The following stocks were used: 10XSTAT92E-GFP (gift from E. Bach, Bach et al., 2007), UAS-dome-GFP (gift from C. Ghiglione), shg-lacZ (gift from JP. Vincent), UAS-Stat92E-GFP (gift from C. Janke); fru-Gal4 (Borensztejn et al., 2013), tj-Gal4 (gift from V. Mirouse), nos-Gal4 (gift from A. Boivin), upd-Gal4 (Khammari et al., 2011), shg::GFP (gift from Yang Hong); RNAi strains are from BDSC: UAS-shgRNAi (#38207, #32904, named #2 and #3, respectively, in the text), and VDRC: UAS-shgRNAi (#27082, named #1 in the text), UAS-updRNAi (#3282), UAS-hopRNAi (#102830); and w ¹¹¹⁸ (#3605), Canton S (#1), UAS-shgS (#65589) and tub-Gal80 ^TS (#7016) are from BDSC. 10XSTAT92E-GFP and shg-lacZ were maintained at 25°C. All crosses using the Gal4-UAS system were performed at 25°C, or at 18°C when tub-Gal80 ^ts was used (with fru-Gal4 and tj-Gal4 lines), until transfer of pharate pupae to 29°C. Females were dissected at 3–4 days of age.

Ovary dissection and immunostaining were performed according to the work of Torres et al. (2017). For native GFP observation of 10XSTAT92E-GFP females, ovaries were fixed in 4% formaldehyde, washed briefly 3x in PBS, 15 min in PBS-Triton 0.3%, incubated with DAPI and phalloidin for 30 min with agitation and rinsed 3x in PBS. Mounting was performed in Dabco (Sigma), and observations were carried out straightaway.

The following primary antibodies were used: rabbit anti-GFP (1:200, Interchim), rabbit anti-Stat92E (1:1000, D. Montell, used for PLA), rabbit anti-Stat92E (1:1000 S. Hou, used for immunostaining), and from DSHB: rat anti-E-cadherin (1:200), mouse anti-β-galactosidase (1:200), mouse anti-GFP (1:50), mouse anti-N-cadherin (1:200), and mouse anti-Fasciclin 3 (1:20). Secondary antibodies were all used at 1:200: anti-mouse Cy3 and anti-rat Alexa 647 (Jackson Laboratories), anti-rabbit Alexa 488, and anti-rat Alexa 647 (Thermo Fisher Scientific). Phalloidin (Sigma) and DAPI were used at 1:200 to label F-Actin and DNA, respectively.

PC counting and BC migration were assessed using a Leica DMRB epifluorescence microscope. Confocal images were acquired with a Leica SP8 inverted microscope driven by Las-X software, using 63x or 40x oil immersion lenses. Quantifications on confocal stacks were performed using FIJI and annotated with Affinity Designer. For GFP + FC number, we used only one confocal section passing through the middle of each follicle (visible PCs) to have consistent data regarding cell positions. For GFP quantification, total GFP signal sum projections were done. In Figure 1, measures correspond to “Raw Integrated Density” quantification of 25% follicle length at poles, whereas in Figure 3 and Supplementary Figure S4, they correspond to “Mean Grey Value” quantification of the entire follicle. All statistical analyses were performed using GraphPad PRISM.

RT-qPCR was performed on 20 pairs of ovaries, as previously described (Parvy et al., 2012), using the following primers:

Mirror—5′-GAACACCGAGGATAACGATCTG-3’/5′-CGGTCATCATGTCGCCAA-3′

H15–5′-GGTGAAGTTGACAAACAACGAG-3’/5′-TTGGGACTGCCGGGTAT-3′

mRNA amounts were normalized with RPL49 mRNA values, used as a reference.

The Duolink PLA was performed according to the manufacturer’s instructions (Sigma-Aldrich), using mouse anti-GFP and rabbit anti-Stat92E (D. Montell) antibodies. During dissection, the muscular sheath was removed manually to allow penetration of the components in FCs, and Tween 0.1% was used instead of Triton 0.3% that we normally use for regular immunostaining.

Proteins were chemically cross-linked with DSP (Thermo Fisher Scientific) prior to cell lysis. Co-immunoprecipitation assays were performed on ovary extracts from 50 females according to the manufacturer’s procedure, using GFP magnetic agarose beads (Chromotek) or Dynabeads™ protein G (Invitrogen) coated with an anti-E-cadherin antibody with 10 mM Tris-HCl pH7.5, 300 mM NaCl, 0.05% NP40 and 0.5 mM EDTA or PBS-Triton 0.02% wash buffer respectively. Proteins were resolved on NuPAGE™ 4–12% Bis-Tris gels (Invitrogen) and visualized by immunoblotting using mouse anti-GFP (1:1000; Roche) or rat anti-E-cadherin (1:250, DSHB) primary antibodies, and HRP anti-mouse (1:1000; Invitrogen) or anti-rat (1:5000; Sigma) secondary antibodies.

All experiments have been conducted between two and three times independently, and similar results were obtained in all cases.

In order to study JAK-STAT signaling regulation during oogenesis, we made use of a 10XSTAT92E-GFP reporter, which drives the expression of GFP-encoding sequences in cells where the pathway is physiologically active (Bach et al., 2007). We observed that JAK-STAT activity presents a gradient in FCs with the highest activity at each follicle pole as expected, but activity did not seem identical at all stages, with a striking difference between anterior and posterior poles from mid-oogenesis, suggesting that signaling activity is finely regulated (Figures 1B–D). It is of note that we only analyzed follicles up to stage 8 since, after this stage, the anterior JAK-STAT active BCs delaminate from the follicular epithelium and migrate through nurse cells towards the oocyte, rendering analyses more difficult. We quantified the reporter activity at each pole between stages 4 and 8 using two complementary approaches: the first by determining the mean numbers of GFP-positive (GFP+) cells (Figure 1E), and the second by quantifying GFP intensities (Figure 1F). During follicle maturation, the mean GFP + FC number indicates a small reduction of the JAK-STAT gradient size at the anterior pole over time. The same tendency of fewer GFP + FCs is observed at the posterior pole from stage 7, which becomes statistically significant and more important than at the anterior pole at late stage 8, thus establishing an asymmetry in JAK-STAT signaling between poles (Figure 1E). In addition, while the gradient size decreases, there is an increase in signaling intensity at both poles between stages 7 and 8, but signaling intensity was significantly lower at the posterior pole (Figure 1F). Both techniques thus confirmed that JAK-STAT signaling activity is symmetrical at both poles up to stage 7 but becomes asymmetrical at stage 8. Our results, therefore, uncover an asymmetry of JAK-STAT signaling in the number and intensity of active cells between anterior and posterior poles.

To verify this result, we tested another readout for JAK-STAT signaling. While in some cases the JAK-STAT receptor is active when it is localized at the plasma membrane (Blouin et al., 2016), receptor endocytosis can sometimes be a requirement for signaling activity (Marchetti et al., 2006). In all cases, receptor endocytosis reflects active transduction. In the follicular epithelium, the accumulation of Dome-GFP vesicles has been described as an indicator of efficient endocytosis, thus being an indicator of JAK-STAT activity (Silver et al., 2005; Devergne et al., 2007; Sahu et al., 2017). We expressed a GFP-fused form of Dome in all FCs with the fruitless-Gal4 (fru-Gal4) driver (Supplementary Figure S1A for driver pattern) and assessed JAK-STAT signaling by counting the number of cells containing Dome-positive vesicles (Figures 1G–I). A progressive decrease in mean numbers of FCs containing Dome vesicles is observed at both poles, confirming that JAK-STAT gradient size decreases with time. Consistently with our previous results, the decrease is smaller at the anterior pole than at the posterior pole, creating an asymmetry in JAK-STAT signaling between poles. There is a slight lag between the translation of the signaling asymmetry into a visible output between the two reporters, the asymmetry being visible with Dome endocytosis from stage 7 and with 10XSTAT92E-GFP from stage 8, reflecting an expected difference in the kinetics between these two types of reporters.

Taken together, and as schematized in Figure 1M, our findings demonstrate that, although anterior and posterior follicle poles display similar levels of JAK-STAT signaling activity up to stage 6, fewer posterior than anterior FCs are active as of stage 7. This creates an asymmetry between poles in the JAK-STAT signaling gradient size and intensity that must be precisely regulated.

Upd, secreted from PCs at each follicle pole, activates the JAK-STAT pathway in surrounding FCs. At stage 7, when we showed signaling becomes asymmetric, PCs have reached their final number of two at each pole. Hence, the asymmetry in JAK-STAT signaling activity cannot be explained by a difference in the number of ligand secreting cells. Rather, we hypothesized that a regulator could be acting within the follicular epithelium to control JAK-STAT signaling activity. In that case, the regulator should display similar dynamics in its expression, with an asymmetry between poles from stage 7 or slightly earlier to allow time for the JAK-STAT pathway to become asymmetric in response to its regulation. We tested several proteins that might respond to such criteria and found E-cadherin as a good candidate. During oogenesis, as reported previously (Godt and Tepass, 1998), E-cadherin is present in all follicle and germline cells, which can be observed with an E-cadherin immunostaining (Figures 1J,K). At the subcellular level, E-cadherin is detected differentially along the apico-basal axis: strong levels are observed at the apical membrane, where FCs not only interact together but also with germ cells (GCs), which represents an unusual E-Cadherin-mediated kind of adhesion as it occurs between different cell types (Godt and Tepass, 1998; González-Reyes and St Johnston, 1998), whereas lower levels are detected at the lateral AJs where FCs only interact with other FCs (Figure 1J). At the tissue level, although E-cadherin is distributed homogenously along the anterior-posterior axis until stage 5, stronger accumulation is observed in anterior compared to posterior FCs, the difference being accentuated at stages 7 and 8, while follicle size is increasing (Figures 1J,K). This result shows that E-cadherin distribution in FCs becomes asymmetric between anterior and posterior follicle regions from stage 6.

To determine whether the asymmetric distribution of E-cadherin in FCs along the anterior-posterior axis is due to post-translational modifications, affecting for instance protein stability or recycling, or if asymmetry is rather regulated at the transcriptional level, we used a transcriptional reporter (shotgun-lacZ). In this line, β-galactosidase detection reflects the activity of regulatory sequences of shotgun (shg), the gene encoding the E-cadherin protein in Drosophila. We found homogenous expression of the shg-lacZ reporter in FCs until stage 5, which becomes differential between anterior and posterior regions at stage 6, except for expression in PCs that are maintained high at all stages (Figure 1L). This demonstrates that the E-cadherin asymmetry in FCs between anterior and posterior follicle regions being set at stage 6 is regulated at the transcriptional level.

Altogether our findings show that JAK-STAT activity and shg expression display similarities in their spatial and temporal dynamics in the follicular epithelium during oogenesis. Notably, by mid-oogenesis, both E-cadherin levels and JAK-STAT activity are high in the anterior follicles, while they are significantly lower in the posterior (Figure 1M), raising the hypothesis that they could be involved in similar processes.

We, thus, wondered if E-cadherin and JAK-STAT signaling correlate in function during oogenesis. We investigated this hypothesis by using RNAi lines targeting shg and upd. To assess whether our tools were efficient, we first tested if we could observe known phenotypes associated with shg and upd knock-downs with these RNAi lines. It has been independently shown that E-cadherin (Niewiadomska et al., 1999; Pacquelet and Rørth, 2005) and JAK-STAT signaling (Beccari et al., 2002; Silver et al., 2005) are involved in BC migration, a cellular process occurring between stage 9 and 10. In addition, a genetic interaction between stat92E and shg was reported in BC migration (Silver and Montell, 2001). We used the upd-Gal4 driver to target RNAi against upd specifically in PCs (Supplementary Figure S1B for driver pattern), thus decreasing JAK-STAT signaling in FCs from stage 2 onward. We assessed BC migration in the resulting follicles at stage 10, when the migration process is complete in controls. 97% of follicles exhibit BC migration delay in an upd knock-down context (Supplementary Figure S2B,E). We next used two RNAi lines against shg, which we showed were efficient at decreasing E-cadherin levels in FCs (#1) or in GCs (#2) when driven with the fru-Gal4 or the nos-Gal4 lines respectively (Supplementary Figure S1A,B for driver patterns), as revealed by the absence of E-cadherin signal detection in the corresponding tissue (Supplementary Figure S2C,D). In both cases, we observed BC migration defects in more than 90% of stage 10 follicles (Supplementary Figure S2C–E). The RNAi tools with the Gal4 lines used are therefore efficient to affect JAK-STAT signaling and E-cadherin functions, leading to the expected BC migration delay phenotypes.

We, thus, used the same tools to investigate whether E-cadherin expression and JAK-STAT signaling are also involved in another common process during oogenesis. PC apoptosis takes place between stages 2 and 5 of oogenesis, resulting in only two PCs at each pole (Besse and Pret, 2003, Figure 2A). JAK-STAT signaling has been shown to regulate morphogenesis leading to PC apoptosis by monosis in a non-cell-autonomous manner (Borensztejn et al., 2013; Torres et al., 2017, Figures 2B,E). We knocked down shg using the fru-Gal4 driver, which results in the absence of E-cadherin staining in FCs, without disturbing epithelial integrity as N-cadherin is also expressed in FCs thereby ensuring the maintenance of the follicular epithelium even in the absence of E-cadherin (Supplementary Figure S3A,B). We then assessed whether completion of PC apoptosis occurred by counting the number of PCs at each pole from stages 4–8. An excess of PCs (>2 per group at stage 7) was observed in the shg knock-down, with a significant difference from the control at all stages, reflecting a defect in the apoptotic process (Figures 2C,F). Similar results were obtained using yet another RNAi line against shg (#3) with a second FC driver (traffic jam-Gal4, tj-Gal4) (Supplementary Figure S1D for driver pattern and Supplementary Figure S4A). E-Cadherin, like the JAK-STAT pathway, is therefore required non-cell-autonomously in FCs for correct PC apoptosis.

E-cadherin can be found apically in FCs, at the interface with GCs (FC-GC contact) and laterally, at the interface between FCs (at FC-FC AJs only). To gain more insight as to which FC interface is involved in PC apoptotic extrusion, we decreased E-cadherin levels in GCs by knocking down shg from the germline with the RNAi line #2 driven by the nos-Gal4 driver, which is expected to disturb junctions at the FC-GC interface. As with the somatic shg knock-down, epithelial integrity is maintained in the germline knock-down (Supplementary Figure S3C), and extra PCs, indicative of apoptosis defects, are observed with significant differences at all stages compared to controls (Figures 2D,G). Our data thus show that decreasing E-cadherin in FCs or in GCs disturbs PC apoptosis in a non-cell-autonomous manner, similarly to decreasing JAK-STAT pathway activity in the follicular epithelium. Hence, E-cadherin and JAK-STAT signaling are both involved in two distinct morphogenesis processes that occur at different stages of oogenesis, namely PC apoptotic extrusion and BC migration, as schematized in Figure 2H. Altogether, our data demonstrate that E-cadherin accumulation and JAK-STAT signaling partially overlap in dynamics and share functions during oogenesis.

As E-cadherin is distributed in a larger domain than the one covered by JAK-STAT activity, and given that E-cadherin asymmetry between poles starts at stage 6, just before the asymmetry in JAK-STAT activity is visible, at stage 7, we wondered whether E-cadherin could be a general regulator of JAK-STAT signaling in the follicular epithelium. To test this hypothesis, we decreased E-cadherin levels by knocking down shg in FCs, in flies carrying the 10XSTAT92E-GFP transcriptional reporter, which allows for the assessment of JAK-STAT activity levels. As a control, we first decreased JAK-STAT activity by targeting RNAi against hop, encoding the kinase of the pathway, in FCs. As expected, when hop was knocked down, the JAK-STAT pathway activity reporter was strongly reduced at all stages, compared to controls (Figures 3A,B). Mean GFP intensity was quantified on sum projections of total GFP in the whole follicular epithelium, and, accordingly, a significant decrease was found at all stages considered (Figure 3E). When E-cadherin levels were decreased in FCs by knocking down shg (RNAi line #1), similar results were obtained, with the 10XSTAT92E-GFP reporter activity being reduced in the whole epithelium and mean GFP intensity being significantly decreased at all stages analyzed as compared to controls (Figures 3C,F). We confirmed that knocking down shg in FCs results in reduced JAK-STAT activity by expressing the shg RNA line #3 with tj-Gal4 (Supplementary Figure S1D for driver pattern and Supplementary Figures S4B–D).

As we found that E-cadherin is required in FCs and GCs to regulate PC apoptosis, we wondered if both FC and GC E-cadherin were also involved in the regulation of JAK-STAT signaling. We thus knocked down shg in the germline with the nos-Gal4 driver. Again, we found the JAK-STAT transcriptional reporter activity reduced in the whole follicular epithelium, with mean GFP intensity being significantly decreased at all stages compared to control (Figures 3D,G). This suggests that E-cadherin is required at FC-FC and FC-GC interfaces to regulate JAK-STAT activity in the follicular epithelium. In summary, our findings, which were confirmed with three different RNAi lines and three different Gal4 drivers, demonstrate that shg expression is required in FCs and in GCs for proper JAK-STAT activity in the follicular epithelium, suggesting that E-cadherin is a positive regulator of JAK-STAT signaling during oogenesis.

To confirm these findings by monitoring JAK-STAT pathway activity in a different way, we quantified the expression of JAK-STAT known target genes by RT-qPCR. mirror (mirr) is negatively regulated in response to JAK-STAT signaling (Xi et al., 2003), whereas H15 is positively activated by the pathway (Fregoso Lomas et al., 2016). Accordingly, we found that mirr was upregulated and H15 downregulated in ovarioles in which RNAi was used to knock-down upd in PCs, thus decreasing JAK-STAT signaling in the follicular epithelium (Figure 3H). When E-cadherin levels were decreased by knocking down shg in FCs or in GCs, similar mirr upregulation and H15 downregulation were obtained (Figures 3I,J). Given that similar changes in the expression of JAK-STAT target genes are detected upon decreasing JAK-STAT signaling and E-cadherin levels, this further confirm that E-cadherin is a positive regulator of JAK-STAT signaling during oogenesis.

The importance of Stat92E localization for efficient signaling activity has been shown in the embryo ectoderm epithelium, where its apical localization is ensured by interaction with the polarity protein Bazooka (Baz), a Drosophila Par3 homolog (Sotillos et al., 2008). To find out if the apical localization of Stat92E is conserved in ovaries, we looked at control follicles stained with an anti-Stat92E antibody. Besides the expected nuclear staining at follicle poles and in border cells at stage 9, where JAK-STAT activity is the highest, we also detected endogenous Stat92E apically in FCs (Figure 4A). Similarly, when we overexpressed a GFP-tagged version of Stat92E, besides the enrichment in nuclei at follicle poles, Stat92E was also localized apically and laterally in FCs from stage 3 to stage 9, partially overlapping with E-cadherin (Figure 4B). To gain more insight into the molecular mechanism by which E-cadherin could regulate JAK-STAT activity in follicles, we tested whether E-cadherin and Stat92E were in close proximity in FCs by performing a proximity ligation assay (PLA). For this, we used anti-Stat92E and anti-GFP antibodies on ovaries from shg::GFP Knock-In flies to test for E-cadherin-GFP/Stat92E proximity. Although it is not easy to detect PLA signals in ovaries due to penetration issues (Mannix et al., 2019), we found a significantly higher number of follicles with PLA foci in shg::GFP ovaries, as compared to the negative control that was nearly devoid of PLA signal (Figures 4C–E), suggesting that E-cadherin and Stat92E can be found in close proximity in FCs.

Finally, we tested whether E-cadherin and Stat92E could interact together in ovaries by performing co-immunoprecipitation assays. For this, Stat92E-GFP and shg were overexpressed in FCs with the tj-Gal4 driver, and co-immunoprecipitation was tested by using beads bound with anti-GFP antibodies (Figure 4F,F′). When only shg is overexpressed, E-cadherin is not immunoprecipitated from ovary extracts incubated with anti-GFP beads (Figure 4F, lane 4). However, when both Stat92E-GFP and shg are co-overexpressed, E-cadherin is co-immunoprecipitated with Stat92E-GFP from ovary extracts incubated with anti-GFP beads (Figure 4F, lane 2). Endogenous E-cadherin was also co-immunoprecipitated from flies expressing only Stat92E-GFP in FCs (Figure 4F’, lane 2). To confirm these results, we also performed the reverse immunoprecipitation on flies expressing only Stat92E-GFP in FCs and succeeded in co-immunoprecipitating Stat92E-GFP from E-Cadherin-coated beads, whereas no Stat92E was immunoprecipitated with the beads alone (Figure 4G). Our results indicate that Stat92E-GFP and endogenous E-cadherin are part of the same physical complex in FCs.

Altogether, our study unravels that E-cadherin is a new positive regulator of the JAK-STAT pathway during oogenesis and that E-cadherin and Stat92E are part of a complex in FCs, suggesting that E-cadherin could modulate JAK-STAT signaling in these cells through physical binding to Stat92E.