Multiple steps in the localization of bicoid RNA to the anterior pole of the Drosophila oocyte

The anterior-posterior pattern of the Drosophila embryo is determined in response to maternal factors that are deposited in the egg during oogenesis. Three groups of maternal genes are required to specify distinct regions of this pattern; the head and thorax (the anterior group), the abdomen (the posterior group) and the acron and telson (the terminal group) (reviewed in Niisslein-Volhard and Roth, 1989; Nüsslein-Volhard et al. 1987). The anterior portion of the body plan depends on the product of the bicoid (bed) gene (Frohnhöfer and Nüsslein-Volhard, 1986). In bicoid^- embryos, the head and thorax are absent, and the abdominal fate map expands anteriorly. In addition, a duplicated telson forms at the anterior end of the embryo. Transplantation experiments have shown that bcd⁺ activity is localized to the anterior pole of the egg, and when transplanted to more posterior positions, can induce the formation of anterior structures at ectopic sites (Frohnhöfer and Nüsslein-Volhard, 1986; Frohnhöfer et al. 1986).

In agreement with the results of the transplantation experiments, bed RNA is localized to the anterior pole of the egg (Frigerio et al. 1986; Berleth et al. 1988). The protein that is translated from this RNA forms a concentration gradient which extends throughout the anterior two thirds of the embryo (Driever and Niisslein-Volhard, 1988A). This protein gradient appears to determine anterior positional values in a concentrationdependent manner, since changes in the maternal bcd⁺ gene dosage produce complementary shifts in both the protein gradient and the fate map (Driever and Niisslein-Volhard, 1988b). Further support for the role of bed protein as the anterior morphogen comes from the results of experiments in which bed RNA synthesized in vitro is injected into early embryos. When the RNA is injected into the middle of a recipient embryo, it can induce the development of ectopic head and thoracic structures (Driever, Siegel and Niisslein-Volhard, in preparation). The most anterior pattern elements form closest to the site of injection, with more posterior (thoracic) structures developing on either side. This result indicates not only that the bed protein determines the anterior quality of the structures that develop, but also that the slope of the protein gradient specifies the polarity of the pattern. In addition, the ability of bed RNA to organize anterior pattern in the middle of the embryo shows that no other anteriorly localized molecules are required for this process.

The bicoid protein contains a homeodomain, suggesting that bicoid encodes a DNA-binding protein (Fri-gerio et al. 1986). This makes it attractive to suppose that bed protein specifies anterior positional values by activating or repressing zygotic target genes in a concentration-dependent manner. Genetic and molecular evidence indicates that bed regulates the anterior zygotic expression of the gap gene, hunchback (Lehmann and Nüsslein-Volhard, 1987; Schröder et al. 1988; Tautz, 1988). bed protein binds to several sites in the hunchback promoter, and acts as a transcriptional abtivator of hunchback (Driever and Nüsslein-Volhard, 1989). Driever et al. (in press) have made constructs containing synthetic-bcd-binding sites fused to the hsp70 promoter and a reporter gene. When transformed into flies, the bed-binding sites drive the anterior expression of the reporter gene in the embryo. The size of this anterior domain of expression is reduced when binding sites with lower affinities for the bed protein are used. The observation that promoters with low-affinity bed-binding sites are only activated at high protein concentrations suggests a model for how the bed protein gradient could activate several zygotic target genes in distinct anterior domains, and thereby determine several levels of anterior development.

The formation of the wild-type protein gradient, and thus the determination of a normal anterior pattern, depends upon the localization of bed RNA to the anterior pole. Mutations in the maternal genes exuper-antia (exu) and swallow (sww) disrupt this localization during oogenesis, and lead to an almost uniform distribution of bed RNA in the early embryo (Berleth et al. 1988; Stephenson et al. 1988). Both of these mutations result in phenotypes in which anterior structures of the head are deleted, and the thoracic region is expanded (Frohnhôfer and Niisslein-Volhard, 1987). Embryos derived from females mutant for the posterior group gene staufen show similar but weaker head defects, in addition to the abdominal deletions characteristic of all mutations in this class (Schiipbach and Wieschaus, 1986; Nüsslein-Volhard et al. 1987; Lehmann, 1988). These embryos also show an anterior shift in the fate map, which can be seen in a shift in the position of the cephalic furrow and the first fushi tarazu stripe (Schiipbach and Wieschaus, 1986; Carroll et al. 1986; Lehmann, 1988). The observation that the bed protein gradient is shallower in the mutant embryos suggests that staufen mutations may also alter the distribution of bed RNA (Driever and Niisslein-Volhard, 1988b).

In this report, we use a non-radioactive, enzyme-linked in situ technique developed by Tautz and Pfeifle (1989) to examine the process of bed RNA localization during oogenesis and early embryogenesis. This technique has the advantage that one can perform hybridizations on wholemount preparations. We have used these wholemount stainings to make direct comparisons between the bed RNA distribution and the protein distribution, as revealed by antibody stainings.

The localization of bed RNA in wild-type embryos has previously been described by Frigerio et al. (1986) and Berleth et al. (1988). We have repeated these investigations in order to gain a clearer understanding of the three-dimensional distribution of bed RNA during the first stages of embryogenesis. In very early (stage 1) embryos, bed RNA staining resembles a flattened ball which is closely apposed to the anterior pole, and which frequently occupies a slightly dorsal position (Fig. 1A). The RNA does not seem to be specifically bound to the cortex of the egg, since much of it is in the interior. As development proceeds through pole cell formation (Fig. IB) to syncytial blastoderm (Fig. 1C), most of the RNA becomes localized to the periphery, in the clear cytoplasm that surrounds each nucleus. This movement to the cortex sometimes results in a slight posteriorwards shift in the RNA distribution. By early nuclear cycle 14, bed RNA begins to disappear (Fig. 1D) and midway through cellularization the signal is no longer detectable (Fig. 1E).

Fig. 2 compares the distributions of bed RNA and protein at the syncytial blastoderm stage. The levels of RNA and protein were quantified using the procedure described by Driever and Niisslein-Volhard (1988a), and these results are presented graphically in Fig. 3. These measurements clearly demonstrate that the RNA is more tightly localized to the anterior end of the embryo, than is the protein, bed RNA forms a steep gradient at the anterior end of the embryo, in which 90% of the RNA is restricted to the region anterior to 82% egg length (0% EL is the posterior pole). In contrast, bed protein is distributed in a much shallower gradient, in which only 57% of the protein falls within this anterior region. The protein must therefore move posteriorly after it has been synthesized. One of the functions of bed protein is to activate the anterior zygotic expression of hunchback (Schrôder et al. 1988; Tautz, 1988; Driever and Niisslein-Volhard, 1989). This anterior hunchback domain is shown in Fig. 2C, and illustrates that the bed protein gradient must extend to at least 55 % egg length, a position at which bed RNA is not detectable above background.

In early (stage 1) embryos derived from exu homozygous mothers, bed RNA is uniformly distributed throughout the egg cytoplasm (Fig. 4B). However, by syncytial blastoderm a shallow anterior-posterior gradient has formed (Fig. 4C). Berleth et al. (1987) have found that bed RNA remains uniformly distributed in eggs laid by mothers which are mutant for both exu and the posterior group gene vasa. This suggests that the late bed RNA gradient is created by the degradation of the RNA in the posterior region of the embryo, due to the activity of the posterior organizing centre (Nüsslein-Volhard et al. 1987).

Embryos laid by sww homozygotes have a more variable distribution of bed RNA. While some embryos show no localization, most contain a weak anterior-to-posterior gradient (Fig. 4D). As is the case for exu mutant embryos, the gradient becomes more pronounced as development proceeds, due to the posterior degradation of the RNA. The variability of the bed RNA distribution in sunv mutant embryos is reflected in the variability in the final cuticular phenotype (Frohn-höfer and Niisslein-Volhard, 1986).

In embryos derived from staufen mutant mothers, bed RNA forms a gradient in the anterior region of the embryo (Fig. 4E). This phenotype is clearly distinct from that produced by exu or sww mutations, and cannot depend upon the posterior degradation of the RNA, since there is no localized posterior activity in stau mutant embryos (Niisslein-Volhard et al. 1987; Lehmann and Nüsslein-Volhard, unpublished results). This partial mislocalization of bed RNA leads to a shallower protein gradient (Driever and Nüsslein-Volhard, 19886). The bed RNA distributions in exu, sww, and stau mutant embryos are compared to the wild-type distribution in Fig. 5. Since these comparisons were performed using syncytial blastoderm embryos, exu and sww both show a shallow anterior-to-posterior RNA gradient.

bed RNA can first be detected in late previtellogenic follicles (stages 5–6 of King (1970)) accumulating in a single, posteriorly located cell of the germ cell cluster (Fig. 6A). As oogenesis proceeds and the oocyte grows larger than the fifteen nurse cells, it becomes clear that this cell is the oocyte. At these stages, very little RNA can be seen in the nurse cells, and the RNA forms a ring around the anterior margin of the developing oocyte (Fig. 6B). This ring increases in size as the follicle grows. During stages 9–10a, large amounts of bed RNA accumulate in the nurse cells (Fig. 6C,D). Unlike other maternal RNAs, bed RNA is not uniformly distributed within the nurse cell cytoplasm, but is concentrated in a peripheral region adjacent to each nurse cell nucleus. During stages 10b and 11, the localization within the nurse cells gradually disappears, as the nurse cells contract and transfer their cytoplasm to the oocyte (Fig. 6E). Upon entering the oocyte, the bed RNA becomes localized to the cortex of the anterior pole. At this stage, the RNA is no longer restricted to a ring around the anterior pole, and instead covers most of the anterior end of the oocyte. By stage 12, when the nurse cells are degenerating, bed RNA is localized in a cap at the anterior end of the egg, with more of the RNA being found ventrally than dorsally (Fig. 6F). During the final few’ hours of oogenesis, stages 13 and 14, the follicle cells that surround the oocyte secrete the chorion (King, 1970). We have been unable to analyze bed RNA localization during these stages because the vitelline membrane and chorion prevent the entry of the probe into the oocyte. However, the distribution of the RNA in stage 12 oocytes is different from that observed in very young eggs. In the oocytes, the RNA is localized to the cortex and is more concentrated ventrally, whereas in the early embryo, the RNA is found in a spherical region that extends into the interior of the egg and which is often located slightly dorsally. These differences suggest that there is a redistribution of bed RNA, either during stages 13 and 14 of oogenesis, or immediately after fertilization.

In order to understand how the mutations that alter the bed RNA distribution in the embryo affect the four phases of RNA localization during oogenesis, we have performed in situ hybridizations on exu, wv and stau mutant ovaries. The earliest differences from wild-type bed RNA localization are observed in exu mutant ovaries. The initial accumulation of bed RNA during stages 5–7 occurs normally. As the oocyte increases in size, the RNA still forms a ring at the anterior end, but this ring often appears more diffuse (Fig. 7A). The first major deviation from normal oogenesis becomes apparent during stage 10a, when bed RNA is not localized to the apical regions of the nurse cells, and instead shows a uniform distribution in the nurse cell cytoplasm (Fig. 7B). When this RNA enters the oocyte, it is not retained at the anterior pole, and from stage 10 onwards no localization within the oocyte is visible.

In the ovaries of stov homozygous females, the process of bcd RNA localization appears entirely normal up to stage 10a. The ring of RNA forms at the anterior end of the oocyte, and RNA accumulates in the apical regions of the nurse cells (Fig. 7C). During stages 10b and 11, the anterior ring of RNA seems to slip posteriorly and become more diffuse (Fig. 7D). In addition, much of the RNA that enters the oocyte during this time is not localized to the cortex. By stage 12, all bed RNA appears to have been released from the cortex and forms a shallow anterior-to-posterior gradient.

In homozygous stau ovaries, the process of bed RNA localization appears completely normal up until stage 12 (Fig. 7E,F), the latest stage that we have examined. Since the RNA is distributed in a gradient in early embryos, it must be released from the anterior pole after stage 12 of oogenesis.

The experiments presented in this report use a wholemount in situ technique to provide a detailed picture of the process of bed RNA localization during oogenesis and embryogenesis. These results confirm and extend the previous analyses of bcd RNA localization, which were performed using radioactive probes on sectioned material (Frigerio et al. 1986; Berleth et al. 1987; Stephenson et al. 1988). The wholemount procedure allows direct comparisons between the bed RNA and protein distributions. As shown in Figs 2 and 3, the two distributions are quite different. Given that no RNA can be detected posterior to 60% egg length even in overstained embryos, whereas the protein gradient extends at least to 30 % EL, the shape of the protein gradient must depend to some extent on the movement of protein molecules towards the posterior after they have been translated. Simple diffusion of the protein can account for this distribution (Driever and Nüsslein-Volhard, 1988a).

The in situ hybridizations to embryos derived from staufen mutant females reveal that, like exuperantia and swallow, staufen is required for the correct localization of bed RNA to the anterior pole. However, staufen mutations only cause a partial mislocalization of bed RNA. The anterior RNA gradient produced by these mutations results in a shallower bed protein distribution, in which the anterior levels of bed protein are strongly reduced compared to wild-type, and more posterior levels are slightly increased (Driever and Niisslein-Volhard, 1988b). The anterior reduction in bed protein concentration accounts for the loss of anterior head structures observed in staufen embryos (Schiipbach and Wieschaus, 1986). In addition to the head defects, staufen mutations produce a typical posterior group phenotype in which the abdomen is deleted and pole cells do not form. The phenotypes produced in double mutant combinations between staufen and other maternal effect mutations strongly suggest that the posterior effects of staufen are due to a failure to transport pole plasm constituents to the posterior pole (R. Lehmann, personal communication). Thus the staufen gene product is implicated in the localization of maternal factors to both the anterior and posterior poles of the egg.

Based on the geometry of the follicle, Frohnhöfer and Nüsslein-Volhard (1987) have proposed a simple model for bed RNA localization, in which RNA entering at the anterior end of the oocyte is trapped and attached to the cytoskeleton by factors that are uniformly distributed in the egg. The present data suggest that the process of localization is likely to be more complex. The in situ hybridizations to wild-type ovaries reveal at least four phases of bed RNA localization. In the first phase, which extends from stage 6 to early stage 9 of oogenesis, bed RNA is found in a ring at the anterior end of the oocyte. During stages 9–10a, RNA is still found in this ring but bed RNA also accumulates to high levels in the apical regions of the nurse cells. Stephenson et al. (1988) have also noticed the nonuniform distribution of bed RNA in the nurse cells of sectioned ovaries. The apical localization of bed RNA within the nurse cells is unusual since all other maternal RNAs that have been examined show a uniform distribution in the nurse cell cytoplasm, and no specialized cytological structures have been observed in these regions (for example Kobayashi et al. 1988; Sprenger et al. 1989). In the third phase, during stages 10b–12, the nurse cell localization disappears and all bed RNA becomes localized to the cortex at the anterior pole of the egg. A final phase of RNA redistribution must occur after stage 12, to produce the spherical pattern of bed staining seen in early embryos. The bed RNA distributions during these four phases are presented schematically in Fig. 8. The second and fourth phases of bed RNA localization cannot be explained by the simple model, since, during the second phase, the RNA is localized in the nurse cells before it enters the egg, and, in the fourth phase, the RNA is redistributed within the egg.

Since exu mutations lead to a uniform distribution of bed RNA in the nurse cell cytoplasm as well as in the oocyte, it seems likely that in wild-type ovaries both localizations occur by similar mechanisms. The apical region of a nurse cell can be considered to be the anterior end of the cell, since it lies on the opposite side of the cell from the ring canals which connect to the other nurse cells and the oocyte. Thus bed RNA may be transiently localized to the anterior ends of the nurse cells in a similar fashion to its localization within the oocyte, suggesting that the nurse cells also possess an intrinsic anterior-posterior polarity. The molecules that are required for bed RNA localization in the oocyte are most probably synthesized in the nurse cells. These molecules could therefore mediate the transient RNA localization within the nurse cells, before they themselves are transported into the oocyte. Since bed RNA is synthesized within the nurse cells, this localization cannot depend upon the polar entry of the RNA into one side of the cell. Thus the localization mechanism can function, at least within the nurse cells, in the absence of an asymmetric source of the RNA. If the process of localization within the egg is similar to that in nurse cells, the anterior accumulation of bed RNA within the oocyte may involve a more active mechanism than the simple trapping of the RNA as it enters at the anterior pole.

None of the mutations examined in this study completely disrupts all phases of bed RNA localization. The earliest phenotypes are seen in exu mutant ovaries, in which the RNA does not become restricted to the apical regions of the nurse cells during stages 9–10a. exu mutations also seem to have a weak effect on the first phase of bed RNA localization, since the anterior ring often appears more diffuse. Although we have used three different strong exu alleles, which all produce the same phenotype, it is possible that none of these is an amorphic mutation. In an exu null genotype, even the earliest phase of bed RNA localization might be abolished. Berleth et al. (1988) have proposed that the exu gene product binds directly to bed RNA, and mediates its attachment to the cytoskeleton. Since exu mutations abolish the localization in nurse cells, if exu does bind the RNA, it must first do so within the nurse cells. This raises the possibility that bed RNA enters the oocyte as part of a ribonucleoprotein complex that also contains exu protein. If this is the case, one might expect exu protein to colocalize with bed RNA at the anterior pole of the embryo.

In sww mutant ovaries, the third phase of bed RNA localization is disrupted. As Berleth et al. (1988) and Stephenson et al. (1988) have previously noted, the RNA is gradually released from the cortex of the oocyte during stages 10b—11 of oogenesis. Since stvtv mutations also cause defects in nuclear migration and cellulariz-ation during embryogenesis, Frohnhöfer and Nüsslein-Volhard (1987) and Stephenson et al. (1988) have suggested that the swallow protein is a component of the cytoskeleton. In .swtv mutants, the lack of this protein would cause a destabilization of the cytoskeletal elements that anchor bed RNA to the cortex of the oocyte and lead to a gradual release of the RNA. Our observation that sw mutations have no effect on the apical localization of the RNA in nurse cells provide support for the idea that swallow encodes an oocytespecific component of the cytoskeleton.

In staufen homozygotes, the process of bed RNA localization appears completely normal up until stage 12 of oogenesis, yet the RNA is not correctly localized in early embryos. The difference between these distributions indicates that the RNA must be released from the anterior pole sometime between stage 12 and egg deposition. This suggests that the staufen gene product may be required for the movement of bed RNA from the anterior/ventral cortex to the more dorsally located spherical region observed in early embryos, and provides further evidence for a fourth phase of bed RNA localization. The anterior gradient of RNA observed in staufen mutant embryos most probably results from the diffusion of the RNA during a short period of time between its release from the anterior pole and the start of embryogenesis. In general, there is a correlation between the stage at which a mutation affects bed RNA during oogenesis and the distribution of the RNA in embryos, exu mutations show the earliest phenotype (stage 9) and result in a uniform distribution in the embryo, xww mutations disrupt localization during stages 10–11 and lead to a very shallow embryonic RNA gradient, while staufen alleles seem to cause a release of the RNA after stage 12, producing a much steeper gradient at the anterior pole.

This study has identified four phases of bed RNA localization during oogenesis, and has demonstrated that exu, sww, and stau mutations each affect a different stage of this process. At present, we still lack any information on how this localization is achieved at a biochemical or cell biological level. Macdonald and Struhl (1988) have identified a 3′ untranslated region of of the bed RNA that is sufficient for localization to the anterior half of the embryo. As the molecular analysis of the tran-acting factors involved in this process advances, we may discover how this region of the bed RNA is recognized, and what components of the cytoarchitecture of the oocyte participate in each phase of the localization of bed RNA to the anterior pole.

We thank Maria Leptin, Leslie Stevens, Frank Sprenger and Dominique Ferrandon for their constructive criticisms of the manuscript, and Roswitha Grômke-Lutz and Doris Eder for help with the Figures. This work was supported by an EMBO fellowship to D. St. J, and by the Leibniz program of the Deutsche Forschungsgemeinschaft.