Drosophila valois encodes a divergent WD protein that is required for Vasa localization and Oskar protein accumulation

valois (vls) was identified as a posterior group gene in the initial screens for Drosophila maternal-effect lethal mutations. Despite its early genetic identification, it has not been characterized at the molecular level until now. We show that vls encodes a divergent WD domain protein and that the three available EMS-induced point mutations cause premature stop codons in the vls ORF. We have generated a null allele that has a stronger phenotype than the EMS mutants. The vlsnull mutant shows that vls+ is required for high levels of Oskar protein to accumulate during oogenesis, for normal posterior localization of Oskar in later stages of oogenesis and for posterior localization of the Vasa protein during the entire process of pole plasm assembly. There is no evidence for vls being dependent on an upstream factor of the posterior pathway, suggesting that Valois protein (Vls) instead acts as a co-factor in the process. Based on the structure of Vls, the function of similar proteins in different systems and our phenotypic analysis, it seems likely that vls may promote posterior patterning by facilitating interactions between different molecules.


Introduction
The embryonic body axes are specified in Drosophila melanogaster during oogenesis, when cytoplasmic determinants localize to different regions of the developing oocyte. This initiates the formation of positional information centers, which define polarity and pattern the body plan along the anteroposterior (AP) and dorsoventral (DV) axes during embryogenesis. This developmental control mechanism is based on mRNA localization and anchoring to specific subcellular compartments. In conjunction with tight translational control of localized mRNAs this is an efficient means with which to generate a local source of polarity determinants, one that is widely used throughout phyla for various purposes (Kloc et al., 2002). In Drosophila oocytes, posteriorly localized oskar (osk) mRNA is locally translated starting in mid-oogenesis (stage 8-9) and nucleates the assembly of the pole plasm (or germ plasm). The pole plasm specifies the germline at the posterior end of the embryo, and it patterns the abdomen along the AP axis (Johnstone and Lasko, 2001;Riechmann and Ephrussi, 2001).
The osk ribonucleoprotein (RNP) complex has been characterized, and many conserved factors are known to function in mRNA localization and/or translational control in different systems across phyla (Farina and Singer, 2002;Hachet and Ephrussi, 2004;Roegiers and Jan, 2000;Wilhelm et al., 2000). Because restriction of osk activity to the posterior is crucial for normal development (Ephrussi and Lehmann, 1992), both pre-and post-translational control mechanisms regulate Osk protein accumulation. Osk protein is actively degraded by the ubiquitin-proteasome pathway, but protected from it by phosphorylation by Par-1 specifically at the posterior (Riechmann et al., 2002). Translational control of osk involves the coordinate action of repressors and derepressors interacting with discrete elements of osk transcripts during transport and at the posterior pole (Gunkel et al., 1998;Kim-Ha et al., 1995;Webster et al., 1997;Yano et al., 2004). Additional factors that do not function as derepressors are also required for stimulating osk translation (Wilson et al., 1996). In addition, Oo18 RNA-binding protein (Orb) polyadenylates osk transcripts at the posterior pole once derepression has been achieved (Castagnetti and Ephrussi, 2003).
Two isoforms of Osk (Long and Short Osk) are produced by initiation at two different in-frame start codons. Short Osk has long been known as the active isoform for pole plasm assembly which recruits downstream components of the pathway such as Vasa (Vas) (Markussen et al., 1995), and recently, Long Osk has been shown to be responsible for anchoring osk mRNA and Short Osk at the posterior (Vanzo and Ephrussi, 2002). Short Osk is likely to anchor Vas directly at the posterior (Breitwieser et al., 1996;Vanzo and Ephrussi, 2002). Vas is an ATPdependent RNA-helicase from the DEAD-box family and has been implicated in translational activation of several maternal transcripts, including osk (Styhler et al., 1998;Tinker et al., 1998;Tomancak et al., 1998). tudor (tud) acts downstream of vas and is followed in the cascade by additional genes whose products localize to the pole plasm and mark the separation of valois (vls) was identified as a posterior group gene in the initial screens for Drosophila maternal-effect lethal mutations. Despite its early genetic identification, it has not been characterized at the molecular level until now. We show that vls encodes a divergent WD domain protein and that the three available EMS-induced point mutations cause premature stop codons in the vls ORF. We have generated a null allele that has a stronger phenotype than the EMS mutants. The vls null mutant shows that vls + is required for high levels of Oskar protein to accumulate during oogenesis, for normal posterior localization of Oskar in later stages of oogenesis and for posterior localization of the Vasa protein during the entire process of pole plasm assembly. There is no evidence for vls being dependent on an upstream factor of the posterior pathway, suggesting that Valois protein (Vls) instead acts as a cofactor in the process. Based on the structure of Vls, the function of similar proteins in different systems and our phenotypic analysis, it seems likely that vls may promote posterior patterning by facilitating interactions between different molecules. germline establishment and abdominal patterning activities (Golumbeski et al., 1991). Pole cell formation depends on the localization of germ cell less (gcl) mRNA (Leatherman et al., 2002) and mitochondrial large ribosomal RNA (Iida and Kobayashi, 1998). Abdominal patterning relies on the vasdependent translation of nanos (nos) mRNA at the posterior pole. This results in a concentration gradient of Nos protein along the AP axis, which acts as the primary posterior morphogen (Riechmann and Ephrussi, 2001).
One more posterior group gene, valois (vls), had been identified in the initial screen for maternal-effect steriles (Schupbach and Wieschaus, 1986), but has neither been cloned nor studied genetically in detail yet. Based on three EMSinduced alleles of vls, it was classified as a member of the 'grandchildless-knirps-like' group that also includes vas, stau and tud. Their phenotype is characterized by a lack of pole cells at the posterior and various degrees of abdominal segment deletions. Pole cell transplantation experiments demonstrated that vls functions in the germline (Schupbach and Wieschaus, 1986) and vls mutants were shown to have a non-functional pole plasm (Lehmann and Nusslein-Volhard, 1991). Until now, the position of vls in the posterior pathway has remained controversial. vls was tentatively placed downstream of osk and vas, but upstream of tud. This was based on the observation that osk mRNA and Vas protein are initially correctly localized to the posterior of the oocyte in vls EMS mutants. Vas then detaches from the posterior of the embryo soon after fertilization (Ephrussi et al., 1991;Hay et al., 1990;Lasko and Ashburner, 1990) and Tud localization is disrupted in embryos from vls mothers (Bardsley et al., 1993). However, conflicting data were reported subsequently. Assembly of an ectopic pole plasm at the anterior of the oocyte, caused by overexpressing osk (6xosk) (Smith et al., 1992) or by targeting osk transcripts specifically to the anterior margin (osk-bcd3′UTR) (Ephrussi and Lehmann, 1992), results in progeny embryos with ectopic pole cells and duplication of the abdomen at the anterior. vls function was found to be required for the expression of the 6xosk phenotype, confirming its position downstream of osk, but not for the expression of the osk-bcd3′UTR phenotype (Ephrussi and Lehmann, 1992;Smith et al., 1992).
Here, we report the cloning and characterization of vls. We have created a null mutant for vls that shows stronger phenotypes than the presently available vls EMS alleles. In contrast to previous models, this tool allows us to demonstrate that vls acts upstream of vas. Furthermore, vls dramatically affects the levels of Osk protein, even though localization of osk mRNA and initial accumulation of Osk do not require vls function. vls encodes a novel protein with significant similarity to WD domain proteins. The presented data suggest that Vls may act as a co-factor in assembling protein-protein and/or protein-RNA complexes.

Generation of anti-Vls antibody
A partial vls cDNA was recovered by PCR amplification from an ovarian cDNA library (Larochelle and Suter, 1995). Rabbit polyclonal antibodies were raised against an E. coli expressed Vls polypeptide (amino acids 166-367; Fig. 1) that contained an N-terminal 6xHis tag. The fusion protein was purified by affinity chromatography and gel electrophoresis prior to injection. Antiserum was affinity purified with a MBP::Vls fusion protein (expressed from the same vls cDNA) that was coupled to CnBr-activated sepharose beads (Pharmacia Biotech).

Western blotting
Ovary, embryo, larvae and whole fly extracts were prepared as described previously (Masrouha et al., 2003). Protein samples were resolved by 10% SDS-PAGE electrophoresis, transferred to nitrocellulose membranes and probed with antibodies. Horseradishperoxidase-conjugated anti-rabbit/mouse IgG were used at 1:2000 (Amersham Pharmacia Biotech).

RNA in situ and immunostaining
In situ hybridizations to osk mRNA on ovaries were performed as described previously (Suter and Steward, 1991). The osk probe was generated by random priming with the DIG High Prime digoxigenin labeling system (Roche Applied Science). Immunostaining on ovaries were performed as described previously (Findley et al., 2003) with α-Osk at 1:3000 and secondary Alexa Fluor anti-rabbit 488nm.

Vls-eGFP and Vas-eGFP observations
A Zeiss confocal microscope was used for Vls-eGFP and a Leica confocal microscope for Vas-eGFP observations. Two-to three-dayold females were used for ovaries and embryo collections. For live observation of Vls-eGFP, ovaries were dissected in halocarbon oil 27, separated and dragged onto a cover slip. They were then covered with petriPERM 50 hydrophobic membrane dishes (Vivascience). Samples were used for no more than 15 minutes after dissection. vas-eGFP ovaries were dissected in Ringer's buffer, fixed for 20 minutes in phosphate-buffered saline (PBS) with 4% paraformaldehyde, rinsed three times, washed twice for 5 minutes in PBST (PBS + 0.2% Tween 20) and mounted in 60% glycerol. Embryos were collected for 1 hour periods, aged accordingly, dechorionated by rolling them over doublesided sticky tape and mounted on slides in halocarbon oil 27. Background signal was evaluated for all types of samples by using yw 118 ovaries or embryos as a control. The wavelength window of detection was adjusted to reduce background signal produced by autofluorescent particles.

Results
Cloning and identification of vls Df(2L)be408 removes part of the barr gene, and the entire coding sequences of chk2 and CG10728 ( Fig. 1) (Masrouha et al., 2003). This deficiency does not complement vls EMS alleles, indicating that one of these three transcription units corresponds to vls. A barr + transgene rescues the barr but not the vls phenotypes. Similarly, a transgene containing chk2 alone does not rescue the vls phenotypes either. By contrast, a transgene containing a wild-type copy of CG10728 ( Fig. 1) was able to rescue the maternal-effect lethal phenotype of chk2 null CG10728 null double mutants and the grandchildless phenotype associated with vls EMS /Df(2L)be408 mutants. These valois in posterior development results strongly suggest that CG10728 corresponds to vls. Indeed, sequencing this genomic region in the three EMS alleles vls PG65 , vls RB71 and vls HC33 finds a single nucleotide substitution in each of them, resulting in premature stop codons in the predicted open reading frame (ORF) of CG10728 ( Fig. 1). This identifies CG10728 as vls and Df(2L)be408 / Df(2L)pr2b, P[w + , barr + ] constitutes a true null mutant for vls and chk2. As the chk2 + construct does not rescue any of the vls phenotypes, but the vls + transgene rescues all of them, we will refer to this mutant as vls null hereafter.

Overlapping genes and repression of chk2 by vls
Interestingly, chk2 and vls are encoded by opposite strands and cDNA sequence data shows that their 3′UTRs are complementary over 127 nucleotides (Fig. 1). chk2 is translationally repressed by orb during oogenesis (Masrouha et al., 2003), and because translational control often relies on the binding of trans-acting factors to sequences in the 3′UTR of mRNAs, we were curious to know whether vls could also play a role in chk2 translational control. Indeed, Chk2 levels increase about 6-fold in vls PG65/HC33 and vls PG65/RB71 ovaries compared with wild type (Fig. 2) and this is close to the 10fold upregulation reported for orb mutants (Masrouha et al., 2003). This indicates that vls is also involved in the regulation of Chk2 levels. However, orb does not simply function to control Vls levels because these are normal in orb mutants (data not shown).
This peculiar genomic organization of chk2 and vls appears somewhat conserved among Drosophilidae because in D. pseudoobscura, vls and chk2 are also on opposite strands next to each other. However, based on the gene predictions only (no cDNA sequences are available), there is no evidence that the mRNAs overlap. In the more distant diptera Anopheles gambiae, the genes are still on the same chromosome, but much further apart. In the human genome, finally, the predicted homologues of these genes are unlinked (data not shown).

Valois belongs to a family of divergent WD domain proteins
vls encodes a novel protein and PROSITE predicted the existence of two WD domains. Database searches reveal the best sequence similarity with the human methylosome protein 50 (MEP50; 20.4% identity; Fig. 3 and Table 1) and alignment of Vls and MEP50 shows that the two predicted WD domains of Vls correspond closely to the predicted WD domains 2 and 3 of MEP50. With the exception of the WD domain 5, the predicted WD domains of MEP50 show elevated similarity with corresponding Vls regions compared to the alignment of the entire proteins (Table 1). This suggests that Vls may have five to six domains that have a similar structure or function as WD domains, and it may mean that Vls has evolved from a WD domain protein. The six WD domains of MEP50 are thought to fold into a β-propeller structure, which serves as a platform for recruiting the Arg-methyltransferase JBP1/ PRMT5 and its substrates, the Sm proteins (Friesen et al., 2002). This event is required for assembling the splicing

Df(2L)be408
l (2) 2. chk2 is translationally repressed by orb and vls during oogenesis. The indicated ovary extracts were probed with α-Chk2 (Masrouha et al., 2003) and α-eIF4A antibodies as a loading control. Protein levels were quantified using a phosphorimager. The upper band in OreR and chk2 null mutants is crossreacting material of unknown identity. Chk2 levels were normalized to eIF4A levels and expressed as a percentage of normalized Chk2 levels in wild type (OreR), which is arbitrarily set at 100%. machinery prior to import into the nucleus (Friesen et al., 2001).
To test whether Vls may be the Drosophila ortholog of MEP50, we searched by BLAST the translated Drosophila genome for sequences similar to MEP50 and used the CLUSTALW multiple alignments tool to analyze the results. The search revealed an entire family of Drosophila WD domain proteins with comparable sequence similarity to MEP50 and clearly a greater level of conservation over the WD domains than in regions outside. Although different algorithms give slightly different alignments, we find that the protein products of CG6486, Lis-1 and vls have the highest levels of similarity over the WD domains of MEP50 (Table 1 includes the six most similar proteins). The observed differences are too small to predict which one of the Drosophila proteins is more likely to be the ortholog of MEP50.

Valois is a maternal product
Northern analysis detected a transcript of 1.5 kb for vls expressed in ovaries, early embryos and adult females, but absent from pooled larval instars and adult males (Butler et al., 2001). In situ hybridization to OreR ovaries with a vls probe detected signal throughout the germ cell cytoplasm from early oogenesis onwards. The signal showed no specific localization pattern. Surprisingly, we detected an equally strong signal in vls PG65/RB71 ovaries, and only in the vls null ovaries the signal is at background levels (data not shown).
On Western blots, polyclonal anti-Vls antibodies do not detect any Vls in vls null , vls PG65 , vls RB71 and vls HC33 ovary extracts (Fig. 4A). This shows that the antibody specifically recognizes the Vls protein and that the EMS mutants do not make significant levels of stable full length Vls. However, because we do not know which epitopes are recognized by the polyclonal antibody, it is still possible that the EMS alleles produce truncated forms of Vls. In wild-type flies, Vls is abundant in ovaries, early embryos and adult females, but reduced in adult males (Fig. 4B). The fact that it is present in ovaries and in 0to 1-hour-old embryos indicates that Vls is a maternally provided protein and this is consistent with the maternal-effect phenotype of vls mutants.

Vls as a co-factor of the posterior pathway
Comparing OreR and control vls null ovaries stained with α-Vls antibodies reveals Vls signal at low levels along parts of the oocyte cortex of wild-type stage 10 egg chambers, and also a stronger signal in the nurse cells, where it appears to be concentrated in nuage (data not shown). Because our anti-Vls antibodies do not work well for immunostaining, we also   -eGFP] transgene rescues the female sterile phenotype of vls null mutants, proving that Vls-eGFP possesses vls + activity (data not shown). Vls-eGFP localization in vls null background is indistinguishable from that in wild-type background (data not shown). Vls-eGFP signal is cytoplasmic and stronger in the germline than in somatic cells, but in contrast to the immunostaining, specific localization patterns of Vls-eGFP were usually not observed (Fig. 5). Only in the germarium did we occasionally observe Vls-eGFP concentrating in perinuclear aggregates that disappear by stage 2 of oogenesis. At later stages, Vls-eGFP signal is uniformly distributed in the nurse cells and oocyte, as well as in young embryos, with no particular enrichment at the posterior or inside the pole cells.
Western blot analysis of Vls as well as localization studies of Vls-eGFP in other posterior group mutants (osk, vas, tud, gus and orb) and in grk failed to identify potential upstream factors of vls that could control its expression levels, potential post-translational modifications, or its spatiotemporal distribution patterns (data not shown). Together with the uniform distribution pattern of Vls-eGFP, this argues that Vls may act as a co-factor in the posterior pathway.

Posterior localization of Osk protein in late oogenesis depends on vls
To investigate further the position of vls in the pathway, we examined the distribution of posterior products in vls null ovaries. osk mRNA is efficiently localized at the posterior of vls null mutant oocytes (Fig. 6A), consistent with previous reports for embryos from vls PE36 mothers (Ephrussi et al., 1991). Osk protein accumulates at the posterior pole of the oocyte during stages 8-10. During this phase, we observe similar patterns in wild type and vls mutants ( Fig. 6B-E,B′-E′). However, at later stages (stage 11, Fig. 6B′′-E′′), Osk levels at the posterior seem somewhat reduced in vls null oocytes compared with OreR and vls null vls + , and we often do not detect any signal for Osk in vls null oocytes. This reduction of Osk levels at the posterior is also observed in hemizygous vls PG65 , albeit to a lesser extent (Fig. 6D′′).
Confirming these observations, Western blot analyses revealed lower levels of Osk in vls null mutants and vls PG65 hemizygotes compared with OreR and vls null vls + ovary extracts, and vls null mutants again show a stronger reduction than vls PG65 hemizygotes do (Fig. 7). A similar decrease of Osk levels was reported previously for vls PE36/RB71 (Rongo et al., 1995). The Long and Short Osk isoforms are affected differently in vls mutants. Whereas Long Osk is only slightly reduced, the Short, indispensable form of Osk, is strongly reduced in vls null and vls PG65 hemizygotes. Moreover, we observe an isoform-specific reduction of the hyperphosphorylated (upper) Short Osk compared with the hypophosphorylated (lower) form in vls (and vas) mutants. This difference is more clearly seen in hemizygous vls PG65 than in vls null , as the hypophosphorylated form of Short Osk is also practically undetectable in vls null . This effect had been described previously for vas mutants (Markussen et al., 1995). We note that vls mutants cause a clear reduction of Long Osk compared with vas mutants, which have relatively normal levels. Furthermore, tud 1 mutant extracts contain lower levels of both isoforms of Short Osk with no isoform-specific reduction, and it seems that the Long isoform might also be slightly reduced although it did not appear to be affected in another study (Markussen et al., 1995).
Taken together, anti-Osk immunostaining and western analyses suggest that vls is required for normal levels of Osk to accumulate at the posterior pole while the pole plasm is assembling. Starting around stage 11, Osk signal progressively disappears from the posterior in the absence of vls function, and by later stages, Osk accumulation at the posterior is probably greatly reduced, explaining the drastic reduction of overall Osk levels observed on western blots.

vls is essential for posterior localization of Vasa
Vas protein is the next factor in the posterior pathway to localize to the posterior end of the oocyte after osk mRNA and protein. This osk-dependent Vas localization remains stable at the posterior pole during the early stages of embryogenesis and Vas is later incorporated into pole cells (Lasko and Ashburner, 1990). In vls null ovaries and in embryos from vls null mothers, Fig. 4. Western analysis of Vls. Ponceau Red staining serves as a loading control and is shown below the corresponding blot. (A) Anti-Vls antibodies detect a protein of ~42 kDa in wild-type ovaries and in ovaries expressing one copy of the CG10728 + /vls + rescue construct. This antigen is not detected in vls null mutants, nor in vls EMS hemizygotes and trans-heterozygotes. Genotypes: vls null -Df(2L)be408 / Df(2L)pr2b, P[w + barr + ]; vls null vls + -Df(2Lbe408, P[w + CG10728 + /vls + ] / Df(2L)pr2b, P[w + barr + ]; and Df -Df(2L)be408. (B) Developmental expression profile of Vls. vls null ovary extracts serve as a control for the specificity of the α-Vls antibodies. Other protein extracts were collected from wild-type flies. The faint low molecular weight band does not appear to originate from ovaries or embryos.
anti-Vas antibody staining showed very little or no accumulation of Vas at the posterior end (data not shown). This observation was further confirmed by analyzing the distribution of Vas-eGFP (Styhler et al., 1998) in vls null and vls PG65 ovaries and embryos. Although the early localization pattern of Vas in nuage of the mutant nurse cells is normal (Fig.  8A-D), the posterior localization in stage 10 oocytes is not observed in the null mutants, and appears very weak in vls PG65 hemizygotes (Fig. 8A′-D′).
Later in development, Vas-eGFP signal is detected at the posterior end and then inside the pole cells of embryos from wild-type and vls null vls + mothers, but not from vls PG65 hemizygotes and vls null mothers (Fig. 8E-H′). These results contrast with previous reports where Vas localization defects in vls mutants (vls PE36 and vls RB71 ) were observed only slightly after fertilization, before pole cell formation (Hay et al., 1990;Lasko and Ashburner, 1990). Our data for the vls null and vls PG65 alleles implicate vls in the late localization or anchoring of Vas to the posterior cortex during oogenesis.

Discussion
The posterior gene vls encodes a maternal protein and is essential for the late localization of Vas to the posterior of the oocyte as well as for the accumulation of Osk, which orchestrates pole plasm assembly. Unlike many other members of the posterior pathway, vls transcripts and Vls protein are not localized to the posterior but accumulate uniformly in the nurse cells and oocyte throughout oogenesis. Similar to vas, vls transcripts and proteins are also detected in adult males even though they have no essential function in males or fly spermatogenesis (Lasko and Ashburner, 1990;Snee and Macdonald, 2004). By contrast, in mice and probably other mammals, vas is important for male gametogenesis and has no essential function in female fertility (Raz, 2000). It would thus be interesting to know whether the same is true for vertebrate vls.

Nature of vls mutants
Specification of the germline in Drosophila is more sensitive to pole plasm activity than is abdominal patterning. This is illustrated by the fact that weak alleles of posterior group mutants display a grandchildless phenotype caused by the lack of pole cells, while stronger alleles cause additional abdominal patterning defects that result in embryonic lethality (Lehmann and Nusslein-Volhard, 1991). In our hands, the hemizygous EMS alleles vls PG65 , vls RB71 and vls HC33 are only partially maternal-effect lethal and 100% grandchildless. vls null , however, is 100% maternal-effect lethal. The stronger phenotype of the null mutant suggests that the EMS alleles may be hypomorphs. However, the initial work on vls produced strong genetic evidence that the EMS alleles are actually nulls (Schupbach, 1986). It is therefore also possible that the EMS allele stocks accumulated maternal-effect modifiers that allow them to survive to adulthood.

vls mRNA escapes NMD
Although vls EMS alleles contain premature stop codons in the vls ORF, the corresponding mutant mRNAs seem to escape nonsense-mediated mRNA decay mechanisms (NMD). Even though premature stop codons are recognized differently in Drosophila and vertebrates, the NMD components are conserved (Gatfield et al., 2003). Given that vls + is translated during oogenesis, it seems unlikely that the mutants are protected because of lack of translation . It would thus be interesting to find out why vls EMS transcripts accumulate to normal levels.

Collapse of the pole plasm in the absence of vls function
Because all aspects of the vls mutant phenotype observed in embryos, including abdominal segment deletions, lack of pole cells, gastrulation defects and weak ventralization are rescued completely by a vls transgene and not even partially by a chk2 transgene, we concluded that vls alone has a developmental requirement. Furthermore, we have demonstrated elsewhere that chk2 function is only clearly required upon activation of cell cycle checkpoints (Masrouha et al., 2003). The vls phenotypes are reminiscent of a collapse of pole plasm Development 132 (3) Research article Vls-eGFP signal is generally fairly weak in the oocyte and more difficult to detect against the autofluorescent yolk particles (D). However, the signal is still significantly higher than in control yw ovaries without vls-eGFP (E).
(F,G) 0-to 1-hour-old embryos from vls-eGFP (F) and yw (G) mothers. (H,I) 1-to 2-hour-old embryos from vls-eGFP mothers. Vls-eGFP signal is detected in somatic and germ cell cytoplasm. pc, pole cells; sc, somatic cells. valois in posterior development assembly that seems to occur around stage 10 of oogenesis in our vls null mutants. vas is crucial for the pole plasm to assemble properly and recruit the mRNAs and proteins required for pole cell specification and abdominal patterning.
Genetic evidence implicates vas in the translational activation of several targets during oogenesis, including osk, grk and, in particular, nos at the posterior pole of the embryo (Carrera et al., 2000;Gavis et al., 1996;Johnstone and Lasko, 2004;Markussen et al., 1995;Rongo et al., 1995;Styhler et al., 1998;Tomancak et al., 1998). Vas levels directly correlate with pole plasm activity, pole cell formation being more vulnerable to decreased Vas levels than abdominal patterning is (Ephrussi and Lehmann, 1992). Previous immunostaining for Vas has been reported to show indistinguishable Vas accumulation at the posterior pole of vls mutant and wild-type oocytes, and young embryos. These studies, performed with the homo-and hemizygous EMS mutants, showed a loss of posterior localization in the embryos from vls mothers sometime between fertilization and pole cell formation (Hay et al., 1990;Lasko and Ashburner, 1990). We used vas-eGFP transgenes to assess the posterior localization of Vas in vls null and hemizygous EMS alleles in detail. Maximal localization was still very weak and was found in oocytes and embryos from vls EMS mothers. In vls null mutants we observed a nearly complete failure to localize Vas-eGFP at the posterior pole. This failure coincides with the collapse of the pole plasm and is probably the cause for the various embryonic phenotypes mentioned above. Consistent with this, the observed Vas localization defects parallel the severity of the phenotypes that we report for these vls alleles. The weak accumulation of Vas at the posterior of vls PG65 hemizygous oocytes gives rise to a grandchildless phenotype, whereas the almost complete absence of Vas from the posterior of vls null oocytes results in a fully penetrant maternal-effect lethal phenotype.
vls is thus required during oogenesis for the localization (transport or anchoring) of Vas to the posterior cortex of the oocyte. The fact that Vls is not specifically enriched at the posterior may suggest that it acts to modify or transport pole plasm components before they reach the posterior pole. Preliminary experiments also failed to produce evidence that Vls and Vas are part of the same protein complex (not shown). This suggests that the mode of action of vls on Vas localization is transient or indirect. The fact that osk mRNA and protein are initially correctly localized implies that oocyte polarity is normal in vls mutants and that vls is not required for osk mRNA localization. Levels of Osk protein isoforms are then reduced in later stages and western analysis reveals a much more drastic decrease of overall Osk levels than immunostaining does for both types of vls alleles. This suggests that most of the drop in Osk levels occurs during the late stages of oogenesis, when the vitelline membrane prevents antibody staining for oocyte Osk. Therefore, it seems that shortly after initiating pole plasm assembly, Osk fails to be maintained at the posterior of vls mutants and progressively disappears, concurrent with a complete collapse of the pole plasm.

vls acts upstream of Vas and Osk
Several lines of evidence implicate the Short Osk isoform in directly anchoring Vas. Short Osk interacts strongly with Vas Fig. 6. osk mRNA and Osk protein distribution appear normal initially in vls mutant ovaries but Osk protein disappears from the posterior pole after stage 11. (A) In situ hybridization with an osk probe to a vls null stage 10 egg chamber. osk mRNA is correctly concentrated at the posterior of the oocyte in vls null mutants. (B-E′′) Immunostaining with α-Osk antibodies on wild-type and vls mutant ovaries. Osk signal is tightly concentrated at the posterior of stage 9-10 oocytes of wild type (B-B′) and vls null vls + (C-C′), as well as in vls PG65 (D-D′) and vls null (E-E′) females. However, around stage 11, Osk signal appears slightly weaker at the posterior of vls PG65 (D′′) and significantly weaker in vls null oocytes (E′′). This defect is even more pronounced in vls null after stage 11, where usually no posterior Osk signal is detected (not shown).
in the two-hybrid system and recruits Vas when ectopically localized in the oocyte (Breitwieser et al., 1996;Cha et al., 2002;Vanzo and Ephrussi, 2002). Because Vas-eGFP mislocalization patterns in stage 10 oocytes are indistinguishable in vls and osk 54 mutants (not shown), vls could act directly at the level of Osk accumulation (e.g. in stimulating translation of osk), which is necessary for anchoring Vas at the posterior pole. On the other hand, it is also possible that vls acts primarily on Vas protein localization. Because Vas also seems to act in a positive feedback loop back on Osk protein accumulation (Markussen et al., 1995), the lack of Vas localization in vls mutants would then also preclude maintenance of posterior accumulation of Osk protein. In vls mutants, Osk levels appear to decrease just slightly after Vas should have localized to the posterior pole, thus it appears that the failure to localize Vas could be the cause of the pole plasm collapse in vls mutants. To investigate these issues further, we compared Osk levels in vas and tud mutants with those in vls mutants by western analysis where we detect a more significant drop than by immunostaining. This analysis revealed generally stronger phenotypes for vls than for vas and tud mutants. We observed a comparable decrease of Short Osk levels on western blots of vls, vas and tud mutant extracts, but with slight differences in the extent of reduction of the hyper-and hypophosphorylated forms, both of which are more severely affected in vls mutants. In addition, we observed a clear reduction of Long Osk levels in vls, a minor reduction in tud, but none in vas mutant extracts (Fig. 7). However, this analysis is complicated by the fact that the vas and tud alleles that are useful and available, respectively, for these experiments are not nulls (Bardsley et al., 1993;Hay et al., 1988;Lasko and Ashburner, 1990). Their residual activity may therefore maintain Osk at the posterior for a longer period of time. These data are thus consistent with the idea that vls acts on either pathway target, Vas or Osk, in a process which could involve additional intermediates that remain to be identified.  Why is vls not required for expression of the oskbcd 3′UTR phenotype? vls was tentatively placed downstream of vas in the posterior pathway based on studies reporting that Vas localization is correct initially in vls EMS mutants (Hay et al., 1990;Lasko and Ashburner, 1990), and because vls was found to be required for the expression of the 6xosk phenotype (Smith et al., 1992). Surprisingly, however, vls is not required for the expression of the osk-bcd 3′UTR phenotype (Ephrussi and Lehmann, 1992). As the 3′UTR is present in the 6xosk transgenes but not in the osk-bcd 3′UTR transgene, one explanation for this discrepancy could be that vls is required to relieve translational repression mediated by the osk 3′UTR. It is also possible that differences in osk mRNA levels and concentration at the anterior between the two systems might explain the discrepancy. In fact, Vas protein accumulation at the posterior pole and the number of pole cells that develop afterwards correlate directly with the osk gene copy number (Ephrussi and Lehmann, 1992). Besides, 6xosk produces lower levels of osk mRNA at the anterior than osk-bcd3′UTR (Smith et al., 1992). Therefore, the ectopic pole plasm induced by osk-bcd3′UTR mRNA is probably more resistant to defects in localization/anchoring of downstream components such as Vas or to defects in the maintenance of Osk itself. By contrast, the 6xosk system seems to represent a more sensitized background where the collapse of an ectopic pole plasm is more likely to occur in the absence of vls. Supporting this idea, the bicaudal phenotype of the progeny from transgenic mothers is 100% penetrant with osk-bcd 3′UTR (Ephrussi and Lehmann, 1992), but only 73% penetrant with 6xosk (Smith et al., 1992). vls might thus function as an enhancer of pole plasm assembly, which is dispensable when osk pole plasm-inducing activity is already extensively deployed at the anterior. This is consistent with our observation that vls dose also correlates with pole plasm activity in the same way that osk does. One copy of a wild-type vls + transgene rescues almost completely the phenotypes described for vls null , but we sometimes noted minor defects compared with wild-type flies, and reduced hatching rates of embryos (not shown).
Speculations on the molecular function of Vls vls differs in many respects from the other long-known members of the posterior pathway and seems to encode a cofactor acting on Osk protein accumulation, Vas localization and possibly on another, yet unknown, component of this pathway. Two lines of evidence suggest that vls facilitates the process of pole plasm assembly but is not absolutely essential: some residual Vas localization is possible even in the null mutant; and an ectopic pole plasm can assemble in the absence of vls function provided that the system is set up excessively or through different 3′UTR control elements (osk-bcd3′UTR vs. 6xosk). How could Vls perform this function at the molecular level?
Vls is a divergent WD domain protein. The β-propeller structure of WD proteins is thought to arise from the folding of at least four WD domains and to promote several simultaneous protein-protein interactions (Smith et al., 1999). Because computer predictions only found two or three such domains in Vls, we tested in preliminary experiments whether Vls forms homodimers. However, we did not detect any untagged Vls in immunoprecipitations performed with functional Vls-eGFP and Vls-6xHis fusion proteins (not shown). Whether Vls forms heterodimers with other WD domain-containing proteins remains to be tested. Sequence alignments point to a more likely interpretation. Vls and a whole family of Drosophila WD domain proteins show similarities to MEP50, which contains six WD domains and facilitates the interactions between a methyltransferase and its substrates, the Sm proteins (Friesen et al., 2002). Notably, the regions corresponding to the WD domains of MEP50 are better conserved than the others, suggesting that these domains are under greater selection pressure and may therefore fold in similar structures that can fulfill similar functions. This sequence comparison also shows that Vls might not be the ortholog of MEP50 and that different members of this family might fulfill the function of MEP50 in different Drosophila tissues.
It is therefore possible that Vls also acts as a mediator of molecular interactions between proteins and possibly also mRNAs. Future experiments will have to focus on identifying the interactors of Vls to determine how precisely vls facilitates the pole plasm assembly process. The Vls interactions may turn out to represent an activating step in pole plasm assembly that involves a methyltransferase or another protein modification enzyme and their substrates. With this information it should then also be possible to clarify how directly this mechanism acts on the targets Vas and Osk.