fringe and Notch specify polar cell fate during Drosophila oogenesis

Oogenesis in Drosophila involves collaboration between two adjacent cell populations, the germline cells, and the somatic follicle cells, which surround them. Throughout oogenesis, many follicle cells adopt specialized fates and perform crucial functions in egg chamber morphogenesis, follicle cell patterning, or signaling between somatic and germline cells (reviewed in Dobens and Raftery, 2000). The polar cells are pairs of cells at the anterior and posterior tips of each follicle. Although they are among the first somatic cells to adopt a distinct fate, the molecular basis for establishment of polar cell fate has not previously been identified.

Both germline and somatic cells derive from stem cells that are located in the germarium. Germline stem cells give rise to cystoblasts, which then undergo four rounds of division with incomplete cytokinesis to produce a cyst of 16 interconnected cells. One of these cells will become the oocyte, and the remainder develop into nurse cells. Each germline cyst becomes surrounded by approximately 16 somatic follicle cells, forming an individual follicle, or egg chamber, which then buds off from the germarium (Fig. 1). Follicles remain connected to each other by short stalks of somatic cells (stalk cells), and a chain of follicles at different developmental stages, together with the germarium, comprise an ovariole (Fig. 1; reviewed by King, 1970; Spradling, 1993). Several genes that are required for follicle formation have been identified, one of these is Notch.

Notch is a transmembrane receptor protein that is involved in a wide range of cell fate decisions during animal development (reviewed by Artavanis-Tsakonas et al., 1999). In Drosophila, two ligands are known that regulate its activity, Delta and Serrate. In imaginal discs, the ability of these ligands to activate Notch is modulated by Fringe (FNG) (reviewed by Irvine, 1999; Panin and Irvine, 1998). Expression of FNG inhibits the ability of cells to respond to Serrate, and potentiates the ability of cells to respond to Delta. In wings, eyes and legs, these effects of FNG, together with the distribution and regulation of Notch ligands, result in the establishment of stripes of Notch activation along borders of fringe expression (reviewed by Irvine, 1999). FNG is a glycosyltransferase that can directly modify Notch, and this modification of Notch presumably influences its interactions with ligands (Bruckner et al., 2000; Moloney et al., 2000).

Requirements for Notch signaling during oogenesis have been analyzed previously using conditional alleles of Notch and Delta, and by ectopic expression of Delta or constitutively activated forms of Notch. Notch and Delta are required for the pinching-off of individual follicles from the germarium, and for the generation of stalks between adjacent follicles (Bender et al., 1993; Xu et al., 1992). Overexpression of Delta or activated Notch generates abnormally long interfollicular stalks, and the cells in these long stalks do not exhibit a normal stalk fate (Keller Larkin et al., 1999; Larkin et al., 1996). Reduction of Delta or Notch function also results in a hyperplasia and loss of epithelial character among posterior follicle cells (Goode et al., 1996; Ruohola et al., 1991). Notch function is also required for the correct establishment of terminal follicle cell fates at both the anterior and posterior ends of the follicle (Gonzalez-Reyes and St Johnston, 1998; Keller Larkin et al., 1999), and for the establishment of developmental axes of the oocyte (Jordan et al., 2000).

However, these studies have not determined in which follicle cells Notch is actually required for any of these steps, and they have all been conducted with conditional alleles that do not completely eliminate Notch signaling. In imaginal discs, analysis of the requirements for fng have proved invaluable for elucidating the function and regulation of Notch signaling (reviewed by Irvine, 1999). In this work, we describe an analysis of the expression and requirements for fng during early stages of Drosophila oogenesis. Our results show that fng is required to specify polar cell fate. Analysis of clones of cells mutant for a null allele of Notch, or of clones that express a constitutively activated form of Notch, imply that this influence of fng reflects its modulation of Notch signaling, and demonstrate that Notch too is required for polar cell fate. Our studies further show that the specification of polar cells in the germarium is necessary for follicle formation.

fng² and fng⁵² are hypomorphic alleles, fng¹³ and fng⁸⁰ appear to be null alleles (Irvine and Wieschaus, 1994). fng^M69 is a temperature-sensitive allele that behaves as a strong allele at 29°C (V. Papayannopoulos and K. D. I., unpublished), and was originally isolated by J. Jiang and G. Struhl. Flies carrying fng^M69 were raised at 18°C and then switched to 29°C after eclosion, all other crosses were conducted at 25°C. fng-lacZ^RF584 is an enhancer trap line isolated by U. Gaul, and confirmed as an insertion upstream of the fng ORF by inverse PCR and DNA sequencing (X. Wang and K. D. I., unpublished). N^55e11 is a null allele (Heitzler and Simpson, 1991). The following fly strains were also used: Oregon-R, 93F/TM3 (Ruohola et al., 1991), A101/TM3 (Bier et al., 1989), EQ1Gal4/TM3, T155Gal4, E4Gal4, 55BGal4 and e22cGal4 (Queenan et al., 1997), E(spl)mβ-lacZ (Cooper et al., 2000).

fng¹³ or N^55e11 clones were generated by Flipase-mediated mitotic recombination on FRT80-3 or FRT18A chromosomes, respectively, and marked using π-Myc or arm-lacZ transgenes (Xu and Rubin, 1993). Ectopic expression of fng and activated Notch was performed by generating Flip-out Gal4 clones in animals carrying UASfng27 (Kim et al., 1995) or UAS-ΔN34a (Doherty et al., 1996) and AyGal4 UASGFP (Ito et al., 1997) transgenes. Flipase expression was induced by heat shocking 2-day-old females at 38°C for 1 hour to generate mutant clones, and at 32.5°C for 30 minutes to generate Flip-out clones.

β-galactosidase activity detection and Hoechst staining were carried out as described previously (Grammont et al., 1997). F-actin was labeled with Texas Red X-phalloidin (1:20 Molecular Probes). Whole-mount in situ hybridization to follicles and antibody staining was performed as previously described (Sahut-Barnola et al., 1995), using goat anti-β-galactosidase (1:1000, Biogenesis), rabbit anti-Myc (1:100, Santa Cruz), mouse anti-Fasciclin III (1:25, Patel et al., 1987), mouse anti-Bicaudal D (1:40, Suter and Steward, 1991), rat anti-E-Cadherin (1:50, Oda et al., 1993) and rat anti-Serrate (1:1000, Papayannopoulos et al., 1998).

To assess the requirements for fng during oogenesis, we examined the phenotypes of three hypomorphic alleles: fng², fng⁵² and fng^M69. Null alleles of fng die as first instar larvae, but these alleles are viable. Examination of the ovarian morphology of young (<5 days old) fng mutant females revealed three consistent defects: compound follicles, long and abnormal interfollicular stalks, and loss of epithelial structure at the posterior of the follicle. The frequency and severity of these defects ranged among the genotypes examined, and the mutations appear to form an allelic series comparable with that derived from studies of wing phenotypes (Table 1 and data not shown; Irvine and Wieschaus, 1994).

In wild-type ovaries (Fig. 2A,B), each follicle contains 16 germline cells. In stronger fng mutants, follicles often contain an abnormal number of germline cells (Table 1, Fig. 2C,D). In most cases (31/36), these abnormal follicles include more than 16 germline cells (Fig. 2C), although we also found occasional examples (5/36) of follicles containing less than 16 germline cells (Fig. 2D). Follicles with multiples of 16 germline cells contain extra oocytes (Fig. 2C). Follicles that contain less than 16 germline cells are always adjacent to follicles that contain more than 16 germline cells, with the total number of germline cells in adjacent follicles being a multiple of 16. Along the length of the entire ovariole, the oocyte:nurse cell ratio is normal (1:15). Phalloidin staining reveals that all of the germline cells are connected by normal ring canals (not shown). Together, these observations indicate that the presence of abnormal numbers of germline cells derives from defects in the enclosure of germline cysts by somatic cells.

In wild-type ovaries, a single column of four to six cells separates adjacent follicles (Fig. 2B). In fng mutant ovaries, long stalks can form that contain, on average, 10 cells (Fig. 2E). Thirteen of 58 stalks examined exhibit an additional defect, in which the stalk cells were not correctly organized. Instead of forming a single column of cells, they form clusters (Fig. 2F), or a double stalk (Fig. 2G). Thus, fng influences both the number and the organization of stalk cells.

In wild type, follicle cells form an even, single-layered columnar epithelium around the germline cells (Fig. 2A,I). By contrast, in fng mutant ovaries the posterior follicular cells can form extra cell layers (Fig. 2H,J). This defect is most obvious in stage 9 and later follicles, but can also be detected earlier (Fig. 2H). Thus, fng is required to maintain a monolayer of epithelial cells around the oocyte.

Although some aspects of fng expression during oogenesis have been described recently (Jordan et al., 2000; Zhao et al., 2000), our analysis is more detailed, and has been performed using both in situ hybridization to mRNA and an enhancer trap insertion, fng-lacZ^RF584, that appears to mimic fng expression throughout development (K. D. I., unpublished). fng expression is first detected in the germarium. This staining appears in some germaria as a solid block in somatic cells throughout region II (Fig. 3A), while in others it appears as a broad stripe throughout region IIa, absence of staining in region IIb, and a thin stripe of staining between regions IIb and III (Fig. 3B). Oogenesis is a dynamic process in which germline cysts move towards the posterior of the germarium as they mature. The variability in fng expression presumably reflects this dynamism, and derives from individual germaria having been frozen by fixation at different points of follicle development. fng expression along the IIb-III border corresponds to the somatic cells that invaginate between and separate adjacent cysts. At the posterior end of a stage 1 follicle (region III of the germarium), fng expression appears to be restricted to the polar cells and stalk cells (Fig. 3B). Between stage 2 and 3 follicles, fng expression is detected in the polar and stalk cells (Fig. 3C,E,G). From stages 3 to 6 of oogenesis, fng expression is restricted to the polar cells (Fig. 3C,D). Overall, fng-lacZ^RF58expression appears similar to that of fng mRNA. However, the enhancer trap also exhibits staining in somatic cells at the anterior of the germarium (Fig. 3E) that is more difficult to detect convincingly above background by in situ hybridization. In addition, some enhancer trap staining is detected in main body follicle cells up to stage 3; we attribute this to the broad transcription of fng in the germarium, and the perdurance of β-galactosidase in these cells.

After stage 6, fng expression expands at the posterior end and by stage 8 is expressed in all of the follicle cells that contact the oocyte (Fig. 3F,J). At the anterior end, fng expression remains restricted to the polar cells (Fig. 3I,J). However, at stage 9 and 10a of oogenesis, fng expression is also detected in all of the follicle cells that will migrate towards the oocyte (Fig. 3J). Among the border cells (Fig. 1), fng is expressed only in the polar cells (Fig. 3I,J). In stage 10b follicles, fng expression becomes repressed in dorsal anterior follicle cells above the oocyte nucleus, but is maintained in all other follicle cells around the oocyte (Fig. 3K). As fng expression expands, fng-lacZ^RF58expression again lags fng expression detected by in situ hybridization, as during stages 7 and 8, not all cells that contact the oocyte express lacZ (Fig. 3I), and staining in most follicle cells is not detected until stage 10a (data not show).

Although occasionally some faint staining was detected in germline cells by in situ hybridization, the weak and variable nature of this staining leads us to infer that it reflects a background of nonspecific hybridization. This inference is also supported by the absence of any detectable expression in germline cells of fng enhancer trap lines (Fig. 3E,I).

Prior stage 5, Notch and Delta are expressed in all somatic cells (Bender et al., 1993; Xu et al., 1992). During stages 6 to 8, Notch is expressed most strongly in posterior cells, where fringe is, while Delta is repressed in these cells. Using antibodies directed against Serrate (Papayannopoulos et al., 1998), we determined that it is expressed similarly to Delta: in the germarium and stage 2 follicle it is expressed in all somatic cells (Fig. 3H), and later it is repressed in posterior cells (not shown).

To elucidate the relationship between the requirements for fng in different populations of follicle cells and fng mutant phenotypes, and to analyze the consequences of complete loss of fng, we constructed genetic mosaics. These were generated by FLP-FRT-mediated mitotic recombination in animals heterozygous for fng¹³ (Irvine and Wieschaus, 1994). The resulting fng mutant clones were marked by the absence of expression of a Myc-tagged gene (π-Myc) or lacZ. Follicles exit the ovariole after they complete their development, hence clones that are induced in developing follicles are lost in about 7 days (transient clones) (Margolis and Spradling, 1995). However, recombination in stem cells generates a continuous supply of mutant tissue (permanent clones). To ensure that all mutant cells examined derived from clones that had been lacking in fng function throughout oogenesis, we restricted our analysis to permanent clones by waiting 10 days after the induction of recombination.

To rule out the possibility that fng functions in the germline, we identified mosaic follicles in which the somatic cells carried a wild-type copy of fng, while the germline cells were homozygous mutant. These follicles develop normally (Fig. 4A). By contrast, mosaics in which the germline cells carry a wild-type copy of fng but some somatic cells are homozygous mutant for fng exhibit defects (see below). Thus, fng is required only in the soma for oogenesis.

Since each germarium contains two somatic stem cells, permanent clones result in ovarioles in which approximately half of the cells are homozygous and half are heterozygous. In wild-type controls, these clones are distributed such that individual follicles can be composed of a mixture π-Myc⁺ and π-Myc⁻ cells (mosaic follicles), or they can be composed entirely of π-Myc⁺ cells or π-Myc⁻ cells (Fig. 4B). Indeed, up to three consecutive π-Myc⁺ or π-Myc⁻ follicles in a row can be observed (not shown). In contrast, among several hundred ovarioles analyzed, follicles composed entirely of fng mutant somatic clones were never observed. Thus, fng is absolutely required in the soma for follicle formation. Follicles that are mosaic for fng yet appear morphologically normal can sometimes be recovered. However, in all cases (>50 follicles scored) the fng- cells are excluded from the termini of the follicle. Conversely, follicles that include fng⁻ cells at their termini, or in regions that we infer would normally have contributed to terminal structures, always exhibit defects in follicle formation or the development of polar and stalk cells. We have focused on requirements for fng during early oogenesis, and analysis of the relationship between the locations of wild-type and mutant cells and the defects observed define distinct requirements for fng in different subpopulations of cells.

Among mosaic follicles with defects in follicle formation, 75% (40/53) contained multiple 16-cell cysts (Fig. 4C). In all of these compound follicles (40/40), fng⁺ cells are located at both extremities of the follicle (Fig. 4C,D), and fng⁻ cells contribute only to non-terminal regions. These observations imply that fng⁻ cells that would normally have contributed to terminal regions are impaired in their ability to migrate in between and separate germline cysts. In the remaining mosaic follicles with defects, germline cells are separated by follicle cells, but this separation is abnormal, leading to the formation of stalkless follicles. In 10% of cases (5/53), 16 cell germline cysts were separated into follicles by a bilayered epithelium of somatic cells (Fig. 4E). However, the epithelial cell layers were fused, and no stalk formed. In every case, the fused epithelial bilayer is composed entirely of fng⁻ cells in the older follicle and a mixture of fng⁺ and fng⁻ cells in the younger follicle. In 15% of cases (8/53), cysts were separated by a monolayered epithelium that is composed entirely of fng⁻cells (Fig. 4F). Altogether, these observations indicate that fng is required for the normal separation of germline cysts into follicles.

In region IIb-III of the germarium, fng expression is maintained in the cells that separate adjacent follicles. The polar and stalk cells presumably belong to this population, as they end up between follicles in the vitellarium (Fig. 1). Polar and stalk cells derive from a shared population of precursor cells, and together they form a lineage distinct from that of other follicle cells (Tworoger et al., 1999). This allows for the formation of mosaic follicles in which only polar and stalk cell precursors are mutant for fng, and all other follicle cells are wild-type (or vice versa). We found 11 mosaic follicles in which we were able to infer that only polar-stalk precursors were mutant for fng, whereas all other follicle cells were wild type. These follicles were separated from adjacent follicles by at least some fng⁻ stalk cells, although adjacent egg chambers appeared to have fng⁺ polar cells (Fig. 5A). These follicles exhibit a consistent defect in which the epithelium is incomplete at the termini next to the fng⁻ stalk cells, leaving some germline cells uncovered by follicle cells (Fig. 5A). Since this gap in the follicle epithelium occurs where the polar cells are normally found, it suggests that polar-stalk precursor cells mutant for fng are unable to differentiate into polar cells.

To test this, we used expression of Fas III as a marker of polar cell fate. Although a neuralized-lacZ enhancer trap line (A101), appears to be a more specific marker of polar cell fate during oogenesis (Ruohola et al., 1991), we were unable to construct the genotypes that would allow us to examine it in fng mosaics. Nonetheless, Fas III is expressed specifically in the polar cells from stages 3 to stage 10A of oogenesis in wild-type egg chambers (Ruohola et al., 1991), and even in Notch mutants appears to be a reliable marker of polar cell fate from stages 3 to 7 (see below). Fas III staining was absent from fng⁻ follicle cells, and importantly was absent from mosaic follicles with gaps in the anterior or posterior follicular epithelium (Fig. 5B). This observation implies that fng is required for normal polar cell fate.

We also identified six mosaic follicles in which all of the main-body follicle cells were fng⁻, and only the polar and stalk cells at each extremity were fng⁺ (Fig. 5C). Strikingly, all of these follicles appear normal and include Fas III-expressing polar cells (Fig. 5C). Thus, expression of fng exclusively in the polar-stalk lineage is sufficient for follicle formation. Moreover, we found 15 examples in which fng⁺ polar cells formed at both ends of a stalk, but most or all the stalk cells were fng⁻ (Fig. 6A). Thus, fng is not required within the stalk for polar cell formation either. The strict correlation between Fas III staining and the presence of two terminal fng⁺ cells indicates that fng is required cell-autonomously within the polar cells for their differentiation. This conclusion is also supported by the observation that only two fng⁺ cells are required to begin or end a compound follicle (Fig. 4D).

fng is also transiently expressed in stalk cells, and fng hypomorphs can have abnormally long and disorganized stalks. When all of the main-body follicle cells are fng⁻, but all of the polar and stalk cells are fng⁺, the stalk develops normally (6/6). In contrast, mosaics in which all of the main-body follicle cells are fng⁺ but some of the stalk and polar cells are fng⁻ usually have elongated and/or disorganized stalks (Fig. 5A). Thus, the requirement for fng for stalk development maps to the polar-stalk lineage.

To determine where fng is required, we compared the structure of stalks in the presence or absence of polar cells. Stalks in which the polar cells at both ends appeared to be fng⁺, but most or all the stalk cells were fng⁻, were normal or displayed weak defects. Among 15 abnormal examples (Fig. 6A), the length ranged from eight to ten cells, except for one stalk of 17 cells. Among 19 mosaics in which polar cells appeared to be absent from one end of the stalk, the length varied from nine to 22 cells. Moreover, organization of the stalk was more severely disrupted, as the cells can form clusters and their nuclei are not properly oriented (Figs 5A, 6B). These morphological defects are observed in both fng⁺ and fng⁻ cells, indicating that the influence of fng on stalk development is not strictly autonomous. We also examined the fate of these cells by looking at an enhancer trap line, 93F, that directs expression specifically in stalk cells from stage 2 through the end of oogenesis (Ruohola et al., 1991). Cells within mosaic stalks can express 93F (Fig. 6B) Analysis of 93F expression also identifies ‘pseudo-stalk’ cells, which express 93F but do not form a morphologically visible stalk, within the epithelium of stalkless follicles (Fig. 6C).

To determine whether expression of fng could be sufficient to specify polar cells, we used the Flip-out Gal4 technique (Ito et al., 1997) to generate clones of cells ectopically expressing fng. These clones were marked by co-expression of GFP, and polar cell fate was assessed by examining the expression of Fas III or A101; both gave similar results. These markers can be expressed in more than two cells at the end of a follicle prior to stage four, but in stage four and older follicles are only expressed in the two polar cells. In over 100 follicles examined, control (GFP-expressing) FLP-out clones we never associated with extra polar cells (not shown). Strikingly, 12% (19/157) of stage 4 and later follicles with Flip-out fng⁺ clones contained three polar cells (Fig. 7A) and 0.6% (1/157) contained four. These polar cells are always adjacent to each other, and are at the termini of the follicle. This suggests that the ability to respond to the presence of fng overexpression is restricted to the polar-stalk cell lineage. Consistent with this, in the remaining 138 follicles, main-body follicle cells over-expressed fng, but they did not adopt a polar cell fate regardless of whether fng overexpression was (Fig. 7B) or was not (Fig. 7C) present in polar cells. These results also indicate that overexpression of fng in polar cells does not exert a non-autonomous induction of polar fate in main-body follicle cells. Moreover, the ability of fng to induce polar cells is restricted even within the polar-stalk lineage; as apart from one exception, only a single additional cell was transformed.

Intriguingly, extra polar cells only form when the polar-stalk lineage is mosaic for ectopic fng expression. If all of the polar and stalk cells overexpress fng, then only the normal complement of two polar cells form at each end of the stalk (>100 follicles scored). Moreover, in all cases where an extra polar cell forms, one of the polar cells does not overexpress fng, and the remaining two or three do.

Surprisingly, aside from the induction of an extra polar cell, follicles and stalks in which fng is mis-expressed are usually completely normal. We have also ectopically expressed fng under the control of a variety of different Gal4 drivers that are expressed in specific patterns (see Materials and Methods) without observing any consistent defects in follicle development. Thus, despite the apparently complex regulation of fng, it appears that the spatial and temporal patterning of fng expression is not essential for normal oogenesis.

Earlier studies of Notch function during oogenesis have ascribed to it an influence on polar cell fate that is exactly opposite to what we have observed for fng. They have suggested that Notch functions to limit the number of cells that can adopt the polar cell fate, analogous to its role in limiting the number of proneural cells during neurogenesis (Ruohola et al., 1991). This inference was based on the analysis of ovaries mutant for a hypomorphic allele of Notch, N^ts1. To define the requirements for Notch more clearly, we generated somatic clones for a null allele of Notch, N^55e11, and analyzed their influence on the expression of three different polar cell markers: Fas III, A101, and E Cadherin (Niewiadomska et al., 1999; Ruohola et al., 1991). We were unable to recover follicles containing Notch mutant cells among permanent clones, presumably because Notch has essential functions in the germarium. Consequently, we induced transient clones, and analyzed the follicles two days later.

In stage 8 or older follicles, all Notch⁻ cells, regardless of their location, autonomously express Fas III (Fig. 8A). Based on the normal rate of development (King, 1970; Spradling, 1993), the progenitors of these cells were in stage 1 or older follicles at the time of clone induction. As polar cells stop dividing around region IIb of the germarium (Tworoger et al., 1999), we infer that the progenitors were main-body follicle cells and not polar cells. The expression of Fas III in Notch⁻ cells does not correspond to a transformation to polar cell fate, as Fas III expression is not detected until stage 8, the cells continue to proliferate (Fig. 8A), and they do not express A101 (Fig. 8B).

In younger follicles (stages 4-5), Notch mutant clones at either end of the follicle result in loss of polar cells (Fig. 8C and data not shown). The progenitors of these cells were in regions IIb-III of the germarium at the time of clone induction. Thus, Notch is required to define polar cell fate during follicle formation. We were unable to identify any mosaic follicles in which only polar cells or only main-body follicle cells were Notch⁻, presumably because these types of mosaics are much rarer when transient clones are generated. However, we found two mosaics with very few Notch⁺ cells at each extremity. In both cases, the polar cells were Notch⁺ and the follicles looked normal (Fig. 8D). This suggests that, as for fringe, Notch activity is required autonomously in the polar cells, and that its function in these cells is sufficient for follicle formation. Consistent with the hypothesis that Notch is activated specifically within the polar cells, a transcriptional target of Notch signaling in many different tissues, E(spl) mβ-lacZ (Cooper et al., 2000), is expressed specifically within the polar cells from stages 2 to 5 of oogenesis (Fig. 7D).

To determine whether activation of Notch could be sufficient to induce polar cell fate, we induced Flip-out clones that expressed a constituitively activated form of Notch. Eight out of 29 (27%) of follicles expressing activated-Notch had one or two extra-polar cells (Fig. 7E). As for fng, these cells are found at the termini of the follicle, adjacent to the normal polar cells, and are observed only when polar cells are mosaic for Notch activation. Main-body follicle cells that overexpress activated Notch do not adopt a polar cell fate (Fig. 7E). Thus, ectopic activation of Notch can lead to extra polar cells, but only within a small group of competent cells.

Earlier studies have suggested that stalkless follicles result from reduction of Notch function, whereas long, abnormal stalks result from ectopic activation of Notch (Bender et al., 1993; Larkin et al., 1996). However, we have observed that long, abnormal stalks can form in fng mutants. As our analysis implies that this fng phenotype is influenced by polar cell specification, and Notch is required to specify polar cells, we analyzed stalk formation in Notch mutant clones. Indeed, long (on average 15 cells), morphologically abnormal stalks occur when one of the polar-stalk precursors is mutant for Notch (Fig. 6D).

fringe is expressed in a complex and dynamic pattern throughout oogenesis, and we have investigated its functions by analyzing hypomorphic alleles, and by creating clones mutant for a null allele. Although a suggestion that fng functions during early oogenesis has been reported recently in a study that focussed on expression of an antisense construct (Zhao et al., 2000), our studies provide the first clear determination of the requirements for fng during early oogenesis. Our results further provide new insights into the earliest steps in follicle cell patterning, the requirements for different follicle cell types in separating cysts into follicles, and the role and regulation of the Notch signaling pathway during early oogenesis.

Cell-lineage studies have revealed that somatic cells in the germarium become subdivided into two groups: the precursors of the polar and stalk cells, and the precursors of the main-body follicle cells (Tworoger et al., 1999). The molecular basis for this subdivision is unknown, and it has not been possible to distinguish the two groups until after follicles exit the germarium. However, our studies reveal that fng is expressed in a group of cells along the border between regions IIb and III of the germarium. As this corresponds to a group of cells that separate cysts, and fng is expressed exclusively in the polar and stalk cells from the earliest time at which they are morphologically recognizable (between stage 1 and 2 follicles), we suggest that this stripe of fng expression in the germarium includes the progeny of polar-stalk precursors. Although this would make fng the earliest known gene whose expression is largely restricted to polar-stalk cells, the observation that fng⁻ cells can be recovered in stalks at later stages of development indicates that fng is not actually required to specify this lineage.

However, fng is required autonomously to specify polar cell fate. The observations that fng expression becomes restricted to the polar cells, and that overexpression of fng can generate an extra polar cell, further imply that fng is a key player in polar cell specification. Nonetheless, other factors must also contribute to polar cell specification, as fng is expressed in a larger group of precursors within the germarium, and as overexpression of fng in most cases can generate only a single extra polar cell.

In genetic mosaics, the requirement for fng in follicle formation maps to the polar cells. Our analysis of fng mutant clones thus identifies key requirements for the polar cells in enclosing and separating germline cysts. However, the phenotypes observed depend not just on the presence of polar cells in a single follicle, but also on the presence of polar cells in the adjacent follicle.

The main-body follicle cells of each egg chamber are separated from the next by a polar-stalk unit consisting of a pair of polar cells, a stalk, and another pair of polar cells (Fig. 1). As more than one polar-stalk precursor participates in the formation of each unit (Tworoger et al., 1999), a polar-stalk unit could be mosaic for fng function in our experiments. When only one pair of polar cells forms, adjacent follicles are still separated. However, one of two morphological defects occurs. In some cases, the egg chamber without polar cells appears open, with a gap in the epithelium. In other cases, stalkless follicles form, separated by a bilayered epithelium. Notably, the observation that normal egg chamber formation actually depends on the specification of two pairs of polar cells per polar-stalk unit implies that polar and stalk cells acquire distinct fates prior to follicle formation.

More severe phenotypes are observed when all of the cells in a polar-stalk unit are fng⁻, and no polar cells can form. In most cases, this results in the formation of a compound follicle. The compound follicle ends whenever polar cells form. Moreover, normal follicles appear to form as long as fng⁺ polar cells form, even if all other cells are fng⁻. These observations imply that the polar cells have the capacity to recognize distinct cysts, to migrate in between them, and to separate them. Although in some cases adjacent cysts can be separated even without any polar cells, this separation is composed of only a monolayer of epithelial cells. Thus, the occasional ability of the main-body follicular cells to migrate between cysts is not sufficient to generate independent follicles.

Several other genes, including brainiac, egghead, toucan, daughterless, hedgehog and cut, have been identified as playing important roles in the packaging of cysts into separate follicles (Bender et al., 1993; Cummings and Cronmiller, 1994; Forbes et al., 1996; Goode et al., 1996; Goode et al., 1992; Grammont et al., 1997; Jackson and Blochlinger, 1997; Xu et al., 1992). hedgehog is required for the normal proliferation of somatic cells in the germarium (Forbes et al., 1996; Zhang and Kalderon, 2000), so it is possible that hedgehog mutation leads to the formation of compound follicles because of a decrease in the number of polar-stalk precursors. Three of these genes, brainiac, egghead and toucan, are required in the germline (Goode et al., 1996; Goode et al., 1992; Grammont et al., 1997). Determination of their roles may provide new insights into the interactions between germline and somatic cells required to organize a follicle.

fng is also required for stalk formation, and this requirement maps in genetic mosaics to the polar-stalk lineage. However, the requirement is complex in two respects. First, the influence of fng on stalk formation is non-autonomous, and appears to map largely to the polar cells. In the complete absence of polar cells, no stalk forms, while stalks that are composed almost entirely of fng mutant cells can form as long as wild-type polar cells exist at each end. Moreover, stalks exhibit more severe defects whenever polar cells at one end are missing. Even in cases where abnormally long stalks have fng⁺ polar cells at each end, it is conceivable that the long stalk phenotype relates to a requirement for fng in the polar cells. That is, we hypothesize that fng⁻ cells that would ordinarily have been fated to be polar cells may instead be incorporated into the stalk, and that polar-stalk precursors may continue to join the stalk until fng⁺ polar cells arrive.

Second, in some cases fng mutant clones result in stalkless follicles, but in other cases they result in long stalks. This observation, together with our analysis of Notch mutant clones, indicates that there is no direct correlation between stalk length and Notch activity. Instead, we suggest that the number, timing and/or location of polar cells within the polar-stalk lineage at the time of stalk formation is crucial to stalk organization.

The phenotypes observed in fng mutant clones are also observed in Notch mutant clones. Moreover, as for fng, the requirements for Notch appear to map to the polar cells. While our results confirm that Notch plays important roles in separating follicles and organizing stalks, they overturn previous conclusions about the nature of this requirement. Most notably, it had been suggested that loss of Notch function leads to extra polar cells (Ruohola et al., 1991). Although after stage 8 Fas III becomes expressed in Notch mutant cells, this occurs regardless of their location and in the absence of other aspects of polar cell fate. Instead, our results imply that Notch is required for polar cell specification, and that Notch is required only in these cells for the separation of germline cysts into distinct follicles. Furthermore, the specific expression of E(spl)mβ-lacZ in the polar cells confirms that Notch is activated within these cells, and the induction of extra polar cells by activated Notch demonstrates that the activation of Notch can be sufficient to specify polar cells within a competent subpopulation of somatic cells.

In the Drosophila wing and eye, as well as in cultured mammalian cells, fng has been shown to potentiate the activation of Notch by Delta and to inhibit the activation of Notch by Serrate (Fleming et al., 1997; Hicks et al., 2000; Panin et al., 1997; Papayannopoulos et al., 1998). Notch, Serrate and Delta all appear to be expressed ubiquitously in follicle cells throughout the germarium. In many circumstances, expression of Notch ligands can exert an autonomous inhibition of Notch activation (e.g. de Celis and Bray, 1997; Micchelli et al., 1997). It has been hypothesized that fng may potentiate Delta signaling by allowing Notch activation within ligand-expressing cells (Irvine and Vogt, 1997; Panin and Irvine, 1998). Similarly, we propose that in the germarium, fng expression overrides autonomous inhibition to allow Notch activation by Delta within ligand-expressing cells, thereby positioning Notch activation to a subset of cells (the polar cells) within a broad domain of Notch and ligand expression. This is consistent with the observation that Notch and fng are both positively required within polar cells for them to adopt their fate, and that fng expression soon becomes restricted to these cells.

Although fng is expressed in a discrete pattern that correlates with its genetic requirements, ectopic expression of fng during oogenesis has surprisingly little effect. This contrasts with the very dramatic effects on Notch activation associated with fng mis-expression during imaginal development (reviewed in Irvine, 1999). The limited consequences of fng mis-expression are also surprising in light of the observation that mirror (mirr) mutant ovaries have compound follicles (Jordan et al., 2000). It has been hypothesized that the mirr phenotype derives from de-repression of fng transcription. mirr expression is complementary to fng, and ectopic expression of mirr represses fng expression and generates phenotypes similar to those observed in fng mutants (Jordan et al., 2000). We suggest that the simplest resolution to this discrepancy would be the existence of co-factors that are required for fng function and are co-regulated by mirr during oogenesis. As FNG is involved in the synthesis of an O-linked tetrasaccharide (Moloney et al., 2000), the enzymes that are responsible for catalyzing other steps in its synthesis are candidate FNG co-factors, and it will be interesting to determine whether they too are expressed in discrete patterns.

We thank N. Hawkins for an initial fng in situ hybridization, S. Bray, U. Gaul, Y. Hiromi, Y. Jan, J. Jiang, T. Schupbach, R. Steward, G. Struhl, J. Treisman, The DSHB and The Bloomington Stock Center for antibodies and Drosophila stocks, and J.-L. Couderc and R. Steward for comments on the manuscript. This work was supported by a fellowship from the Institut National de la Santé et de la Recherche Medicale to M. G. and by the Howard Hughes Medical Institute and NIH grant R01-GM54594.