Notch-Delta signaling induces a transition from mitotic cell cycle to endocycle in Drosophila follicle cells

The Notch signaling pathway is evolutionarily conserved and is widely used in cell fate determination and cell differentiation (Artavanis-Tsakonas et al., 1999; Bray, 1998; Greenwald, 1998). In humans, defects in this pathway are implicated in multiple diseases, including leukemia and heart disease (Joutel and Tournier-Lasserve, 1998; Selkoe, 2000). The basic components of the pathway are the Notch receptor, the two Notch ligands Delta and Serrate (Delta and Jagged in vertebrates), the transcription factor Suppressor of Hairless (Su(H); CBF1/RJBk in mammals) and the bHLH transcription factors encoded by the Enhancer of Split complex genes E(spl). In addition to differentiation, the Notch pathway also regulates some aspects of growth control (Kimble and Simpson, 1997; Artavanis-Tsakonas et al., 1999). In the developing Drosophila wing, for example, Notch induces cell cycle arrest indirectly through modulating the function of Wingless, a secreted signaling molecule, in the nonproliferative boundary zone (Johnson and Edgar, 1998). Other evidence suggests that the cell cycle function of Notch in the wing is more direct (Baonza and Garcia-Bellido, 2000). No direct molecular link, however, has been made between Notch and cell cycle regulation.

During Drosophila oogenesis, the Notch pathway plays multiple roles in specifying follicle cell groups, the terminal follicle cells, dorsal anterior follicle cells and polar cells (Ruohola et al., 1991; Xu et al., 1992; Bender et al., 1992; Larkin et al., 1996; Keller Larkin et al., 1999; Gonzalez-Reyes and St Johnston, 1998; Jordan et al., 2000). Our analyses of Notch function in the ovary have revealed a connection between Notch signaling and regulation of a transition from a mitotic cell cycle to an endoreplication cycle (this study). The endoreplication cycle, or endocycle, is a variation of the normal mitotic cell cycle, in which cells increase their genomic DNA content without dividing. The endocycle is widespread among multicellular organisms and uses much of the same G1/S regulatory machinery as mitotic cycles. The mechanism responsible for skipping over M phase during the endocycle is the subject of much current investigation (Edgar and Orr-Weaver, 2001). The follicle cells in the Drosophila ovary offer an excellent system in which to study the mitotic-to-endocycle transition. The Drosophila germline is surrounded by an epithelial follicle cell layer. The follicle cells undergo multiple rounds of mitotic divisions during the early stages of oogenesis. At stage 6, these cells switch from the normal mitotic cycle and undergo three rounds of endocycles. After the endocycles, most DNA is no longer replicated except the chorion genes, which continue to be amplified in a specific, tightly regulated fashion (Royzman and Orr-Weaver, 1998).

String is a Drosophila Cdc25-type phosphatase that triggers mitosis by dephosphorylating and thereby activating the Cdk1/Cyclin B kinase (Edgar et al., 1994). Differential expression of string regulates mitosis during most stages of Drosophila development: loss-of-function mutations in the gene cause G2 arrest in both embryos and imaginal disc cells (Edgar and O’Farrell, 1989; Neufeld et al., 1998), whereas ectopic string expression in both embryo and discs drives G2 cells into mitosis (Edgar and O’Farrell, 1990; Neufeld et al., 1998; Johnston and Edgar, 1998). These results suggest that String is rate limiting for mitotic initiation and therefore regulation of string expression is a key in controlling mitosis. The abundance of String protein is mainly controlled at the transcriptional level. Moreover, mutations in many patterning genes alter string expression in specific ways, indicating that string transcription is controlled by the same network of factors that controls cell fates (Arora and Nüsslein-Volhard, 1992; Edgar et al., 1994; Johnston and Edgar, 1998). The large upstream control region of the string gene has been dissected and shown to consist of many modular elements with separable activities (Lehman et al., 1999). These elements pattern mitoses in the embryo and in neural lineages at many stages of development. In some cases Drosophila String is functionally redundant with another Cdc25 homolog, Twine (Edgar and Datar, 1999). In addition to transcriptional regulation, String protein turnover and Twine translational levels can affect cell cycle transition (Mata et al., 2000; Seher and Leptin, 2000; Grosshans and Wieschaus, 2000; Maines and Wasserman, 1999).

We show that the transition from mitotic cycle to endocycle in the Drosophila follicle cell epithelium is regulated by the Notch pathway. Loss of Notch or its downstream transcription factor Su(H) in follicle cells, or its ligand Delta in the underlying germline, disrupts the normal transition of the follicle cells from mitotic cycle to endocycle, leading to overproliferation of these cells. A target of Notch and Su(H) is likely to be the G2/M cell cycle regulator string, as string is normally repressed in the follicle cells just before the endocycle transition but is expressed when Notch is inactivated. Analyses of the activity of string enhancer elements in follicle cells reveal the presence of an element that promotes expression of string until just before the onset of polyploidy in the wild-type follicle cells, but much beyond that in egg chambers that are lacking germline Delta. These results suggest that the string gene may be at least one of the targets of the endocycle-promoting activity of the Notch pathway.

The following fly-stocks were used: N^55e11FRT101 (N^55e1 is an amorphic allele of Notch, gift from the laboratory of Y. N. Jan), FRT82B Dl^rev10 e (Dl^rev10 is an amorphic allele of Delta that is produced by excision of the promoter region, transcription start site and first exon) (Heitzler and Simpson, 1991; Zeng et al., 1998), FRT82B Df(3R)gro^b32.2P[gro⁺]E8 (this combination removes all seven bHLH-encoding genes of E(spl)-C but provides a wild-type copy of groucho) (Heitzler et al., 1996), fng¹³ FRT 80B (fng¹³ is a strong loss-of-function allele of fringe) (Irvine and Wieschaus, 1994), lgl⁴ FRT40A (lgl⁴ is a null allele) (Mechler et al., 1985; Woods and Bryant, 1991), FRT82B stg^7B, STG-15.3; FRT82B stg^7B (stg^7B is a strong loss-of-function allele of string) (Lehmann et al., 1999), Su(H)^SF8 FRT40A (Su(H)^SF8 is a strong loss-of-function (but not a null) allele of suppressor of hairless, a gift from S. Blair), hs FLP; +/+; Ubi-GFP FRT 80B, hs FLP; +/+; FRT 82B Ubi-GFP and yw Ubi-GFP FRT 101; MKRS P{ry=hs FLP}86E / TM6B Tb. String 15.3 kb and 31.6 kb transgenes were tested for function in follicle cells using the following flies: hs-FLP122/+; P[w⁺STG-15.3]/+; FRT82B stg^7B/FRT82B Ubi-GFP and hs-FLP122/+; P[w⁺STG-31.6]/+; FRT82B stg^7B /FRT82B Ubi-GFP. The string-lacZ-fusion (Table 2 and Fig. 6A) and string-rescue transgene constructs were a kind gift from the laboratory of B. Edgar (Lehmann et al., 1999). Multiple insertion lines were analyzed when available.

Drosophila melanogaster stocks were raised on standard cornmeal-yeast-agar medium at 25°C. To obtain follicle cell clones, 1- to 5-day-old flies were heat-shocked as adults for 50-60 minutes at 37°C and put in freshly yeasted vials with new males for 2 or 3 days. To obtain germline clones, flies were heat shocked as second and third instar larvae for 2 hours on 2 consecutive days. Once they emerged as adults, females were placed in vials with fresh yeast paste for 1-5 days with young males in preparation for dissection.

Ovaries were dissected in 1× phosphate-buffered saline (PBS) and fixed shaking for 20 minutes in 200 μl of 4% paraformaldehyde in PMES (0.1 M Pipes pH 6.9, 2 mM MgSO₄, 1 mM EGTA) with 0.5% NP40 and 600 μl of Heptane. They were rinsed for 5 minutes with NP40 wash (50 mM Tris 7.4, 150 mM NaCl, 0.5% NP40, 1 mg/ml bovine serum albumin (BSA), 0.2% sodium azide) and blocked in 20% goat serum (in NP40 wash) for 2 hours at room temperature. The tissue was incubated with primary antibodies overnight at 4°C. The next day they were rinsed with NP40 wash four times, for 20 minutes each. They were then incubated with a secondary antibody for 2 hour at room temperature and stained with DAPI (1 μg/ml in NP40 wash), washed three times with NP40 wash (20 minutes each) and dissected onto slides in 70% glycerol, 2%NPG, 1×PBS. Confocal microscopy, X-gal staining and in situ hybridization was performed as described previously (Keller Larkin et al., 1999; Tworoger et al., 1999). stg cDNA (a gift from B. Edgar) was labeled with digoxigenin. A two-photon laser-scanning microscope (Leica TCS SP/MP) was used in this study.

The following antibodies were used: mouse anti-Armadillo (1:20, Developmental Studies Hybridoma Bank), mouse anti-FasciclinIII (1:20, Developmental Studies Hybridoma Bank), mouse anti-Cyclin B (1:20, Developmental Studies Hybridoma Bank), mouse anti-Notch intracellular domain (1:20, Developmental Studies Hybridoma Bank), mouse or rabbit anti-β-gal (1:5000, Sigma), mouse anti-Broad-Complex Z1 (1:50, a gift from G. Guild), mouse anti-Delta (1:100) (Bender et al., 1993), rabbit anti-PH3 (1:200, Upstate Biotechnology), Alexa 488, 568 or 633 goat anti-mouse (1:500) and Alexa 568 or 633 goat anti-rabbit (1:500) (Molecular Probes).

Follicle cells are descendents of somatic stem cells located in region 2 of the germarium (Fig. 1A) (Margolis and Spradling, 1995). In region 3 of the germarium, the follicle cells encapsulate the germline cyst, which contains 15 nurse cells and an oocyte to form a stage 1 egg chamber that then exits from the germarium (stage 1). The follicle cells carry out a mitotic division program from stage 2 to 6 (∼30 hours), giving rise to approximately 1000 follicle cells. At stage 6, follicle cells cease mitosis and undergo endocycles (also called endoreplication or endoreduplication cycle) in which the mitotic stage of the cell cycle is bypassed (Fig. 1A). This cell cycle transition can be visualized with anti Phospho-Histone3 (PH3) (Bradbury, 1992; Handzel et al., 1997) and anti-Cyclin B (CycB, a G2 cyclin) antibodies (Fig. 1B). Until stage 6, phosphorylation of Histone3 (indicative of mitosis) and expression of CycB is observed in the follicle cells, while no obvious staining is detected after stage 6 (Fig. 1B). The expression of CycB is observed in groups of 5-10 follicle cells (Fig. 1B), suggesting that mitosis in the follicle cell layer is controlled in small domains, reminiscent of mitotic domains observed in early embryos. However, it is not clear yet how synchronized these domains are, what controls their size and whether the positions of the domains are sporadic or developmentally controlled. Data obtained by flow sorting of follicle cell nuclei show that follicle cells undergo three rounds of endocycles, giving rise to 16 copies of genomic DNA (Lilly and Spradling, 1996). BrdU-labeling of stage 7-9 egg chambers demonstrates that the timing of endocycles is not synchronized among follicle cells within an egg chamber (Calvi et al., 1998). By the end of stage 10A, all follicle cells complete their overall endoreplication program but continue chorion gene amplification (Spradling, 1993; Royzman and Orr-Weaver, 1998; Bosco et al., 2001).

The abrupt and relatively synchronous transition from mitotic cycle to endocycle makes the follicle cell epithelium an excellent model system to study this cell cycle control process. In other cell types, the mitotic-to-endoreplication switch involves transcriptional repression of G2/M regulators such as CycA, CycB or string, which encodes a CDC25 phosphatase (Edgar and Orr-Weaver, 2001). We therefore analyzed transcription of string in oogenesis using whole-mount in situ hybridization. string expression is detected in a patchy manner in mitotic follicle cells (Fig. 1C). Downregulation of string expression in follicle cells was observed at stage 7 (Fig. 1C, arrow), suggesting that the transition from mitotic cell cycle to endocycle is regulated in part by repression of string at the transcriptional level.

To investigate the possibility that the Notch pathway is involved in the regulation of the mitotic cycle-to-endocycle transition, we first analyzed the expression of the Notch ligand, Delta, in oogenesis. Weak Delta expression is observed in follicle cells and germline from the germarium to stage 5 in oogenesis. At around stage 6, a dramatic upregulation of Delta expression is observed in the germline cells reaching the highest level at stage 7 (Fig. 2A) (Bender et al., 1993; Keller Larkin et al., 1999). This upregulation of Delta in the germline coincides with the transition from mitotic cell cycle to endocycle in the follicle cells (Fig. 2A, note the lack of PH3 in the egg chamber with high germ line Delta at stage 7).

To determine if Delta plays a role in regulating the mitotic-to-endocycle-transition, we undertook a mosaic analysis. We employed mitotic recombination (Xu and Rubin, 1993) to create clones of cells lacking Delta function, then analyzed cell cycle stage by examining PH3 and CycB. The mutant cell clones were marked by lack of green fluorescence protein (GFP) expression, while the wild-type cells were labeled by GFP. As wild-type follicle cells lack PH3 and CycB after stage 6, owing to the transition into the endocycle, we analyzed the expression of these markers in follicle cells adjacent to Delta germline clones after stage 6. This analysis revealed defects in cell cycle regulation in follicle cells surrounding the Delta germline clones. In almost all cases (99%, n=68), follicle cells surrounding the Delta germline clones expressed PH3 and CycB beyond stage 7, suggesting that these cells remained in the mitotic cell cycle program (Fig. 2C,D; Table 1). Although dividing, these follicle cells appeared to maintain their apical-basal polarity, as shown by normal apical localization of Armadillo, a β-catenin homolog in Drosophila (Fig. 2C) (Peifer et al., 1993) (some abnormalities were observed in later stages). In Delta follicle cell clones, no PH3 and CycB staining was detected in follicle cells after stage 6 (Fig. 2B; Table 1 and data not shown), suggesting that germline but not follicle cell contribution of Delta is essential for the follicle cell mitotic cycle transition into the endocycle program. These results differ from earlier studies (Ruohola et al., 1991), which failed to detect a role for Delta in the germline in regulating follicle cell division, because germline clones were made with a weaker allele and/or the extent of proliferation of the follicle cells was not analyzed because of a lack of suitable markers.

As Delta signal from the germline is required for the mitotic-to-endocycle transition in follicle cells, we tested whether Notch is the receptor for the signal in the follicle cells. Notch is predominantly localized to the apical side of the follicle cell epithelium (facing the germline cells; Fig. 3A) (Xu et al., 1992), making it a possible candidate to receive Delta signal from the germline. We generated Notch mutant follicle cell clones to analyze Notch function in the transition by inducing mutant clones in the adult flies and allowing development to proceed at 25°C for 2 days. Many smaller nuclei were observed in Notch follicle cell clones (Fig. 3B) compared with the polyploid wild-type neighbors. Furthermore, PH3 and CycB expression were detected in Notch follicle cell clones in a cell autonomous manner after stage 6 (94%, n=65), suggesting that these cells remain in a mitotic cell cycle (Fig. 3C-E; Table 1). These results show that Notch acts cell autonomously to regulate the transition from mitotic cycle to endocycle in follicle cells.

If the follicle cells in Notch mutant clones proceed through full mitotic cell cycles, more cells should be observed in mutant clones than in twin spots carrying wild-type copies of the Notch gene (sister clone). We tested this prediction by counting the number of cells in mutant and sister clones. More cells were observed in the mutant clones than in the associated twin spots (Fig. 3B,G,H, the ratio varied from 1 to 3.3, mean=1.8), suggesting that the mutant cells have undergone more mitotic cycles than their endocycling sister clones. In addition, as expected, the DNA content was higher in sister clones than in mutant clones, based on DAPI staining, and the wild-type cells were larger on average than the mutant cells (Fig. 3B). Importantly, the ratio between the cell number in mutant and sister clones did not change as a function of the clone size (Fig. 3G,H; G is plotted as a function of increasing sister clone size, H is the mutant:sister ratio of G as a function of increasing sister clone size), suggesting that there are no defects in division in the Notch mutant before the transition to the endocycle. Therefore the cells in Notch mutant clones divide at a normal rate before the endocycle switch.

The glycosyltransferase Fringe is a modifier of the Delta-Notch interaction in imaginal discs (Bruckner et al., 2000; Moloney et al., 2000; Munro and Freeman, 2000; Blair, 2000), and in oogenesis of earlier Notch-dependent processes that allow pinching off to occur in the germarium and define the number of polar cells (Jordan et al., 2000; Zhao et al., 2000; C. A., W.-M. D., K. Jordan and H. R.-B., unpublished). When Fringe is present in Notch-expressing cells, Notch can be activated by low levels of Delta expression. Because Delta expression is upregulated in stage 6 germline cells, we suspected that Fringe would not be required for Delta-Notch signaling in this later case. Indeed, no CycB or PH3 staining was detected after stage 6 of oogenesis in fringe mutant clones (Table 1), suggesting that Fringe is not essential in the Delta-Notch interaction that regulates the transition from mitotic cycle to endocycle in follicle cells.

Does Notch exert its effect on the mitosis-endocycle transition through transcriptional regulation? Suppressor of Hairless (Su(H)) is a transcription factor that interacts with the cleaved-Notch-intracellular domain after Delta binds to the Notch extracellular domain. This interaction leads to transcription of downstream target genes, including Enhancer of Split (Espl) complex genes. We made follicle cell clones of a Su(H) hypomorphic allele and E(spl) null allele, then probed the egg chambers with anti-PH3 and CycB antibodies. As in Notch follicle cell clones, small nuclei, PH3 and CycB were observed in Su(H) clones (Fig. 4A-C; Table 1; 76%, n=46). No extra cell division was detected, however, in follicle cell clones mutant for E(spl) (Fig. 4D; Table 1). These results suggest that Notch and Su(H) regulate the follicle cell mitotic-to-endocycle transition but do not act through the E(spl) genes.

In Drosophila, mutations in three neoplastic genes, discs large (dlg), lethal(2)giant larvae (lgl) and scribble (scrib), cause loss of apical-basal polarity accompanied by hyperproliferation (Jacob et al., 1987; Woods and Bryant, 1991; Bilder and Perrimon, 2000; Bilder et al., 2000). It is not understood how loss of apical-basal polarity results in hyperproliferation in these mutants. One attractive hypothesis, however, is that overproliferation is caused by mislocalization of critical plasma membrane proteins that are involved in cell cycle control. We found that the Notch protein was indeed mislocalized throughout lgl mutant cells (Fig. 5D); this defect could reduce responsiveness to Delta from the germline and thus lead to overproliferation. In addition, many lgl mutant follicle cells existed in multilayered structures where they were less likely to receive the Delta signal because of lack of direct contact with the germline (Fig. 5C, arrow). By contrast, in Notch mutant follicle cells, the cell cycle defect was observed before any obvious polarity defect; normal Armadillo localization was observed in Notch but not in lgl clones (Fig. 5A,B). Additional factors probably also contributed to the overproliferation in lgl mutant clones because significantly more proliferation was observed in lgl than in Notch clones (compare Fig. 5A with B,C).

Notch is known to act on follicle cell differentiation multiple times during oogenesis. To test whether Notch acts primarily on follicle cell differentiation at stage 6, we used a marker that begins to be expressed in all follicle cells at around stage 6, the Broad-Complex (BR-C) (Deng and Bownes, 1997; Tzolovsky et al., 1999). BR-C was expressed at similar levels in Notch mutant follicle cell clones and in wild-type follicle cells (Fig. 5E), suggesting that at least on this level differentiation at stage 6 occurred in a normal manner. However, some markers showed elevated expression in the mutant clones compared with the wild-type cells neighboring the mutant cells. One marker, Fasciclin III, is normally expressed in all cells at early stages of oogenesis and then later is restricted to the polar cells (Ruohola et al., 1991). We observed elevated levels of FasIII (Fas3 – FlyBase)in a cell-autonomous fashion in Notch clones (Fig. 5F). Although elevated expression of FasIII may indicate potential differentiation defects in Notch mutant clones, there is also an alternative explanation. Normally, high FasIII expression occurs up until stage 1 in oogenesis, yet we observed elevated levels of FasIII in two-cell clones that were generated at stage 4-5 in oogenesis. In addition, many other markers that have no general early expression were elevated in the clones. One possible explanation for these expression abnormalities is that the highly mitotically active cells in Notch clones show elevated translational activity compared with the neighboring wild-type cells. More experiments are required to distinguish between these possibilities.

Mitosis in most Drosophila cells is triggered by brief bursts of transcription of string, which encodes a Cdc25-type phosphatase that activates the mitotic kinase, Cdk1 (Cdc2). Transcriptional control is an important mode of regulation for the activity of this G2/M controller. Fig. 1C shows that string is expressed in a patchy fashion when follicle cells are in the mitotic cell cycle but repressed when the follicle cells are in the endocycle. This repression requires Notch activity, as string mRNA is observed in egg chambers with Notch clones beyond the normal stage 6 in oogenesis (Fig. 3F).

To test whether String is essential for follicle cell mitotic cycles, we generated string mutant clones. The mutant clones had far fewer cells than sister clones (zero to two cells in mutant clones, 3-15 in sister clones, n=24, mutant:sister=0.1; Fig. 6F,H) and showed no expression of G2/M stage markers, suggesting that String is required for the follicle cell mitotic cycle. In control experiments, the mutant:sister ratio was 1.0 (n=10).

Previous studies have revealed that many cis-acting elements distributed over >30 kb upstream of string that control string transcription in different cells and tissue types (Lehman et al., 1999). To map the transcriptional control elements that regulate string in follicle cells, we took advantage of gene-fusion constructs in which the string locus was dissected into fragments and fused to lacZ reporter genes containing the basal 0.7 kb string promotor (Lehman et al., 1999). Of 11 different gene fusions tested, two showed specific expression in follicle cells. These constructs define the regions of the string promoter that are required for string expression and therefore for controlling the cell cycle in follicle cells (Table 2). Interestingly, R4.9 (4.9 kb region) exhibited expression in follicle cells of the germarium to around stage 2 in oogenesis, while R6.4 (6.4 kb region) showed expression beginning in stage 3 and abruptly terminating at the mitotic-to-endocycle transition in stage 6 (Fig. 6B-E; Table 2).

If the two enhancer elements (4.9 and 6.4; Fig.6A) are required for controlling two different bursts of cell divisions in follicle cells, then a rescue construct that covers only the first element should only partially rescue the division defects in string clones. That is exactly what we observed: when string clones were produced in the background of the 15.3 kb rescue construct (see Fig. 6A), mutant clones of half the size of sister clones were detected (one to six cells in mutant clones, 2-15 in sister clones, n=21, mutant:sister=0.49). This result suggests that the 4.9 kb element completely contained within the 15.3 kb rescue construct supports only the early expression and rescues the first but not the second burst of cell divisions in the follicle cell layer (Fig. 6G,I). Interestingly, the size of most of the nuclei in the clones was twice that of the nuclei typically observed in the sister clones or in the neighboring cells (Fig. 6G; 77%; wild-type nuclear volume/mutant nuclear volume=2.2), possibly suggesting an extra endoreplication cycle. After the action provided by the first string promoter element, the cells may have stopped dividing and as a default prematurely entered the endocycle program. However, we cannot rule out the possibility that the abnormal size of nuclei is a result of abnormal condensation of the DNA. We also produced string clones in the background of 31.6 kb rescue construct (Fig. 6A) and, as expected, observed mutant clones the size of sister clones and a comparable size of nuclei in each group.

These data suggest that Notch pathway activity will impinge in (6.4 kb region)-enhancer element. To test this, we crossed both the R6.4-lacZ and R4.9-lacZ gene fusions to the line mutant for Delta and analyzed the expression of these elements in egg chambers with germline mutant for Delta. In controls, the expression of R6.4 element was observed in stage 5 but not in stage 9 egg chambers (Fig. 7A-B). However, in Delta germ line clones, expression of R6.4 element was observed in domains of 5-20 cells through the follicle cell layer past stage 6 (Fig. 7C-D, n=36). By contrast, when R4.9-element was analyzed in Delta germ line clones, no expression beyond the normal early expression was observed (data not shown). These data suggest that Notch pathway activity does control String at the 6.4 kb element but not at the 4.9 kb element.

We have shown that the mitotic-to-endocycle transition in Drosophila follicle cells is controlled by the Notch pathway. Notch at the apical surface of the follicle cells is activated by Delta from the germline, which is expressed at high levels just before the switch. Disruption of the Notch pathway results in follicle cell over-proliferation but does not primarily affect their apical-basal polarity. This process does not require Fringe, probably because Delta levels are high enough to activate the Notch pathway in its absence. Notch activation leads through Su(H) to repression of the transcription of string, a regulator of the G2/M transition. Analysis of the activity of string enhancer elements in wild-type and Notch mutant follicle cells reveals the presence of an element that promotes expression of string until just before the onset of polyploidy in wild-type follicle cells but well beyond this stage in follicle cells defective for Notch pathway, suggesting that it may be a target of the endocycle promoting activity of the Notch pathway. A second element that is insensitive to Notch regulation promotes string expression earlier in follicle cell development, explaining why Notch only represses string at the mitotic cycle-endocycle transition and not in the germarium, where the Notch pathway is also active.

Endocycles are widespread in plants and animals including humans and Drosophila. In human placenta, subgroups of trophoblast cells enter the endocycle and terminally differentiate to become trophoblast giant cells. It is known that over-proliferation of the trophoblast cells causes choriocarcinoma, placental cancer, but the signal that triggers the endocycle transition is not known. Another example of a mammalian cell type that becomes polyploid by endocycling is the megakaryocyte, a blood cell that is specialized to produce platelets (Zimmet and Ravid, 2000). These cells undergo endomitosis, and the transition from mitotic to endomitotic cycle is influenced by the secreted signal thrombopoietin. Thrombopoietin upregulates Cyclin D3 protein and Cyclin D3 overexpression has been found to increase megakaryocyte ploidy in transgenic mice (Zimmet et al., 1997). In addition, recent data has shown that constitutively active form of Notch can induce CycD expression and that CSL (stands for CBF1, Su(H), and Lag-1) binding site is identified in the CycD promoter (Ronchini and Capobianco, 2001).

In Drosophila, cells adopt the endocycle program during both embryonic and adult stages. Embryonic endocycling cells are found in the gut, fat body, malpighian tubules and salivary glands (Smith and Orr-Weaver, 1991), and in adult tissues, endoreplicating cells are found in the ovary (nurse cells and follicle cells) and in the wing (sensory neurons). The transition from mitotic to endocycle during embryogenesis coincides with transcriptional repression of the mitotic regulators Cdk1, Cyclin A, Cyclin B, Cyclin B3 and Cdc25/string (Sauer et al., 1995). Control of this transcriptional repression is not yet understood. In addition to transcriptional level control, embryonic cells control their mitotic machinery at the protein level to ensure the transition to endocycles (Sigrist and Lehner, 1997).

It is plausible that Notch also regulates other components of the cell cycle in addition to String, because the transition to the endocycle requires not only the repression of a mitosis promoter (string) but the upregulation during G2 of an S phase promoter, such as CycE and/or CycD (for comparison, during early embryogenesis cells that lack String pause in G2 rather than undergoing endocycles). However, when string was restored to clones under the control of the 15.3 kb rescue construct, which contains the control element active in the germarium and stage 1 but not the control element active between stage 3 and stage 6, some cell divisions occurred before premature stopping of the mitotic cycle, and interestingly, the nuclei were 2.2 fold bigger in the mutant clones than in the sister clones or in the neighboring wild type cells (Fig. 6G) suggesting that the mutant cells that stopped dividing too early also entered the endocycle too early. Thus, in contrast to embryonic cells, cessation of string expression may lead the follicle cells to endocycle rather than simply arrest, and thus Notch might act solely on string to stop mitosis and the subsequent transition to the endocycle could be due to other factors constitutively present in follicle cells. However, we cannot rule out the possibility that Notch also acts on other cell cycle components, and based on the example of megakaryocytes, in which thrombopoietin promotes the mitotic-endocycle switch and also elevates the levels of CyclinD, it will be interesting to examine the role of Cyclin D in the follicle cell transition.

We have shown that in Drosophila follicle cells string mRNA expression coincides with the mitotic cycle of follicle cells. Furthermore, we have shown that two separate controlling elements are found in the promoter region of string for early follicle cell expression, one of which is shut off just at the mitotic cycle-endocycle transition and is likely to be the target of regulation by the Notch pathway. Previous work suggests that string transcription is altered in embryos mutant for axis, gap, pair rule, segment polarity, homeotic, neurogenic and proneural genes (Arora and Nüsslein-Volhard, 1992; Edgar et al., 1994), suggesting that the upstream regulatory elements of string is a sophisticated integrator of patterning signals. It will be interesting to determine how Su(H) functions to repress string expression and leads to the transition to the endocycle.

Signaling between the somatic follicle cells and the underlying germline plays a crucial role in coordinating the complex process of oogenesis. Previous work has shown important roles for epidermal growth factor receptor signaling from the oocyte to the follicle cells at two stages of oogenesis, and revealed that the follicle cells signal back to the oocyte to help define the oocyte anteroposterior axis later in oogenesis (Schupbach, 1987; Gonzalez-Reyes et al., 1995; Roth et al., 1995; Anderson, 1995; Ruohola et al., 1991; Deng and Ruohola-Baker, 2000; Ray and Schüpbach, 1996; Rongo and Lehmann, 1996). The new oocyte-germline signaling event from Delta in the germline to Notch in the surrounding follicle cells defined in this paper adds a new twist to our understanding of germline-follicle cell signaling during oogenesis and plays an important role in coordinating the division of follicle cells with crucial steps in the maturation of the oocyte. The next challenge in this area is to understand what regulates the key regulatory step in the process, the elevation of Delta protein level in the germline.

We thank Dr B. Edgar for helpful discussions throughout this study, Drs C. Berg, K. Jordan, B. Edgar and D. Baker and anonymous reviewers for extremely helpful comments on the manuscript, and Drs M. Bownes, Y. N. Jan, S. Younger-Shepherd, N. Baker, M. Freeman, B. Edgar, S. Blair, G. Guild, C. Doe and M. Muskavitch and Developmental Studies Hybridoma Bank for fly lines and/or antibodies. This work was supported by grants from the National Institutes of Health (R01-GM62748-06), American Heart Association, American Cancer Society and Pew Memorial Trust, Pew Scholarship in the Biomedical Sciences to H. R.-B.