vasa is required for GURKEN accumulation in the oocyte, and is involved in oocyte differentiation and germline cyst development

Segregation of the germline from the soma is a central feature of animal development. In Drosophila, the germline is determined through the activities of maternally expressed RNAs and proteins which colocalize in the pole plasm at the posterior pole of the egg (reviewed by Rongo and Lehmann, 1996). Pole cells, the progenitors of the germline, form very early in embryogenesis, then, beginning at gastrulation, they migrate into the interior of the embryo and ultimately associate with the gonadal mesoderm to form the embryonic gonads (reviewed by Williamson and Lehmann, 1996). Beginning in larval development, germ cells proliferate and differentiate in order to carry out spermatogenesis and oogenesis; among the structures assembled during oogenesis is new pole plasm, which specifies the germline for the subsequent generation of individuals.

Genetic and molecular studies have identified numerous genes which are required for pole plasm assembly and subsequent posterior segment specification and germ cell formation; many of these genes are expressed during oogenesis and produce mRNAs and/or proteins which localize in pole plasm or in polar granules, specialized organelles contained within the pole plasm (reviewed by Rongo and Lehmann, 1996). Analysis of the expression of these genes supports an early hypothesis (Mahowald, 1968) that translational control is a major mechanism regulating Drosophila germline development. The product of the vasa (vas) gene, a DEAD-box-family protein which is localized in polar granules and which shares the enzymatic functions of the translation initiation factor eIF4A (Hay et al., 1988; Lasko and Ashburner, 1988; Liang et al., 1994), is a candidate germline-specific translational regulator. For instance, levels of the short isoform of OSKAR protein (OSK), a molecule central to pole plasm assembly (Ephrussi et al., 1991; Kim-Ha et al., 1991; Ephrussi and Lehmann, 1992), are greatly reduced in vas mutant ovaries (Markussen et al., 1995; Rongo et al., 1995). Another pole plasm mRNA whose translation may be activated by VAS is nanos (nos), as nos RNA carrying an intact translational regulation element in its 3′UTR is completely inactive in embryos derived from vas mutant ovaries (Gavis et al., 1996; Dahanukar and Wharton, 1996).

While the activities of pole plasm components such as VAS have been most thoroughly studied with respect to their function in pole cell formation and specification of the posterior soma, clearly some genes involved in pole plasm assembly also function in other stages of germline development. For instance, females homozygous for either of two strong nos alleles exhibit defects in germ cell proliferation (Lehmann and Nüsslein-Volhard, 1991; Wang et al., 1994). Furthermore, pole cells lacking maternal nos function fail to complete migration and do not associate with the embryonic gonadal mesoderm (Kobayashi et al., 1996), indicating a role for nos in the transition from pole cell to functional germ cell.

Similarly, various vas alleles have defects in oogenesis and lay few or no eggs (Lasko and Ashburner, 1988, 1990; Lehmann and Nüsslein-Volhard, 1991; Schüpbach and Wieschaus, 1991). Females trans-heterozygous for Df(2L)A267 and Df(2L)TE116-GW18, two large deletion mutations which both include vas, were reported to be blocked in early vitellogenic stages of oogenesis (Lasko and Ashburner, 1988). Analysis of whether this phenotype was caused solely by loss of vas function has been confounded by the fact that these trans-heterozygous deficiency lines are haploid for a large number of genes, but that, aside from large deficiencies, a clearly null allele of vas did not exist. Four EMS-induced alleles of vas, vas^D1, vas^Q6, vas^Q7 and vas^D5, also lead to greatly reduced fertility, with many egg chambers blocked as for the trans-heterozygous deficiency females (Lehmann and Nüsslein-Volhard, 1991). The few eggs produced by females homozygous for these alleles often lack dorsal appendages and have the micropyle, a specialized vitelline membrane structure normally found only at the anterior of the egg, duplicated at the posterior (Lehmann and Nüsslein-Volhard, 1991). Again, whether these phenotypes represent the results of a complete loss of vas function is unknown. vas^Q6 and vas^D5 are missense mutations which alter single amino acids of VAS and both alleles produce substantial amounts of mutant protein (Liang et al., 1994), so neither of these mutations is likely to be null. For vas^D1 and vas^Q7 the molecular nature of the mutation is unknown, but the vas coding region is unaffected in these mutant alleles.

In this paper, we have used a new vas null allele, vas^PH165, a small deletion which we generated by imprecise P-element excision, to investigate in detail the role of vas in events of oogenesis prior to pole plasm assembly. We found that abrogation of vas function results in defects in many aspects of oogenesis including control of cystocyte divisions, oocyte differentiation, and specification of posterior and dorsal follicle cell-derived structures. Furthermore, vas^PH165 oocytes only weakly concentrate many oocyte-localized RNAs, although some oocyte-specific molecules, including gurken (grk; Schüpbach, 1987; Neuman-Silberberg and Schüpbach, 1993) RNA, remain concentrated in the oocyte in vas mutant ovaries. However, in the case of grk, translation is severely reduced in the absence of vas function. This provides evidence that VAS is involved in translational control mechanisms operating in early stages of oogenesis.

To create a null allele of vas, excision lines were generated through the introduction of the Δ2-3 transposase source into vas^P(ry[+])LYG2 (Rittenhouse and Berg, 1995). To do this, +/Y; vas^P(ry[+])LYG2cn; ry⁵⁰⁶ males were crossed to w; Bic-D^PA66Su(Bic-D^PA66) cn/CyO; Δ2-3Sb/TM3 Ser virgin females. w/Y; vas^P(ry[+])LYG2cn/CyO; Δ2-3Sb/ ry⁵⁰⁶ males were then crossed to +/+; l(2)05084^{P(ry[+]l(2)05084}/CyO; ry⁵⁰⁶ virgin females. Individual ry F¹ males representing excisions of vas^P(ry[+])LYG2were then individually crossed to l(2)05084^{P(ry[+]l(2)05084}/CyO; ry⁵⁰⁶ virgin females. Balanced ry stocks were generated and females homozygous for an excision chromosome were crossed to Oregon R males to test for fertility. Excision lines were screened for deletions through Southern analysis using a 1.9-kb EcoRI genomic fragment from the vas first intron, and which includes the vas^P(ry[+])LYG2 insertion site, as a probe. VAS protein levels from excision lines were determined by western analysis.

vas^P(ry[+])LYG2 was provided by Celeste Berg (University of Washington, Seattle); l(2)05084^{P(ry[+]l(2)05084}/CyO was provided by the Berkeley Drosophila Genome Project; nos^RC, which disrupts a splice donor site and is null for RNA and protein expression (Wang et al., 1994; Curtis et al., 1997) was received from Takahiro Akiyama (Azabu University, Kanagawa, Japan); and grk^HK36, a strong grk allele (Schüpbach, 1987) was obtained from Trudi Schüpbach (Princeton University). The wild-type strain employed was Oregon R.

In situ hybridizations with DIG-labeled RNA probes and antibody stainings were carried out on ovaries and embryos as described by Kobayashi et al. (1998), except that DMSO was omitted from the fixation solution used for ovaries. Primary antibodies were used at the following dilutions: α-ORB, 1:20; α-BIC-D, 1:10; α-NOS, 1:150; α-GRK, 1:3000; α-ADD-87, 1:20. α-NOS was pre-adsorbed with embryos from nos^BN females, and α-GRK was preadsorbed with grk^HK36 ovaries. For bright-field microscopy, antibody stainings were detected with DAB, enhanced using the Vectastain ABC or ABC Elite kits (Vector Laboratories) and biotinylated secondary antibodies. For confocal microscopy, antibody stainings were detected using Texas Red-conjugated secondary antibodies (Molecular Probes). β-gal staining of khc:lacZ ovaries was carried out as described by Clark et al. (1994).

Ovaries were dissected in PBS and fixed for 20 minutes in a mixture of 600 μl heptane, 200 μl 4% paraformaldehyde in PBS + 0.2% Tween-20, and 20 μl DMSO. Following fixation, samples were rinsed 3 times with PBT (PBS + 0.1% Tween-20), incubated for 1 hour in RNase A (100 mg/ml), and rinsed again 3 times with PBT. Ovaries were incubated for 1 hour in OliGreen (Molecular Probes, 1:1000 dilution) and Texas Red-phalloidin (Molecular Probes), washed with PBT, and mounted in 70% glycerol in PBS.

Identification of additional vas cDNA clones has indicated the presence of a 127-bp exon upstream of the previously reported 5′ end of vas, which extends the 5′ UTR of the gene. This exon is separated by a large intron of 6603 bp from the remainder of the gene (Fig. 1A,B; Berkeley Drosophila Genome Project, unpublished results, submitted to GenBank under accession numbers L81347, L81348, L81449, AC000466, and AC000469). vas^LYG2 is a P-element-induced vas allele (Rittenhouse and Berg, 1995), which we mapped to the large first intron of vas (Fig. 1A), and which produces about 2% of the wild-type level of VAS (Fig. 1C), a level essentially undetectable in tissue staining experiments. Despite the very low level of VAS in vas^LYG2 ovaries its phenotype is hypomorphic; oogenesis in vas^LYG2 females is less severely compromised than in Df(2L)A267/Df(2L)TE116-GW18 flies, and vas^LYG2 females lay numerous eggs. This suggested to us either that trace amounts of VAS are sufficient for oogenesis to often proceed to completion, or that the more severe phenotype observed in the double-deficiency lines results from the effects of a reduction in dosage of a second gene which enhances the vas phenotype.

To distinguish between these possibilities, it was necessary to obtain a molecularly characterizable null vas allele. For this purpose, we generated a series of derivative lines by mobilizing the P element in vas^LYG2. Out of 201 excision events, stable lines were generated from 129. Of these, 119 caused reversion to a wild-type phenotype, indicating that the vas^LYG2 phenotype results solely from the P-element insertion, and that the vas^LYG2 chromosome is free of other female-sterile mutations. Three other derivatives were recessive lethal, and seven, although having excised the ry⁺ marker on the P element of vas^LYG2, remained defective in oogenesis. These were checked by PCR and Southern blots to determine whether they carried a deletion confined to the vas gene, and by western blots using anti-VAS antiserum to determine whether they expressed VAS. From these analyses, one line, vas^PH165, was identified, which produced no detectable VAS protein and carried a 7343-bp deletion which removes the entire coding region of vas (Fig. 1A,C). We confirmed the breakpoints of vas^PH165 by nucleotide sequencing, comparing the mutant sequence to wild-type sequence provided by the Berkeley Drosophila Genome Project. From the nature of the vas^PH165 mutation and the fact that no VAS protein is detectable on western blots even on long overexposures (Fig. 1C), we conclude that vas^PH165 is null for VAS. The vas^PH165 deletion is mostly limited to vas, as one of its breakpoints lies within vas and the other is 1270 bp downstream of its 3′ end. A nested gene may be located within the 3.5-kb third intron of vas, which is deleted in vas^PH165, as a 3-kb transcript present throughout all developmental stages is detected on northern blots using genomic probes including this intron but not with vas cDNA probes (Lasko and Ashburner, 1988). However, any gene other than vas which may be disrupted in vas^PH165 is almost certainly irrelevant to the discussion below, as a vas-GFP transgene, constructed from a vas cDNA fused to the vas promoter, and therefore lacking any nested or 3′-flanking genes, rescues vas^PH165 homozygotes, and Df(2L)A267/ Df(2L)TE116-GW18 flies to fertility. The phenotypes of vas^PH165/vas^PH165, vas^PH165/ Df(2L)A267, and Df(2L)A267/ Df(2L)TE116-GW18 ovaries are essentially identical.

Upon cursory examination of vas^PH165 ovaries under the light microscope, we noticed an obvious atrophy of the germaria as compared with the wild-type, suggesting that fewer germ cells were present. To investigate this more closely, we used an antibody recognizing ADD-87 protein (Lin et al., 1994; Zaccai and Lipshitz, 1996) which stains spectrosomes and fusomes, specific structures present in stem cells, cystoblasts and dividing cystocyte clusters. In vas^PH165 ovaries, a reduction in the number of ADD-87-staining structures, and therefore a reduction in the number of developing cysts, is readily apparent as compared with wild-type (Fig. 2A-D). This phenotype has high expressivity, and increases in severity with the age of the female (Fig. 2C-D). In germaria from 7-day-old vas^PH165 females, region 1 frequently consists of only a few stem cells and developing cysts. Posterior to these there is often what appears to be an extended interfollicular stalk (compare Fig. 2D with Fig. 2C), likely formed by follicle cells in the absence of cystocyte clusters. This suggests that the cystocyte clusters remaining in the germarium have ceased to develop further and have degenerated, while the follicle cells continue to divide.

Two other genes that function in the same pathway as vas in posterior patterning, nos and pum, have also been implicated in germ cell proliferation (Lehmann and Nüsslein-Volhard, 1991; Wang et al., 1994; Lin and Spradling, 1997), raising the possibility that they also may be interacting with vas in the germarium. pum mutants produce ovarioles which contain only two or three clusters of undifferentiated germ cells which lack spectrosome/fusome structures or ring canals (Lin and Spradling, 1997), a phenotype we never observed in vas mutants. Upon examination of nos^RC mutant ovaries with the α-ADD-87 antibody, we observed a phenotype which appears similar, but somewhat more severe, than that of vas^PH165. Many nos^RC ovarioles consist of a germarium with one to three cysts (Fig. 2E), followed by an extended stalk and one to three normal-looking egg chambers. In more extreme cases, only remnants of spectrosome/fusome material can be detected in the anterior of the germarium (Fig. 2F), consistent with the conclusion that these germline cells have arrested development.

At a low frequency (aprox. 1% for each), we observed defects in germline differentiation and oocyte determination in vas^PH165 ovaries, including tumorous egg chambers (Fig. 3A), egg chambers with 16 nurse cells and no oocyte, others with two oocytes, and again others with a mislocalized oocyte (Fig. 3B-D). Far more frequently the normal 15 nurse cells and one oocyte are present; however, by at least two criteria the oocytes produced in vas^PH165 egg chambers are not fully differentiated. In wild-type development, the nurse cell nuclei endoreplicate during pre-vitellogenic oogenesis and become highly polyploid, whereas the oocyte nucleus remains diploid and condenses into a tight karyosome (Mahowald and Kambysellis, 1980). However, in vas^PH165 the oocyte nucleus appears more diffuse than does the wild-type oocyte nucleus (Fig. 3E,F).

This would be consistent either with a failure to form the karyosome structure, perhaps involving a premature meiotic arrest in diplotene rather than in metaphase, or an increase in ploidy in vas^PH165 oocytes. A very similar nuclear morphology has been observed in spindle (spn) mutant oocytes, which has been interpreted as resulting from a delay in oocyte determination (González-Reyes et al., 1997).

Secondly, vas^PH165 oocytes do not efficiently accumulate at least four oocyte-localized RNAs. In wild-type ovaries, the Bicaudal-D, orb, osk and nos RNAs all accumulate efficiently in the oocyte within the germarium and remain concentrated therein throughout the early stages of oogenesis (Ephrussi et al., 1991; Kim-Ha et al., 1991; Suter and Steward, 1991; Lantz et al., 1992; Wang et al., 1994; Fig. 4A,B,D,E,I,J,M). Oocyte accumulation of these RNAs is much less pronounced in vas^PH165 egg chambers in which Bic-D RNA localization is essentially undetectable (Fig. 4C), and the concentration of orb RNA in the oocyte is only slightly above the levels in the nurse cells (Fig. 4F). In vas^PH165 egg chambers, osk RNA is also not observed to accumulate in the oocyte in the germarium or first vitellarial stages of development. Rather, osk RNA tends to concentrate in a somewhat diffuse manner in the center of the egg chamber (Fig. 4K,L). Finally, loss of vas function has the least pronounced effect on the accumulation of nos RNA into the oocyte, but even in this case oocyte localization is incomplete and poorly maintained (Fig. 4N,O).

Despite the poor localization of orb RNA, ORB protein accumulates efficiently in vas^PH165 oocytes, although somewhat later in oogenesis than in the wild-type (compare Fig. 4H with Fig. 4G). NOS protein expression is also broadly similar to wild-type in vas^PH165 ovaries. The reduced number of developing cysts in vas^PH165 germaria complicates the analysis, but the peak of NOS protein expression in 4-to 8-cell cysts at the posterior of region 1 (Wang et al., 1994) remains apparent in vas^PH165 (Fig. 4P,Q). OSK protein is not detectably translated in early oogenesis in wild-type nor in vas^PH165 egg chambers (Kim-Ha et al., 1995; Rongo et al., 1995; data not shown).

In wild-type oocytes, BIC-D and EGALITARIAN (EGL) proteins colocalize in the oocyte cytoplasm, in a posterior crescent from stages 2-7 (Suter and Steward, 1991; Mach and Lehmann, 1997; Fig. 5A,C,D). In vas^PH165, BIC-D and EGL are efficiently translated and are localized to the oocyte. However, the two proteins are tightly concentrated in a focus which, in higher magnification in both bright-field and confocal microscopy, appears to be adjacent to the oocyte nucleus (Fig. 5B,E, and data not shown).

Most vas^PH165 egg chambers develop into germline cysts containing 15 nurse cells and a posteriorly localized oocyte. These appear normal until about stage 6, whereupon all the nuclei undergo pycnosis and further development ceases (Fig. 6A). Approximately 70% of all vas^PH165 egg chambers examined terminate development in this manner. A minority (approx. 20%) of vas^PH165 oocytes continue developing beyond stage 6; most of these are blocked at stage 9-10, after which the oocyte often loses its integrity at the anterior end, and nurse cell nuclei invade the region of the egg chamber normally occupied only by the oocyte (Fig. 6B). Finally, we found that a small number of oocytes produced by vas^PH165 females complete oogenesis, but that these frequently have duplicated micropyles at both the anterior and posterior ends (Fig. 6C). These eggs also often have dorsal appendage defects consistent with ventralization of the chorion: of 73 eggs analyzed, 14 (19%) lacked dorsal appendages; 48 (66%) had a single fused dorsal appendage; and only 11 (15%) had two dorsal appendages.

The phenotypes of late-stage vas^PH165 oocytes suggest that the GRK signalling pathway (González-Reyes et al., 1995; Roth et al., 1995) may be inactive in these ovaries, and led us to investigate the expression and distribution of grk RNA and protein in vas^PH165. Unlike Bic-D, orb, osk and nos RNA, localization of grk RNA to the oocyte remains highly efficient in vas^PH165 egg chambers (Fig. 7). However, while in wild-type oocytes from about stage 2-3 the distribution of grk RNA forms an obvious posterior crescent (Fig. 7A), grk RNA remains tightly concentrated in an extremely small area in vas^PH165 oocytes (Fig. 7B), similar to the BIC-D/EGL localization pattern in these oocytes (Fig. 5). Later in oogenesis, grk RNA becomes anteriorly localized in both wild-type and vas^PH165 oocytes, although in the mutant its distribution may extend further ventrally (Fig. 7C and D).

Despite the relatively normal accumulation of grk RNA in vas^PH165 oocytes, and unlike the other proteins discussed above, the effects of a loss of vas activity are striking with regard to GRK protein accumulation; essentially no localized GRK is observed in vas^PH165 oocytes (Fig. 7E and F). Furthermore, as measured on western blots, the level of GRK protein is greatly reduced in vas^PH165 ovaries as compared with wild-type (Fig. 7G). Upon overexposure of such western blots a small amount of GRK can be detected in both grk^HK36 and vas^PH165 extracts (Neuman-Silberberg and Schüpbach, 1996; data not shown). As grk is required for specification of dorsal and posterior follicle structures, the duplicated micropyles and dorsal appendage defects found in vas^PH165 eggs (and in eggs produced by other vas alleles; Lehmann and Nüsslein-Volhard, 1991; Schüpbach and Wieschaus, 1991) are likely to be caused by the reduced level of GRK in vas mutants. These results also suggest that VAS may activate GRK translation in wild-type oocytes.

We examined the organization of the microtubule cytoskeleton in vas^PH165 oocytes using a khc:lacZ reporter gene construct (Clark et al., 1994). In the minority of vas^PH165 egg chambers that develop normally through stage 10 and can thus be assessed in this way, the distribution of KHC-β-gal in oocytes is similar to that in the wild-type (Fig. 7H-M), with the exception that the intensity of signal is much lower in mutant egg chambers. The reduced signal may indicate that only a subset of microtubules are correctly organized in vas^PH165 oocytes. An alternative explanation, that VAS may be required for efficient translation of KHC-β-gal protein in oocytes, is formally possible.

We have isolated a null allele of vas, vas^PH165, and its analysis reveals that VAS is involved in many stages of oogenesis, including cystocyte differentiation, oocyte differentiation, and specification of anterior-posterior polarity in the developing cysts. However, the phenotype of vas^PH165 is inconsistent with a total failure of any of these processes, indicating that VAS has a partially redundant role in bringing them about. Some vas-null egg chambers, namely those with two oocytes or with mislocalized oocytes (Fig. 3C,D) are similar to those produced in spn-A, spn-B, spn-C, spn-D and homeless (spn-E) mutants (González-Reyes and St Johnston, 1994; Gillespie and Berg, 1995; González-Reyes et al., 1997). Furthermore, the vas-null oocyte nucleus appears to be very similar to those observed in spn mutants (Fig. 3F, González-Reyes et al., 1997), and defects in anterodorsal accumulation of grk mRNA in stage 9 have also been reported for spn mutants (Gillespie and Berg, 1995; González-Reyes et al., 1997). It is likely that vas and the spn genes have overlapping functions in oocyte determination, and it is noteworthy in this context that homeless, which, like vas, is required in the germline, encodes a DE-H-box protein somewhat related to VAS. A role for the spn gene products in translational activation of grk has been suggested (González-Reyes et al., 1997). Construction of multiple mutants with vas^PH165 and spn mutations will be important in understanding the degree to which their functions overlap.

We have observed a decrease in the number of developing germline cysts in vas^PH165 ovaries, and a similar, but more severe, phenotype occurs in ovaries from females null for nos function (Fig. 2; Wang et al., 1994; Curtis et al., 1997). Because of the similarities between this aspect of the vas and nos mutant phenotypes, and because VAS has been implicated in translational activation of nos in the pole plasm (Gavis et al., 1996; Dahanukar and Wharton, 1996), it is tempting to speculate that VAS and NOS function in the same pathway to promote germline cyst development, and that VAS may activate (but not be absolutely essential for) translation of NOS in this stage of germline development as well. Our immunostaining for NOS neither supports nor refutes this possibility. While we see no obvious lack of NOS in vas^PH165 germaria (Fig. 4Q), it is presently impossible to detect differences in NOS levels in stem cells or cystoblasts, as even in wild-type ovaries these cells are not detectably stained with NOS antibody (Wang et al., 1994; Fig. 4Q). The higher level of NOS in 4-8 cell cysts, which appears to be unaffected by vas^PH165, would not be responsible for cyst development, as it occurs at a later developmental stage than that at which the cysts are blocked in nos^RC ovaries.

The reduction in the number of developing germline cysts we observe in vas and nos null mutants could result from a failure of the pole cells to migrate to the gonad during embryogenesis, a failure of the germ cells to regulate the cell cycle or to respond to cell division signals, or from inappropriate differentiation of stem cells into cystoblasts, depleting the germarium of proliferative stem cells. Of these three possibilities, we do not think the defect involves pole cell migration, as we never observe completely agametic ovaries in vas or nos mutants. Agametic ovaries are observed even in cases where pole cell migration is incompletely blocked, such as in the progeny of AS-Pgc females (Nakamura et al., 1996). Direct assessment of the role of zygotic VAS in pole cell migration is confounded by the fact that maternal VAS perdures until embryonic stage 15-16, long after pole cell migration is complete.

In wild-type flies, vas is abundantly expressed in the female germline, and, while VAS is concentrated at the posterior pole of the oocyte from stage 10 of oogenesis, easily detectable levels of VAS are present uniformly throughout the early embryo (Lasko and Ashburner, 1990). These uniform levels are sufficient for ectopic posterior somatic structures to develop in circumstances where OSK is present outside of the pole plasm, for instance in females overexpressing osk or carrying the P[oskBRE^–] transgene (Smith et al., 1992; Kim-Ha et al., 1995). For embryos produced by females carrying a Bic-D dominant allele, ectopic posterior segments form dependent on vas activity (Mohler and Wieschaus, 1986), but ectopic pole cells do not form, suggesting that the threshold level of VAS necessary for determination of the posterior soma is less than that necessary to induce formation of pole cells (Wharton and Struhl, 1989). A much lower threshold level of VAS is required in early oogenesis than for either later function, since vas^LYG2 homozygotes, which produce very low levels of unaltered VAS (Fig. 1C), lay a far larger number of eggs than do vas^PH165 homozygotes, and show defects in oogenesis at lower penetrance and of less severity than for the null (S. S. and P. L., unpublished observations).

A comparison of the distribution of RNAs and proteins in vas-null egg chambers suggests that localization of RNAs, such as Bic-D and orb, is largely redundant for targeting their proteins to the oocyte. Furthermore, as most vas-null egg chambers proceed through development at least to stage 6, despite the very weak localization of such RNAs, it can also be argued that RNA localization is relatively unimportant for the functions of these genes. Therefore translational control, or protein targeting by means independent of RNA localization, must have a critical role in establishing the correct distribution of these molecules. This is perhaps similar to the interplay between translational control and RNA localization apparent in later oogenesis and in embryos; targeting of OSK and NOS to the posterior pole is accomplished through a combination of RNA localization and translational repression of the unlocalized RNA (Kim-Ha et al. 1995; Markussen et al., 1995; Rongo et al., 1995; Gavis et al., 1996; Smibert et al., 1996), and, at least for nos, it has been argued that RNA localization comes about as a consequence of translational regulation (Dahanukar and Wharton, 1996).

In vas^PH165 ovaries, grk RNA and the BIC-D/EGL complex still localize efficiently to the oocyte, but these molecules are concentrated in a tight focus rather than being distributed more widely in the oocyte cytoplasm. This suggests that in wild-type egg chambers, localization of these molecules may occur in two steps: a vas-independent step, which concentrates them in the oocyte, and a second step which distributes them to a specific region within the oocyte, depending directly or indirectly on vas activity. Alternatively, loss of vas function results in the formation of a novel structure or structures in the oocyte cytoplasm which trap grk RNA and the BIC-D/EGL complex. The distribution of BIC-D and EGL is unaffected in grk^HK36 ovaries, indicating that the abnormal localization of these two proteins in vas^PH165 oocytes is not a consequence of the failure to accumulate GRK in vas^PH165 (data not shown). We have found that vas-null mutations affect some of the same processes as do Bic-D and egl mutations; namely, oocyte determination (in rare 16-nurse-cell and two-oocyte egg chambers) and establishment of dorsal-ventral polarity (Swan and Suter, 1996; Mach and Lehmann, 1997; this paper). In this context, it is possible that these vas phenotypes are a result of the altered distribution of BIC-D/EGL we observed. The altered distribution of grk RNA within the early oocyte is similar to that observed in maelstrom (mael) mutants which affect the organization of oocyte microtubules (Clegg et al., 1997). However, unlike in vas-null egg chambers in which many RNAs, such as osk, are poorly transported from the nurse cells to the oocyte (Fig. 4), mael only affects RNA distribution within the oocyte (Clegg et al., 1997). Furthermore, posterior localization of KHC-βGAL in vas-null oocytes (Fig. 7) suggests that the organization of the microtubule cytoskeleton is at least partially maintained.

In the yeast S. cerevisiae, a protein highly similar to VAS, DED1p, has recently been implicated in translation initiation (Chuang et al., 1997; de la Cruz et al., 1997). While the loss-of-function experiments presented here cannot distinguish whether VAS is required directly or indirectly for GRK accumulation, a simple model which takes into account the molecular nature of VAS and explains the reduction of GRK protein expression in vas-null ovaries is that VAS interacts directly with grk RNA and activates its translation. This is consistent, as well, with the striking reduction we observed in GRK accumulation in vas-null oocytes despite the high level of grk RNA in these oocytes. Importantly, other RNAs, such as Bic-D and orb, whose localization to the oocyte is more severely perturbed in vas-null ovaries, are still translated at levels comparable to wild-type in vas-null oocytes. Nevertheless, demonstration of a direct effect of VAS on grk translation must await more detailed experimentation. Bacterially expressed VAS has been demonstrated to bind RNA in vitro (Liang et al., 1994). However, no specificity for RNA binding by VAS in vitro has thus far been demonstrated, though grk was not among the RNAs tested in these in vitro binding experiments. There appears to be elaborate posttranscriptional regulation of GRK in the developing oocyte, involving the activities of numerous genes. In addition to vas, the genes aubergine and encore have been implicated as essential for efficient GRK accumulation (Wilson et al., 1996; Hawkins et al., 1997).

Duplication of anterior vitelline membrane structures occurs in strong grk mutant alleles, and GRK signalling has been implicated in the specification of posterior polar follicle cells (González-Reyes et al., 1995; Roth et al., 1995). We demonstrate in this paper that many of the oocytes that complete oogenesis in a loss-of-function vas mutant have duplicated micropyles and/or fused or reduced dorsal appendages, and four other vas alleles (PW72, QS17, AQB3, RG53) have previously been shown to produce eggs with fused and reduced dorsal appendages when hemizygous over Df(2L)osp29 (Schüpbach and Wieschaus, 1991). All of these phenotypes resemble those observed in grk mutants, making it likely that these aspects of the vas mutant phenotype are brought about by a reduction of GRK activity, particularly since we show that GRK accumulation is negligible in vas-null oocytes. However, aspects of the vas-null phenotype which manifest themselves earlier in oogenesis, such as the reduced number of developing cysts, the defects in RNA localization to the oocyte, and the aberrant localization of BIC-D and EGL within the oocyte, are not found in grk mutant ovaries (Fig. 2; S.S. and P.L., unpublished results). This suggests that VAS may regulate the activity of numerous genes. Evidence similar to that presented here implicates VAS in translational activation of Short OSK in the pole plasm (Markussen et al., 1995). Furthermore, it is possible that VAS regulates translation of other, presently unknown, RNAs whose products are involved in establishing cytoskeletal polarity in the early cyst and/or in localizing RNAs to the oocyte.

We are grateful to Takahiro Akiyama, Celeste Berg, Ruth Lehmann, John Roote, and Trudi Schüpbach for fly stocks, to Howard Lipshitz for α-ADD-87 antibodies, to Paul Schedl for α-ORB antibodies, to Shira Neuman-Silberberg and Trudi Schüpbach for α-GRK antibodies, and to Robin Wharton for α-NOS antibodies. Thanks also to Antoine Guichet, Pavel Tomancak, Anne Ephrussi, Jean-Louis Couderc, Don Rio, Lin Yue, Frank Laski, and the Berkeley Drosophila Genome Project for sharing unpublished data and reagents, to an anonymous reviewer for comments on the manuscript, and to Cate Lévesque for fly food. This research was supported by operating grants to P. L. from the National Cancer Institute of Canada (NCIC) with funds from the Canadian Cancer Society, the Natural Science and Engineering Research Council of Canada (NSERC), and by a team grant from the Fonds pour la formation de chercheurs et l’aide à la recherche du Québec (FCAR). A. N. was supported by a Japan Society for the Promotion of Science postdoctoral fellowship, and a Medical Research Council of Canada postdoctoral fellowship. A. S. and S. S. acknowledge graduate bursaries from FCAR. P. L. and B. S. are Research Scientists of the National Cancer Institute of Canada.