Cytokine exocytosis and JAK/STAT activation in the Drosophila ovary requires the vesicle trafficking regulator α-Snap

Paracrine signaling relies on the proper spatiotemporal release of activators from signaling cells, typically by exocytosis, to elicit distant responses. During trafficking, fusion of a vesicle to the appropriate target membrane requires coordination through multiple conserved proteins, including Snap receptor (SNARE) superfamily members (reviewed in Rothman, 1996; Zhao et al., 2007; Bombardier and Munson, 2015; Zhao and Brunger, 2016; Wickner and Rizo, 2017). Generally, an R-Snare in the vesicle membrane tethers to three Q-Snares on the target membrane, bringing the membranes close and docking them for fusion. Different SNARE proteins have distinct subcellular localizations, mediating specific fusion events. α-soluble NSF attachment protein (α-Snap) binds to the cis-SNARE complex and recruits the ATPase N-ethylmaleimide-sensitive Factor (NSF). Upon ATP hydrolysis, NSF and α-Snap promote Snare complex disassembly, which is required for another round of fusion.

While much of vesicle trafficking occurs constitutively, additional specialized mechanisms are needed for certain cases, such as cytokine or histamine release from immune cells, neurotransmitter release from neurons, or insulin secretion from β-pancreatic cells. (Burgoyne and Morgan, 2003; Südhof, 2012; Scheller, 2013; Xiong et al., 2017). In these cases, secretory cargo is separated and stored in vesicles, followed by rapid release that is independent from constitutive exocytosis. This is clearest in synaptic signaling, for which NSF and α-Snap are required.

Although a few specific roles have been suggested for NSF and α-Snap, their distinct developmental functions remain largely unknown. NSF and α-Snap are required for Drosophila neurotransmission (Kawasaki et al., 1998; Babcock et al., 2004; Yu et al., 2011; Li et al., 2015), and in hypothalamus and neuron development in zebrafish and mice, respectively (Chae et al., 2004; Hong et al., 2004; Kurrasch et al., 2009). Two genes encode NSF proteins in Drosophila – comatose (comt) and NSF2 – with some tissue-specific requirements (Siddiqi and Benzer, 1976; Boulianne and Trimble, 1995; Pallanck et al., 1995; Golby et al., 2001; Zhao et al., 2012). Notably, expression of a dominant-negative NSF2 mutant can phenocopy the loss of Wingless/WNT and Notch signaling, and can genetically interact with components of these pathways (Stewart et al., 2001) and other developmental signaling cascades (Laviolette et al., 2005). We found a specific requirement for α-Snap and an NSF component in ovaries to regulate Janus kinase/signal transducer and activator of transcription (JAK/STAT) signaling.

JAK/STAT signaling is well-conserved, required for development and immune function, and misregulated in disease (Amoyel et al., 2014; O'Shea et al., 2015; Villarino et al., 2015). In Drosophila, an Unpaired (Upd) protein binds to a cytokine receptor, which results in JAK activation, STAT phosphorylation and the regulation of target genes (Zeidler and Bausek, 2013; Chen et al., 2014). STAT signaling is required for the specification and migration of ovarian follicle cells that are called border cells (reviewed in Saadin and Starz-Gaiano, 2016a).

Developing egg chambers are encased by a monolayer of follicle cells that differentiate into multiple fates. The anterior and posterior-most follicle cells become polar cells, which are essential for patterning (reviewed in Wu et al., 2008; Duhart et al., 2017). Upd secreted from polar cells activates JAK/STAT signaling (Van de Bor et al., 2011; Hayashi et al., 2012). Early on, this signaling reduces the number of polar cells and specifies stalk cells (Baksa et al., 2002; McGregor et al., 2002; Assa-Kunik et al., 2007; Borensztejn et al., 2013, 2018). Later, follicle cells with the highest STAT activity become border cells and migrate to the oocyte (Fig. 1A,B) (Montell et al., 2012; Saadin and Starz-Gaiano, 2016a). Border cell specification and migration requires multiple inputs that control transcription, signaling, cytoskeletal organization, adhesion and protein trafficking. Here, we demonstrate a crucial role for α-Snap, Syntaxin1A (Syx1A) and NSF/NSF2 in mediating trafficking event(s) required to induce JAK/STAT signaling.

To assay vesicle trafficking in developmental signaling, we disrupted trafficking regulators by utilizing RNA interference (RNAi) (Table S1). We downregulated α-Snap, ɣ-Snap, comt/NSF, NSF2 and ten syntaxins (Q-Snares) (Littleton, 2000; Lloyd et al., 2000; Zhao et al., 2012) in a subset of follicle cells, including polar cells and border cell precursors by using c306-Gal4 (Fig. 1A,B).

Strikingly, depletion of α-Snap resulted in an unusual phenotype, i.e. egg chambers without border cell clusters, as assayed by expression of border cell-enriched proteins (Fig. 1C; Fig. S3C; see Materials and Methods). We observed this phenotype in 20–50% of stage 10 egg chambers by using two α-Snap RNAi lines (Fig. 1M; Table S1). Depletion of Syx1A resulted in the same phenotype (Table S1; Fig. 1D). Downregulation of other vesicle trafficking regulators produced wild-type egg chambers or other defects (Table S1). Thus, different vesicle trafficking genes are likely to have different key roles in follicle cells.

We confirmed the on-target effect of α-Snap RNAi by qRT-PCR (Table S2) and by rescuing the no-border cell phenotype through α-Snap overexpression (>3-fold rescue compared to controls, Fig. S1A,B,J,M). Overexpression of α-Snap-HA alone caused no obvious defects (Fig.S1C). We verified the temporal requirement for α-Snap by reducing its expression specifically in adults by using a temperature-sensitive (ts) version of the transcriptional regulator Gal80 (tub-Gal80^ts) and also rescuing the mutant phenotype in adults (Fig.S1D–F,K). By using α-Snap antiserum (Babcock et al., 2004), we detected the protein at or near the plasma membrane of follicle cells, consistent with it functioning there (Fig. 1E).

α-Snap recruits NSF, which hydrolyzes ATP to provide the energy for conformational changes that disassemble the SNARE complex, thereby resetting the components for a later round of fusion. However, recent studies have revealed that α-Snap can act independently of NSF activity, for example, to promote yeast vacuolar fusion (Zick et al., 2015; Schwartz et al., 2017; Song et al., 2017) or to regulate other cellular behaviors (Naydenov et al., 2012, 2014; Miao et al., 2013). Depletion of comt/NSF or NSF2 in follicle cells yielded wild-type phenotypes (Tables S1 and S2); thus, the function of α-Snap in these cells could be NSF-independent or the two NSFs might function redundantly (Golby et al., 2001). Consistent with the latter, depletion of both genes together caused lethality (Table S2). Different cell types show different levels of NSF/NSF2 expression (Ordway et al., 1994; Boulianne and Trimble, 1995; Pallanck et al., 1995; Golby et al., 2001; Mohtashami et al., 2001), indicating they have different demands for its function. We detected both genes by qRT-PCR from ovaries, and found high expression levels of NSF protein in the early germline and follicle cells, including polar cells, by antibody staining (Li et al., 2015) (not shown). Next, we overexpressed NSF2 in follicle cells. While overexpression of NSF2 alone yielded no mutant phenotype, it significantly rescued the phenotype caused by reduced α-Snap (Fig. S2). NSF2 overexpression may promote vesicle machinery resetting in the presence of a small amount of α-Snap through less favored interactions, or it may promote fusion independently of α-Snap at high concentrations – like overexpressed yeast Sec18 (Zick et al., 2015; Song et al., 2017). Since the two NSF proteins are more than 80% identical and both are detected in the ovary, we suspect that α-Snap normally functions with either in follicle cells. These results, and the role of Syx1A at the plasma membrane (Schulze et al., 1995), suggest that α-Snap, NSF and/or NSF2 and Syx1A are necessary for exocytosis during border cell specification.

α-Snap could be required in signaling cells for signal release or in responding cells to traffic receptors to the cell surface. To distinguish between these, we depleted α-Snap in different cell types: polar cells early (upd-Gal4, Fig. 1F,G) (McGregor et al., 2002), border cell precursors (not polar cells) (fruitless-Gal4, Fig. 1J,K) (Borensztejn et al., 2013), or later in specified border cells (not polar cells) (stage 8/9, slbo-Gal4, not shown) (Rørth et al., 1998). α-Snap depletion in border cells or their precursors yielded wild-type phenotypes (Fig. 1L,M, and data not shown). In contrast, depletion of α-Snap by using upd-Gal4 led to the absence of border cells in >80% of egg chambers (Fig. 1H,M), and a lack of polar cells by stage 10. This severe phenotype was rescued by α-Snap overexpression (Fig. S1G–I,L). Similarly, reduction of Syx1A in polar cells resulted in impaired polar cell maintenance and no border cells (Fig. 1I,M).

These results suggest that α-Snap and Syx1A are required in polar cells to promote border cell specification but are largely dispensable in border cells themselves. This is unlike depletion of exocyst components, which affects cell migration but not specification (Assaker et al., 2010; Laflamme et al., 2012; Ramel et al., 2013). Together, these results indicate that different vesicle trafficking regulators have cell-type specific key functions.

The lack of border cells upon reduction of α-Snap expression could be due to cell death or failed specification. A primary marker for polar cells, fasciclin III (FasIII) (Ruohola et al., 1991), was present early but not detected later when α-Snap was depleted, indicating the cells are specified but not maintained (Fig. 2A). We do not believe that this reflects a general requirement for α-Snap in cell viability since depletion in other follicle cells does not result in obvious defects (Fig. 1L). Different cell types differ in their ability to survive α-Snap disruption (Babcock et al., 2004; Chae et al., 2004; Hong et al., 2004; Tomes et al., 2005; Bátiz et al., 2009; Naydenov et al., 2014; Arcos et al., 2017) and might require different amounts of this regulator. To overcome polar cell loss, we expressed the anti-apoptotic gene p35 (Hay et al., 1994) in α-Snap-depleted polar cells. Polar cells were apparent at all stages in this genotype, indicating that α-Snap is essential for their maintenance (Fig. 2 compare panels A and B with panels D and E, respectively). Interestingly, >70% of these egg chambers still lacked border cells (Fig. 2F), revealing a separable, essential function for α-Snap to enable polar cells to specify border cells.

In mammalian cells, α-Snap is required for localization of E-cadherin (E-Cad), β-catenin, and apical proteins (Chae et al., 2004; Hong et al., 2004; Andreeva et al., 2005; Naydenov et al., 2012). Since early depletion of α-Snap resulted in the absence of polar cells, we could not investigate protein localization there. However, the localization patterns and/or levels of E-Cad and Armadillo (Arm) were normal in posterior cells upon disruption of α-Snap or Syx1A (Fig. 1C,D,H,I; Fig. S3A–C) and normal upon α-Snap disruption in border cells (not targeting the polar cells, Fig. 1L). Additionally, reduction of α-Snap affects border cells differently than loss of adhesion and polarity molecules: reduction of E-Cad results in cluster separation (Cai et al., 2014) and disruption of apical proteins yields poor border cell migration (Niewiadomska et al., 1999; Pinheiro and Montell, 2004).

Polar cells activate several signaling pathways but most do not appear to require α-Snap. Depletion of α-Snap did not result in fused egg chambers, stalk defects or oocyte polarity defects (Figs 1C and 2B), as would disruption of Hedgehog, Hippo and/or Notch signaling (Lopez-Schier and St Johnston, 2001; Grammont and Irvine, 2002; Chen et al., 2011). Furthermore, although polar cell fate requires Notch activity, we found no genetic interaction between Notch (N^ts2) and α-Snap (α-Snap^G8) in polar cell fate (Fig. S3D). These observations are consistent with cell culture studies linking α-Snap only to certain signaling cascades (Baeg et al., 2005; DasGupta et al., 2005; Nybakken et al., 2005). Notably, genetic disruptions of α-Snap dramatically compromise the functions of some cell types while not obviously affecting others (Babcock et al., 2004; Chae et al., 2004; Hong et al., 2004; Bátiz et al., 2009; Arcos et al., 2017), suggesting that some cells can overcome this defect. It is possible that less-sensitive cells synthesize SNAREs rapidly to reduce demand for recycling or have a ‘bypass’ mechanism that alleviates the need for α-Snap and/or NSF activity (Thorngren et al., 2004), or that NSF functions with a not-yet identified co-factor to promote exocytosis.

Because border cell specification requires JAK/STAT signaling, we tested whether α-Snap could regulate components of this pathway. We assayed the recessive mutant alleles Stat92E³⁹⁷ (Silver and Montell, 2001) and α-Snap^G8 (Babcock et al., 2004). Single heterozygotes have normal numbers of border cells but minor cell migration defects (<5% of stage 10 egg chambers had border cell migration defects; Fig. 3A,B,D) (Silver and Montell, 2001; Silver et al., 2005). Trans-heterozygous (Stat92E³⁹⁷/α-Snap^G8) flies had a significantly enhanced mutant phenotype (>20% of stage 10 egg chambers had border cell migration defects; Fig. 3C,D), indicating that α-Snap and Stat92E are likely to function in the same genetic pathway.

Next we examined the secreted activator Upd. In epistasis tests, we found that overexpression of upd in the polar cells significantly rescued the phenotype caused by one α-Snap RNAi line (Fig. 3E) and to a lesser extent by another (not shown), resulting in normal border cells. These results are consistent with a model in which α-Snap and Syx1A are necessary in polar cells to regulate Upd release.

During early oogenesis, Upd in polar cells promotes stalk cell fate (Lopez-Schier and St Johnston, 2001; Baksa et al., 2002; McGregor et al., 2002; Assa-Kunik et al., 2007) and eliminates excess polar cells (Borensztejn et al., 2013; Torres et al., 2017). Although we disrupted α-Snap and Syx1A early, we did not see stalk defects or extra polar cells (Fig. 2B), or changes in stretch or centripetal cells, which require low levels of STAT signaling (Xi et al., 2003). We speculate that α-Snap is not depleted early enough to cause these defects, or that low levels of α-Snap function permit low levels of STAT activity. Upd is not required for polar cell fate, and the survival signal(s) that α-Snap regulates to maintain polar cells is unknown.

We propose that α-Snap mediates the controlled exocytosis of Upd separately from constitutive trafficking. Our results add another control node to JAK/STAT signaling. Given the strong conservation of this pathway, it will be interesting to see whether the human α-Snap homolog NAPA, which is 60% identical to fly α-Snap (Lemons et al., 1997), regulates STAT-mediated immune responses.

Some secreted molecules are sequestered and released in a Ca²⁺-dependent, regulated manner (Südhof, 2012; Xiong et al., 2017), including secretory granules from neutrophils, neurotransmitter release and the sperm acrosome reaction (Littleton et al., 1993; Tomes et al., 2005; Stow et al., 2009; Scheller, 2013; Sheshachalam et al., 2014). The latter two events are known to be mediated by α-Snap, NSF and certain syntaxins. Since our data suggest that STAT activation depends on these regulators in polar cells, we examined intracellular Ca²⁺ flux in egg chambers using the verified Ca²⁺ sensors GCaMP6s and GCaMP6m, which fluoresce in response to high intracellular Ca²⁺ (Chen et al., 2013). Interestingly, we observed periodic, transient increases of intracellular Ca²⁺ signals in polar cells prior to border cell specification by using three different Gal4 drivers (Fig. 4A–C; Movies 1–3). Consistent with a role for Ca²⁺, we observed significantly fewer border cells when free Ca²⁺ was reduced in polar cells by overexpression of the human parvalbumin protein (PVALB, hereafter referred to as PV ) (Harrisingh et al., 2007) (Fig. 4D–G). The number of border cells is known to drop with decreasing STAT activity (Silver et al., 2005; Starz-Gaiano et al., 2008; Van de Bor et al., 2011). Thus, we favor a model in which Upd is released in a regulated non-constitutive manner in response to Ca²⁺ (Fig. 4H). Notably, cation concentrations govern morphogen release during fly wing development (Dahal et al., 2012, 2017). Thus, ion-regulated exocytosis might represent a widespread means to mediate specificity of signal release during animal development.

The GD11986 and KK107910 transgenic α-Snap RNAi lines were from Vienna Drosophila Research Center (Dietzl et al., 2007). The polar cell driver (upd-Gal4) (Khammari et al., 2011), and UAS-upd (21.2) (Silver et al., 2005) transgenic and stat92E³⁹⁷ mutant flies were provided by D. J. Montell (University of California Santa Barbara, Santa Barbara, CA). UAS-NSF2 transgenic fly line was a generous gift from L. Pallanck (University of Washington, Seattle, WA). All other transgenic and mutant fly lines including UAS-mCD8-GFP/CyO (Lee and Luo, 1999), UAS-GFP.nls 14 (Neufeld et al., 1998), the anterior follicle cell drivers: c306-Gal4 (Manseau et al., 1997), slbo-Gal4 (Rørth et al., 1998) and fruitless-Gal4 (Hayashi et al., 2002), tub-Gal80^ts/TM6 (McGuire et al., 2003), UAS-p35 (Hay et al., 1994), Ep-Dome (P(EPgy2)EY22614) (Bellen et al., 2011), αSnap^M4 and αSnap^G8 (Babcock et al., 2004), N^l1N-ts2 (Shellenbarger and Mohler, 1975), UAS-GCaMP6m and UAS-GCaMP6s reporters (Chen et al., 2013), tub-Gal4/TM3 (Lee and Luo, 1999), UAS-PV-Myc (Harrisingh et al., 2007), JF03266 and HMS00872 RNAi lines for α-Snap, and the rest of RNAi lines were from the Bloomington Drosophila Stock Center (Ni et al., 2008).

Genetic expression of transgenes in follicle cells was performed as previously described (Saadin and Starz-Gaiano, 2016b). All genetic crosses were established at 25°C. When offspring of a cross involving Gal4 were not viable at 25°C, the cross was maintained at 18°C or the temperature-sensitive allele of tub-Gal80 was introduced to the cross while the cross was maintained at 20–21°C (permissive temperature range for Gal80^ts) to repress Gal4. Female offspring (3–7 days old) were cultured on yeast-supplemented food and maintained at either 25°C, 29°C or 31°C, depending on their genotype, for different periods of time prior to dissection. Unless otherwise stated in the figure legends, flies bearing Gal4 were incubated at 29°C for 12–16 h. Flies bearing Gal4 and Gal80^ts were incubated at 31°C (non-permissive temperature for Gal80^ts) for 54 h, and flies that did not bear either Gal4 or Gal80^ts, were incubated at 25°C for 12–16 h before dissection.

Genetic crosses were established and maintained at 25°C. Since follicle cell Gal4 drivers only impact a fraction of the cells in the ovary, we tested RNAi efficiency in embryos. tub-Gal4 females were crossed with males from the indicated RNAi lines. Total RNA was isolated from 0–24-hour-old F1 embryos (25°C) by using the Qiagen RNeasy Plus Micro Kit. BioRad iScript was used for cDNA synthesis. qRT-PCR was performed in triplicates, using BioRad iTaq SYBR Green Supermix and the BioRad CFX96 thermocycler. Tubulin-Gal4 embryos were used as the calibrating sample, and relative changes in gene expression were calculated using rp49 as a reference gene. Fold-changes were determined using the Livak 2^-ΔΔCT method.

Primers used, were α-snap ex1-2 forward: 5′-TGGGTGACAACGAACAGAAG-3′, reverse: 5′-CCCAGCTTTGTCCAGTTTT-3′; α-snap ex2-3 forward: 5′-GCTGCCAAACATCACCAAAG-3′′, reverse: 5′-CCACCTTCAACATGCACTTG-3′; comt set2 forward: 5′-CCCGTGAAAATCAGCAAGAATC-3′′, reverse: 5′-ACCAACATTCACTATAGCCCG-3′; comt ex4-6 forward: 5′-ACACGGATATCTTTAGCAAGGG-3′′, reverse: 5′-ACAGGAACTTTATGGCCCG-3′; NSF2 ex2-3 forward: 5′-AGTTCCTCATGCAGTTCGC-3′′, reverse: 5′- GAGTCTTCGGTAGGGAATCG-3′; NSF2 ex3-4 forward: 5′-AGTTGGAGGGTCTGGTTAGAG-3′′, reverse: GCTTGATGTCGTTGTCCAATG-3′.

Fly ovaries were dissected to ovarioles in Schneider's medium supplemented with 10% fetal bovine serum (FBS) and 0.6× penicillin–streptomycin (Pen–Strep) (Prasad et al., 2007), and fixed in 4% paraformaldehyde diluted in 0.1 M potassium phosphate buffer. The egg chambers were immunostained following a previously established protocol (McDonald et al., 2006). Briefly, fixed egg chambers were washed in NP40 wash buffer (0.05 M Tris-HCl pH 7.4, 0.15 M NaCl, 1 mg/ml BSA, 0.5% Nonidet P-40 (Igepal CA-630, Sigma-Aldrich), 0.02% sodium azide) (McDonald et al., 2006) and incubated with primary antibodies (see below) diluted in NP40 wash buffer either overnight at 4°C or for 2–3 h at room temperature. The egg chambers were then washed four times in NP40 wash buffer and incubated with diluted (1:400) secondary antibodies (Alexa Flour 488 and Alexa Flour 568; Life Technologies) at room temperature for 2–3 h. Nuclei of the immuno-stained egg chambers were then stained with DAPI (1:1000; D1306, Invitrogen,) for 10 min, washed and mounted in 70% glycerol. Stage-10 egg chambers were scored for defects in border cell specification characterized by the absence of border cell markers, including the Drosophila β-catenin, Armadillo (Arm) (Peifer et al., 1993) (Fig. 1C), Eyes Absent (Eya) (Bai and Montell, 2002) (Fig. 1C), Apontic (Apt) (Starz-Gaiano et al., 2008), Slow border cells (Slbo) (Montell et al., 1992) and E-cadherin (E-Cad) (Niewiadomska et al., 1999) (Fig. S3C and data not shown), as well as defects in migration, where the border cell clusters had not reached the oocyte. The absence of the border cells was further confirmed using optical sections to examine the whole depth of the tissue. Data for c306-driven and fruitless-driven JF03266 RNAi lines in Fig. 1M are the average penetrance of four and two independent experiments, respectively. Statistical significance was established using the Mann–Whitney U test for border cell numbers and Fisher's exact test for proportions by using online tools (http://vassarstats.net).

The following primary antibodies were used: mouse anti-Eya (10H6, DSHB; 1:100) (Bonini et al., 1993), mouse anti-Armadillo (N27A1, DSHB; 1:40) (Riggleman et al., 1990), rabbit anti-Apt (provided by S. Hirose, National Institute of Genetics, Mishima, Japan; 1:1000) (Liu et al., 2003), guinea pig anti-Apt from GST-Apt (Pocono Rabbit Farm, 1:5000), rat anti-Slbo (provided by P. Rorth, Institute of Molecular and Cell Biology A-STAR, Proteos, Singapore; 1:1000) (Beccari et al., 2002), anti-rabbit GFP (Life Technologies; 1:250), rat anti-Drosophila E-Cad (DCAD2, DSHB;1:20) (Oda et al., 1997), rat anti-human influenza hemagglutinin (HA) (11867423001, Roche; at 0.5 µg/ml), mouse anti-FasIII (7G10, DSHB;1:50) (Patel et al., 1987), rabbit anti-NSF (provided by J. Han, Southeast University, Nanjing, China; 1:50; Li et al., 2015) and α-Snap (provided by L. Pallanck, University of Washington, Seattle, WA; 1:600; Babcock et al., 2004). Images were acquired with a Carl Zeiss AxioImager Z1 and Apotome optical sectioning with Axiovision acquisition software. Image size adjustments and figure assembly were completed using Photoshop CS6 Adobe.

Drosophila melanogaster α-Snap cDNA clone (LD21601) was obtained from Drosophila Genomic Research Center in pOT2 vector. The coding sequence – excluding the stop codon – to allow expression of the HA tag was amplified using primers that contained attB sites following the GateWay cloning protocol by Invitrogen/ThermoFisher. The Kozak sequence (CAAC) (Cavener, 1987) was also added to the forward primer just upstream of the start codon. The BP reaction was carried out using a pDNOR221 vector (Invitrogen, 12536-017) and Gateway BP Clonase II enzyme mix (Invitrogen, 11789020) followed by heat shock-induced transformation into One Shot OmniMax 2T1 competent E. coli cells (Invitrogen, C854003). Successful cloning of α-Snap into pDONR221 was confirmed by sequencing using M13F(-21) and M13R primers (Genewiz). The LR reaction was then carried out using a pDONR221 vector containing α-Snap and pUASg-HA.attB (Bischof et al., 2007, 2013) (Gene Bank Accession number KC896837.1) with a Gateway LR Clonase II enzyme mix (Invitrogen, 11791020). The product of the LR reaction was transformed into One Shot OmniMax 2T1 competent E. coli cells. Successful cloning was confirmed by sequencing, using the primers Forward 5′-CGTCGCTAAGCGAAAGCTAAGC-3′ and Reverse 5′-AGCCTGCTGCTACACTTGCC-3′. The Destination vector (pUASg-HA.attB) containing α-Snap was used for PhiC31 injection through BestGene, Inc, plan H.

Live imaging was performed following the previously established protocol by Manning and Starz-Gaiano, 2015. Briefly, fly ovaries were dissected in Schneider's medium containing 0.2 mg/ml insulin, 10% FBS and 0.6× Pen–Strep. Egg chambers were cultured in insulin-containing medium supplemented with 9 µM FM4-64 dye (Invitrogen) during imaging. The dissected ovarioles were imaged every 10–20 s at 1–2 s exposure time for 0.5–2 h using a Carl Zeiss AxioImager Z1 microscope with AxioVision acquisition software. Changes in fluorescence intensity were measured using ImageJ, if needed after background correction, by defining regions of interest (ROIs) around polar cells and by using Multi Measure on a stack of time-lapse images. The ROI of a nearby cell was defined as background. Change over background (ΔF) was normalized against fluorescence intensity (F) at the start of the imaging procedure (Macleod, 2012).

Stocks obtained from the Bloomington Drosophila Stock Center were used in this study. We thank the Bloomington Stock Center for RNAi and other transgenic flies, the Vienna Drosophila RNAi Center for RNAi fly stocks, and the Developmental Studies Hybridoma Bank, Drs D. Montell, L. Pallanck, J. Han, S. Hirose and P. Rorth for Drosophila mutant stocks and antibodies. Further, we thank Flybase for providing database information (Gramates et al., 2017). We thank Drs Jocelyn McDonald and Fernando Vonhoff for helpful comments on the research and manuscript. We acknowledge technical assistance from Grant Wunderlin.