Establishment of dorsal-ventral polarity of the Drosophila egg requires capicua action in ovarian follicle cells

Dorsal-ventral (DV) patterning of the Drosophila embryo begins during oogenesis and relies on a series of intercellular signaling events in the egg chamber between the germline-derived oocyte and the somatically derived follicle cells (Ray and Schüpbach, 1996; Nilson and Schüpbach, 1999). The initial asymmetry for embryonic DV polarity is established at mid-oogenesis in the oocyte and is subsequently conveyed to the surrounding follicle cells. Spatial information generated within the follicular epithelium is then transmitted back to the germline, and is interpreted by the embryo after fertilization.

The initial polarizing signal from the oocyte to the follicle cells is determined by gurken (grk), which encodes a secreted transforming growth factor α (TGFα)-like protein (Neuman-Silberberg and Schüpbach, 1993). During mid-oogenesis, grk mRNA associated with the oocyte nucleus is localized to the dorsal anterior corner of the oocyte, resulting in restricted distribution of Grk protein and the activation of the Epidermal growth factor receptor (Egfr) in dorsal follicle cells.

Activated Egfr transduces the dorsal signal to follicle cell nuclei via the Ras-Raf-mitogen-activated protein (MAP) kinase kinase - MAP kinase cascade (Brand and Perrimon, 1994; Hsu and Perrimon, 1994; Schnorr and Berg, 1996; Wasserman and Freeman, 1998; Peri et al., 1999). The pattern of Grk-mediated Egfr activation can be visualized by target gene RNA expression in a dorsal-anterior patch of follicle cells centered on the dorsal midline (Ruohola-Baker et al., 1993; Musacchio and Perrimon, 1996; Sapir et al., 1998). Egfr activation in the follicle cells produces two delayed consequences: generating DV polarity of the embryo and patterning the eggshell dorsal appendages (Roth, 1998).

The DV pattern of the embryo is ultimately determined by the domain of pipe expression in the follicle cells during oogenesis (Nilson and Schüpbach, 1998; Sen et al., 1998). Expression of pipe RNA is normally restricted to ventral follicle cells, as a result of repression in dorsal follicle cells by Egfr activation (Sen et al., 1998). Pipe is expressed in the Golgi (Sen et al., 2000) and is predicted to act as a glycosaminoglycan-modifying enzyme on an undetermined substrate. This modified molecule is hypothesized to ventrally activate an extracellular serine protease cascade, resulting in the proteolytic processing of Spätzle protein, which is thought to act as a ligand for the embryo transmembrane receptor Toll (Morisato and Anderson, 1995; LeMosy et al., 1999). Toll signaling is relayed by cytoplasmic components to effect nuclear translocation of the Dorsal protein, a member of the NF-κB/rel family (Roth et al., 1989; Rushlow et al., 1989; Steward, 1989). The maternal DV pathway culminates in a ventral-to-dorsal gradient of nuclear Dorsal, resulting in the transcriptional activation or repression of zygotic genes that respond to distinct nuclear concentrations of Dorsal and thereby subdividing the DV axis into discrete stripes (Rusch and Levine, 1996).

Mutations in grk or Egfr cause an expansion of ventral structures at the expense of dorsal pattern elements in both the eggshell and embryo (Schüpbach, 1987). Conversely, mutations in genes encoding regulators of the Egfr pathway produce a dorsalized eggshell and embryo. This phenotype is apparently caused by ectopic Egfr signaling, as constitutive activation of the Egfr by artificial dimerization also produces a dorsalized eggshell and embryo (Queenan et al., 1997). The dorsalizing mutations can be divided into two groups. The first class of genes, which includes K10 and sqd (Wieschaus et al., 1978; Kelley, 1993), is required in the germline for grk mRNA association with the oocyte nucleus. Loss of function in these genes results in grk mRNA mislocalization around the entire anterior cortex of the oocyte (Neuman-Silberberg and Schüpbach, 1993), and ectopic Egfr activation in the follicle cells (Ruohola-Baker et al., 1993; Kelley, 1993). The second class of genes, which includes kek1 and Cbl, acts in the follicle cells to negatively regulate the Egfr pathway (Ghiglione et al., 1999; Pai et al., 2000). In these mutants, Egfr signaling fails to be downregulated, resulting in ectopic activation of Egfr pathway target genes.

The grk-Egfr signaling pathway coordinately establishes the polarity of both the eggshell and embryo. Yet, the analysis of genes that act downstream of Egfr signaling in the follicle cells indicates that the pathway bifurcates into one branch which regulates eggshell pattern and one branch which establishes embryo polarity. For example, while mutations in argos change spacing of the dorsal appendages without affecting the pattern of the embryo (Wasserman and Freeman, 1998), mutations in pipe lead to dorsalized embryos without affecting the eggshell pattern (Stein et al., 1991).

We describe new maternal effect mutations in a locus named fettucine (fet) that we show are alleles of the capicua (cic) gene. Mutant fet females lay dorsalized eggs due to a requirement for fet in ovarian follicle cells. Unlike other dorsalizing mutations, fet does not act as a negative regulator of the Egfr pathway in patterning the embryo. Rather, fet appears to have distinct functions in each pathway downstream of the branchpoint for embryo and eggshell patterning. In embryo patterning, fet is required for pipe RNA expression in ventral follicle cells. In eggshell patterning, fet is required for modulating Egfr signaling later during oogenesis. cic has previously been shown to encode an HMG-box transcription factor (Jiménez et al., 2000). In support of the genetic analysis described above, we show that Cic protein is expressed in ovarian follicle cell nuclei. In addition to the maternal effect phenotypes, mutant fet flies exhibit zygotic wing defects, suggesting that cic is required in several developmental processes.

We isolated P element alleles of fet in a hybrid dysgenic screen for suppressors of spz^D1 sterility (Morisato and Anderson, 1994). P-type π2 spz^D1/ TM6 males were crossed to M-type DTS/ TM3, Sb females. Male progeny of the genotype * spz^D1/ TM3, Sb were crossed to TM1/ TM3, Ser females to generate * spz^D1/ TM1 and * spz^D1/ TM3, Ser females. From 37,000 females screened for fertility, 16 lines were recovered: five lines mapped to spz and 11 lines mapped to chromosome III, but were unlinked to spz. Five of the 11 second-site dominant suppressors, including fet^E11, were alleles of fet.

We isolated EMS alleles of fet in an F₁ screen for wing defects. Females of the genotype fet^E11/ TM3, Sb were crossed to mutagenized st ry e males, and 20,000 st ry e */ fet^E11 flies were screened for extra wing veins or blisters. Five mutant lines were recovered, including fet^T6 and fet^U6, and all exhibited recessive maternal effect phenotypes affecting embryonic pattern. The fet^U6 allele behaves genetically like a null allele, as the maternal effect and zygotic wing defects observed with fet^E11 in trans to fet^U6 flies were indistinguishable from fet^E11 in trans to Df(3R) Dl^BX12, a deficiency that uncovers the fet locus.

For mosaic analysis, fet alleles were recombined onto a chromosome containing FRT82B (Xu and Rubin, 1993). For making germline clones, FRT82B fet/ TM3, Sb females were crossed to hs-FLP/ Y; FRT82B ovo^D1/ TM3, Sb males (Chou et al., 1993), and the progeny were heat-shocked as third instar larvae for 2 hours at 37°C on 3 consecutive days. Eggs were collected from non-stubble female progeny crossed to wild-type males. For making unmarked follicle cell clones, e22c-GAL4, UAS-FLP/ CyO; FRT82B flies (Duffy et al., 1998) were crossed to FRT82B fet/ TM3, Sb flies. Eggs were collected from non-curly, non-stubble female progeny crossed to wild-type males. Even under optimal conditions, only a fraction of the egg chambers were expected to contain follicle cell clones. For making marked follicle cell clones, eggs laid by mosaic females of the genotype dec^VA28 hs-FLP/dec^VA28; FRT82B fet^U6/FRT82B P[dec⁺] were inspected for the presence of dec-marked clones and associated embryonic DV patterning defects, recognized by reduction or absence of ventral cuticular markers (Nilson and Schüpbach, 1998).

For rescue of the fet phenotype, we used a P-element transgene encoding the cic transcription unit, which was inserted on chromosome II (kindly provided by Gerardo Jiménez and Jordi Casanova, Institut de Biologia Molecular de Barcelona).

Enhancer trap lines were used for detecting the expression of kek1 (15A6) (Musacchio and Perrimon, 1996) and mirror (mirr^6D1) (Ruohola-Baker et al., 1993). Briefly, ovaries were dissected from aged females, separated into ovarioles and fixed in formaldehyde/heptane for 10 minutes. The expression of β-galactosidase was detected by incubating the ovaries in 0.2% X-Gal staining solution overnight at 37°C. Stained ovaries were dissected into individual egg chambers and mounted in 75% glycerol.

For analysis of grk RNA expression in ovaries, a digoxigenin-labeled DNA probe was prepared using Boehringer Mannheim reagents from a cDNA template, kindly provided by Trudi Schüpbach (Princeton University). Ovaries were fixed, treated with 50 μg/ml proteinase K for 4 minutes, and hybridized after a modification of the embryo hybridization protocol (Neuman-Silberberg and Schüpbach, 1993; Tautz and Pfeifle, 1989). For detection of fringe and pipe RNA expression in follicle cells, antisense RNA probes were prepared from cDNAs, kindly provided by Ken Irvine (Waksman Institute; fringe) and David Stein (University of Texas, Austin; pipe). Ovaries were dissected into individual egg chambers, treated with 50 μg/ml proteinase K for 3 minutes, and hybridized according to Suter and Steward (Suter and Steward, 1991) with modifications (Hong and Hashimoto, 1995).

Analysis of Cic expression in ovaries was carried out with rat anti-Cic antibodies generated against the C-terminal half of Cic protein, kindly provided by Gerardo Jiménez and Jordi Casanova. Preabsorbed antiserum was used at a dilution of 1:900 with ovaries that were fixed and pretreated as described (Peri et al., 1999). Primary antibodies were visualized with biotin-conjugated anti-rat antibodies and streptavidin-horseradish peroxidase (HRP) using Vectastain ABC (Vector Laboratories). DAB staining was enhanced with NiCl as described (Patel, 1994). For analysis of cross-sections, egg chambers were embedded in Spurr (Polysciences) and 10 μm sections containing the oocyte nucleus were identified.

For analysis of RNA expression in embryos, RNA probes were generated from zen, sog and tll cDNAs, kindly provided by Mike Levine (UC, Berkeley; zen), Ethan Bier (UC, San Diego; sog) and Jordi Casanova (tll), and in situ hybridization carried out following the protocol of Tautz and Pfeifle (Tautz and Pfeifle, 1989).

For analysis of Twist expression, embryos were stained with rabbit anti-Twist antibodies kindly provided by Siegfried Roth (Universität zu Köln). Primary antibodies were visualized with biotin-conjugated anti-rabbit antibodies and streptavidin-HRP using Vectastain ABC.

Ovarian RNA was isolated from dissected ovaries by homogenizing in 10% Sarcosyl/0.5 mg/ml proteinase K and extracting in 50:49:1 phenol:chloroform:isoamyl alcohol. Poly(A)⁺ RNA was selected by oligo (dT)-cellulose chromatography. The RNA blot was probed with a 3.5 kb NotI-EcoRI cic cDNA fragment containing the HMG box, kindly provided by Celeste Berg (University of Washington).

For sequencing the fet alleles, six overlapping PCR products (approximately 1.3 kb each) spanning the cic gene were amplified from genomic DNA, leaving gaps of approximately 50 bp uncovered in the second intron and 720 bp uncovered in the large fifth intron. Primers were chosen by consulting the cic cDNA sequence (Jiménez et al., 2000) and the cic genomic sequence CG5067 from the annotated genome sequence (Adams et al., 2000). Genomic DNAs for use as templates in the PCR were isolated (Jowett, 1986) from three sources: the parental stock st ry e, homozygous fet^T6 pupae and heterozygous fet^U6/ st ry e adults. All six PCR products from each genotype were sequenced directly by automated PCR sequencing reactions using an ABI 373S DNA sequencer in the BCMP Biopolymers Laboratory.

In comparing the st ry e sequence with the published sequences, nine polymorphisms within the coding region were detected: five silent changes, three changes producing amino acid changes, and an insertion of nine nucleotides coding for an additional three glutamines in a run of seven glutamines in the N terminus. Only a single change was found between the parental chromosome and the T6 or U6 chromosome, as described in the Results. In the case of T6, there is a 5 bp repeat around the splice donor and acceptor sites of cic intron 3, making the junction sequence difficult to predict. Unlike all other cic introns, the intron 3 junction does not conform to the GT-AG rule that specifies the 5′ and 3′ ends of the intron. The closest fit to this rule positions the T6 mutation as the last nucleotide of exon 3.

We originally isolated a class of dominant suppressors of a weakly ventralizing mutation (spz^D1) in a dysgenic screen (Materials and Methods). These suppressors mapped to 92D and defined a new locus that we called fettucine (fet). The fet^E11 mutation is a representative allele caused by the insertion of a P element. In a subsequent screen, the EMS-induced alleles fet^T6 and fet^U6 were generated. While embryos laid by spz^D1 / + females were ventralized and failed to hatch, about 10% of eggs laid by fet spz^D1/++ females hatched. Flies carrying the fet^E11 allele were viable as homozygotes and as transheterozygotes with fet^U6 and fet^T6, and these females exhibited a recessive maternal effect phenotype in which the eggshell and embryo were dorsalized. The fet^U6 allele behaved genetically like a null allele and was larval lethal, while the fet^T6 allele was a strong hypomorph, and homozygotes died as pharate adults.

The fet eggshell morphology was dorsalized, as assessed by a lateral shift of broadened dorsal appendages. In the strongest mutant combinations, ectopic dorsal appendage base material was secreted around the anterior circumference of the egg (Fig. 1A,B). Embryos produced by fet females (we will hereafter refer to these embryos as fet mutant embryos) exhibited an expansion of dorsal cell fates around the circumference of the embryo. These fet embryos failed to hatch and secreted a cuticle which consisted entirely of dorsal epidermis lacking any structures derived from lateral or ventral regions (Fig. 1C,D). This cuticular phenotype was preceded at the cellular blastoderm stage by expanded expression of the dorsal marker zen around the circumference of the embryo at the expense of the expression of the ventrolateral and ventral markers sog and Twist (Fig. 1E-J). Interestingly, the morphology of dorsalized fet cuticles was distinct from the dorsalized phenotype of pipe mutant cuticles (Anderson and Nüsslein-Volhard, 1986), in that the fet cuticles appeared broader with fewer constrictions. This observation might be accounted for by the superimposition of a segmentation defect on the dorsalized phenotype in fet mutants (see below).

Flies that lacked zygotic fet showed a spectrum of wing defects. Flies that carried weaker fet alleles showed ectopic vein material and blisters, while greatly thickened veins and crumpled wing blades were observed with stronger fet alleles (Fig. 1K,L). The severity of the wing defects correlated with the strength of dorsalization of the maternal effect phenotype.

The observation that fet mutations affected both eggshell and embryo pattern suggested the involvement of fet in the grk-Egfr signaling pathway. To determine whether fet acts in the oocyte or the follicle cells of the egg chamber, we used FLP-FRT recombination systems to generate genetically mosaic egg chambers (Materials and Methods; Table 1), in which mutant clones were restricted to either the germline (Chou et al., 1993) or follicle cells (Duffy et al., 1998). Females with a fet^– germline produced embryos showing a DV pattern that appeared normal by the expression of Twist, zen and sog (Fig. 2A,B; data not shown). However, these embryos failed to hatch, indicating an essential role for fet in the germline. The cuticles of these embryos (Fig. 2C) resembled a gain-of-function torso phenotype in which the embryonic termini were expanded at the expense of segmentation (Duffy and Perrimon, 1994) (see Fig. 6).

By contrast, mosaic females with a wild-type germline but fet^– follicle cell clones produced dorsalized embryos. A few embryos were completely dorsalized (Fig. 2D), displaying a phenotype identical to embryos laid by pipe^– females, while many more showed regionally dorsalized phenotypes (data not shown). In order to determine if fet were required in a specific region of the follicular epithelium, we generated fet^– clones that were marked with a defective chorion 1 (dec-1) mutation, which allowed visualization of mitotic clones within the mature eggshell (Nilson and Schüpbach, 1998). None of the dorsal or lateral clones (0/8) showed defects in the embryonic DV pattern (Fig. 2E). By contrast, ventral clones (22/24) produced local dorsalization in the embryo (Fig. 2F), implying that fet is required cell autonomously to establish DV polarity of the embryo. Thus, the localized requirement for fet resembles the behavior of pipe and windbeutel, which were previously shown to be required on the ventral side of the follicular epithelium (Nilson and Schüpbach, 1998).

The eggshell phenotypes resulting from these mosaic studies were more difficult to interpret (Table 1). Dorsalized eggshells were produced by females in which follicle cell clones had been induced, and in eggshells containing dec-marked clones, ectopic dorsal appendage material was observed at the anterior boundary of some fet clones (data not shown). Dorsalized eggshells were also produced by females carrying a mutant germline, although this observation could be attributed to the simultaneous induction of follicle cell clones as a result of hs-FLP action.

Taken together, these findings indicate that fet is required in the ventral follicle cells for normal DV pattern of the embryo. These results also suggest a requirement for fet in the follicle cells for patterning the eggshell, although we cannot formally rule out a possible germline contribution.

We further investigated the role of fet in grk-Egfr signaling by examining the spatial distribution of grk mRNA in the oocyte. In fet mutant ovaries, grk mRNA was properly localized to the dorsal anterior corner of the oocyte (Fig. 3A,B), an observation consistent with the normal DV pattern of fet germline clone embryos.

To determine the effect of fet mutations on Egfr signaling in follicle cells, we analyzed expression patterns of known Egfr target genes. Mutations in the negative regulator Cbl result in ectopic activation of the Egfr pathway, leading to expanded expression of Egfr target genes and the production of a dorsalized eggshell and embryo (Pai et al., 2000). Although mutant fet eggs showed a similar morphological phenotype, Egfr signaling appeared normal in the follicle cells of fet ovaries, as determined by the stage 10 expression pattern of Egfr target genes kek1 (Fig. 3C,D) and PZ5650 (data not shown). This result suggests that fet action is mechanistically distinct from that of Cbl, and that fet is not normally required for negatively regulating the Egfr pathway.

We also examined fet mutant ovaries for expression of the homeobox gene mirror, which is expressed in dorsal follicle cells (Jordan et al., 2000; Zhao et al., 2000). mirror represses fringe transcription, establishing a complementary domain of ventral follicle cells that expresses fringe RNA. Like other targets of Egfr activation, mirror-lacZ was expressed normally at stage 10 in fet ovaries (Fig. 3E,F), and fringe RNA expression in ventral follicle cells at this stage appeared normal (Fig. 3G,H). Interestingly, however, this dorsal restriction of mirror-lacZ was not maintained at slightly later stages of oogenesis, when misexpression was observed around the anterior circumference of the egg (Fig. 3I,J). Taken together, these results suggest that fet is not required for initial reception of the Grk signal by Egfr in follicle cells.

We constructed double mutants of fet and Egfr to test the genetic epistasis relationship between these two genes. Females homozygous for the Egfr1 mutation laid eggs with a ventralized eggshell and embryo (Schüpbach, 1987) (Fig. 4A,B), while fet females laid eggs with a dorsalized eggshell and embryo (Fig. 4C,D). Embryos laid by Egfr1; fet double mutant females were dorsalized (Fig. 4E), indicating that Egfr requires fet activity to specify ventral embryonic fates and that fet acts downstream of Egfr for this function. By contrast, fet appears to act additively with Egfr for patterning the eggshell, as the eggshell dorsal appendages of the double mutant were broadened like fet dorsal appendages, but were positioned closer to the dorsal midline like Egfr mutant appendages (Fig. 4F).

The genetic observation that fet acts downstream of Egfr for defining embryo polarity led us to ask whether fet acts upstream or downstream of pipe transcription. Egfr activation establishes embryonic DV polarity by restricting pipe transcription to ventral follicle cells. In the wild-type ovary, expression of pipe RNA is first detected at stage 9 (Fig. 5A) and continues through stage 10B (Fig. 5C). In fet mutant ovaries, pipe RNA was never detected in ventral follicle cells, although expression of pipe RNA was observed in a few posterior cells at later stages (Fig. 5B,D). These results indicate that fet is required to establish pipe expression in the ventral follicle cells for patterning the DV axis of the embryo.

We mapped the fet mutations to the interval 67.2-68.6 on chromosome III by meiotic recombinational mapping and to 92D by P element hybridization to polytene chromosomes. As both cic and bullwinkle (bwk) mutations also map in this region (Jiménez et al., 2000; Rittenhouse and Berg, 1995), we examined the relationship between fet mutations and cic and bwk mutations.

The cic1 allele has a maternal effect terminal class phenotype similar to the torso gain-of-function phenotype, in which tll expression is expanded from the termini towards the center of the embryo, resulting in suppression of segmentation (Jiménez et al., 2000). fet mutant embryos and fet germline clone embryos also showed an expansion of tll expression (Fig. 6A-C), revealing a terminal class phenotype for fet embryos. This result accounted for the segmentation defect observed in the cuticles of fet germline clone embryos, and suggested that the fet mutations and cic1 could be allelic.

In addition to its requirement for normal terminal development, analysis of the cic1 allele revealed a role for cic in the germline as a co-factor necessary to convert Dorsal from a transcriptional activator to a repressor (Jiménez et al., 2000). In cic1 mutant embryos, zen RNA expression is expanded from its normal dorsally restricted domain to more lateral positions, although Twist expression appears wild type (Jiménez et al., 2000). By comparison, zen RNA expression was only slightly broadened in embryos produced by females with a fet germline (fet^E11 germline clone embryos exhibited a 15% expansion of zen expression compared with wild type; Fig. 2B and data not shown).

We determined complementation between cic1 and fet alleles. Transheterozygous cic1/fet females produced embryos with a torso gain-of-function phenotype (Fig. 6D), indicating that both alleles affect the same germline function. However, cic1 complemented the fet follicle cell defect, as eggs laid by transheterozygotes showed apparently normal DV polarity of both the eggshell and embryo. cic1/fet flies also developed wild-type wings.

The cic1 phenotype was rescued by a 9 kb transgene which expressed a transcript encoding an HMG-box transcription factor (Jiménez et al., 2000). The maternal and zygotic fet phenotypes were successfully rescued by a single copy of the same transgene. Rescued fet flies developed wild-type wings and produced eggs with wild-type dorsal appendages and hatching larvae.

We also tested the ability of fet to complement bwk. Maternal effect mutations in bwk cause abnormal dorsal appendages and bicaudal embryos with normal DV pattern (Rittenhouse and Berg, 1995); bwk8482 flies also show wing defects characterized by ectopic veins. bwk was shown to be required in the germline for normal eggshell and embryo patterning. P-element alleles of bwk complemented the fet eggshell and wing defects, while embryos laid by fet/bwk transheterozygotes were bicaudal (data not shown).

We analyzed transcripts expressed in the cic/bwk region on RNA blots (Fig. 6E). The level of a 5.7 kb transcript was reduced in both fet and cic1 ovaries. Expression of this transcript was restored in the ovaries of fet females carrying one copy of the transgene. The level of this 5.7 kb transcript was not affected in bwk mutant ovaries, and the bwk phenotype was not rescued by the 9 kb transgene (Jiménez et al., 2000).

Finally, we identified the lesions responsible for fet^T6, fet^U6 and fet^E11 in the open reading frame of the cic transcript (Fig. 6F). The cic gene (CG5067) contains 13 exons (Adams et al., 2000) (Fig. 6F), with the HMG box mapping to exon 3 and the conserved C-terminal domain being split between exons 11 and 12. The T6 and U6 mutations were each caused by a single nucleotide change in the coding region. For the T6 allele, a G to A transition at position 2861 at the putative exon 3/intron 3 junction produces a S to G amino acid change, and, more importantly, is predicted to reduce splicing efficiency of intron 3, introducing an in-frame translational stop codon in the unspliced intron (see Materials and Methods). For the U6 allele, a C to T transition at position 3007 produces an amber mutation in exon 4. For the E11 allele, the P element was inserted with 3′ to 5′ orientation after nucleotide 6510 in exon 10, introducing an in-frame stop codon that is predicted to produce a truncated protein that lacks the conserved C-terminal domain.

After determining that fet and cic were allelic, we used anti-Cic antibodies (Jiménez et al., 2000) to visualize the distribution of Fet/Cic protein in the follicular epithelium. Cic expression appeared quite dynamic. In stage 9 egg chambers, Cic was expressed uniformly in follicle cell nuclei (data not shown); at stage 10A, Cic began to be downregulated on the dorsal side (Fig. 7A,C), and Cic was completely absent in dorsal follicle cells by stage 10B (Fig. 7B,D; see Discussion). As predicted from the preceding genetic analysis, Cic protein was expressed in cic1 ovaries (Fig. 7E) and absent in fet ovaries (Fig. 7F).

Establishment of DV polarity in the Drosophila egg requires fet activity during oogenesis in the follicle cells. Genetic analysis suggests that fet is required downstream of Egfr signaling, and acts in two different pathways to pattern the eggshell and embryo. We show that fet is allelic to the cic gene, which encodes an HMG-box transcription factor.

Previous studies have described several dorsalizing mutations that all affect some aspect of Egfr signaling, consistent with its role in coordinating DV polarity of the eggshell and embryo. Thus, it has been surprising to find that both the production of the Grk signal in the oocyte and initial activation of the Egfr in follicle cells appear unaffected in mutant fet ovaries. As a consequence of Egfr signaling, the follicular epithelium is normally partitioned into dorsal and ventral domains through the spatially restricted expression of mirror and fringe transcripts, respectively (Jordan et al., 2000; Zhao et al., 2000). Concomitant with this early response, Egfr modulates its own signaling during the course of oogenesis by inducing the expression of genes (e.g. rho, kek1 and Cbl) that encode regulators, which act at the level of receptor activation. At early stage 10, which corresponds to the period when Egfr activation leads to the transcription of target genes, the expression patterns of mirror, fringe, rho and kek1 (Fig. 3; data not shown) in mutant fet ovaries are indistinguishable from wild type.

The dorsalized phenotype observed with the loss of fet activity in follicle cells is apparently caused by a requirement for fet in each branch of the Egfr pathway that separately patterns the embryo and eggshell. In establishing DV polarity of the embryo, fet is required as an essential transcriptional regulator of pipe RNA expression in the ventral follicle cells (Fig. 5). By epistasis analysis, fet acts downstream of Egfr and causes a dorsalized phenotype even when Egfr signaling is reduced.

In establishing DV polarity of the eggshell, fet appears to be involved in refining Egfr activity later during oogenesis. Although mirror-lacZ is initially expressed correctly, the domain later expands around the anterior circumference of the egg chamber (Fig. 3I,J), suggesting a role for fet after the first round of Egfr signaling by Grk. This interpretation is supported by analysis of the double mutant. In contrast to the strictly linear relationship observed for establishing embryonic polarity, the eggshell phenotype produced by Egfr1; fet females shows contributions from each of the individual phenotypes. The distance between dorsal appendages is reduced, as observed for the ventralizing Egfr1 mutation, but the individual appendage structure is broadened, as seen for the dorsalizing fet mutation.

In the wild-type eggshell, dorsal appendage pattern is achieved through refinement of the Egfr activation profile by both positive and negative feedback regulation (Freeman, 2000). Expression of rhomboid RNA is induced as a result of Egfr activation (Ruohola-Baker et al., 1993) and positively regulates continued Egfr signaling (Sapir et al., 1998). Induction of argos RNA expression leads to negative feedback on Egfr signaling at the dorsal midline, causing refinement of one domain of Egfr activation into two laterally symmetric domains that specify the placement of the paired dorsal appendages (Wasserman and Freeman, 1998).

This dorsal appendage pattern can be genetically altered in different ways. The K10 and squid mutant phenotype is characterized by an eggshell with a fused cylindrical dorsal appendage around the anterior circumference of the egg deposited around a dorsalized embryo (Roth and Schüpbach, 1994). The fet and Cbl mutant eggshell phenotype appears distinct; rather than a single cylindrical structure, two laterally placed broadened dorsal appendages form, often associated with circumferential dorsal appendage base material. A change in either the strength or timing of Egfr signaling might translate into these different phenotypes.

The fet mutations map to the cic transcript, which encodes an HMG-box transcription factor. Analysis of the cic1 maternal effect phenotype revealed two germline-encoded functions (Jiménez et al., 2000). The Cic protein acts as a repressor of tll transcription in the embryo termini and as a Dorsal-dependent repressor along the DV axis.

The experiments described show that establishment of DV polarity in the embryo requires maternal cic activity at two distinct stages. The characterization of the fet phenotype indicates that cic is first required during oogenesis to activate pipe transcription in the ventral follicle cells. The cic1 phenotype shows that cic is then required during embryogenesis as a co-repressor with Dorsal. In cic1 mutant embryos, zen RNA expression is expanded from its normal dorsally restricted domain to more lateral positions, although Twist expression appears wild type (Jiménez et al., 2000). By contrast, in fet mutant embryos, lack of pipe expression during oogenesis leads to a more severely dorsalized phenotype in the embryo, with zen RNA expressed circumferentially and Twist staining absent (Fig. 1).

In addition to the maternal effect eggshell and embryo phenotype, viable fet alleles exhibit wing phenotypes and strong fet alleles are lethal. This range of fet phenotypes, as compared with the cic1 phenotype, may be accounted for by the molecular nature of the mutations. The cic1 mutation is caused by a hobo mobile element insertion in the 5′ untranslated region of the cic transcript (Fig. 6F), while the bwk8482 allele, which also has a maternal effect phenotype caused by a germline defect, is associated with a P-element insertion in the same region (Jiménez et al., 2000). These mobile elements may contain cryptic promoters which allow sufficient expression of the cic transcript to rescue the somatic but not the germline functions of this gene (Kiger et al., 1999). The analysis of fet mutations, which alter the coding region of the cic transcript, suggests that cic is required in a number of signaling pathways during Drosophila development.

In regulating terminal development in the embryo, cic acts by repressing tll transcription (Jiménez et al., 2000). In controlling DV asymmetry of the follicular epithelium, two models of cic action are formally possible: either cic represses an unknown negative regulator of pipe expression or cic positively regulates pipe. cic is unlikely to be a direct negative regulator of pipe transcription, as pipe expression is abolished in fet mutant ovaries. However, cic also acts in the follicle cells in the eggshell patterning pathway, where it is possible that cic acts as a negative regulator at the mirror promoter, because ectopic mirror expression is observed in fet mutants at later stages of oogenesis.

In the blastoderm embryo, the receptor tyrosine kinase (RTK) Torso positively regulates tll expression at the termini by repressing cic action, resulting in the absence of Cic protein at the termini; uniform distribution is observed in embryos produced by torso^– females (Jiménez et al., 2000). In an intriguing parallel, Cic expression in the follicular epithelium is absent from the dorsal side; uniform distribution is observed in grk^– ovaries (data not shown), suggesting that this downregulation could be a consequence of signaling by the RTK Egfr. However, Cic asymmetry is detected after the initiation of pipe transcription. Thus, the pattern of pipe transcription cannot be determined by the physical distribution of Cic protein, although it remains possible that Cic is inactivated as a result of Egfr signaling prior to its disappearance from dorsal follicle cells.

Cic shows features of a family of HMG-box proteins that bind DNA in a sequence-specific manner, although DNA-binding activity has not been detected yet (Jiménez et al., 2000). As shown for the case of Drosophila TCF (Cavallo et al., 1998), a target of the wingless signaling pathway, some HMG-box proteins can act as positive or negative transcriptional regulators, depending on promoter contexts and interaction with co-factors. As positive regulators of transcription, HMG-box proteins bind in the minor groove of DNA and induce bending, which can bring distal regulatory elements closer to promoters to enhance transcription. As negative regulators of transcription, HMG-box proteins interact with co-repressors, such as Groucho (Gro). Gro does not bind to DNA directly, but rather is recruited to the template by DNA-bound transcription factors (Fisher and Caudy, 1998; Parkhurst, 1998; Chen and Courey, 2000). Repression probably occurs by several mechanisms, including recruitment of the histone deacetylase Rpd3 (Chen et al., 1999), which presumably silences transcription by altering local chromatin structure.

Cic and Gro have been shown to interact in vitro (Jiménez et al., 2000). For the Torso pathway, the biological significance of this physical association is supported by the analysis of gro germline clones, which show mutant phenotypes that are identical to those caused by the germline cic1 defects (Paroush et al., 1994; Paroush et al., 1997). It remains to be determined whether recruitment of Gro co-repressor by Cic is conserved in follicle cells and the wing disc.

D. M. acknowledges the guidance of K. V. Anderson in the nascent stages of this work. We are grateful to C. Berg and J. Casanova for generously sharing information and reagents. We also thank D. Bilder for advice on making ovarian clones, M. Walczak for assistance in generating dec-marked clones, and W. Bender, G. Denton and C. Hashimoto for their helpful suggestions on the manuscript. This work was funded by grant GM52084 (awarded to D. M.) from the NIH; D. J. G. was supported by postdoctoral fellowship PF4443 from the American Cancer Society.