Embryonic patterning in Drosophila is regulated by maternal factors. Many such factors become localized as mRNAs within the oocyte during oogenesis and are translated in a spatio-temporally regulated manner. These processes are controlled by trans-acting proteins, which bind to the target RNAs to form a ribonucleoprotein (RNP) complex. We report that a DEAD-box protein, Me31B, forms a cytoplasmic RNP complex with oocyte-localizing RNAs and Exuperantia, a protein involved in RNA localization. During early oogenesis, loss of Me31B causes premature translation of oocyte-localizing RNAs within nurse cells, without affecting their transport to the oocyte. These results suggest that Me31B mediates translational silencing of RNAs during their transport to the oocyte. Our data provide evidence that RNA transport and translational control are linked through the assembly of RNP complex.
INTRODUCTION
Cell polarity is often generated by asymmetric distribution of mRNAs (St Johnston, 1995; Bashirullah et al., 1998). Such mRNAs act as the localized source for translation, and allow specific proteins to be synthesized in the subcellular region where they are required. As these mRNAs are synthesized in the nucleus and travel through the cytoplasm to their final destinations, their translation must be silenced during their localization.
Drosophila oogenesis provides an excellent model system with which to study translational regulation of localized mRNA. The Drosophila egg chamber is made up of 16 germline cells surrounded by a monolayer of somatic follicle cells (Spradling, 1993). The 16 cells are generated from a single germline cell by four successive mitotic divisions. As cytokinesis is incomplete during these divisions, the cells remain connected to each other through specialized cytoplasmic bridges called ring canals. The oocyte is determined from one of 16 germline cells, while the remaining 15 cells differentiate into nurse cells. Nurse cells synthesize large amounts of RNAs and proteins, which are transported to the developing oocyte through ring canals.
The mRNAs encoding determinants for embryonic polarity such as bicoid (bcd), oskar (osk) and nanos (nos) are synthesized in nurse cells, transported to the oocyte and localized within the oocyte during oogenesis. Translation of these mRNAs is silenced during their localization, and is activated when and where the protein is required. It has been reported that premature or ectopic translation of these mRNAs causes severe developmental defects (Ephrussi and Lehmann, 1992; Gavis and Lehmann, 1992; Smith et al., 1992; Kim-Ha et al., 1995; Dahanukar and Wharton, 1996; Gavis et al., 1996). For example, translation of osk mRNA is repressed during its posterior localization in the oocyte by an RNA-binding protein, Bruno, which binds to a repeated sequence in the 3′ untranslated region (3′UTR), or Bruno response elements (BREs) (Kim-Ha et al., 1995; Webster et al., 1997). When BRE-mutated osk mRNA (oskBRE–) is expressed in oogenesis, the Osk protein is produced prematurely in oocytes (Kim-Ha et al., 1995). The premature translation of oskBRE– mRNA leads to a maternal-effect lethal phenotype. Similarly, translation of unlocalized nos mRNA in embryos is repressed by Smaug protein, which binds to the regulatory element in the 3′UTR. Ectopic translation of unlocalized nos mRNA causes a defect in anteroposterior axis formation (Dahanukar and Wharton, 1996; Gavis et al., 1996; Smibert et al., 1996, 1999; Dahanukar et al., 1999). Thus, localization-coupled translational control of these mRNAs is an essential mechanism for embryonic patterning.
Although the translational repression of unlocalized mRNAs in oocytes and embryos has been well analyzed, little is known as to how these mRNAs are translationally silenced during their transport from nurse cells to the oocyte. In egg chambers that express oskBRE− mRNA, Osk protein is present throughout the oocyte, yet is undetectable in nurse cells (Kim-Ha et al., 1995). Furthermore, additional introduction of a mutation in the osk 5′UTR causes a more pronounced premature translation of a reporter gene within nurse cells than BRE− alone (Gunkel et al., 1998). This suggests that BRE-mediated translational repression is not the only mechanism to prevent premature translation of osk mRNA.
Several elegant studies in diverse organisms have shown that localized RNAs are organized into particles, which move through the cytoplasm in a microtubule-dependent manner (Ainger et al., 1993; Ferrandon et al., 1994; Bertrand et al., 1998). This suggests that localization of RNAs are controlled by trans-acting proteins, which bind these RNAs to form a ribonucleoprotein (RNP) complex. One of the candidates for such trans-acting proteins in Drosophila oogenesis is Exuperantia (Exu). Green fluorescent protein (GFP)-tagged Exu protein forms cytoplasmic particles, which move to the oocyte in a microtubule-dependent manner (Wang and Hazelrigg, 1994; Theurkauf and Hazelrigg, 1998). Genetic studies have shown that Exu is essential for the initial step of bcd mRNA localization to the anterior of the oocyte (St Johnston et al., 1989). A recent study has further shown that Exu is also required for proper localization of osk mRNA at the posterior pole of the oocyte (Wilhelm et al., 2000). Exu is highly concentrated in distinctive RNA-rich regions in the cytoplasm of nurse cells and oocytes, called the sponge bodies (Wilsch-Bräuninger et al., 1997). These observations have led to the proposal that Exu-containing particles are specialized RNP complexes that are involved in the transport of maternal RNAs to the oocyte.
Similarly, trans-acting proteins for translational control of maternal mRNAs may be assembled into a cytoplasmic RNP complex. This complex could be visualized, like Exu, as particles using light microscopy. To isolate such proteins, we conducted a visual screen with a GFP-cDNA expression library. Here, we report that a putative RNA-binding protein, Maternal expression at 31B (Me31B; De Valoir, 1991), forms cytoplasmic particles during oogenesis. Me31B also forms a complex with Exu in an RNA-dependent manner, and is concentrated in the sponge bodies. Furthermore, we show that several oocyte-localizing RNAs are colocalized with Me31B during their transport from nurse cells to the oocyte. In early egg chambers that lack Me31B, at least two mRNAs in the particles, osk and Bicaudal-D (BicD) mRNAs, are prematurely translated in nurse cells, though the transport of these RNAs to the oocyte is Me31B independent. Our data suggest that Me31B plays an essential role in translational silencing of oocyte-localizing mRNAs during their transport to the oocyte.
MATERIALS AND METHODS
Drosophila stocks
Fly stocks were raised at 25°C on standard cornmeal and agar medium. P[w+ gfp::exu] line (Wang and Hazelrigg, 1994) was a gift from Dr P. Lasko (McGill University, Montreal, Canada). P[w+mc=lacW]l(2)k06607k06607/CyO (Spradling et al., 1999), hsFLP; CyO/Sco, FRT40A; ry, and ovoD-2L FRT40A/CyO lines were obtained from Bloomington stock center.
Construction of GFP-cDNA fusion library
A germline-specific vas promoter is located within ∼2 kb upstream of the vas transcription start site (A. N. and P. Lasko, unpublished). This genomic region was isolated from λ phage clone v3.04 (Lasko and Ashburner 1988). The P[w+ Pvas-gfp] plasmid was constructed in pCaSpeR2 (Thummel and Pirrotta, 1992) by multiple subcloning steps, and contains the vas promoter, 5′ UTR and first 16 amino acids coding region of vas, enhanced green fluorescent protein (egfp, Clontech), vas 3′ UTR and the subsequent ∼550 bp of the 3′ flanking region. This vector has unique XhoI and NotI sites between the 3′ end of egfp and the termination codon. Detailed construction procedure will be provided upon request.
Ovaries from y w females were used as the RNA source for library construction. Poly(A)+ RNA (5 μg) was used to generate cDNA with Superscript Plasmid System kit (Gibco BRL). This procedure yielded cDNA fragments with a SalI adapter at the 5′ end and oligo-dT/dA-NotI at the 3′ side. NotI digested cDNAs were ligated into XhoI/NotI-digested P[w+ Pvas-gfp] plasmid. An unamplified cDNA library was generated from 2.2×106 independent clones. Of 168 random clones tested, all clones contained an insert and the average insert size was estimated to be 2.0 kb with a range of 0.1-6.0 kb.
P element-mediated germline transformation was performed according to a standard method (Spradling and Rubin,1982), using y w embryos as recipients. Ovaries from transformed females were hand-dissected, fixed in 4% paraformaldehyde in PBS (7 mM Na2HPO4, 3 mM NaH2PO4, 130 mM NaCl) and mounted in 50% glycerol in PBS. Spatial distributions of GFP signals in ovaries were observed under a laser confocal microscope (Leica TCS-NT) with an Ar/Kr laser unit. The cDNA insert of the fusion gene was amplified by PCR from genomic DNA of the transformant, and the nucleotide sequence was determined using PCR product as template.
Antibody generation
The N-terminal half (Met2-Gln294) of Me31B-coding region was subcloned into pProExHTa (Gibco BRL) to produce a His6-tagged Me31B protein. The coding region of Me31B cDNA was amplified by PCR with primers 5′-CCCCCATGGATATGACTGAAAAGTTAAATTCTGG-3′ and 5′-CCCCTCGAGTTATTTGCTAACGTTGCCCTCC-3′ using an EST clone LD27473 (Berkeley Drosophila Genome Project) as a template. Amplified cDNA fragment was then digested with NcoI and PstI, and ∼0.9 kb fragment was ligated into pProExHTa. Histidine tagged Me31B protein was expressed in Escherichia coli BL21 cells by IPTG induction. After centrifugation at 5000 g for 10 minutes, E. coli cells were treated with Bugbuster regent (Novagen) containing 100 μg/ml of lysozyme and 10 μg/ml of DNase I. Me31B protein was recovered as an insoluble form, solubilized with 6M guanidine-HCl in buffer A (100 mM sodium phosphate pH 8.0, 10 mM Tris-HCl, 0.5 M NaCl) and purified with Ni-NTA agarose resin (Qiagen). Purified protein was loaded on 10% SDS-polyacrylamide gels and protein bands were visualized with 0.2 M KCl on ice. Me31B band was cut and dialyzed extensively against water. About 1 mg of the purified protein was emulsified with Titer Max Gold adjuvant (CytRx) and injected into two rabbits. Boosting was carried out with the same amount of protein for 4 weeks intervals. Being boosted twice, antisera from two rabbits both recognized a single ∼50 kDa band on immunoblot of ovary extract. Antisera were further affinity purified with the protein immobilized on PVDF membrane (Millipore). The affinity-purified antibodies from antisera from two rabbits showed identical staining pattern in wild type ovaries.
Fluorescence in situ hybridization
Whole-mount in situ hybridization to ovaries was performed as described (Kobayashi et al., 1999) with several modifications. Hybridized signals were detected with mouse anti-digoxygenin monoclonal antibody (Roche), followed by Alexa568 conjugated anti-mouse IgG (Molecular Probes). The ovaries were double stained with an affinity-purified rabbit anti-Me31B antibody and Alexa488-conjugated anti-rabbit IgG. The ovaries were mounted in Vectashield (Vector Laboratories) and observed under a laser confocal microscope (Leica TCS SP2).
Immunostaining
Immunostaining of ovaries were carried out as described (Kobayashi et al., 1999). Primary antibodies and antisera used were mouse anti-BicD 1B11 (Suter and Steward, 1991), rabbit anti-Exu (Wilsch-Bräuninger et al., 1997), rabbit anti-GFP (Clontech), mouse anti-GFP 3E6 (Wako Pure Chemicals, Osaka, Japan), rabbit anti-Me31B, rabbit anti-Osk (Kim-Ha et al., 1995), and rabbit anti-Yps (Wilhelm et al., 2000). Secondary antibodies used were Alexa488-conjugated anti-rabbit IgG, Alexa568-conjugated anti-rabbit IgG, Alexa568-conjugated anti-mouse IgG (Molecular Probes), FITC-conjugated anti-mouse IgG (Jackson Immuno Research), biotinylated anti-rabbit IgG (Vector Laboratory) and 15 nm gold-conjugated anti-rabbit IgG (Biocell). To visualize cell boundaries and nuclei, ovaries were incubated with Texas Red-X phalloidin and Hoechst 33258 (Molecular Probes), respectively.
Immunoprecipitation of Me31B and Yps
Immunoprecipitation was performed as described by Wilheim et al. (Wilheim et al., 2000) with slight modifications. Affinity-purified antibodies (5 μg) or normal rabbit IgG (Santa Cruz Biotechnology) were crosslinked to ∼50 μl of protein G sepharose 4B Fast Flow beads (Sigma) with dimethyl pimelimidate at a concentration of 25 mM in 0.2 M triethanolamine pH 8.0. After the beads were washed in 0.2 M ethanolamine pH 8.0, they were pre-eluted with 0.2 M glycine-HCl pH 2.7. The beads were then equilibrated with PBS.
Ovaries from y w females were homogenized on ice in DXB-50 (25 mM Hepes-KOH pH 6.8, 250 mM sucrose, 1 mM MgCl2, 1 mM DTT, 50 mM KCl, 0.1% Triton X-100) containing complete protease inhibitor cocktail EDTA free (Roche), and centrifuged at 10000 g for 20 minutes. The supernatant was pre-absorbed twice in ∼100 μl of protein G sepharose beads for 1 hour at 4°C. The protein concentration of the extract was then adjusted to 10 mg/ml and 200 μl of the extract was immunoprecipitated with the antibody-coupled beads for overnight at 4°C on a rotator. To evaluate RNase sensitivity for the interaction between Me31B and Yps-Exu complex, 20 μl of 10 mg/ml RNase A was added during this incubation period. The beads were washed with DXB-50. Bound proteins were then eluted with 0.2 M glycine-HCl pH 2.7. The eluate was immediately neutralized with 1 M Tris-HCl pH 8.0, and residual beads were removed by passing through a spin column (Samprep C02-LH; Millipore). The samples were then precipitated with trichloroacetic acid. The precipitates were suspended in 40 μl of sample buffer and boiled for 5 minutes. 10 μl of denatured samples were run on a 10% SDS-polyacrylamide gel and transferred onto PVDF membrane (Millipore). The membrane was then probed with the primary antibodies. Proteins were detected by an ECL system (Amersham-Pharmacia Biotech).
Generation of me31B-null alleles
l(2)k06607 is a viable placw insertion at 31B (Spradling, 1999) but the chromosome carried separable lethal mutation(s). Homozygous viable and fertile lines were generated by meiotic recombination between chromosomes with and without k06607 insertion. Excision lines were generated by crossing virgin k06607 females with males possessing Δ2–3 transposase. From 360 independent excision events, 351 white-eyed balanced stocks were established. These were checked by PCR to determine whether they carried a deletion within me31B locus. Breakpoints of the three deletion lines were determined by a direct sequencing of the PCR products. To generate germline clones, FRT40A was introduced on me31BΔ1 and me31BΔ2 chromosomes. They were crossed with hsFLP; gfp-vas FRT40A or hsFLP; ovoD FRT40A/CyO males. To induce mitotic recombination in proliferating germ cells, progeny were heat shocked twice at 37°C for 2 hours during early pupal stage. Germline clones in hsFLP/w; me31BΔ FRT40A/vas-gfp FRT40A females were identified by loss of anti-Me31B or anti-GFP staining. The two me31B mutations, me31BΔ1 and me31BΔ2, showed identical phenotypes.
Transgenic constructs for a rescue of me31B mutants
A ∼13.7 kb genomic DNA containing an entire me31B locus was isolated from a Drosophila genomic DNA library in λFIXII (a gift from B. Suter, McGill University, Montreal, Canada). A 9 kb SalI fragment, which contains the entire me31B mRNA-coding region and ∼4.5 kb of upstream promoter region, was subcloned into pCaSpeR4 to yield the P[w+ me9k] plasmid. This me31B+ transgene and two deletion derivatives were used as substrates for P element-mediated transformation. At least three independent transformed lines were checked for their ability to rescue me31BΔ phenotype.
RESULTS
A DEAD-box protein, Me31B, forms cytoplasmic particles in germline cells during oogenesis
We conducted a visual screen with an ovarian GFP-cDNA library, in which fusion genes are expressed in germline cells during oogenesis. We generated transgenic flies with this library and identified proteins that distribute in a granular pattern during oogenesis. Screening ∼3000 independent lines, we isolated one, in which GFP signals were detected as cytoplasmic particles during oogenesis (Fig. 1A). The particles were dispersed in the cytoplasm of both nurse cells and oocytes but never detected within nuclei. We frequently observed the particles passing through ring canals (Fig. 1B), suggesting that the particles are assembled in nurse cell cytoplasm and transported to the oocyte.
The cDNA from this line was identified as me31B (De Valoir et al., 1991). In the cDNA fusion, almost the entire coding region of me31B, which lacks only the first four codons, was fused in frame with that of gfp. Me31B, a DEAD-box protein and therefore a putative ATP-dependent RNA helicase, has been isolated as a gene expressed extensively during oogenesis (De Valoir et al., 1991). Me31B is a part of an evolutionally conserved DEAD-box protein group, which includes human RCK/p54 (71% identical), Xenopus Xp54 (73%), Caenorhabditis elegans C07H6.5 (76%), Schizosaccharomyces pombe Ste13 (68%) and Saccharomyces cerevisiae Dhh1 (68%). Furthermore, Me31B is phylogenetically close to two evolutionally conserved proteins, eIF4A and Dbp5/Rat8p but far from Vasa, which functions in germline development (data not shown).
To examine distribution of the endogenous Me31B, we generated antibodies that specifically recognized Me31B. As shown in Fig. 1C-F, the distribution pattern of endogenous Me31B was identical to that of GFP-Me31B. No detectable signal in somatic follicle cells was observed at any stages of oogenesis. Me31B was first detected at a low level in germarium region 2B, where the signal was concentrated in the pro-oocytes (Fig. 1C). The signal remained concentrated in the oocyte until mid-oogenesis (Fig. 1D,E). In early egg chambers, a low level Me31B signal was detected in nurse cell cytoplasm. In both nurse cells and oocytes, the signal appeared to be granular. Me31B signals in nurse cell cytoplasm became more evident from stage 5-6, when Me31B expression was drastically increased (Fig. 1D,E). In addition, Me31B was frequently enriched around nurse cell nuclei. Later, Me31B accumulates at the posterior pole of stage 10 oocytes (Fig. 1F). However, this posterior accumulation was transient, as revealed by uniform distribution of the signal in cleavage embryos (Fig. 1G). By cellular blastoderm stage, Me31B became undetectable in the entire embryonic region (Fig. 1H). No zygotic expression of Me31B was detected during embryogenesis (data not shown).
Me31B colocalizes with oocyte-localizing RNAs
Because Me31B is probably an RNA-binding protein that is transported to the oocyte, we next asked whether Me31B formed a complex with oocyte-localizing RNAs. We first examined colocalization of osk mRNA with Me31B. As has been reported previously (Ephrussi et al., 1991; Kim-Ha et al., 1991), osk mRNA starts to accumulate in oocytes from germarium region 2B, with the concentration of osk increasing over time. Posterior accumulation of osk mRNA in the oocyte began from stage 8 onwards. By fluorescent in situ hybridization, osk mRNA exhibited particulate signals in the cytoplasm of both nurse cells and oocytes, and was frequently concentrated around nurse cell nuclei (Fig. 2B). This distribution pattern of osk mRNA was essentially identical to that of Me31B (Fig. 2C), with colocalization present until osk mRNA localized to the posterior pole of stage 10 oocytes (data not shown).
We also examined colocalization of Me31B with other RNAs. We double stained ovaries for Me31B and BicD mRNA. In early egg chambers, BicD mRNA also produced particulate signals, and appeared to localize in Me31B-containing particles (data not shown). This colocalization became apparent from stage 5-6, when BicD mRNA expression was elevated (Fig. 2D-F). The oocyte-localizing RNAs examined (bcd, nos, oo18 RNA-binding (orb), Polar granule component (Pgc) and germ cell-less (gcl)) all produced particulate signals in the cytoplasm of both nurse cells and oocytes, and colocalized with Me31B (data not shown). In contrast, vasa (vas) mRNA, which is not specifically transported to the oocyte (Hay et al., 1988; Lasko and Ashburner, 1988), did not appear to be colocalized with Me31B (Fig. 2G-I). These results indicate that Me31B forms cytoplasmic particles that contain oocyte-localizing RNAs.
Me31B is colocalized with Exu and Yps
We next examined whether Me31B and Exu were colocalized, because distribution of GFP-Exu fusion protein was very similar to that of Me31B (Wang and Hazelrigg, 1994). GFP-Exu was concentrated to the oocytes from as early as stage 1, and transiently accumulated at the posterior pole of stage 10 oocytes (Wang and Hazelrigg, 1994; data not shown). From stage 5-6 onwards, particulate GFP-Exu signals in the cytoplasm became apparent (Fig. 3A; Wang and Hazelrigg, 1994). To compare the distribution between Exu and Me31B directly, we immunostained ovaries expressing GFP-Exu with an anti-Me31B antibody. We found that almost all GFP-Exu particles contained Me31B (Fig. 3A-C). Colocalization between Exu and Me31B throughout oogenesis was further confirmed by immunostaining of ovaries expressing GFP-Me31B with an anti-Exu antiserum (Fig. 3D-F).
We next tested the colocalization of Me31B with Ypsilon Schachtel (Yps), which has been shown to be an Exu-binding protein in oogenesis (Wilhelm et al., 2000). In nurse cells and oocytes, Yps signals were observed in GFP-Me31B-positive cytoplasmic particles (Fig. 3G-I). These results indicate that Me31B, Exu and Yps are all components of a cytoplasmic RNP complex in nurse cells and oocytes.
Exu has been reported to be highly concentrated in a specialized RNA rich regions in the cytoplasm of nurse cells and oocytes, called the sponge bodies (Wilsch-Bräuninger et al., 1997). Sponge bodies have been proposed to be involved in the localization of RNAs in the oocytes. We examined whether Me31B was also concentrated in the sponge bodies using immunoelectron microscopy (Fig. 3J). Me31B signals were detected in distinctive cytoplasmic regions, consisting of string-like materials embedded in an electron dense mass. The regions were in close proximity to mitochondria but not surrounded by membrane. All of these features completely satisfied the reported criteria of the sponge body (Wilsch-Bräuninger et al., 1997). Therefore, we conclude that Me31B is highly enriched in the sponge bodies, and that the sponge body is the region where cytoplasmic RNPs for oocyte localization are concentrated.
Me31B forms a complex with Exu and Yps in an RNase-sensitive manner
To characterize Me31B-containing particles in detail, the particles were immunoprecipitated with an anti-Me31B antibody from extracts of wild-type ovaries. The resulting immunoprecipitates were separated by SDS-PAGE and probed with antibodies against Me31B, Yps and Exu. We also tested whether Me31B was coimmunoprecipitated with Yps. As shown in Fig. 3K, Yps and Exu were coimmunoprecipitated with Me31B, and Me31B was coimmunoprecipitated with Yps. Control experiments showed that none of these proteins was coimmunoprecipitated with a normal rabbit IgG.
It has been reported that Exu binds directly to Yps, and that these proteins coimmunoprecipitate several unknown proteins in an RNase-sensitive manner (Wilhelm et al., 2000). We therefore examined the effect of RNase treatment on the nature of the Me31B-Exu/Yps complex. In the presence of RNase, Me31B-Yps and Me31B-Exu, coimmunoprecipitation was diminished (Fig. 3K). These results indicate that Me31B forms a complex with Yps and Exu via RNA.
me31B is an essential gene
To investigate the function of Me31B during oogenesis, we identified me31B mutations. We found that a P element insertion, k06607, is located ∼100 bp upstream of the 5′ end of me31B (Fig. 4; Spradling et al., 1999). By mobilizing the k06607 insertion, we isolated three partial deletions of me31B (me31BΔ1, me31B Δ2 and me31B Δ3). All the deletions started from the original P element insertion point, and their other breakpoints were within the me31B-coding region (Fig. 4).
The three me31BΔ lines were all recessive lethal. They did not complement each other, and were also lethal in trans to a deficiency, Df(2L)J2, which deletes the entire me31B locus (De Valoir et al., 1991; data not shown). Homozygous me31BΔ individuals died during the second- or third-instar larval stages without expressing any discernible morphological defects. The lethality associated with the me31BΔ chromosome was rescued by introducing a genomic DNA (Pme9k), which contains the entire me31B locus and ∼4.5 kb of the upstream region (Fig. 4). A 3.7 kb AflII/AflII fragment (PmeAfl), including only ∼700 bp upstream and 12 bp downstream of the me31B mRNA-coding region, also complemented the me31B mutations. In contrast, Pme5k, which deletes most of me31B locus from Pme9k, failed to rescue the lethality. From these results, we conclude that the three me31BΔ mutations affect only me31B function, and that me31B is a vital gene.
Me31B is essential for oogenesis
To analyze the me31B function in oogenesis, we used the FLP-DFS mitotic recombination system (Chou and Perrimon, 1996) to produce me31B– germline clones. None of the egg chambers homozygous for me31B mutations completed oogenesis (Fig. 5). Most me31B– egg chambers degenerated during mid-oogenesis (Fig. 5D-F). In these egg chambers, cell membranes of germline cells were collapsed, and ring canals and nurse cell nuclei were concentrated to form large aggregates within egg chambers. Nurse cell nuclei in these egg chambers were then fragmented into small pieces. Occasionally, we found egg chambers that progressed as far as stage 10 (Fig. 5G,H). In these egg chambers, oocytes did not grow normally, and the nurse cell nuclei were displaced posteriorly relative to the anterior border of columnar follicle cells, which normally surround the oocyte at stage 10 (Fig. 5A-C).
Me31B is dispensable for the transport of the associated molecules to the oocytes
The complicated and redundant phenotypes we observed in me31B– egg chambers in mid-oogenesis are unlikely to be the primary effect of loss of me31B function. We examined earlier phenotypes of me31B– egg chambers using a FLP/FRT system to generate homozygous germline clones that are marked by the loss of Vas-GFP fusion protein (see Materials and Methods). Based on Hoechst and phalloidin staining, me31B– egg chambers were morphologically normal until stage 4-5. From stage 6 onwards, oocytes in me31B– egg chambers failed to grow normally. At this stage, these egg chambers began to degenerate. In early me31B– egg chambers, Exu signal was concentrated to the oocytes (Fig. 6B, arrowheads). We next examined distribution of osk and BicD mRNAs in me31B– egg chambers. As shown in Fig. 6C-F, both osk and BicD mRNAs also accumulated in the oocytes of me31B– egg chambers until the chambers degenerated (arrowheads). Particulate signals for these RNAs were detectable in nurse cell cytoplasm in these egg chambers. These results indicate that Exu, osk and BicD mRNAs can be transported to the oocyte even in the absence of Me31B. We conclude that in early egg chambers, Me31B is dispensable for the transport of the molecules that form a complex with Me31B.
Me31B is essential for the translational silencing of osk and BicD mRNAs during their transport to the oocyte
We next examined whether loss of Me31B affects translation of osk and BicD mRNAs. We immunostained ovaries with an anti-Osk antiserum (Fig. 7A-F). Although osk mRNA is expressed during almost all stages of oogenesis, its translation is repressed to keep Osk protein level very low during early oogenesis (Kim-Ha et al., 1995; Markussen et al., 1995; Rongo et al., 1995). In me31B– egg chambers, Osk signal was significantly increased compared with that in the neighboring me31B+ egg chambers (Fig. 7B,E).
A similar increase of BicD signal in me31B– egg chambers was more evident (Fig. 7G-L). In wild-type egg chambers, BicD protein, like BicD mRNA, is highly concentrated in the oocytes starting from germarium region 2B (Wharton and Struhl, 1989; Suter and Steward, 1991). In the egg chambers lacking me31B, increased BicD signal was detected in nurse cell cytoplasm (Fig. 7H,K). These results suggest that loss of Me31B in germline cells causes derepression of osk and BicD mRNA translation during their transport to the oocyte.
DISCUSSION
In this report, we have identified Me31B as a new component of cytoplasmic RNP particles that function in Drosophila oogenesis. The particles contain several oocyte-localizing RNAs, suggesting that these RNAs are transported to the oocyte with Me31B through the formation of an RNP complex. Consistent with this, the particles also contain Exu, a factor involved in the localization of bcd and osk mRNAs to the anterior and the posterior pole of the oocyte, respectively. We have further shown that Me31B is required for translational silencing of two mRNAs in the particle: osk and BicD mRNAs. This is, to our knowledge, the first evidence for a link between transport and translational silencing of localized RNAs through the assembly of an RNP complex.
Me31B forms RNP particles with oocyte-localizing RNAs
We have shown that Me31B is integrated in large cytoplasmic particles containing many oocyte-localizing RNAs and Exu (Figs 2, 3; data not shown). This suggests that each Me31B-containing particle is an aggregate of numerous RNP complexes that are composed of different RNA species. It has been shown that mRNA localization during Drosophila oogenesis is microtubule dependent (Pokrywka and Stephenson, 1995) and that GFP-Exu particles move to the oocyte in a microtubule-dependent manner (Theurkauf and Hazelrigg, 1998). Thus, it is likely that the particles act as vehicles for RNA transport from nurse cells to the oocyte. We speculate that RNA transport to the oocyte is operated by a common mechanism, which involves assembly of an RNP complex, integration of each RNP complex into large aggregates and targeting of the aggregates to the oocyte along microtubule networks.
These RNAs usually have cis-regulatory elements for oocyte localization in their 3′UTR (Bashirullah et al., 1998). Trans-acting protein(s) required for oocyte localization would bind these elements to form a link between localizing RNAs and microtubule networks. Such protein(s) should be able to bind target RNAs in the absence of Me31B, as the transport of osk and BicD mRNAs to the oocyte is Me31B independent (Fig. 6). It has been shown that Exu is required for proper localization of bcd and osk mRNAs within oocytes (St Johnston et al., 1989; Wilhelm et al., 2000), and that it is able to bind RNA in vitro (Macdonald et al., 1995). Therefore, Exu might bind to the cis-regulatory elements in oocyte-localizing RNAs and promote their transport. Me31B forms a complex with Exu and Yps in an RNA-dependent manner (Fig. 3K), whereas binding between Exu and Yps is direct (Wilhelm et al., 2000). This suggests that Me31B and the Exu-Yps complex bind different regions of the same RNA molecule. These data lead us to speculate that the assembly of a cytoplasmic RNP complex is achieved by binding of functionally different proteins to discrete regions of an oocyte-localizing RNA.
It remains unanswered where Me31B is assembled to form an RNP complex. It has been shown that shuttling hnRNP proteins remain associated with several localized RNAs, even in the cytoplasm (Hoek et al., 1998; Cote et al., 1999; Lall et al., 1999; Norvell et al., 1999), suggesting that cytoplasmic RNPs could be nucleated within the nucleus. Our results show that Me31B is frequently enriched around the nuclear envelope in the nurse cell cytoplasm but is never detected within nuclei (Fig. 1). Thus, we favor the idea that Me31B is assembled into RNP complexes containing oocyte-localizing RNAs in the perinuclear region of the nurse cell cytoplasm. It is notable that in yeast and mammalian cells, Dbp5p/Rat8p, a DEAD-box protein that is phylogenetically closed to Me31B, is also highly enriched in the perinuclear region, and is required for mRNA export from the nucleus (Snay-Hodge et al., 1998; Tseng et al., 1998; Schmitt et al., 1999). Although the Drosophila counterpart of Dbp5 has not yet been characterized, similar distribution of the two related proteins suggests an important link between nuclear RNA export and assembly of cytoplasmic RNP complexes. The perinuclear region would be the site where nuclear RNP complexes are reorganized into a cytoplasmic form, in which additional proteins involved in RNA transport and translational silencing are incorporated.
Me31B is involved in translational silencing of oocyte-localizing RNAs
Significant accumulation of Osk and BicD proteins in nurse cells of me31B– egg chambers suggests that Me31B plays a role in translational silencing of these mRNAs during their transport to the oocyte. Alternatively, Me31B may regulate protein turnover such that these protein levels remain low in nurse cells. Both ideas are formally possible, but we favor the former one, because osk and BicD mRNAs are colocalized with Me31B during their transport from nurse cells to the oocyte. Interestingly, among DEAD-box proteins found in the Drosophila genome, Me31B is most similar to eIF4A, which is essential for translation initiation (Merrick and Hershey, 1996). Sequence similarity between the two proteins suggests that Me31B might antagonize the function of eIF4A such that translation of oocyte-localizing RNAs is silenced.
Egg chambers that lack me31B degenerate during mid-oogenesis (Fig. 5). As Me31B colocalizes with many oocyte-localizing RNAs other than osk and BicD mRNAs, Me31B is likely to silence translation of these RNAs. Therefore, me31B– phenotypes in mid-oogenesis could be caused simply by translational derepression of many mRNAs in the particles. Similar pleiotropic effects, including germ cell degeneration during mid-oogenesis and invading of nurse cell nuclei into the oocyte, has been observed in mutations for vas, which encodes a germline-specific DEAD-box protein implicated in translational activation of several maternal mRNAs (Markussen et al., 1995; Rongo et al., 1995; Dahanukar and Wharton, 1996; Gavis et al., 1996; Styhler et al., 1998; Tinker et al., 1998; Tomancak et al., 1998; Carrera et al., 2000). We propose that translational control plays a fundamental and widespread role to advance oogenesis. Nevertheless, it is still an unanswered question whether Me31B has another function in oogenesis. Isolation of weak mutations of me31B may reveal additional functions, if any, of Me31B.
In Xenopus oocytes, Xp54, a protein homologous to Me31B, has been identified as a major integral component of stored RNP particles (Ladomery et al., 1997), in which maternal mRNAs are masked from translation, degradation and polyadenylation/deadenylation (Spirin, 1996). Although the function of Xp54 in the RNP particles remains elusive, RNA masking is thought to involve the structural rearrangement of an RNP complex into a condensed form. Therefore, Xp54 may function to unwind the secondary structures of mRNAs to facilitate binding of RNA-packaging proteins such as FRGY2 (Ladomery et al., 1997; Matsumoto and Wolffe, 1998; Sommerville, 1999). Interestingly, Yps, a component of Me31B-containing particles, shares a conserved sequence motif with FRGY2 (Thieringer et al., 1997; Matsumoto and Wolffe, 1998; Wilhelm et al., 2000). Although the function of Yps in the particles remains to be analyzed, these observations suggest that similar proteins mediate the translational silencing during oogenesis between two diverse organisms.
Me31B is highly concentrated in distinctive cytoplasmic regions, the sponge bodies (Fig. 3J). A similar electron-dense structure known as the mitochondrial cloud has been found in Xenopus oocytes (Heasman et al., 1984; Wilsch-Bräuninger et al., 1997). Numerous maternal mRNAs that are involved in the axis determination and germ plasm assembly, co-migrate with this structure to localize the vegetal cortex in growing oocytes (Kloc et al., 1993; Ku and Melton, 1993; Forristall et al., 1995; Kloc and Etkin, 1995; Houston et al., 1998; Hudson and Woodland, 1998; Mowry and Cote, 1999; MacArthur et al., 2000; Pannese et al., 2000). It has recently been shown that some of the RNA components in the mitochondrial cloud are translationally silenced during oogenesis and translated only after the onset of embryogenesis (MacArthur et al., 1999; Houston and King, 2000). These data strongly suggest that the sponge body and the mitochondrial cloud are related cytoplasmic structures in terms of morphology and function, although it has not been determined whether they contain related protein components. Further analysis of composition and the role of the maternal RNP complexes in Drosophila and Xenopus will uncover the molecular mechanism of how localization and translation of maternal mRNAs is spatio-temporally regulated.
Acknowledgements
We thank B. Suter, K. Bohmann, C. Nüsslein-Volhard, P. Macdonald, N. Hom-Booher, J. E. Wilheim and T. Hazelrigg for providing antibodies, S. Styhler and P. Lasko for a GFP-Exu stock, and B. Suter for a genomic DNA library. We also thank P. Lasko and members of Kobayashi’s laboratory. for critical reading of this manuscript. This work was supported by a Grant-in-aid from the Ministry of Education, Science, Sports and Culture, Japan; the Tsukuba Advanced Research Alliance project; the University of Tsukuba Research Project; the Nissan Science Foundation; a CREST project from Japan Science and Technology Corporation; and a Research Project for Future Program from the Japan Society for the Promotion of Sciences.