Evidence for a highly selective RNA transport system and its role in establishing the dorsoventral axis of the Drosophila egg

The dorsoventral pattern of the Drosophila embryo is generated by the sequential action of eighteen maternal effect genes (reviewed in Rushlow and Arora, 1990; Govind and Steward, 1991). The first six genes of this hierarchy, K10, capu, spir, cornichon (cni), gurken (gk) and torpedo (top), act during oogenesis to specify somatic cell fates along the dorsoventral axis of the egg chamber. The egg chamber consists of the oocyte, 15 nurse cells, which are connected to the oocyte’s anterior end via cytoplasmic bridges, and a surrounding epithelium of somatic follicle cells (King, 1970). Through the action of the remaining dorsoventral maternal genes, dorsal and ventral follicle cells differentially signal the oocyte, thus establishing the ventral and dorsal identities of the future embryo.

Genetic mosaic analyses show that K10, capu, spir, cni and gk are required exclusively in germ cells (Wieschaus, 1979; Schüpbach, 1987; Manseau and Schüpbach, 1989; Schüpbach and Wieschaus et al., 1991), indicating that follicle cell differentiation is dependent on factors synthesized in germ cells. The top gene appears to encode the receptor for one such factor (ligand); top is required exclusively in somatic cells (Schüpbach, 1987), and top protein, which is found on the oocyte-facing surface of all follicle cells (R. Schweitzer, N. Zak and B. Shilo, personal communication), has extensive sequence homology to the mammalian epidermal growth factor receptor (Price et al., 1989; Schejter and Shilo, 1989). Females lacking wild-type top activity produce ventralized egg chambers, i.e., egg chambers whose dorsal and ventral follicle cells behave and differentiate like the ventral follicle cells of wild-type egg chambers (Schüpbach, 1987). This result shows that top protein is required for dorsal-like follicle cell differentiation, but not for ventral-like follicle cell differentiation. Females lacking wild-type K10, capu or spir activity produce dorsalized egg chambers, indicating that these genes are required to repress the delivery of top ligand to ventral follicle cells (Wieschaus, 1979; Manseau and Schüpbach, 1989). The alternative view that K10, capu and spir encode factors that actively promote ventral-like follicle cell differentiation is disfavored, since ventral follicle cells differentiate normally in females doubly mutant for top and K10, capu or spir (Schüpbach, 1987; Manseau and Schüpbach, 1989).

Although it is not clear how top ligand is specifically targeted to dorsal follicle cells, it is tempting to speculate, as previously suggested, that the necessary asymmetry in the system comes from the dorsal migration of the oocyte nucleus at stage 9 of oogenesis (Parks and Spradling, 1987; Manseau and Schüpbach, 1989). The first visible sign of dorsal versus ventral follicle cell differentiation appears during stage 10. In this model, K10 protein is thought to repress (partially) the transcription of the top ligand in the oocyte nucleus, such that the ligand is synthesized in amounts only sufficient to induce those follicle cells that lie closest (i.e., dorsal follicle cells) to the ligand source. Consistent with the idea that K10 encodes a transcription factor, K10 protein is localized to the oocyte nucleus and contains structural motifs similar to several procaryotic and eucaryotic transcription factors (Prost et al., 1988).

In this paper we investigate the mechanism of K10 RNA accumulation in the oocyte. In contrast to a previous study, which reported that K10 RNA is synthesized directly in the oocyte (Haenlin et al., 1987), our data provide strong evidence that K10 RNA is synthesized principally in nurse cells and, beginning at stage 2, rapidly transported into the oocyte. K10 RNA transport must be a selective process, since most of the RNAs synthesized in nurse cells are stored there until stage 11 (Mahowald and Tiefert, 1970; Mahowald and Kambysellis, 1980). Sequences responsible for K10 RNA transport have been mapped to the 3′ end of the gene, which includes a long ∼ 1400 bp transcribed nontranslated region with potential to form multiple stemloop structures. K10 genes lacking this sequence produce wild-type or near wild-type amounts of RNA, but the RNA is not transported into the oocyte until stage 11.

An unexpected finding of the studies presented here is that K10 RNA becomes localized to the anterior margin of the oocyte at stage 8, just before or coincident with the onset of K10 protein synthesis. In capu and spir mutant ovaries, K10 RNA accumulates in the oocyte on schedule, but never localizes to the anterior margin. This result indicates that capu and spir act upstream of K10 in dorsoventral patterning and that the localization of K10 RNA to the anterior margin of the oocyte is important for K10 function, perhaps facilitating the subsequent localization of K10 protein to the oocyte nucleus.

The wüd-type stock was Oregon R. The K10^LMOO RNA null mutation was provided by T. Schüpbach. The capu^RK12, capu^G7, spir^PJ56 mutations are described in Manseau and Schüpbach (1989). ry⁵⁰⁶ flies served as recipients for P element transformation experiments. All mutant chromosomes carried visible markers for identification. Flies were reared under standard conditions. Staging of egg chambers was according to King (1970).

Embryos were prepared for microinjections as previously described (Cohen and Meselson, 1985). Test DNAs (Fig. 1) were cloned into the Carnegie 20 transformation vector (Rubin and Spradling, 1983) and injected at a concentration of ∼ 300 μ g/ml. Transposase was provided by the helper plasmid, pl37rwc (Rubin and Spradling, 1982; Spradling and Rubin, 1982; Cohen and Meselson, 1985), which was coinjected with the test constructs at a concentration of ∼ 75 μ g/ml. Between 3 and 8 transformed lines were generated and analyzed for each construct. Transformants were maintained as homozygous or balanced heterozygous stocks.

Ovaries were dissected in PBS* (Phosphate-buffered saline containing 2 mM MgCb, 2 mM spermidine, 0.02% sodium deoxycholate and 0.02% Nonidet P-40) and fixed for 10 minutes in PBS* containing 2% freshly made paraformaldehyde. Fixed tissues were rinsed with PBS* and then placed in PBS containing 0.2% X-gal, 1.1 mM K₃Fe(CN)₆ and 1.1 mM K4Fe(CN)₆ and 2 mM MgCb (Glaser et al., 1986). Staining reactions were carried out at 37°C until the desired intensity was obtained, typically 15-30 minutes. Stained tissues were mounted in Polyaquamount (Polysciences, Inc.) and photographed with Nomarski optics.

The KZK construct (Fig. 1) was made in several steps to yield the following: K10 sequences from – 483 bp to +333, where + 1 corresponds to the first nucleotide of K10 mRNA, and +318 corresponds to the first nucleotide of the translation start site; E. coli lacZ sequences from just downstream of the translation start site to just downstream of the translation stop site; K10 sequences from +2628 to +4861, where the K10 protein coding region ends at nucleotide body, and where the poly(A) addition site corresponds to +3997 (ref. 8). All K10 sequences derive from plasmid Pst22 (kindly provided by E. Mohier). K10 coordinates come from Prost et al. (1988). The lacZ sequence derives from plasmid pMC1871 (Pharmacia). The hsp26 nurse cell enhancer-SgsJ promoter DNA fragment of the 26gZS and 26gZK constructs consists of two copies of the hsp26 sequence – 543 to – 373 linked to the Sgs3 sequence – 127 to +38 (L. Frank et al., 1991). The Sgs3 sequence contains a translation start site (Garfinkel et al., 1983). The Sgs8 sequence, which constitutes the 3′ end of the KZS, KKZS, and 26gZS constructs (Fig. 1), derives from plasmid ploxba (kindly provided by V. Corbin) and includes a functional poly(A) addition site (Garfinkel et al., 1983). Further details of these constructs will be made available upon request.

Probe labeling with digoxigenin-dUTP and whole-mount in situ hybridization was carried out according to Tautz and Pfeifle (1988), except that dimethyl sulfoxide was added to the initial fixation buffer to a final concentration of 10%. In addition, the length of proteinase K treatment was increased to 30 minutes. K10 RNA was detected using two different random-primed probes, each of which gave identical results.

For the experiment presented in Fig. 4, the probe corresponded to the 3′ end of the K10 mRNA (nucleotide residues +2628 to +3997). The other K10 probe derived from the 5′ end of the K10 transcription unit. The lacZ hybridization probe was generously provided by P. Feinstein. Hybridization was visualized using an alkaline phosphatase-conjugated anti-digoxigenin antibody (Boehringer Mannheim) and color development with X-phosphate according to manufacturer’s instructions. Stained tissues were mounted and photographed as described above for X-gal-stained tissues.

To identify sequences responsible for the early accumulation of K10 RNA in the oocyte, we employed a transgenic fly assay system. We started with a 5.3 kb DNA fragment, consisting of K10 genomic and cDNA sequences previously shown to rescue completely K10^- flies upon P element transformation (Haenlin et al., 1987) . We reintroduced this fragment into the Drosophila germ-line after substituting the structural portion of the bacterial lacZ gene for the bulk of the K10 protein coding region to create KZK (Fig. 1). The ovarian expression pattern of the KZK gene was analyzed in several independently transformed lines using a histochemical assay for β -galactosidase (β -gal) activity. As seen in Fig. 2, β -gal-staining activity is greatly enriched in oocytes as compared to nurse cells, and is not detected in somatic follicle cells. β -gal staining activity is first detectable in oocytes at ∼ stage 6 and steadily increases in intensity thereafter. This expression pattern is very similar to that reported for authentic K10 protein (Prost et al., 1988) and we therefore conclude that the KZK construct contains most or all of the cis-acting sequences required for wildtype K10 expression during oogenesis.

To delimit further the sequences required for wildtype K10 expression, we replaced the 3′ K10 portion of the KZK gene, which includes ∼ 1400 bp of transcribed non-translated sequence and ∼ 800 bp of 3′ flanking DNA, with the poly(A)/transcription termination region of the Drosophila Sgs8 gene to create KZS (Fig. 1). In contrast to KZK transformants, the oocytes of KZS transformants exhibit no β -gal-staining activity until stage 11, when nurse cells indiscriminately donate their entire cytoplasmic contents to the oocyte (Fig. 2 and data not shown). Prior to stage 11, the β -gal-staining activity of KZS transformants is restricted to nurse cells (Fig. 2). Thus, the 3′ end of the K10 gene is required for the early (pre-stage 11) accumulation of K10 gene products in the oocyte.

Because the KZK. and KZS genes encode identical proteins (Fig. 1), the difference in the ovarian β -gal staining patterns of KZK and KZS transformants should reflect differences in the distribution patterns of KZS and KZK RNA. Indeed, in situ hybridization analyses show that prior to stage 11 KZK RNA accumulates predominantly in oocytes, while KZS RNA accumulates predominantly in nurse cells (Fig. 3). It is also clear from such analyses that the absolute amount of KZK RNA in pre-stage 11 oocytes is much greater than the absolute amount of KZS RNA in prestage 11 oocytes. This is true even though Northern blots indicate that the total amount of KZS RNA in the ovaries of KZS transformants is several fold higher than the total amount of KZK RNA in the ovaries of KZK transformants (data not shown). Whether the difference in the steady state concentrations of KZK and KZS RNAs reflect differences in the efficiency at which these genes are transcribed or differences in the stability of the respective RNAs is not clear.

To test whether the 3′ K10 sequence can cause a heterologous RNA to accumulate in the oocyte prior to stage 11, we substituted this sequence for the 3′ end of an hsp26-lacZ-Sgs8 fusion gene (26ZS) to create 26ZK. While the β -gal staining activity of 26ZS transformants is concentrated in nurse cells prior to stage 11, we find that β -gal-staining activity of 26ZK transformants is concentrated in oocytes during such stages (Fig. 2). We interpret this result to mean that the 3′ K10 sequence is sufficient to cause a heterologous RNA to accumulate in the oocyte prior to stage 11.

The 3′ end of the K10 gene could control the early accumulation of K10 RNA in the oocyte at the level of transcription initiation and/or post-transcriptionally. In the case of control at the level of transcription initiation, the 3′ sequence would have to contain sequences that both promote transcription in the oocyte and repress transcription in nurse cells, since, as noted above, a deletion of the 3′ sequence leads both to an increase in the amount of K10 RNA found in nurse cells and a decrease in the amount of K10 RNA found in oocytes. To distinguish between control at the level of transcription initiation and post-transcriptionally, we investigated the positional requirements of the 3′ sequence. If the sequence functions post-transcriptionally, it should work only when incorporated into the transcribed portion of a gene. Conversely, if the sequence functions at the level of transcription initiation, it should work when placed outside or inside of a transcription unit. As seen in Fig. 2, the 3′ K10 sequence has no effect on the expression pattern of the KZS gene when placed 5′ to the promoter (compare the β -gal-staining patterns of the KKZS and KZS genes). We tentatively conclude from this experiment that the 3′ K10 sequence functions post-transcriptionally.

To rule out the possibility that the 3′ K10 sequence affects oocyte expression at the level of transcription initiation, but only when placed 3′ to a gene, we moved the sequence back to its original location, 3′ to the lacZ gene, but stripped the promoter down to a minimal sequence (sequences from – 127 to +38, which includes a TATA box and transcription start site, were retained). Transformants carrying this construct (gZK) exhibit no β -gal staining whatsoever (Fig. 1 and data not shown). Together with our previous findings, we conclude that the 3′ K10 sequence functions post-transcriptionally.

The simplest interpretation of the data presented above is that K10 is transcribed principally in nurse cells, and that the resulting transcripts are rapidly transported into the oocyte. Although previous in situ hybridization experiments using ³⁵S probes only detected K10 RNA in oocytes (Haeiüin et al., 1987), we thought that we might be able to detect K10 RNA in nurse cells prior to their transport into the oocyte with digoxigenin-labeled probes, which in our hands are more sensitive than radiolabeled probes. To control for background labeling, we also examined ovaries from flies carrying various K10 mutant alleles. One such allele, K10^LM00, proved to be an RNA null (Fig. 4). During early oogenesis (e.g. stage 1) in wild-type ovaries, we find that K10 transcripts are equally abundant in nurse cells and oocytes (Fig. 4). Thereafter, K10 RNA decreases in abundance in nurse cells, approaching background levels by stage 3, and increases in abundance in the oocyte through stage 8, when it becomes localized to the periphery of the oocyte’s anterior margin. K10 RNA remains localized to the oocyte’s anterior margin at least through the end of stage 10. (K10 RNA cannot be detected reliably in stage 11 and older oocytes using the whole-mount hybridization procedure employed here, due to the secretion of the chorion by the somatic follicle cells). K10 RNA reappears in nurse cells at ∼stage 9 and accumulates there until stage 11, when it is apparently transported into the oocyte with the rest of the nurse cell cytoplasm.

We also examined K10 RNA distribution in capu and spir mutant ovaries (Fig. 4 and data not shown). Two strong mutant alleles of capu (capu^RK12, capu^G7) and one of spir (spir^PJ56) were examined (Manseau and Schüpbach, 1989). In each case, K10 RNA accumulates normally through stage 7, i.e., the RNA preferentially accumulates in the oocyte. In contrast to what is seen in wild-type oocytes, however, K10 RNA never localizes to the anterior margin of capu and spir mutant oocytes. Instead the RNA remains uniformly distributed throughout the ooplasm or, less frequently, organizes itself into several, randomly distributed, diffuse islands. Weak alleles of capu and spir, which give anteroposterior pattern defects, but no dorso ventral pattern defects were not examined. Mutations in other dorsoventral group genes also were not examined, since epistasis tests indicate that all other such genes act downstream of K10.

The first visible sign of dorsoventral patterning during Drosophila development is the migration of the oocyte nucleus towards the dorsal surface of the cell at stage 9 of oogenesis (King, 1970). During stage 10, the follicle cells lying on the dorsal surface of the oocyte differentiate from those lying on its ventral surface (Margaritis et al., 1980; Parks and Spradling, 1987). Such differentiation is dependent on wild-type K10 gene activity (Wieschaus, 1979). K10 protein specifically accumulates in the oocyte nucleus beginning at stage 8 or 9 and increases in abundance through stage 12 (Prost et al.,1988), where it has been suggested to partially repress the synthesis of factors that stimulate dorsal-like follicle cell differentiation (Manseau and Schüpbach, 1989). As a result of such repression, only those follicle cells that lie closest to the oocyte nucleus are stimulated.

Here we show that the accumulation of K10 protein in the oocyte nucleus is a multistep process involving: (1) the synthesis of K10 RNA in nurse cells, (2) the rapid transport of K10 RNA from nurse cells into the oocyte, (3) the localization of K10 RNA to the anterior margin of the oocyte and (4) the translation of K10 RNA and the localization of KI0 protein to the oocyte nucleus.

Tritiated uridine incorporation studies indicate that the vast majority of the RNAs present in the mature egg and early embryo originate in nurse cells (Mahowald and Tiefert, 1970). This is not a surprising finding, since nurse cells replicate their DNA throughout oogenesis, reaching an average chromosome number of over 1,000 by stage 10, while the oocyte remains diploid (Mahowald and Kambysellis, 1980). In addition, there is some evidence that oocyte transcription is actively repressed during stages 5-8 (Mahowald and Tiefert, 1970). Based on these findings alone, one might expect that the preferential accumulation of K10 transcripts in the oocyte during oogenesis results from K10 transcription in nurse cells and the transport of its transcripts into the oocyte. The data presented here meet two key predictions of this model, and argue against the alternative view (Haenlin et al., 1987) that most K10 transcripts are synthesized directly in the oocyte. Specifically, we show that the 5′ end of the K10 gene contains transcriptional control sequences active in nurse cells, and that the 3′ end of the gene contains sequences necessary for the preferential accumulation of K10 transcripts in the oocyte. Moreover, the 3′ sequences function only when incorporated into the transcribed portion of the gene.

Presumably the early (pre-stage 11) transport of K10 RNA into the oocyte is important for the early (prestage 11) accumulation of K10 protein in the oocyte nucleus, and hence, for K10 function. Consistent with this idea, we have recently shown (unpublished) that K10 genes specifically deleted for sequences required for the early accumulation of its RNA in the oocyte, do not, or only weakly, rescue K10 mutants.

The transport of K10 RNA into the oocyte must be a fairly selective process, since most RNAs are retained by nurse cells until stage 11 (Mahowald and Tiefert, 1970). The RNAs of the bicoid (bed) and oskar (osk) genes also preferentially accumulate in the oocyte prior to stage 11 (St. Johnston et al., 1989; Kim-Ha et al., 1991; Ephrussi et al., 1991) and may, therefore, be transported into the oocyte via the same system that transports K10 RNA into the oocyte. If so, mutations that disable the system should disrupt anteroposterior patterning as well as dorsoventral patterning; bed and osk are required for anterior and posterior patterning, respectively (Nüsslein-Volhard et al., 1987). We have examined several mutations that disrupt dorsoventral, or both dorsoventral and anteroposterior patterning, but have not found any defects in K10 RNA transport (data not shown). Perhaps the genes that regulate K10 RNA transport and, possibly, bed and osk RNA transport, also regulate the transport of molecules required for oocyte differentiation. Consistent with this idea, we have found that loss-of function Bic-D and egal mutations lead to the uniform accumulation of K10 transcripts in all 16 germ cells (data not shown, also see Suter and Steward, 1991; Mohler and Wieschaus, 1986). Bic-D loss-of-function mutations also lead to the uniform accumulation of osk RNA in all 16 germ cells, although prior to egg chamber formation, osk RNA temporarily localizes to a single germ cell, i.e., to the presumptive oocyte (Suter and Steward, 1991). Also, gain-of-function Bic-D mutations interfere with the translocation of osk RNA from the oocyte’s anterior to posterior end during stage 9 (Kim-Ha et al., 1991; Ephrussi et al., 1991). Nevertheless, it remains unclear whether the absence of efficient RNA transport in BicD loss-of-function mutants reflects a direct involvement of Bic-D in such transport, or simply the fact that no oocyte is made in such mutants.

Beginning at ∼stage 8, K10 RNA becomes localized to the anterior margin of the oocyte, where it remains through at least the end of stage 10. Interestingly, bed and osk RNAs also localize to the oocyte’s anterior end at ∼stage 8 (St Johnston, 1989; Kim-Ha et al., 1991). In the case of osk RNA, such localization is only transient; osk RNA stably relocalizes to the oocyte’s posterior pole during stage 9 (Kim-Ha et al., 1991; Ephrussi et al., 1991). Sequences required for the anterior localization of bed mRNA have been mapped to a putative stem loop structure at the 3′ end of the molecule (Macdonald and Struhl, 1988; Macdonald, 1990). Computer-assisted analyses of K10 RNA secondary structure reveal multiple possible stem-loop structures, but none of these are similar in primary sequence to the bed localization sequence (data not shown). Similar analyses of osk RNA also reveal no striking sequence similarities to the bed localization sequence (data not shown). It is likely, therefore, that the genes that regulate bed RNA localization are different than those that regulate K10 and osk RNA localization. Consistent with this idea, mutations in exuperantia and swallow, which disrupt bed RNA localization (St. Johnston et al., 1989), specifically disrupt anterior pattern formation (Frohnhofer and Niisslein-Volhard, 1987). (As alluded to above, mutations that interfere with K10 and osk function would be expected to interfere with dorsoventral and posterior pattern formation, respectively). Similarly, the capu and spir mutations that disrupt osk and K10 RNA localization (Kim-Ha et al., 1991; Ephrussi et al., 1991; this paper), specifically interfere with dorsoventral and posterior pattern formation (Manseau and Schüpbach, 1989). The fact that capu and spir mutations disrupt osk and K10 RNA localization raises the possibility that these RNAs are localized to the oocyte’s anterior end in a similar fashion.

The finding that K10 RNA does not localize to the oocyte’s anterior end in capu and spir mutants suggests that such localization is important for K10 function. The localization of K10 RNA to the oocyte’s anterior end could, for example, facilitate K10 protein synthesis and/or the subsequent localization of this protein to the oocyte nucleus, which fies in the anterodorsal quadrant of the oocyte. At present, however, we can rule out neither the possibility that K10 function is independent of K10 RNA localization within the oocyte, nor the possibility that capu and spir mutations disrupt dorsoventral pattern formation in ways other than the mislocalization of K10 RNA.

We thank Trudi Schüpbach and Christiane Niisslein-Volhard for sending mutant strains, Paul Macdonald for sharing unpublished data, and Vicki Corbin, Jane Dodd, Richard Mann, Gary Struhl, Robin Wharton, and members of the Cohen lab for comments on the manuscript. This work was funded by grants from the NIH and the NSF to R.S.C.