Dynamic myosin activation promotes collective morphology and migration by locally balancing oppositional forces from surrounding tissue

A challenge for migrating collectives is to respond to physical changes in local environments. Border cells migrate collectively in the Drosophila ovary and require dynamic myosin to maintain their morphology. Border cells elevate active myosin in response to tissue compression. Myosin tension counteracts tissue constraints for collective movement.

Abstract

Migrating cells need to overcome physical constraints from the local microenvironment to navigate their way through tissues. Cells that move collectively have the additional challenge of negotiating complex environments in vivo while maintaining cohesion of the group as a whole. The mechanisms by which collectives maintain a migratory morphology while resisting physical constraints from the surrounding tissue are poorly understood. Drosophila border cells represent a genetic model of collective migration within a cell-dense tissue. Border cells move as a cohesive group of 6−10 cells, traversing a network of large germ line–derived nurse cells within the ovary. Here we show that the border cell cluster is compact and round throughout their entire migration, a shape that is maintained despite the mechanical pressure imposed by the surrounding nurse cells. Nonmuscle myosin II (Myo-II) activity at the cluster periphery becomes elevated in response to increased constriction by nurse cells. Furthermore, the distinctive border cell collective morphology requires highly dynamic and localized enrichment of Myo-II. Thus, activated Myo-II promotes cortical tension at the outer edge of the migrating border cell cluster to resist compressive forces from nurse cells. We propose that dynamic actomyosin tension at the periphery of collectives facilitates their movement through restrictive tissues.

INTRODUCTION

Cells frequently move in interconnected groups, termed collectives, to form and shape organs during development, remodel vessels during angiogenesis, and heal wounds (Friedl and Gilmour, 2009). Many types of cancer cells also invade as collectives, greatly contributing to tumor progression (Friedl et al., 1995, 2012; Alexander et al., 2013; Bronsert et al., 2014; Clark and Vignjevic, 2015). The ability of cells to migrate in vivo, as single cells or as collectives, is influenced by the architecture of the three-dimensional tissue, whether it is loose or dense extracellular matrix (ECM) or even other cells (Friedl and Wolf, 2010; Wolf and Friedl, 2011; Wolf et al., 2013; Clark and Vignjevic, 2015). Cells moving within tissues dynamically respond to physical changes in the local environment through deformability and adaptability of the cell itself (Friedl and Wolf, 2010; Doyle et al., 2013; Wolf et al., 2013). Cells in collectives often arrange into specific overall shapes, such as clusters, strands, or sheets, which need to be maintained for optimal movement of the entire group (Friedl and Gilmour, 2009; Vedula et al., 2012; Choi et al., 2013; Leong et al., 2013). Unlike individually migrating cells, cells in a collective must also stay together to coordinate their response to the microenvironment as a single unit. Because it is technically difficult to recreate the composition of tissues in vitro, the mechanisms that influence the maintenance of collective shape and organization in vivo and their effect on motility are still poorly understood.

Drosophila border cells are a relatively simple and genetically tractable model in which to address how collectives withstand the impact of surrounding tissue, stay together, and migrate within a native microenvironment. Border cells undergo a guided collective migration during late oogenesis (Montell et al., 2012). The major subunit of the ovary is the egg chamber, which consists of a somatic, monolayer, follicular epithelium surrounding the germ line–derived nurse cells and the large oocyte at the posterior (Spradling, 1993) (Figure 1A). The polar cells, a specialized pair of follicle cells located at the anterior end of the egg chamber, secrete a cytokine signal to activate JAK/STAT signaling in adjacent follicle cells (Silver and Montell, 2001; Beccari et al., 2002; Ghiglione et al., 2002). The four to eight follicle cells with highest levels of JAK/STAT activation surround the polar cells and become migratory border cells. The border cell cluster then detaches from the epithelium and navigates between the 15 large nurse cells to eventually reach the oocyte (Figure 1A). Importantly, border cells stay together as a group during their entire migration. The cell polarity proteins Par-3 and Par-6, along with Jun kinase (JNK) signaling, promote cohesion of border cells and prevent individual cells from pulling away from the main group (Pinheiro and Montell, 2004; Llense and Martín-Blanco, 2008; Melani et al., 2008). E-cadherin–based adhesion between border cells is critical to mechanically couple cells within the cluster (Cai et al., 2014). Such adhesion induces front-to-back tension within the border cell cluster, allowing the group to respond to guidance cues and migrate directionally to the oocyte. E-cadherin also organizes the cluster by promoting adhesion of individual border cells to the centrally located polar cells.

FIGURE 1:

Nurse cells influence border cell cluster migration. (A) Frames from a DIC movie (Supplemental Movie S2) showing border cells (false-colored yellow) migrating between the nurse cells (nc; pink outlines) to reach the oocyte (n = 5 movies). The follicle cell epithelium surrounds the germ line. Scale bar: 40 μm. (B) Merged three-dimensional z-stack image of a live stage 9 egg chamber stained with the lipophilic FM 4-64 dye; the dye labels many, though not all, cell membranes. This image is taken from Supplemental Movie S1. Border cells, distinguished by enriched FM 4-64 staining (false-colored magenta), are closely surrounded by neighboring nurse cell membranes. The oocyte is to the far right. Scale bar: 20 μm. (C) Model of tissue-level forces between the border cell cluster (blue arrows) and the nurse cells (red arrows). Border cells are proposed to resist compression from nurse cells. Border cells are blue; central polar cells are pink; nurse cells are yellow; dashed lines indicate nurse cell nuclei. The direction of migration is to the right (black arrow). (D–E′) 2P-Sqh expression in nurse cells. Stage 9 egg chambers stained for 2P-Sqh (green in D and E) and phalloidin to label F-actin (magenta in D and E). (D and D′) Control egg chambers have low levels of 2P-Sqh at nurse cell membranes. (E and E′) Elevated 2P-Sqh staining (arrowheads) at cortical nurse cell membranes in a nurse cell (n.c.) > RhoGEF2 egg chamber (n = 24). Filamentous 2P-Sqh within nurse cell cytoplasm (asterisks in D′ and E′) is likely nonspecific staining. Scale bars: 20 μm. (F and G) RhoGEF2 expression in nurse cells causes the cells to become rounder than normal (asterisks). Late stage 9 control (F; nurse cell GAL4/+) and stage 10 RhoGEF2 (G; nurse cell GAL4/UASp-RhoGEF2) egg chambers stained with phalloidin to label F-actin. Scale bar: 20 μm. (H and I) Measurements of nurse cells, represented as box-and-whisker plots. The whiskers represent the minimum and maximum, the box extends from the 25th to the 75th percentiles, and the line indicates the median measurement. Nurse cells from control egg chambers (n = 29) and nurse cells from nurse cell > RhoGEF2 egg chambers (n = 26) were analyzed; not significant (n.s.), p ≥ 0.05; ****, p < 0.0001; unpaired t test. (H) Quantification of the area of nurse cells in control and nurse cell > RhoGEF2 egg chambers. (I) Quantification of the circularity of nurse cells from control and nurse cell > RhoGEF2 egg chambers. A value of 1.0 is a perfect circle, whereas a number close to 0.0 is an extremely elongated shape. (J) Quantification of border cell migration in egg chambers expressing RhoGEF2 in nurse cells (nurse cell GAL4/UASp-RhoGEF2; n = 162 egg chambers) compared with control (nurse cell GAL4/+; n = 217), shown as the percentage of stage 10 egg chambers with complete, incomplete, or no migration to the oocyte. A diagram illustrating the migration distance is shown in Supplemental Figure S1C. At least three trials were performed per genotype; n ≥ 50 egg chambers per trial. **, p < 0.01; unpaired t test. Error bars: SEM. In this and all subsequent figures, anterior is to the left and the direction of migration is to the right.

Border cells remain as a cohesive cluster despite moving within the dense cellular environment of the egg chamber. To move forward, the border cell cluster needs to navigate between the closely packed nurse cells. The nurse cells serve as the migratory substrate for the border cell cluster, with differential and dynamic E-cadherin adhesion between the outer cluster membranes and nurse cell membranes providing optimal traction for border cells to move forward (Niewiadomska et al., 1999; Cai et al., 2014). The arrangement of nurse cells does not appear to allow much space for migration and further suggests that nurse cells apply force onto border cells. Despite this seeming limitation, the border cell cluster moves without difficulty while maintaining cohesion of the group. How the structural organization of the border cell group is preserved, and the extent to which this facilitates its movement through the confined setting of the egg chamber, is unclear.

Cells can withstand external mechanical pressure through the functions of internal force-generating proteins (Mohan et al., 2015) such as nonmuscle myosin II (Myo-II). Myo-II is a hexameric protein complex that consists of three subunits: two heavy chains, two essential light chains, and two myosin regulatory light chains (MRLC; Drosophila Spaghetti Squash [Sqh]). The RhoA small GTPase activates Myo-II through kinases such as Rho-associated kinase (Rok; Amano et al., 2010). Phosphorylation of MRLC/Sqh leads to maximal Myo-II activation (Vicente-Manzanares et al., 2009). The active Myo-II motor establishes cortical tension of cells through its interaction with membrane-associated F-actin. For example, during mitosis, actomyosin tension promotes the rounding of cells (Stewart et al., 2011; Ramanathan et al., 2015). Myo-II is also essential for migrating cells, polarizing the cell and promoting retraction of the trailing edge (Vicente-Manzanares et al., 2009). The role for Myo-II in migrating collectives, however, appears to be more complex. RhoA-activated Myo-II helps organize mechanical tension at the group level to prevent nonleading cells from forming migratory protrusions (Omelchenko et al., 2003; Friedl et al., 2014; Reffay et al., 2014). Elevated Myo-II at cell contacts between A431 cancer cells inhibits their ability to invade collectively, thus promoting single-cell migration (Hidalgo-Carcedo et al., 2011). Most of these studies were performed with cells in culture, either on two-dimensional surfaces or in simplified three-dimensional matrices. At present, the degree to which these findings apply to collectives moving through complex endogenous tissues and whether Myo-II-dependent cortical tension has additional roles, if any, are still open questions.

We report here that live border cells maintain a specific overall structure throughout their entire migration that requires highly dynamic Myo-II activity. The surrounding nurse cells produce forces that impinge on the border cell cluster. In response, border cells up-regulate activated Myo-II at the cluster periphery. The spherical border cell cluster shape can be perturbed by various genetic manipulations of Myo-II function, either in the nurse cells or in border cells themselves. Such altered cluster morphology correlates with reduced ability of some border cells to migrate. Dynamic phosphorylation of Sqh/MRLC through Rok is required for cluster morphology. Moreover, periodic waves of Myo-II enrichment occur at the outer edges of the cluster coincident with membrane contraction. We propose that cycles of dynamic actomyosin activation and inactivation at the cluster periphery provide optimal levels of cortical tension at the level of the collective. This allows border cells to achieve a specific compact morphology that enables the cluster to withstand forces from the surrounding tissue and may help the cluster to efficiently migrate.

RESULTS

Nurse cells mechanically influence border cell cluster migration

The egg chamber is a cell-dense tissue, raising the question of whether the nurse cell environment mechanically influences the border cell cluster during its migration. We first addressed how nurse cells are organized with respect to border cells during the stages that border cells migrate. We stained live egg chambers with FM 4-64 dye, which labels many, though not all, cell membranes. Three-dimensional reconstruction of z-stack images shows that adjacent nurse cells are in tight contact with one another, with very little apparent space between the cells (Figure 1B and Supplemental Movie S1). The border cell cluster is located directly adjacent to, and is surrounded by, nurse cell membranes (Figure 1B). Live time-lapse imaging using differential interference contrast (DIC) optics to visualize the shapes of unstained cells reveals that border cells move efficiently between the 15 packed nurse cells (Figure 1A and Supplemental Movie S2; Montell et al., 2012). Moreover, the nurse cells visibly deform when border cells pass between them (Supplemental Movie S2; Stonko et al., 2015).

Given the tight association between border cells and the much larger nurse cells (Figure 1C), we asked whether nurse cells are able to mechanically influence border cells. The nurse cells are normally constrained by the surrounding squamous follicle cell epithelium; by the basement membrane located on the outside of the tissue, which serves as a so-called “molecular corset” for egg chamber growth; and by a muscle sheath that surrounds each string of egg chambers (Cetera and Horne-Badovinac, 2015; Andersen and Horne-Badovinac, 2016). Increasing the cellular contraction of nurse cells to increase nurse cell rigidity is expected to increase compressive forces along the migration path (Figure 1C). RhoGEF2 is a known activator of the Drosophila RhoA GTPase, which in turn activates Rok and Myo-II to drive cell shape changes (Hacker and Perrimon, 1998; Rogers et al., 2004). Myo-II activity upon cortical F-actin is known to promote force generation that is transmitted along actin cables to cell membrane anchors (Munjal and Lecuit, 2014). When coupled with intracellular pressure, actomyosin-dependent contraction contributes to cell rounding and cortical tension (Stewart et al., 2011; Clark et al., 2014; Ramanathan et al., 2015). We therefore expressed RhoGEF2 in all nurse cells to ectopically increase the cellular contraction and tension of these cells. We used an antibody to phosphorylated Sqh (2P-Sqh) to confirm that this manipulation increases the levels of activated Myo-II in nurse cells compared with control (Figure 1, D, D′, E, and E′). RhoGEF2 expression altered the shape but not the overall size of nurse cells, causing the cells to appear rounder than control (Figure 1, F and G, and Supplemental Movie S3). To analyze this directly, we measured the area (Figure 1H) and circularity (Figure 1I) of nurse cells. The area of the RhoGEF2-expressing nurse cells is unchanged, but these cells are significantly more circular than control nurse cells (median control = 0.68 vs. median nurse cell > RhoGEF2 = 0.78). Notably, we did not observe any visible gaps between RhoGEF2-expressing nurse cells, nor did the ring canals that connect nurse cells appear grossly abnormal (Supplemental Movie S3).

Elevating nurse cell contraction by RhoGEF2 expression impairs the ability of genetically wild-type border cells to complete their migration (Figure 1J). Whereas all border cells in control egg chambers reach the oocyte by stage 10 of oogenesis, ∼40% of border cells in nurse cell > RhoGEF2 egg chambers either do not initiate migration or fail to reach the oocyte by stage 10. These data indicate that having optimal levels of nurse cell deformability is necessary for border cells to be able to migrate.

Border cells increase levels of activated Myo-II and undergo morphology changes in response to constraint by nurse cells

Live migrating border cell clusters have a distinctly round shape that is maintained, with only minor deviations, for the entirety of their migration (Figure 1A and Supplemental Movie S2). This observation suggests that border cells require a specific morphology during their migration. Therefore we next asked what happens to the cluster in the more rigid environment of RhoGEF2-expressing nurse cells (Figure 1G). Border cells in fixed control egg chambers exhibit typical cluster morphology and length (Figure 2, A, A′, and C; see Materials and Methods). Strikingly, however, egg chambers with nurse cells that express high levels of RhoGEF2 have a significant number of genetically wild-type border cell clusters that visibly become stretched along the migration pathway (Figure 2, B, B′, and C). Increased nurse cell tension thus constricts the shape of the border cell cluster during its migration.

FIGURE 2:

Border cells change shape and increase levels of activated Myo-II in response to nurse cells. (A–C) Elongation of wild-type border cell clusters upon increased contraction of nurse cells. Stage 9 control (nurse cell GAL4/+) and RhoGEF2 (nurse cell GAL4/UASp-RhoGEF2) egg chambers stained with phalloidin to label F-actin (green in A and B) and Fascin to mark border cells (magenta in A and B; white in A′ and B′). Scale bars: 20 μm. (A and A′) Control border cell cluster shape is normal. (B and B′) Example of a border cell cluster that is elongated in a nurse cell > RhoGEF2 egg chamber. (C) Quantification of individual border cell cluster lengths from control (n = 20) and nurse cell > RhoGEF2 (n = 55) egg chambers. The line indicates the mean. *, p < 0.05; unpaired t test. (D–F) Elevated 2P-Sqh levels in border cells when the surrounding nurse cells express RhoGEF2. (D and E) The heat map shows the levels of low to high signal intensity of 2P-Sqh. (D′ and E′) The border cell cluster is defined by Fascin signal, which is expressed at low levels in nurse cells at this stage of egg chamber development (Cant et al., 1994). Scale bars: 5 μm. (F) Quantification of the mean 2P-Sqh pixel intensity within the border cell cluster (defined by Fascin staining) from control (n = 8) and nurse cell > RhoGEF2 (n = 11) egg chambers, represented as a box-and-whisker plot. The whiskers represent the minimum and maximum pixel intensity; the box extends from the 25th to the 75th percentiles; the line indicates the median measurement. ***, p < 0.001; unpaired t test.

Given that Myo-II can function as a mechanosensor and that physical constriction of cells in vitro can recruit Myo-II to the cell cortex (Laevsky and Knecht, 2003; Ren et al., 2009; Luo et al., 2012; Elliott et al., 2015; Kim et al., 2015), we next analyzed what happens to Myo-II when border cells become constrained by nurse cells. Myo-II, as visualized using a green fluorescent protein (GFP)-tagged sqh transgene (Royou et al., 2002), is highly enriched at outer membrane surfaces of the border cell cluster throughout all stages of migration (Supplemental Figure S1; Edwards and Kiehart, 1996; Majumder et al., 2012). Drosophila Sqh is phosphorylated at either Ser-21 (1P) or at both Thr-20/Ser-21 (2P), which are equivalent to Thr-18/Ser-19 on mammalian MRLC (Jordan and Karess, 1997; Zhang and Ward, 2011). Phosphorylation at the 1P and/or 2P sites activates Myo-II. Therefore we examined activated Myo-II in wild-type border cells under conditions of elevated confinement. As described above, we increased nurse cell tension and contraction by expressing RhoGEF2 only in nurse cells. Next we analyzed border cells from nurse cell > RhoGEF2 egg chambers stained for 2P-Sqh. The border cell–specific marker, Fascin, allowed us to define and assess the levels of activated Myo-II only in border cells (Figure 2, D, D’, E, and E′).

When contraction of nurse cells is increased through expression of RhoGEF2, border cell clusters exhibit more 2P-Sqh staining compared with control (Figure 2, D and E). Moreover, the most intense 2P-Sqh staining is found close to the cluster periphery (Figure 2E). Quantification of the mean pixel intensity for 2P-Sqh confirms that the border cells in nurse cell > RhoGEF2 egg chambers have significantly elevated 2P-Sqh levels (Figure 2F). Thus, elevated pressure from the nurse cells increases the levels of activated Myo-II in border cells. These data support the idea that the nurse cells physically influence the morphology of the border cell cluster and that border cells respond through Myo-II activation. Together our results suggest a model in which the border cell cluster has a distinct morphology that is influenced by the nurse cell environment and that may be important for its movement within the tissue.

The border cell cluster requires Myo-II to maintain morphology during migration

We next asked how border cells attain their characteristic cluster shape. We hypothesized that the specific morphology of the border cell cluster contributes to the ability of border cells to resist forces from the surrounding tissue (Figure 3A). Myo-II is required for border cell migration and is highly expressed in the cluster (Figure 3B; Edwards and Kiehart, 1996; Fulga and Rørth, 2002; Majumder et al., 2012). Moreover, phosphorylated Myo-II is elevated in response to nurse cell–driven constriction (Figure 2, D–F). We thus reasoned that localized cortical actomyosin tension within the border cell group could serve as a mechanism to oppose compressive forces imposed on them by adjacent nurse cells (Figure 1C and Figure 3A).

FIGURE 3:

Myo-II maintains the migratory morphology of the border cell cluster. (A) Illustration of proposed mechanism by which Myo-II (red) and F-actin (blue) generate cortical tension to contract border cell cluster membranes (arrows). (B) Localization of Sqh:GFP in border cells during migration. Sqh:GFP is enriched in foci at the cluster periphery (arrowheads) and in polar cells (asterisk). Scale bar: 5 μm. (C–G) Knockdown or loss of sqh disrupts cluster shape. (C and D) Fascin labels border cells of stage 9 control (c306-GAL4, tsGAL80/+) and stage 10 sqh RNAi egg chambers (c306-GAL4, tsGAL80/+; UAS-sqh RNAi/+). (D and G) Dashed line indicates the anterior oocyte border. (E) Quantification of cluster length (l) along the axis of migration (schematic), shown as the percentage of control (n = 30) and sqh RNAi (n = 75) midmigration border cells in stage 9 egg chambers. (F and G) Examples of stage 9 FRT 19A control (F) and stage 10 sqh^AX3 (G) mosaic mutant clusters (n = 38), costained for E-cadherin (E-cad; green) to label cell membranes and DAPI (blue) to show nuclei. Loss of nuclear RFP (red fluorescent protein; red) marks cell clones (arrowheads). Scale bars (C, D, F, and G): 20 μm. (H and I) Frames from control (H) and sqh RNAi (I) live movies (Supplemental Movies S4 and S5) showing migrating border cells (mCD8:GFP). (J) Quantification of the variability in cluster length from individual movies over time (∼1 h duration), shown as box-and-whisker plots. The whiskers represent the minimum and maximum length within one movie; the box extends from the 25th to the 75th percentiles; the line indicates the median length measurement within that movie.

To investigate the extent to which Myo-II contributes to the shape of the migrating border cell cluster, we knocked down Myo-II function using a GAL4-driven upstream activation sequence–RNA interference (UAS-RNAi) transgene that significantly reduces sqh RNA levels in vivo (Supplemental Figure S2, A and B). We also validated specific knockdown in border cells. Expression of sqh RNAi, driven by c306-GAL4 in border cells and anterior follicle cells (Supplemental Figure S3), shows reduced levels of phosphorylated Sqh in border cells (Supplemental Figure S2C). In agreement with previous studies, sqh RNAi impairs the detachment of border cells from the follicle cell epithelium, border cell migration to the oocyte, and retraction of leading-edge protrusions (Supplemental Figure S4, A–E; Edwards and Kiehart, 1996; Fulga and Rørth, 2002; Majumder et al., 2012). A large proportion of sqh RNAi border cell clusters visibly elongate along the migration pathway compared with control (Figure 3, C and D). We then measured the length of border cell clusters along the migration axis from the front to back of the group, excluding visible cellular protrusions, in fixed egg chambers (see Materials and Methods). Control border cell clusters are compact and measure ∼20–30 μm in length (Figure 3, C and E). In contrast, sqh-deficient clusters can reach lengths up to 80 μm, more than twice the control cluster length (Figure 3, D and E). We confirmed the requirement of sqh for proper cluster morphology by producing mosaic clones of a loss-of-function sqh allele (sqh^AX3). We find that 18% of border cell clusters containing at least two or three sqh^AX3 homozygous mutant cells are elongated, whereas control mosaic clusters are compact (Figure 3, F and G). Clusters that contain only mutant sqh^AX3 border cells are rarely found (Majumder et al., 2012), which may account for the lower frequency of elongated clusters compared with sqh RNAi. The stretched-out cluster shape produced by loss of Myo-II is distinct from mutants that disrupt adhesion between individual cells and cause the border cell cluster to partially or completely pull apart (Pinheiro and Montell, 2004; Llense and Martín-Blanco, 2008; Cai et al., 2014). Indeed, border cells deficient for sqh do not break away from the cluster (Figure 3, D and G), further indicating that Myo-II does not play a major role in keeping border cells adhered together in a group.

Next we performed live time-lapse imaging to address the role for Myo-II in maintaining the dynamic morphology of the cluster during migration (Figure 3, H and I; Supplemental Movies S4 and S5). Once migration begins, sqh RNAi border cell clusters noticeably become more elongated than control. We quantified the variability in cluster length during the entire movie, primarily focusing on border cell clusters that had detached or were in the process of detaching from the epithelium (Figure 3J). Control border cells consistently retain a compact shape, with minimal variation in the length of the cluster. In contrast, when sqh RNAi border cells migrate, cluster length is highly variable and can stretch up to 80 μm long, in agreement with measurements from fixed egg chambers (Figure 3, E and J). Together these data show that loss of Myo-II disrupts the compact morphology of border cell clusters during their migration.

Activated Myo-II is required for cell and cluster morphology of border cells

The results described above suggest that Myo-II promotes cluster shape and show that nurse cell constriction recruits activated Myo-II to the cluster periphery (Figure 2, D–F). Next we asked to what extent is specific regulation of Myo-II activity required for normal border cell cluster shape. The regulatory light chain component Sqh (MRLC) undergoes dynamic cycles of phosphorylation by kinases and dephosphorylation by myosin phosphatase to, respectively, activate and inactivate Myo-II (Vicente-Manzanares et al., 2009). Rok is one of a number of serine–threonine kinases that phosphorylate the 1P and 2P sites of Sqh/MRLC (Amano et al., 1996; Ueda et al., 2002). Because Rok is required for border cell migration (Supplemental Figure S4A; Majumder et al., 2012), we wanted to know whether Rok is the relevant kinase for Myo-II function in cluster morphology. Activated Myo-II (1P- and 2P-Sqh) localizes primarily in intense foci restricted to the cluster periphery, adjacent to contacts with the nurse cells (Figure 4, A–A″). We therefore examined Rok localization with respect to activated Myo-II. Using a functional Rok:GFP transgene (Bardet et al., 2013), we find that Rok specifically localizes to the outer edge of the border cell cluster in a discontinuous pattern (Supplemental Figure S5, A″ and A′′′). Rok:GFP coincides with some 1P and 2P Sqh foci (Supplemental Figure S5, A–A′′′), consistent with the ability to activate Myo-II. In agreement with other migrating collectives (Hidalgo-Carcedo et al., 2011; Reffay et al., 2014), very little active Myo-II or Rok localizes at membrane contacts between border cells.

FIGURE 4:

Cell and cluster morphology of border cells requires activated Myo-II. (A–A″) Representative example of a stage 9 wild-type border cell cluster stained for 1P-Sqh (green) and 2P-Sqh (red). The 1P-Sqh and 2P-Sqh colocalize (yellow in A″) and are enriched in discrete foci (arrowheads, dashed line) at the cluster periphery (n = 17). Armadillo (Arm; white in A″) labels all cell membranes, including the central polar cells (asterisk). Scale bar: 5 μm. (B) Example of an elongated Rok KD cluster (bracket) in a slbo-GAL4/UAS-Venus:Rok KD stage 10 egg chamber. The border cells are stained for Fascin (red), Venus:Rok KD (green), and DAPI (blue nuclei). (C) Example of a cluster with stretched-out, trailing Rok² mutant border cells (loss of nuclear RFP; arrowheads) in a stage 9 egg chamber stained for Fascin (green) and DAPI (blue nuclei). Scale bars (B and C): 50 μm. (D–G) Activated RhoA causes rounded border cells and disrupts cluster shape. Stage 9 control (slbo-GAL4, UAS-mCD8:GFP/+) and active RhoA (slbo-GAL4, UAS-mCD8:GFP/UAS-Rho1^V14) border cells labeled by GFP (D and F) and F-actin (D′ and F′). (E and G) The 1P- and 2P-Sqh localize to the periphery of control (E) but are mislocalized in constitutively active RhoA (G; n = 19) border cell clusters. Dashed lines indicate the cluster boundary. (H–H′′) Altered shape of Mbs RNAi (c306-GAL4/+; UAS-Mbs RNAi/UAS-mCD8:GFP) border cells (n > 7), marked by GFP (H; green in H″) and F-actin (H′; magenta in H″). Scale bar (D–H″): 5 μm. (I) Mean pixel intensity of 1P- and 2P-Sqh measured in control (n = 10) and active RhoA (n = 6) border cell clusters. Error bars: SEM. **, p < 0.01; unpaired t test.

Given the specific localization of Rok to the cluster periphery, and colocalization with activated Myo-II, we next analyzed the requirement for Rok in the maintenance of border cell cluster morphology. Border cell clusters that express a kinase-dead version of Rok (Rok KD), or that contain mosaic mutant clones of a strong loss-of-function Rok allele (Rok²), result in elongated clusters (Figure 4, B and C). Thus the Myo-II–activating kinase Rok regulates border cell cluster shape during migration.

We next analyzed the degree to which the appropriate levels and/or localization of activated Myo-II contribute to cell and cluster shape of border cells. The small GTPase RhoA (Drosophila Rho1) activates Rok and thus influences Myo-II activation (Amano et al., 1996; Ishizaki et al., 1996; Matsui et al., 1996). We therefore expressed a constitutively active form of RhoA (RhoA^V14) in border cells. Compared with control clusters, RhoA^V14 individual border cells are much rounder, and the overall cluster appearance is altered (Figure 4, D–G). Consistent with a previous report (Bastock and Strutt, 2007), 40% of RhoA^V14 border cells do not complete their migration (Supplemental Figure S4A). The overall levels of 1P- and 2P-Sqh are also elevated in RhoA^V14 border cell clusters compared with control (Figure 4, D–G and I). Moreover, activated Myo-II (1P- and 2P-Sqh) is mislocalized in RhoA^V14 clusters, with ectopic enrichment at membrane contacts between border cells (Figure 4, E and G). We confirmed that RhoA protein is misexpressed (Supplemental Figure S5, B–C′). Specifically, RhoA^V14 clusters have high levels of RhoA protein around each border cell, compared with a more general distribution in control clusters. Such altered localization supports the idea that RhoA^V14 activates Myo-II at individual border cell cortical membranes (Figure 4G). Border cells also have altered F-actin distribution (Figure 4F′), consistent with the known ability of mammalian RhoA both to promote F-actin assembly through the formin mDia (mammalian homologue of Drosophila Diaphanous) and Myo-II activity through Rok (Narumiya et al., 2009).

To further address the importance of appropriate levels of activated Myo-II, we next inactivated myosin phosphatase. Myosin phosphatase consists of a catalytic protein phosphatase 1 subunit (PP1c) and a myosin-binding specificity subunit (Mbs) (Grassie et al., 2011). Because myosin phosphatase dephosphorylates Sqh, loss of the phosphatase will result in increased levels of phosphorylated and activated Myo-II (Kimura et al., 1996; Grassie et al., 2011). Mbs is highly expressed in border cells, and RNAi knockdown elevates the levels of 1P-Sqh in the cluster (Majumder et al., 2012). When we knocked down Mbs using RNAi, these clusters contained rounder border cells than normal (Figure 4, H–H′′). Taken together, our results suggest that appropriate levels of activated Myo-II, as well as restriction of activity to the cluster periphery, promote the proper morphology of both individual border cells and the cluster as a whole.

Dynamically regulated Myo-II correlates with cluster shape and membrane deformation

Previous research on collectively migrating cells in vitro indicates that spatially localized Myo-II activation keeps collectives together as a functional unit and prevents their dissociation into single cells (Hidalgo-Carcedo et al., 2011; Friedl et al., 2014; Reffay et al., 2014). In the case of collectives that migrate within tissues, such as border cells, it is unclear whether it is the levels/localization of active Myo-II or the dynamics of active Myo-II that are needed to maintain a migratory morphology. We therefore investigated the extent to which dynamic Myo-II activation contributes to border cell cluster morphology during migration. We used phosphorylation-mutant variants that lock Sqh/MRLC into either unphosphorylated (Sqh-AA) or constitutively phosphorylated (Sqh-EE) forms and that are driven by the endogenous sqh promoter (Jordan and Karess, 1997; Winter et al., 2001). Because we wanted to manipulate the levels of activated Myo-II, rather than Myo-II protein itself, we knocked down the Myo-II activator Rok in border cells. Expression of Rok RNAi strongly depletes Rok mRNA in vivo (Supplemental Figure S2, D and E) and reduces 1P-Sqh levels in border cells (Majumder et al., 2012). Rok RNAi disrupts border cell migration (Figure 5A and Supplemental Figure S4A; Majumder et al., 2012) and produces long protrusions that fail to retract (Figure 5, B and C). Moreover, similar to our observations with Rok mutants (Figure 4, B and C), Rok RNAi border cell clusters are more elongated compared with control (Figure 5F).

FIGURE 5:

Requirement for dynamic cycles of activated Myo-II in cluster shape and membrane deflection. (A) Sqh-EE strongly suppresses the Rok RNAi migration defects. Quantification of complete, incomplete, and no migration in stage 10 Rok RNAi (c306-Gal4/+; UAS-Rok RNAi/+) egg chambers, with or without the indicated Sqh mutant transgenes. n ≥ 50 egg chambers in each of three trials; not significant (n.s.), p ≥ 0.05; ***, p < 0.001; one-way ANOVA with Dunnett test compared with “complete migration.” (B–E) Sqh-EE rescues the Rok RNAi border cell protrusion-length defects. (B) Quantification of mean protrusion length in the indicated genotypes. Dashed line shows the mean protrusion length of control (c306-GAL4/+). n ≥ 86 protrusions per genotype; n.s., p ≥ 0.05; ***, p < 0.001; one-way ANOVA with Dunnett test. Error bars in A and B: SEM. (C–E) Stage 9 Rok RNAi egg chambers, with or without the indicated Sqh mutant transgenes, stained for Fascin to label border cells and protrusions (brackets). Scale bar: 20 μm. (F) Sqh-EE and Sqh-AA did not alter the Rok RNAi cluster elongation defect. Quantification of individual border cell cluster length measurements in the indicated genotypes. The line indicates the mean. n ≥ 46 clusters per genotype; n.s., p ≥ 0.05; **, p < 0.01; one-way ANOVA with Dunnett test. (G) Consecutive frames from a Sqh:GFP movie at the indicated times. Two foci of Sqh:GFP (arrowheads) were tracked at the cluster periphery until they disappeared (arrowhead outlines). A new foci appears in the last frame. (H and H′) Frame from a time-lapse Sqh:GFP movie (Supplemental Movie S6) simultaneously imaged by fluorescence microscopy (H) and DIC (H′) optics to visualize cell membranes (n = 6). Scale bars (G and H): 5 μm. The boxed region corresponds to the kymograph shown in I, with the outside (“o”) and inside (“i”) labeled to provide orientation. (I) Kymograph of the GFP (top) and DIC (bottom) channels from the boxed region in H and H′ over time (minutes). Sqh:GFP (magenta, overlay) correlates with deflection of the cluster membrane (n = 9 egg chambers). Orientation as in H and H′.

We next asked whether the phosphomimetic or nonphosphorylatable mutant forms of Sqh could rescue the Rok-dependent phenotypes. Introduction of one copy of Sqh-EE strongly, though not completely, suppresses Rok RNAi migration (Figure 5A) and protrusion defects (Figure 5, B, C, and E), consistent with the known role for Rok in phosphorylating Sqh (Amano et al., 1996; Winter et al., 2001). As expected, Sqh-AA has little effect (Figure 5, A, B, and D). Surprisingly, neither Sqh-EE nor Sqh-AA is able to alter the length of Rok RNAi clusters (Figure 5F). Importantly, Sqh-EE, visualized by a FLAG tag, localizes to the cluster periphery (Supplemental Figure S5, D–D″′). The distribution of Sqh-EE is similar to that observed for endogenous 2P-phosphorylated Sqh (compare with Figure 4, A–A″), indicating that the inability to rescue the Rok RNAi cluster shape defects is not due to mislocalization of the Sqh-EE transgene. These data together demonstrate that Rok promotes cluster shape through activation of Myo-II.

Our finding that phosphomimetic Sqh cannot rescue the Rok RNAi–dependent elongated cluster phenotype suggests that cycling between phosphorylated active and dephosphorylated inactive forms of Myo-II maintains the shape of the border cell cluster. To further test the idea that dynamic regulation of Myo-II activity is required for cluster morphology, we next analyzed the localization of GFP-tagged Sqh (Sqh:GFP) in live border cells. Our previous study showed that Sqh:GFP coalesces into discrete, highly dynamic foci at the cluster periphery (Figure 5G) that are lost upon inhibition of Myo-II phosphorylation (Majumder et al., 2012). These foci resemble those observed using 1P-/2P-Sqh antibodies in fixed border cells (Figure 4, A–A′′), suggesting that they represent assembly-competent Myo-II that is activated and can bind to F-actin (Vicente-Manzanares et al., 2009). We performed live time-lapse microscopy of Sqh:GFP fluorescence in conjunction with DIC imaging, which allowed us to explicitly image border cell membranes at the outer edge of the cluster where they contact the nurse cells (Figure 5, H and H’; Supplemental Movie S6). Using this method, we analyzed the dynamics of Sqh:GFP foci versus cluster membrane deflections over time (Figure 5I). The appearance of enriched Sqh:GFP signal strongly correlates with cell membranes that pull inward toward the center of the cluster, whereas disappearance of the Sqh:GFP signal correlates with membranes that relax outwardly (Figure 5I). These cycles, or waves, of Sqh:GFP signal are highly dynamic, with relatively short timescales of ∼60 s. Interestingly, Myo-II signal and cluster membrane contractility appear to propagate in both the forward and rearward directions (Supplemental Movie S6). Sqh:GFP foci enrichment thus coincides with contractility of border cell membranes at the cluster periphery during migration. Together these results are consistent with dynamic Myo-II activity producing waves of actomyosin contraction that help maintain border cell cluster morphology during migration.

DISCUSSION

Cells move within complex environments in vivo, interacting with other cells and the ECM. This poses a particular challenge for migrating collectives, because they need to withstand extracellular forces without falling apart or becoming greatly disorganized. The mechanisms that govern the impact of tissues on the structural organization of collectives have been unclear because of the difficulty of studying these processes within the native tissue. Here we show that the border cell collective maintains a specific morphology during its entire migration to the oocyte (Figure 6A). There is very little apparent space separating adjacent nurse cells; nonetheless, border cells efficiently navigate between nurse cells. In contrast to cells that move through the ECM and require proteolytic enzymes to create a pathway (Wolf and Friedl, 2011), the “cell-on-cell” type of migration used by border cells appears to be independent of proteases (Montell et al., 2012). We propose instead that reciprocal mechanical interactions between nurse cells and border cells facilitate the movement of border cells within the cell-dense environment of the egg chamber.

FIGURE 6:

Proposed model for maintaining the collective morphology of border cells migrating within the egg chamber. (A) The border cell cluster maintains a collective morphology while migrating between the large nurse cells. (B) A balance of forces (arrows), through actomyosin activity and cortical tension, between nurse cells (yellow) and border cells (blue) produces a compact cluster shape. (A and B) Right, when forces become unbalanced, for example by altered myosin activity, the collective migratory morphology is disrupted. Either knocking down Sqh in border cells or expressing ectopic RhoGEF2 in nurse cells causes the border cell cluster to elongate and stop migrating. Border cells respond to increased levels of compressive forces from the surrounding nurse cells by elevating the levels of activated Myo-II.

Although nurse cells are known to function as the migratory substrate for border cells, we show that they also impose force upon migrating border cells. The egg chamber is elliptical at this stage and is surrounded by a layer of follicle cells, a basement membrane, and the muscle sheath, which together constrain the overall shape of the tissue (Cetera and Horne-Badovinac, 2015; Andersen and Horne-Badovinac, 2016). We propose that such geometry, in turn, drives an inward, anisotropic force from the nurse cells that imposes tension upon border cells perpendicular to the migration pathway. The border cell cluster actively resists compressive forces from the nurse cells (Figure 6B). The periodic nature of Sqh:GFP indicates that border cells have an intrinsic mechanism to rapidly enrich Myo-II at the cluster periphery during their movement between nurse cells. Specifically, upon nurse cell constriction, local and dynamic activation of Myo-II helps border cells resist this pressure. When forces become imbalanced, either through loss of activated Myo-II in border cells or by elevated cortical tension of nurse cells, the border cell cluster elongates (Figure 6, A and B). Cluster elongation likely occurs along the migration pathway, because the forces imposed by nurse cells are greater in this direction.

Our study also suggests that border cells and their migratory substrate, the nurse cells, each have a certain level of flexibility required for optimal migration. An open question is how nurse cell deformability is regulated at the global level, as well as more locally as border cells move between these large cells. Here we used RhoGEF2 to ectopically elevate Myo-II activation in all nurse cells through actomyosin-based contraction and tension (Figure 6B; Rogers et al., 2004). Endogenous Myo-II levels are not very high in nurse cells (Edwards and Kiehart, 1996), suggesting that Myo-II is unlikely to be the sole regulator of nurse cell membrane pliability. Further work will be needed to identify the specific combination of factors, such as membrane lipids, associated cortical membrane proteins, and/or cross-linked F-actin (reviewed by Diz-Muñoz et al., 2013; Clark et al., 2014), that influence the malleability of nurse cell membranes.

How does actomyosin-driven tension help border cells maintain a compact morphology? Myo-II accumulates at the periphery of cells undergoing mitosis, increasing intracellular pressure and cortical tension; this in turn allows a cell to resist external deformation (Ramanathan et al., 2015). Similar actomyosin contractility at the periphery of the border cell cluster thus could increase surface tension to maintain cellular organization. In physics, surface tension is inversely proportional to surface area. Small groups of cells, such as border cells or tumor clusters, often have a small surface area and pronounced curvature (this study; Yamamoto et al., 1983; Friedl et al., 1995; Hidalgo-Carcedo et al., 2011; Montell et al., 2012). However, Myo-II also is able to minimize cell-surface curvature of branching endothelial cells cultured in a three-dimensional microenvironment (Elliott et al., 2015). Our observation that Sqh:GFP and phosphorylated Myo-II each localizes to discontinuous foci on outer cluster membranes may help resolve these two seemingly contradictory ideas. Local enrichment of Myo-II at the cluster periphery, coupled with F-actin, could produce regions of transient lower curvature that help the cluster keep its shape. Indeed, border cells produce periodic dynamic waves of intense Sqh:GFP signal that correlate with local deformation of outer cluster membranes. Similarly, fibroblasts in suspension undergo oscillating cell shape changes that depend on actomyosin contractility through dynamic distribution of F-actin and Myo-II (Paluch et al., 2005; Salbreux et al., 2007). Periodic enrichment of Myo-II at the cluster periphery therefore could promote sufficient cortical tension to rapidly and reversibly adjust the cell cortex in response to changes in the surrounding tissue. We propose that, at the scale of the collective, such local membrane deformations jointly sustain cluster-wide organization and shape.

Our data support a model in which border cell cluster shape requires dynamic regulation of Myo-II activity. Cycles of Sqh/MRLC phosphorylation and dephosphorylation, which correspond to assembly- and disassembly-competent Myo-II, respectively, influence the ability of Myo-II to form bipolar minifilaments and bind to F-actin (Vicente-Manzanares et al., 2009). Phosphomimetic Sqh-EE locks Myo-II into an active conformation but does not rescue the border cell cluster elongation phenotype caused by loss of the activating kinase Rok. Further highlighting the importance of dynamic cycles of Sqh phosphorylation, Myo-II that contains phosphomimetic Sqh fails to undergo contractile pulses, becomes more stabilized, and cannot support morphogenesis of Drosophila embryonic tissues (Kasza et al., 2014; Vasquez et al., 2014). In biochemical assays, phosphomimetic Sqh/MRLC mutants do not have as high an activity as endogenously phosphorylated Myo-II (Kamisoyama et al., 1994). Nonetheless, Sqh-EE strongly rescues Rok-dependent defects in border cell detachment and protrusion retraction and mildly suppresses migration defects. The failure of Sqh-EE to rescue the Rok RNAi cluster morphology defect therefore suggests that normal cluster shape requires greater cycling of Myo-II activity. At present, however, we cannot rule out a need for higher levels of phosphorylated Sqh. High levels of sustained Myo-II activation, such as through constitutively active RhoA or Mbs RNAi, are detrimental to the structure of the cluster because they induce continuous contraction of cells (Majumder et al., 2012; this study).

Myo-II has a surprising number of functions in migrating border cells, with roles in protrusion retraction (Fulga and Rørth 2002), detachment from the anterior epithelium (Edwards and Kiehart 1996; Majumder et al., 2012), and rotation of the cluster (Combedazou et al., 2016). We suggest that, in addition, dynamic Myo-II at the cluster periphery contributes to efficient border cell migration. How can localized dynamic Myo-II promote the movement of collectives? The restriction of activated Myo-II to outer collective surfaces has been proposed to keep cells together so they move as coordinated units with single leader cells (Hidalgo-Carcedo et al., 2011; Reffay et al., 2014). Our data show that dynamic Myo-II promotes the shape of the collective at the supracellular level. Such a morphology could allow the group to withstand constraints from the surrounding environment to be able to migrate. In this model, dynamic Myo-II creates waves of contraction to help collectives push through the tissue. Cancer cell collectives that migrate in a protease-independent manner do so by pushing through three-dimensional ECM in vitro (Ilina et al., 2011). Moreover, nonadherent migrating cells can use actomyosin cortical flows to propel cells forward in confined environments (Bergert et al., 2015). These studies raise the intriguing possibility that migrating border cells use Myo-II–dependent contraction to squeeze between nurse cells. The phenomenon of balanced opposing forces, coupled to actomyosin contraction, which determines the shapes of collectives that move in tissues, has been difficult to model in vitro. Therefore our results further highlight the importance of studying migrating cells, both single cells and collectives, in their native microenvironment.

MATERIALS AND METHODS

Drosophila strains and genetics

Crosses were performed at 25°C, except tub-GAL80^ts (“tsGAL80”) crosses, which were set up at 18°C. The following Drosophila stocks were obtained from the Bloomington Drosophila Stock Center, Bloomington, IN (unless otherwise indicated), and are described in FlyBase (http://flybase.org): hsp70-GAL4 (hs-GAL4), sqh^AX3 FRT 19A, Rok² FRT 19A, FRT 19A, c306-GAL4, slbo-GAL4, tub-GAL80^ts (“tsGAL80”), matalpha4-GAL-VP16-GAL4 (“nurse cell GAL4”), UAS-mCD8:GFP, UAS-Rho1.V14 (“RhoA^V14”), UAS-Venus:Rok^K116A (“Rok KD”; from J. Zallen, Memorial Sloan Kettering Cancer Center, New York, NY), UASp-T7.RhoGEF2, UAS-sqh RNAi (line 7916; Vienna Drosophila Resource Center [VDRC]; Vienna, Austria), UAS-Mbs RNAi (line 105762; VDRC), UAS-Rok RNAi (line 9774R-2; National Institute of Genetics Fly Stock Center [NIG-Fly]; Kyoto, Japan), UAS-Drak RNAi (line 107263, VDRC), UAS-GFP dsRNA (“GFP RNAi”), sqh-sqh^E20E21:FLAG (“Sqh-EE”) and sqh^AX3; sqh-Sqh:GFP (III) (“Sqh:GFP”; from R. Karess, Paris Diderot University, Paris, France), sqh-sqh^A20A21:FLAG (“Sqh-AA”; from L. Luo, Stanford University, Palo Alto, CA), and ubi-Rok:GFP (from V. Mirouse, Clermont Université, Clermont-Ferrand, France). The w¹¹¹⁸ line was used as a wild-type control.

Mosaic mutant clones of sqh^AX3 FRT 19A, Rok² FRT 19A, and control FRT 19A were produced by the FLP-FRT system (Xu and Rubin, 1993) using the ubi-mRFP.nls hsFLP FRT 19A stock. Homozygous mutant cell clones were marked by loss of nuclear mRFP. Three- to 5-d-old flies of the correct genotype were selected and heat shocked for 1 h at 37°C, twice a day for 2 d; this was followed by recovery at 25°C for 5 d. Clones for sqh^AX3, FRT 19A, and FRT 19A control were generated previously (Majumder et al., 2012) but were reanalyzed independently to assess cluster shape (Figure 3, F and G). Few completely mutant sqh^AX3 or Rok² clusters were found.

GAL4 lines were outcrossed to w¹¹¹⁸ as controls. slbo-GAL4 was used to overexpress proteins uniformly and at high levels in all border cells (Rørth et al., 1998; Laflamme et al., 2012). c306-GAL4 was used to express UAS-RNAi lines in anterior follicle cells and border cells. c306-GAL4 is expressed earlier in oogenesis than slbo-GAL4 and generally achieves more efficient RNAi knockdown (Murphy and Montell, 1996; Aranjuez et al., 2012; Laflamme et al., 2012). To induce optimal GAL4/UAS overexpression and RNAi knockdown, we incubated flies overnight (14–18 h) at 29°C before dissection. matalpha4-GAL-VP16-GAL4 (“nurse cell GAL4”) was used to overexpress RhoGEF2 in nurse cells (Hudson and Cooley, 2014; Spracklen et al., 2014). Expression of RhoGEF2 using the nurse cell GAL4 was achieved by incubating flies at 27°C. RhoGEF2 expression can cause nurse cell membrane breakdown in some egg chambers; these egg chambers were identified using phalloidin, which stains F-actin enriched at nurse cell membranes, and were excluded from our analyses.

The tsGAL80 line was used to suppress c306-GAL4/UAS-RNAi during early stages of oogenesis in which an RNAi line could be cell lethal. UAS-sqh RNAi (VDRC 7196) phenotypes were stronger when using c306-GAL4, tsGAL80 compared with c306-GAL4 (Majumder et al., 2012). The enhanced phenotype is presumably due to increased viability of cells with the greatest knockdown, since the pattern of c306-GAL4 expression with or without tsGAL80 was identical in anterior follicle cells and border cells (Supplemental Figure S3). Crosses with tsGAL80 were kept at 18°C to suppress GAL4-UAS (McGuire et al., 2004a, b). Before dissection, flies were heat shocked at 37°C for 1 h, followed by a shift to 29°C for 18–24 h to completely turn off GAL80 and turn on GAL4/UAS.

Immunostaining

Ovaries were dissected and antibody stained as previously described (McDonald and Montell, 2005). Fixation was generally in 4% methanol-free formaldehyde (Polysciences, Warrington, PA) in phosphate buffer (pH 7.4) for 10 min. For 1P-Sqh and 2P-Sqh staining, fixation was in fresh 4% paraformaldehyde (Polysciences) in PBT (1× phosphate-buffered saline, pH 7.2, 0.1% Triton X-100) for 20 min. Primary antibodies from the Developmental Studies Hybridoma Bank (DSHB; University of Iowa, Iowa City, IA) were used at the following dilutions: 1:100 mouse anti-Arm (N27A1); 1:10 rat anti–E-cad (DCAD2); 1:10 mouse anti-GFP (12A6); 1:10 mouse anti-Rho1 (p1D9); 1:50 mouse anti-Fascin (Sn7C). Other antibodies used in this study were: 1:100 mouse anti-FLAG (M2; Sigma-Aldrich, St. Louis, MO), 1:1000 guinea pig anti–1P-Sqh and 1:500 rat anti–2P-Sqh (from R. Ward, University of Kansas, Lawrence, KS). Secondary antibodies conjugated to Alexa Fluor 488, 568, or 647 (Thermo Fisher Scientific, Waltham, MA) were used at 1:400. Alexa Fluor 568–phalloidin (1:400; Thermo Fisher Scientific) was used to label F-actin, and 4′,6-diamidino-2-phenylindole (DAPI; 0.05 μg/ml; Sigma-Aldrich) was used to label nuclei. Egg chambers were mounted in Aqua-Poly/Mount (Polysciences) before being imaged.

Imaging and analyses

Micrographs were acquired using the Zeiss AxioImager Z1 microscope (Carl Zeiss, Thornwood, NY), Zeiss ApoTome system, and an MRm CCD camera with a 20× Plan-Apochromat 0.75 numerical aperture (NA) or a 40× Plan-Neofluar 1.3 NA, oil-immersion objective. The z-stack images acquired from live imaging of Sqh:GFP were processed using the Zeiss Axiovision 4.8 deconvolution software module. Live imaging was performed as previously described (Prasad et al., 2007; Majumder et al., 2012). For visualization of nurse cell membranes (e.g., Figure 1B and Supplemental Movie S1), the lipophilic dye FM 4-64 (Thermo Fisher Scientific) was added at 9 μM final concentration to the culture media during preparation and imaging of live dissected egg chambers (Bianco et al., 2007; Prasad et al., 2007). Kohler illumination was optimized to perform DIC time-lapse live imaging. Movies were created using the Axiovision 4.8 Inside 4D software module or ImageJ. Image brightness and/or contrast of images were adjusted in the ImageJ distribution Fiji (http://fiji.sc; Schindelin et al., 2012), Adobe Photoshop (Adobe, San Jose, CA), or Zeiss Axiovision 4.8 software.

Measurements of nurse cell shape, border cell cluster length, protrusion length, and mean pixel intensity were done using Fiji. Nurse cell shape was measured using the Analyze Particles function of Fiji. Briefly, individual nurse cell regions of interest (ROIs) were obtained by manually filling in the area bounded by the nurse cell membranes and were subjected to the Analyze Particles function. The function provides values for nurse cell ROI area and circularity. A circularity value of 1.0 indicates a perfect circle, while values approaching 0.0 represent extremely elongated shapes. For measuring cluster length, the outlines of clusters and protrusions in midmigration were defined by Fascin immunofluorescence (fixed samples) or mCD8:GFP expression (live samples). The overall length of the cluster was measured along the migration axis but excluded cellular protrusions. Protrusions and the cluster body were defined similar to the criteria of Poukkula et al. (2011). Briefly, the entire cluster, including protrusions, was identified by Fascin staining. The cluster body was defined by the unstained border cell nuclei and surrounding Fascin-positive cytoplasm. The protrusions were defined as cellular “extensions” away from the cluster body that lack nuclei and are narrower than the cluster body (Poukkula et al., 2011). Backward extensions were not included in these analyses. Protrusion length was measured manually and defined as the distance from the tip of the protrusion to its base, where it meets the main cluster body (Majumder et al., 2012).

For measuring mean pixel intensity, control and RhoA^V14 clusters were stained for 1P-Sqh and 2P-Sqh simultaneously and imaged using the same exposure settings. The cluster was defined by border cell–specific slbo-GAL4 > mCD8:GFP expression. For measuring the mean pixel intensity of 2P-Sqh staining in border cells from nurse cell > RhoGEF2 egg chambers, the border cell cluster was defined by Fascin staining. A new image was created to measure pixel intensity only within the area of the border cell cluster. The pseudocolor heat map images of 2P-Sqh–stained border cell clusters were generated using the Fiji Lookup Table function. The Fiji Reslice function was used to generate kymographs from movies after a common-line ROI was defined in the GFP and DIC channels. RNAi knockdown efficiency was measured from reverse transcription PCR (RT-PCR) products run on gels using the Fiji Gel Analyzer function.

RT-PCR

For measuring RNAi efficiency (Supplemental Figure S2), UAS-RNAi transgenes were expressed in whole flies using hs-GAL4 to achieve ubiquitous knockdown. Flies were subjected to heat shock for 1 h at 37°C twice a day on 2 consecutive days; this was followed by 29°C recovery overnight. Total RNA was isolated from 10 to 15 whole-fly carcasses. Trizol-extracted RNA was used for cDNA synthesis followed by RT-PCR, as previously described (Aranjuez et al., 2012). RT-PCR was performed using the Superscript III One-Step RT-PCR system (Thermo Fisher Scientific). PCR settings were as follows: 50°C for 30 min during the cDNA synthesis step; 55°C for 30 s during the annealing step; 72°C for 1 min during the extension step. The number of PCR cycles was empirically determined for each reaction and primer set to avoid the plateau phase of PCR amplification; 29 cycles were used for both sqh and Rok. Band intensities of the RT-PCR products were measured using ImageJ (Schneider et al., 2012). GAPDH was used as the reference gene. The gene-specific primers used were as follows: sqh fwd, TCACACTTGGCCTTCTCGTC; sqh rev, CGAGATAGTCGTCCGTTGGG; Rok fwd, AGGAACGCGTCTCACTCAAG; Rok rev, GTGAGGGAGAGCAGAGAGGA; GAPDH fwd, ACTCATCAACCCTCCCCCG; GAPDH rev, GCGGACGGTAAGATCCACAA.

Figures, graphs, and statistics

Figures and illustrations were created in Adobe Illustrator CS5. GraphPad Prism (GraphPad Software, La Jolla, CA) was used to generate graphs and statistical analyses (unpaired t test and one-way analysis of variance [ANOVA]).

Abbreviations used:

DIC

differential interference contrast

DSHB

Developmental Studies Hybridoma Bank

ECM

extracellular matrix

GFP

green fluorescent protein

JNK

Jun kinase

Mbs

myosin-binding specificity subunit

MRLC

myosin regulatory light chain

Myo-II

nonmuscle myosin II

NIG-Fly

National Institute of Genetics Fly Stock Center

PP1c

catalytic protein phosphatase 1 subunit

ROI

region of interest

Rok

Rho-associated kinase

Sqh

Drosophila Spaghetti Squash

UAS-RNAi

upstream activation sequence–RNA interference

VDRC

Vienna Drosophila Resource Center.