Dorsoventral axis formation in Drosophila depends on the correct dosage of the gene gurken

The two Drosophila genes gurken (grk) and torpedo (top)/DER (Drosophila EGF receptor homolog, Schejter and Shilo, 1989; Price et al., 1989), are involved in dorsoventral patterning during oogenesis. Females homozygous for mutations in either top/DER or grk produce ventralized eggs and embryos, indi-cating that loss of either grk or top function leads to a shift in cell fates, both in the somatically derived follicle cell epithelium surrounding the oocyte and in the embryo that develops inside the mutant egg. Analysis of the mutant phenotypes, as well as studies involving mosaic egg chambers, have shown that grk and top are components of a signaling process between the germ line and the somatic follicle cells, and form a putative ligand-receptor pair. The grk-top/DER signal is required for dorsal follicle cell differentiation and controls embryonic dorsoventral patterning (Schupbach, 1987; Price et al., 1989; Neuman-Silberberg and Schupbach, 1993). Dorsoventral asymmetry in the embryo is achieved through asymmetric activation of the Toll receptor present on the embryonic membrane. The grk-top signaling process inhibits the production of an active ligand for Toll on the dorsal side of the egg and thus restricts activation of the Toll receptor to the ventral side (Schupbach, 1987; Hashimoto et al., 1988; 1991; Manseau and Schupbach, 1989; Roth et al., 1989; Steward, 1989; Stein et al., 1991; Stein and Nusslein-Volhard, 1992; Chasan and Anderson, 1994).

It has never been shown, however, that the grk-top signal induces dorsal cell fates in a follicle cell epithelium that does not yet possess any dorsoventral polarity. It is possible, and in fact has been proposed, that initial patterning of the follicle cells depends on other spatially regulated factors in the egg chamber (see e.g. Ruohola-Baker et al., 1993). Although the grk RNA is localized to the dorsal anterior corner of the oocyte, the grk-top signal might only trigger a differentiation response in a prepatterned follicle cell epithelium. In this case, the restricted accumulation of the grk RNA would be fortuitous, or at least not an essential aspect of the grk function. Mislo-calization of the grk RNA was observed in egg chambers from females mutant for fs(1)K10, squid, cappuccino and spire which produce dorsalized eggs and embryos (Neuman-Silberberg and Schupbach, 1993). This observation does not, however, prove that grk is the limiting factor in dorsoventral patterning, since the effects of the dorsalizing mutations might not be restricted to grk. If other unknown localized factors establish dorsoventral polarity in the egg chamber, they might also be aberrantly localized in the dorsalizing mutants. That maternal mutants might affect the localization of more than one factor in the egg chamber is not without precedent. Mutations in at least two of these genes, cappuccino and spire, are known to affect the localization of a number of maternal components (for review see Lehmann, 1992).

Identificaton of the key regulated components that establish dorsoventral polarity in the egg chamber requires a procedure whose initial effects are restricted to a single component or gene product. In the experiments described in this paper, we have constructed females with extra copies of the gurken gene. We show that an increase in the amount of grk transcripts is sufficent to dorsalize the follicle epithelium. These results argue that gurken plays a crucial instructive role in dorsoventral patterning.

Females carrying extra copies of the gurken transgene were obtained from stocks carrying two different P-element-mediated insertions of the genomic gurken region over balancer chromosomes (Neuman-Silberberg and Schupbach, 1993). Transgenic females carrying more than three copies of the gurken transgene are sterile. We were unable to obtain flies with more than four copies of the transgene. The reason for this lethality is unknown.

For characterization of the egg shell phenotypes, freshly laid eggs were mounted in Hoyer’s medium. For inspection of the embryonic cuticle, fully developed embryos were dechorionated in bleach, manually removed from the vitelline membrane, and mounted in Hoyer’s/lactic acid (7:3) (Wieschaus and Nusslein-Volhard, 1986).

Immunological staining of embryos was performed as previously described (Macdonald and Struhl (1988) using the Avidin/biotin ABC system (Vector). Anti-twist antibody was kindly provided by S. Roth. Stained embryos were mounted in 50% methyl salicylate/50% Canada Balsam.

For in situ hybridization, whole-mount ovaries were hybridized with digoxigenin-labeled DNA (grk) or RNA (rho) probes prepared with the Boehringer Mannheim kit. rho probe was provided by S. Roth. Fixation and hybridization were done according to the procedure described by Tautz and Pfeifle (1989) with modifications (Suter and Steward, 1991), hybridizing at 45°C for DNA and 55°C for RNA probes.

To determine the effects of grk dosage on dorsoventral patterning, we analyzed the eggs and embryos produced by females that carried multiple copies of a grk transgene, inserted at different places in the genome.

The transgenes are, most likely, not transcribed at the level of the endogenous wild-type genes, since a single copy of a particular transgene restored wild-type morphology to about 50-70% of the eggs produced by a mutant female homozygous for the strong allele grk^HK36. With two copies of this transgene crossed into a grk^HK36 background, we found that approximately 90% of the eggs showed wild-type morphology. This is similar to females heterozygous for grk^HK36, which typically produce 80-90% wild-type eggs (grk does show a slight hap-loinsufficiency, which can be enhanced in some genetic back-grounds). We therefore estimate that this particular transgene may produce somewhat more than half as much grk product as one endogenous gene. Females carrying four extra copies of the gurken transgene were chosen for the analysis, since we were unable to obtain flies with higher copy numbers of the gurken transgene. We would predict that these females contain about two-to three-fold higher levels of grk RNA than wild-type females.

Females that carry two copies of the gurken transgene, in addition to the two endogenous copies, produce a small fraction of partially dorsalized eggs. Females with four copies of the transgene lay a significant fraction of partially dorsal-ized eggs, and a few severely dorsalized eggs (Fig. 1). There is a variability in the amount of dorsalization observed in eggs produced by genotypically identical transgenic females (or even by a single female). On average, in 33% of the eggs, the dorsal appendages are spaced considerably further apart than in wild type (Fig. 1B). In 51% of the eggs, there is an excess of dorsal appendage material which usually results in a broad fusion of appendage material over the dorsal side (Fig 1C). 4% of the eggs have a strongly dorsalized phenotype with dorsal appendage material being secreted around the entire egg circumference (Fig. 1D). Few embryos hatch from the eggs laid by the females with four extra copies of the gurken transgene, and the majority of the unhatched embryos show a partial dorsalization. This dorsalization appears more extreme in the head and thoracic region, whereas the abdomen and telson often remain relatively normal (Fig. 2). In general, it appeared that there was a good correspondence between the degree of dorsalization of the embryonic cuticle and of the egg shell. In order to assess the dorsalization phenotype of the embryos more carefully, we stained a collection of such embryos with antibodies to the twist protein, which in wild type is expressed along the entire ventral region of the cellular blastoderm embryo. In the embryos produced by the transgenic flies, we observed a narrowing or deletion of the ventral twist domain particularly in the anterior part of the embryo (Fig. 3). There is, however, a considerable variability in the reduction of the twist domain ranging from embryos in which twist-expressing cells are missing only in an anterior domain, or in anterior and middle regions of the embryo, to the most severe cases in which only the poles still express the twist protein. Since expression of twist at the poles is regulated by the terminal genes (Ray et al., 1991), this extreme phenotype represents a complete loss of ventral mesoderm. The progressive loss of twist expression from anterior to posterior correlates well with the anterior-to-posterior dorsalization observed in the embryonic cuticle preparations.

To test how the addition of extra grk gene copies would affect the RNA localization, we performed in situ hybridization experiments on ovaries from females with four copies of the transgene (Fig. 4). These ovaries stain more intensely than the wild-type control at all stages of oogenesis. In early stages (stages 1-7), the grk RNA accumulates normally in the posteriorly situated oocyte. In wild-type oocytes, the grk RNA briefly accumulates along the anterior border of the oocyte during stage 8 of oogenesis, before assuming a dorsally restricted accumulation pattern in early stage 9. In ovaries from transgenic females, grk RNA requires longer to assume the tightly localized perinuclear pattern since in the majority of stage 9 oocytes the grk RNA is still found in a ring around the anterior margin of the oocyte (Fig. 4). During stage 10 of oogenesis, the grk RNA appears normally localized. Extra copies of the grk gene lead to an increase in the amount of grk RNA present in the oocyte, which presumably causes the observed lag in the localization of grk RNA to the dorsal side of the oocyte.

The gene rhomboid (rho), which is expressed in the dorsal follicle cells during oogenesis, is required for dorsoventral patterning (Ruohola-Baker et al., 1993) and can be used as a molecular marker for dorsal follicle cell fate. We tested whether rho expression changes in ovaries containing an increased level of grk. The wild-type expression pattern of rho undergoes dynamic changes during development (Ruohola-Baker et al., 1993). In our experiments, rho expression in wild-type ovaries was first detected in early stage 10 of oogenesis. The RNA is expressed on the dorsal side of the egg chamber in a patch of follicle cells forming an apron-like shape overlying the oocyte nucleus (Fig. 5A). In addition, the RNA seems to be expressed in a single row of follicle cells forming a lateral to ventral semicircle along the anterior margin of the oocyte. In mid-stage 10 egg chambers, the initial broad staining on the dorsal anterior side is resolved into a pattern of two dorsolateral stripes (Fig. 5B). This pattern persists till later stages of oogenesis. In females carrying four extra copies of grk both the early and the late expression patterns of rho are altered on the dorsal side (Fig. 5C,D,E). The initial patch of staining on the dorsal side becomes broader expanding more ventrally and, in extreme cases, encircling the entire egg chamber. When the broad pattern resolves, the two dorsolateral stripes are further apart with a larger dorsal gap between them than in wild type (Fig. 5E). This result indicates that a higher dosage of grk leads to an expansion in the dorsal population of follicle cells that initially express rho and are later bordered by the lateral rho stripes.

The pattern alterations observed in eggs and embryos from females carrying multiple copies of the gurken gene indicate that gurken is a limiting factor in the dorsoventral patterning of follicle cells. The dorsalization of the egg shell and the changes in rho expression pattern demonstrate that an increased level of grk product induces dorsal fates in a larger group of follicle cells than in wild type. The size of the follicle cell population that assumes a dorsal fate depends on the local amount of grk product. Therefore, activation of the top/DER receptor by gurken produces instructive information for dorsal follicle cell determination.

In egg chambers from females with four copies of the grk transgene, the localization of grk RNA to the anterior-dorsal corner of the oocyte appears to be delayed as compared to wild type, and grk RNA can still be found along the anterior margin of the oocyte during stage 9 of oogenesis. This suggests that the grk RNA localization process is saturable by a two-to three-fold increase in the amount of grk RNA. The increased amount of grk RNA, possibly in conjunction with some unlo-calized RNA, leads to a variable degree of dorsalization of egg shell and embryo. The variability of the observed dorsalization most likely indicates that the dosage level of grk in our experiments just barely reached a threshold of dorsalization, where small physiological differences between different egg chambers lead to a big range of egg shell and embryonic phenotypes. Unfortunately we were unable to obtain females with even higher copy numbers of the grk transgene. Such females would have allowed us to test at which level of extra grk RNA the dorsalization of egg shell and embryo would have become completely penetrant.

The aberrantly localized grk RNA at the (Grossniklaus et al., 1989; Schupbach and Wieschaus, 1991, Spradling, 1994,). This ring of rho expression is present in wild type and in grk mutant ovaries (data not shown, but visible in anterior end of the stage 9 egg chambers from transgenic females resembles the alteration of grk RNA localization seen in egg chambers homozygous for mutations in fs(1)K10, squid, cappuccino and spire (Neuman-Silberberg and Schupbach, 1993). Similarly, the dorsal-ization pattern of the embryos derived from transgenic females also resembles the dorsal-ization of embryos produced by females mutant for fs(1)K10. A stronger dorsalization of these embryos was observed at the anterior end (Wieschaus, 1979). The direct biochemi-cal targets of fs(1)K10, squid, spire or cap-puccino in oogenesis are presently unknown. At least in the case of cappuccino and spire it has been shown that these mutations affect the localization of several gene products in oogenesis (for review see St. Johnston and Nusslein-Volhard, 1992). Here we show that increasing the dosage of grk alone is sufficient to produce dorsalization suggesting that grk is the major factor whose mislocalization causes the observed dorsalization in these mutant genotypes.

Rhomboid is expressed in a population of follicle cells overlying the anterior dorsal side of the egg chamber, in proximity to the oocyte nucleus (Ruohola-Baker et al., 1993). Loss of rho expression results in ventralized eggs and embryos, and rho is therefore necessary for dorsoventral pattern formation during oogenesis. Ectopic expression of rho in the follicle cells was shown to dorsalize the egg and the embryo. Rho appears to act as a limiting factor of dorsal signal reception in the follicle cells (Ruohola-Baker et al., 1993). Our data indicate that expression of rho on the dorsal side of the egg chamber is most likely induced in the follicle cells by the grk-top/DER signaling process. Rho, in turn, may be necessary to facilitate the ligand-receptor interactions (see discussions in Ruohola-Baker et al., 1993, and Sturtevant et al., 1993). The dorsoventral signaling process therefore utilizes a limiting factor both in the production as well as in the reception of the signal. We also observed expression of rho in a single row of cells along the ventral and lateral sides of the egg chamber which was not affected by increasing the grk gene dosage. This ring of follicle cells are situated precisely at the border between the nurse cells and the oocyte, and they have been previously shown to specifically express a number of lacZ enhancer insertions, indicating that they constitute a distinct subpopulation of follicle cells Fig. 6F of Ruohola-Baker et al., 1993), indicating that this part of the rho expression pattern is independent of grk. We believe that this anterior subpopulation of follicle cells is determined by an anterior-posterior patterning system operating in the follicle cell epithelium, and is independent of the dorsoventral system.

The effects of gurken dosage are similar to the effects of bicoid and oskar, two other localized and limiting maternal factors in Drosophila development, which cause pattern abnormalities when present in extra maternal gene copies (Driever et al., 1988; Smith et al., 1992) bicoid acts as a morphogen whose concentration specifies several different cell fates in the target tissue (Struhl et al., 1989). oskar, in contrast, is not thought to act as a morphogen, but serves to localize the morphogen nanos into a translation competent complex (Gavis and Lehmann, 1992; Lehmann, 1992). Extra copies of oskar lead to ectopic sources of nanos protein, and this causes embryonic abnormalities. Our studies demonstrate that extra gene doses of gurken affect follicle cell patterning and expand the population of follicle cells assuming a dorsal cell fate. The experiments do not determine whether specific follicle cell fates are directly controlled by different Grk protein concentrations. It may well be that gurken induces a relatively unpatterned dorsal primordium in the follicle cells, and that subsequently secondary patterning mechanisms operating within and between the dorsal and ventral follicle cell primordia specify the exact follicle cell fates. In this view, the initial expression of rho in a dorsal domain might be a relatively direct response to the grk-top signaling process, whereas the later lateral stripes of rho expression would be regulated by secondary patterning mechanisms possibly requiring interactions between dorsal and ventral follicle cells.

In summary our data show that grk is the major regulated factor in oogenesis whose concentration and spatial distribution are regulated by the germ line to establish the drosoven-tral fate of the follicle cells, and subsequently, of the embryo.

We would like to thank Siegfried Roth for the twist antibody, for the rhomboid probe and for advice on embryo staining techniques. We are grateful to Eric Wieschaus, Robert Ray and Ken Irvine for comments on the manuscript, and our colleagues in the lab, in particular S. Roth, for stimulating discussions. F. S. N.-S was supported by an American Cancer Society Postdoctoral Fellowship. The work was supported by grant DB 23A from the American Cancer Society and by grant GM40558 from the US Public Health Service.