Oskar anchoring restricts pole plasm formation to the posterior of the Drosophila oocyte

The asymmetric transport of mRNA to precise subcortical regions of the cell is an evolutionarily conserved mechanism for the segregation of cytoplasmic determinants involved in cell-fate specification or patterning (Bashirullah et al., 1998). For example, in Saccharomyces cerevisiae, the asymmetric sorting of Ash1p, the repressor of cell-type switching, to the daughter cell relies on actin-based transport of Ash1 mRNA to the bud tip during mitosis (Takizawa et al., 1997). Here, cytokinesis occurs soon after Ash1 mRNA transport, ensuring segregation of the localized determinant away from the mother cell. However, in other organisms, an extensive delay can exist between localization of determinants and their ultimate entrapment by membranes. One strategy to achieve durable localization over longer periods is to anchor determinants robustly to their target site.

One prominent example in which a delay of several hours occurs between segregation and subsequent partition of cell-fate determinants into cells concerns the process of germline formation. In Drosophila melanogaster, this process begins during mid-oogenesis and progresses over many hours through the successive recruitment of RNAs and proteins, leading to formation of a specialized cytoplasm, the germ plasm, at the posterior tip of the mature oocyte (for a review, see Mahowald, 2001). It is only during cellularization of the embryo that germ plasm determinants are incorporated into pole cells, specifying the primordial germ cells of the fly. Delayed cellularization causes a germ cell-less phenotype due to the delocalization of the determinants before their internalization (Iida and Kobayashi, 2000). This underscores the importance of cortical anchoring of the determinants after localization. Similarly, anchoring most probably ensures maintenance of germ plasm localization at the vegetal cortex of full-grown Xenopus laevis oocytes awaiting maturation (Alarcon and Elinson, 2001). Although long-term cortical maintenance of determinants seems decisive for germline formation both in invertebrates and vertebrates, little is known of the mechanisms governing this process.

Germline formation and posterior patterning of the Drosophila embryo are regulated by a common genetic pathway. A crucial actor in this pathway is the maternally provided oskar (osk) mRNA, which localizes to the posterior pole of the oocyte at mid-oogenesis (Ephrussi et al., 1991; Kim-Ha et al., 1991). Posterior localization of osk mRNA requires a polarized microtubule network and an intact microfilament-based cytoskeleton, as revealed by mutants in the plus-end motor Kinesin I or in the actin-related proteins Profilin and Tropomyosin II, in which the RNA localization process fails (Brendza et al., 2000; Manseau et al., 1996; Erdélyi et al., 1995). Cis-acting elements located in the osk mRNA 3′ untranslated region (3′UTR) mediate RNA transport (Ephrussi and Lehmann, 1992; Kim-Ha et al., 1993). Two RNA-binding proteins, Staufen and Y14, as well as Mago nashi and Barentsz, are thought to be part of the transport complex (St Johnston et al., 1991; Hachet and Ephrussi, 2001; Mohr et al., 2001; Micklem et al., 1997; van Eeden et al., 2001). Until posterior localization, osk mRNA translation is repressed by several trans-acting factors bound to its 3′UTR (Kim-Ha et al., 1995; Gunkel et al., 1998; Lie and Macdonald, 1999). Upon localization, osk translation is activated by a still unknown mechanism involving a derepressor element in the 5′ part of osk mRNA and Staufen (Gunkel et al., 1998; Micklem et al., 2000). Throughout the rest of oogenesis and until early embryogenesis, both osk mRNA and Oskar (Osk) protein remain concentrated at the posterior cell cortex. Here, Osk nucleates formation of the germ plasm, also known as pole plasm because of its location, recruiting the components for both germline formation and posterior patterning. In contrast to germ cell determinants, which are maintained in a narrow subcortical region at the posterior of the egg until their encapsulation into pole cells, the posterior determinant Nanos diffuses towards the anterior, forming a gradient that governs abdominal segmentation of the embryo (for reviews, see Williamson and Lehmann, 1996; Wang et al., 1994).

The decisive role of Osk in induction of pole plasm assembly is revealed by genetic loss- and gain-of-function experiments. Embryos that lack maternal osk activity develop neither posterior segments nor pole cells (Lehmann and Nüsslein-Volhard, 1986). Conversely, mis-expression of osk at the anterior of the embryo causes formation of an ectopic pole plasm, interfering with normal anterior development, because of production of Nanos in the anterior of the embryo (Ephrussi and Lehmann, 1992; Smith et al., 1992). Observed defects range from loss of head structures to duplication of the abdomen and occasionally germ cells at the anterior, the latter requiring higher doses of Osk.

Despite the obvious contribution of RNA localization in generating a concentrated source of osk mRNA at the posterior pole, another mechanism must be responsible for its maintenance. Indeed, the polarized microtubule network, which is established at stages 7/8 and mediates the bulk of osk mRNA localization, is disassembled during stage 10, and the vigorous cytoplasmic flow of vesicles, called the oocyte streaming, ensues (Gutzeit and Koppa, 1982; Theurkauf et al., 1992). From then onwards, maintenance of osk localization requires Osk protein, Staufen and BicD (Kim-Ha et al., 1991; Ephrussi et al., 1991; Rongo et al., 1995; Swan and Suter, 1996). Of the two Osk isoforms translated from alternative in-frame initiation codons, Long Osk has been proposed to mediate osk mRNA maintenance, based on the tight localization of this isoform at the oocyte posterior (Markussen et al., 1995). Consistent with this, it has been reported that osk mRNA is maintained at the posterior of the early embryo only in the presence of Long Osk (Rongo et al., 1997). In contrast, Short Osk function appears restricted to pole plasm assembly (Markussen et al., 1995; Breitwieser et al., 1996).

We show that Long Osk anchors both osk mRNA and Short Osk at the posterior cortex of the oocyte, ensuring their stable maintenance during late oogenesis when cytoplasmic streaming occurs. Through this anchoring, Long Osk plays a dual role in Drosophila development, guaranteeing both spatial restriction and integrity of the pole plasm throughout oogenesis, ensuring proper patterning and efficient germline formation. This analysis unravels the crucial association of transport and anchoring mechanisms when localization of developmental determinants must be durably maintained.

The wild-type reference stock was Oregon R. The osk RNA null allele, osk^A87, is described elsewhere (M. Erdélyi, D. St Johnston and A. E., unpublished). Df(3R)p^XT103, which uncovers osk, has been previously described (Lehmann and Nüsslein-Volhard, 1986). The recipient stock for P element transformation was w¹¹¹⁸. Transformation was performed according to standard procedure (Rubin and Spradling, 1982). Transgene oskM139L was previously described (Markussen et al., 1995). The substitution (ATG to CGC) at the first translation initiation codon of the osk open reading-frame in the transgene oskM1R was introduced by oligonucleotide-directed mutagenesis using PCR. All constructs were cloned into the pCaSpeR4 vector (Pirrotta, 1988) for germline transformation.

Construction of the osk-K10 3′UTR transgene was as previously described (Riechmann et al., 2002). oskM1R-K10 3′UTR was generated by subcloning a BamHI-XbaI osk cDNA fragment containing the substitution ATG→CGC at the first translation initiation codon, into pUASP2 (Rorth, 1998) linearized by the same enzymes. Construction of the transgene m1⁴¹⁴lacwt has been described previously (Gunkel et al., 1998).

Whole-mount antibody staining of ovaries and embryos was performed according to Tomancak et al. (Tomancak et al., 2000) and RNA in situ hybridization of embryos was as described by Mata et al. (Mata et al., 2000). Preparation of ovarian extracts (Markussen et al., 1997) and western blotting analysis have been previously described (Markussen et al., 1995). The polyclonal rabbit anti-Osk antibody used in this work was generated against an EcoRV-SphI fragment of the Blue-osk cDNA (Ephrussi et al., 1991) and was a gift from Andreas Jenny. Membranes were stripped and re-probed with an α-tubulin monoclonal antibody DM 1A (1:2000, Sigma). Rat anti-Vasa polyclonal serum, used to visualize pole cells, was previously described (Tomancak et al., 1998). β-Galactosidase activity was detected in situ in ovaries as described in elsewhere (Clark et al., 1994). Stained ovaries were mounted in 80% glycerol.

An osk RNA probe was synthesized by run-off transcription from the plasmid oskTAAXba linearized with XbaI (Ephrussi and Lehmann, 1992) and labeled with digoxigenin-11-UTP ribonucleotides (Roche). For in situ hybridization, ovaries were fixed in 4% paraformaldehyde/phosphate-buffered saline (PBS) for 30 minutes and rinsed three times 10 minutes in PBT (PBS + 0.5% Tween 20). Subsequent steps were as previously described for RNA in situ hybridization of embryos (Wilkie et al., 1999). The osk digoxigenin-labeled probe was visualized by TSA detection using FITC-tyramide (NEN-life Sciences, UK). After overnight equilibration in N-propylgallate, ovaries were mounted on slides and images were captured with a Leica laser scanning confocal microscope.

FISH/antibody double-labeling was performed according to Hughes and Krause (Hughes and Krause, 1999) with the following modifications. Ovaries were fixed in 4% paraformaldehyde/PBS for 30 minutes and rinsed three times for 10 minutes in PBT. Ovaries were successively pre-hybridized 10 minutes in HYB/PBT (1:1) (see Hughes and Krause, 1999) and 1 hour in HYB at 64°C. Hybridization was carried out at 64°C. After the post-hybridization washes, the ovaries were directly incubated with anti-digoxigenin (1:200) and rabbit anti-Osk antibodies (1:1000) in PBT during 12-15 hours at room temperature. Ovaries were washed three times for 10 minutes in PBT and incubated with a Rhodamine-coupled anti-rabbit secondary antibody (1:300, Jackson ImmunoResearch Laboratories) during 2 hours before proceeding with the TSA-based detection of osk RNA.

For rescue analysis, eggs were collected on apple juice plates and hatch rates scored by counting hatched and unhatched eggs, after aging for 24 hours at 25°C. Fertility of hatchers was tested by mating the female offspring with wild-type males.

To directly investigate the process of osk RNA maintenance during oogenesis, we expressed the two Osk isoforms separately in osk mRNA null females (M. Erdélyi, D. St Johnston and A. E., unpublished) and observed osk mRNA distribution in ovaries. We expressed each isoform independently from the oskM1R and oskM139L transgenes, in which the first or the second start codon was mutated, respectively (see Materials and Methods) (Markussen et al., 1995). In situ hybridization reveals that both transgenic osk mRNAs accumulate in the oocyte during the early stages of oogenesis and concentrate at the posterior pole during mid-oogenesis (Fig. 1A), as is the case for endogenous osk mRNA in wild-type egg chambers (Kim-Ha et al., 1991; Ephrussi et al., 1991). During stage 10, however, a striking difference in the distribution of the two osk mRNAs is observed. In the case of transgenic lines expressing only Long Osk, osk mRNA remains tightly localized at the posterior pole of late stage oocytes and in the early embryo (lower panels), as it does when both isoforms are expressed from a wild-type rescuing transgene (upper panels). In contrast, in oocytes expressing only Short Osk, the bulk of osk mRNA detaches from the posterior cortex of the oocyte and disperses massively (middle panels). The penetrance of delocalization is high: 95% of late stage 10 oocytes (n=118) show significant RNA detachment, resulting in only trace amounts of osk mRNA detected at the posterior pole of freshly laid eggs (see arrow). Notably, this pattern of delocalization is distinct from that observed in the osk mutants osk⁸⁴ and osk³⁴⁶, which are predicted to produce only truncated Osk peptides (Kim-Ha et al., 1991). Indeed, whereas osk⁸⁴ mRNA spreads diffusely during stage 10 and remains visible only in a restricted area close to the posterior pole (Fig. 1B, top panel), oskM1R mRNA disperses in tight aggregates that are readily detected in the anterior half of the oocyte (Fig. 1B, bottom panel). Western blotting analysis of ovary extracts reveals that similar quantities of Short and Long Osk are produced from the two transgenes oskM1R and oskM139L (Fig. 1C). This indicates that a qualitative, rather than a quantitative, difference accounts for the differential ability of Short and Long Osk to maintain osk mRNA at the posterior pole of the oocyte after mid-oogenesis. Thus, we conclude that the two Osk isoforms differ in their intrinsic competence to maintain osk mRNA and that only Long Osk fulfils this function.

Possible explanations of the failure of Short Osk to maintain osk mRNA at the oocyte posterior might be that Short Osk lacks elements crucial for its association with an anchoring target at the cortex and/or lacks the binding activity necessary for RNA sequestration. To investigate these possibilities, we first analyzed the distribution of the two Osk isoforms when expressed separately during oogenesis, in the same osk RNA null mutant background (Fig. 2). Both isoforms are detected during mid-oogenesis at the posterior pole of the oocyte where osk mRNA is localized (Rongo et al., 1995; Markussen et al., 1995). However, during late stage 10 of oogenesis, Short Osk disperses in aggregates from its initial cortical location in 97% of egg chambers (n=84), in contrast to Long Osk, which persists tightly anchored at the oocyte cortex (Fig. 2A, middle and lower panels). Higher magnification of the posterior pole at this stage shows that only small amounts of Short Osk are maintained at the oocyte cortex, in comparison with Long Osk (Fig. 2B). Hence, Long Osk, but not Short Osk, has the capacity to anchor to the oocyte cortex, where osk mRNA is localized and translated. To assess whether Short Osk has the capacity to associate with its own mRNA, we then examined the released Short Osk and osk mRNA aggregates in oskM1R oocytes, by simultaneous detection of RNA and protein. The merged confocal image reveals that Short Osk and osk mRNA co-localize in the released particles (Fig. 2C, middle panel). This demonstrates that Short Osk, although unable to anchor at the cortex, is closely associated with its own mRNA as is Long Osk (Fig. 2C, lower panel).

We next investigated localization of other pole plasm components in egg chambers expressing a single Osk isoform. We examined the distribution of Staufen, an RNA-binding protein associated with osk mRNA throughout oogenesis, and of Vasa, a DEAD-box RNA helicase recruited by Short Osk and required for subsequent recruitment of the other pole plasm components (St Johnston et al., 1991; Lasko and Ashburner, 1990; Hay et al., 1990; Breitwieser et al., 1996). In oocytes expressing only Short Osk, Staufen and Vasa are detected in the same Short Osk/osk RNA detaching granules (Fig. 2D, middle panel, Fig. 2E), showing that Short Osk detachment causes pole plasm delocalization. By contrast, in ovaries expressing only Long Osk, Staufen remains associated with Long Osk at the posterior pole of the oocyte (Fig. 2D, lower panel), whereas Vasa is not recruited (data not shown) (Breitwieser et al., 1996).

To determine whether expression of Long Osk would suffice to rescue the osk mRNA and Short Osk anchoring defects observed in oskM1R transgenic ovaries, we generated flies expressing both Short and Long Osk from the oskM1R and oskM139L transgenes. Remarkably, co-expression of Long Osk entirely suppresses the granular detachment of both oskM1R mRNA and its protein product, Short Osk (Fig. 3B,D). In situ hybridization reveals a pattern of osk mRNA maintenance indistinguishable from that observed in stage 10 oocytes expressing the two Osk isoforms from a wild-type osk transgene (compare Fig. 3B with Fig. 1A). Similarly, the characteristic crescent-shaped distribution of Osk isoforms observed in wild-type oocytes is detected at the posterior pole of oocytes produced by the doubly transgenic females (Fig. 3D). This result demonstrates that the nucleotide substitutions in oskM1R mRNA, replacing the first translation initiation codon, do not interfere with RNA cortical anchoring. It also rules out that ongoing translation from M1 is required per se for RNA maintenance. Rather, we conclude that expression of Long Osk from the oskM139L transgene is responsible for anchoring, in trans, of both oskM1R mRNA and Short Osk after mid-oogenesis.

To determine if cortical anchoring influences Short Osk accumulation during oogenesis, we compared the amount of Short Osk produced in transgenic females expressing oskM1R either alone or together with oskM139L. Western blotting of ovary extracts reveals that significantly higher levels of Short Osk are produced in females also expressing Long Osk (Fig. 3E). Indeed, wild-type levels of Short Osk are detected when the two transgenes are co-expressed, demonstrating that Long Osk promotes Short Osk accumulation during oogenesis. Long Osk presumably controls Short Osk levels by maintaining osk mRNA at the posterior pole of the oocyte, where it is translated; it is also possible that Long Osk acts directly to stabilize Short Osk.

Given that a localized source of Osk activity is required for correct anteroposterior patterning and fertility, it seemed likely that loss of Short Osk anchoring would affect these processes. However, conflicting results have been reported concerning the extent to which Short Osk alone can support patterning and germline formation in the embryo (Markussen et al., 1995; Rongo et al., 1997). We therefore took advantage of the oskM1R transgene to evaluate carefully the consequence of expressing Short Osk alone on these processes.

We first analyzed the hatching rate of embryos produced by two lines expressing oskM1R. Both of the lines exhibit a moderate reduction in hatch rate (71% and 78%) compared with the control line expressing a wild-type osk transgene (88%) (Fig. 4A). In contrast, the progeny of oskM139L females, which only express Long Osk, do not hatch, as previously shown (Markussen et al., 1995). However, females bearing both oskM1R and oskM139L produce embryos whose hatch rate is very similar to that of lines expressing a wild-type osk transgene, showing that Long Osk is required to achieve wild-type hatch rates. We further looked for cuticle defects in unhatched embryos produced by oskM1R females. Approximately half of the abnormal cuticles exhibit strictly posterior group defects (Fig. 4A,B, left panel). Strikingly, the remaining half shows either only anterior group defects or both anterior and posterior group defects, ranging from subtle reduction to complete deletion of head structures (Fig. 4A,B, middle and right panels). In some cases, anterior duplication of the filtzkorper is also observed (data not shown), revealing posteriorization of the anterior in these embryos. Such anterior defects can be caused by translation of unlocalized osk mRNA, which is recognizable by the diffuse distribution of Osk throughout oocyte before stage 7 (Kim-Ha et al., 1995). Indeed, immunostaining for Short Osk in early oskM1R oocytes, using our anti-Osk serum at unusually high concentrations, reveals a very faint, but reproducible, diffuse signal above the nonspecific background (data not shown). However, although ovaries co-expressing both oskM1R and oskM139L still show premature Osk signal (data not shown), both anterior and posterior patterning defects are suppressed in the progeny (Fig. 4A,C). We therefore conclude that the patterning defects of oskM1R embryos are primarily the result of Short Osk delocalization from the posterior pole of the oocyte, which is rescued by co-expression of Long Osk (Fig. 3). Thus, both posterior and anterior defects can be explained by the delocalization of pole plasm and its abdomen-inducing activity from the posterior of the oocyte.

To determine whether germline formation is also impaired in the absence of Short Osk anchoring, we first examined the pole cells of embryos produced by oskM1R females. Embryos from the two different oskM1R lines show a dramatic reduction in the number of pole cells formed at the posterior pole, compared with embryos of the osk wild-type transgene control (Fig. 5A,C). In contrast, embryos from females co-expressing oskM1R and oskM139L develop pole cells similar in number to the wild-type osk control (Fig. 5B). We directly evaluated the fertility of adult female progeny, by scoring their ovaries (Fig. 5D). Notably, only half of the progeny of oskM1R females develop wild-type ovaries. The rest are sterile, developing either rudimentary ovaries consisting only of the terminal filaments or, in rare cases, tiny ovaries (‘atrophic’). More importantly, wild-type fertility is restored in both lines in which the oskM1R transgene is co-expressed with oskM139L. This demonstrates that Long Osk is required to achieve wild-type fertility, despite its inability to assemble the pole plasm (Markussen et al., 1995; Breitwieser et al., 1996). Given our previous results showing that Long Osk anchors Short Osk and osk mRNA at the posterior pole of the oocyte, we conclude that the anchoring activity of Long Osk is required to generate the high local source of Short Osk necessary for wild-type germline formation.

It has been shown that when osk mRNA is translated throughout the oocyte, as can be achieved by replacing the osk 3′UTR with that of K10 (osk-K10), an ectopic accumulation of Osk around the entire oocyte cortex is observed (Fig. 6A) (Riechman et al., 2002). Although both Osk isoforms are recognized by the antibody in this experiment, our results concerning Osk anchoring at the posterior pole would predict that the cortical signal is dependent on the ectopic expression of Long Osk. To test this possibility, we ectopically expressed Short Osk alone, by fusing the oskM1R-coding region to the K10 3′UTR (oskM1R-K10), and analyzed its distribution in a wild-type background. Strikingly, Short Osk is no longer detected around the oocyte cortex (Fig. 6B). Western blot analysis of protein extracts from osk-K10 and oskM1R-K10 ovaries shows that similar amounts of Short Osk are produced from the two transgenes (Fig. 6C), indicating that, even when unlocalized, Short Osk is stable. Thus, we conclude that Long Osk is required for Osk anchoring to all regions of the oocyte cortex.

Because Long Osk contains an N-terminal extension not present in Short Osk, we tested whether this difference is responsible for the unique property of Long Osk to anchor at the oocyte cortex. A previous study has shown that a fusion between the N-terminal extension of Long Osk and β-galactosidase (Osk-β-galactosidase fusion) accumulates at the posterior tip of the oocyte in wild-type egg chambers (Gunkel et al., 1998). This accumulation relies on the Long Osk region of the fusion, as wild-type β-galactosidase is not enriched at the posterior pole, even when translated from a posterior-localized transgenic mRNA (N. Gunkel and A. E., unpublished). To examine if endogenous Osk plays a role in the anchoring of the Osk-β-galactosidase fusion protein in wild-type egg chambers, we expressed this fusion in a background in which full-length Osk is not produced (osk⁸⁴/Df(3R)p^XT103). The fusion protein is tightly localized at the posterior of osk⁸⁴/Df(3R)p^XT103 oocytes at stage 10 (Fig. 7C). However, after stage 10, Osk-β-galactosidase diffuses away from the posterior pole and is no longer detected (Fig. 7D). In contrast, the fusion protein remains localized at the posterior pole of wild-type oocytes during late oogenesis (Fig. 7A,B), as previously reported (Gunkel et al., 1998). Thus, we conclude that the N-terminal extension of Long Osk is required but is not sufficient for posterior anchoring during the late stages of oogenesis.

After mid-oogenesis, at the posterior tip of the Drosophila oocyte, a region morphologically distinct from the remaining ooplasm forms under the control of Osk. This spatially restricted region, the pole plasm, contains the information required for specification of the abdomen and of the germ cells of the future embryo. In this study, we have shown that Long Osk anchors both osk mRNA and Short Osk at the posterior cortex of the oocyte. Unanchored Short Osk, the pole plasm-inducing isoform, nucleates a functional but untrapped pole plasm whose components, such as Staufen or Vasa, disperse in the bulk ooplasm at the onset of oocyte streaming. The loss of pole plasm components from the posterior pole of the oocyte results in a significant impairment of fertility in the progeny. In addition, the erratic dispersion of the posterior-inducing activity of Osk throughout the oocyte ooplasm causes anteroposterior patterning defects during embryogenesis.

At mid-oogenesis, the transport of osk mRNA to the posterior pole of the oocyte requires the plus-end motor Kinesin I, a polarized microtubule network and an intact actin cytoskeleton (for a review, see Riechmann and Ephrussi, 2001) (Brendza et al., 2000; Theurkauf et al., 1992; Manseau et al., 1996; Erdélyi et al., 1995; Baum et al., 2000). The polarized and sustained transport of osk mRNA can also account for its maintenance at the posterior pole of the oocyte from stage 7 to 10. In contrast, at stage 10, an active process of mRNA maintenance must exist, as the polarized microtubule network is disassembled and a subcortical array of microtubules forms and promotes vigorous cytoplasmic streaming. Indeed, several lines of evidence indicate that, as of stage 10, Osk protein maintains osk mRNA localization. First, osk mRNA delocalizes during stage 10 in the three osk nonsense mutants osk⁵⁴, osk⁸⁴ and osk³⁴⁶, predicted to produce truncated Osk peptides of 179, 253 and 323 amino acids, respectively (Kim-Ha et al., 1991). However, our failure to detect these peptides by western analysis (N. V. and A. E., unpublished) suggests that they are unstable and the nonsense osk alleles are protein null. Second, the maintenance of a transgenic nos-osk3′UTR mRNA at the posterior pole of stage 10B oocytes requires Osk protein (Rongo et al., 1995). Finally, posterior accumulation of fluorescent osk mRNA, injected into living oocytes at stage 10-11, occurs by a trapping mechanism dependent on endogenous Osk protein (Glotzer et al., 1997). Our analysis shows that, although it supports expression of Short Osk, oskM1R transgenic mRNA detaches from the posterior pole during late stage 10. In contrast, we show that Long Osk, which is dispensable for pole plasm formation (Markussen et al., 1995), is competent and required to persistently confine osk mRNA at the posterior pole of the oocyte during late oogenesis until early embryogenesis. Thus, osk mRNA maintenance is an active process mediated by Long Osk. Maintenance of the three nonsense osk mRNAs is rescued in heterozygous females (Kim-Ha et al., 1991). Our present work reveals that this rescue in trans can only be attributed to the Long Osk isoform encoded by the wild-type osk gene. Consistent with this, we show that Long Osk can also maintain localization of the transgenic oskM1R mRNA, which encodes only Short Osk.

In the absence of Long Osk, Short Osk also detaches from the posterior cortex of stage 10 oocytes, in concert with oskM1R mRNA. Both Short Osk and osk mRNA delocalize in dense aggregates, suggesting that they might be associated. Consistent with this, we have shown that they co-localize in the same released aggregates. It is noteworthy that the delocalizing pattern of oskM1R mRNA is significantly different from that of osk⁸⁴ mRNA, which diffuses without forming aggregates. Because osk⁸⁴ mRNA encodes an unstable Osk peptide, we conclude that aggregate formation is dependent on Short Osk. The pole plasm protein Vasa, which is a component of the polar granules, the germline granules of Drosophila, is also detected in these aggregates. This suggests that the aggregates contain nascent but untethered polar granules, whose assembly might be initiated by Short Osk-mediated clustering of osk mRNA. The ability of Short Osk to package macro-molecular complexes is supported by our observation that it can oligomerize, in a yeast two-hybrid assay (A. Jenny and A. E., unpublished). Given the underexpression of Long Osk relative to Short Osk in wild-type ovaries, multimerization of Short Osk could also explain the apparent non-stoichiometric competence of Long Osk to anchor Short Osk at the oocyte cortex.

Because both Long and Short Osk can sequester osk mRNA, it is likely that the same region in the two isoforms mediates RNA association. However, no RNA-binding activity has been reported for Osk, which does not exhibit any predicted RNA-binding motif in its coding sequence. Thus, the association of the two Osk isoforms with osk mRNA most probably involves adaptor(s). One such candidate could be Staufen, a double-strand RNA-binding protein suspected to bind osk mRNA directly (Micklem et al., 2000). Staufen is required for posterior maintenance of osk mRNA, as revealed by its delocalization in the temperature-sensitive mutant stau^C8 (Rongo et al., 1995). We have shown that, in ovaries expressing each Osk isoform individually, Staufen either co-localizes with the Long Osk/osk mRNA complex at the posterior pole of the oocyte or co-segregates with the Short Osk/osk mRNA complex in the released aggregates, as expected of an adaptor factor. Previous analysis has shown that Staufen binds to Short Osk in a yeast two-hybrid assay, but binds quite poorly to Long Osk (Breitwieser et al., 1996), which does not reflect the robust ability of Long Osk to maintain osk mRNA localization that our study reveals. Thus, whether the association of the two Osk isoforms with osk mRNA relies on a direct interaction with Staufen or with another adaptor factor remains to be elucidated.

The observation that Long Osk but not Short Osk can anchor at the cortex suggests that the N-terminal extension of Long Osk mediates anchoring. Surprisingly, the extension is not sufficient for this function, as revealed by its failure to maintain an Osk-β-galactosidase fusion in Osk protein-null oocytes at the onset of oocyte streaming. This suggests that at least two separate docking modules, one in the N-terminal extension of Long Osk and a second in the region shared by the two Osk isoforms, cooperate to form a robust anchoring domain. Two observations support this hypothesis. First, the Osk-β-galactosidase fusion remains localized during stages 8 to 10 of oogenesis in Osk protein-null oocytes, whereas native β-galactosidase translated from a posterior localized RNA fails to accumulate at this location. Second, even in the absence of Long Osk, a residual amount of Short Osk remains localized at the posterior pole of the oocyte during late oogenesis and supports substantial posterior patterning and partial fertility of the progeny. Thus, two docking modules appear to be involved in Long Osk anchoring, but neither alone is sufficient for this process.

An important result of this work is that Long Osk anchors Short Osk, the pole plasm-inducing isoform, at the posterior pole of oocyte. Strikingly, although it contains the entire Short Osk sequence, Long Osk can not recruit pole plasm components (Markussen et al., 1995; Breitwieser et al., 1996). It has previously been proposed that the N-terminal extension of Long Osk exerts an inhibitory effect on downstream protein-interaction domains (Breitwieser et al., 1996). We propose that this inhibition is caused by folding of the robust anchoring domain of Long Osk, masking the pole plasm-recruiting activity of this isoform. By contrast, the absence of the N-terminal anchoring module would allow Short Osk to nucleate pole plasm assembly. Hence, anchoring and pole plasm nucleation might be structurally mutually exclusive activities.

Our results show that, in the absence of Long Osk-anchoring activity, which causes a massive dispersion of Short Osk from the posterior pole of the oocyte, effective abdominal patterning in the embryo can nonetheless be achieved. Consistent with this, strong impairment of osk mRNA localization and translation in mutants in Tropomyosin II, an actin-binding protein, and Barentsz, a putative component of the transport machinery, has little consequence on abdominal development (Erdélyi et al., 1995; van Eeden et al., 2001). This demonstrates that, with regard to abdomen formation, a substantial excess of Short Osk is present at the posterior pole of the wild-type oocyte. However, our results indicate that dispersion of untethered Short Osk from the posterior of oskM1R oocytes can hinder anterior development. We cannot exclude a contribution of the weak premature translation of oskM1R to these anterior patterning defects. However, the complete suppression of these defects by co-expression of oskM139L, demonstrates that the anchoring activity of Long Osk can restrict the pole plasm-inducing activity of Short Osk to the posterior. Anchoring of the bulk of Short Osk to the posterior pole would lead to titration of limiting pole plasm components from any residual ectopically localized Osk.

In contrast, Short Osk delocalization causes a significant reduction of germ cell precursors and fertility of the progeny. Our results provide yet another demonstration of the correlation between Osk protein dose and the number of pole cells formed. Indeed, whereas overexpression of Osk at the posterior pole increases pole cell number (Ephrussi and Lehmann, 1992; Smith et al., 1992), its underexpression, caused by defects in RNA localization or translation, impedes pole cell formation (Erdélyi et al., 1995; Lie and Macdonald, 1999; Chang et al., 1999; van Eeden et al., 2001). In our analysis, restoration of Short Osk anchoring by co-expression of Long Osk enhances Short Osk accumulation and restores germ-plasm integrity. This demonstrates that Long Osk guarantees accumulation of high levels of Short Osk in the subcortical region of the egg that is subsequently incorporated into pole cells. It has been shown that when the D. virilis Osk homolog is expressed in Drosophila melanogaster oocytes, it efficiently rescues the posterior patterning defects of osk mutants, but does not support pole cell formation in the embryos (Webster et al., 1994). Transgenic D. virilis Osk fails to maintain osk mRNA localization in D. melanogaster, which led Webster et al. to hypothesize that virilis Osk is not competent to anchor at the posterior pole of the D. melanogaster oocyte. This conclusion is entirely consistent with our demonstration that Osk-mediated anchoring of the pole plasm is a critical step during Drosophila germline formation.

Given that both efficient germ cell formation and, to a lesser extent, proper patterning rely on Long Osk-mediated anchoring of the pole plasm, an important issue in the future will be the characterization of the mechanism by which Long Osk is tethered to the cortex.

We thank Daniel St Johnston for the osk^A87 mutant and Paolo Filardo for the osk-K10 transgenic line. We are particularly grateful to Shoko Yoshida and the other members of the Ephrussi laboratory for advice and discussions during the course of this work, to Anna Cyrklaff for help with fluorescent RNA in situ hybridization and to Ann-Mari Voie for P element transformation. N. V. has been supported by EMBO and Human Frontier Science Program Organization long-term fellowships.