The Drosophila cytokine receptor Domeless controls border cell migration and epithelial polarization during oogenesis

The temporal and regional control of epithelial morphogenesis underlies the assembly of many organs in metazoans. In Drosophila, the egg chamber provides a useful model to study genes and signaling networks controlling numerous stages of epithelia formation, from stem cell development to the final organ shape (Dobens and Raftery,2000). The egg chamber is a composite unit that develops in a succession of discrete developmental stages(Spradling, 1993). Each egg chamber is made of a central germline cyst, from which the oocyte is selected,surrounded by a monolayer epithelium, that ultimately makes the external envelope of the egg. Several signaling pathways are essential to control egg chamber development, including Notch, EGFR, Hedgehog, TGFβ, steroid hormones and JNK cascades (Dequier et al.,2001; Dobens et al.,2001; Dobens and Raftery,2000; Nilson and Schupbach,1999; Riechmann and Ephrussi,2001; Suzanne et al.,2001; van Eeden and St Johnston, 1999). Among the diverse cellular processes taking place during oogenesis, the delamination and migration of border cells (BCs), a group of anterior follicle cells, represent a powerful model to study cell invasion in a normal tissue (Montell,1999; Montell,2001). During stage 9 of oogenesis, BCs detach from the outer epithelium to invade the germline cyst compartment(Fig. 1A). The BC cluster contains two centrally located polar cells surrounded by approximately six outer border cells and undergoes a nearly 6-hour long posteriorward migration to reach the anterior part of the growing oocyte. Together with centripetal cells, they assemble the micropyle, a specialized structure required for sperm entry. Recent work has shown that the EGFR and its ligand Gurken, together with a novel PDGF/VEGF transduction pathway are involved in the directional control of BC migration to the anterior-dorsal corner of the oocyte(Duchek and Rorth, 2001;Duchek et al., 2001). In addition, a novel co-activator of the ecdysone receptor, Taiman, has been shown to be essential for BC migration, thus providing an interesting parallel with the known role of steroids in mammary gland metastasis(Bai et al., 2000). In order to identify new genes and signaling pathways that are involved in egg chamber development and BC migration, we performed a tissue-targeted mosaic screen using a collection of X-linked lethal mutations (unpublished results; see Materials and Methods). Here, we report the identification ofdomeless, which encodes a putative receptor of the JAK/STAT pathway.

The JAK/STAT signaling pathway is involved in many developmental and physiological processes in mammals, particularly in blood cell differentiation or proliferation and in mammary gland development(Darnell et al., 1994;Ihle, 1995;Leonard and O'Shea, 1998;O'Shea et al., 2002). Moreover, constitutive activation of STAT3 and STAT5, for example, is responsible for numerous human epithelial cancers and leukemia(Bowman et al., 2000;Levy and Gilliland, 2000). Signaling through the JAK/STAT cascade involves membrane receptors the intracellular domain of which holds a tyrosine kinase of the JAK family. Ligand binding promotes receptor dimerization and its subsequent transphosphorylation by the associated JAK molecules. Phosphorylation of the receptor on specific tyrosine residues creates docking sites for the STAT transcription factors, which in turn are phosphorylated and activated by receptor-bound JAK. The final step in the transduction cascade is the release of activated STAT, which then translocates as a dimer into the nucleus to regulate target gene transcription.

In Drosophila, homologs of the mammalian JAK/STAT pathway components have been isolated. hopscotch (hop) encodes aDrosophila JAK homolog (Binari and Perrimon, 1994) and marelle (mrl) codes for a STAT factor (also known as Stat92E)(Hou et al., 1996;Yan et al., 1996). The only known ligand for the Drosophila JAK/STAT pathway is encoded by theunpaired (upd/os) gene(Harrison et al., 1998). Only recently, Domeless (Dome; also known as Mom)(Chen et al., 2002) has been identified as the first putative receptor for the JAK/STAT pathway inDrosophila embryos (Brown et al.,2001). The reduced number of JAK/STAT pathway genes inDrosophila contrasts with the multiplicity of JAK/STAT homologs found in mammals, making Drosophila a suitable system to study JAK/STAT signaling (Luo and Dearolf,2001; Mathey-Prevot and Perrimon, 1998; Zeidler et al., 2000). The Drosophila JAK/STAT pathway was originally identified for its role in embryonic segmentation. Further work has shown that this pathway also participates in blood cell determination and proliferation, eye and wing development(Luo and Dearolf, 2001;Zeidler et al., 2000;Zeidler et al., 1999), sex determination (Jinks et al.,2000; Sefton et al.,2000; Zeidler and Perrimon,2000) and stem cell differentiation(Kiger et al., 2001;Tulina and Matunis, 2001). More recently it has been shown to play a role in oogenesis for stalk and BC differentiation (Baksa et al.,2002; Beccari et al.,2002; McGregor et al.,2002; Silver and Montell,2001). Thus, as in mammals, the Drosophila JAK/STAT pathway has a pleiotropic role and a common function in cell differentiation and blood cell development.

Here we show that dome is involved in BC determination and migration. In addition, dome is required for follicle cell polarization through the control of the apical determinant Crumbs. The Dome protein is associated with apicolateral membranes in follicle cells, and becomes internalized upon exposure to Upd, its putative ligand. Our loss- and gain-of-function analyses show that dome is autoregulated in follicle cells, and that both the intracellular and extracellular domains are required for Upd transduction and BC migration. Our data show that Dome is an essential receptor molecule for Upd and JAK/STAT signaling during oogenesis.

A description of genetic markers and chromosome balancers can be found in Lindsley and Zimm (Lindsley and Zimm,1992) and FlyBase(http://flybase.bio.indiana.edu).dome has been identified in an ongoing P element screen using a targeted mosaic approach in ovaries (unpublished). Homozygous follicle cell clones for dome or hop were induced using the UAS-FLP method(Duffy et al., 1998).dome FRT19A/FM7 or hop^C111 FRT101/FM7 females are crossed to UB-GFP, FRT19A/Y; e22c-GAL4, UAS-flp/CyO or UB-GFP, FRT101/Y; e22c-GAL4, UAS-flp/CyO males, and mosaic egg chambers are analysed in dome, FRT19A/UB-GFP, FRT19A;e22C-GAL4, UAS-flp/+ or hop^C111,FRT19A/UB-GFP, FRT19A; e22C-GAL4, UAS-flp/+ females. The expression of the FLP recombinase under the control of the e22c-GAL4 line induced mutant follicle cell clones, which were identified by the absence of GFP expression (Duffy et al.,1998). The following Drosophila stocks have been used:dome^PL100 FRT19A/FM7i; dome¹²⁰³⁰FRT19A/FM7i; dome^PG14 FRT19A/FM7i. Other stocks used in this study are: hop^C111 FRT101/FM7; UAS-hop;UAS-Stat92E; UAS-upd (a gift from M. Zeidler and N. Perrimon). S. Brown and J. Castelli-Gair kindly provided us with UAS-dome and UAS-domeΔCYT(Brown et al., 2001). TheCY2-GAL4 and slbo-GAL4 lines used in this study have been described elsewhere (Queenan et al.,1997; Rorth,1998).

Several independent P-element insertions (PL58, PL100, PG14, 12030, PG5,PG35) (Bourbon et al., 2002)have been identified in domeless and mapped in the 5′UTR region of the gene between position -20 and +405 of the putative transcription start site. A full-length cDNA (LD46805) was obtained from the Berkeley Drosophila Genome Project (BDGP) and fully sequenced using an automated ABI DNA sequencer(Accession Number, AY147847). The dome cDNA is 4805 bp long and contains 478 bp of 5′UTR, 3846 bp of coding region and 480 bp of 3′UTR. It encodes a protein of 1282 amino acids, whose sequence is identical to the one described previously(Brown et al., 2001;Chen et al., 2002). The pUAS-dome construct was made by cloning a full-length domecDNA (LD46805) cut with EcoRI and XhoI and ligated into theEcoRI and XhoI sites of the transformation vector pUAST. pUAS-domeΔEXT was constructed by cloning aEcoRI-XhoI fragment containing the 5′UTR, and the sequence encoding amino acids 1-27 (signal peptide) fused to amino acids 873-1281 (transmembrane and intracellular region) into the transformation vector pUAST cut with EcoRI and XhoI.

pUAS-domeGFP was constructed by single step ligation of the following DNA fragments: (i) EcoRI-XhoI fragment containing the 5′UTR and coding region (1-1281), (ii) a XhoI-XbaI fragment encoding EGFP and (iii) the transformation vector pUAST cut withEcoRI and XbaI. This results in the insertion of GFP at the C terminus of a wild-type Dome protein. For each construct, several independent transgenic lines have been generated and tested.

A glutathione S-transferase (GST) fusion protein containing Dome amino acids 918-1113 was produced in E. coli and used to immunize female New Zealand rabbits, according to standard protocols. The sera were collected and tested by western blot analysis to ensure specificity (data not shown). To purify the sera, a His-tagged-Dome (aa 918-1113) protein was produced inE. coli. 1 mg of this protein was separated by SDS-PAGE, and blotted onto a nitrocellulose filter. A strip containing the purified protein was used to affinity purify the Dome antibodies. Following extensive washing, the bound antibodies were eluted with elution buffer (200 mM glycine and 1 mM EDTA, pH 2.8). The eluate was neutralized immediately with one-tenth volume of 1 M Tris(pH 8.0). This antibody was used on western blots and for immunohistochemistry.

Staining of egg chambers with X-gal or antibodies were performed as described previously (Lasko and Ashburner,1990; Suzanne et al.,2001; Tanentzapf et al.,2000). The following primary antibodies have been used: rabbit anti-Dome (1:200), mouse anti-Fas3 (1:40; 7G10, Developmental Studies Hybridoma Bank-DSHB), mouse anti-β-galactosidase (1:1000, Promega), mouse anti-Crumbs (1:50; Cq4, DSHB); rat anti-DE-cadherin (1:50, a gift from Hiroki Oda), rabbit anti-Stat92E (1:500, a gift from Steven Hou). Secondary antibodies used in this study were anti-rabbit FITC (fluorescein)-tagged(1:400), anti-mouse CY3 (1:400; Molecular Probes). Confocal images were taken with a Leica TCS-SP1 confocal microscope. Other images were taken using a Nikon Coolpix 990 digital camera and processed with Photoshop 6.0 (Adobe).

In a screen for mutations on the X chromosome affecting follicle cell development (Bourbon et al.,2002) (unpublished results), we identified several independent P-element insertions leading to defective migration of border cells(Fig. 1C-E). All P elements are inserted in the 5′UTR of the CG14226 gene, at cytological position 18D13-E1 (see Materials and Methods). While this study was being conducted,this gene was independently identified for its role in JAK/STAT signaling during embryogenesis and named domeless [dome; a.k.a.mom (Brown et al.,2001; Chen et al.,2002]. dome encodes a putative transmembrane protein with distant homology to vertebrate cytokine receptors, and is proposed to function as a receptor molecule in the Drosophila JAK/STAT pathway during embryogenesis (Brown et al.,2001; Chen et al.,2002) (data not shown).

In dome mosaic egg chambers, some BC clusters are not formed,migrate aberrantly and/or contain less cells than normal(Fig. 1D,E). In these cases,dome mutant cells are always found in the anterior region of the egg chamber, indicating that they cannot be determined to become migratory BCs. In mosaic BC clusters, wild-type cells usually migrate normally or ahead ofdome mutant cells (Fig. 1E; data not shown), suggesting differences in the migratory ability and adhesiveness of mutant and wild-type cells.

Interestingly, in heterozygous females the BC cluster shows a slow border cell migration phenotype, aberrant shape and/or incomplete number of cells in the cluster due to the detachment of some BCs(Fig. 1C; data not shown). This suggests a semi-dominant, dose-sensitive effect of dome in this process and a possible role of dome during migration itself.

In addition to an aberrant migration of the BCs, we also found defects in the encapsulation of early egg chambers(Fig. 1F). In some cases, this can lead to the fusion of adjacent egg chambers(Fig. 1G), suggesting thatdome may also be involved in the assembly of egg chambers during early oogenesis (see below).

dome is a maternal effect gene controlling tracheal development and segmentation in embryos where it participates in JAK/STAT signaling(Brown et al., 2001;Chen et al., 2002). Interestingly, members of the JAK/STAT pathway, including upd, hopand Stat92E, have recently been shown to be involved in oogenesis for proper stalk and BC differentiation (Baksa et al., 2002; Beccari et al.,2002; McGregor et al.,2002; Silver and Montell,2001). However, it is not known whether Dome is a receptor for JAK/STAT signaling during oogenesis and whether it is activated by the Upd ligand. In order to address these questions, we tested for genetic interactions between dome and other members of the JAK/STAT pathway using BC migration as an assay. Since dome, upd and hopgenes are all located on the X chromosome, we have only been able to test interactions with Stat92E (Hou et al., 1996; Yan et al.,1996) and the recently identified dpias (a.k.a.Su(var)2-10) gene (Betz et al.,2001), which are positive and negative regulators of the JAK/STAT pathway, respectively. Removing one copy of Stat92E aggravatesdome/+ BC migration phenotypes, while removing one copy ofdpias suppresses them (Fig. 1H). This supports a model in which dome, Stat92E anddpias would participate in the same pathway. In order to test the role of dome in JAK/STAT signaling more directly, we analysed Stat92E expression in follicle cells. Upon activation, Stat proteins translocate into the nucleus to regulate target gene transcription. In dome mutant follicle cells, nuclear Stat92E is dramatically reduced(Fig. 1I,J), as it is in cells mutant for hop^C111 (data not shown). Since the active,nuclear form of Stat92E is strongly affected, these data indicate thatdome is required cell autonomously to activate JAK/STAT signaling during oogenesis.

In egg chambers, upd expression is restricted to two pairs of polar cells, which are located at the most anterior and posterior parts of the follicle cell layer (Fig. 2A,B)(Beccari et al., 2002;McGregor et al., 2002;Silver and Montell, 2001). Interestingly, we found that the upd mRNA is concentrated at the apical side of the polar cells, which is suggestive of a polarized synthesis and secretion of this ligand during oogenesis(Fig. 2A,B). Upd is proposed to signal to neighboring cells, committing them to the BC fate(Beccari et al., 2002;Silver and Montell, 2001). It is thus important to determine the expression pattern of domerelative to its putative ligand. In situ hybridization suggests thatdome is expressed at low level in all germline and follicle cells(Fig. 2C). Indeed,dome transcripts can only be detected over background level after overexpression using the UAS-GAL4 system(Fig. 2D)(Brand and Perrimon, 1993). To determine the subcellular localization of the Dome protein in follicle cells,we raised antibodies directed against an intracellular fragment of Dome (aa 918-1113; see Materials and Methods). Consistent with in situ data,immunostaining of egg chambers reveals that Dome is expressed in all germline and follicle cells. Dome is a membrane protein whose localization is restricted to apicolateral regions (Fig. 3). The membrane staining is specific, since it is absent in cells that are mutant for dome (Fig. 3A,B).

The subcellular localization of Dome protein was further analysed using a Dome-GFP construct which retains wild-type activity (see below;Fig. 5D,G,H). Expression of Dome-GFP in follicle cells targets the fusion protein to apicolateral membranes, in a pattern similar to wild-type Dome (compareFig. 3B and 3C). In addition,the fusion protein accumulates in intracellular vesicles(Fig. 3C). This specific vesicular pattern may reflect an enhanced trafficking of Dome in the secretory and/or endocytic pathway(s) due to overexpression. Consistently, the overexpression of the wild-type Dome protein, or a truncated form(DomeΔEXT, a deletion of the extracellular domain; seeFig. 5D) also gives rise to similar Dome-containing vesicles (Fig. 3D,E).

The presence of Dome in intracellular vesicles after overexpression prompted us to check whether Dome could undergo detectable trafficking in normal conditions. Indeed, detailed analysis of Dome protein localisation in follicle cells reveals a specific accumulation of endogenous Dome in intracellular vesicles, starting from stages 2-3 of oogenesis. In early stages(Fig. 4A,B), Dome vesicles are present in all follicle cells, but later they become restricted to regions where the Upd ligand is most abundant, i.e., close to the polar cells(Fig. 4C,F-I). These vesicles contain Dome protein, since they are absent in dome mutant follicle cell clones (Fig. 4B,H,I). Dome-containing vesicles are present in BCs before and during migration(Fig. 4C-E), thus following the migration pattern of upd-expressing polar cells(Fig. 1A). In the posterior region where polar cells also express upd(Fig. 2A,B), the same pattern of vesicles is visible, though these cells do not migrate(Fig. 4F,H).

Interestingly, the abundance and distribution of Dome vesicles depend on the distance from the polar cells, which are the source of the Upd ligand(Fig. 4F,G). To test whether vesicle formation is Upd dependent, we generated flies in which updwas ectopically expressed. Cells newly expressing upd accumulate a high level of Dome vesicles (Fig. 4J,K). In addition, overexpression of upd in the BCs leads to an elevated number of vesicles(Fig. 4L), as compared to wild-type cells (Fig. 4C). Thus, Upd can promote the internalization of the Dome receptor. Vesicles may likely reflect active endocytosis of the Dome receptor after binding to its ligand, a known mechanism by which receptor-ligand complexes are inactivated or recycled. Consistently, the intracellular domain of Dome contains several tyrosine-based and di-leucine motifs, which have been shown to work as internalization sorting codes in several vertebrate receptors(Trowbridge et al., 1993). It is noteworthy that both endogenous and ectopic Dome vesicles are found preferentially at the apical side of the cells, a region where theupd mRNA, and probably the Upd protein, are most abundant. In this respect, the pattern of endogenous Dome vesicles may well reflect the level and pattern of Dome and JAK/STAT pathway activation in follicle cells.

To assess the contribution of specific domains in Dome activity, we performed a structure-function analysis by expressing truncated Dome proteins in BCs using slbo-GAL4 as a driver. In this study we used UAS-dome, UAS-domeGFP, UAS-domeΔCYT (deletion of the cytoplasmic domain) (Brown et al.,2001), and UAS-domeΔEXT (deletion of the extracellular domain; see Fig. 5D). The fate and migratory phenotype of BCs were assessed using the slbo-lacZ marker (Fig. 5A) (Montell et al.,1992). Overexpression of Stat92E or hop leads to migration defects, while expression of the upd gene causes the formation of extra BCs (Fig. 5B,C,M) (Beccari et al.,2002; Silver and Montell,2001). Excess of Dome or DomeGFP in BCs blocks their migration(Fig. 5E-H), which never takes place, as indicated by the presence of slbo-lacZ-positive cells at the tip of old egg chambers, and female sterility (data not shown). This phenotype indicates a dose-sensitive effect of dome on BC migration,reminiscent of the dose sensitivity of the slbo gene(Rorth et al., 2000), another gene that is crucial for BC migration.

In egg chambers expressing DomeΔEXT or DomeΔCYT, outer BCs are absent and only polar cells are formed and express the slbo-lacZmarker. This phenotype, which is similar to a complete loss of domein BCs (Fig. 1), is consistent with these truncated Dome proteins behaving as dominant negative forms. We confirmed this conclusion by looking at the embryonic cuticle phenotypes induced in early embryos using a maternal GAL4 line(Brown et al., 2001; data not shown).

To date, Dome is the only known receptor for the JAK/STAT pathway inDrosophila. In order to test the absolute requirement ofdome in upd signaling during oogenesis, we used an epistasis test in egg chambers. When the upd gene is overexpressed, extra BCs are recruited and the level of slbo-lacZ expression is strongly enhanced, indicating that slbo is positively regulated in conditions of high JAK/STAT activity (compare Fig. 5A and 5M) (Silver and Montell,2001). The co-expression of upd and a domedominant negative construct (UAS-DomeΔCYT) completely suppresses the gain-of-function phenotypes associated with updoverexpression, including recruitment of extra BCs and enhancedslbo-lacZ expression (compare Fig. 5M and 5N). The resulting phenotype is similar to the simple expression of DomeΔCYT (Fig. 5I). These results demonstrate that dome is downstream ofupd to activate JAK/STAT signaling in follicle cells, and thatdome is essential for upd function.

Either reduction, as in dome heterozygotes(Fig. 1C), or elevation, as in overexpression experiments (Fig. 5E-H), of dome expression leads to BC migration defects,suggesting that dome function is tightly regulated in normal egg chambers. To further test this hypothesis, we examined Dome expression in cells that are mutant for JAK/hop. In hop mutant follicle cell clones, the level of Dome protein is higher than in the neighboring wild-type cells (Fig. 6A),suggesting that hop normally downregulates Dome expression. Since this effect could be due to either transcriptional or post-transcriptional regulation, we tested whether dome gene expression itself is affected. For this purpose, and because endogenous dome transcripts are barely detectable (Fig. 2C), we used a dome-Gal4 line as a reporter ofdome expression. dome-Gal4 is a pGAL4 enhancer trap element that is inserted in the promoter region of the dome gene (20 bp upstream of the transcription start site;Fig. 6E). dome-GAL4 is expressed weakly in anterior and posterior follicle cells, but much more highly in polar cells and BCs (Fig. 6B). Expression of UAS-dome leads to a strong reduction of dome-GAL4 driven GFP expression in BCs(Fig. 6C), suggesting that elevated dome activity can downregulate its own expression. In contrast, expression of the dominant negative form DomeΔEXT did not significantly affect the level of dome-GAL4 activity(Fig. 6D). These data suggest that in normal egg chambers, dome expression is negatively controlled by the JAK/STAT pathway. In agreement with this proposal, the analysis of thedome promoter region (12 kb upstream to the transcription start site)revealed the presence of 2 short sequences that perfectly match the consensus Stat92E binding site, TTCNNNGAA (Fig. 6E) (Yan et al.,1996). These putative Stat92E sites are similar to those present in the eve promoter, which have been shown to respond to JAK/STAT signaling (Yan et al.,1996).

In order to better understand how dome may control egg chamber development, we examined the expression of different follicle cell markers indome mosaic egg chambers. Previous work has shown that the JAK/STAT pathway controls the accumulation of DE-cadherin and Fas3 in follicle cells(Baksa et al., 2002;McGregor et al., 2002;Silver and Montell, 2001). Indome mutant clones, DE-cadherin is strongly reduced(Fig. 7A,B) suggesting thatdome participates in the formation of a correctly differentiated epithelium throughout oogenesis. Up until stage 3, Fas3 is abundant in all follicle cells (Ruohola et al.,1991), while later, Fas3 level drops sharply in all follicle cells except in the polar cells (data not shown). Therefore, Fas3 serves as a differentiation marker to monitor the maturation of developing egg chambers. The removal of dome leads to an abnormally high accumulation of Fas3 in older egg chambers, indicating that these cells did not differentiate normally and remained immature (Fig. 7C-F).

The formation of incompletely encapsulated egg chambers in domemosaic females (Fig. 1F,G) is reminiscent of the loss-of-function phenotype of crumbs mutations(Tanentzapf et al., 2000). Indeed, crumbs is required for the initial polarization of follicle cells, to control their mesenchymal-epithelial transition. In the absence ofcrumbs, precursors of the follicle cells do not encapsulate the germline cells, leading to egg chambers with epithelial discontinuities. In contrast, crumbs mutant cells that have been generated after the formation of the follicular epithelium have no apparent abnormalities,indicating that Crumbs is required for initial polarization of the epithelium,but not for its maintenance (Tanentzapf et al., 2000). Interestingly, dome mutant cells do not express the apical determinant Crumbs in follicle cells(Fig. 7G-J). We thus conclude that dome is necessary for the initial polarization of the follicular epithelium in the germarium through the control, direct or indirect, ofcrumbs expression.

The dome gene has recently been identified because of its segmentation phenotype which is similar to that of upd, Stat92E andhop mutants (Brown et al.,2001; Chen et al.,2002). As dome encodes a putative membrane protein with distant homology to vertebrate type I cytokine receptors, it is proposed to function as a receptor for Upd. This conclusion is supported by experiments made in cell culture showing that nuclear translocation of Stat92E requires Dome in cells treated with Upd-conditioned medium(Chen et al., 2002).

We show here that dome interacts genetically with theStat92E and dpias genes during BC migration, and thatdome phenotypes in ovaries are similar to those found inStat92E and hop mutants(Baksa et al., 2002;Beccari et al., 2002;McGregor et al., 2002;Silver and Montell, 2001). Furthermore, Stat92E nuclear localization is lost in dome mutant follicle cells (Fig. 1I,J),indicating that the mechanisms leading to Stat02E activation and subsequent nuclear translocation require dome. Since dome is epistatic to upd (Fig. 5M,N),our data indicate that dome is required downstream of updand upstream of Stat92E for JAK/STAT signaling in egg chambers. Altogether, these results provide strong evidence that Dome is a receptor molecule for Upd during oogenesis.

Our study shows that Dome is not uniformly distributed at the membrane but is restricted to apicolateral regions. Other receptor molecules have been shown to preferentially localize to apicolateral membranes, such as the EGF and Notch receptors (Lopez-Schier and St Johnston, 2001; Sapir et al.,1998), suggesting that the apical region is an active signaling interface for several receptors in follicle cells. Indeed, the apical localization of upd mRNA, membrane Dome and Dome-containing vesicles support a model in which ligand-receptor interactions take place apically in follicle cells, to activate the JAK/STAT pathway.

Dome is a transmembrane protein with both extracellular and intracellular domains whose functions are unknown. The extracellular part contains a cytokine-binding module (CBM) and 3 fibronectin-type III domains likely participating in ligand binding, while the intracellular domain presumably interacts with Hop, through binding to one or several potentially phosphorylated tyrosines (Brown et al.,2001; Chen et al.,2002; unpublished results). Using truncated forms of Dome we show that both the extracellular and intracellular domains are essential for BC migration and signal transduction (Fig. 5). The dominant negative phenotypes that are observed are consistent with a model in which DomeΔCYT would titrate the ligand Upd,and DomeΔEXT would titrate Hop, therefore inducing a dramatic reduction in signaling strength. Both constructs may also lead to the formation of non-functional Dome-Dome dimers by capturing the wild-type Dome protein in an inactive complex. Further biochemical work will be necessary to understand the molecular mechanisms underlying Dome signal transduction.

Previous studies have shown that the migration of BCs is sensitive to gene dosage, making it useful for genetic screens(Beccari et al., 2002;Duchek and Rorth, 2001;Duchek et al., 2001). The reduction or elevation of slbo, a gene encoding a C/EBP homolog(Montell et al., 1992), is sufficient to produce BC migration defects. Consistently, recent work has shown that Slbo protein levels are tightly regulated by the ubiquitination pathway (Rorth et al., 2000). Our results show that the BCs are also sensitive to changes in Dome protein levels. Indeed, either a decrease (Fig. 1) or an increase (Fig. 5) of Dome causes BC migration defects. There are several mechanisms by which gene activity can be regulated, including post-translational regulation, as with the Slbo protein(Rorth et al., 2000), or transcriptional regulation. Our data suggest that dome expression is regulated in part by a transcriptional negative feedback loop(Fig. 6). Two consensus STAT binding sites (Yan et al.,1996) present in the promoter region of the dome gene may prove to be important for this regulation. Interestingly, it has been shown that vertebrate STAT proteins can have both positive and negative regulatory functions (Ramana et al.,2000). Further work will be necessary to determine whether Stat92E is a direct repressor of dome.

In addition to a transcriptional control of dome, there are also post-translational mechanisms regulating Dome function in follicle cells,through a dynamic pattern of intracellular vesicles. We show that these Dome-containing vesicles are located at a relevant distance from Upd-producing cells, and that Upd can promote the formation of de novo vesicles upon ectopic expression. These results, together with the presence of several tyrosine-based and di-leucine motifs known to sort proteins for internalization (Trowbridge et al.,1993), are consistent with Dome-containing vesicles being the result of endocytosis upon ligand binding. Endocytosis is an important process controlling several signaling pathways during development(Seto et al., 2002), acting on the recycling or desensitization of ligand-receptor complexes. Our work thus provides the first experimental evidence that JAK/STAT signaling in flies may be regulated by endocytosis.

Our analysis of dome reveals several important functions in follicle cells. First, we show that during early oogenesis, dome is required for the encapsulation of germline cells into a functional egg chamber. dome mutant follicle cells made in the germarium are unable to assemble into the nascent follicular epithelium, thus leading to incompletely encapsulated egg chambers at stage 2-3(Fig. 1F,G). The fusion of some egg chambers seen in our and other studies(Fig. 1G and data not shown)(Baksa et al., 2002;McGregor et al., 2002),suggests that follicle cells normally separating adjacent egg chambers in the germarium have rapidly degenerated. This conclusion is supported by the fact that mutant cells cannot be observed in early egg chambers(Fig. 1). This is an alternative to the model in which the formation of fused egg chamber would be associated to stalk cell defects (McGregor et al., 2002).

The dramatic, early follicle cell phenotype contrasts with the essentially normal phenotype of dome mutant cells observed in later stage egg chambers. In this case, follicle cells are viable and divide normally (Figs3,4,6,7). A similar, dual phenotype has been reported in crumbs mutant chambers. After initial polarization of the follicle cells in the germarium, Crumbs is no longer required and its loss has no visible effects(Tanentzapf et al., 2000). Importantly, we show here that dome controls Crumbs expression in follicle cells, thus providing a novel link between the JAK/STAT signaling pathway and epithelial polarity.

In addition to its early function in the germarium, dome is required for the normal expression of several follicle cell markers, including DE-cadherin and Fas3. It is important to note that despite a clear defect in the expression of these markers, dome mosaic egg chambers are morphologically normal. However, because completely mutant egg chambers cannot be obtained because of the early effect of dome in the germarium, one cannot rule out the possibility that large mutant clones would lead to abnormal development of egg chambers.

The pattern of epithelial markers in dome mutant cells indicates that the JAK/STAT pathway is active in all follicle cells(Fig. 7), a notion that is reinforced by the wide expression of nuclear Stat92E(Fig. 1). How is Dome activated during egg chamber development and does this activation follow the same profile at all stages? Given the restricted pattern of upd expression in the egg chamber and its dramatic effect upon overexpression, it is unlikely that Upd is able to signal long distances in the follicular epithelium of late stage egg chambers. Rather, we favor a model by which the JAK/STAT pathway plays a pre-patterning function, acting early during egg chamber development to activate DE-cadherin and Crumbs expression (see alsoMcGregor et al., 2002). This view is consistent both with the expression pattern of upd and the distribution of Dome-containing vesicles described in this study. We have shown that the formation of endogenous vesicles can be promoted by Upd, and that a gradient of such vesicles is present around polar cells. Strikingly,these vesicles, which likely indicate active signaling through Dome, are widespread at early stages and become more restricted later on. We propose that during early development, the Upd signal produced by anterior and posterior polar cells contributes to the differentiation of all follicle cells. At this stage, Upd would be more diffusible than later, as suggested by the pattern of Dome intracellular vesicles(Fig. 4). The study of the mechanisms controlling Dome activation and Upd activity will require additional tools to directly detect Upd, as, for example, Upd-GFP fusion proteins.

In contrast to the situation in main body follicle cells, the role ofdome in BCs is essential. dome and other JAK/STAT pathway components (Beccari et al.,2002; Silver and Montell,2001) promote the differentiation of a selected group of follicle cells into a cohesive migratory cluster, a process requiring several other inputs (Montell, 2001). Mutations in dome induce phenotypes ranging from a complete absence of BCs to non-cohesive BC clusters. This suggests that dome may be required during migration itself, in addition to its role in the recruitment of BCs at stages 8-9. Although only conditional mutants could help to address this question, the pattern of Dome vesicles in the BCs before and during migration supports a sustained activation of the JAK/STAT pathway. Such a requirement could also explain the semi-dominant migration phenotype ofdome heterozygous egg chambers(Fig. 1).

Our study has revealed several new findings about the function ofdome and the JAK/STAT pathway during oogenesis. Future work will help to understand how Upd and Dome initially interact at the cell surface and transduce the signal to downstream JAK/STAT pathway members.

We wish to thank S. Brown, J. Castelli-Gair, D. Harrison, S. Hou, H. Oda,N. Perrimon, P. Rorth, M. Zeidler and the Bloomington Stock Center for providing us with fly stocks and other materials. We also thank J. Castelli-Gair for discussions, and members of the laboratory for critical reading of the manuscript. Our work is supported by grants from Centre National de la Recherche Scientifique (ATIPE-CNRS), Ministère de la Recherche (ACI-Biologie du Développement), Association pour la Recherche sur le Cancer (ARC), Fondation pour la Recherche Médicale(FRM), Ligue Nationale contre le Cancer, EMBO YIP and fellowships from Fondation Human Caritative (E. G.), Ligue Nationale contre le cancer (F. C.)and Ministère de la Recherche (C. M.).