The endocytic pathway acts downstream of Oskar in Drosophilagerm plasm assembly

The polarized targeting and anchoring of specific molecules and organelles to particular subcellular regions are crucial for many cellular processes,including cell-polarity establishment and cell-fate determination. In many animals, germline fate is controlled by maternal factors localized to a specialized cytoplasmic region within the egg, called the germ plasm(Extavour and Akam, 2003). Germ plasm contains germ granules, which are electron-dense, and non-membranous structures consisting of maternal RNAs and proteins required for the formation of germ cells. Drosophila germ plasm, also called pole plasm, forms at the posterior pole of the embryo and is inherited by the germline precursors, or pole cells(Mahowald, 2001). Because the cytoplasmic transplantation of the pole plasm into recipient embryos causes the ectopic formation of pole cells(Mahowald, 2001), the pole plasm contains sufficient factors for germ-cell formation. This observation also highlights the importance of retaining the pole plasm at the posterior cortex of the embryo to ensure the germ cells form at the appropriate location.

In Drosophila, the pole plasm is assembled during oogenesis, which is divided into 14 morphologically distinct stages of egg chamber development(Spradling, 1993). The egg chamber is composed of a single oocyte and 15 nurse cells, surrounded by a monolayer of somatic follicle cells. During oogenesis, most components of pole plasm are synthesized in the nurse cells and transported into the oocyte via ring canals, which are cytoplasmic bridges interconnecting the oocyte with nurse cells. Within the oocyte, these factors become concentrated at the posterior pole and are assembled into the polar (germ) granules. These factors are transported by a polarized microtubule (MT) array that is initially nucleated at the oocyte posterior and extends into the nurse cells through the ring canals (Steinhauer and Kalderon,2006). During stages 6-7, the MT array is reorganized by the transforming growth factor α-like Gurken (Grk) signal(Steinhauer and Kalderon,2006). In the stage-6 oocyte, posteriorly restricted Grk induces neighboring follicle cells to adopt the posterior fate. These cells send back as-yet unknown signals to the oocyte to trigger the reorganization of the MT cytoskeleton. Consequently, the MT array within the oocyte becomes polarized along the anteroposterior (AP) axis, with the minus ends abundant at the anterior of the oocyte and the plus ends extending toward the posterior. This MT organization promotes the migration of the oocyte nucleus and associated grk RNA to the future anterior-dorsal corner, where Grk signals the follicle cells to define the dorsoventral axis. The polarized MT array also directs the localization of bicoid (bcd) RNA to the anterior and oskar (osk) RNA to the posterior within the oocyte. The anterior accumulation of bcd RNA is required for the proper development of the embryonic head and thoracic structures(Ephrussi and St Johnston,2004). The posterior localization of osk RNA is essential for the formation of the germ cells and abdomen(Mahowald, 2001).

osk RNA localization is tightly coupled to translational control:only the posteriorly localized osk message is translated(Wilhelm and Smibert, 2005). The localized Osk protein, in turn, recruits downstream components of the pole plasm, such as Vasa (Vas) and Tudor (Tud) proteins, and the nanos, germ cell-less and polar granule component RNAs(Mahowald, 2001). Misexpression of Osk at the anterior of the oocyte causes ectopic pole plasm assembly and the formation of germ cells at the new site(Ephrussi and Lehmann, 1992),indicating that Osk organizes pole plasm assembly.

Although osk has no known alternatively spliced variants, the osk message produces two protein isoforms, long and short Osk, by translation from in-frame alternative start codons(Markussen et al., 1995). Short Osk shares its entire sequence with the long isoform. Nevertheless,genetic evidence shows that the two Osk isoforms have distinct functions in the assembly of the pole plasm (Markussen et al., 1995; Breitwieser et al., 1996; Vanzo and Ephrussi,2002). Short Osk assembles the pole plasm(Markussen et al., 1995) and is likely to recruit Vas directly(Breitwieser et al., 1996). Long Osk is required for all the components of the pole plasm, including Osk itself, to be anchored to the posterior cortex, preventing their diffusion into the cytoplasm (Vanzo and Ephrussi,2002). However, the mechanism by which long Osk retains pole plasm components at the posterior cortex remains unknown.

A recent immuno-electron microscopic study revealed that the two Osk isoforms localize to distinct organelles in the oocyte posterior: long Osk associates with endosomes and short Osk is concentrated in the polar granules(Vanzo et al., 2007). Long Osk also upregulates endocytosis, which occurs preferentially at the oocyte posterior (Vanzo et al.,2007). Therefore, the endocytic pathway may be involved in pole plasm assembly downstream of long Osk, although data are lacking to show that the association between long Osk and endosomes is functionally significant. Several reports have suggested that vesicular trafficking is involved in pole plasm assembly and germ cell formation(Ruden et al., 2000; Jankovics et al., 2001; Dollar et al., 2002; Coutelis and Ephrussi, 2007; Januschke et al., 2007; Swanson and Poodry, 1980). For example, in mutants for Rab11, which encodes a small GTPase involved in the recycling of endosomes, osk RNA fails to be transported to the oocyte posterior, instead forming aggregates close to the posterior(Jankovics et al., 2001; Dollar et al., 2002). However,the defects in osk RNA localization in Rab11 mutants are thought to be an indirect consequence of the disrupted MT polarization(Jankovics et al., 2001; Dollar et al., 2002).

Here, we show that Drosophila Rabenosyn-5 (Rbsn-5), a Rab5 effector protein involved in the early endocytic pathway, is required for osk RNA localization and pole plasm assembly. Although the primary defect of the rbsn-5 mutation is, as in the Rab11 mutant,caused by the failure to maintain MT polarity, which secondarily affects osk RNA localization, we provide evidence that the endocytic pathway also functions downstream of Osk to anchor the pole plasm components to the oocyte cortex.

Flies were reared in vials containing a standard cornmeal medium at 25°C. yw flies were used as the wild-type strain. Five deficiency lines, Df(2L)spd, Df(2L)Trf-C6R31,Df(2L)Exel7034, Df(2L)TE29Aa-11 and Df(2L)N22-5, were obtained from the Bloomington Stock Center and used to map the C241 region. To analyze the osk and grk mutants, respectively, the following genotypic combinations were used: osk⁵⁴/Df(3R)p^XT103(Kim-Ha et al., 1991) and grk^HK36/Df(2L)ED623(Neuman-Silberberg and Schüpbach,1993). To assess the specificity of newly raised antibody against Rab5, a Rab5-null allele, Rab5²(Wucherpfennig et al., 2003)was used. To analyze MT polarity, yw; Pin/CyO; KZ503 (Clark et al.,1994) was used. A germline-specific matα-Gal4VP16 driver was used to express UASp-based transgenes in oogenesis.

Isogenized w; P{neoFRT}40A males were starved for 12 hours in an empty vial. These flies were transferred into a vial containing filter paper soaked with 1% sucrose solution containing 25 mM ethyl methanesulfonate (EMS) (Sigma) and kept for 16-20 hours at 25°C. We obtained an average of 1.4 lethal hits per chromosome 2. EMS-exposed flies were mated with yw hs-FLP; Gla/CyO females. Single male progenies were crossed to w P{w⁺; Pvas-egfp::vas}; P{w⁺; Ubi-GFP(S65T)nls}2L P{neoFRT}40Afemales. To obtain germline clones (GLCs), third-instar larvae from this cross were heat-shocked at 37°C for 2 hours on two successive days. Subsequently, we hand-dissected ovaries from three to five females of genotype y w hs-FLP/w P{w⁺; Pvas-egfp::vas}; ^* P{neoFRT}40A/P{w⁺;Ubi-GFP(S65T)nls}2L P{neoFRT}40A (where the asterisk represents the mutation) and tested them for GFP-Vas localization in the GLCs that were marked by the lack of nuclear GFP. The lines that showed defects in GFP-Vas localization were recovered by crossing males in the same vials to w; Gla/CyO females to obtain balanced stocks (w; ^* P{neoFRT}40A/CyO).

Chromosomal mapping for mutations was carried out as described(Berger et al., 2001), using PCR product length polymorphisms and restriction fragment length polymorphisms between the P{neoFRT}40A and P{EP} chromosomes. After defining the mutated position within several hundred kb by deficiency mapping,the annotated genes in the region were PCR amplified from w; ^* P{neoFRT}40A/P{neoFRT}40A flies, and the products were direct-sequenced.

To rescue the rbsn-5^C241 mutant, a 4.5 kb genomic fragment containing the rbsn-5 transcription unit and 2.3 and 0.3 kb,respectively, of the 5′ and 3′ genomic flanking regions was cloned into pCaSpeR4. To express Osk at the anterior pole of the oocyte, the entire osk ORF sequence was joined to a modified bcd 3′ UTR,in which the putative NRE sequence (a 45 bp HpaI-EcoRV fragment) was replaced by a 10 bp SpeI linker(Wharton and Struhl, 1991) and cloned into the pUASp vector. To express the long or short Osk isoform specifically, an ATG triplet corresponding to the start codon for the unwanted isoform was changed to CTG by site-directed mutagenesis [M1L for short Osk and M139L for long Osk (Markussen et al.,1995)]. P-element-mediated germline transformation was carried out by standard methods.

A full-length rbsn-5 ORF was cloned into pProExHTa (Gibco) to produce a 6×His-tagged protein. The fusion proteins were purified with Ni-NTA agarose (Qiagen) followed by a disc preparative gel electrophoresis and used to raise rabbit and rat antibodies. The polyclonal rabbit antibodies against Rab5, Rab11 and Rab7 were generated and purified using the synthetic peptides SGTGTAQRPNGTSQNKSC (amino acid residues 11-28 of Rab5),CSQKQIRDPPEGDVIR (amino acid residues 177-191 of Rab11) and CNDFPDQITLGSQNNRPG(amino acid residues 184-200 of Rab7), respectively. Specificities of these antibodies were evaluated by immunoblotting and immunostaining(Fig. 1D and see Fig. S1 in the supplementary material).

Hand-dissected ovaries from 50 females were homogenized in ice-cold lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, pH 7.5). Total ovarian protein (50 μg) was run on SDS-PAGE and transferred onto an Immobilon-P membrane (Millipore). The signals were detected using the ECL system (GE Healthcare).

The full-length rbsn-5 ORF was PCR-amplified and cloned into pTNT(Promega). Recombinant proteins were synthesized in vitro using the TNT Quick Coupled Transcription/Translation system (Promega) in the presence of cold amino acids. The full-length Rab5, Rab7 and Rab11 ORFs were PCR-amplified and cloned into pGEX-5X1 (Pharmacia) to generate glutathione S-transferase (GST) fusion proteins, which were expressed in Escherichia coli and purified with glutathione resin (Clontech). Four micrograms of each purified GST-Rab protein was mixed with 40 μl of 50% glutathione resin and incubated for 1 hour at room temperature in PBS containing 0.1% Tween 20. The GDP or GTP-bound form of each GST-Rab protein was prepared as described(Christoforidis and Zerial,2000). In-vitro translated Rbsn-5 protein (2 μl) was incubated with GDP- or GTP-bound GST-Rab resins for 1 hour at 4°C under rotation in nucleotide stabilization (NS) buffer(Christoforidis and Zerial,2000) containing 10 μM GDP or GTPγS. The resins were washed extensively with NS buffer. Rbsn-5 protein bound to the resins was detected by immunoblotting.

Immunostaining of ovaries was performed by standard procedures. The following primary antibodies were used: rabbit anti-Stau (1:3000; a gift of D. St Johnston, The Gurdon Institute, Cambridge, UK), guinea pig anti-Osk(1:3000; laboratory stock), rabbit anti-Osk (1:8000; a gift of A. Ephrussi,EMBL, Heidelberg, Germany), rabbit and rat anti-Rbsn-5 (1:5000 and 1:2000,respectively), rabbit anti-Rab5 (1:1000), rabbit anti-Rab11 (1:8000), rabbit anti-Rab7 (1:3000), mouse anti-KDEL (1:25; StressGen), mouse anti-Golgi(1:100; Calbiochem), mouse anti-β-gal (1:5000; Promega), rabbit and rat anti-Vas (1:2500 and 1:1500, respectively; laboratory stock), and rabbit anti-Tud (1:2000) (Amikura et al.,2001). Alexa-conjugated secondary antibodies (1:1000; Invitrogen)were used. Ovaries were counterstained with 4 U/ml Alexa-660-conjugated phalloidin to label F-actin and/or with 1 μg/ml DAPI to label nuclei. All images were captured by a laser confocal microscope (Leica TCS SP2 AOBS) using a 63× PL APO water-immersion lens (N.A. 1.2) and processed with Adobe Photoshop.

RNA in situ hybridization was performed using standard procedures. RNA probes were synthesized in the presence of digoxigenin (DIG)- or fluorescein-UTP. DIG-labeled probes were detected by mouse anti-DIG (1:400;Roche) followed by Alexa-568-conjugated anti-mouse IgG (1:1000). Signal amplification of the fluorescein-labeled probes was performed using Alexa-488-conjugated anti-FITC (1:1000; Invitrogen).

The FM4-64 incorporation assay was performed as described(Sommer et al., 2005). Briefly, ovaries were dissected in Drosophila-SFM (Invitrogen) containing 10μM FM4-64 dye (Invitrogen) and kept for 30 minutes at 25°C. Ovaries that incorporated the dye were back-extracted by washing them twice for 15 minutes each in Drosophila-SFM. Immediately after being mounted on slide glass, the ovaries were examined under a confocal microscope.

Vas is a reliable marker for the germline throughout Drosophiladevelopment (Mahowald, 2001). A GFP-Vas fusion protein enables the direct visualization of the pole plasm and germ cells in the living organism(Sano et al., 2002). During oogenesis, GFP-Vas accumulates at the oocyte posterior from stage 9 onward. Using GFP-Vas as a marker, we performed a germline clonal screen targeting chromosome 2L for mutations that disrupted pole plasm assembly (our unpublished data). From 5122 lines mutagenized with EMS, we isolated 66 mutants defective in GFP-Vas localization. Twenty-seven of these were alleles of cappuccino, spire or profilin (chickadee), three genes on 2L that are known to be involved in osk RNA localization(Steinhauer and Kalderon,2006), which validated our screening strategy.

Among the other mutants recovered (Fig. 1A,B) was a recessive lethal mutation, C241, that mapped to 28C2-29E2. Subsequent deficiency mapping and sequencing of the mutant chromosome revealed that the C241 mutation was a single nucleotide substitution in the CG8506 gene, which resulted in a premature stop codon at position 315 of the 505 amino acid open reading frame (ORF)(Fig. 1C, asterisk). The introduction of a transgene containing a genomic DNA fragment with the CG8506 transcriptional unit rescued the C241 mutant phenotypes (described below). These data show that CG8506 corresponds to the gene that was mutated at the C241 locus. Rabbit and rat polyclonal antisera raised against full-length CG8506 did not detect a truncated form of CG8506 in ovarian extracts from C241 heterozygotes(Fig. 1D). Furthermore, neither antibody showed immunoreactivity in C241 homozygous clones (see Fig. S1A-D in the supplementary material), suggesting that the truncated protein was not expressed at detectable levels and/or was unstable. Therefore, C241 appeared to be a strong loss-of-function, presumably a protein-null, allele of CG8506.

CG8506 (Rabenosyn - FlyBase) encodes a protein homologous to Rabenosyn-5 (Rbsn-5) (Fig. 1C) (Nielsen et al.,2000). Rbsn-5 interacts with several Rab proteins, including Rab5,which functions in early endosomal transport(de Renzis et al., 2002; Eathiraj et al., 2005). Several Rbsn-5 protein domains are conserved across species, including the FYVE domain, which binds phosphatidylinositol-3-phosphate(Nielsen et al., 2000). However, invertebrate Rbsn-5 homologs lack the C-terminal domain common to the mammalian homologs of this protein (Fig. 1C). As the C-terminal domain of mammalian Rbsn-5 is responsible for its interaction with Rab5 (de Renzis et al., 2002; Eathiraj et al.,2005), we examined whether CG8506 interacted with Rab5. Pull-down assays showed that GST-Rab5 efficiently pulled down in-vitro-synthesized CG8506 protein in the presence of a GTP analog, GTP-γS, but inefficiently in the presence of GDP (Fig. 1E). The interaction between CG8506 and Rab5-GTP was specific,because the interactions of CG8506 with Rab11 and Rab7 were at background levels (Fig. 1E). Consistent with a physical interaction between CG8506 and Rab5 in vitro, in CG8506^C241 GLCs, neither auto-fluorescent granules derived from endocytosed yolk proteins nor the incorporation of a fluorescent marker for endocytosis, FM4-64, were observed in the oocytes(Fig. 1F-I), suggesting that CG8506 functions cooperatively with Rab5 in the early endocytic pathway. Thus,CG8506 is the Drosophila ortholog of Rbsn-5 and has an evolutionarily conserved function in the endocytic pathway.

In rbsn-5^- GLCs, little or no GFP-Vas was detected at the posterior pole of the stage-10 oocyte(Fig. 1B). The posterior accumulation of Vas depends on the proper localization and translation of osk RNA (Mahowald,2001). We therefore examined whether osk RNA localization was affected in the rbsn-5^- oocyte. In wild-type oocytes,the Staufen (Stau) protein (St Johnston et al., 1991), a marker for osk RNP, accumulates at the posterior region of the oocyte until stage 6(Fig. 2A). During stages 7-8,Stau becomes highly concentrated in the center of the oocyte. Stau is then transported to, and remains at, the posterior pole of the oocyte until oogenesis completes. The distribution patterns of Stau in rbsn-5^- GLCs were normal until stage 8(Fig. 2C). However, Stau was not tightly localized to the posterior cortex in the stage-9 rbsn-5^- oocyte, resulting in its ectopic aggregation in late-stage oocytes (Fig. 2C). As a consequence, Osk was translated very weakly at the posterior pole in rbsn-5^- oocytes (Fig. 2D). These defects were fully rescued by the introduction of the rbsn-5 genomic rescue construct(Fig. 2E,F). The defects in osk RNP localization in the rbsn-5^- GLCs were confirmed by in situ hybridization analysis of osk RNA(Fig. 2G-J). By contrast, the rbsn-5^- oocytes showed normal localization of both bcd and grk RNAs at the anterior and anterior-dorsal corner,respectively (Fig. 2G-J,arrowheads). These results indicate that rbsn-5 is specifically required for the stable accumulation of osk RNP at the posterior cortex of the oocyte.

The transport of osk RNA to the oocyte posterior depends on the MT cytoskeleton, which is thought to align along the AP axis during stages 8-10(Steinhauer and Kalderon,2006). We therefore examined whether MT polarity was affected by the loss of Rbsn-5 function. For this purpose, we used an MT plus-end marker,Kinesin-β-galactosidase (Kin-βgal) fusion protein(Clark et al., 1994), which is normally concentrated at the posterior of stage 8-10 oocytes(Fig. 2K,L). At stage 8, no difference in the Kin-βgal distribution was observed between rbsn-5^- GLC and wild-type oocytes(Fig. 2K,M), indicating that the MT polarity was initially established in the absence of Rbsn-5. However,in the rbsn-5^- oocyte, the posterior localization of Kin-βgal was lost from stage 9 onward. Instead, Kin-βgal, like Stau,formed aggregates in the oocyte cytoplasm(Fig. 2N). These results indicate that rbsn-5 is required for the maintenance of the MT polarity.

Although the above results indicated that defects in osk RNA localization in rbsn-5^- oocytes are primarily attributable to the failure to maintain MT polarity, the distribution of Rbsn-5 within developing oocytes prompted us to investigate additional roles for Rbsn-5 in pole plasm assembly. Double staining for Rbsn-5 and Stau revealed that the distribution patterns of Rbsn-5 were very similar to those of osk RNP(Fig. 3A,B). Although immunolabeling for Rbsn-5 was observed on the entire oocyte cortex, it was concentrated at the center of the oocytes at stage 7, when Stau is also enriched in this region (Fig. 3A). Subsequently, Rbsn-5 became highly concentrated at the posterior pole from stage 8 onward (Fig. 3B). At the posterior pole of the oocyte, Rbsn-5 signals showed nearly complete co-localization with Stau.

To characterize the Rbsn-5-positive structures in detail, the localization of Rbsn-5 was compared with several marker proteins for specific types of endosomes. We found that all the endosomes that were labeled for Rab5 (an early endosome marker), Rab11 (recycling endosomes) or Rab7 (late endosomes)showed polarized distributions that were very similar to those of Rbsn-5 and Stau. Although all the endosomal proteins analyzed were distributed along the entire cortex within the oocyte, they transiently accumulated at the center of stage 7 oocytes and later at the posterior pole of the oocyte(Fig. 3C-H). At the center of stage 7 oocytes, Rbsn-5 was co-localized with Rab5(Fig. 3C), consistent with their interaction in vitro (Fig. 1E). These polarized distributions were specific to endosomal proteins, because other membranous organelles, such as the endoplasmic reticulum (ER), marked by the KDEL epitope, and the Golgi apparatus, marked by gp120 (Stanley et al., 1997),did not show polarized distribution in the oocytes(Fig. 3I-L)(Lee and Cooley, 2007). The posterior accumulation of endosomal proteins correlated well with the preferential uptake of FM4-64 dye at the oocyte posterior(Fig. 1H)(Vanzo et al., 2007). These results demonstrate that endosomal proteins and endocytic activity are highly polarized toward the posterior pole of the oocyte.

To test whether the posterior accumulation of endosomal proteins is governed by osk activity, we examined the localization of Rbsn-5 and Rab11 in osk mutant oocytes. Wild-type oocytes exhibited a clear posterior accumulation of Rbsn-5 and Rab11 during stages 8-10(Fig. 4A,D). In the osk protein-null mutant, Rbsn-5 and Rab11 were initially concentrated at the oocyte posterior, and FM4-64 was preferentially endocytosed at the oocyte posterior at stage 8 (Fig. 4B). However, in agreement with published observations(Dollar et al., 2002; Vanzo et al., 2007), Rbsn-5,Rab11 and FM4-64 uptake lost their polarized distribution in stage 10 oocytes and were instead distributed uniformly on the oocyte cortex(Fig. 4B,E). The same defects were observed in an osk mis-sense mutant (data not shown). These results demonstrate that Osk activity is required for the maintenance, but not the establishment, of polarized endocytosis.

Although the initial polarization of endocytosis was independent of Osk activity, it does depend on Grk, which is required for the establishment of the AP polarity within the oocyte(Steinhauer and Kalderon,2006). In grk mutants, about 90% of stage 8-10 oocytes showed no posterior accumulation of Rbsn-5 or Rab11, and the remaining 10%showed faint posterior signals for these proteins(Fig. 4F). The polarized incorporation of FM4-64 was also lost in grk mutant oocytes(Fig. 4C). These results indicate that the posterior accumulation of endosomal proteins and asymmetric endocytic activity require the determination of oocyte polarity.

Our observations apparently contradict the previous finding that Rab11 remains accumulated at the posterior in grk mutant oocytes(Dollar et al., 2002). However, we noticed that signals for endosomal proteins at the anterior cortex in wild-type oocytes were somewhat higher than those at the lateral cortex(e.g. Fig. 4A), suggesting that endosomal proteins may accumulate, to some extent, at the MT minus ends as well as the MT plus ends. In grk mutant oocytes, MT minus ends are found at both poles (Steinhauer and Kalderon, 2006). It is therefore likely that the weak posterior accumulation of Rab11 in 10% of the grk mutant oocytes is due to accumulation at the MT minus ends.

To examine the roles of Osk in the polarization of endosomal protein localization and endocytosis, we ectopically expressed Osk at the anterior pole of the oocyte. Expression of an osk-bcd 3′ UTR transgene, in which the osk ORF is fused to the bcd 3′ UTR that contains the RNA localization signal for anterior accumulation, promotes the anterior misexpression of Osk in the oocyte and ectopic pole plasm assembly(Ephrussi and Lehmann, 1992). In oocytes expressing the osk-bcd 3′ UTRtransgene, we found that all the endosomal proteins analyzed were recruited to the oocyte anterior (Fig. 5A,Band see Fig. S2A,C,E in the supplementary material). Consistent with the ectopic localization of endosomal proteins, the anterior misexpression of Osk promoted increased uptake of FM4-64 at the oocyte anterior (see Fig. S2G in the supplementary material). These results indicate that the ectopic Osk can recruit endosomal proteins and stimulate endocytosis.

We next examined whether one of the Osk isoforms or both were required for the recruitment of endosomal proteins and the stimulation of endocytosis. When short Osk alone was misexpressed at the oocyte anterior, endosomal proteins did not accumulate there (Fig. 5C and see Fig. S2B,D,F in the supplementary material). By contrast, oocytes expressing a mutant version of the osk-bcd 3′ UTR transgene that produces long Osk alone exhibited obvious accumulation of endosomal proteins at the anterior pole(Fig. 5D-J). Increased FM4-64 uptake at the anterior region of the oocyte was also observed when the long Osk was ectopically expressed (Fig. 5L). The anterior accumulation of endosomal protein caused by ectopic long Osk was not due to the alternation of the MT polarity within the oocyte, as Kin-βgal never concentrated at the anterior pole in the osk-bcd 3′ UTR oocytes(Fig. 5N and see Fig. S2I,J in the supplementary material). These results demonstrate that long Osk alone is sufficient for recruiting endosomal proteins and stimulating endocytosis, even ectopically, without altering the MT polarity.

The above results suggested that the endocytic pathway functions downstream of long Osk in the assembly of the pole plasm. As long Osk has been implicated in retaining pole plasm components at the posterior cortex(Vanzo and Ephrussi, 2002), we supposed that the endocytic pathway might play a role in the anchoring process. However, rbsn-5 is essential for the maintenance of MT polarity (Fig. 2N), making it impossible to examine the involvement of rbsn-5 in the anchoring process during normal pole plasm assembly, which depends on the polarized MT array. By contrast, the anterior localization of bcd RNA was unaffected in rbsn-5^- oocytes(Fig. 2J). We therefore used the osk-bcd 3′ UTR transgene to direct ectopic pole plasm assembly at the anterior pole of the oocyte, and examined the effects in the rbsn-5^- oocytes.

Anteriorly misexpressed Osk was tightly anchored at the cortex in otherwise wild-type oocytes (Fig. 6A,B). However, in the rbsn-5^- oocytes expressing the osk-bcd 3′ UTR transgene, the anterior Osk was not tightly associated with the cortex, but diffused into the cytoplasm, where it formed aggregates (Fig. 6D,arrowheads). Other pole plasm components, such as Vas and Tud, also dispersed into the cytoplasm of the rbsn-5^- oocyte(Fig. 6E-H and data not shown). These results demonstrate that Rbsn-5 plays a crucial role in retaining pole plasm components at the oocyte cortex.

The anchoring of pole plasm components to the posterior cortex depends on the actin cytoskeleton (Jankovics et al.,2002; Polesello et al.,2002; Babu et al.,2004). We next examined whether the F-actin organization was affected in rbsn-5^- oocytes. Osk promotes long F-actin projections from the posterior pole of the oocyte(Vanzo et al., 2007). We found that anterior misexpression of Osk also induced long F-actin projections emanating from cortical actin bundles (Fig. 6A,B). However, in the rbsn-5^- oocytes, the anterior misexpression of Osk induced large aberrant F-actin aggregates, in which pole plasm components, such as Osk, Vas and Tud, were often engulfed(Fig. 6D,H and data not shown). We never observed these F-actin aggregates in rbsn-5^-oocytes that did not express osk-bcd 3′ UTR(Fig. 6C). These results indicate that Rbsn-5 is involved in the Osk-mediated F-actin reorganization.

Intriguingly, endosomal proteins were still recruited at the anterior in rbsn-5^- oocytes expressing the osk-bcd 3′ UTR transgene, although they were accumulated in a diffused manner(Fig. 6J-N). Thus, Rbsn-5, and hence the early endocytic pathway, is dispensable for the recruitment of endosomal proteins by Osk. These results further suggest that the accumulation of endosomal proteins is insufficient for the proper regulation of F-actin dynamics by Osk.

Once pole plasm components are transported to the oocyte posterior, they must be retained in place to ensure proper assembly of the pole plasm and their inheritance by the pole cells. Despite the known importance of long Osk for anchoring pole plasm components to the posterior cortex(Vanzo and Ephrussi, 2002),the underlying mechanism has been elusive. A recent report showed that long Osk promotes polarized endocytosis within the oocyte(Vanzo et al., 2007). Here we show that Osk maintains, but does not establish, the posterior accumulation of endosomal proteins and asymmetric endocytosis(Fig. 4), and that Osk can recruit endosomal proteins and stimulate endocytosis even at an ectopic site(Fig. 5). We further show that the anchoring of the pole plasm components to the oocyte cortex requires the Osk-dependent stimulation of endocytic activity(Fig. 6). These data reveal an interdependent relationship between Osk anchoring and localized endocytic activity at the oocyte posterior (Fig. 7).

In rbsn-5^- oocytes, the anterior misexpression of Osk induces aberrant F-actin aggregates, which diffuse along with pole plasm components into the cytoplasm (Fig. 6A-H). Several lines of evidence suggest that the anchoring of pole plasm components requires the proper organization of F-actin(Jankovics et al., 2002; Polesello et al., 2002; Babu et al., 2004). As endosomal proteins are recruited by long Osk(Fig. 5A-J), we favor the idea that the endocytic pathway functions downstream of long Osk to anchor the pole plasm components at the cortex by regulating F-actin dynamics. Supporting this idea, in addition to its roles in early endosomal sorting, Rab5 acts as a signaling molecule that remodels F-actin networks(Lanzetti et al., 2004). Rab11, which regulates the recycling of endosomes, is also involved in F-actin organization during cellularization in Drosophila blastoderm embryos(Riggs et al., 2003). Intriguingly, the recruitment of endosomal proteins by Osk is not sufficient for proper F-actin reorganization to anchor the pole plasm components at the cortex, because their recruitment occurs even in oocytes lacking Rbsn-5, in which cortical anchoring fails (Fig. 6). We therefore propose that the continuous cycling of endosomes is required for pole plasm components to be anchored to the oocyte cortex. This scenario is compatible with a model in yeasts, which use endocytic cycling coupled with localized exocytosis to maintain their polarity(Valdez-Taubas and Pelham,2003), although it is unclear if F-actin reorganization is involved in this process.

Rbsn-5 is primarily required for the maintenance of MT polarity that directs posterior localization of osk RNA(Fig. 2). Rab11 is also required for MT polarization in the oocyte(Jankovics et al., 2001; Dollar et al., 2002). However,the accumulation of endosomal proteins and upregulation of endocytosis at the oocyte posterior require the oocyte polarization, which promotes the reorganization of the MT array (Fig. 4). Thus, MT polarization and asymmetric activation of the endocytic pathway are probably interdependent as well. Furthermore,maintenance of polarized endocytic activity depends on Osk (Figs 4 and 5). Intriguingly, Osk is also thought to maintain MT polarity, as posterior accumulation of Kin-βgal is partially defective in the absence of Osk(Zimyanin et al., 2007). It is therefore likely that the endocytic pathway and Osk form a positive-feedback loop that maintains oocyte polarity: Osk may maintain MT polarity through recruiting endosomal proteins. Based on these results, we propose a model, in which the endocytic pathway is involved in several distinct steps in pole plasm assembly (Fig. 7).

The localization of bcd RNA to the anterior pole of the oocyte requires the ESCRT-II (endosomal sorting complex required for transport II)complex, which sorts mono-ubiquitinated endosomal transmembrane proteins into multivesicular bodies (Irion and St Johnston, 2007). Furthermore, Vps36p, a component of the ESCRT-II complex, binds bcd 3′ UTR in vitro and co-localizes with bcd RNA at the oocyte anterior, suggesting the direct involvement of ESCRT-II in bcd RNA localization. osk RNA,however, appears to use another mechanism for its posterior localization, as its localization is unaffected in the absence of ESCRT-II function(Irion and St Johnston, 2007). Several lines of evidence suggest that ER organization and RNA localization are linked (Schmid et al.,2006; Deshler et al.,1997; Chang et al.,2004; Wickham et al.,1999; Gautrey et al.,2005). However, we consider it unlikely that the ER directs the posterior localization of osk RNA, because ER components and osk RNP distributed differentially in developing oocytes(Fig. 3I,J). Interestingly, the osk RNP and the endosomal proteins were in close proximity during their transport to the oocyte posterior(Fig. 3A-H). Although their close association may simply be owing to the dynamic rearrangements of the MT array during stages 7-8 (Steinhauer and Kalderon, 2006), these findings suggest that the endocytic pathway may also play a role in the targeting of osk RNP to the posterior pole of the oocyte (Fig. 7). Retroviral genomic RNAs are known to hitchhike on endosomal vesicles to reach the plasma membrane (Basyuk et al.,2003). Therefore, it will be interesting to learn if oskRNA is also transported to the posterior pole of the oocyte along with the endosomes.

We thank C. Nakamoto for technical assistance, A. Ephrussi and D. St Johnston for reagents, A. Guichet and the Bloomington DrosophilaStock Center for fly stocks, and M. Abe, K. Hanyu-Nakamura, S. Ito, K. Kawahashi, T. Otani and K. Takizawa for comments on the manuscript. This work was supported in part by a Grant-in-Aid from MEXT and JSPS, and the RIKEN President Discretionary Fund.