Insulin signaling modulates border cell movement in Drosophila oogenesis

Molecular mechanisms underlying collective cell migration are being investigated in different contexts as it constitutes an integral component of metazoan development (Friedl and Gilmour, 2009). Border cell migration in Drosophila oogenesis has emerged as an excellent ‘ex-vivo’ model system for studying collective cell movement. Drosophila ovaries consist of multiple egg chambers, which go through 14 stages of development to form the mature egg. Stages 1-7 are pre-vitellogenic, whereas stages 8-10 are post-vitellogenic. An egg chamber consists of 16 germline cells (a single oocyte and 15 nurse cells) surrounded by a layer of ∼750 somatic cuboidal follicle cells (Spradling, 1993). The pre-vitellogenic to vitellogenic phase transition is coordinated with an increase in the size of the developing oocyte and transformation of a small group of 6-10 anterior follicle cells (AFCs) into border cells (BCs) (Silver and Montell, 2001). In response to the chemoattractants secreted by the oocyte, the BCs activate the receptor tyrosine kinases, which stimulate the small GTPases Rac and Ras to modulate cluster detachment and guide its movement towards the oocyte (Montell, 2006). Upon reaching the oocyte, BC cluster forms a channel in the eggshell for sperm entry during fertilization (Montell et al., 1992).

During oogenesis, Insulin signaling (InS) allows the transition from the pre-vitellogenic to vitellogenic phase and modulates mitotic-to-endocycle transition of follicle cells (Jouandin et al., 2014; Richard et al., 2005). Binding of one of the eight Drosophila Insulin-like peptides (Ilps) to the insulin receptor (InR) activates the insulin receptor substrate Chico. Activated Chico stimulates phosphatidylinositol-3-kinase (PI3K), which further modulates several effector pathways through various intermediates, including the Protein kinase B/Akt (Garofalo, 2002; Lizcano et al., 2003; Teleman, 2010). Akt phosphorylates and inhibits the transcriptional repressor FOXO, which downregulates the canonical InS targets (Brunet et al., 1999). Moreover, PI3K/Akt function is known to be required for generation of the membrane ruffles in the migrating cells (Doughman et al., 2003; Henderson et al., 2015).

As border cell specification and migration occur concomitantly with the pre-vitellogenic to vitellogenic phase transition, we addressed whether InS influences the movement of BC cluster. Here, we show that canonical InS is required for BC movement. Our live-cell imaging analysis suggests that InS influences the detachment of the cluster from the follicular epithelium and, also controls the protrusive behavior of BCs. Based on these data we propose that InR likely functions through PAR-1 to alter the activity of Drosophila Myosin II [encoded by spaghetti squash (sqh)], which is crucial for the efficient detachment and forward movement of the BC cluster.

To assess the potential involvement of InS during BC development, we first examined the distribution of a pleckstrin homology-GFP fusion protein, tGPH, a reliable marker of phosphoinositide 3-kinase (PI3K) activity, shown to respond specifically to InS (Britton et al., 2002). PI3K is targeted to the membrane upon activation of InS. Curiously, we observed enhanced lateral membrane localization of the tGPH reporter in the AFCs of post stage 7 egg chambers. This pattern was retained in the migrating BCs, which are specified from the AFCs (Fig. 1A-E).

Next, we downregulated InS in the BCs by overexpressing a dominant-negative construct of InR (K1409A), using the c306 Gal4 driver that is expressed in only the BC cluster and surrounding AFCs (Manseau et al., 1997). Satisfyingly, downregulation of InR function inhibited the BC movement in 31.78±2.32% of stage 10 egg chambers, compared with 9.60±2.05% of controls (Fig. 1F-H). It also affected tGPH localization between the InR-depleted border cells in the migrating cluster (Fig. S1). Similarly, overexpression of InR^RNAi with c306 Gal4 driver inhibited BC movement (Fig. S2) and also affected the levels of the dual specificity phosphatase puckered, a known downstream target of FOXO (Bai et al., 2013) (Fig. S3). However, the BC fate was unaltered when InS was downregulated in the anterior follicle cells (Kang et al., 2018). To substantiate our observations, we evaluated migration efficiency of the mosaic BC clusters generated with a putative null allele, inr³³⁹ by employing the mosaic analysis with a repressible cell marker (MARCM) technique (Lee and Luo, 2001). BC clusters with inr³³⁹ mutant cells exhibited significant migration defects compared with the controls (Fig. 1I-J). Specifically, clusters with only one or two mutant cells migrated normally and covered the same distance as the wild type. By contrast, BC clusters with more than two inr³³⁹ mutant cells exhibited significant migration defects (Fig. 1I-J). Confirming the specificity, we did not observe any migration defects in egg chambers (n=972) that lacked inr³³⁹ mutant cells (data not shown). Our data thus suggest that InR signaling functions in the BCs to mediate their efficient movement. As reported previously, the size of the individual inr³³⁹ cells was smaller than the corresponding wild-type cells from the same cluster (Brogiolo et al., 2001).

Typically, Insulin functions through Chico, PI3K and Akt to repress FOXO activity, in turn, to regulate expression of downstream target genes (Böhni et al., 1999; Chang et al., 2005; Coelho et al., 2005). To assess whether canonical InS is involved in BC migration, we generated chico¹ mutant BCs using the MARCM technique and analyzed the migration efficiency of the cluster. Similar to inr³³⁹, BC clusters with multiple chico¹ mutant cells (>2) exhibited migration defects (Fig. 2A,B). This suggests that a downstream component of InS, Chico, is also required for BC migration.

If InS functions through its canonical effectors to modulate BC migration, overexpression of the known downstream mediators should rescue the migration defects observed in InR-depleted clusters. Accordingly, overexpression of wild-type PI3K (PI3K92E) or Akt (Akt1) in the InR-depleted clusters exhibited substantially lower migration defects (11.95±1.05% and 6.79±1.25%, respectively) when compared with that observed in the InR^DN-overexpressing clusters (31.71±1.92% and 32.95±2.43%, respectively) (Fig. 2C-G).

Conversely, simultaneous depletion of a negative regulator of InS, FOXO, mitigated the phenotype. Migration efficiency of InR-depleted clusters in the wild-type background (33.22±3.21%) was significantly rescued in the foxo^Δ94 heterozygous background (6.01±0.95%) (Fig. 2H). Moreover, overexpression of FOXO in the AFCs inhibited BC movement (Fig. S4). Altogether, our data suggest that canonical InS functions in a cell-autonomous manner during BC movement.

Typically, BC movement is affected by several factors, including the number of cells in the migrating cluster, protrusion dynamics and timely detachment of the cluster from the remaining follicular epithelium. We therefore wondered which of these traits are affected in InR-depleted clusters. Interestingly, our MARCM analysis also revealed that InR activity might be required not only in the BC cluster but also in the surrounding AFCs. Supporting this conclusion, in several instances (n=91) we observed BC migration defects, when only the AFCs were mutant for inr³³⁹ (Fig. 3A-B). Importantly, this observation was also recapitulated in chico¹ MARCM analysis (data not shown).

BC migration was also disrupted even when BC cluster itself was wild type, indicating a possible role for InS during the physical attachment between the two cell types, i.e. AFCs and BCs. An early step during successful BC migration is the detachment of the cluster from the AFCs. We therefore wondered whether the BC migration defects due to cell-autonomous loss of InR are induced by delayed detachment between the two cell types.

We employed live-cell image analysis to assess how the dynamics of cluster movement differs between the wild-type (control) and InR-depleted clusters. We simultaneously overexpressed the UAS-GFP Moesin (UAS-GMA) to mark the migrating clusters in the wild-type and InR-depleted clusters to capture several movies of the undetached cluster (Fig. 3C,D) (Movies 1-3) (Dutta et al., 2002; Geisbrecht and Montell, 2002). In wild-type controls (n>6), the detachment of border cell cluster was observed around late stage 8 or early stage 9, concomitant with the initiation of cuboidal to squamous cell shape change in the neighboring AFCs. By contrast, in the InR-depleted background, though the BC cluster formation was normal, its detachment from the AFCs was delayed. In some instances (n=13), the ‘InR-depleted’ clusters were still attached to the AFCs despite the progression through stage 10. In the remaining cases, there was a significant delay in detachment and the cluster severed from the AFCs around late stage 9, by the time 75-80% of the AFCs were flattened (Movie 2). Overall, our MARCM analysis coupled with live-cell imaging suggested that InR modulates the detachment of the cluster from the AFCs.

Next, we compared the migration rates of InR-depleted clusters with that of control. To enhance border cell detachment, female flies were incubated at 25°C instead of 29°C for 14-16 h before imaging the egg chambers. As expected, the control clusters moved at an average speed of 0.47 µm/min (n=20), whereas the clusters depleted for InR, migrated more slowly, with an average speed of 0.26 µm/min (n=20) (Fig. 3E, Movie 3).

As altered protrusive behavior influences the speed of the migrating clusters, we mapped and quantified the parameters of the protrusions, including total number, length, direction and lifetime for the entire duration of movement for the control and InR-depleted clusters. Our analysis suggested that the protrusions formed in the InR-depleted clusters were misdirected. In the control clusters, significant proportion of protrusions were formed at the migrating front (∼54%), whereas the InR-depleted clusters exhibited a lesser proportion of frontal protrusions (∼34%) and a corresponding enhancement in the sideward and backward protrusions (Fig. 3H). Furthermore, the lifetime of protrusions was significantly lower in the InR-depleted clusters (4.8 min) compared with the control (7.3 min) (Fig. 3G). Altogether, the time-lapse imaging analysis suggested that InR is required for polarizing the migrating cluster by modulating the migration and protrusion dynamics of the moving cluster.

Since the formation of the protrusions depends upon actively nucleating F-actin, we stained the InR-depleted clusters with rhodamine phalloidin. In the control clusters, rhodamine staining was detected around the rim of the BCs with several actin fibers within the migrating cluster. Overexpression of InR^RNAi in the BCs exhibited considerably reduced number of actin fibers with a corresponding increase in large number of actin puncta within the cluster. This suggests that InR modulates the actin levels in BCs, which possibly regulate polarized protrusion and efficient movement (Fig. 3I-K). Altogether, these results uncover the temporally separable dual function of InS during BC migration. First, it regulates the detachment of the cluster from the anterior follicular epithelial cells and, subsequently, it modulates forward movement by regulating the protrusion dynamics.

Next, we sought to determine the possible mechanistic basis of the phenotypes induced by the cell-autonomous loss of InS. In this regard, an interesting candidate is PAR-1, a serine-threonine kinase. PAR-1 localizes to the basolateral surface of the epithelial follicle cells and has been shown to regulate both the detachment of the BC cluster and generation of polarized protrusions in the migrating cells (McDonald et al., 2008). As we observed similar phenotypes in the InR-depleted clusters, we wondered whether par-1 and InR regulate BC migration in a cooperative manner. As an initial test of this idea, we evaluated the migration efficiency of the clusters overexpressing PAR-1^RNAi in the wild-type and inr³³⁹ heterozygous background (Fig. 4A-C). Overexpression of PAR-1^RNAi in the wild-type AFCs resulted in migration defects (30.19±2.24%) with a small fraction of BCs lingering at the anterior end (3.14±0.91%) of the egg chambers. Downregulation of par-1 function in the inr³³⁹ heterozygous background exhibited slightly higher (45.74±3.50%) migration defects and more than 15±3.33% of the egg chambers displayed undetached clusters (Fig. 4D). This enhancement in the number of undetached border cell clusters suggests that InR and par-1 genetically interact and together may potentiate the severing of the BC clusters from the follicular epithelium.

Recently, Bai et al. have shown that FOXO represses par-1 levels in adult flies (Bai et al., 2013). Hence, we sought to examine the level and distribution of PAR-1 protein in inr³³⁹ mutant follicle cells. We detected a modest change in the distribution of PAR-1 at the lateral interface between mutant cells (Figs S5,S6). In the follicle cells overexpressing InR^RNAi, we observed a higher cytoplasmic distribution of PAR-1 in 53% of the egg chambers (n=13; Fig. S7) compared with the adjacent control cells.

Altogether, the genetic interaction analysis between PAR-1^RNAi and inr³³⁹ followed by an examination of PAR-1 distribution in InR-depleted follicle cells suggest that InR may regulate the level/distribution and/or activity of PAR-1 in the follicle cells to facilitate detachment and subsequent forward movement of the BC cluster.

As par-1 has been shown to potentiate Myosin II activity, we wondered whether increasing the Myosin II activity could partially suppress the migration defect associated with InR downregulation (Majumder et al., 2012). Myo II consists of three subunits: myosin heavy chain (Zipper), myosin light chain and the myosin regulatory light chain (Sqh) (Karess et al., 1991). Rho-associated kinase (Rok or ROCK) and myosin light chain kinase (MLCK) independently phosphorylate Sqh at Thr18/Ser19 (Thr20/Ser21 in Drosophila) to potentiate Myosin activity by stabilizing Myosin-containing filaments (Edwards and Kiehart, 1996; Jordan and Karess, 1997). sqh^E21 is a gain-of-function transgene, in which serine 21 is replaced by glutamate; it can rescue oogenesis defects in egg chambers containing sqh^AX3 germline clones (Edwards and Kiehart, 1996; Jordan and Karess, 1997). We therefore assessed whether the migration defect induced by InR-depletion could be ameliorated by sqh^E21 transgene. We observed lower migration defect of clusters overexpressing the InR^DN construct in sqh^E21 background (13.53±1.78%) compared with that in the wild-type background (37.47±3.6% of the egg chambers) (Fig. 4H). sqh^E21 alone exhibited minimal migration defect of 9.09±1.17% (Fig. 4H). This suggests that egg chambers carrying a sqh^E21 transgene rescue the migration defect of InR-depleted clusters.

In contrast to ROCK and MLCK kinases, Myosin-binding subunit (Mbs) phosphatase dephosphorylates Sqh to reduce the Myosin activity. A similar reduction in the extent of migration defects was observed when Mbs^RNAi was co-expressed with InR^DN in the BCs (15.33±2.09% versus 41.24±4.48%) (Fig. 4I) (overexpression of Mbs^RNAi alone inhibits migration in 6.58±1.66% egg chambers; Fig. 4I). Altogether, these data are consistent with the possibility that InR regulates Myo II function via par-1 in the BCs to assist their efficient detachment and movement (Majumder et al., 2012; McDonald et al., 2008).

Collective, as well as individual, cell migration are known to be regulated by a number of signaling cascades in different developmental contexts (Gaggioli et al., 2007; Geisbrecht and Montell, 2002; Schmidt et al., 2007; Wolf et al., 2007; Zaidel-Bar et al., 2006). As Insulin has been shown to affect the movement and/or invasion of cells in different cell culture settings, in the present study we attempted to examine the role of InS in collective cell movement employing the ex vivo model of BC migration in Drosophila oogenesis (Wang et al., 2017; Wei et al., 2017). The loss-of-function analysis coupled with live-cell imaging suggests that InS regulates the BC movement first by allowing BCs to detach from the AFCs and then by engineering polarized protrusions for a progressive movement.

Although our data document involvement of InR during BC migration, how InR controls activation of PAR-1 in the migrating BCs remains elusive. We believe that, post-specification, InS plays an additional important role in the detachment of the BC cluster (Fig. 4J). However, we cannot rule out the possibility of other targets downstream of InS in BC behavior, especially in light of the report from Kang et al. and our genetic interaction data with the Ecdysone receptor (Fig. S8) (Kang et al., 2018). Moreover, the potential source and precise identity of the Ilps that regulate the InS also remains unknown. Recent studies elucidating a non-autonomous ‘remote control’ mechanism to achieve activation of InS may prove highly instructive in this regard (Rajan and Perrimon, 2012).

All crosses were carried out at 25°C. The fattening and incubation were performed at 29°C, for a time period of 24-26 h. InR depletion experiments employed c306 Gal4 (Brand and Perrimon, 1993) to drive the expression of UAS InR^DN (BL-8251/ BL-8253) in BC clusters. Insulin/PI3K activity was assessed using the tGPH indicator (BL-8163). Mosaic analysis was carried out using inr³³⁹ (Brogiolo et al., 2001), a loss-of-function allele. MARCM ready line hsFLP; UAS GFP; neoFRT82B tubP-Gal80 was used for generating inr³³⁹ homozygous mutant clones. Heat shocks were given for 3 days, three times per day at an interval of 3 h. For mosaic analysis for chico¹ (Böhni et al., 1999), an amorphic allele, MARCM ready line hsFLP mCD8GFP; (BL-5192; tubP GAL4)/CyO:TM6Be [derived from BL 42725] was used. c306-Gal4 was used for MARCM analysis and progeny flies were fattened at 25°C for 16-18 h, after 2 days of heat-shock regime. Genetic interaction tests between Insulin signaling components in a InR-deficient background were performed using the following stocks: c306 Gal4, UAS Akt1 (BL-8191), UAS Pi3K92E (BL-8286), EcR³¹/ Cyo (BL-4900), UAS Mbs^RNAi (BL- 32516), UAS FOXO GFP (BL-43633), foxo^Δ94 (Slack et al., 2011) (a kind gift from Dr Rakesh Mishra, Centre for Cellular and Molecular Biology, Hyderabad, India), sqh^E21 (BL-25763), UAS PAR-1^RNAi (a kind gift from Dr Majumder, Presidency University, Kolkata, India), puc-LacZ (Ring and Martinez-Arias, 1993) (a kind gift from Prof. L. S. Shashidhara, Indian Institute of Science Education and Research, Pune, India). As controls, UAS GFP, UAS mCD8 GFP and Canton S were used. The flies were procured from Bloomington Drosophila Stock Center (BDSC). UAS InR^RNAi (18402Ra-2) lines were procured from the National Institute of Genetics (NIG, Japan).

Ovaries of 2- to 3-day-old female flies were dissected in Schneider's medium supplemented with fetal bovine serum. Flies were fixed in 4% paraformaldehyde solution (Sigma) for 15 min. Blocking solution was made using 5% bovine serum albumin, 0.3% Triton-X-100 in phosphate-buffered saline (Sigma). PAR-1 antibody staining was carried out as reported earlier (McDonald et al., 2008). Primary antibodies were used at the following concentrations: mouse anti-Armadillo (N2 7A1 at 1:120; Developmental Studies Hybridoma Bank, DSHB), anti-β-galactosidase (40-1a at 1:100; DSHB), rabbit anti-GFP (A11122 at 1:500; Thermo Fisher Scientific) and guinea pig anti-PAR-1 (at 1:10; a kind gift from Dr Jocelyn McDonald, Kansas State University, Manhattan, KS, USA). Secondary antibodies were used at the following concentrations: Alexa Fluor 488 at 1:250 (Molecular Probes, Life Technologies) and Alexa Fluor 568 at 1:400 (Molecular Probes, Life Technologies). Rhodamine phalloidin staining was carried out at 1:500 dilution. DAPI (Sigma) was used to stain nuclei (0.05 µg/ml). The quantitation, imaging and analysis were carried out using an Olympus IX-81 Epi-fluorescence Microscope and Zeiss Apotome and Zeiss (LSM 710) confocal microscope.

For the assessment of Insulin/PI3K activity, tGPH flies were incubated at 29°C, for 16-18 h. The flies were dissected in 1× phosphate-buffered saline and fixed overnight with 4% paraformaldehyde at 4°C. Imaging was carried out using a Zeiss LSM 710 confocal microscope.

Time-lapse imaging of BC migration was performed using c306 Gal4; UAS GMA, as control, and c306 Gal4; UAS InR^DN/UAS GMA, as experimental, genotypes. Time-lapse experiments were performed as described previously (Prasad et al., 2007). Each frame was taken at an interval of 2 min. Both the control and experimental movies are captured for 3-4 h, on average. Imaging was carried out using an Olympus IX-81 Epi-fluorescence microscope.

The speed of the migrating cluster was calculated only for the duration of its movement. The length of the protrusions was measured by drawing a line from the center of the cluster to the tip of the extending protrusion. The extensions greater than the radius of the cluster were considered as protrusions. The direction of the protrusion with respect to the moving cluster was evaluated by measuring the angle from x-axis in a two-dimensional plane. The measured angles of protrusion were categorized into front (0-45° and 0-315°), side (45-135° and 225-315°) and back (135-225°). The number of frames during which the protrusion lasted was used to determine the lifetime of protrusions. The analysis was carried out using ImageJ software.

A two-tailed t-test of unequal variance was employed to evaluate the significance of the experiments, with their respective controls. The level of significance is classified into: *P=0.01-0.05, **P<0.01 and ***P<0.001.

The numbers of actin puncta and fibers were calculated from the z section of the BC clusters imaged at 63× magnification in the LSM 710 confocal microscope. Precaution was taken to avoid overlapping puncta and fibers in z sections. The fibers greater than 0.09 μm in length were considered for analysis. At least five BC clusters were analyzed for control and experiment to calculate the average number of actin puncta and fibers.

In order to quantify levels of puc-LacZ in the cluster, the whole border cell cluster was captured with Apotome (Zeiss) using identical camera settings (exposure was 1600 ms) for both control and experimental clusters. The mean intensity of puc-LacZ per nuclei was calculated from maximum intensity projections made using the Zen Blue software. Data are the average of the total number of nuclei considered for intensity calculation±s.e.m.

While this work was in progress, InS signaling was shown to modulate actin dynamics to influence border cell migration (Ghiglione et al., 2018).

We thank Drs Pralay Majumder, Jocelyn McDonald and Rakesh Mishra, and Prof. L. S. Shashidhara for providing the crucial fly stock lines. We thank Bloomington Drosophila Stock Center (BDSC), the National Institute of Genetics (NIG, Japan) and Developmental Studies Hybridoma Bank (DSHB) for providing reagents. We thank IISER Kolkata imaging facility and, in particular, Ritabrata Ghosh for capturing the images in the confocal LSM 710.