Valois, a component of the nuage and pole plasm, is involved in assembly of these structures, and binds to Tudor and the methyltransferase Capsuléen

In Drosophila, the formation and fate of the germ cells depend on components distinctly located at the posterior pole of the egg – the pole plasm (Geigy, 1931). Functional and structural characteristics of the germ plasm confer specific properties to this region of the egg. First, transplantation of posterior plasm from oocytes or early cleavage embryos induces ectopic formation of pole cells within the recipient embryos(Illmensee and Mahowald, 1974; Okada et al., 1974). Second,the pole plasm is devoid of yolk particles, but exhibits electron-dense,ribonucleoprotein-rich organelles without delimiting membranes, called polar granules (Huettner, 1923; Mahowald, 1962; Illmensee et al., 1976). Third, the respiratory activity of the mitochondria located in the pole plasm is higher than that detected in other regions of preblastoderm embryos(Akiyama and Okada, 1992). Fourth, in addition to its ability to induce formation of ectopic pole cells,the pole plasm also contains an activity required for abdominal development,designated as the posterior determinant(Frohnhöfer et al., 1986; Lehmann and Nüsslein-Volhard,1986).

Although germline formation and abdominal development represent distinct developmental processes, molecular and genetic analyses of the posterior-grandchild-less group of genes indicate that both processes share a common pathway for producing functional determinants localized at the posterior pole of the embryo. So far four genes are known to be directly required in pole plasm function, including osk(Lehmann and Nüsslein-Volhard,1986), vasa (vas)(Schüpbach and Wieschaus,1986), tudor (tud)(Boswell and Mahowald, 1985)and valois (vls)(Schüpbach and Wieschaus,1986). Mutants in all four genes are characterized by the absence or strong reduction of discernible polar granules(Schüpbach and Wieschaus,1986; Boswell and Mahowald,1985).

Formation of pole plasm initially depends upon the localization of osk mRNA (Ephrussi and Lehmann,1992), which becomes concentrated at the posterior end of the oocyte in stage 8 egg chambers (Ephrussi et al., 1991; Kim-Ha et al.,1991). Shortly thereafter, osk mRNA is translated into two isoforms by use of different initiation codons(Markussen et al., 1995; Rongo et al., 1995), and the short form of Osk induces the assembly of polar granules by recruiting directly the Vas protein, a member of the DEAD-box family of putative RNA helicases (Lasko and Ashburner,1988; Hay et al.,1988b) at the posterior pole of the oocyte(Breitwieser et al., 1996). Vas may contribute to the synthesis of specific polar granule components, as suggested by its requirement for promoting translation of nanos(nos) (Gavis et al.,1996) and gurken(Styhler et al., 1998; Tomancak et al., 1998). The other factor involved in pole plasm function is Tud(Golumbeski et al., 1991),which accumulates in polar granules but is also detected in the nuage of nurse cells and at the periphery of embryonic nuclei(Bardsley et al., 1993). Although the molecular nature of vls remained elusive, this gene appears to be involved in pole plasm function(Schüpbach and Wieschaus,1986).

Formation of abdomen and pole cells can be mechanistically uncoupled. The nos gene contributes to abdominal patterning, albeit not to pole cell formation (Lehmann and Nüsslein-Volhard, 1991). The localization of nosmRNA to the posterior pole of the embryo(Wang and Lehmann, 1991)depends upon genes acting in pole plasm and produces a transient gradient of Nos proteins emanating from the posterior pole(Wang et al., 1994). Unlike the posterior determinant, the determinant involved in pole cell formation is likely to consist of multiple elements. The gene germ cell-less(gcl) is one of them (Jongens et al., 1992), and ectopic expression of gcl mRNA at the anterior pole of the embryo produces nuclei with characteristics of pole cell nuclei with which Gcl protein becomes associated on the inner rim of the nuclear pores (Jongens et al.,1994). Furthermore, mitochondrial ribosomal RNA (mtlrRNA) also contributes to germ-line determination(Iida and Kobayashi, 1998). Transport of the mtlrRNA from mitochondria to the surface of polar granules is mediated by Tud protein through a yet unknown mechanism(Amikura et al., 2001).

During our analysis of the capsuléen gene (csul; CG3730), we found one major protein partner by using the yeast two-hybrid system. In this report, we show that the protein interacting with Csul is encoded by the vls gene and describe its function in the process of pole cell formation.

Plasmid constructs were generated by amplifying selected fragments by PCR(High Fidelity PCR Master, Roche) and cloned into appropriate vectors. vas and tud cDNA plasmids were kindly provided by P. Lasko and R. Boswell, respectively. vls cDNA (clone GM04727, BDGP-EST project) was purchased from Invitrogen and gus cDNA (clone LD34464,BDGP-EST project) was obtained from the Drosophila Genomics Resource Center (DGRC; Bloomington, IN). The plasmid pSM 492 (3×HA) was kindly provided by S. Michaelis.

To rescue the vls phenotype a 2.04 kb genomic DNA fragment containing the vls transcription unit and 535 and 195 nucleotides of the 5′ and 3′ genomic regions, respectively, was cloned into the P-element vector CaSpeR4. The construct was named P-[vls]. Details will be provided upon request. To generate a Vls-HA fusion protein construct, a BglII site was created at the 3′end of the vls-coding sequence of a vls cDNA containing identical 5′ and 3′ untranslated regions as the genomic transgene. A fragment with 3×HA tags was cut with BglII from plasmid pSM 492 and cloned into the vsl-coding sequence. Details will be provided upon request. A clone containing six copies of HA was selected for microinjection.

The vls-coding region was amplified by PCR (High Fidelity PCR Master; Roche) from genomic DNA obtained from single vls¹,vls² and vls³ flies (primers: sense 5′-gttgctcttccttgctggccgattctc-3′, located at position –58 from the translation initiation codon; and antisense,5′-tgcatttaaactgggctgctgcttcac-3′). The PCR products of 1.46 kb were cloned into pBluescript. Two clones from independent PCR reactions were sequenced for each allele using primers covering the coding region. Nucleotide differences from the wild-type vls sequence were distinguished from PCR errors by their appearance in both independent clones.

Standard yeast two-hybrid methods were employed, using the GAL4 system(Saccharomyces cerevisiae strain Y190). The Csul-coding region was cloned into the yeast vector pGBT9 to produce a bait construct. This construct was used to screen the Drosophila embryo MATCHMAKER cDNA library(Clontech). About 1.2 millions clones were screened.

Full-length vls and vas cDNAs were subcloned into pGEX6P2(Pharmacia), full-length csul, gus and tud cDNAs into pCITE-4 (Novagen). Recombinant proteins were synthesized in vitro using the TNT T7 Coupled Reticulocyte Lysate System (Promega) in the presence of unlabeled amino acids. GST-fusion proteins expressed in E. coli were purified with glutathione sepharose (Pharmacia) and washed with the binding buffer [20 mM HEPES (pH 7.8), 10% glycerol, 300 mM NaCl, 0.1% sodium deoxycholate, 0.1% NP40 and 0.1% Triton X-100] plus protease inhibitors(Complete EDTA free from Roche; 1:50 dilution). Recombinant proteins (10% of the reaction volume) were added to this mixture (in 1 ml) and incubated for 3 hours at room temperature. The beads were washed six times (10 minutes each)with binding buffer, boiled in loading buffer, and the proteins were separated by electrophoresis on 8% or 12% SDS-polyacrylamide gels. After transfer to a polyvinylidene difluoride (PVDF) membrane, the bound proteins were detected by Western blotting using an S-protein Alkaline Phosphatase conjugate(Novagen).

Fly ovaries were homogenized using a plastic pestle in ice-cold IP buffer[145 mM NaCl, 10% glycerol, 1 mM MgCl₂, 1.5 mM NaH₂PO₄, 8 mM Na₂HPO₄ (pH 7.4) and 0.5% NP40] containing protease inhibitors. Lysates were clarified by centrifugation at 14,000 g for 10 minutes at 4°C. The supernatants were removed and mixed with 400 units of RNasin (Promega), 30μl of anti-HA antibody-coupled protein A beads for 5 hours at 4°C on a rotator. The IP mix was centrifuged at low speed to remove the supernatant. Protein A-beads were then washed four times with IP buffer for 10 minutes each and TAP-Csul was detected on a western blot by using alkaline phosphatase-conjugated IgG (Sigma). TAP-Csul purification was essentially performed as described by Rigaut et al.(Rigaut et al., 1999).

Immunostaining was performed with primary antibodies, including anti-Mael from rabbit (Findley et al.,2003), anti-Osk from rabbit (gift of A. Ephrussi), anti-Vas from rabbit and rat (gift of P. Lasko and A. Nakimura, respectively), anti-Tud from rabbit (TUD56; gift of S. Kobayashi), and monoclonal anti-HA (clone 16B12;BAbCO). To reduce background, we incubated first the anti-HA antibodies(1:400) for 3 hours at room temperature with wild-type ovaries and then overnight with experimental ovaries. For immunostaining, the ovaries were fixed for 10 minutes in 4% paraformaldehyde. To visualize components of the nuage, we used the protocol of Findley et al.(Findley et al., 2003). Chromatin was visualized by staining with Oli-Green (Molecular Probes). Cy3-,Cy5- and FITC-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:200.

The capsuléen gene (csul; CG3730) is required for pole cell formation in the pathway controlled by osk(J.A., R. Ollo, A. Ephrussi and B.M.M., unpublished). To identify Csul-interacting proteins active in pole plasm, we performed a yeast two-hybrid screen. The bait construct consisted of the sequence of the Csul protein (610 amino acid residues) fused to the DNA binding domain (BD) of Gal4. By screening a Drosophila cDNA library, we obtained six independent interacting clones of which four of them corresponded to the CG10728 coding sequence predicted by the Drosophila genome project. Conceptual translation of CG10728 cDNA revealed a protein of 367 amino acids. Two cDNAs encoded a nearly full-size protein lacking the first 10 amino acid residues, whereas the other two isolates encoded a protein in which 37 residues were missing from the N terminus. Confirmation of the interaction was obtained by performing the reciprocal two-hybrid assay,showing that BD-CG10728 could strongly bind the activation domain of Gal4 fused to Csul (data not shown). Sequence analysis on the Protein Sequence Analysis Server(bmerc-www.bu.edu/psa/)indicated that CG10728 contains four WD repeats(Fig. 1A).

The CG10728 gene is located at chromosomal band 38B2. As vls was genetically assigned to this region(Schüpbach and Wieshaus,1986), we investigated whether the CG10728 sequence was altered in vls mutants. To determine the lesions in vls¹, vls² and vls³, we amplified by PCR genomic DNA fragments encompassing the CG10728-coding sequence, then cloned and sequenced the amplified fragments from each mutant. For each vls mutant we found single nucleotide substitutions that result in premature termination of the coding sequence (Fig. 1A). In vls¹ and vls² the nucleotide substitutions changed a TGG Trp codon into TAG and TGA stop codons, producing truncated proteins of 227 and 52 amino acids, respectively. In vls³, the nucleotide substitution transformed a CGA Arg codon into a TGA stop codon producing a truncated protein of 69 amino acids. To confirm that we cloned vls, we constructed two transgenes containing the presumed vls sequence, including its 5′regulatory sequences. The first transgene contained a genomic DNA fragment comprising the CG10728 transcription unit(Fig. 1B). The second transgene harboured the same sequence to which we fused six copies of the hemaglutinin tag (HA), in frame with the vls-coding sequence. Western blots of proteins extracted from ovaries producing HA-Vls probed with anti-HA antibodies displayed a single polypeptide with the expected mass of 50 kDa(Fig. 1C). Both transgenes restored full viability and fertility to embryos produced by vls¹ and vls³ homozygous females (data not shown). Taken together, our results demonstrate that CG10728corresponds to vls.

Among the nearest homologues of Vls we identified in protein databases the methylosome protein 50 (MEP50) (Fig. 1D). MEP50, the human homologue of Vls, contains six WD repeats and interacts with PRMT5 (Friesen et al.,2002), the human homologue of Csul. The finding of an interaction between Csul and Vls in yeast prompted us to analyze whether such binding occurs in Drosophila ovaries. To this end, we constructed transgenic flies bearing both the HA-vls transgene and a Tandem Affinity Purification (TAP) (Rigaut et al.,1999) tagged csul transgene (TAP-csul) shown to restore csul development (data not shown). Protein extracts of P{TAP-csul}; vls³; P{HA-vls} ovaries were immunoprecipitated using rabbit IgG-sepharose beads and the bound Csul-complexes were released by treatment with recombinant TEV protease. Blots of released proteins probed with anti-HA antibodies showed an immunoreactive band of 50 kDa exhibiting the predicted mass of HA-Vls fusion protein(Fig. 2A). Reciprocally, blots of ovarian proteins immunoprecipitated with anti-HA antibodies and probed with IgG antibodies displayed an immunoreactive band of the mass predicted for TAP-Csul (90 kDa; Fig. 2B). These results indicate that Csul and Vls can associate in a protein complex in Drosophila ovaries.

To further characterize the interaction between Csul and Vls, we mapped the binding domains in each protein by using a GST-pull down assay. For this purpose we fused Vls or fragments of Vls to GST, and Csul or fragments of Csul to S•Tag (Kim and Raines,1993). In vitro translated S•Tag-Csul polypeptides were incubated with immobilized GST-Vls proteins and, after washing, the bound S•Tag-Csul proteins were detected by using S-protein coupled to alkaline phosphatase. As shown in Fig. 2C, Csul bound specifically to immobilized full-length Vls, or to the 30 amino acid C terminal region of Vls, but not to the rest of the protein. GST alone displayed no binding to S•Tag-Csul (data not shown). Conversely, we mapped the region of Csul mediating the interaction with Vls to the C terminal region of Csul (Fig. 2D). Neither the N-terminal region nor the central region of Csul showed strong binding to Vls. Thus, our data indicate that Csul and Vls interact through their C-terminal regions.

To determine the distribution of Vls during oogenesis, we analyzed the localization of HA-Vls protein using anti-HA antibodies. Immunostaining of vls³; P{HA-vls} egg chambers revealed that Vls accumulates at the posterior pole of the oocyte in stage 10 egg chambers(Fig. 3A,C) and in the pole plasm of early syncytial embryos (Fig. 3E,F). We also found Vls in pole cells(Fig. 3H,I) and in migrating primordial germ cells (data not shown). During oogenesis, Vls decorates a region surrounding the nurse cell nuclei(Fig. 3K), which bears strong similarity to the nuage (Findley et al.,2003). By contrast, immunostaining of wild type ovaries using anti-HA antibodies displayed only an overall residual staining(Fig. 3B,D,G,J,L). These data show that Vls is a component of both the nuage and the pole plasm.

Embryos produced by vls females fail to form pole cells at the posterior pole (Schüpbach and Wieschaus, 1986) (Fig. 4B). To check whether pole cell formation strictly depends on vls activity, we introduced an osk-bcd 3′UTR transgene into vls¹ and vls³ backgrounds, and found no evidence for ectopic formation of pole cells at the anterior pole(Fig. 4D), although we could detect residual Vas staining in this region of the egg (data not shown). This result concurs with the suppressor effect of vls¹ on ectopic pole cell formation observed after overexpression of osk(Smith et al., 1992), but contrasts the data using an osk-bcd 3′UTR transgene expressed in vls²/Df(2R)TW2 flies by Ephrussi and Lehmann(Ephrussi and Lehmann, 1992). From our data, we conclude that vls function is essential for pole cell formation.

To determine vls hierarchical position among known components of the pole plasm, we investigated the distribution of Osk, Vas and Tud in vls¹ and vls³ egg chambers. Use of specific antibodies against Osk revealed that Osk was normally localized at the posterior pole from stage 9 onwards egg chambers in vls(Fig. 4F). However, we often noticed that the amount of Osk protein was reduced in vls in comparison with wild type. To determine whether vls regulates Osk protein accumulation, we examined by immunoblotting the amount of Osk protein produced in wild-type and vls ovaries. We found that the level of Osk short form was markedly reduced in vls in comparison with that in wild type (Fig. 4G)(Rongo et al., 1995) and even more reduced in aubergine (aub)(Wilson et al., 1996). However, the phosphorylation of this isoform seems unaffected in vls. In addition, we noticed that the slower migrating form of the short Osk protein was particularly reduced in abundance in aub protein extract,suggesting a defective phosphorylation in aub ovaries. This result extends the previous observation of Wilson et al.(Wilson et al., 1996), showing a strongly reduced level of short Osk in aub ovaries and may explain the aub requirement for osk activity induced anteriorly by the OB1 osk-bcd 3′UTR transgene. Examination of Vas distribution showed a normal localization at the posterior pole in stage 10 vls egg chambers (Fig. 4I), albeit at a reduced level compared with wild-type egg chambers (Fig. 4H).

Checking the distribution of Tud during oogenesis revealed that Tud was absent from the posterior pole of vls mutant oocytes in stage 10 egg chambers (Fig. 4K), indicating that Vls may play a role in pole plasm accumulation of Tud. As Tud is also a component of the nuage encircling nurse cell nuclei(Bardsley et al., 1993), we looked at this structure by examining the distribution of three components of the nuage including Tud, Vas (Hay et al.,1988a) and Maelstrom (Mael)(Findley et al., 2003). The use of a transgene expressing a GFP-Vas fusion protein(Sano et al., 2002) associated with the vls chromosome revealed that GFP-Vas was normally localized in the nuage of vls nurse cells, albeit at a lesser degree compared with wild type (Fig. 5B)(Hay et al., 1990; Lasko and Ashburner, 1990). Examination of Mael showed that this protein is also present in the nuage,although its distribution diverges from the wild-type pattern. We found that Mael is concentrated in brighter spots in vls than in wild type at the periphery of the nurse cell nuclei(Fig. 5D). By contrast, we were unable to detect Tud in the nuage of vls nurse cells(Fig. 5F), although Tud accumulates in vls oocytes (Fig. 5F) and transiently localizes at their anterior margin (data not shown). These findings indicate that: (1) vls acts downstream of Vas for the recruitment of Tud in the nuage; and (2) events occurring in the nuage are dispensable for the transport of Tud from the nurse cells to the oocyte but are required for Tud recruitment in the pole plasm.

Examination of vls egg chambers revealed also a modification in the structure of the oocyte nucleus in which the karyosome is fragmented into two fuzzy spots (Fig. 5H)instead of forming a single compact dot in wild-type nuclei(Fig. 5G). A similar phenotype was described for spindle(González-Reyes et al.,1997) and vas mutants(Tomancak et al., 1998). As vls oocytes undergo normal meiosis and support normal embryonic development, this finding suggests that Vls may exert a dispensable function on the karyosome organization.

The presence in Vls of four WD repeats able to interact with other proteins(reviewed by Smith et al.,1999) prompted us to examine whether Vls could physically interact with proteins localized in the nuage. For this purpose, we performed pull-down assays using tagged Tud, Vas and Gustavus (Gus)(Styhler et al., 2002)polypeptides.

Five segments of Tud fused to the S•Tag, including JOZ (amino acid residues 3-273), 9A1 (residues 198-1199), 3ZS+L-N (residues 1198-1981) and 3ZS+L-C (residues 1941-2515) (Golumbeski et al., 1991), together comprising the complete Tud protein(Fig. 6A), were in vitro translated and the synthesized polypeptides were incubated with immobilized full-length and truncated GST-Vls proteins. Of the five Tud fragments, we found that only the 9A1 fragment could interact with Vls, whereas JOZ and the two subfragments of 3ZS+L showed only weak binding(Fig. 6C). The 9A1 fragment was further divided into two fragments. The 9A1-N and 9A1-C fragments (residues 198-770 and 751-1199, respectively) displayed a strong binding to Vls(Fig. 6C).

In the reciprocal experiment, we found that the Tud 9A1-N fragment can bind to the region encompassing residues 90-190 of Vls. When this segment was further divided into two A and B fragments, containing residues 90-139 and 139-192, respectively, we discovered that only the fragment A was able to bind to Tud 9A1-N. These results indicate that the first WD repeat of Vls can directly interact with Tud (Fig. 7).

No specific Vls binding with Vas and Gus polypeptides was detected in GST pull-down assays using GST, GST-Vls and GST-Vls (residues 90-190). Similarly,no interaction could be uncovered between Vls and Vas in the yeast two-hybrid assay (data not shown). These data indicate that the binding between Vls and Tud represents a specific interaction.

By analyzing potential interactors of the putative Csul methyltransferase,we isolated the vls gene. The confirmation that we cloned vls was obtained by determining the molecular lesions in three EMS-induced vls mutations and by showing that development and fertility of vls mutant flies is restored by transgenes containing the vls genomic sequence.

Attempts to isolate additional partners of Vls by using the yeast-two hybrid system were unsuccessful because of the occurrence of a too large number of clones growing on selective medium (data not shown), presumably reflecting the occurrence of WD repeats that are known to mediate interactions with numerous proteins in eukaryotic cells(Smith et al., 1999). By using direct binding assays with proteins involved in pole plasm function, we found that Vls interacts with Tud.

The feature characterizing the superfamily of WD-repeat proteins is the WD motif, a ∼40 amino acid stretch typically containing a GH dipeptide 11-24 residues from its N-terminus and a WD dipeptide at its C terminus, albeit exhibiting only limited amino acid sequence conservation(Yu et al., 2000). When present in a protein, the WD motif typically occurs in multiple tandem repeated units. Based on structural analysis, the conformation of the WD repeat is defined as a series of four-stranded anti-parallel β sheets,which fold into a higher-order structure termed a β-propeller. At least four repeats are required to constitute a β-propeller(Chothia et al., 1997). We found that Vls contains four conserved WD repeats and can potentially form aβ-propeller structure. Interestingly, the nearest mammalian homologue of Vls, the human MEP50 protein, displays six WD repeats. Although the sequence of the two additional repeats is conserved overall in Drosophila Vls,no characteristic GH and WD di-peptides could be found in these repeats(Fig. 1D).

WD-repeat proteins act as scaffolding/anchoring proteins for a number of binding partners (Smith et al.,1999). WD-repeat motifs within one protein can simultaneously bind several proteins and foster transient interactions with other proteins. Moreover, WD-repeat proteins occur in relatively stable protein complexes in which they play a structural role. A similar function can be assigned to Vls in promoting either permanent or transient interaction with other proteins.

The human homologue of Vls, the MEP50 protein, was identified through its association with PRMT5 (Friesen et al.,2002), the human homologue of Drosophila Csul methyltransferase. We show that the domain of Vls required for binding to Csul is small and confined to its C-terminal end. By contrast, the domain of MEP50 specifically involved in the interaction with PRMT5 remains unknown.

MEP50 acts as an adaptor binding to a subset of the Sm proteins and contributing to their methylation by the PRMT5 methyltransferase(Friesen et al., 2002). Biochemical assays indicate that MEP50 is necessary for the methyltransferase activity of the methylosome, as anti-MEP50 antibodies significantly reduce the methylation of Sm proteins (Friesen et al., 2002). However, the precise role of MEP50 remains elusive. Its possible functions include a regulation of the enzymatic activity of PMRT5 and the control of the positioning of the substrate for methylation.

The human methylosome complex is involved in the assembly of spliceosomal U-rich small nuclear ribonucleoproteins (snRNPs) mediated by the survival motoneuron (SMN) protein, a gene product that is affected in spinal muscular atrophy (Lefebvre et al.,1995; Meister et al.,2002). SMN is produced ubiquitously and contains a single Tudor domain that associates with SmB/B′, SmD1-D3 and SmE proteins of snRNPs(Liu et al., 1997; Buhler et al., 1999; Selenko et al., 2001). The assembly of snRNPs mediated by SMN occurs in the cytoplasm and is stimulated by the PMRT5-methylosome complex that converts specific arginine residues in the Sm proteins into dimethylarginines(Meister at al., 2001; Friesen et al., 2001a; Branscombe et al., 2001),facilitating the binding of the Sm proteins to SMN and their association with snRNA molecules (Friesen et al.,2001b; Brahms et al.,2001). Ultimately, the assembled snRNPs are released and targeted to the nucleus, whereas the SMN-PRMT5 complex may dissociate before its components associate again for a new round of assembly.

In Drosophila, the Tud protein differs from SMN by containing eight Tudor domains and two Tudor-like domains(Callebaut and Mornon, 1997),whereas SMN contains a single Tudor domain(Pontig, 1997). It is thus possible to envisage that Drosophila Tud may bind different categories of cytoplasmic RNPs through its multiple Tudor domains. However, in contrast to the PMRT5/MEP50 complex that apparently binds SMN through other protein(s) present in the complex, we found that Vls can directly bind to Tud through its first WD repeat. As the C-terminal tail of Vls binds to Csul, it is possible that the Drosophila Csul/Vls methylosome associates with Tud through Vls. Thus, we propose that the association between the methylosome and Tud promotes binding and assembly of specific RNPs on Tud. Further experiments are needed to unravel the relationship between these proteins, the targets of Csul methyltransferase activity and the nature of the RNPs associated to Tud.

vls mutations causing a grandchild-less phenotype are characterized by agametic larvae exhibiting defects in abdominal patterning(Schüpbach and Wieschaus,1986). Eggs laid by homozygous vls females are devoid of polar granules and consequently the embryos produce no pole cells(Schüpbach and Wieschaus,1986). In these embryos, Tud is absent from the posterior pole(Bardsley et al., 1993), and Vas rapidly disappears from this location during the period of nuclear cleavage (Hay et al., 1990; Lasko and Ashburner, 1990). Our analysis reveals that localization of Tud is already absent from the nuage surrounding the vls nurse cell nuclei. However, the occurrence of Vas and Mael in the nuage of vls nurse cells indicates that aspects of this structure can be made independently of Tud.

A pivotal role in the organization of the nuage was assigned to Vas on the basis of its involvement in localizing specific components such as Aub(Wilson et al., 1996) and Mael(Clegg et al., 1997) in this structure (Harris and Macdonald,2001; Findley et al.,2003). To confirm a Vas-dependence of Tud localization in the nuage, we examined Tud distribution in vas egg chambers and found that the localization of Tud around nurse cell nuclei is fully abolished (data not shown). Similarly the absence of Tud in the nuage of vls nurse cells shows that Tud localization in the nuage is vls dependent. However, we detected significant amounts of Tud in vls oocytes and at their anterior border, indicating that Tud accumulation in the nuage is not required for Tud transport into the oocyte and its anterior margin. Furthermore, as Vls-HA does not accumulate in early oocytes nor localize at their anterior margin (data not shown), we conclude that the interaction between Vls and Tud should be spatiotemporally regulated.

We show that Vls is a component of the nuage and pole plasm. Only a limited number of proteins display a similar pattern of distribution, including Vas(Hay et al., 1988a), Aub(Harris and Macdonald, 2001)and Tud (Bardsley et al.,1993). The dual location of these proteins indicates that they either perform distinct functions at each location or exert a function in the nuage required for their accumulation in the pole plasm. The finding that Vas absence in the pole plasm correlates with its absence in the nuage supports the latter possibility (Liang et al.,1994).

Inactivation of vls function exerts a further effect on the production of the short form of Osk protein. As osk mRNA localization and amount seems normal in vls embryos(Ephrussi et al., 1991), we assume that the lower amount of this form detected by immunoblotting in vls ovaries corresponds to either a defect in Osk synthesis or stability. We notice, however, that the level of Osk abundance varies considerably between individual vls oocytes with an apparently normal level in a small number of oocytes and a markedly reduced level in the majority of oocytes. The lower amount of Vas detected at the posterior pole of stage 10 vls or vls²/Df(2L)TW2 oocytes(Hay et al., 1990), can be interpreted as a consequence of the reduced amount of Osk protein, as Vas is absent from the posterior pole of osk oocytes(Hay et al., 1990; Lasko and Ashburner,1990).

The mechanism by which Vls regulates Osk synthesis and/or stability remains unknown. However, on the basis of Vls localization during oogenesis, we envisage that vls could regulate the production of the short form of Osk by two distinct mechanisms. One the one hand, vls can regulate Osk synthesis by recruiting specific enhancing factors in the pole plasm. On the other hand, Osk synthesis may be dependent on events mediated by vls occurring in the nuage. Similarly, efficient osk mRNA translation in the pole plasm could also be mediated by Aub in the nuage(Harris and Macdonald, 2001). Furthermore, recent data point out that the nuage may function in assembling or reorganizing ribonucleoprotein complexes, particularly those involving localized or translationally regulated mRNAs(Snee and Macdonald,2003).

The formation of polar granules fully depends upon vls activity(Schüpbach and Wieshaus,1986), but only partially upon tud function as polar granules in reduced number and altered morphology are observed in amorphic tud pre-blastoderm embryos(Thomson and Lasko, 2004). This raises the question of what the targets of vls function are in addition to tud and osk. Further experiments will reveal the components required for vls-dependent formation of polar granules.

As this work was being completed, another group reported the cloning of vls by positional mapping (Cavey et al., 2005). Although Cavey and colleagues identified only two WD repeats in Vls, they clearly recognized that its nearest homologue is MEP50. As the human methylosome is formed by MEP50 and PMRT5 homologous to Drosophila Vls and Csul, respectively, our finding that Vls can specifically bind to Csul indicates that it is the orthologue of MEP50 and not a `divergent WD protein'. In contrast to Cavey and colleagues, we found a restricted and dynamic Vls distribution during oogenesis, first in the nuage and then at the posterior pole of the growing oocyte. Finally, we observed that Vls is preferentially incorporated in the forming pole cells. These findings show that vls may crucially act in the nuage, germ plasm and pole cells, and are consistent with the vls mutant phenotype. Similar to Cavey and colleagues, we detected a lower amount of Osk at the posterior pole of growing vls oocytes, but, in a discrepancy that we cannot explain, we found that Osk levels were already lower in stage 9 egg chambers,whereas Cavey et al. only observed a decrement of Osk later during oogenesis in stage 11 egg chambers.

In conclusion, we demonstrate that Vls can interact physically with at least two proteins, Csul and Tud, which are specifically involved in germ-line determination. Vls, in association with Csul, constitutes the first example of a partner of a dimethylarginine protein methyltransferase whose function has been characterized in vivo. These findings reinforce their cardinal function in a pathway first elucidated through genetic investigations. Our work sets the basis for further investigations on the role of Vls, its dependence upon Csul and its involvement in specific localization of cytoplasmic proteins during the formation of a functional pole plasm.

We are grateful to R. Boswell, A. Ephrussi, S. Findley, S. Kobayashi, P. Lasko, R. Lehmann and A. Nakimura, for providing reagents. We are particularly thankful to Anne Ephrussi for her numerous inputs given to this work. We also thank the members of the DKFZ group, present and past, especially Gunter Merdes for his advice, and Joachim Marhold, Rolf Schmitt, the students of the Fachhochschule für Technik und Gestaltung (Mannheim), Tanja Hönig,Sylvia Lux and Jens Hohmann for their help.