Identifying loci required for follicular patterning using directed mosaics

From a single cell with finite resources, the process of development calls forth an astonishing array of cellular diversity. How is such a task accomplished? The use of a single molecule in multiple developmental processes provides a simple means to create diversity from limited resources. Recently, discoveries that particular proteins or entire signaling pathways are used in multiple biological processes have become increasingly common. As a result, this concept of molecular reutilization has moved to the forefront of developmental biology. Experimentally, this presents significant technical difficulties. The activity of a molecule in one developmental pathway can obscure its role(s) in other developmental processes. This knowledge has necessitated the generation of more sophisticated methods with which to address the role of such proteins and pathways in diverse biological processes (Perrimon and Gans, 1983; Golic and Lindquist, 1989; Golic, 1991; Chou and Perrimon, 1992; Brand and Perrimon, 1993; Chou et al., 1993; Xu and Rubin, 1993; Struhl and Basler, 1993; Calleja et al., 1996; Rørth, 1996).

Historically, biologists have relied upon the ability to create mosaics, organisms in which the genotype varies in a cell- or tissue-specific manner, to circumvent such problems. In organisms where genetic analysis is feasible, the induction of mitotic recombination has been used to produce mosaics. When a cell heterozygous for a particular mutation undergoes mitotic recombination, the resulting chromosomal exchange coupled with assortment and division can result in homozygosity for one of the progenitors. Classically, mitotic recombination has been induced by X-irradiation. However, X-irradiation results in a low frequency of mosaicism and high frequency of cell death. These technical problems have been circumvented in Drosophila melanogaster using the yeast site-specific recombinase, FLP (Golic and Lindquist, 1989; Golic, 1991; Chou and Perrimon, 1992; Xu and Rubin, 1993). FLP catalyzes recombination at sequence motifs termed FRTs. Integration of FRTs in a cis or trans configuration within the genome provides a direct means to initiate site-specific mitotic recombination. As such, induction of FLP expression via a ubiquitous (heat-shock, hs) promoter is capable of significant increases in the frequency of mosaicism without the associated lethality of X-irradiation. However, as with X-irradiation, this approach is still limiting. Ubiquitous induction of FLP via heat shock provides only a limited degree of spatial and temporal control. While the use of heat shock can direct mosaic patches that are widely distributed throughout the animal, it typically does not result in the population of any one tissue with all or mostly mutant tissue. Furthermore, the induction of mosaic patches by heat-shock-directed FLP may cause lethality by virtue of the extensive amount of tissues that are made mosaic.

By selectively targeting FLP activity to specific cells or tissues using the GAL4 system (Brand and Perrimon, 1993), we have developed a genetic system that overcomes these limitations. With this approach, the loss of gene activity can be directed to any process of interest. As we document below, the ‘directed mosaic’ approach provides an efficient way to bypass the vital requirements of lethal mutations and assay their effect on specific tissues during later stages of development.

Previous work has suggested that establishment of the body axes in Drosophila melanogaster requires cellular communication between the germline-derived oocyte and its overlying somatic-derived follicular epithelium (Ray and Schupbach, 1996). The germline contribution of any gene to these processes can be examined through hs-induced mosaics and selected for using the dominant-female sterile mutation, ovo_D1 .(Chou and Perrimon, 1992). In contrast, no such directed or selective system exists for the analysis of gene requirements within the follicular epithelium. To identify vital loci required for the somatic components of this intercellular signaling and to complement the germline mosaic system, we focused our efforts with the directed mosaic system on the follicular epithelium. Using two different assays for follicular clones, we first document the parameters of the system and establish it as an efficient approach to control the production of mitotic clones in a temporal and spatial fashion. Next, vital loci whose effects on follicular development include alterations in A/P and D/V axis formation, egg chamber organization, cell migration and cell shape were identified using this approach. Support for a requirement of receptor tyrosine kinase (RTK) signaling within posterior follicle cells for axis formation and the specification of embryonic dorsal/ventral (D/V) pattern in part by limited diffusion of a ventralizing activity has also been provided. Finally, to extend the usefulness of this system, we have generated FLP/GAL4/FRT stocks that allow ∼95% of the genome to be tested in follicular mosaics.

The en-GAL4 and e22c-GAL4 lines used in this study were generated in our laboratory and are described in Sanson et al. (1996) and Yoffe et al. (1996), respectively. The lacZ FLP-out cassette line is described in Struhl and Basler (1993). The zeste-white3 allele, zw3^M11, was used for making wing clones and is described in Siegfried et al. (1992). The T155-GAL4 line was provided by J. Urban and G. Technau. The UAS-Green fluorescent protein (UAS-GFP) is described in Yeh et al. (1995). The Ubiquitin-GFP (Ub-GFP) is described in Davis et al. (1995).

The FLP gene was excised from the vector pDM420 (kindly provided by K. Golic) as a SalI (nt 5280) to XbaI (nt 6643) fragment. This fragment was then cloned into the XhoI and XbaI sites of the vector pUAST to generate pUAST-FLP (UAS-FLP). This construct was injected into y, w; Δ2-3, Sb/TM6, Ubx (Spradling, 1986; Robertson et al., 1988). From these injections, a homozygous viable second chromosome line was established. Using this line as a starter and Δ2-3-mediated transposition, a homozygous viable third chromosome line was generated.

For the F₁ .screen for adult phenotypes using T155-GAL4, the 2L FRT chromosomes used are described in Harrison and Perrimon (1993). Males carrying the 2L FRT (E24-4) and the T155-GAL4 were fed 15 mM EMS in 1% sucrose for 12-20 hours, allowed to recover and then mated to females homozygous for the 2L FRT (E24-1) and a second chromosome UAS-FLP insert. From this cross, 6665 F¹ .progeny were scored for adult abnormalities. Any F₁. adults with a phenotype (total = 835) were backcrossed to the homozygous 2L FRT (E24-1) and UAS-FLP line. 527 of these crosses gave rise to progeny. The F² .from these crosses were scored for the presence of the appropriate phenotype (seen in the F¹.) to determine its heritability and penetrance. Only 58% of these 527 lines bred true in the second generation. This is most likely a result of the mutagenic mechanism by EMS. A single EMS lesion results in the modification of only one DNA strand (Ashburner, 1989). As a result, mutational events that occur during meiosis may be seen in only half the cells of a resulting zygote. Consequently, the mutated DNA strand may not be represented in the germline of that mutant and thus will not be inherited by its offspring. Nevertheless, 58% of the mosaic animals in our screen had heritable lesions. The use of X-irradiation or other mutagens that affect both DNA strands should result in a higher proportion of lesions breeding true.

The X chromosome lacZ enhancer-trap line AN365 was provided by T. Schüpbach and is expressed in follicle cells. A recombinant X chromosome carrying AN365 and the FRT¹⁰¹ .was provided by R. Hsu. For the second assay, three different recombinant lines containing the tsl allele tsl⁶²¹ .and the third chromosome FRT^2A .were provided by A. Amon and R. Lehmann.

For the X chromosome-directed mosaic screen, a subset (containing only germ cell viables) of lethals from Chou and Perrimon (1992) was used. Virgin females of the genotype: y, w, lethal, FRT¹⁰¹./FM7 were mated to w, AN365, FRT101/Y; e22c-GAL4, UAS-FLP/CyO males. F₁ .virgin females of the genotype: y, w, lethal, FRT¹⁰¹./w, AN365, FRT¹⁰¹.; e22c-GAL4, UAS-FLP were collected and mated to OreR males. From 2-4 days posteclosion, individual vials were examined for deposited eggs, which were then examined for defects. After this, the ovaries of directed mosaic females representing each lethal line were dissected, stained for lacZ activity and DAPI, mounted in 70% glycerol/PBS and examined for the presence of follicular clones and ovarian defects.

To mark MEK⁷⁵⁶ .follicle cell clones with GFP, the X chromosome line Ub-GFP^33C1 .was recombined onto an X chromosome containing the FRT¹⁰¹.. Virgin females of the genotype: y, w, MEK⁷⁵⁶., FRT¹⁰¹./FM7 ; e22c-GAL4, UAS-FLP/CyO were mated to y, w, Ub-GFP33C1, FRT101/Y. F₁ females were collected and ovaries were dissected from 1-6 days posteclosion. Ovaries were then processed for GFP fluorescence by fixation in 4% methanol-free formaldehyde diluted from a 16% solution with phosphate-buffered saline plus 0.1% Tween 20 or alternatively for in situ hybridizations with a digoxigenin-labelled antisense RNA probe for bicoid and mounted in 70% glycerol/PBS plus Slow-fade (Molecular Probes) (Gonzalez-Reyes et al., 1995). Directed mosaic females for GAP1 were generated in the following manner. Virgin females of the genotype: w; GAP1^B2., FRT^2A./TM3, Sb were mated to w; e22c-GAL4, UAS-FLP/CyO; FRT2A/TM3, Sb males. F₁ .virgin females of the genotype: to w; e22c-GAL4, UAS-FLP/ + +; GAP1^B2., FRT^2A./FRT^2A .were collected and mated to OreR males. Embryos were collected from these F₁.-directed mosaic females and processed for cuticle preparations or immunohistochemistry (Kania et al., 1990; Duffy et al., 1996) using a digoxigenin-labelled antisense RNA probe for twist or a mouse monoclonal antibody to Dorsal. The α-Dorsal antibody was provided by R. Steward, used at dilution of 1/10, detected with fluorescein-labelled α-mouse IgG (Vector Labs) at a dilution of 1/300 and mounted in 70% glycerol/PBS plus Slow-fade (Molecular Probes).

To overcome the limitations of hs-induced FLP expression, we took advantage of the GAL4 gene expression system (Brand and Perrimon, 1993). This binary gene activation system utilizes the yeast activator GAL4 to quantitatively and qualitatively control the transcription of target genes. By fusing the gene of interest to a GAL4-responsive promoter (UAS), its expression can then be directed with an appropriate GAL4 enhancer-trap line.

By building a GAL4-responsive FLP (UAS-FLP), we hoped to provide a high degree of spatial and temporal control over recombination. Fig. 1 documents such control. Using an en-GAL4 line, we directed expression and activity of FLP to the posterior compartment of imaginal discs (Sanson et al., 1996). Fig. 1A documents this directed cis (intrachromosomal) recombination with β-galactosidase (lacZ) expression from a FLP-out cassette developed by Struhl and Basler (1993). Their cassette is organized in the following fashion: actin5C promoter-FRT-Draf₊ .-FRT-lacZ. The combined presence of en-GAL4 and UAS-FLP should result in cis recombination at the flanking FRTs, removal of the Draf₊ .fragment and expression of lacZ due to its subsequent juxtaposition to the actin5C promoter. As expected, lacZ expression was observed and restricted to the posterior compartment indicating that FLP activity was spatially targeted by the en-GAL4 driver (Fig. 1A). In addition, the presence of UAS-FLP and en-GAL4 was required to observe lacZ expression indicating that both components are necessary for directed recombination (data not shown). Similarly, directed trans (interchromosomal) recombination was also demonstrated. Mutant clones of the segment-polarity gene zw3 produced in the developing wing imaginal disc result in the production of ectopic bristles (Simpson et al., 1988). Mutant clones of zw3 generated via en-GAL4 and UAS-FLP resulted in the production of ectopic wing bristles only within the posterior compartment (Fig. 1B). Thus, directed cis and trans recombination can be obtained with the GAL4-responsive UAS-FLP.

Screening a series of random lethal mutations for phenotypes is a powerful genetic tool for dissecting developmental processes. However, the classical approach to recovering such mutations requires generating lethal mutations to be maintained as individual stocks before analysis. Such F2 screens are both time consuming and labor intensive. A more efficient approach is to analyze the first generation of progeny directly for mutant phenotypes (an F1 screen). Due to the recessive and/or lethal nature of most mutations, however, this is not possible without mosaic techniques. As described above, the lack of tissue specificity associated with previous methods for inducing mitotic clones may be a limitation in F1 screens for lethal mutations. Restricting mutant clones to the tissue of interest should retain high viability of mosaic animals.

To address the feasibility of such a directed screen in adult tissues with the UAS-FLP, we carried out the following scheme. Males carrying a FRT inserted at the base of 2L and the T155-GAL4 line were mutagenized with EMS. These males were then crossed to females also carrying the 2L FRT and the UAS-FLP construct. In the progeny of this cross, T155-GAL4-directed UAS-FLP expression results in adults mosaic for mutations induced on 2L. The GAL4 insert (T155) used in this screen directs GAL4 activity to the imaginal discs during mid and late larval stages. 6665 F1 adults from this screen were scored for visible phenotypes and the results are summarized in Table 1. Over 12% (835) of the adults recovered in the screen had easily scored visible phenotypes. These phenotypes represented a broad spectrum of developmental abnormalities, ranging from eye defects to homeotic transformations. From this, we conclude that directed mosaicism provides a potent approach for genetic investigations of specific developmental processes.

Having demonstrated the general feasibility of this system, we wanted to concentrate our efforts on the primary objective of this approach: adult tissue-specific screens for vital loci.

To complement the germline mosaic technique, we chose to further develop this method to direct mutant clones to the follicular epithelium of the adult female. We screened for and identified GAL4 enhancer-trap lines that are somatically expressed in region 2A of the germarium, the location of the follicular stem cells (Margolis and Spradling, 1995), and not in the germline (Fig. 2, data not shown) using UAS-LacZ, UAS-GFP and lacZ FLP-out cassettes as reporters. Using one of these GAL4 lines, e22c-GAL4, the following questions were addressed: (1) can clones be directed to the follicular epithelium and (2) how efficient is clone production within the epithelium (what are the size and frequency of clones)?

To answer these questions, we used two viable assays. In the first assay, one of the FRT chromosomes was marked with an enhancer-trap line that is expressed within the follicular epithelium. This allowed the identification of follicular clones on the basis of lacZ staining. The results of this analysis are shown in Fig. 3. From these data, a number of points can be made. First, clones can be directed to the follicular epithelium. Second, since egg chambers and ovarioles whose entire follicular epithelium is monoclonal can be recovered, e22c-GAL4 must be expressed in the follicular stem cells (or their precursors). Third, the frequency of monoclonal egg chambers and ovarioles changed as the females aged, reaching a maximum of 36% and 31%, respectively.

These results compare favorably with the results of Margolis and Spradling (1995), who observed an increase in monoclonal ovarioles from initially mosaic ovarioles induced via hsFLP. They also conclude that there are 2-3 follicular stem cells per ovariole and that follicular stem cells have a finite lifespan. With e22c-GAL4-directed mosaics, we observed an increase in monoclonal ovarioles (either homozygous state) as the females aged. This strongly suggests that when FLP is driven by e22c-GAL4, the follicular stem cells were initially heterozygous and after eclosion FLP-mediated mitotic recombination occurred during subsequent rounds of stem cell division. The parent stem cell and its progenitor then adopted contrasting homozygous states. The conversion of the stem cell to a homozygous state, coupled with its finite lifespan resulted in the transformation of initially mosaic ovarioles to a monoclonal state as the females aged. This mechanism is also supported by our observation that none of the day 1 chambers (and thus ovarioles) scored were completely clonal, as might be predicted if conversion were taking place prior to the establishment of a stem cell population. Instead 99% of these day 1 egg chambers were mosaic (data not shown). Small clones were present in these day 1 egg chambers indicating that mitotic recombination must also occur within the follicular cells during S1-S6. Consistent with this, e22c-GAL4 activity can be detected during these stages. Finally, no follicular clones were seen in control progeny lacking any one of the four components of the system (FRTs, UAS-FLP, e22c-GAL4) demonstrating that all elements are necessary for directing mosaics.

To further characterize the parameters of this system, a second assay employing a FRT chromosome bearing a mutation in the gene torsolike (tsl) was used. Tsl activity is required in the follicle cells for the establishment of the embryonic anterior and posterior termini (Stevens et al., 1990; Martin et al., 1993; Savant-Bhonsale and Montell, 1993). Loss of Tsl activity within the follicle cells results in the loss of these embryonic terminal structures. By scoring the cuticular phenotype of embryos derived from directed tsl₋ .mosaic females, we could indirectly assay the frequency and size of mutant clones of tsl within the follicular epithelium (Fig. 3B). As shown in Fig. 3B, results similar to those seen in the lacZ assay were obtained. Moreover, the results with tsl clones also suggest that monoclonal frequencies reach a maximal level of ∼30% around 6 days posteclosion. From these two assays, we conclude that GAL4-directed FLP-mediated mitotic recombination is an efficient approach to regulate the spatial and temporal induction of mosaics.

The results above indicate that mosaics can be generated within the follicular epithelium with a frequency and efficiency suitable for genetic screens. However, both of these assays involved the use of a non-lethal marker (lacZ) and mutation (tsl). To use the directed mosaic approach to identify vital loci required for follicular patterning, we must be able to obtain mosaic adults with clones of lethal mutations within the follicular epithelium. If e22c-GAL4 (or any other GAL4 line used to make follicular mosaics) was expressed during earlier developmental stages, this could result in the production of clones in other tissues and lethality.

Therefore, we carried out a small-scale directed mosaic screen with e22c-GAL4 as the clone driver (see Materials and Methods). A collection of lethal mutations from Chou and Perrimon (1992) induced on an X chromosome carrying FRT₁₀₁ .(14A-B) was screened for: (1) the presence of mosaic adults, (2) follicular clones and (3) phenotypic defects. Out of the 93 lines carrying recessive lethal mutations, adult females of the appropriate genotype were recovered for all lines. Similar results were obtained with another GAL4 line (J. Duffy, unpublished). Of these 93 lines, 89 produced adult mosaic females with follicular clones. From these lines, a wide variety of defects, ranging from alterations in A/P and D/V polarity, egg chamber organization, cell migration and cell shape were observed (Figs 4, 5; data not shown). As described below, three of these lines, 667, 817, 756, correspond to previously identified components of RTK signaling. In addition, two lines, 968 and 1070, exhibiting defects in the overall shape of the chorion and follicular cells were genetically mapped to the region containing myospheriod, the β subunit of the position-specific integrins PS1 and PS2 (Fig. 4G,H). Consistent with the possibility that these lines have mutations in mys, FLP-induced imaginal clones of 968 and 1070 produced wing blisters similar to that seen with mys clones (Brower and Jaffe, 1989).

With the successful results for this X chromosome screen, we have now generated FLP/GAL4/FRT stocks that allow for directed mosaics of mutations on 2L, 2R, 3L and 3R (Table 2). Each line carries an UAS-FLP, a follicular GAL4 driver and a FRT insertion. With a single cross of the appropriate FLP/GAL4/FRT line to a FRT-bearing mutant line, follicular mosaics for that mutant can be generated in the F1. This provides a means to screen ∼95% of the genome for roles in follicular development via directed mosaics.

Establishment of both A/P and D/V polarity within the oocyte depends upon signal transduction through Torpedo, the Drosophila epidermal growth factor receptor (Egfr) (Schupbach, 1987; Gonzalez-Reyes et al., 1995; Roth et al., 1995). Evidence exists supporting roles for RTK downstream molecules within the follicular epithelium for the establishment of axis polarity. Polarity disruptions, evident when the activity of these molecules is reduced within both the germline and the follicular epithelium, are not observed when the germline is lacking their activity; thereby supporting the notion that they are functioning within the follicular epithelium. Consistent with this, our screen uncovered previously unidentified alleles of Drosophila Map kinase kinase (DMEK) and the phosphatase Corkscrew (Csw). For three of the lethals that we tested, directed mosaic females laid eggs with ventralized chorions and had egg chambers with disruptions in A/P polarity. Complementation analysis with two of the lethal lines, 817 and 667, indicated that they were alleles of csw, while genetic mapping, complementation and transposon-mediated rescue analysis of the third lethal, 756, indicated that it was an allele of DMEK. DMEK has been previously implicated in D/V patterning through the use of a temperature-sensitive mutation (Hsu and Perrimon, 1994), while the role of Csw in establishing polarity has been demonstrated through hs-regulated Csw expression (Perkins et al., 1996). By identifying alleles of these loci through a directed mosaic screen, we have confirmed a tissue-specific and spatial requirement for these molecules in patterning both the A/P and D/V axes, consistent with previous hypotheses. Correlation of the location of MEK clones with the presence of a disrupted A/P axis also provided evidence to support the notion that the A/P axis organizing activity is generated by the posterior follicle cells (Fig 5).

A similar case for a somatic role of a negative regulator of RTK signaling, GAP1, has also been made for D/V patterning (Chou et al., 1993). During mid-oogenesis ventral follicular fates are specified by inactivation of the Ras-Raf-MEK signaling cassette, proposed in part by the inhibitory action of GAP1. This in turn results in the production of a ventralizing activity during early embryogenesis. It is this activity that directs the graded nuclear to cytoplasmic transport of the Dorsal (Dl) protein, which regulates embryonic D/V axis formation (Morisato and Anderson, 1995). As a negative regulator of the Ras-Raf-MEK signaling pathway reduced GAP1 activity in the germline and follicular epithelium results in embryonic dorsalization. In contrast, germline reduction alone has no effect on D/V polarity, supporting its proposed role in follicular patterning (Chou et al., 1993). To extend these observations, we generated e22c-GAL4-directed mosaics of GAP1 in the follicular epithelium. From GAP1-directed mosaic females, we observed dorsalization of the embryonic cuticle. However, in addition to complete dorsalization, we also observed cases of extremely localized dorsalization (Fig. 6). Consistent with these observations, expression of the ventral molecular marker twist was lost in ventral regions in embryos derived from these females (data not shown).

These results suggested that a localized loss of GAP1 activity within the follicular epithelium led to a localized inactivation of the embryonic ventralizing activity. Previous reports using embryos lacking the Toll receptor have suggested that this ventralizing activity is generated as a diffusible gradient from a ventral source (Stein et al., 1991; Stein and Nusslein-Volhard, 1992). Alternatively, recent results support the possibility in which the embryonic gradient of Dl localization could be generated by a ventral gradient within the follicular epithelium that transmits information in a cell-autonomous fashion (Queenan and Schupbach, 1997). This information would only be capable of diffusion in the absence of a receptor, but not in the presence of a wild-type receptor. By generating localized GAP1 clones (localized dorsalization) within the follicle cells and observing the embryonic nuclear translocation pattern of Dl, we were able to distinguish between these potential D/V patterning mechanisms. In embryos derived from directed mosaic GAP1 mutant females, localized disruptions in the nuclear translocation of Dl were observed (Fig. 7B,C). Specifically, ventral regions without detectable nuclear Dl localization were observed. However, in these regions, sharp boundaries of Dl translocation were not present as is predicted from the cell-autonomous model. Rather, these disruptions were always accompanied by transition zones showing a graded translocation of Dl into the wild-type pattern, indicating a ventralizing activity capable of diffusion even in the presence of a wild-type Toll receptor (Fig. 7B,C).

As exemplified by the seminal screens for lethal mutations affecting embryonic pattern in Drosophila (Nusslein-Volhard and Wieschaus, 1980), the ability to genetically dissect the process of development has provided enormous insight into its underlying molecular mechanisms. It is this embryonic requirement for many of the molecules and signaling pathways that also contribute to later developmental stages that has hampered a similar systematic analysis of post-embryonic development. For such an analysis, one must be able to efficiently bypass the early/vital requirements of any locus to determine its function at later stages of development. By targeting the activity of the FLP recombinase in a spatial and temporal fashion, we have demonstrated that it is possible to efficiently generate directed mosaics for vital loci in most adult tissues.

Although the primary goal of this work was to establish a more directed approach for producing mosaics than previously available, our general adult screen compares favorably with an untargeted mosaic screen using FLP driven by a heat-shock promoter (Xu and Rubin, 1993). While that screen used X-irradiation to generate mutations and we used EMS, the frequency of lethal mutations was the same as for the screen reported here, one lethal mutation per chromosome arm. In contrast, in the Xu and Rubin screen, ∼1% of the adult flies had visible defects – with ∼0.6% having eye defects, while in our directed screen ∼12% of the adults had visible defects, and 6.8% had eye defects. Thus, the directed mosaic technique yielded detectable phenotypes with ∼12-fold higher frequency than the heat-shock generation method. While the rate of eye defects is sufficiently high by either method to feasibly conduct a screen, rarer classes of phenotypes, such as homeotic transformations and pattern duplications, were only detected with the directed mosaic method. Furthermore, another advantage in using the directed mosaic method for screening was the consistency with which mutant clones were produced. In contrast to heat-shock induction with which variation is seen both within an individual and a population, the use of GAL4-mediated FLP expression results in consistent and reproducible clones both within individuals and populations. For example, in flies with eye phenotypes, both eyes routinely displayed defects in our screens. Such consistency makes it easier to distinguish true genetic lesions from non-heritable developmental abnormalities, as well as to carry out phenotypic analyses.

Our two assays using viable markers gave consistent results regarding the production of mosaic tissue within the follicular epithelium. Temporal analysis of clonal production via e22c-GAL4 yielded a number of relevant observations. First, shortly after eclosion most (>90%) egg chambers are mosaic containing small clones. Second, as the females age the frequencies of monoclonal egg chambers and ovarioles increase to reach maximal monoclonal frequencies of ∼30%. These results indicate that temporally the system can be considered efficient from two distinct standpoints: the induction of small clones or monoclonal epithelial tissue. In addition, this temporal aspect provides us with a mechanism to control clonal size simply by regulating the age of females examined.

Previous work from a number of laboratories has provided an emerging model of how axis formation is established during oogenesis (for review see (Morisato and Anderson, 1996; Ray and Schupbach, 1996). Prior to stage 7 of oogenesis, a germline-derived signal, encoded by the TGF-alpha-like molecule, Gurken, specifies posterior follicle cell fates through the activity of the Top/DER RTK. This appears to occur through the activation of the downstream Ras-Raf-MEK signaling cassette within the posterior follicle cells. In response to this Grk-mediated signal, a secondary signal is sent from the posterior follicle cells back to the oocyte. The reception of this secondary signal then results in a reorganization of the microtubule network of the oocyte and the definition of the A/P axis. Concomitant with these events is a relocalization of the oocyte nucleus with its associated-Grk signaling activity to the dorsal-anterior cortex. In this new position, Grk and Top/DER-mediated signaling specifies dorsal follicle cell fates. The absence of this signal on the ventral surface results in the assumption of ventral follicular fates. From these follicle cells, a ventralizing activity is generated that is responsible for the activation of the Toll receptor. The activation of Toll then creates a ventral/dorsal gradient of nuclear to cytoplasmic Dorsal within the developing embryo to specify fates along the D/V axis. While a firm basis exists for this model of the serial relay of positional information, numerous points remain to be confirmed and questions answered with respect to the creation and relay of these signaling asymmetries.

Towards this end, we have used the directed mosaic system to provide further insight into how the follicular epithelium contributes to pattern formation. Specifically, we have directly supported the requirement for two members of the DER RTK signaling pathway, MEK and Csw, within the posterior follicular epithelium for axis organization. Furthermore, by limiting clones to the posterior of the follicular epithelium, we have established that it is posterior somatic cells that require RTK signaling activity for an A/P axis-organizing activity to be sent back to the germline-derived oocyte.

In addition, our mosaic analysis of GAP1 has provided the following observations. First, in embryos derived from GAP1-directed mosaic females a loss of Dl nuclear localization can be observed on the ventral surface. In addition, these ventral disruptions in Dl localization were always bordered by transition zones showing a graded translocation of Dl into the wild-type pattern. If the ventralizing signal were not diffusible then any disruptions in Dl nuclear localization would be bordered by sharp boundaries. In contrast, if the ventralizing signal was freely diffusible one would not expect to observe the loss of Dl nuclear localization in these embryos. In such cases, active signal would diffuse into regions unable to generate their own signal. Rather than support either of these extreme alternatives, our observations suggest a more moderate view: that active ventralizing signal is capable of short-range diffusion even in the presence of the Toll receptor and that it is present in limiting amounts.

Finally, we have identified vital loci affecting a number of processes involving the development and organization of the follicular epithelium. As some of these loci have chorion phenotypes that represent D/V or A/P alterations, but have no effects on embryonic axis formation they are excellent candidates for downstream targets responsible for elucidating phenotypic subsets of the axis-forming signals. With directed mosaic stocks that allow for the identification of vital loci throughout ∼95% of the genome, a genetic approach to follicular development and its contributions to oocyte/embryonic patterning is feasible. Moreover, as the specificity of clonal production relies upon a GAL4 driver, this approach takes advantage of an extensive collection of GAL4 lines throughout the Drosophila community. Thus, we anticipate this system providing an efficient, genetic means to study most developmental processes.

We thank Angelika Amon and Ruth Lehman, Ilan Davis and Pat O’Farrell, Gabrielle Boulianne, Andrea Brand, Trudi Schüpbach and Ken Yoffe for providing fly strains. We thank Ruth Steward for providing the Dorsal antibody and Robert Ray for the twist DNA. J. B. D. thanks Bob Boswell for providing space and support for the final stages of this work. We thank Bob Boswell and Petra Belady for comments on the manuscript. This work was supported by a Damon Runyon-Walter Winchell Cancer Fund Fellowships DRG #1213 (J. B. D.), and DRG #1154 (D. A. H.). N. P. is an Investigator of the Howard Hughes Medical Institute.