Graded Egfr activity patterns the Drosophila eggshell independently of autocrine feedback

The follicular epithelium of the Drosophila ovary provides a well-characterized genetic system that allows a detailed analysis of how spatial and temporal regulation of conserved signaling pathways can generate a complex pattern of cell fate (Berg,2005; Horne-Badovinac and Bilder, 2005; Nilson and Schupbach, 1999). The follicular epithelium surrounds the germline of each developing egg chamber and ultimately produces the eggshell(Spradling, 1993). The patterning of the follicular epithelium is thus reflected in the eggshell it produces, which exhibits pronounced axial asymmetries in its shape and in the presence of specialized structures, such as two dorsal respiratory appendages.

Central to the patterning of this tissue are two inductive signals: an Epidermal growth factor receptor (Egfr) ligand encoded by the gurken(grk) gene, and Decapentaplegic (Dpp), a member of the bone morphogenetic protein group of transforming growth factor β signaling molecules (Berg, 2005; Horne-Badovinac and Bilder,2005; Nilson and Schupbach,1999). In mid-oogenesis, Grk is localized to the dorsal side of the oocyte, where it induces spatially restricted activation of Egfr in the overlying follicle cells. This localized Egfr activation is essential for determination of the two distinct primordia that produce the two eggshell appendages and for their proper positioning along the dorsal-ventral (DV) axis(Neuman-Silberberg and Schupbach,1993; Neuman-Silberberg and Schupbach, 1996; Schupbach,1987). Dpp is produced by the anterior-most follicle cells, which overlie the nurse cells, and regulates patterning along the anterior-posterior(AP) axis (Dequier et al.,2001; Twombly et al.,1996). The combined action of these two ligands during these stages establishes and positions the appendage primordia(Chen and Schupbach, 2006; Deng and Bownes, 1997; Peri and Roth, 2000; Shravage et al., 2007).

An interesting aspect of this patterning process is that the single Grk/Egfr signal initially reaches a broad field of dorsal anterior follicle cells, yet ultimately leads to the establishment within this field of two dorsolateral appendage primordia separated by a distinct dorsal midline region. This pattern has been proposed to arise through a feedback loop in which the initial Grk/Egfr signal induces the expression of the Rhomboid (Rho)intracellular protease, leading to the production of a second Egfr ligand,Spitz (Spi), which amplifies the initial Egfr activation. According to this model, high levels of Egfr activity at the dorsal midline in turn induce the expression of Argos (Aos), a secreted inhibitor of Spi, leading to decreased Egfr activity in this region and accounting for the observed decrease in activated MAPK (Rolled - FlyBase) at the dorsal midline(Klein et al., 2004; Wasserman and Freeman, 1998). Negative-feedback regulation is thus proposed to split the initial field of Grk/Egfr signaling, creating a low-signaling region that generates the appendage-free dorsal midline domain that separates the two appendage primordia (Wasserman and Freeman,1998).

Much of our understanding of follicle cell patterning has been derived from analyzing eggshell patterning defects. Looking directly at the follicular epithelium during oogenesis, however, reveals that some aspects of this process remain unclear. For example, although downregulation of Egfr activity by Aos at the dorsal midline has been proposed to split the initial dorsal field of Egfr activity into two dorsolateral peaks, comparing the spatial and temporal profile of MAPK activation with that of an appendage fate marker shows, instead, that high (not low) Egfr activity is associated with the appearance of dorsal midline fate; decreased Egfr activity in these cells is detected only later, after the midline fate is established(Kagesawa et al., 2008; Peri et al., 1999). Also inconsistent with an Aos-mediated negative-feedback model is the observation that midline aos expression is observed after, not before, the establishment of the dorsal midline region(Wasserman and Freeman, 1998; Yakoby et al., 2008a). Such apparent discrepancies suggest that current models of Egfr signaling and eggshell patterning are incomplete.

Here we investigate the contribution of several regulators and effectors of Egfr signaling to the patterning of the follicular epithelium. We directly analyze follicle cell fate by examining the effect of mutant follicle cell clones on the establishment and positioning of the appendage primordia,enabling characterization of patterning defects with a resolution not possible from analysis of the mature eggshell. Since defects could potentially arise downstream of patterning, for example during appendage morphogenesis, we also look at the eggshells produced by females with follicle cell clones. Analyzing such eggshells has been inherently problematic because mosaics produce mixed populations of eggs, and there has been no easily detectable marker to indicate which eggs are derived from egg chambers with completely mutant follicular epithelia. We therefore developed a novel eggshell marker system that utilizes a follicle cell-dependent dominant female-sterile mutation to allow eggshells produced by mutant epithelia to be reliably identified,analogous to the P[ovoD1] system for marking germline clones(Chou and Perrimon, 1996). This tool has enabled us to correlate definitively the effect of mutations on cell fate patterning with their effect on the resulting eggshell.

Unexpectedly, our data show that the Spi/Aos feedback loop is not required to establish the normal pattern of follicle cell fates or for production of normally patterned eggshells. By contrast, we show that Mirror (Mirr) and Pointed (Pnt), transcription factors that function downstream of Egfr signaling, and Sprouty (Sty), an inhibitor of receptor tyrosine kinase (RTK)signaling, play distinct roles in follicle cell patterning. Mirr is required for the establishment of all dorsal anterior fates, whereas Pnt is required to generate the dorsal midline region. Sty restricts the width of the midline region, consistent with a role in negative regulation of Egfr signaling. Importantly, each transcription factor is required cell-autonomously. Together, these data suggest that the DV pattern of follicle cell fates is determined directly by the Grk gradient, without a requirement for intervening autocrine feedback.

Mosaic females were generated as described(Laplante and Nilson, 2006),using standard FRT sites and genotypic markers obtained from the Bloomington Stock Center. The mutant strains used were: rho⁶ P{neoFRT}80B,rho^7M43 P{neoFRT}80B (gift from S. Roth)(Wasserman and Freeman, 1998), aos^Δ7 P{neoFRT}80B (gift from N. S. Moon)(Voas and Rebay, 2003), spi¹ P{neoFRT}40A, spi³ P{neoFRT}40A (gift from T. Schüpbach, Princeton University), P{neoFRT}82B cic^U6 (Goff et al.,2001), P{neoFRT}82B pnt^Δ78(O'Neill et al., 1994), P{neoFRT}82B pnt^Δ88 (gift from H. Ruohola-Baker)(Scholz et al., 1993), mirr^e48 P{neoFRT}80B (gift from S. Roth), mirr^l(3)6D1 (Jordan et al., 2000), sty^Δ5 P{neoFRT}80B,sty²²⁶ P{neoFRT}80B, P{neoFRT}42D wind^RP54(Nilson and Schupbach, 1998)and pip⁶⁶⁴ P{neoFRT}80B (L.A.N., unpublished). The presence of rho, spi and aos mutations on these chromosomes was confirmed by complementation test. Additionally, rho mutations were confirmed by the generation of wing clones with loss of wing veins, which is characteristic of loss of rho(Sturtevant et al., 1993). spi¹ is also known as spi^A14 and spi³ as spi^OE72.

The P[decDN] transgene consists of a pCaSpeR4 P-element vector containing a cDNA encoding a version of the Dec-1 fc106 pro-protein with an engineered deletion of the s20 region (fc106ΔV288-E473)(Spangenberg and Waring,2007). Chromosomes bearing an FRT site and a P[decDN]insertion were generated by standard injection of the pCaSpeR4-decDNplasmid into strains containing existing FRT transgenes(Xu and Rubin, 1993) or by P-element-mediated male recombination (P{neoFRT}42D P[decDN]77.1)(Preston and Engels, 1996). Strains that contained P[decDN] insertions, maintained FRT function,and exhibited complete penetrance of the P[decDN] collapsed egg phenotype were retained for use in mosaic studies.

Mitotic recombination was induced as described(Laplante and Nilson, 2006) in females containing an FRT-P[decDN] chromosome in trans to a chromosome containing the same FRT and a mutation of interest. Eggs were collected on apple juice agar plates supplemented with dry yeast. The collapsed phenotype ranged in severity from completely collapsed eggshells to eggshells that retained a relatively normal three-dimensional shape but were translucent, often deformed, and tended to collapse upon storage at 4°C. For all genotypes, the data in Tables 1 and 2 were compiled from multiple independent trials.

Ovaries were fixed and stained as described previously(Laplante and Nilson, 2006). Primary antibodies used were mouse anti-BR-C core hybridoma supernatant 25E9.D7 (1:200, Developmental Studies Hybridoma Bank), rat anti-Ed (1:1000)(Laplante and Nilson, 2006)and rabbit anti-c-Myc supernatant sc-789 (1:200, Santa Cruz Biotechnology). Secondary antibodies (Molecular Probes) used were goat anti-mouse Alexa Fluor 555 (1:2000), goat anti-mouse Alexa Fluor 594 (1:1000), goat anti-rabbit Alexa Fluor 488 (1:2000) and goat anti-rat Alexa Fluor 647 (1:2000).

Images were acquired using a Zeiss Axioplan 2 wide-field fluorescence microscope or a Zeiss LSM 510 confocal microscope (McGill Cell Imaging and Analysis Network Facility). Apart from minor adjustments to brightness and contrast in Figs 2 and 4, no image manipulations were performed. Measurements of inter-appendage distance represent the distance between the dorsal-most edges of the appendages. All measurements and cell counts are reported as mean ± s.e.

The proposed role of Spi and Aos in eggshell patterning(Wasserman and Freeman, 1998)predicts that loss of spi or rho function will block amplification of the initial Grk/Egfr signal, resulting in a reduced domain of dorsal follicle cell fates, with reduced or absent dorsal midline fates and a single dorsally localized appendage primordium. Loss of aos function is predicted to eliminate the inhibition of Egfr signaling in the dorsal-most follicle cells, blocking the establishment of the dorsal midline region and yielding a single broad primordium.

To test these predictions, we generated follicle cell clones homozygous for mutant alleles of spi, aos or rho, and looked directly at follicle cell patterning using Broad Complex [BR-C; Broad (Br) - FlyBase]expression as a cell fate marker (Deng and Bownes, 1997; DiBello et al.,1991). In wild-type ovaries, BR-C protein expression is uniform in all follicle cells at early stages of oogenesis, but at early stage 10, BR-C becomes strongly reduced in a dorsal anterior T-shaped group of cells that will contribute to the eggshell operculum(Fig. 1A,A′, bars). Although indistinguishable in terms of BR-C expression, we will refer to the anterior portion of this `T' as the operculum region and the dorsal portion as the dorsal midline region to reflect our observation (see below) that they are differentially sensitive to genetic manipulations. Slightly later, BR-C levels are elevated in the presumptive roof cells of the two appendage primordia(Fig. 1B,B′)(Deng and Bownes, 1997; Dorman et al., 2004; Tzolovsky et al., 1999). The remaining posterior and ventral cells exhibit lower, `basal' BR-C levels.

Contrary to the above prediction, egg chambers with follicle cell clones homozygous for the amorphic rho^7M43 allele exhibited a wild-type BR-C pattern, with a dorsal midline region of normal width(Fig. 1C-D′). This normal phenotype was observed even in egg chambers in which the follicular epithelium was composed entirely of rho mutant cells (23/23)(Fig. 1D′), and was completely penetrant (55/55). Normal BR-C expression was also evident at later stages of oogenesis when dorsal appendages were being formed(Fig. 1D,D′). Midline width was normal at these late stages, even though the epithelial morphogenesis associated with appendage formation can give the appearance of a reduced distance between the appendage primordia. We found the same lack of effect on BR-C expression for the rho⁶ allele (10/10). Consistent with our observations for rho, large spi-null mutant follicle cell clones had no effect on the pattern of BR-C expression(Fig. 1E-F′) (12/12 for spi¹ and 12/12 for spi³). Similarly,although loss of aos is predicted to abolish the appendage-free dorsal midline region, egg chambers with complete(Fig. 1G,G′) (11/11) or very large (Fig. 1H,H′)(16/16) clones homozygous for the amorphic aos^Δ7allele displayed two normal domains of high BR-C expression separated by a clear midline region of normal width (3.8±0.2 cells versus 4.0±0.2 cells for non-mosaic egg chambers from the same females). Together, these data indicate that feedback modulation of Egfr signaling by Rho, Spi and Aos is not required for normal patterning of the follicular epithelium.

Although the above analysis showed no effect on BR-C expression, we asked whether rho, spi or aos mutant follicular epithelia exhibiting a normal BR-C expression pattern could nevertheless produce the previously reported eggshell appendage defects(Wasserman and Freeman, 1998). To identify which of the eggs produced by mosaic females were derived from egg chambers with homozygous mutant follicle cell clones, we developed a marker that incorporates information about the follicle cell genotype into the phenotype of the eggshell. This marker is a transgene bearing a dominant-negative form of the defective chorion 1 (dec-1)gene, which is required in the follicle cells for eggshell integrity(Bauer and Waring, 1987; Hawley and Waring, 1988). The dominant-negative version, P[decDN], bears an engineered deletion within the dec-1 coding region that causes dominant female sterility and the production of dec-like collapsed eggs(Spangenberg and Waring,2007). We reasoned that the P[decDN] transgene could function as a dominant genotypic marker for eggs derived from egg chambers with mutant follicle cell clones, analogous to the ovoD system for identification of eggs derived from mutant germline clones(Chou and Perrimon, 1992; Chou and Perrimon, 1996). Induction of FRT-mediated mitotic recombination in females transheterozygous for a mutation of interest and P[decDN] would generate egg chambers containing homozygous mutant follicle cells that lack P[decDN]. Eggs derived from egg chambers with large or complete mutant follicle cell clones,which lack P[decDN], should therefore be readily identifiable because they would remain intact, whereas those produced by non-mutant follicular epithelia retaining P[decDN] would collapse.

Initial tests of this system showed that a fraction of the eggs produced by mosaic females bearing an FRT-P[decDN] chromosome in trans to a wild-type FRT chromosome remained intact, whereas non-mosaic females bearing the same FRT-P[decDN] chromosome produced only collapsed eggs,suggesting that the intact eggs were the product of egg chambers with homozygous follicle cell clones (Table 1). Moreover, all intact eggs produced by females mosaic for P[decDN] and a mutant allele of capicua (cic),which is required in the follicle cells for normal eggshell patterning(Atkey et al., 2006; Goff et al., 2001), exhibited the characteristic cic dorsalized eggshell phenotype(Table 1), confirming that this marker allows the selection of eggs derived from egg chambers with homozygous mutant follicle cell clones.

To determine whether intact eggs are derived exclusively from complete follicle cell clones, in which all follicle cells are homozygous mutant, or whether epithelia with partial clones can also produce intact eggs, we looked for localized defects in eggs produced by mosaic females. Intact eggshells produced by cic mosaic females exhibited no evidence of partial dorsalization (Table 1),demonstrating that they were produced by follicular epithelia with mutant clones involving at least the entire anterior region. We also generated females mosaic for windbeutel (wind; wbl) and pipe (pip) (Stein et al., 1991; Schupbach et al.,1991), as partial wind or pip clones cause localized dorsalization of the embryo(Nilson and Schupbach, 1998). However, all intact eggs produced by wind or pip mosaic females bearing the P[decDN] marker gave rise to embryos that were completely dorsalized (Table 1); no cases of localized dorsalization were observed. Although we cannot rule out the possibility that epithelia with small groups of P[decDN]-bearing follicle cells can produce non-collapsed eggshells,these data strongly suggest that only egg chambers with completely homozygous mutant follicular epithelia can give rise to intact eggs. These data validate P[decDN] as an eggshell marker for complete follicle cell clones.

To ask whether rho, spi or aos play a role in eggshell patterning that is not reflected in the normal pattern of BR-C expression, we assessed the phenotype of eggshells produced by mosaic females bearing the P[decDN] marker. Consistent with their normal BR-C pattern(Fig. 1C-D′), virtually all intact eggs recovered from rho^7M43, rho⁶,spi¹, spi³ or aos^Δ7mosaics appeared normal and exhibited two dorsal appendages(Table 2). Non-mosaic sibling controls produced virtually no intact eggs, confirming the penetrance of the P[decDN] collapsed egg phenotype(Table 2). Control females with the identical genotype, but in which mitotic recombination was not induced,produced no intact eggs, arguing that the presence of the mutant rho,spi or aos alleles did not alter the penetrance of the P[decDN] phenotype (see Table S1 in the supplementary material). Moreover, none of the collapsed eggs exhibited fused dorsal appendages (see Table S2 in the supplementary material), indicating that no patterning defects were overlooked owing to unanticipated problems with the P[decDN]marker. These data indicate that Spi-mediated autocrine amplification of Egfr signaling and Aos-mediated negative feedback are not required for eggshell patterning.

We then took a candidate approach to identify additional genes involved in follicle cell patterning. We first looked at pnt, which encodes two Ets transcription factors, the expression of which is induced by Egfr activity in the ovary (Brunner et al.,1994; Gabay et al.,1996; Klambt,1993; Morimoto et al.,1996; Rebay,2002). Eggshells from pnt mutant epithelia exhibit a single dorsal eggshell appendage that is considerably wider than a wild-type appendage (Morimoto et al.,1996), but the requirement for pnt in the establishment of follicle cell fate had not been investigated.

Intact eggshells produced by females mosaic for either pnt^Δ78 or pnt^Δ88 and bearing the P[decDN] marker exhibited a single thick appendage(Fig. 2B; Table 2). Unexpectedly, some intact eggs with a wild-type eggshell phenotype were also recovered. However,such eggs were also observed in the non-mosaic sibling controls(Table 2), suggesting that the penetrance of the P[decDN] collapsed egg phenotype was reduced in this genetic background and that these eggs were the products of non-mutant epithelia. Consistent with this interpretation, follicle cell fate defects associated with pnt mutant follicle cell clones are completely penetrant (see below). These data, together with previous observations(Morimoto et al., 1996),demonstrate that loss of pnt function in the follicular epithelium results in a loss of the appendage-free dorsal midline of the eggshell.

We then analyzed follicle cell fate directly and found that loss of pnt function from the dorsal midline region, which in the wild type lacks BR-C expression by stage 10B (Fig. 2C,C′), results in ectopic high-level BR-C expression indistinguishable from that observed within the appendage primordia(Fig. 2D-F″). This phenotype is completely penetrant (54/54 clones examined for pnt^Δ88 and 30/30 clones for pnt^Δ78), and coincides precisely with the pnt clone borders (Fig. 2D″,E″,F″); even single-cell pnt mutant clones exhibited this effect on cell fate(Fig. 2F-F″). These data indicate that pnt function is required cell-autonomously to determine midline fate and that in the absence of pnt, the cells instead adopt an appendage-producing fate. Interestingly, the lack of BR-C expression in the operculum region is not affected by the loss of pnt(Fig. 2D-F″), indicating that this region is specified through a different mechanism than the dorsal midline, even though both exhibit identical BR-C levels.

In contrast to the dorsal midline, pnt mutant clones within the appendage primordia region had no effect on BR-C levels(Fig. 2D-F″). Interestingly, loss of pnt in follicle cells just posterior to the dorsal midline, which normally exhibit basal BR-C levels(Fig. 2C′), led to high BR-C levels, similar to those in the appendage primordia(Fig. 2E,E′). This effect was limited to the 2-3 rows of cells posterior to the normal primordia region,defining a domain within the follicular epithelium that is competent to adopt a primordium fate in the absence of pnt.

We next looked at the requirement for mirr, an Egfr target that encodes a homeodomain transcription factor expressed in dorsal anterior follicle cells during mid-oogenesis(Jordan et al., 2000; McNeill et al., 1997; Zhao et al., 2000). Loss of mirr has been associated with dorsal appendage defects, and ectopic mirr expression leads to ectopic high-level BR-C expression and ectopic appendage material (Atkey et al.,2006; Jordan et al.,2000; Zhao et al.,2000), but the changes in follicle cell patterning that lead to these defects have not been analyzed.

In contrast to the wild type (Fig. 3A,B), mirr mutant clones in the dorsal midline(Fig. 3C′, arrow) or operculum (Fig. 3C′,small arrow) regions failed to downregulate BR-C and instead exhibited a basal level of expression (Fig. 3C-D″). In the normal appendage primordia region, mirr mutant cells failed to upregulate BR-C, and thus also exhibited a basal level of BR-C expression, although the level was occasionally somewhat higher (Fig. 3C′,arrowhead). In all three regions, this effect was cell autonomous(Fig. 3C″,D″)(34/34). mirr mutant cells throughout the dorsal anterior region thus resemble ventral or posterior follicle cells that do not receive the Grk/Egfr signal, indicating that mirr is essential for determination of dorsal anterior fates. The fact that some cells in the primordia region can exhibit a mild elevation of BR-C expression, particularly in later stages(Fig. 3E,E′, arrowhead),suggests that this mirr allele is not a null, or that there is also a mirr-independent contribution to BR-C regulation in this region.

Consistent with this effect on follicle cell fate, the majority of intact eggshells produced by mirr mosaic females bearing the P[decDN] marker lacked dorsal appendages(Fig. 3F; Table 2). Intact eggs were recovered from these females at low frequency, likely reflecting the requirement for mirr in cyst encapsulation in early oogenesis,reducing the number of egg chambers with complete follicle cell clones that complete development (Jordan et al.,2000). Some eggs did exhibit appendages, possibly owing to decreased penetrance of the P[decDN] phenotype in this genetic background (see Discussion). Interestingly, eggs lacking appendages still retained a recognizable DV asymmetry in their shape(Fig. 3F, arrowhead), unlike the most severely affected eggs from grk mutant females(Schupbach, 1987), consistent with residual mirr activity or a possible mirr-independent contribution to follicle cell patterning (see above).

Because mirr and pnt have markedly different patterning phenotypes, it seems unlikely that they interact in a straightforward regulatory relationship. However, we found that expression of a mirrenhancer trap is decreased in pnt mutant follicle cells (see Fig. S1 in the supplementary material), suggesting that mirr is downstream of pnt and that pnt positively regulates mirrexpression. Since mirr levels are highest at the dorsal midline in wild-type egg chambers (see Fig. S1 in the supplementary material)(Atkey et al., 2006), one interpretation is that pnt mutant cells at the dorsal midline fail to adopt a midline fate because they do not express sufficiently high levels of Mirr. In such a scenario, high Mirr levels would determine midline fate,whereas a broad range of lower mirr levels would determine appendage primordium fate. Mirr would thus play an important role as an effector of the Grk gradient.

The dynamic pattern of MAPK activation in the follicular epithelium during mid-oogenesis (Kagesawa et al.,2008; Nakamura and Matsuno,2003; Peri et al.,1999; Wasserman and Freeman,1998) is consistent with a decrease in Egfr activity at the dorsal midline during dorsal anterior patterning. We therefore looked at the requirement for sty, which is induced by Egfr activation and encodes an intracellular inhibitor of RTK signaling(Casci et al., 1999; Hacohen et al., 1998; Reich et al., 1999).

Follicular epithelia completely homozygous for the sty^Δ5 allele, or with large clones that included the entire dorsal anterior region, exhibited a pronounced increase in width along the DV axis of the dorsal midline domain(Fig. 4A-A″,C-C″,arrows) (9.6±0.3 cells versus 3.1±0.2 cells in non-mosaic egg chambers from the same females), as well as a reduction in the total number of cells per primordium (Fig. 4A-A″,C-C″) (28.2±1.6 cells versus 50.8±3.4 cells in controls). This lateral expansion of the dorsal midline domain can also be visualized by examining the apical distribution of the cell adhesion molecule Echinoid (Ed), which at these stages is cortically localized in all follicle cells except for those expressing high levels of BR-C (Fig. 4A″,C″,small arrows) (Laplante and Nilson,2006).

These egg chambers also exhibited a reduction in the width along the DV axis of each appendage primordium (7.0±0.5 cells versus 8.8±0.3 cells in controls), suggesting that the cells within roughly the dorsal-most third of each presumptive primordium instead adopted a dorsal midline fate. Smaller sty mutant clones in this region also showed this phenotype(Fig. 4E-G″, arrowheads)and revealed that it is cell autonomous. Interestingly, when the border of such a clone fell within the dorsal region of the primordium, this cell fate change generated an ectopic midline region(Fig. 4E′,F′,arrowheads) flanked by two mini-primordia(Fig. 4E″″,F″″, arrows). Indeed, Ed staining revealed the apical constriction of these small, high-level BR-C-expressing domains, presumably as they prepared to produce appendages(Fig. 4F″″,arrows). Overall, the DV width of the dorsal domain (midline plus appendage primordia) was increased by about three cells in sty mutant epithelia, taking into account the increased width of the midline and the modest decrease in the width of the laterally shifted appendage primordia.

These sty mutant epithelia also exhibited changes in cell fate pattern along the AP axis. The AP length of the operculum domain of low BR-C expression (2.5±0.2 cells) was indistinguishable from that of non-mosaic controls (2.6±0.2 cells), but the posterior limit of the dorsal midline domain and appendage primordia was located more anteriorly(3.4±0.3 cells from the anterior, versus 5.9±0.2 cells in control egg chambers). This effect was seen most clearly in egg chambers in which roughly half of the dorsal anterior region was mutant, facilitating a direct comparison of wild-type and sty mutant regions within a single egg chamber (Fig. 4E-E″,G-G″, bars). This phenotype suggests that some cells in the posterior portion of the presumptive appendage primordia instead adopt a basal BR-C fate. Identical phenotypes were observed for the sty²²⁶ allele.

Consistent with this effect on follicle cell fate, intact eggshells produced by sty^Δ5 mosaic females bearing the P[decDN] marker exhibited an expanded dorsal midline region(Fig. 4B,D; Table 2) that was significantly wider than that of control eggs (57.3±2.3 μm versus 40.0±1.4μm for intact eggs produced by wild-type mosaic females bearing the P[decDN] marker). Interestingly, the appendages of these eggshells did not appear dramatically reduced in size compared with the wild type,despite the reduced total number of cells in the high-level BR-C-expressing domain, suggesting that morphogenesis might somehow compensate for the reduced number of cells.

As part of our analysis of patterning, we developed the P[decDN]marker, which allows the identification of laid eggs derived from egg chambers with homozygous mutant follicular epithelia. A recessive dec-1 mutant allele was previously developed as an eggshell marker(Nilson and Schupbach, 1998)but is of limited utility for analysis of complete follicle cell clones because the mutant cells are marked by the production of defective eggshell. By contrast, in the P[decDN] system, the mutant cells are marked by the loss of P[decDN], allowing us to recognize eggs that are the products of follicular epithelia that are largely or entirely mutant.

A potential complication of this system, not apparent in our initial tests(see Table 1), is the production of occasional intact eggshells by some negative control genotypes(see Table 2). Our current data suggest that certain P[decDN] transgenes exhibit a slightly decreased penetrance in certain genetic backgrounds. As discussed above, such low levels of background complicate, but do not prevent, the analysis of mutations associated with patterning defects, given that appropriate negative controls are included.

Our analysis of BR-C expression demonstrates that rho, spi and aos are not required for patterning of the dorsal anterior follicular epithelium, contrary to the current model of eggshell patterning(Wasserman and Freeman, 1998). Moreover, our analysis of mature eggshells using the P[decDN]eggshell marker revealed that loss of rho, spi or aos does not lead to eggshell patterning defects. Consistent with these data, eggs produced by spi mutant follicular epithelia bearing a recessive dec-1 allele as an eggshell marker(Nilson and Schupbach, 1998)exhibit two normally positioned dorsal appendages (T. Schüpbach, personal communication). Collectively, these observations rule out a straightforward role for the Spi/Aos feedback loop in patterning of the dorsal anterior follicular epithelium.

In addition to our observations of rho and spi mutant clones, our data also argue more generally against the requirement for a diffusible factor that amplifies the Grk/Egfr signal. If Grk/Egfr signaling were to induce such a factor, then this induction would be predicted to increase in sty mutant clones, which presumably experience increased Egfr pathway activity. The increased production of the diffusible factor would be predicted to have non-autonomous effects on follicle cell fate, but sty clones have strictly cell-autonomous effects. Together, these data support a model in which spatial information is provided directly by the distribution of Grk, without induction of an intervening diffusible ligand. Consistent with this idea, recent analysis of internalization of Grk-Egfr complexes by the follicle cells, as well as quantitative modeling, indicate that the Grk signal extends at least to the lateral limits of the dorsal anterior domain (Chang et al.,2008; Goentoro et al.,2006).

According to such a model, midline fates are determined in cells that receive high levels of Grk/Egfr signaling, whereas appendage primordium fates are determined in cells that experience intermediate levels of signaling. This interpretation is supported by the distribution of Grk, which has a dorsal high point (Neuman-Silberberg and Schupbach, 1996), and the fact that a variety of manipulations that lead to increased Egfr activity, such as Grk overexpression or loss of negative regulators such as sty, Cbl or kekkon-1, lead to an expansion of dorsal midline fates (this study)(Ghiglione et al., 1999; Neuman-Silberberg and Schupbach,1994; Pai et al.,2000; Zartman et al.,2009). Determination of distinct fates by distinct levels of Grk/Egfr signaling does not itself exclude a requirement for negative feedback in response to high-level Egfr signaling, but the observation that high levels of activated MAPK are restricted to the dorsal-most anterior follicle cells at the same time that BR-C is lost from the dorsal midline region(Kagesawa et al., 2008; Nakamura and Matsuno, 2003; Peri et al., 1999) suggests that feedback inhibition of Egfr activity is not required to specify dorsal midline fate.

A role for pnt in the specification of dorsal midline fate has been proposed previously (Morimoto et al.,1996; Yakoby et al.,2008b). Our clonal analysis demonstrates that pnt is required cell-autonomously in the dorsal midline domain to repress appendage primordium fate (Fig. 5A,C). Even single-cell pnt mutant clones exhibit ectopic, high BR-C levels,indicating that midline fate does not require pnt-mediated expression of a diffusible factor. Cells in the operculum region of the T-shaped operculum/midline domain are not affected by loss of pnt function,consistent with the presence of an operculum in the resulting eggshells. This differential effect of loss of pnt function on the operculum and midline domains is consistent with the observation that Dpp signaling is required in the operculum region, but not in the midline, to downregulate BR-C(J.-F.B.L. and L.A.N., unpublished observations)(Yakoby et al., 2008b).

We also found that pnt is required to repress appendage primordium fate in the follicle cells just posterior to the dorsal midline domain,indicating that follicle cells posterior to the normal midline region are competent to adopt an appendage primordium fate(Fig. 5C). Loss of pntthus converts two different domains to an appendage primordium fate,suggesting that Pnt interacts with different inputs and/or effectors in the normal dorsal midline region than in the cells just posterior to this region.

Our data demonstrate that mirr is required for determination of all dorsal anterior follicle cell fates(Fig. 5B). Loss of grkor Egfr function also leads to loss of these fates, and mirris expressed in the dorsal anterior follicle cells in response to Grk/Egfr signaling (Jordan et al.,2000; Zhao et al.,2000). A straightforward interpretation of this phenotype is,therefore, that mirr is a critical target of Egfr signaling in dorsal anterior follicle cell fate determination, and that loss of mirrimpairs the ability of the follicle cells to adopt the appropriate fate in response to the Grk/Egfr signal.

sty mutant follicular epithelia exhibit an expanded dorsal midline region and a largely corresponding reduction in the DV width of the appendage primordia (Fig. 5D). Given the role of Sty as a negative regulator of RTK signaling, this shift in fates is consistent with a model in which high levels of Egfr signaling induce dorsal midline fate. However, the total extent of dorsal fates (midline plus primordia) along the DV axis exhibits a modest expansion at best, suggesting that patterning occurs within roughly the same domain along the DV axis, even though the distribution of cell fates within this region is markedly different. This observation might indicate that the slope of the Grk gradient is sharper near the lateral edges of this domain.

Unexpectedly, along the AP axis, loss of sty function leads to a reduction in the length of the primordia, suggesting that the profile of an AP determinant is affected, either directly or indirectly. The best candidate for such a determinant is Dpp, which influences follicle cell fate along the AP axis. Dpp is required to determine operculum fate (J.-F.B.L. and L.A.N.,unpublished results) (Yakoby et al.,2008b), and Dpp overexpression, or removal of its negative regulator brinker (brk), causes an AP expansion of the operculum domain at the expense of appendage primordium fate(Chen and Schupbach, 2006; Deng and Bownes, 1997). The latter results seem to implicate Dpp in the repression of appendage fates, but some reports have concluded that appropriate levels of Dpp and Egfr activity cooperate to promote primordium fate (Peri and Roth, 2000; Shravage et al., 2007), whereas others have shown that loss of Dpp signaling does not affect this fate (J.-F.B.L. and L.A.N., unpublished results)(Yakoby et al., 2008b). Interestingly, more mild overexpression of Dpp seems to lead to a posterior shift of appendage primordium fates (Deng and Bownes, 1997; Dequier et al., 2001); whether these fates lie within the normal primordium region was not determined in these studies, but ectopic determination of primordium fate could suggest a role for Dpp beyond a repressive effect.

The AP defects in sty mutant clones cannot be explained by a simple model in which high levels of Egfr signaling cooperate with Dpp to induce dorsal anterior fates. Given the apparent DV expansion of the domain of high-level Egfr signaling, one would predict an expansion or no effect. Instead, the AP length of primordia is decreased, whereas the length of the operculum domain along the AP axis is unaffected. Another possible explanation is that the loss of sty function leads to decreased Dpp signaling. For example, Egfr activation induces the expression of brk(Campbell and Tomlinson, 1999; Chen and Schupbach, 2006; Minami et al., 1999; Shravage et al., 2007),suggesting that increased Egfr signaling in sty mutant epithelia might lead to ectopic brk expression, which in turn would modify the profile of the Dpp gradient and alter patterning along the AP axis. Alternatively, increased Egfr activity could impact Dpp signaling at another level of the pathway, or loss of sty could influence an as yet unrecognized determinant of appendage primordium fate.

Together, our data support a model in which graded Grk/Egfr signaling determines distinct dorsal anterior cell fates without a requirement for intervening autocrine feedback. Moreover, the changes in cell fate we observe in various mutant conditions reveal the ways in which the pattern can be altered, which is itself informative. The identification and characterization of such shifts in cell fate pattern, using BR-C and other markers, will contribute to the dissection of the spatial and temporal components that pattern this tissue during this developmental period.