aPKC is a key polarity determinant in coordinating the function of three distinct cell polarities during collective migration

During animal development, cells display various forms of polarity or asymmetry, and how they establish, maintain and remodel their polarity is a fundamental question in cell biology. One of the most common forms of cell polarity observed in developing organisms is the apical-basal polarity found in epithelial tissues (Tepass, 2012). Early discoveries of key polarity complexes are crucial to our molecular understanding of apical-basal polarity. They include the apically localized Par complex (aPKC/Par6/Par3/Cdc42) and Crb complex (Crb/Sdt/Patj), and the lateral complex of Scribble/Dlg/Lgl in Drosophila. However, recent work has found that individual components within the same complex, such as the Crb or Par complex, may function differently in varying developmental stages or contexts (Knust and Bossinger, 2002; Morais-de-Sá et al., 2010; Penalva and Mirouse, 2012; Pocha and Knust, 2013; St Johnston and Ahringer, 2010; Tepass, 2012). As the epithelia undergo morphogenesis or collective cell movement, the original apical-basal polarity has to be remodeled and other types of cell polarity, such as front-back, need to form (Tepass, 2012; Veeman and McDonald, 2016). However, the molecular mechanisms underlying such polarity remodeling is not well understood.

The border cells in the Drosophila ovary represent an excellent developmental system with which to study cell polarity generation and remodeling. They are genetically tractable and amenable to live cell imaging and optogenetic manipulation (Montell et al., 2012), and have been used as an established in vivo model in the field of collective cell migration (Friedl and Gilmour, 2009). The border cells are derived from a group of about six follicle cells, which are part of the follicle epithelial layer that maintains its own apical-basal polarity (Fig. 1A) (Montell, 2003; Montell et al., 2012). As these follicle cells detach from their neighbors, round up, and form a migratory cluster, a dramatic polarity remodeling takes place, resulting in three distinct cell polarities (Montell et al., 2012; Veeman and McDonald, 2016). First, the border cell cluster adopts a front-back polarity in the distribution of the actin cytoskeleton, signaling molecules and trafficking vesicles (Montell et al., 2012; Wan et al., 2013). The leading border cell extends a major lamellipodial protrusion enriched in F-actin and actin dynamics regulators including active Rac and cofilin, whereas border cells at the side and back of the cluster only occasionally extend minor protrusions that contain less of these factors (Wang et al., 2010; Zhang et al., 2011). Such front-back polarity of the actin cytoskeleton is established and maintained as a result of guidance signaling from two receptor tyrosine kinases (RTKs), namely PVR and EGFR, with PVR making the major contribution (Montell, 2003; Montell et al., 2012). In addition, PVR is also mainly responsible for the asymmetric distribution of recycling endosomes, exocysts and phosphotyrosine (p-Tyr) staining (Wan et al., 2013), which serves as a well-tested marker for intracellular signaling molecules.

The second polarity inherent in each of the six outer border cells and the two central polar cells is apical-basal polarity (Fig. 1A), which has its origin in the follicle epithelial cells but is significantly remodeled to reflect the part-epithelial and part-mesenchymal characteristics of cells undergoing collective migration (Felix et al., 2015; Friedl and Gilmour, 2009; Niewiadomska et al., 1999; Pinheiro and Montell, 2004). Interestingly, outer border cells only extend protrusions in their lateral, not apical or basal, membrane regions (see Fig. 2) (Montell et al., 2012; Veeman and McDonald, 2016), although the functional significance of such restriction of protrusion formation is unclear.

The third cell polarity within each outer border cell is inside-outside polarity (see Fig. 3D, Fig. 4H), which is much less well understood. The inside region refers the membrane area where an individual border cell attaches to polar cells and other neighboring border cells, while the outside region refers to the membrane region where border cells contact the surrounding substrate – the large nurse cells (Montell et al., 2012; Veeman and McDonald, 2016). F-actin, myosin II and E-cad display clearly polarized distributions at the outside or inside membrane (Cai et al., 2014; Combedazou et al., 2017; Lucas et al., 2013), which allows strong actin dynamics and protrusion formation to occur only at the outside membrane of border cells.

How does the original apical-basal polarity of follicle cells give rise to the three distinct polarities in border cells? Is there a common mechanism that coordinates the generation of all three cell polarities? Are these cell polarities interconnected? These questions largely remain to be addressed. In this study, we show that the apical complex molecule aPKC is a key determinant in coordinating the formation of the three polarities and that one of the functions of the Crb complex, Par complex and the endocytic recycling machinery is to ensure a polarized distribution of aPKC at the inside apical junction and the outside lateral membrane.

Previous studies have analyzed the roles of several apical and basolateral polarity components in border cells, including the Par complex components Baz (Par3) and Par6, the adherens junction molecule E-cadherin (E-cad; Shotgun), and the lateral complex components Dlg (Dlg1) and Lgl [L(2)gl] (Goode and Perrimon, 1997; Niewiadomska et al., 1999; Pinheiro and Montell, 2004; Szafranski and Goode, 2004). Here, we utilized both in vivo RNAi and genetic mosaic approaches to determine the functional roles of the Crb complex.

Loss of function of Crb complex components resulted in significant border cell migration defects. Expressing sdt or Patj RNAi using a border cell-specific Gal4 driver (slbo-Gal4) caused significant migration delay compared with the wild-type control (Fig. 1B, Fig. S1A,B). The migration index (MI) of the control is close to 1 (0.96), whereas those of sdt RNAi and Patj RNAi are 0.79 and 0.83, respectively (Fig. 1B). Importantly, multiple stocks for Patj RNAi (four) and sdt RNAi (three), which were obtained from different sources (see Materials and Methods), were tested and displayed similar migration delay (data not shown). Furthermore, immunostaining of egg chambers containing flip-out clones expressing Patj RNAi revealed that the Patj protein level was strongly reduced in the clones, indicating effective knockdown (Fig. S1C). Together, these results strongly support the conclusion that the phenotypes of Patj RNAi and sdt RNAi are due to loss of function of Patj and sdt, respectively.

The generation of mosaic border cell clusters containing homozygous crb^82-04 or crb^j1B5 mutant clones revealed more severe defects (MI=0.49, Fig. 1B, Fig. S1D; MI=0.68, Fig. S1E,F). Furthermore, sdt RNAi and Patj RNAi in the heterozygous background of sdt^K85/+ (Krahn et al., 2010) and Patj^△1/+ (Sen et al., 2012), respectively, resulted in more severe migration defects than the RNAi alone (MIs of 0.56 and 0.44, respectively; Fig. 1B). These results clearly indicate that the three Crb complex components are all required for border cell migration.

Staining with the F-actin-binding dye phalloidin revealed that those border cell clusters with migration delay typically displayed ectopic F-actin-enriched patches on the side and back of the clusters (Fig. 1C,D, Fig. S1G), indicating a reduction in polarized F-actin distribution at the front. Moreover, mosaic clusters containing two wild-type polar cells and all crb^82-04 border cells demonstrate that the phenotype of ectopic actin patches is autonomous to border cells (Fig. S2). Using a rapid dissection method, we were able to better preserve the dynamic actin structures and found that the actin patches resembled lamellipodial protrusions (Fig. S3). Indeed, live imaging of stage 9 egg chambers confirmed the presence of ectopic dynamic protrusions at the side and back of the clusters, in addition to the leading protrusion at the front (Fig. 1E,F, Movies 1-3). It is known that a normal front-back polarity for border cell clusters involves confining a predominant protrusion to the leading cell and suppressing protrusions in the side cells. Together, our results demonstrate that loss of function of Crb complex components disrupts such front-back polarity.

Upon closer examination, we found that Patj- or Sdt-deficient border cell clusters stained for the apical marker Baz exhibited a major defect in the positioning of the entire cluster with respect to the direction of migration. In the wild-type control, the border cell cluster was positioned in such a way that its apical-basal axis was always perpendicular to the migration direction (along the anterior-posterior axis) (Fig. 2A,B). This is presumably due to the fact that the outer border cells, especially the leading cells, almost always (90% of the time) extended protrusions at the lateral region of the outer cortex (Fig. 2C-E) (Montell et al., 2012; Veeman and McDonald, 2016). However, migrating border cell clusters expressing Patj RNAi or sdt RNAi were positioned such that their apical-basal axis was no longer at a 90° angle but at random angles to the anterior-posterior axis (Fig. 2A,B; data not shown). Furthermore, co-staining with Baz and phalloidin revealed that the ectopic protrusions referred to above were distributed randomly in the top, middle and bottom sections of the outside membrane (Fig. 2C-E), which roughly correspond to the apical, lateral and basal regions, respectively. This indicates that loss of apical polarity molecules leads to loss of exclusive and lateral localization of lamellipodial protrusions and hence to random positioning of the protrusions along the apical-basal axis, which in turn is likely to have caused the apical-basal axis of entire border cell clusters to appear at random angles relative to the migration direction. Such positioning could conceivably lead to inefficient collective migration of the cluster. Together, these results demonstrate that the Crb complex components Patj and Sdt are required for the polarized positioning of protrusions along the apical-basal axis.

The phenotype of ectopic protrusions that resulted from loss of Crb complex components resembles that of border cells defective in cell-cell communication (Ramel et al., 2013; Wang et al., 2010). Previous reports have shown that border cell clusters deficient in JNK signaling display defects in cell adhesions between border cells that results in loss of cell-cell communication, which was revealed using a method involving photoactivatable Rac (PA-Rac) (Llense and Martin-Blanco, 2008; Wang et al., 2010). Border cells expressing a PA-Rac transgene can respond to blue laser light and activate the exogenous PA-Rac. When Rac was photoactivated in the lagging border cell, ectopic protrusion was induced by the active Rac in the back of the cluster, and the leading border cell, which is at least one cell distance away, somehow sensed the communication from the cells at the back and side of the cluster and retracted the leading protrusion within 30 min (Fig. S4A) (Wang et al., 2010). It was reported that in border cells in which JNK signaling is downregulated, cell-cell communication is abolished and ectopic protrusions form in the back and side of the cluster (Wang et al., 2010), similar to clusters lacking the Crb complex components. It was previously proposed that the homotypic and transmembrane Crb might act in the apical junction between adjacent border cells to mediate cell-cell adhesion (Thompson et al., 2013). So, the Crb complex could in theory mediate non-autonomous communication between cells.

However, we found that photoactivation of Rac in the back of Patj RNAi border cell clusters resulted in induction of a new protrusion at the back, retraction of the leading and ectopic protrusions, and redirection of collective migration to the opposite direction (Fig. S4A-C). These results demonstrate that Patj RNAi clusters possess normal cell-cell communication, and thus that lack of cell-cell communication is not the cause of the ectopic protrusions and disruption in front-back polarity.

To investigate the cause of the disruption of front-back and apical-basal polarities in border cells lacking Crb complex components, we performed immunostaining for a variety of polarity and cytoskeletal markers. In the wild-type border cell cluster, Crb complex components such as Crb, Sdt and Patj, and Par complex components such as Par6, aPKC and Baz, were primarily localized in the apical junction between any two adjacent cells (Fig. 3A-F, Fig. S5), while a second pool of these molecules was distributed at moderate levels in the cytoplasm and near the outer cell periphery of each cell (Fig. 3A,B, Fig. S5). Furthermore, the second pool of apical complex components displayed a front-polarized distribution pattern and was often observed to be enriched in the leading protrusions (Fig. 3A-C, Fig. S6).

Loss of function of Sdt, Patj or Crb disrupted the apical localization of the Crb complex and Par complex components except for Baz (Fig. 3A-F, Fig. S7A; data not shown), supporting the previous proposal that the Crb complex and Par complex together act as one super-complex to regulate apical polarity and that depletion of one complex affects the apical distribution of the other complex (Tepass, 2012; Thompson et al., 2013). It is important to note that Patj RNAi caused strong disruption in the distribution of Crb and Sdt in the apical junctions (Fig. 3E,F). This indicates that Patj is required for the stability of the Crb complex in the apical junctions, suggesting that it is indeed a functional component of the Crb complex in border cells. Interestingly, the distribution pattern of Baz in the apical junction was not grossly affected, consistent with previous findings that Baz functions independently of the Par6-aPKC complex and is localized subapically below the aPKC apical domain in follicle cells (Morais-de-Sá et al., 2010). In addition, the subapical and lateral junctional distribution of E-cad and the lateral junctional distribution of Dlg were not grossly affected (Fig. 3E,F). Such loss of function of Crb complex components also increased the distribution of Crb and Par complex components throughout the cytoplasm in the form of large ectopic foci (Fig. 3E,F).

Among these components, we found that aPKC and Par6 displayed a unique ectopic distribution pattern, whereby they often colocalized with the aforementioned ectopic actin patches that were shown to be the ectopic protrusions (Fig. 3G,H, Fig. S9B). Such colocalization is specific since large internal actin patches induced by reduction of cofilin (an actin depolymerizing factor) displayed no colocalization with ectopic aPKC spots (Fig. S7B). The unique association between ectopic protrusions and aPKC and Par6 spots suggested that ectopically localized aPKC and Par6 caused the extra protrusions. To test this idea, we reduced the function of aPKC or Par6 in the background of Patj loss of function, and found that aPKC but not Par6 loss of function rescued the phenotype of ectopic actin patches (see Fig. 5A,B, Movie 4). Furthermore, knockdown of Par6 by itself still resulted in ectopic actin patches colocalizing with large aPKC spots (Fig. 3G,H), whereas knockdown or dominant-negative inactivation of aPKC (aPKC-DN) resulted in a very different phenotype, namely the virtual absence of non-leading protrusions or actin patches and the reduced length of leading protrusions (Fig. 4A,B, Movie 5). Together, these results demonstrate that aPKC is required for protrusion formation in the wild type and that mislocalized aPKC, as results from loss of Crb complex components, causes the formation of ectopic protrusions and the disruption of front-back and apical-basal polarities.

This conclusion raised the possibility that aPKC is sufficient for the formation of ectopic protrusions. To test this, we overexpressed wild-type and mutant forms of aPKC including the membrane-tethered aPKC-CAAX and the constitutively active aPKC-CA (with an N-terminus deletion) in border cells (Lee et al., 2006). This resulted in large ectopic protrusions as well as increased F-actin levels in these border cell clusters (Fig. 4B,C, Movie 5). Moreover, overexpressing active forms of aPKC in individual border cells within a cluster using the flip-out technique resulted in the autonomous generation of large protrusions (Fig. 4D). Lastly, expressing aPKC-CA in single-cell clones within an otherwise wild-type cluster revealed that a local or autonomous increase in aPKC activity in a single cell causes that cell to behave much more like a leading cell, as compared with a single cell with only GFP expression (Fig. 4E).

We found that overexpressing the active forms of aPKC (aPKC-CA and aPKC-CAAX) in border cells by slbo-Gal4 caused a reduction in the cell-cell contact area between adjacent border cells and an enlargement of the membrane area contacting the substrate nurse cells, resulting in each border cell appearing to stretch away from the central polar cells (Fig. 4F, Movie 5). This result clearly shows that a significant part of the inside membrane (Fig. 3D), which normally displays strong cell adhesion and little membrane dynamics, has taken on the features of outside membrane (Fig. 3D), which typically displays high membrane and actin dynamics. As a result, the border cells do not behave as a coherent cluster undergoing collective migration, but act more like cells migrating individually, even though they are held together by their adhesion to the central polar cells (Fig. 4F, Movie 5).

This phenotype resembles that of the crb mutant, in which individual border cells become more separated and ectopic actin patches form between adjacent border cells (Fig. 1C, Fig. 4G,H). Indeed, in border cells with Patj and Sdt reduction, ectopic actin patches that were colocalized with aPKC not only formed in the outside cell cortex but were also observed to sometimes form in areas near cell-cell contacts (Fig. 3G,H). By contrast, overexpressing Crb in the border cells resulted in substantially fewer actin-rich protrusions, little membrane dynamics and much reduced F-actin levels in the outside membrane (Fig. 4G,H), which appeared to take on the characteristics of inside membrane. Interestingly, reducing aPKC activity also resulted in similar phenotypes, including much reduced front protrusions, no side and back protrusions, and significantly reduced F-actin levels in the outside membrane (Fig. 4A,C,F,H). Taken together, these results suggest that the Crb complex is required to maintain inside-outside polarity, probably by restricting the actin-promoting function of aPKC to the outside membrane only.

Rac and its positive regulators are crucial for actin polymerization and protrusion formation in migratory cells including border cells (Heasman and Ridley, 2008; Montell, 2003; Montell et al., 2012). We examined whether they mediate the generation of ectopic protrusions that result from a lack of Crb components. Using a similar genetic approach to that mentioned above, we reduced the function of Rac in the background of Patj or Sdt loss of function (Fig. 5A,B; data not shown). We found that the ectopic actin patch phenotype was successfully rescued (Fig. 5A,B). Furthermore, using a previously reported Rac-fluorescence resonance energy transfer (FRET) sensor and live imaging (Wang et al., 2010), we observed that strong Rac activity was present in the ectopic protrusions (Fig. 5C-E), suggesting that increased Rac activity in the side and back of border cells was responsible for the ectopic protrusions.

Previous studies reported that Vav and Mbc (Myoblast city), both of which are guanine exchange factors (GEFs) and hence positive regulators of Rac, promote protrusion formation and migration (Bianco et al., 2007; Duchek et al., 2001; Fernandez-Espartero et al., 2013). We tested whether reducing the function of Vav and Mbc, as well as Sif, another Rac GEF and a Drosophila homolog of mammalian Tiam1 (Sone et al., 1997), could produce similar rescue as Rac. Knockdown of Sif but not of Vav or Mbc rescued the Patj RNAi phenotype (Fig. 5A,B, Movie 4). Since individual knockdown of Sif, Vav or Mbc each caused reduction in the length of leading protrusions and a migration delay of the border cell cluster (Fig. S8), the rescue by Sif of ectopic protrusions is specific. Together, these results demonstrate that Sif and Rac mediate the ectopic protrusions that result from lack of Patj.

It is known that the Par complex interacts with and activates mammalian Tiam1, and that aPKC interacts with and phosphorylates Tiam1 in cultured mammalian cells (Chen and Macara, 2005; Matsuzawa et al., 2016; Wang et al., 2012; Zhang and Macara, 2006). We tested whether aPKC could act through Sif to promote Rac activity and protrusions. First, we found that knockdown of Sif but not of Vav or Mbc in border cells could suppress the ectopic protrusions that resulted from overexpression of active aPKC (Fig. 6A,B, Movie 6). Moreover, Sif knockdown also suppressed the phenotypes of increased F-actin levels and delayed migration (Fig. 6B-D). By contrast, expression of a dominant-negative form of aPKC (aPKC-DN) failed to rescue the ectopic protrusion phenotype of Sif overexpression (Movie 7). Moreover, overexpressing Sif also mimicked the phenotypes of active aPKC overexpression (Fig. 6A,C, Movie 6). Together, these results indicate that aPKC acts genetically upstream of Sif. Second, a dominant-negative form of Rac (RacN17) could strongly suppress the phenotype of ectopic protrusions or increased F-actin levels that resulted from active aPKC or Sif overexpression, respectively (Fig. 6C,E-G). Overexpressing active aPKC resulted in a marked increase in Rac activity as detected by the Rac-FRET sensor (Fig. 6H,I), similar to that which resulted from Patj RNAi (Fig. 5C,D). Third, co-immunoprecipitation experiments demonstrated that Myc-tagged Sif (Myc-Sif) physically interacts with aPKC-GFP (Fig. 6J).

Lastly, we generated a transgene encoding HA-tagged Sif (HA-Sif) to determine its localization pattern in border cells (Fig. 6K,L). Staining with HA antibody revealed that HA-Sif is distributed in the cytoplasm and localized near the membrane, with a polarized distribution toward the front in the leading cells (Fig. 6K). Interestingly, most of the HA-Sif staining was localized in the basolateral region of border cell clusters, whereas the predominant population of membrane-bound aPKC is localized in the apical junctions, a region that is spatially segregated from the HA-Sif localization (Fig. 6L, middle and bottom). However, the non-apical population of aPKC near the outside membrane at the basolateral position displayed significant overlap with HA-Sif (Fig. 6L, top). This suggests that the aPKC that is localized in apical junctions could not interact with Sif strongly in the apical region due to the lower level of available Sif there. However, Sif overlap with the non-apical aPKC pool near the outside lateral membrane could promote the formation of protrusions there. Co-expression of HA-Sif and aPKC-GFP in Drosophila S2 cells showed that HA-Sif partially colocalizes with aPKC-GFP in both cell membrane and cytoplasm (Fig. 6M). Taken together, these results indicate that aPKC interacts with and acts through Sif near the outside membrane at the basolateral region to promote Rac activity and actin-based protrusions.

Immunostaining revealed that Crb complex components and Par complex components each displayed a distribution into two distinct populations (Fig. 3A-C). The major pool resided in the apical junctions, whereas the minor pool was present close to the outside and lateral cell membrane of border cells. Interestingly, some components, such as aPKC, Patj and Crb, even displayed a front-back asymmetry for the second, minor pool (Fig. 3C), and the front-polarized distribution of aPKC requires the function of the guidance receptor PVR (Fig. 7A,B). Live imaging also revealed that punctate spots of Par6-GFP (a genomic transgene driven by the par6 promoter), which is enriched in the apical junction, could be found trafficking from apical junction to the outside membrane, suggesting that there might be dynamic exchange between the two pools (Fig. S9A).

We attempted to decipher the mechanism underlying the distribution of the two distinct pools of apical components. Previous studies from our and other labs have shown that the endocytic recycling machinery is required for front-back polarity and the collective migration of border cell clusters (Assaker et al., 2010; Jékely et al., 2005; Laflamme et al., 2012; Wan et al., 2013). We tested whether vesicle trafficking could provide the means to generate the polarized distribution of these two pools. Interestingly, loss of function of the endocytic regulators dynamin and Rab5, the recycling regulator Rab11, and the exocytosis regulator Sec3 each resulted in severe disruption in the distribution of, and balance between, the two pools of apical polarity molecules (Fig. 7C-H). For instance, reduction of Rab11 or Sec3 caused dramatic disruption of Crb and aPKC at the apical junction, accompanied by a strong increase of their staining in the cytoplasm (Fig. 7C-F). Loss of function of Rab5 or dynamin (Shibire) resulted in a strong sharp line of membrane staining of Crb and aPKC or of Baz and aPKC, respectively, at the outside lateral membrane (Fig. 7G,H), which is very different from their diffuse staining pattern near the outside membrane of wild-type border cells (Fig. 7G,H, Fig. 3A-C). Moreover, the apical localization pattern of Crb, Baz and aPKC was also disrupted and their cytoplasmic stainings were increased. This result suggests that in the wild-type border cells, membrane-bound apical polarity molecules at the outside membrane need to be constantly trafficked to the cytoplasm by dynamin- and Rab5-mediated endocytosis. Together, these results demonstrate that the endocytic recycling machinery is required for the polarized distribution of two distinct pools of apical polarity proteins, but whether the localization of these polarity molecules is directly or indirectly regulated by endocytic recycling is unclear.

We then sought to determine whether polarized recycling and exocytosis were affected by Crb complex disruption. Our previous work indicated that Rab11-labeled recycling endosomes and Sec5-labeled exocysts displayed a front-polarized distribution near the leading protrusion, which depends on signaling from the guidance receptor PVR (Wan et al., 2013). However, in Patj RNAi border cell clusters (in which the Crb complex was disrupted; Fig. 3E,F), Rab11 and Sec5 staining was enriched in the ectopic protrusions, with no apparent polarized distribution at the front of clusters (Fig. 7I,J,M), implying that the locally enriched recycling endosomes and exocysts were somehow involved in mistrafficking of aPKC and hence the formation of ectopic protrusions (Fig. 7M). Indeed, locally enriched Rab11 staining was shown to partially colocalize with large and ectopically localized aPKC spots in Patj RNAi border cell clusters (Fig. S10). Reducing the function of Rab11 effectively suppressed the formation of ectopic actin patches or protrusions that resulted from Patj RNAi (Fig. 7K,L). Together, these results demonstrate that disruption of the Crb complex also affects the distribution pattern of recycling endosomes and exocysts and that redistribution of aPKC is likely to depend on recycling endosomes.

This study demonstrates that the Crb complex is required for the collective migration of border cells. Loss of function of Crb, Sdt or Patj each delayed border cell migration, which was likely to be a result of the combined effect of disrupting three distinct cell polarities (Fig. 7M). Most importantly, the front-back polarity of the border cell cluster was disrupted, as demonstrated by the ectopic formation of large actin-rich protrusions in border cells located at the side and back of the cluster (Fig. 7M). Furthermore, Patj RNAi or sdt RNAi caused border cell clusters to extend major protrusions at random angles relative to the apical-basal axis, unlike the wild-type clusters that restrict protrusion formation to the lateral region, thus extending the protrusions perpendicular to their inherent apical-basal axis (Fig. 7M). Such restriction of lateral protrusion formation would ensure that protrusions are parallel to the migration direction, resulting in efficient forward movement of the entire cluster. Mutation in crb or the expression of active forms of aPKC expanded the outside membrane area, and overexpressing Crb or reducing aPKC activity suppressed the outside membrane characteristics, causing disruption in inside-outside polarity for each border cell. Interestingly, crb mutant border cells sometimes exhibited ectopic actin patches (containing large aPKC spots) between the adjacent cells (Fig. 7M), where the inside membrane is normally located. Taken together, these results raise the following question: is there a common mechanism that is affected during the disruption of all three cell polarities? In other words, are these cell polarities interconnected and coordinated by the same mechanism?

We found that a common feature of loss of Crb complex components is that mislocalized aPKC generates ectopic Rac-dependent protrusions in border cells at the side and back of the cluster and at the apical and inside (junctional) region of individual border cells, leading to disruption of all three cell polarities. This indicates that there is a common mechanism involving aPKC that organizes all three polarities. First, the ectopic protrusions and the loss of these three polarities as a result of loss of Patj are likely to be mediated by the ectopically localized aPKC, since reduction of aPKC was able to rescue the ectopic protrusions. Interestingly, loss of other apical polarity proteins (Crb, Sdt, Par6, Cdc42) except for aPKC and Baz also led to similar phenotypes, including disrupted aPKC localization in the apical junctions, ectopic actin patches colocalized with large aPKC spots, and increased F-actin levels and Rac activity at or near the outside membrane. By contrast, loss of aPKC resulted in few protrusions and reduced F-actin levels at the outside membrane, while overactivation of aPKC led to increased F-actin levels and Rac activity, which are mediated by the downstream Sif. These results suggest that an important role of the Crb and Par complexes is to sequester most of the aPKC in the apical junction, leaving only a moderate level near the outside membrane to promote protrusion formation. The major pool of aPKC at the apical junction (together with Crb and Par complex components) is likely to function similarly to its classical role in epithelial cells, which is to promote apical polarity and integrity of apical and subapical junctions. However, the minor aPKC pool near the outside lateral membrane might function differently in that it can activate Sif to increase Rac-mediated actin dynamics. Such a difference might arise if complexes at the apical junction restrict or inhibit the Sif-promoting activity of aPKC. Conceivably, such inhibition would not apply to aPKC near the outside lateral membrane.

As summarized in our model (Fig. 7M), a crucial function of the Crb complex and Par complex is to produce a high level of membrane-bound aPKC at the inside apical junction and a moderate level of cytoplasmic aPKC near the outside lateral membrane so that the three distinct, but related, cell polarities can be properly established. Furthermore, polarized endocytic recycling of vesicles associated with aPKC and other apical polarity molecules ensures the polarized distribution of two aPKC pools within each border cell. Finally, it is interesting to note that the front-polarized recycling and exocytosis within the wild-type cluster, as mediated by PVF-PVR guidance signaling (Wan et al., 2013), could cause aPKC to be much more enriched at the outside membrane of the leading edge (to promote leading protrusion) than at the outside membrane at the side and back (to promote minor side protrusions) of the border cell cluster. When cells migrate collectively under developmental, physiological and pathological contexts, the migrating sheets or clusters of cells often display part-epithelial and part-mesenchymal characteristics. It will be interesting to determine whether aPKC together with Crb and Par complexes and the endocytic recycling machinery also play conserved roles in coordinating these three cell polarities in other types of collective migration.

Flies were cultured following standard procedures at 25°C, except for RNAi experiments at 29°C. All strains were obtained from the Bloomington Drosophila Stock Center (BDSC), National Institute of Genetics Stock Center (Japan), Vienna Drosophila RNAi Center, and Tsinghua University RNAi Stock Center (Table S1), except for the following: crb^82-04 (Ling et al., 2010), Patj^△7 (Zhou and Hong, 2012), DEcad::GFP (Huang et al., 2009), crb-HA (Huang et al., 2009), Patj^△1 (Sen et al., 2012), sdt^K85 (Krahn et al., 2010), UAS-aPKC-CAAX (Lee et al., 2006), UAS-aPKC-DN (Lee et al., 2006), UAS-aPKC-CA (Lee et al., 2006), UAS-aPKC (Lee et al., 2006), Sec3^PBac (Wan et al., 2013), UAS-RacFRET (Wang et al., 2010), UAS-PARacQ61L (Wang et al., 2010) and UAS-PARacQ61L-C450M (Wang et al., 2010). To generate the UAS-HA-sif transgenic line, a full-length cDNA of the sif gene was amplified from the genome of the UAS-sif fly stock (stock #9127, BDSC), then the PCR product was fully sequenced and subcloned into the pUAST-HA vector. The HA fragment was inserted at the N-terminus of the Sif protein. Then, the recombined vector was injected into embryos according to standard procedures. The primers for HA-sif were: sif-F1, 5′-AGCGCGTTACCACATAGATCTATGGGTAACAAACTGAGCTGC-3′; sif-R1, 5′-TCCTCTAGAGGTACCCTCGAGCTTAATTTTTCACATCGTCTTTGC-3′.

FRT clone was induced by hs-FLP. Heat shock was applied starting from late third instar larval or early pupal stage at 37°C for 2 h per day for 3 days. After eclosion, flies were raised on yeast-supplemented media for 2 days before dissection. To perform flip-out experiments, AyGal4 UAS transgenes were crossed to hs-FLP. Newly eclosed flies were heat shocked in a 37°C water bath for 5 min. For analysis of single-cell clones, mosaic clusters with only one border cell expressing GFP were used.

Ovary dissection was carried out in phosphate-buffered saline (PBS) and then fixed in devitellinizing buffer (7% formaldehyde) and heptane (Sigma) mixture (1:6) for 10 min (Zhang et al., 2011). After three washes in PBS, ovaries were incubated in PBT (PBS with 3% Triton X-100) and blocking solution (PBT, 10% goat serum) for 30 min and then stained overnight at 4°C. Fast dissection was performed in less than 5 min and involved fewer ovaries than the normal dissection method, which typically took ∼20 min. Primary antibodies were as follows: mouse anti-phospho-Tyr (4G10, 1:200, Millipore), mouse anti-Rab11 (1:200, BD Transduction Laboratories), mouse anti-Enable [5G2, 1:100, Developmental Studies Hybridoma Bank (DSHB)], rat anti-E-cad (DCAD2, 1:50, DSHB), goat anti-Arp2 (sc-11968, 1:10, Santa Cruz), mouse anti-Dlg (4F3, 1:100, DSHB), mouse anti-Crb (Cq4, 1:10, DSHB), mouse anti-Armadillo (N2 7A1, 1:50, DSHB), rabbit anti-PKCζ (C-20, 1:200, Santa Cruz), rabbit anti-phospho-Myosin II (3671S, 1:100, Cell Signaling), rat anti-HA (3F10, 1:100, Roche), rabbit anti-Baz (1:400, gift from A. Wodarz, University of Göttingen), mouse anti-Sec5 (1:50, gift from T. Schwarz, Boston Children's Hospital), mouse anti-Patj (1:800, gift from Yang Hong, University of Pittsburgh) and rabbit anti-Sdt (1:500, gift from E. Knust, Max Planck Institute, Dresden). Methanol treatment was used before anti-Crb staining as described (Niewiadomska et al., 1999). After washes in PBT, ovaries were incubated with secondary antibodies (1:200, Jackson ImmunoResearch) for 2 h at room temperature. F-actin was labeled by Rhodamine-phalloidin (1:200, Sigma). Confocal images were obtained using a Leica TCS SP5 II, an Olympus FV1200, or a Zeiss 880 microscope (with Airyscan technology) and images were processed by ImageJ, Imaris (Bitplane) and MATLAB (MathWorks) software.

Egg chambers were dissected from ovaries and mounted for live imaging as described previously (Prasad et al., 2007). Rac-FRET imaging and analysis were performed as described by Wang et al. (2010). CFP and YFP images were acquired with the Zeiss 880 confocal microscope and processed by ImageJ. A Gaussian smoothing filter was applied to both channels, with background subtracted. The CFP image was then used to create a binary mask with the background set to zero. The final images were generated from YFP/CFP ratios. The FRET index was calculated for the entire border cell cluster by measuring the average intensity of FRET. Heat maps of FRET indices were generated with MATLAB. Rac photoactivation was performed as described previously (Wang et al., 2010). To photoactivate, the 458 nm laser was set at 8% laser power and illuminated in a 7 μm diameter spot. The photoactivation scan lasted ∼30 s. The border cells were then imaged using a 568 nm laser. This series of steps was repeated for the duration of the time-lapse experiment. Live imaging of Lifeact-GFP was carried out using a Leica TCS SP5 II confocal microscope equipped with an HyD detector. Live imaging of Par6-GFP was performed using an Olympus FV1200 with a GaAsP detector.

Border cells were labeled by expression of UAS-GFP using the slbo-Gal4 driver. To quantify border cell migration, stage 10 egg chambers were used. Depending on the positioning of the cluster along the migratory route (between the anterior tip of the egg chamber and the oocyte border), the extent of border cell migration was measured and categorized into five classes: 0% (no migration), 25%, 50%, 75% and 100% (arriving at border). Migration index (MI) was then calculated according to the following formula to evaluate migratory ability, as described previously (Assaker et al., 2010): |$\hbox{MI} = [\hbox{1} \times n(100\% ) + 0.\hbox{75} \times n(75\% ) + 0.5 \times n(50\% ) + $| |$0.\hbox{25} \times n(25\% ) + 0 \times n(0\% ) + 0.\hbox{5} \times n(\hbox{dis})]/ N.}\!\!\!$| For example, n(75%) represents the number of stage 10 egg chambers in which border cell clusters were found at 75% of the migration distance. n(dis) represents the number of border cell clusters that were disassociated and not coherent. Disassociated clusters were very rare in our experiments. N is the total number of stage 10 egg chambers examined.

For fixed samples, a z-series of confocal sections was taken for each border cell cluster. After checking through the z-series, the confocal section with the longest protrusion visible was selected and measurement was performed for that protrusion. The length of phalloidin-labeled protrusions was determined by the distance from the protrusion tip to the boundary between the basal region of the protrusion (enriched with phalloidin staining) and cell body (much less phalloidin staining), as described previously (Zhang et al., 2011). Those longer than 2 μm and with significant width were qualified as protrusions and used for calculation. To determine the distribution of actin patches or protrusions, border cell clusters were divided into four quadrants: one front, two middle, and one back quadrant. Protrusions extended from the front quadrant were defined as leading protrusions, whereas those extended from other regions were considered non-leading protrusions. To determine the position of actin patches or protrusions along the apical-basal axis, Baz staining was used to label the apical side of the border cell cluster. Confocal z-stacks of border cell clusters were acquired and processed by Imaris for 3D reconstruction, from which protrusion position along the apical-basal axis was determined.

For live imaging, Lifeact-GFP was used to label actin-enriched protrusions (Riedl et al., 2008). A z-series of ∼25 confocal sections was taken every 2 min for each border cell cluster, and these confocal images were subjected to maximum projection. The length of the protrusions was measured and determined from the final projected image. Protrusions at least 3 µm in length (from base to tip) and 3 µm in width at the base were counted. The extension lifetime of protrusions was determined from the onset of extension to the final moment of disappearance (complete retraction). The full 2-min interval would be included if the protrusion disappeared during the interval.

To be considered ectopic, a protrusion (at least 3 µm from base to tip and 3 µm in width at the base, according to the Lifeact-GFP labeling) extended from side or back positions. When two or more protrusions were situated at similar positions around the front of the cluster, the larger one was designated the leading protrusion. If two protrusions were similar in size, that situated more toward the front or leading position was designated the leading protrusion.

The ectopic actin patches are large and intense phalloidin-stained patches at least 4 µm² in area (as measured by ImageJ). We do not count leading lamellipodial protrusions that contain fine actin structures (with an elongated lamellipodial shape) in wild-type and mutant clusters as ectopic actin patches for quantification purposes.

To determine the alignment of the entire border cell cluster with respect to the anterior-posterior axis of the egg chamber, Baz staining was used to label its apical side. A z-series of confocal sections was taken for each cluster with 0.4 μm between successive sections. The images were then reconstructed and analyzed in Imaris. The reconstructed images were rotated along the anterior-posterior axis to facilitate measurement of the angle between the apical-basal axis of the border cell and the anterior-posterior axis.

For Fig. 1B, all 19 mosaic clusters that we used are mutant (homozygous crb^82-04) for all the outer border cells, and 14/19 (74%) of mutant clusters demonstrate various degrees of migration delay (Fig. S1D). However, migration defects could also be observed when mosaic clusters contained clones of two or more mutant border cells (data not shown).

To measure the fluorescence intensity of phalloidin staining, w¹¹¹⁸ (considered as wild type) egg chambers without GFP label were mixed with control or other genotypes that were labeled with GFP in the same vial for phalloidin staining. Average fluorescence intensity (FI) of w¹¹¹⁸ border cell clusters and FI of the genotype of choice was measured by ImageJ, and the normalized FI was determined by: FI (genotype)/FI (w¹¹¹⁸).

To measure the front/back ratios for Fig. 3A,C, Fig. 6K and Fig. 7A,B, an area around the leading edge of the cluster was chosen as the front region, and an area including the lagging end was chosen as the back region. Average FI was measured by ImageJ for each region; the front/back ratios were calculated as: front FI/back FI. slbo-Gal4;UAS-GFP was chosen as the GFP control for Fig. 3A,C and Fig. 6K, and showed no front/back bias (ratio ∼1).

Drosophila S2 cells were transfected with UAS-Myc-sif or UAS-aPKC-GFP, or both, and cell lysates used for the immunoprecipitation assay. To generate the UAS-Myc-sif construct, a full-length cDNA of the sif gene was amplified from the genome of the UAS-sif fly stock (stock #9127, BDSC), then the PCR product was fully sequenced and cloned into the pUAST-Myc vector. The Myc fragment was inserted at the N-terminus of the Sif protein. Primers for Myc-sif were: sif-F2, 5′-GAGCAGATCTGCGGCCGCGGCATGGGTAACAAACTGAGCTGC-3′; sif-R2, 5′-CCTTCACAAAGATCCTCTAGTTAATTTTTCACATCGTCTTTGC-3′. To generate the UAS-aPKC-GFP construct, the full-length cDNA of aPKC was amplified from the FI03288 clone [Drosophila Genomics Resource Center (DGRC)] and then subcloned into the pUAST vector. Coding sequence of GFP was inserted into the C-terminus of aPKC. Primers for aPKC-GFP were: aPKC-F, 5′-CGGAATTCATGCAGAAAATGCCCTCGCA-3′; aPKC-R, 5′-ACAGAGACCTCCTAACGCAGTCTAGAAG-3′; GFP-F, 5′-TGGGAATTCGTTAACAGATCTATGGTGAGCAAGGGCGAGGA-3′; GFP-R, 5′-TCCTCTAGAGGTACCCTCGAGTTACTTGTACAGCTCGTCCATGC-3′.

S2 cell lysates were analyzed by immunoblotting with the following primary antibodies: rabbit anti-GFP (FL, 1:500, Santa Cruz, SC-8334), mouse anti-c-Myc (9E10, 1:500, Santa Cruz, SC-40), rabbit anti-PKCζ (C-20, 1:500, Santa Cruz). Secondary antibody was HRP conjugated (Vazyme). For immunoprecipitation, lysates were incubated with 5 ml primary antibodies overnight at 4°C. Then, 50% protein A/G agarose beads (Protein A/G PLUS-Agarose, Santa Cruz, SC-2003) working solution (in PBS) was added to the lysate and rocked for 4 h at 4°C. Lysates and immunoprecipitates were then immunoblotted.

Statistical analyses were performed with GraphPad Prism, version 5.01. Statistical comparisons of means were made using the unpaired two-tailed Student's t-test. P<0.05 was considered statistically significant.

We thank Yang Hong for fly stocks and critical reading of the manuscript; Michael P. Krahn, Shian Wu, Cheng-yu Lee, the Bloomington Drosophila Stock Center, Tsinghua University RNAi Stock Center, National Institute of Genetics Stock Center (Japan), and Vienna Drosophila RNAi Center for fly stocks; Andreas Woodarz, Tom Schwarz, Yang Hong and Elisabeth Knust for antibody reagents.