The Drosophila insulin pathway controls Profilin expression and dynamic actin-rich protrusions during collective cell migration

Motility of different cell types is essential for proper embryogenesis. Later during development, cell migration plays a crucial role in the immune response, during inflammation and wound-healing (Montell, 2003; Ridley, 2003). It is well established that cancer cells can re-activate embryonic migratory programs, leading to their escape from the tumour through metastasis. In all these processes, cells can migrate either alone or as cohorts; in the latter case, cells can show complex organization into mixed populations with specific functions. Cell motility can either be permanent (i.e. immune cells) or only transient, for example during an epithelial-to-mesenchymal transition or when cells re-epithelialize after reaching their target tissue. Cell migration is therefore a highly heterogeneous phenomenon that is common to both normal and pathological processes, requiring the development of genetically amenable models to identify the different molecules and signalling pathways at work.

The migration of border cells provides a unique system with which to genetically dissect the mechanisms regulating invasive cell migration in vivo (Montell, 2006; Rorth, 2009). Border cells are a group of about eight somatic cells composed of two central polar cells and six to eight surrounding outer border cells (Montell et al., 1992). At stage 9 of oogenesis, border cells form a cohesive cluster that undergoes an epithelial-to-mesenchymal transition, leading to its delamination from the surrounding epithelium and its posterior-ward migration through the egg chamber (Fulga and Rørth, 2002) (Fig. 1A).

Genetic as well as genome-wide profiling studies (Borghese et al., 2006; Wang et al., 2006) have identified several signalling pathways controlling distinct processes involved in border cell assembly and/or migration: whereas the early specification of pre-migratory cells requires JAK/STAT signalling and its target gene slow border cell (slbo) (Beccari et al., 2002; Devergne et al., 2007; Ghiglione et al., 2002, 2008; Silver and Montell, 2001; Van de Bor et al., 2011), the timing and guiding of cell migration depend on Ecdysone signalling (Jang et al., 2009) and PVR/EGFR pathways, respectively (Duchek and Rorth, 2001; Duchek et al., 2001; McDonald et al., 2006).

Migration of border cells is initiated by the formation of a single actin-rich ‘long cellular extension’ (LCE) that enables motility through a ‘grapple and pull’ mechanism (Fulga and Rørth, 2002). Additionally, in the course of their migration, border cells extend and retract actin-rich protrusions dynamically through cycles of F-actin assembly and disassembly (Prasad and Montell, 2007). During this ‘treadmilling’ process, the dynamic actin cytoskeleton is regulated by a number of molecules, including highly conserved actin-monomer-binding proteins such as Profilin, which promotes actin polymerization, and Cofilin, which enhances filament depolymerization. Mutations in these two actin-regulating proteins lead to border cell migration defects (Verheyen and Cooley, 1994). Membrane ruffling and actin protrusions are therefore important for border cell migration; however, how exactly actin dynamics is controlled in border cells to drive cell locomotion is still not fully understood.

In this study, we identify the Drosophila Insulin/Insulin-like growth factor signalling (IIS) pathway as an important new regulator of border cell migration. The IIS pathway couples growth with nutrition (Andersen et al., 2013; Edgar, 2006), and, during oogenesis, it has been shown to regulate germline stem cell division, cyst growth, vitellogenesis and epithelial cell cycle progression (Drummond-Barbosa and Spradling, 2001; Jouandin et al., 2014; LaFever, 2005; LaFever et al., 2010). Here, we show that specific loss of IIS activity in border cells leads to an immotile phenotype. Our results show that the function of IIS in border cells is independent of its role on cell growth control. We reveal that activation of the Drosophila Insulin Receptor at the onset of migration relieves the repressive activity of FoxO on the Drosophila Profilin-encoding gene chickadee (chic), therefore promoting actin polymerization and the formation of protrusions that are essential to initiate and support migration. These results demonstrate that the IIS pathway controls local cell migration through the regulation of dynamic, actin-rich protrusions.

To identify new genes involved in border cell migration, we performed an RNAi-based screen using available collections of UAS-RNAi transgenic flies (see Materials and Methods; C.G., D.C. and S.N. unpublished). RNAi lines were individually crossed to a composite Gal4-expressing strain (USG>), carrying both the upd-Gal4 (UPD>; polar cell driver) and slbo-Gal4 (SLBO>; outer border cell driver) drivers, therefore targeting the expression of UAS-RNAi constructs in both polar and outer border cells (Fig. 1A′). This combination of Gal4 lines allows the targeting of all cell types that make up the border cell cluster.

Among the candidate lines identified by this screen, the RNAi construct targeting the unique Drosophila Insulin Receptor (InR) (Fernandez-Almonacid and Rosen, 1987; Petruzzelli et al., 1986) led to strong defects in border cell migration. Border cell clusters from USG>inr-RNAi stage 10 egg chambers were not migrating (about 75%) or migrated only partially (about 15%) (Fig. 1B,C). In addition, we found that the frequency and the severity of border cell migration defects were aggravated in a inr^ex15 (an amorphic allele for inr) heterozygous mutant background (USG>inr-RNAi; inr^ex15/+), with the proportion of clusters not migrating rising from 75% to 95% (Fig. 1C).

These results were confirmed by generating mosaic clones for inr^ex15 (see Materials and Methods) (Xu and Rubin, 1993), which led to mutant clusters failing to migrate (Fig. 1E, compare with D). Detailed analysis showed that migration was blocked when outer border cells were mutant for inr (Fig. 1G), while clusters with only mutant polar cells migrated normally (Fig. 1H). This suggests that inr does not play a role in polar cells for border cell migration. These results were further confirmed by selectively expressing inr-RNAi in either outer border cells (using SLBO>) or polar cells (using UPD>) (Fig. 1C).

To rule out any indirect effect of inr depletion on border cell differentiation, we stained mosaic inr^ex15 clusters with the two well-established border cell fate markers Slbo and Singed, and found that mutant border cells were specified normally (Fig. 1D-F; Fig. S1). Altogether, these results indicate that InR is required specifically in outer border cells for normal migration of the cluster.

Signalling downstream of the Drosophila Insulin Receptor involves Chico (the Drosophila homolog of vertebrate insulin receptor substrate), PI3K and AKT, which lead to the phosphorylation and subsequent repression, through cytoplasmic retention, of the Forkhead transcription factor FoxO (Puig et al., 2003).

Expression of chico-RNAi in BCs using the USG> driver led to impaired border cell migration (Fig. 2A), consistent with our InR genetic analysis (see above). However, it has been reported that targeted expression of a dominant-negative form of the PI3K catalytic subunit (dp110^D945A) using the SLBO> driver has no effect on border cell migration (Duchek and Rorth, 2001; Fulga and Rørth, 2002). We repeated this experiment using USG> and obtained the same result (data not shown). Thus, our identification of InR and chico requirement was surprising and suggested that InR may act in a non-canonical pathway to control migration. To test this hypothesis further, we first analysed whether a constitutively activated form of PI3K (PI3K*, dp110^CAAX) could suppress inr-RNAi border cell migration defects. Expression of PI3K* rescued the inr-RNAi migration phenotype (Fig. 2A), indicating that InR signals through PI3K during border cell migration. In addition, expression of either P60 (a PI3K regulatory subunit whose overexpression dominantly blocks PI3K; Weinkove et al., 1999) or the inhibitory phosphatase PTEN led to severe migration phenotypes (Fig. 2A,B). Finally, generation of mutant mosaic clusters for the dp110^1C1 amorphic allele or for the akt¹ allele and expression of akt-RNAi led to impaired border cell migration (Fig. 2A,C,D). These results therefore indicate that PI3K is involved in border cell migration. The discrepancy with previous results (Duchek and Rorth, 2001; Fulga and Rørth, 2002) is likely due to a poor efficiency of the dp110^D945A construct in blocking PI3K activity when driven by SLBO>. Consistently, we observed weak but significant migration defects when dp110^D945A was expressed using a stronger Gal4 driver (C306>; data not shown).

To further assess the role of the IIS pathway in border cell migration, we monitored the intracellular localization of the tGPH reporter gene (a GFP fused to the PH domain, binding to phosphoinositides) (Britton et al., 2002). In wild-type egg chambers, GPH is localized to the membranes of all the germ and follicle cells, with a stronger accumulation in germ cells (Fig. 2E) (Fulga and Rørth, 2002). We noticed a clear increase in GPH membrane localization in border cells compared with other follicle cells, starting at around stage 8-9 and persisting during cell migration, indicating higher activation of the IIS pathway in border cells (Fig. 2E-E″″).

This GPH expression pattern suggested that the IIS pathway is required in border cells both to initiate the detachment from the follicular epithelium and also during their migration. To test this hypothesis further, we took advantage of the temperature-inducible tub-Gal80^ts/Gal4 TARGET system (McGuire, 2003) (see Materials and Methods) to turn on the expression of P60 in border cells after they initiated their migration. Time-controlled conditional expression of P60 led to 47% of stage 10 egg chambers showing partial border cell migration compared with 7% in the absence of induction (Fig. 2F). Altogether, these results indicate that the canonical InR/Chico/PI3K/dAKT signalling pathway is required both for the detachment of the border cell cluster and throughout migration.

Border cells initiate their migration through the formation of a single long cellular extension (LCE) that promotes movement through a ‘grapple and pull’ mechanism (Fulga and Rørth, 2002). Because border cells mutant for the IIS pathway stall before they initiate migration, we tested a possible role of IIS in LCE formation on fixed tissues, using a lacZ reporter whose expression is under the control of the slbo regulatory region (slbo-lacZ) (Fulga and Rørth, 2002). At early stage 9, we observed cytoplasmic extensions in about 65% of the control clusters (Fig. 3A,E), with an average size of 11.5 µm (Fig. 3F), as previously reported (Colombié et al., 2017). Interestingly, inhibiting IIS through overexpression of P60 led to a reduction in the number and size of cytoplasmic extensions: only about 20% of early stage 9 clusters had an LCE (Fig. 3B,E), the size of which was much shorter with an average length of 3.9 µm (Fig. 3F). Furthermore, among the occasional P60-expressing clusters able to partially migrate, we observed that the protrusions were shorter and less frequent when compared with control (Fig. 3C,D).

To better assess the function of the IIS pathway in protrusion dynamics during border cell migration, we performed live time-lapse imaging of egg chambers in culture, with border cell clusters expressing mCD8::GFP, as previously described (Prasad and Montell, 2007). In most live control egg chambers (71%), border cells extended protrusions in the direction of migration, delaminated from the follicular epithelium around late stage 8/early stage 9, and migrated towards the oocyte over the course of 4-6 h (Movie 1). In contrast, expression of P60 (Movie 2) or FoxO (Movie 3) strongly inhibited border cell delamination and migration. Indeed, although cluster formation was normal, 83% of P60 and 80% of FoxO-expressing clusters had not yet delaminated at stage 10 (Movies 2 and 3). Notably, some occasional P60- or FoxO-expressing clusters were able to delaminate, allowing us to determine their migration speed. Whereas control clusters migrated at an average speed of 0.53 µm/min, P60- and FoxO-expressing clusters showed reduced velocity with an average speed of 0.22 µm/min and 0.32 µm/min, respectively (Fig. 3J).

We next quantified protrusion features and dynamics from live egg chambers (Fig. 3G-M; Movies 4-6). Fig. 3G-I show still images extracted from 1 h movies from control (Movie 4), P60 (Movie 5) or FoxO-expressing (Movie 6) border cells. Results show that the number, length and lifetime of protrusions are all reduced in conditions of reduced IIS signalling. The average number of protrusions/h is 11.8 in control border cells, which is reduced to 6.8 and 7.6 in P60- and FoxO-expressing clusters, respectively (Fig. 3K). These limited protrusions also show a reduction in their average length, which is 5.56 µm for P60 and 4.74 µm for FoxO-expressing clusters, compared with 9.74 µm in control border cells (Fig. 3L). Finally, although the average lifetime of protrusions reaches 6.05 min in control border cells, it is reduced to 4.32 min and 4.38 min in P60 and FoxO-expressing clusters, respectively (Fig. 3M).

The actin cytoskeleton plays a crucial role for membrane ruffling and protrusion dynamics during border cell migration (Prasad and Montell, 2007). Interestingly, we observed that loss of inr function in follicle or border cells disrupted the polymerization of the actin cytoskeleton as shown by a strong reduction of F-actin structures (Fig. 4A-B′″″). Similarly, P60-expressing clusters also show a reduction of F-actin levels when compared with control (Fig. 4C-E). These findings suggest that the IIS pathway is important for the polymerization of the actin cytoskeleton and for the formation of dynamic protrusions that are essential for driving border cell delamination and migration.

Actin polymerization largely relies on the Profilin protein to maintain pools of monomeric actin. Drosophila Profilin is encoded by the chickadee (chic) gene, whose loss of function leads to border cell migration defects (Verheyen and Cooley, 1994).

Interestingly, we found that Profilin is ubiquitously expressed in egg chambers with a stronger accumulation in border cells throughout their migration (Fig. 5A,B; Fig. S2). Clonal analysis using the chic^D5203 mutation confirmed the importance of Profilin for the formation of F-actin structures and for border cell migration, which are both absent in mutant cells (Fig. 5C,D) (Verheyen and Cooley, 1994).

To test for a possible functional interaction between chic/Profilin and IIS, we performed genetic, molecular and rescue experiments. First, we observed that removing one functional copy of the inr gene led to a strong enhancement of a chic border cell migration phenotype. Although flies harbouring a heteroallelic combination for chic mutations (chic¹³²⁰/chic^D5203) had weak migration defects (13% of mutant egg chambers not migrating normally), the phenotype was greater than 50% when the flies were heterozygous for the inr^ex15 mutation (chic¹³²⁰/chic^D5203; inr^ex15/+) (Fig. 5E,F).

Second, we found that IIS controls chic expression autonomously in both follicle and border cells, as shown by the strong reduction of Profilin protein accumulation in inr^ex15 mutant cells (Fig. 5G-I). In addition, these mutant cells showed a strong reduction of a chic-lacZ reporter line (Fig. 5J,K), indicating that the IIS pathway controls chic expression transcriptionally.

Finally, we asked whether this activation of chic expression represents a major response to Insulin signalling during migration, by testing the ability of chic to rescue a loss of IIS activity. Interestingly, chic overexpression was able to rescue inr-RNAi border cell migration defects by a factor of 6.25 (frequency of normal border cell migration raising from 4% to greater than 25%; Fig. 5F). The partial phenotypic rescue suggests that IIS likely controls other processes that are necessary for full border cell migration, in addition to controlling Profilin accumulation. Taken together, these results show that IIS controls the levels of chic/Profilin expression, which in turn is important for specific and dynamic F-actin structures that are essential for the migration of border cells.

To further characterize the molecular mechanism controlling chic expression, we looked at the role of the FoxO transcription factor, the activity of which is inhibited by AKT downstream of IIS activation (Junger et al., 2003; Puig et al., 2003). We first generated mosaic border cells for the foxo^Δ94 amorphic allele and found that mutant clusters migrated normally (Fig. 6A-A″), thus indicating that foxo is dispensable for migration. However, overexpression of either a wild-type or a constitutively active nuclear form of FoxO (UAS-hfoxO3a; Junger et al., 2003) in border cells led to severe migration defects, with about 60% to 100% of clusters not migrating, respectively (Fig. 6B,C). These results clearly indicate that FoxO activity suppresses border cell migration.

To further assess whether FoxO controls cluster migration downstream of IIS, we generated clones of border cells that overexpressed the P60 subunit and that were also mutant for foxo (see Materials and Methods). The P60-induced border cell migration phenotype was suppressed in the absence of foxo (Fig. 6D), indicating that IIS downregulation prevents border cell migration through FoxO.

We next analysed FoxO protein expression and found that the protein is present in both germline and follicle cells during oogenesis (Fig. 6E-I). Interestingly, detailed analysis revealed a dynamic expression pattern of FoxO in border cells, as follows: (1) before or at early stage 8, FoxO expression is similar in presumptive border cells and surrounding follicle cells; (2) starting at the time of cluster formation (stage 8-9), during (late stage 9) and until the end of migration, FoxO levels decrease constantly and become visibly undetectable (Fig. 6F-I). Quantification of FoxO signal intensity (ratio between border and neighbouring follicle cells) at these four different stages reveal a gradual reduction of FoxO protein during cell migration (Fig. 6E). When the border cell cluster reaches the oocyte (stage 10), the relative amount of FoxO has dropped by about 70% of that of stage 8 levels (Fig. 6E). Of note, the dynamic pattern of FoxO expression well mirrors the pattern of IIS activity, as revealed by tGPH staining (Fig. 2E-E″″).

The decrease of FoxO and the concomitant increase of Profilin in border cells suggest that FoxO could negatively control chic expression to promote migration. To test this hypothesis further, we analysed the levels of chic-lacZ in border cells following FoxO overexpression and found that the levels of this reporter line were strongly reduced (Fig. 6J-M). Additionally, we analysed chic mRNA expression following foxo overexpression and found that the levels of chic transcripts were also strongly reduced (Fig. 6N). Altogether, these results show that IIS-mediated inhibition of FoxO at the onset of migration is essential to allow chic expression in border cells, thereby promoting the assembly of dynamic actin-rich protrusions necessary for cluster delamination and migration (Fig. 6O).

In this study, we identify the Insulin/IGF-Signalling (IIS) pathway as a key regulator of border cell migration during Drosophila oogenesis. We demonstrate that activation of InR at the onset of migration promotes actin dynamics in the outer border cells, the subpopulation of cells known to drive migration. In this process, the canonical IIS pathway is shown to act through the inhibition of the transcription factor FoxO, which leads to the de-repression of chic/profilin. High levels of Profilin in turn facilitate actin polymerization and the formation of dynamic protrusions and of specific, long actin cellular extensions that are required for delamination and proper migration of the invasive cell cluster (Fig. 6O).

The conserved IIS pathway couples nutritional cues with cellular metabolism, which in turn is essential for coordinating development with growth conditions. The systemic action of the IIS pathway thus makes it difficult to discriminate between chronic versus more acute or specific roles in particular cellular processes and during morphogenesis. In this context, border cells provide a powerful model with which to specifically address the role of the IIS pathway on cellular motility. During Drosophila oogenesis, the IIS pathway acts both in the germline and somatic cells to adjust egg chamber maturation rates to protein availability (Drummond-Barbosa and Spradling, 2001, 2004; Ikeya et al., 2002; Jouandin et al., 2014; LaFever, 2005). We show, using the FLP/FRT system, that chronic downregulation of IIS in border cells impairs their migration, a process that can be associated with metabolic defects. Interestingly, our acute manipulation of IIS in border cells, using the Gal4/Gal80^ts system, shows that IIS downregulation can also block cluster migration specifically, a phenotype that can be rescued partly by restoring Profilin expression (Fig. 5). These data argue for an active control of cell migration by IIS, independently of cellular fitness. This view is consistent with previous work showing that in ex vivo experiments, Insulin-containing culture medium is necessary to support egg chamber development and border cell migration (Bianco et al., 2007; Prasad and Montell, 2007).

Border cells migrate towards the oocyte to make the micropyle: an opening that allows oocyte fertilization through the chorion. In this process, border cell migration needs to be synchronized with oocyte growth. We propose that the dual role of IIS for both egg chamber growth and border cell migration could help to coordinate migratory events with organ maturation, thereby ensuring the robust morphogenesis that is important for fertility.

Actin dynamics are essential to a multitude of cellular and morphogenetic processes; therefore, understanding the diverse modes of actin regulation is of prime interest. Members of the IIS pathway have been linked to actin regulation in a number of normal and pathological processes (Xue and Hemmings, 2013). For example, IIS plays an important role in neuronal guidance (Liu et al., 2014; Song, 2003) or wound healing (Kakanj et al., 2016). Additionally, PI3K has been shown to couple glycolytic flux with actin dynamics (Hu et al., 2016), whereas AKT participates in the epithelial-to-mesenchymal transition required to drive mesoderm formation during gastrulation (Montero et al., 2003; Yang et al., 2008). Accumulating evidence also indicates that PI3K/AKT controls the migratory phenotype of metastatic cells (Xue and Hemmings, 2013). In breast cancer cells, AKT enhances cell migration and invasion through increased filopodia formation, which can be blocked with a specific AKT inhibitor (Yang et al., 2004). These observations suggest a model in which AKT activation potentially influences cell motility through direct modulation of actin, which is supported by studies showing that actin preferentially binds to phosphorylated AKT at pseudopodia sites (Amiri et al., 2007; Cenni et al., 2003). Despite this evidence, the view is fragmented and data are lacking that demonstrate a clear role for the full canonical pathway in cytoskeleton plasticity. In particular, the requirement of IIS transcriptional regulation in this process remained elusive. In this report, we reveal that canonical IIS acts through inhibition of the transcription factor FoxO to control a major actin regulator: Profilin. These data reveal a molecular mechanism for FoxO-mediated control of actin remodelling, which may be generalized to other processes where actin dynamics is particularly important. For example, during wound healing in Drosophila larvae, formation of an acto-myosin cable has been shown to depend on PI3K activation and redistribution of the transcription factor FoxO (Kakanj et al., 2016).

In conclusion, our findings establish the canonical IIS pathway as a gene regulatory network important for collective cell migration (Sharma et al., 2018). The data also identify a novel mechanism by which actin homeostasis and organization is regulated transcriptionally in a dynamic migratory process. By this mechanism, the formation of actin-rich protrusions is constitutively and negatively controlled by the transcription factor FoxO, the inhibition of which by IIS signalling can generate peak levels of actin polymerization required for delamination and migration. It will be interesting to establish whether the control of Profilin expression through IIS signalling represents a general mechanism that controls actin remodelling in cell and tissue morphogenesis.

Drosophila culture and crosses were performed following standard procedures at 25°C, except where indicated. The following Drosophila strains were used: inr-RNAi (National Institute of Genetics, NIG-Fly); akt-RNAi and chico-RNAi (VDRC, Vienna Drosophila Resource Center) (Dietzl et al., 2007); UAS-foxo and UAS-dp110^CAAX (Bloomington); slbo-Gal4 (Rørth et al., 1998); upd-Gal4 and slbo-Gal4, UAS-mCD8 GFP (a gift from D. Montell, University of California, Santa Barbara, USA); slbo-Gal4, tGPH and slbo-Gal4, slbo-flac (referred to as slbo-lacZ) (Fulga and Rørth, 2002); UAS-hfoxO3a (Junger et al., 2003); chic^D5203FRT40A, chic¹³²⁰ and UAS-chic (a gift from L. Cooley, Yale University, New Haven, USA); UAS-P60, UAS-PTEN (a gift from P. Léopold, Institut de Biologie Valrose, Nice, France); inr^ex15 FRT82B (Song, 2003) (a gift from L. Pick, University of Maryland, USA); dp110^1C1FRT82B (a gift from H. Stöcker, ETH, Zurich, Switzerland; Willecke et al., 2011); foxO^Δ94 FRT82B (Slack et al., 2011); and akt¹ FRT82B (Rintelen et al., 2001).

For expression induced using the Gal4/UAS system, 2-day-old females with the designated genotypes were incubated for 2 days at 29°C before dissection. w¹¹¹⁸ flies were used as control.

Mutant follicle cell clones were generated by mitotic recombination using the FLP/FRT technique (Xu and Rubin, 1993). Flies with mutations on FRT82B chromosomes were crossed with hsFLP;; FRT82B, UbiGFP flies. chic^D5203FRT40A flies were crossed with hsFLP; FRT40A, UbiGFP flies.

UAS-P60; FRT82B foxo^Δ94 and UAS-P60; FRT82B flies were crossed with yw, tubGal4-UAS GFP; FRT82B, tub Gal80^ts flies (a gift from A. Ephrussi, EMBL, Heidelberg, Germany) to generate MARCM clones (Lee and Luo, 1999).

Mosaic clones were generated as follows: females with the required genotypes were heat-shocked for 1 h at 37°C, twice a day for 3 days and dissected 2 days later. Mosaic mutant clones were marked by the absence (FLP/FRT) or by the presence of GFP (MARCM).

Flip-out clones were made from UAS-foxO flies crossed to hsFLP;;act<CD2>GAL4, UAS-GFP (a gift from D. Montell). The progeny was heat-shocked for 1 h at 37°C and dissected 2 days later.

Conditional expression of UAS-P60 was achieved using the TARGET system (McGuire, 2003), combining the UAS/Gal4 bipartite expression system with a thermosensitive form of the Gal4 negative regulator, Gal80^ts, to switch on and off expression of the UAS-construct at desired developmental time points.

Briefly, the tub-Gal80^ts; slbo-GAL4>P60 flies were obtained and raised at 18°C for 1 day. At this restrictive temperature, Gal80^ts is active and blocks Gal4 function, leading to an absence of UAS-P60 expression. Flies were then raised for 3-4 h at permissive temperature (30°C), leading to Gal80^ts inactivation and hence allowing Gal4 to activate UAS-P60 expression and consequently IIS pathway inhibition in border cells. Flies were dissected immediately after temperature shifts.

This screen was performed by crossing the NIG-Fly UAS-RNAi transgenic fly collection with the USG> driver (a combination of UPD> and SLBO>), which allows the targeted expression of UAS constructs in both polar and outer border cells, together with a UAS-GFP transgene allowing the visualization of the border cell cluster (De Graeve et al., 2012). USG>UAS-RNAi females were dissected after 2 days of incubation at 29°C, and candidate UAS-RNAi lines were selected when abnormal border cell migration was observed in more than 20% of stage 10 egg chambers.

Ovary dissection, fixation and staining with antibodies, phalloidin and DAPI were performed as described previously (Devergne et al., 2007; Ghiglione et al., 2008). The primary antibodies used were: rabbit anti-Slbo (1:1000; De Graeve et al., 2012); mouse anti-Singed [1:50, SN7C, Developmental Studies Hybridoma Bank (DSHB), University of Iowa, USA]; mouse anti-FasIII (1:100, 7G10, DSHB); mouse anti-Profilin (1:100, DSHB); rabbit anti-FoxO (1:500; Shin et al., 2011; Slaidina et al., 2009); and chicken anti-β-Galactosidase (1:1000, Gene Tex).

Alexa Fluor-conjugated goat antibodies were used as secondary antibodies (Molecular Probes). Phalloidin conjugates and DAPI (1:1000, Sigma) were used to outline cells and label DNA, respectively. The samples were mounted in Mowiol and images were taken on a Zeiss LSM510 or LSM710 confocal microscope. Quantification of protrusion length was performed using the ImageJ software.

Imaging of live border cell migration was performed using an Olympus spinning-disc confocal microscope (20×objective+1.5×lens) coupled with an Andor camera. Acquisition was carried out using Metamorph software. Egg chambers were cultured in S2 culture medium with 1 µg/ml Insulin at 22°C and processed as described previously (Prasad and Montell, 2007). Movies were captured over a period of 1 h (1.30 min. interval) or 8 h (10 min. interval). Cluster velocity and protrusion length, frequency and dynamics were analysed on Slbo-Gal4>UAS-CD8::GFP (control) and Slbo-Gal4>UAS-CD8::GFP>UAS-P60 or Slbo-Gal4>UAS-CD8::GFP>UAS-FoxO flies. Analysis was carried out using ImageJ software.

Total RNA extraction from dissected ovaries and real-time RT-PCR for chickadee were performed as previously described (Slaidina et al., 2009). Three independent biological experiments each with a triplicate measurement were conducted. Primers (left 5′-ctgcatgaagacaacacaagc-3′ and right 5′-caagtttctctaccacggaagc-3′) were designed using the Primer Express software (Applied Biosystems) and tested by standard curve experiments.

We thank all members of S.N.’s laboratory for fruitful discussions; members of P. Léopold's laboratory for providing numerous reagents and helpful advice; A. Popkova for teaching us time-lapse imaging of egg chambers in culture; L. Cooley, A. Ephrussi, P. Gallant, E. Hafen, D. Montell, W. J. Lee, L. Partridge, L. Pick, P. Rorth and H. Stöcker for reagents; the DSHB for antibodies; the Bloomington Drosophila Stock Center, the National Institute of Genetics Fly (NIG-Fly) and Vienna Drosophila RNAi Center (VDRC) for providing Drosophila fly lines. We thank F. Bondi, D. Chobert, F. De Graeve, M. A. Derieppe, G. Gozzerino, L. Parel, M. Pierret, A. Samuel and J. Soltys for their help with the UAS-RNAi screen; and the iBV PRISM platform for providing state of the art imaging resources and advice. Some data are reproduced from P.J.’s PhD thesis, defended at Institut de Biologie Valrose in 2013.