Genome Project. The P-element (SUPor-P)was mobilized in the presence of P transposase and internal deletions of the P-element were recovered. Several female fertile revertants were also recovered among which two were sequenced and found to have retained a small portion of the P-element ends (33 bp and 39 bp,respectively). As the insertion point is downstream of the start codon and the remaining short insertions are coding and in frame with Wisp in both of the revertants, we believe these revertants produce slightly longer functional Wisp proteins. A 1.7 kb genomic region overlapping the conserved domains in wisp^12-3147, was PCR amplified and sequenced. Two point mutations, T1151I and I1292V, were found. The stock y¹fs(1)K10⁴ cv¹ v¹ f¹/FM0 that was generated in the same mutagenesis as wisp^12-3147(Mohler, 1977) was used as a source of parental chromosome, and the same region of wisp was sequenced in this stock. The conservative I1292V mutation was present in this control stock, whereas T1151I was not.

Analysis of poly(A) tail length by PCR (the PAT assay) and RT-PCRs were performed as reported previously (Benoit et al., 2005; Zaessinger et al.,2006). RNA preparations were from 20 embryos, 10 ovaries, and from dissected germarium-to-stage 8, stage 9-10, and stage 14 egg chambers from 10 ovaries. RT-PCRs were performed on the same RNA preparations used for the PAT assays, using serial dilutions of the cDNAs. Dilutions 1:10 are shown. Oligos for PAT assays were (5′ to 3′): osk,AAGCGCTTGTTTGTAGCACA; CycB, GCTGGCCGAACACATCGGCG; nos,TTTTGTTTACCATTGATCAATTTTTC; bcd, CATTTGCGCATTCTTTGACC; cort,GGCCAAGGACAAGTGCAGCTC; and sop, GGATTGCTACACCTCGGCCCGT. Oligos for RT-PCR were (5′ to 3′): osk, GCCATATTGCTGAGCCACGCCC and CCAGTAGCGTGGAGGTGCTCG; nos, CGATCCTTGAAAATCTTTGCGCAGGT and TCGTTGTTATTCTCACAAAAGACGCA; bcd, CTGGGTCGACCAATGTCAAT GGCG and GCTCTTGTCCAGACCCTTCAAAGG; and sop, CACCCCAATAAAGTTGATAGACCT and ATCTCGAACTCTTTGATGGGAAGC. Whole-mount RNA in situ hybridizations were performed by standard methods. The RNA antisense probes were made from pKSbcdwt (bcd), pN5 (nos) and osk cDNA clones.

To express MS2 fusion proteins in Xenopus oocytes, the C-terminal half of Wisp (residues 702-1373) and full-length Homo sapiens GLD-2(HsGLD-2) were cloned into the pCSMS2 vector using NheI and XhoI (Rouhana et al.,2005). Wisp D1031A and HsGLD-2 D215A mutations were created by site-directed mutagenesis. For in vitro transcription, Wisp and HsGLD-2 clones were linearized with XbaI and NotI,respectively, and transcribed using the SP6 Megascript Kit (Ambion). Xenopus oocyte manipulation, injection, luciferase assays and labeled RNA analysis were performed as described(Dickson et al., 1999; Kwak et al., 2004). To test MS2 fusion protein expression, oocytes were harvested 6 hours after mRNA injection and analyzed by western blotting using α-HA₁₁antibody (1:2000; Covance). Two oocytes were loaded per lane.

GST recombinant proteins were produced by cloning the C-terminal half of Wisp (residues 702-1373) into the pBAH vector using EcoRI and XhoI, and the N-terminal region of Wisp (residues 11-547) digested with BglII into pGEX-5X-2 digested with BamHI. In vitro interactions were performed as described(Benoit et al., 2002), in the presence of 0.2 μg/μl RNase A. Orb was in vitro translated from the orb D5 cDNA cloned into pBluescript using EcoRI and HindIII.

Antibodies against Wisp were obtained by cloning a portion of wispcDNA LD18468 encoding residues 702 to 1373 into the pMAL vector. The Wisp-MBP fusion protein was expressed in Escherichia coli and purified using amylose beads (NEB). The purified fusion protein was injected into guinea pigs. Western blots and immunostaining were performed as described(Benoit et al., 2005; Benoit et al., 1999). Antibody dilutions for western blots were: anti-Wisp 1:3000, rabbit anti-BicC(Saffman et al., 1998) 1:1000,rat anti-PAP (Juge et al.,2002) 1:500, anti-Orb 6H4 (Developmental Studies Hybridoma Bank)1:20, rabbit anti-Cyclin A (Whitfield et al., 1990) 1:10,000, anti-Cort(Pesin and Orr-Weaver, 2007)1:2000, anti-α-tubulin (Sigma T5168) 1:10,000. Dilutions for immunostaining were: anti-Wisp 1:2000, anti-Osk(Kim-Ha et al., 1995) 1:500,anti-Nos (gift from A. Nakamura, RIKEN Center for Developmental Biology, Kobe,Japan) 1:1000, anti-Bcd (Kosman et al.,1998) 1:200, mouse anti-C(3)G(Anderson et al., 2005) 1:500. To visualize meiotic or mitotic spindles in embryos, methanol fixation was performed as described (Brent et al.,2000) and dilution of anti-α-tubulin (Sigma T9026) was 1:200. Meiotic spindles in stage 14 oocytes were visualized as described(Endow and Komma, 1997) using FITC-conjugated anti-α-tubulin (Sigma F2168).

Immunoprecipitations (IPs) were performed as described(Zaessinger et al., 2006). Each IP was with 60 ovaries from well-fed 3- to 4-day-old females and 5 μl of serum (immune or pre-immune) or 10 ml of hybridoma supernatant (Orb 6H4 or irrelevant 12CA5).

Two genes, GC5732 and CG15737, encoding GLD-2 homologs are present in the Drosophila genome. The corresponding proteins share the characteristics of GLD-2 family members in other species. They have a catalytic DNA polymerase β-like nucleotidyltransferase domain containing three conserved aspartic acid residues that is included in a larger conserved central domain, a PAP/25A-associated domain, and they lack an RNA-binding domain (Fig. 1A). The region that is N-terminal to the central domain is variable in size and non-conserved in the Drosophila GLD-2. Several CG5732 cDNAs described in FlyBase are from adult testis, indicating that CG5732 is expressed in this tissue. We verified by RT-PCR that CG5732 is not expressed in ovaries (data not shown). We focused on CG15737 (Fig. 1B), which was expressed in ovaries.

An antibody against the C-terminal half of the protein (residues 702-1373)was developed. In western blots, this antibody revealed a band of 180 kDa that was present in adult females and ovaries and absent from adult males(Fig. 1C). This band corresponds to the protein encoded by CG15737 as it was absent when the gene was mutated (see below). In embryos, expression of the protein peaked at 0-2 hours of development, decreased at 2-4 hours and was undetectable in later stages, consistent with a maternal expression. Immunostaining of ovaries showed that the protein is expressed throughout oogenesis(Fig. 1D). It was cytoplasmic,present both in nurse cells and the oocyte and accumulated in the oocyte from stage 5 onwards. In stages 9 and 10, protein accumulation was visible at the posterior pole of the oocyte. Staining of ovaries mutant for CG15737 was reduced to background levels, indicating that the antibody specifically recognized the protein encoded by this gene.

A P-element insertion in CG15737(CG15737^KG5287) was available at the Berkeley Drosophila Genome Project (BDGP), in which the insertion occurred 36 residues downstream of the initiation codon(Fig. 1B). Homozygous CG15737^KG5287 males were viable and fertile, whereas homozygous CG15737^KG5287 females were viable and sterile. Oogenesis did not appear affected and these females laid a normal number of eggs, but none of the embryos from mutant females crossed with wild-type males showed any development (Fig. 2A,B). Because the protein encoded by CG15737 was lacking in ovaries from CG15737^KG5287 homozygous females(Fig. 1C,D) and the phenotypes of CG15737^KG5287 homozygotes were identical to those of CG15737^KG5287/Df(1)RA47 (a deficiency overlapping the region) females, we believe that CG15737^KG5287 is a null allele. Mobilization of the P-element in CG15737^KG5287 generated either new mutant alleles(CG15737⁴⁰ and CG15737⁸⁹)corresponding to internal deletions of the P-element, or nearly complete deletions of the P-element that restored female fertility. A close examination of CG15737^KG5287 phenotypes showed that the gene is required at metaphase of meiosis I.

These phenotypes were similar to those of wisp mutants, a gene previously identified genetically and which is located in the same chromosome region as CG15737 (Brent et al.,2000; Mohler,1977). Complementation tests between CG15737^KG5287 and wisp^12-3147 showed that CG15737 is wisp. Sequencing of the conserved domains in wisp^12-3147 identified a point mutation, T1151I, in the central domain of the protein, changing a residue that is conserved in the other species (Fig. 1A). This mutation did not prevent the production of Wisp in ovaries(Fig. 1C).

Staining of embryos from wisp^KG5287 or wisp^KG5287/Df(1)RA47 females (thereafter called wisp embryos) with anti-α-tubulin and DAPI to visualize meiotic figures showed that these embryos did not complete meiosis(Fig. 2A-E). In most embryos(74%), a single meiotic spindle was visible, which could be thin or distorted and on which the chromosomes were scattered or separated in asymmetric pools(Fig. 2C,D). A second small spindle was sometimes present, which was nucleated by one or several lost chromosomes (8% of embryos). These figures (81%, Fig. 2E) correspond to a block in meiosis I. Among the remaining embryos, 15% were identified as blocked in meiosis II by visualization of two meiotic spindles, which in most cases were abnormally arranged. This phenotype is stronger than that of wisp^12-3147/Df(1)RA47 embryos, most of which were arrested at or after metaphase II (Brent et al.,2000). Meiotic figures in wisp mutant embryos correspond to either abnormal metaphase I or anaphase I. In the wild type, a meiotic arrest occurs at metaphase I in mature oocytes(Fig. 2F,G). Meiosis I resumes at egg activation, driven by egg laying, and meiosis II is rapidly completed without further arrest.

Meiotic figures were analyzed in wisp^KG5287/Df(1)RA47mature (stage 14) oocytes and we found that in most oocytes (79%), metaphase I arrest was not maintained properly (Fig. 2H-J). The chromosomes separated along the spindle and did so asymmetrically, the fourth chromosomes often migrating out of the spindle. The spindle appeared thin or irregular (Fig. 2H,I). Therefore, wisp mutants showed a defect in metaphase I arrest, and meiosis is then blocked at this stage as abnormal meiotic figures are similar before and after egg activation in wispmutant embryos. Because a defect in metaphase I arrest could result from a defect earlier in meiosis, we analyzed earlier aspects of meiosis. We found that the karyosome in wisp^KG5287/Df(1)RA47 stage 8 oocytes appeared as in the wild type (Fig. 2K,L). Entry into meiosis, as well as meiosis restriction to one oocyte in the germarium, as visualized by the formation of the synaptonemal complex using anti-C(3)G antibody (Anderson et al., 2005), were also as in the wild type (data not shown).

We conclude that the GLD-2-type poly(A) polymerase in the Drosophila female germline is encoded by wisp and is required for metaphase I arrest and for progression of meiosis after this stage.

A tethering assay was previously developed in Xenopus oocytes to analyze the poly(A) polymerase activity of candidate proteins(Kwak et al., 2004). The tested protein is tethered to mRNAs through MS2, an exogenous RNA-binding protein, and poly(A) addition as well as translational activation of the targeted mRNA are assayed. A chimeric mRNA encoding an MS2-Wisp fusion protein was injected into Xenopus oocytes. Two reporter mRNAs, luciferase with MS2 binding sites and β-galactosidase lacking MS2 binding sites,were then co-injected. Translational activation was calculated by comparing luciferase activity to β-galactosidase activity(Fig. 3A). HsGLD-2 was used as a positive control and induced translational activation that was abolished when the catalytic site was destroyed by mutation of an essential aspartic acid residue (HsGLD-2 DA). Wisp protein strongly stimulated translation of the reporter RNA and the stimulatory effect was abolished by a point mutation in the catalytic domain (Wisp DA)(Fig. 1A). Poly(A) tail addition to a short RNA containing MS2 binding sites was measured and a robust polyadenylation, which depended on the catalytic domain integrity, was observed in the presence of either MS2-HsGLD-2 or MS2-Wisp(Fig. 3B). These results demonstrate that Wisp is a poly(A) polymerase.

We confirmed the role of Wisp in poly(A) tail elongation in vivo using PAT assays, an RT-PCR-based technique that allows measurements of poly(A) tail length. Cytoplasmic polyadenylation of osk and CycB mRNAs during oogenesis has been reported and this depends on Orb(Benoit et al., 2005; Castagnetti and Ephrussi,2003). We determined whether it also depended on the Wisp poly(A)polymerase. Cytoplasmic polyadenylation of nanos (nos) and bcd mRNAs was also analyzed. Egg chambers of progressive stages were dissected and poly(A) tails were measured(Fig. 3C). For all tested mRNAs, poly(A) tails lengthened during oogenesis in the wild type, with a moderate lengthening between early stages (germarium to stage 8) and stages 9-10, and a pronounced elongation between stages 9-10 and stage 14. For bcd mRNA, this lengthening is not sufficient for translational activation, which occurs after egg activation(Salles et al., 1994). In wisp^KG5287 egg chambers, the lengthening at stage 9-10 was slightly affected and the strong elongation at stage 14 was completely abolished.

To determine whether Wisp has a general role in regulating mRNAs, we analyzed the poly(A) status of 30 mRNAs that we have identified to be regulated by cytoplasmic polyadenylation and the poly(A) tails of which undergo a robust lengthening in mature oocytes (I. Busseau and M.S.,unpublished). For all tested mRNAs, the poly(A) tail lengthening in wisp^KG5287 mature oocytes was impaired (see Fig. S1 in the supplementary material).

Together, these data show that Wisp is the poly(A) polymerase responsible for cytoplasmic polyadenylation of mRNA targets during late oogenesis.

To determine whether the meiotic defect observed in wisp mutants resulted from defects in mRNA polyadenylation and translational activation, we identified a Wisp target required for meiotic progression. cortencodes a female meiosis-specific activator of the APC that is necessary for female meiosis (Chu et al.,2001). Eggs from cort mutant females show aberrant chromosome segregation in meiosis I and eventually arrest in metaphase II(Lieberfarb et al., 1996; Page and Orr-Weaver, 1996). Cort is required for sequential degradation of CycA, B and B3 during meiosis,with CycA being degraded earlier, prior to metaphase I arrest(Pesin and Orr-Weaver, 2007; Swan and Schupbach, 2007). Cort protein expression peaks during oocyte maturation (stages 13 and 14) and correlates with cort mRNA poly(A) tail lengthening(Pesin and Orr-Weaver, 2007). We found that cort poly(A) tail elongation was abolished in wisp^KG5287 stage 14 oocytes(Fig. 4A). This led to a defect in Cort protein accumulation at this stage(Fig. 4B). The earliest target of Cort is CycA, the degradation of which fails in cort mutants by metaphase I arrest (Pesin and Orr-Weaver,2007). We verified that the lack of Cort accumulation in wisp^KG5287 mature oocytes resulted in a defect of CycA destruction: CycA levels were elevated in wisp^KG5287mature oocytes, consistent with impaired destruction(Fig. 4C).

These results identify cort, an mRNA required for meiotic progression, as a Wisp target, the poly(A) tail elongation and translation of cort being dependent on Wisp. They strongly suggest that the poly(A)polymerase function of Wisp is required during the progression of meiosis.

Cytoplasmic polyadenylation of osk and CycB mRNAs during oogenesis is Orb-dependent, and GLD-2 is in a complex with CPEB in Xenopus oocytes (Barnard et al.,2004; Rouhana et al.,2005). We therefore analyzed whether Wisp and Orb were present in a complex, by co-immunoprecipitations in ovary extracts. Orb was able to co-precipitate with Wisp and this interaction was RNA-independent(Fig. 5A). Conversely, Wisp co-precipitated with Orb, and the interaction again was RNA-independent(Fig. 5B). A direct interaction between Wisp and Orb was confirmed by in vitro binding assays, in which in vitro translated Orb bound to recombinant GST-Wisp(702-1373), but not to GST-Wisp(11-547) or to GST alone (Fig. 5E).

In C. elegans, GLD-2 directly interacts with the RNA-binding protein GLD-3 (Wang et al.,2002), a homolog of Drosophila Bicaudal C (BicC). In a co-immunoprecipitation between Wisp and BicC, Wisp co-precipitated with BicC. This interaction depended on the presence of RNA, suggesting that BicC and Wisp can be present in common RNP complexes, but do not interact directly(Fig. 5C).

We reported previously that poly(A) tail elongation of osk mRNA in ovaries, and Osk protein accumulation at the posterior pole of oocytes, depend on Orb and the canonical PAP (Juge et al.,2002). We now find that cytoplasmic polyadenylation of osk mRNA requires Wisp, another poly(A) polymerase. To understand the role of two different poly(A) polymerases in poly(A) tail lengthening of the same targets, we investigated the presence of PAP in the cytoplasmic polyadenylation complex. PAP co-precipitated with Orb in ovary extracts,independently of the presence of RNA (Fig. 5B). More strikingly, Wisp co-precipitated with PAP and the interaction was maintained in the absence of RNA(Fig. 5D).

These data show that Wisp is recruited to mRNAs through a direct interaction with Orb, and are consistent with the role of both PAP and Wisp in cytoplasmic polyadenylation with Orb.

To investigate the respective roles of PAP and Wisp in oogenesis, we analyzed genetic interactions between orb and wisp and between orb and the PAP-encoding gene hrg. Strong hrg mutants are lethal and do not produce late egg chambers in germline clones (Juge et al.,2002). Therefore, the role of PAP and Orb in cytoplasmic polyadenylation had been inferred from strong genetic interactions between heterozygous hrg alleles and the homozygous weak orb allele, orb^mel. The combination of both mutants strongly reduces female fertility and oogenesis stops around stage 8 in a number of ovarioles(Juge et al., 2002).

Because the phenotype of the orb^mel mutant tended to become stronger with time, we backcrossed orb^mel in a new background (Canton S) to produce a weaker phenotype. In this background, orb^mel females produced a low level of maternal-effect embryonic lethality (15%), which was increased to 31% and 45% in the hrg^-/+; orb^mel combinations, consistent with the previously described interactions (Fig. 6A). Embryonic lethality from wisp^-/+;orb^mel females also increased to similar levels of 35-50%. These genetic interactions corroborate the physical interactions between Wisp and Orb (Fig. 5) and provide additional support for the proposed combined role of Wisp and Orb in cytoplasmic polyadenylation. A number of embryos from orb^mel females are ventralized owing to defects in gurken (grk) mRNA localization and translation during mid-oogenesis (Chang et al.,2001). Whereas the percentage of ventralized embryos did not increase in interactions between wisp and orb, they were enhanced for embryos from hrg^-/+; orb^melfemales (26-31%) (Fig. 6A). These results suggest that hrg and orb interact during early or mid-oogenesis (before or during stages 9-10), whereas wisp and orb interact later (after stage 10).

We then compared the phenotypes of genetic interactions between orb and hrg and between orb and wispduring oogenesis. Oogenesis stopped at stage 8 in a number of ovarioles from hrg^PAP21/+; orb^mel females (34%)(Fig. 6B). Staining with anti-C(3)G to visualize the oocyte was impaired in arrested hrg^PAP21/+; orb^mel egg chambers, indicating either degeneration or the lack of oocyte determination(Fig. 6C). Lack of oocyte determination was confirmed by the presence of sixteen nurse cells (10% of ovarioles). Consistent with the possible function of PAP with Orb during early oogenesis, PAP and Orb expression overlapped at these stages (see Fig. S2 in the supplementary material). Note that these data do not preclude a role for PAP later in oogenesis.

By contrast, early defects did not appear in wisp^-/+;orb^mel ovaries (Fig. 6B). Moreover, translation of osk mRNA has been quantified previously in the orb^mel mutant and found to be strongly affected in stage 9-10 oocytes(Castagnetti and Ephrussi,2003). We found that Osk protein accumulation was not affected in wisp^KG5287/Df(1)RA47 oocytes at these stages(Fig. 6D) (n=128).

We conclude that PAP is required with Orb during early oogenesis, whereas Wisp has an essential function with Orb after stage 10 of oogenesis.

Upon egg activation, cytoplasmic polyadenylation leads to a robust poly(A)tail elongation and translation of a number of maternal mRNAs, including bcd. In wisp mutant embryos, poly(A) tail elongation of bcd mRNA was abolished (Fig. 7A). This short poly(A) tail was stable during the first 3 hours of embryogenesis and, consistently, bcd mRNA was not completely destabilized. bcd mRNA was also detected in embryos by in situ hybridization; however, the transcript was delocalized in the anterior region and to a lesser extent in the rest of the embryo(Fig. 7B). Maternal mRNA delocalization in wisp mutant embryos has been reported previously,but was weaker, probably owing to the utilization of weaker alleles(Brent et al., 2000). Consistent with previous data establishing that poly(A) tail elongation of bcd mRNA is necessary and sufficient for its translation in embryos(Salles et al., 1994), the lack of poly(A) lengthening in wisp mutant embryos prevented Bcd protein accumulation (Fig. 7C,see Fig. S3 in the supplementary material).

We measured poly(A) tail lengths of osk and nos mRNAs in embryos. In wild-type embryos, pools of osk and nos mRNAs are localized at the posterior pole, but another fraction of these mRNAs is widespread in the bulk cytoplasm (Bergsten and Gavis, 1999) and then degraded. In agreement with this, both osk and nos mRNAs analyzed by PAT assays and RT-PCR were deadenylated and destabilized during the first 3 hours in wild-type embryos(Fig. 7A). We have shown previously that nos mRNA destabilization in the bulk cytoplasm of the embryo depends on deadenylation by CCR4 (twin - FlyBase)(Zaessinger et al., 2006). In wisp mutant embryos, osk and nos mRNAs were destabilized prematurely in the first hour of embryogenesis and pools of these mRNAs were not stabilized at the posterior pole(Fig. 7A,B, see Fig. S3B,C in the supplementary material). Accordingly, Osk and Nos proteins were completely lacking in wisp mutant embryos(Fig. 7C). We propose that premature destabilization of osk and nos mRNAs in wisp mutant embryos results from increased shortening of poly(A)tails in the absence of poly(A) tail elongation by Wisp. As Smaug protein is required for nos mRNA destabilization, through the recruitment of the deadenylation complex (Zaessinger et al.,2006), we verified that Smaug levels were unaffected in wisp mutant embryos (data not shown).

These results demonstrate that Wisp is responsible for cytoplasmic polyadenylation in early embryos and that poly(A) tail elongation is required for the production of the major determinants in embryo anteroposterior patterning.

We have characterized Wisp, one of the two GLD-2-type poly(A) polymerases in Drosophila, which has a function in the female germline. We show that Wisp is a bona fide poly(A) polymerase: it has poly(A) polymerase activity in a tethering assay that depends on a conserved residue in the catalytic domain. Wisp is required for poly(A) tail lengthening of a pool of mRNAs in late stages of oogenesis. GLD-2 poly(A) polymerases do not have an RNA-binding domain; instead, they interact with RNA through their association with RNA-binding proteins. In Xenopus oocytes, GLD-2 interacts with CPEB in a complex that is active in cytoplasmic polyadenylation(Barnard et al., 2004; Rouhana et al., 2005). We find that Wisp interacts directly with Orb. Consistent with a role of Wisp and Orb together in an ovarian cytoplasmic polyadenylation complex, wispmutants are dominant enhancers of a weak orb allele. In C. elegans, GLD-2 has been reported to interact with the KH-domain RNA-binding protein GLD-3, which has homology with Drosophila BicC. Although we find Wisp and BicC together in an ovarian RNP complex, their association is mediated by RNA, suggesting that the proteins do not interact directly. We recently reported that BicC functions in deadenylation: BicC recruits the CCR4-NOT deadenylase complex to mRNAs(Chicoine et al., 2007). However, we also found a role of BicC in poly(A) tail elongation during oogenesis (Chicoine et al.,2007).

In addition to its function in oogenesis, Wisp-dependent cytoplasmic polyadenylation is required for the translation of essential determinants of the anteroposterior patterning of the embryo. bcd mRNA poly(A) tail elongation was known to be required for the deployment of the Bcd gradient from the anterior pole of the embryo(Salles et al., 1994). We now show that Osk and Nos accumulation at the posterior pole also depends on Wisp. This highlights the general role of poly(A) tail length regulation in Drosophila early development.

In Drosophila, meiosis starts in the germarium, where several cells per germline cyst enter meiotic prophase. Meiosis is then restricted to a single oocyte that remains in prophase I during most of oogenesis. Progression to metaphase I (oocyte maturation) occurs in stage 13, with maintenance of metaphase I arrest in mature stage 14 oocytes(King, 1970). Arrested oocytes are then activated by egg laying, which induces the resumption of meiosis(Heifetz et al., 2001).

The earliest phenotypes in wisp-null mutant are defects in metaphase I arrest and in the progression beyond this stage. This suggests that Wisp-dependent cytoplasmic polyadenylation and translational activation are essential for meiosis during and after metaphase I (but not for oocyte maturation). Consistent with this, massive translation appears to be dispensable for the completion of meiosis(Page and Orr-Weaver, 1997),but translational activation of specific mRNAs, at least of cort, is required (Pesin and Orr-Weaver,2007). We identify cort as a Wisp target: cortpoly(A) tail elongation and Cort accumulation in mature oocytes require Wisp. Moreover, defects in Cort accumulation in wisp mutant oocytes result in impaired CycA destruction, an event thought to be critical for meiotic progression (Pesin and Orr-Weaver,2007; Swan and Schupbach,2007). We find that Wisp regulates many mRNAs at oocyte maturation, several of which might be involved at various steps of meiosis. Identification of these specific targets will be necessary to fully unravel the role of Wisp during meiosis.

Cytoplasmic polyadenylation has been linked to meiotic progression at egg activation given that some maternal mRNAs undergo poly(A) tail elongation at egg activation (Benoit et al.,2005; Salles et al.,1994; Vardy and Orr-Weaver,2007). Moreover, bcd polyadenylation is affected in mutants that are defective in meiosis, such as cort mutants(Lieberfarb et al., 1996). It has been proposed that the link between cytoplasmic polyadenylation and egg activation results from the inactivation of canonical PAP activity by phosphorylation via the MPF (Mitotic promoting factor: Cdc2/CycB)(Chu et al., 2001). CycB degradation by APC-Cort would both induce meiotic progression and release PAP inactivation, leading to polyadenylation.

This model can be adapted with results presented here and in the recent literature (Pesin and Orr-Weaver,2007; Vardy and Orr-Weaver,2007). Two waves of cytoplasmic polyadenylation occur successively, one during oocyte maturation and one at egg activation. They both depend on Wisp poly(A) polymerase. The first wave is Orb-dependent and the pathway that triggers its activation is unknown. This polyadenylation induces the synthesis of Cort (and probably other proteins), which in turn is required for the second wave of cytoplasmic polyadenylation at egg activation. Cort could act in this process through the destruction of cyclins or of other proteins more specifically involved in the regulation of the polyadenylation machinery.

A striking result in this paper is the requirement of two poly(A)polymerases for cytoplasmic polyadenylation during oogenesis. Since the discovery of GLD-2 poly(A) polymerases, it has been assumed that these proteins were responsible for cytoplasmic polyadenylation. Our data reveal a higher level of complexity to this regulation. The phenotypes of wispmutants indicate a function of Wisp late in oogenesis. We find that entry into meiosis and restriction of meiosis to one oocyte, as well as DNA condensation in the karyosome, are unaffected in wisp mutants. By contrast, orb-null mutants arrest oogenesis in the germarium, with defects in the synchronous mitoses of cystoblasts and in the restriction of meiosis to one oocyte (Huynh and St Johnston,2000) (I. Busseau and M.S., unpublished). We find that orb phenotypes corresponding to early defects in oogenesis, including oocyte determination and dorsoventral patterning, are dominantly enhanced by hrg mutants, strongly suggesting that canonical PAP and Orb act together in cytoplasmic polyadenylation during the first steps of oogenesis. Because Orb forms complexes with both PAP and Wisp, the same pools of mRNAs can be regulated by the two different complexes, at different steps of oogenesis. The inclusion of one or other poly(A) polymerase could allow for different types of regulation. In addition, it is possible that the presence of both poly(A) polymerases together in the complex could be required for some step of oogenesis.

In Xenopus, GLD-2 catalyzes polyadenylation during oocyte maturation (Barnard et al.,2004; Rouhana et al.,2005), but the enzymes involved after fertilization have not been identified. Moreover, polyadenylation at earlier stages of oogenesis remains unexplored.

CPEB function has been addressed genetically in mouse and the defect in the female germline of Cpeb-knockout mice was found to be during prophase I (Tay and Richter, 2001). By contrast, GLD-2 expression in the oocytes appears to start at metaphase I(Nakanishi et al., 2006). Moreover, no female germline defective phenotype was observed in GLD-2 knockout mice (Nakanishi et al.,2007). This demonstrates some level of redundancy in poly(A)polymerase function in mouse female meiosis, and indicates that the involvement of different types of poly(A) polymerase for translational activation in oogenesis and meiotic progression is common to other species.

We thank A. Nakamura, P. Macdonald, P. Lasko, J. Reinitz, S. Hawley, D. Glover, J. Pesin and T. Orr-Weaver for gifts of antibodies; the Bloomington Stock Center for Drosophila stocks; M. Wolfner for sharing unpublished data; and I. Busseau for comments on the manuscript. This work was supported by the CNRS UPR1142, `ANR Blanche' (ANR-06-BLAN-0343) and FRM(Equipe FRM 2007) to M.S., and by NIH (GM31892 and 50942) grants to M.W. P.B. held awards from the Ministère de la Recherche and from the Association pour la Recherche sur le Cancer. J.E.K. held a Mary Shine Peterson Fellowship.