Guidance receptor promotes the asymmetric distribution of exocyst and recycling endosome during collective cell migration

During chemotaxis, migrating cells are often guided by external guidance cues, which are probably distributed in a graded fashion along the migratory route. Upon sensing such an extracellular gradient, the guidance receptors of migratory cells signal intracellularly to convert a usually shallow external gradient of guidance cues into a robust intracellular front-back polarity, which is essential for chemotactic migration of individual cells (Devreotes and Janetopoulos, 2003). However, cells do not always migrate individually; they often migrate collectively as a cluster, a sheet or a strand of cells in physiological, developmental and cancer metastatic conditions (Friedl and Gilmour, 2009; Rørth, 2011). During collective cell migration, each group of either tightly connected or loosely associated cells also needs to interpret an external gradient of chemotactic factors and convert it into a strong front-back asymmetry across a collective group rather than an individual cell, which is often manifested as one cell in the leading edge extending a protrusion and leading the rest of group forward (Friedl and Gilmour, 2009; Rørth, 2011). Both single cell migration and collective migration require a strong amplification process that results in an asymmetric distribution of cytoskeletal and adhesive molecules within an individual cell or within a group of cells, but the underlying mechanism is not clearly understood. Recently, evidence began to emerge pointing to intracellular trafficking as a probable mechanism for such amplification.

Endocytosis is a common mechanism employed by eukaryotic cells to internalize extracellular materials, receptors and other proteins and lipids in the plasma membrane (Grant and Donaldson, 2009; Maxfield and McGraw, 2004). The endocytic vesicles containing these cargoes first fuse with the early endosome or sorting endosome, where the proteins and lipids are then sorted for transport via different routes. They are either transported to the late endosome and then lysosome for degradation or recycled back to the plasma membrane via the recycling endosome. Previous studies have shown that the endocytic pathway is crucial for both individual and collective cell migration. For example, tissue cell culture experiments demonstrated that during single cell migration, internalization and recycling of integrin not only promotes the dynamic disassembly and assembly of adhesive contacts, but also restricts the localization of integrin to the front of the cell (Caswell and Norman, 2008; Caswell et al., 2007; Ulrich and Heisenberg, 2009). This is probably achieved when localized endocytic recycling of integrin at the leading edge outcompete the slow lateral diffusion of integrin toward the cell body.

Similarly, during collective migration of Drosophila border cells, endocytic recycling was shown by several studies to be required for restricting phosphotyrosine [pTyr, a measurement of receptor tyrosine kinase (RTK) activity] to the leading edge of each migrating cell cluster (Assaker et al., 2010; Janssens et al., 2010). During Drosophila oogenesis, a coherent cluster of about eight border cells delaminate from the anterior follicle epithelium and migrate collectively through the germline-derived nurse cells of a stage-9 egg chamber (Fig. 1A). The stereotypic and directed migration of border cells are guided by several external guidance cues secreted from the posteriorly located oocyte, acting on two RTKs, PDGF/VEGF receptor (PVR) and Epidermal growth factor receptor (EGFR), to transduce guidance signaling. It has been shown that dynamin and other regulators of the early steps of endocytosis, including a guanine exchange factor for Rab5 (promoting cargo entry into early endosome), were involved in localizing pTyr to the front of border cell clusters (Jékely et al., 2005). In addition, Rab11 (a positive regulator of recycling endosome) and its effector, the exocyst component Sec15, were also shown to be required for the asymmetric localization of pTyr (Assaker et al., 2010; Janssens et al., 2010).

Based on these studies, it had been suggested that there might be a directional recycling of active RTKs to the leading edge of border cells, amplifying the initial small difference between RTK signaling levels at the front and those at the back (Janssens et al., 2010; Jékely et al., 2005; Montell et al., 2012). Recently, it was shown that the active (tyrosine phosphorylated) form of PVR was enriched at the front of cluster when PVR is overexpressed, but the front-back asymmetry of active PVR is significantly less than that of pTyr. This discrepancy raises the question of whether kinases and phosphorylated proteins other than active PVR are directionally recycled to the front. Moreover, how this recycling process is regulated is unclear. In this study, we showed that there was a polarized distribution of Rab11-labeled recycling endosome and Sec5-labeled exocyst near the leading edge of migrating border cells, which is mediated by the guidance receptor PVR. Interestingly, we found PVR signaled through the small GTPase Rac to locally increase recycling endosome levels. Elevated recycling endosome levels were proximal to strong F-actin stainings, which normally occurred at the leading edge of the migrating cluster, suggesting that directional recycling could be mediated by the asymmetrically localized F-actin. Conversely, the exocyst component Sec3 was also required for asymmetric localization of RTK activity and F-actin. These results suggested the presence of a positive feedback loop in border cells, where RTKs such as PVR acted to induce a higher level of vesicle recycling and tethering activity in the leading cells, enabling active RTKs or other phosphorylated proteins to be concentrated in a more polarized fashion at the front. Lastly, we also provided evidence that Sec3 promoted recycling and tethering of vesicles containing E-cadherin (Shotgun - FlyBase) in border cells and follicle cells.

In a search for genes essential for border cell migration, we screened a collection of flippase (Flp) recognition target (FRT)-P-element lethal mutations described previously (Chen et al., 2005). Using genetic mosaic methods (Flp-FRT), we identified a P-element line, 10199, that displayed border cell migration defects. Detailed genetic analyses including P-element excision, deficiency mapping and DNA sequencing, had uncovered that the defects were not caused by the P-element insertion but by a point mutation in the CG3885 locus, which encodes the Drosophila homolog of Sec3, an exocyst component that had not been genetically characterized in Drosophila (Boyd et al., 2004; Finger et al., 1998; He and Guo, 2009). The exocyst is a complex composed of eight subunits, which were shown to mediate the tethering of secretory or recycling vesicles to the plasma membrane before the step of exocytic fusion during various polarized cellular processes in yeast, Drosophila and mammals (Finger et al., 1998; He and Guo, 2009; Langevin et al., 2005). The isolated allele was named sec3^GT because of the G to T mutation that disrupts the stop codon (supplementary material Fig. S1A), and it failed to complement a lethal PiggyBac insertional allele of sec3 (sec3^PBac). Both alleles were shown to be loss-of-function alleles, as the homozygous mutants died at third instar larval stage and displayed strong reduction of sec3 RNA levels compared with the wild type (supplementary material Fig. S1). A sec3 cDNA construct expressed by UAS/Gal4 system was able to rescue the larval lethality for both alleles, indicating they were bona fide sec3 mutations (supplementary material Table S1). Lastly, expressing sec3 RNAi in the border cell clusters by c306-Gal4 produced similar migration defects (Fig. 1D,F).

Detailed phenotypic analyses showed that mosaic border cell clusters containing clones mutant for either sec3^GT or sec^PBac displayed significant migration delay. In wild-type stage-10 egg chambers, almost all (97%) of the border cell clusters reached the oocyte border, whereas 35% (sec3^GT) or 43% (sec^PBac) of the mutant mosaic border cell clusters failed to reach the border by stage 10 (Fig. 1B,C,F). A significant percentage of these delayed border cell clusters stopped at the starting point of the migration route (Fig. 1F). Furthermore, we found that within the mosaic border cell clusters sec3 mutant cells tended to stay at the lagging end, whereas the heterozygous border cells tended to stay at the leading edge (Fig. 1E,E′,G). By contrast, increasing sec3 expression in a single border cell enabled it to become much more likely to be a leading cell within a cluster (supplementary material Fig. S2). Together, these data indicate that Sec3 promotes both the collective migration of border cell clusters and the leading ability of individual border cells.

A prominent feature of guided migration of a border cell cluster is the presence of a polarized lamellipodial protrusion that is rich in F-actin and is extended by one or two border cells at the leading edge. Such a polarization of F-actin within the border cell clusters was lost in sec3 mutant or RNAi knockdown clusters, and large lamellipodial protrusions were rarely observed at the leading edge (Fig. 2A-C). This result suggests that the overall front-back polarity of border cell clusters was affected. Indeed, sec3 mutant or RNAi border cell clusters displayed an absence of front-back asymmetry of pTyr staining, which has been validated as a reliable local readout of endogenous RTK activity levels within border cell clusters (Assaker et al., 2010; Jékely et al., 2005). In these clusters, pTyr staining was somewhat uniform, whereas it was strongly enriched at the front of the wild-type border cell cluster (Fig. 2A,B,D). Consistently, recent work had shown that front-back polarity of pTyr staining was similarly disrupted in sec15 RNAi border cells, and RNAi knockdown of sec15 and three other genes encoding some of the subunits (sec5, sec6, sec8) of exocyst complex also displayed migration defects (Assaker et al., 2010). It should be noted that front-back asymmetry of F-actin and pTyr is most prominent during early phase of border cell migration, during later phase this asymmetry is moderate but still significant (supplementary material Fig. S3A′ and Fig. S4B,D), hence the above data have all been collected on clusters initiating migration. Together, these data indicate that Sec3 or the exocyst complex as a whole is required for front-back polarity of RTK signaling and F-actin in border cell clusters.

Previous works had suggested that endocytic recycling was involved in enriching pTyr in the leading edge of border cell clusters (Assaker et al., 2010; Janssens et al., 2010). As components of exocyst have been shown to interact with Rab11 (Beronja et al., 2005; Langevin et al., 2005; Wu et al., 2005; Zhang et al., 2004), a major marker of recycling endosomes, we decided to determine the distribution patterns of exocyst and Rab11 within the border cell clusters. A Rab11-GFP transgene (expressed by slbo-Gal4) has been previously reported to display no obvious front bias in border cells (Assaker et al., 2010). However, we found that a specific antibody against Rab11 (shown to be specific for endogenous Rab11 in supplementary material Fig. S3A-D′) revealed a clear front-back asymmetry of endogenous Rab11 within border cell clusters (Fig. 3A,A′); the front-back ratio was measured to be 1.80±0.12 (n=16) by fluorescence quantification (Fig. 3I). In addition, a different Rab11-GFP transgene (different from the above one and obtained from the Bloomington Stock Center, stock# 8506) expressed in border cells by slbo-Gal4 gave a similar asymmetric staining pattern as the Rab11 antibody (supplementary material Fig. S4), confirming the antibody staining result. Note that all front-back ratios described here were measured for clusters undergoing the first phase of migration (0-50% position); a lower but significant asymmetry for Rab11 and Rab11-GFP was also detected for clusters during the second half of migration (50-100%; supplementary material Fig. S4). Interestingly, immunostaining using a well-documented Sec5 (an exocyst subunit) (Beronja et al., 2005; Langevin et al., 2005; Murthy et al., 2003; Murthy et al., 2010) antibody, also uncovered a similar front-back ratio of 1.80±0.15 (n=16) in its distribution pattern (Fig. 3E,E′,J). Both Rab11 and Sec5 were shown to be enriched near the leading edge (Fig. 3A′,E′).

We next tested the possibility that the asymmetry of Rab11 and Sec5 was regulated by the guidance signaling. Remarkably, front-back ratios of both Rab11 and Sec5 were reduced to 1.13±0.03 (n=16) and 1.0±0.03 (n=15), respectively, in border cells expressing the dominant-negative forms of both PVR and EGFR (DN-PVR + DN-EGFR, Fig. 3B′,F′,I,J) by slbo-Gal4, which effectively removed directional guidance in border cells (Duchek et al., 2001; Prasad and Montell, 2007). Furthermore, expressing DN-PVR alone resulted in a strong reduction of front-back ratios for both Rab11 (1.25±0.04, n=13) and Sec5 (1.17±0.09, n=13), whereas DN-EGFR expression resulted in no significant reduction (Fig. 3C′,G′,I,J), suggesting that PVR is the major RTK that mediated their polarization. Interestingly, expressing a constitutive active form of PVR (λ-PVR) (Duchek et al., 2001) in all outer border cells also caused a remarkable reduction of front-back ratios for Rab11 and Sec5 (Fig. 3D′,H′-J), accompanied by a complete block of migration (data not shown) (Duchek et al., 2001; Zhang et al., 2011); however, the overall levels of Rab11 (but not Sec5) were significantly elevated (Fig. 3D′,H′; Fig. 4K; quantification data not shown for Sec5). By contrast, λ-EGFR (Queenan et al., 1997) expression resulted in no effects on the front-back ratio or the overall levels of Rab11 (Fig. 3I,J; Fig. 4K). Lastly, to show that this loss of asymmetry is specific for disruption of guidance and not merely a consequence of just any migration delay, we demonstrated that deficiency in either Rho1 (the RhoGTPase) or Par-1 (epithelial polarity regulator) resulted in no reduction of Rab11 asymmetry but significant migration delay (Fig. 4J,M; supplementary material Fig. S5). Taken together, these results suggested that the guidance signaling in response to graded chemotactic factors caused a polarized vesicle recycling and tethering activity.

The above data on λ-PVR suggest that PVR signals increase recycling endosome levels. To further test this hypothesis, we overexpressed PVR or λ-PVR in follicle cells by slbo-Gal4, which drove Gal4 expression not only in all the migratory border cells (outer border cells, excluding the central nonmigratory polar cells) but also in a subset of centripetal, posterior and a few sporadically positioned follicle cells (Jékely et al., 2005). Interestingly, Rab11-labeled recycling endosome was strongly upregulated in all follicle cells expressing λ-PVR (Fig. 4B-B′,L), consistent with that of the Rab11 staining increase in λ-PVR-expressing border cells (Fig. 3D′; Fig. 4K). Likewise, PVR overexpression caused a milder but still significant increase in a few posterior follicle cells, and the increase in Rab11 staining was proximal to the basal region of follicle epithelium (Fig. 4E,E′). As the overexpressed PVR also had a basal junctional localization (Fig. 4E′) (Jékely et al., 2005), the close proximity between PVR and Rab11 suggests that PVR sends a signal to locally increase the recycling endosome levels. By contrast, stainings for Sec5 and Rab5 (early endosome marker, supplementary material Fig. S3E) were not affected in λ-PVR-expressing follicle cells (Fig. 4C-D′), indicating that PVR specifically affects recycling endosome.

To test whether λ-PVR increases Rab11 expression or promotes only the formation of recycling endosome, we expressed UAS-Rab11-GFP in the border cells using slbo-Gal4. We found that λ-PVR strongly increased the total levels of Rab11-GFP in border cells and follicle cells (supplementary material Fig. S6), in the same way that it increased endogenous Rab11 stainings. This indicates that the increase in Rab11-GFP levels is unlikely through transcription upregulation by PVR signaling, as the UAS-Rab11-GFP is only subject to transcriptional regulation of the yeast Gal4 protein. This result suggests that ectopic PVR signaling does not upregulate the expression level of Rab11, but it promotes the formation of recycling endosomes, which are manifested as strong aggregates or large masses that are labeled by intense Rab11 stainings or strong Rab11-GFP signals (supplementary material Fig. S6B′,D′; Fig. 3D′; Fig. 4B′). This enlargement of Rab11-stained structures is consistent with a recent report, which showed that increase of Rab11 activity resulted in enlargement of Sec15-labeled vesicles in the border cells (Laflamme et al., 2012).

We next sought to identify the downstream signal of PVR that mediates this effect. One candidate is the small GTPase Rac, which had been previously shown to act genetically downstream of PVR to promote actin polymerization and lamellipodial protrusion in both follicle cells and border cells (Duchek et al., 2001). Interestingly, reducing the function of Rac by co-expressing a dominant-negative form of Rac (RacN17) (Geisbrecht and Montell, 2004; Zhang et al., 2011), in the λ-PVR expressing background, dramatically rescued the phenotype and reverted the high levels of Rab11 back to almost wild-type levels and pattern (Fig. 4G-G′,K,L). This rescue is specific and not due to an extra copy of UAS transgene, as co-expression of UAS-lacZ transgene has no rescuing effect (Fig. 4L). Furthermore, expression of RacN17 alone did not reduce overall levels of Rab11, but significantly reduced its front-back ratio (Fig. 4I,K,M). By contrast, increasing the activity of Rac by expressing a constitutive active form of Rac (RacV12) (Duchek et al., 2001) elevated the levels of Rab11 in border cells and follicle cells (Fig. 4G-H′,K,L). We found that after RacV12 expression or λ-PVR expression, 82% (n=94) or 90% (n=88) of GFP-labeled follicle cells, respectively, had significantly higher Rab11 levels.

A noted feature from λ-PVR (or RacV12) overexpression in both border cells and follicle cells is the strong increase in actin polymerization, and RacN17 co-expression suppressed both the F-actin and Rab11 staining increases. These data raised a possibility that actin polymerization is directly responsible for the elevated recycling endosome levels in both border cells and follicle cells. To test this idea, we treated wild-type egg chambers with an F-actin destabilizing drug Latrunculin A (Lat-A) to reduce F-actin levels (Zhang et al., 2011). Stage-9 egg chambers incubated with control medium for 1 hour displayed a normal morphology and wild-type phenotype, including F-actin and Rab11 enrichment at the leading edge of border cell cluster (Fig. 5A-A′). Interestingly, incubation with Lat-A (2 μM) containing medium for 1 hour resulted in a loss of front bias of Rab11 staining while causing a dramatic reduction of F-actin levels in border cells (Fig. 5B-B′). Moreover, reducing the actin depolymerizing activity by knocking down the Cofilin (Actin depolymerizing factor; Twinstar - FlyBase) (Zhang et al., 2011) encoding gene tsr in the border cells and follicle cells resulted in ectopic strong F-actin regions, which also led to a strong increase in the Rab11 levels (Fig. 5D-E′), further demonstrating that F-actin levels directly affected recycling endosome levels. Interestingly, the elevated Rab11 stainings or Rab11-GFP signals were often proximal to or partially overlapping the strong F-actin stainings. Such a complementary or overlapping pattern was often observed around the leading edge of wild-type border cells (Fig. 5A-A′; supplementary material Fig. S4B and Fig. S6A′), and near the ectopic F-actin regions within the λ-PVR, RacV12 or tsr RNAi expressing follicle cells and border cells (Fig. 4B′,G′,H′; Fig. 5D-E′; supplementary material Fig. S6B′,D′). Lastly, we treated the λ-PVR expressing egg chambers with Lat-A for 1 hour but found no significant rescuing effect (data not shown). The difference in results from the RacN17 rescue and the Lat-A treatment is likely to be due to the difference in the rescue timings. RacN17 is expressed at the same time as λ-PVR expression, whereas Lat-A treatment is applied several hours later (than λ-PVR expression), when the strong inducing effect from λ-PVR has been fully manifested and the massive recycling endosome structure cannot be easily reverted back to its normal formation. Taken together, our results showed that PVR signaled through Rac to strongly induce recycling endosome formation in border cells, probably in an F-actin-dependent manner.

Next, we sought to understand what kinds of cargo are transported by the Rab11-labeled vesicles and what role exocyst plays in regulating transport of these vesicles during border cell migration. Previous studies have shown that the exocyst promotes polarized recycling of epithelial polarity molecules such as E-cadherin onto the plasma membrane of epithelial cells, probably via the vesicle-tethering function of the exocyst (Blankenship et al., 2007; Classen et al., 2005; Langevin et al., 2005). E-cadherin is not only an essential component of the subapically localized adherens junction in follicle epithelium but also a major adhesive molecule required for the adhesion of border cells to the substrate (nurse cells) and for their migration (Niewiadomska et al., 1999). It was previously reported that E-cadherin pattern is disrupted by Rab11 deficiency in border cells (Cobreros-Reguera et al., 2010). To determine whether exocyst regulates recycling of E-cadherin or other epithelial polarity proteins during border cell migration, we carried out a detailed phenotypic analysis of sec3 mutant border cells. Mosaic border cell clusters containing sec3 homozygous mutant clones showed that cytoplasmic dots of E-cadherin staining appeared enlarged and irregularly shaped compared with the wild-type border cells and were strongly elevated in levels in 15% of mutant clones examined (60 mosaic clusters examined; Fig. 6A-B′). A high proportion of these enlarged and irregular structures either colocalized or overlapped with Rab11-labeled structures, which were also strongly upregulated and appeared enlarged and irregular as compared to the wild type (Fig. 6A-B′). By contrast, none of the sec3 mutant clones displayed significant increase or disruption in stainings for Baz/Par-3 (n=48; Fig. 6D,D′) and Dlg1 (n=45; Fig. 6E,E′), which are polarity molecules associated with apical and lateral junctions, respectively, and are involved in regulation of border cell migration (Goode and Perrimon, 1997; Pinheiro and Montell, 2004; Szafranski and Goode, 2004). Furthermore, sec15 mutant border cells displayed a similar phenotype to that of the sec3 mutant (supplementary material Fig. S7A-A′). Likewise, follicle epithelium containing sec3 or sec15 mutant clones also exhibited a dramatic simultaneous increase in both E-cadherin and Rab11 stainings at the subapical cytoplasmic region (15 of 84 sec3 clones, Fig. 6C-C′; 12 of 30 sec15 clones examined, supplementary material Fig. S7B-B′). Upon closer examination, the colocalization pattern between the two highly elevated stainings varies from almost total colocalization (supplementary material Fig. S7B-B′,D-D′) to strong overlap (Fig. 6C-C′). Consistent with the border cell data, the distribution pattern and amount of Baz/Par-3 (n=62) and Dlg1 (n=57) were not affected in the sec3 mutant follicle cells (Fig. 6F-G′). To determine whether the colocalization of E-cadherin is specific only to recycling endosome and not to other compartments, Golgi, early endosome and late endosome were stained with Lva (Lava lamp), Rab5 and Hrs, respectively. We found no significant colocalization between the highly elevated E-cadherin staining and Lva, Rab5 or Hrs staining (supplementary material Fig. S8; data not shown). Together, the above data show that E-cadherin is a specific cargo transported by the recycling endosome in border cells and follicle cells.

Lastly, we tested whether Sec3 promoted the tethering and delivery of E-cadherin-containing vesicles onto the plasma membrane, as expected for a functional component of the exocyst complex. We utilized a previously reported assay (Langevin et al., 2005) that detected only the membrane-bound E-cadherin but not the intracellular E-cadherin pool in the follicle epithelium. Stage-9 or -10 egg chambers were incubated with the DCAD2 antibody that detects an epitope in the extracellular region of E-cadherin, and the incubation was performed at 4°C for 30 minutes to prevent endocytosis of the antibody. After incubation, fixing (cell permeabilization) and staining with a secondary antibody would allow visualization of the pool of E-cadherin delivered onto the lateral membrane only. Indeed, within a sec3^GT mosaic follicle epithelium, sec3 mutant follicle cells displayed a significant decrease of membrane E-cadherin staining compared with the adjacent wild-type cells (Fig. 6H-H′). Taken together, these results indicated that exocyst promoted the tethering and delivery of E-cadherin-containing vesicles onto the plasma membrane and the deficiency of exocyst function probably blocked such delivery and caused an accumulation and enlargement of Rab11-labeled recycling endosomes, which contained E-cadherin as a specific cargo during both border cell migration and follicle cell morphogenesis.

It has been proposed that repeated cycles of endocytosis of RTKs (or active RTKs) and recycling of them back to the membrane would effectively concentrate active RTK in the front of the migrating border cells (Assaker et al., 2010; Jékely et al., 2005; Montell et al., 2012). However, if the levels of endocytic recycling remain uniform in all the outer border cells during migration, a fast amplification of RTK activity levels between front and back would be difficult to achieve. Here, we show that there is a polarized endogenous distribution of the recycling endosome and exocyst in the leading border cells within the migrating cluster, which could conceivably make such amplification faster and more efficient in the leading cells. It was also shown previously that Sec15-GFP has an asymmetric localization at the front, when it is overexpressed in border cells (Assaker et al., 2010). Along their migrating route, the border cells often tumble or rotate as a cluster, resulting in position changes such as front cells becoming lateral and back cells and vice versa (Bianco et al., 2007). In such a scenario, a fast and robust amplification process would be essential to relocalize active RTKs. Indeed, we found that overexpressing Sec3 or Rab11-GFP, but not Sec5-GFP (Kakihara et al., 2008), in a single cell clone within a mosaic border cell cluster significantly promoted the likelihood of such a cell being positioned at the leading position (supplementary material Fig. S2), suggesting that this cell utilizes its increased recycling and tethering to amplify and relocalize active RTKs faster and more efficiently than other wild-type neighbor cells. The difference in promoting effect from Sec3 and Sec5 is interesting, suggesting that when overexpressed the Sec3 subunit is more able to enhance the overall exocyst function than Sec5. This is consistent with a Sec3 study in budding yeast, which shows that as a unique subunit of exocyst Sec3 serves as a spatial landmark on the bud tip to recruit a subcomplex (comprising seven subunits) of exocyst containing all subunits but Sec3 (Boyd et al., 2004; Finger et al., 1998). Only when the subcomplex along with the associated vesicle arrives at the bud tip, can Sec3 be joined with it to form a fully functional tethering complex.

The next question is how the polarized distribution of recycling and tethering activity is initiated in border cells. We demonstrated that this was likely to be induced by the guidance receptors in response to the external gradient of guidance cues, as removing guidance signaling by DN-PVR and DN-EGFR expression abolished Rab11 and Sec5 polarized distribution, and DN-PVR expression alone markedly reduced the polarization. These data suggested the presence of a positive feedback loop of active RTKs-endocytic recycling-active RTKs in border cells, as Rab11 and exocyst components (Sec3 and Sec15) were shown to be conversely required for polarized pTyr or active RTK localization at the front (this study; Assaker et al., 2010; Janssens et al., 2010). Interestingly, we find PVR signals downstream through Rac and then polymerized actin to promote recycling endosome levels, providing mechanistic details to this feedback loop (Fig. 7). Interestingly, it was recently shown that Rab11 interacts with Rac and actin cytoskeleton regulator moesin during border cell migration (Ramel et al., 2013). Furthermore, we found that strong Rab11 stainings were proximal to or partially overlapping with strong F-actin staining in the leading edge of wild-type border cells and around the ectopic F-actin regions in the λ-PVR, RacV12 or tsr-RNAi expressing follicle cells and border cells. F-actin appears to be the direct cause rather than the effect of recycling endosome accumulation, because manipulating its levels by Lat-A or tsr RNAi leads to either up- or downregulation of the levels of recycling endosome. However, we cannot rule out the possibility that Rac can somehow act on recycling endosome-associated regulators directly (independently of F-actin) to affect their function. It was previously shown that actin polymerization is required for recycling of cargo back to plasma membrane, possibly through F-actin serving as a track for the movement of vesicles (Grant and Donaldson, 2009; Radhakrishna and Donaldson, 1997; Weigert et al., 2004). However, how F-actin induces recycling endosome formation and organization is not clear and remains to be elucidated.

It was previously proposed that recycling of active RTKs needs to be directional (toward the front) to achieve polarized RTK activity (Janssens et al., 2010; Jékely et al., 2005; Montell et al., 2012). If active RTKs in the leading edge are endocytosed and then recycled to new regions in the membrane, RTK activity would be delocalized. What causes the recycling to be directed toward the front membrane is not clear. Our proposed feedback loop via F-actin suggests that the active PVR (RTK) in the leading edge could locally induce higher levels of recycling endosome through Rac and enhanced actin polymerization (by Rac). As a result, the directional recycling could be achieved with the localized actin filaments serving both as a recycling endosome inducing agent and as tracks for movement of vesicles (carrying active RTKs) toward the front membrane, which prevents the active RTKs from being recycled to elsewhere and becoming delocalized. Indeed, inhibiting actin polymerization in the border cells by Lat-A treatment abolished both the polarized F-actin and the elevated Rab11 stainings proximal to F-actin, which are normally present in the leading edge of the wild-type cluster.

Lastly, our work also provides some insight into the kinds of cargo that are recycled during border cell migration. We show that E-cadherin is a specific cargo. E-cadherin is the major adhesion molecule required for border cell migration (Niewiadomska et al., 1999), whereas integrin plays only a minor role and is not required in border cells (Devenport and Brown, 2004; Dinkins et al., 2008). Our finding suggests that cycles of endocytosis and recycling of E-cadherin could promote the dynamic assembly and disassembly of E-cadherin-mediated adhesion on the substrate (nurse cell E-cadherin), similar to how the turnover of integrin at the focal adhesion is regulated by endocytic recycling in mammalian cells (Caswell and Norman, 2008; Caswell et al., 2007; Ulrich and Heisenberg, 2009). Interestingly, we often observed that elevated intracellular E-cad stainings tend to be localized below the cell membrane that juxtaposes nurse cell membrane (supplementary material Fig. S7A,C), suggesting that E-cadherin is normally delivered to or recycled back to this membrane region by Rab11 and exocyst during adhesion and migration. Another important candidate cargo to be determined is PVR. However, we have been unable to detect significant colocalization between Rab11 with PVR or active PVR with the previously reported PVR or pPVR antibody (Janssens et al., 2010; Jékely et al., 2005). Therefore, the definitive role of PVR or active PVR as a cargo for recycling still awaits further determination.

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, except for the following: sec3^GT/TM6B (Chen et al., 2005), Rab11^j2D1, Rab5^k08232, sec3^PBac/TM6B (Drosophila Genetic Resource Center), UAS-sec5-GFP (Hayashi’s lab), sec3 RNAi (Vienna Drosophila RNAi Center), UAS-DNPVR/Cyo (Montell’s lab), UAS-DNPVR, UAS-DNEGFR/Cyo, UAS-λPVR/TM3 (Rørth’s lab). To generate UAS-sec3 transgenic line, a full-length cDNA of the sec3 gene was subcloned into a modified pUAST-attB vector and was injected into embryos according to standard procedures. Mutant FRT clones were induced using hs-FLP. Flies were heat shocked for 1 hour per day at 37°C for 3 days before eclosion and 1 day after eclosion, then dissected 2-3 days after the last heat shock. To perform flip-out experiments, AyGAL4 UAS-GFP (or UAS-sec3, or UAS-rab11-GFP, or UAS-sec5-GFP) was crossed to hs-Flp. Newly eclosed flies were heat shocked at 37°C for 4.5 minutes, and dissected after 2-3 days. Only mosaic border cell clusters with one-cell flip-out clone were used for analysis.

Ovary dissection was carried out in PBS and then fixed in devitellinizing buffer (7% formaldehyde) and heptane (Sigma) mixture (1:6) for 10 minutes. After washes in PBS, ovaries were incubated in blocking solution (PBT, 10% goat serum) for 30 minutes and then stained overnight at 4°C. Note that we and some labs used the formaldehyde/heptanes method for fixing egg chambers, whereas other labs used formaldehyde only method, which may result in variability in staining patterns. But, we have used both methods for E-cadherin staining and found both produced similar staining patterns (data not shown). Primary antibodies and their concentrations were as follows: mouse anti-phospho-Tyr (1:200, 4G10, Millipore), mouse anti-Rab11 (1:200, BD Transduction), rat anti-E-Cad (DCAD2, 1:50, DSHB), rabbit anti-Baz (1:400, gift from A. Wodarz), mouse anti-Dlg1 (1:100, DSHB), mouse anti-Sec5 (1:50, gift from T. Schwarz), Rabbit anti-Rab5 (1:200, Abcam) and rat anti-PVR (1:200, gift from P. Rørth). After washes in PBT, ovaries were incubated with secondary antibodies (Jackson ImmunoResearch) for 2 hours at room temperature. F-actin was labeled by Rhodamine phalloidin (1:100, Sigma). Confocal images were obtained using a Leica TCS SP5 II (with HyD detector) or an Olympus FV1000 confocal microscope.

The quantification methods are similar to that described previously (Zhang et al., 2011). For measurement of the front/back ratios, an area around the leading edge of the cluster, but excluding polar cells (labeled by lack of GFP from slbo-Gal4), was chosen as the front region, and an area including the lagging end and excluding polar cells was chosen as the back region. Fluorescence intensity (FI) and area were measured in ImageJ software (NIH) for each region; the front/back ratios were calculated as [front FI/front area] divided by [back FI/back area]. For measurement of Rab11’s levels in the slbo-Gal4:UAS-GFP marked border cells and follicle cells, nurse cells or follicle cells (not marked by GFP) adjacent to the GFP-marked border cells or follicles respectively are used for normalization of fluorescence, because their constant and uniform levels of Rab11 staining are not affected by slbo-Gal4:UAS-transgene. FI and area were measured in ImageJ for each area as mentioned above and were used for the following calculations: normalized Rab11 levels in border cells=[border cells FI/border cells area] divided by [nurse cells FI/nurse cells area]; normalized Rab11 levels in Gal4 expressing follicle cells=[Gal4 FCs FI/Gal4 FCs area] divided by [adjacent WT FCs FI/adjacent WT FCs area].

Ovaries were dissected and cultured in Schneider’s medium cocktail as described previously for use of live imaging (Chu et al., 2012; Prasad and Montell, 2007). After dissection, egg chambers were incubated in Latrunculin A (2 μM in Schneider’s medium cocktail, Invitrogen) for 1 hour before fixation, and were then stained as described above.

Egg chambers were incubated with the anti-DE-Cadherin (DCAD2 diluted at 1:10) antibody in the Schneider’s medium cocktail for 30 minutes at 4°C. Following three washes in Schneider’s medium at 4°C, egg chambers were fixed and then stained as described above, the procedure is similar to that previously described (Langevin et al., 2005).

We thank Bloomington Drosophila Stock Center, Drosophila Genetic Resource Center, Vienna Drosophila RNAi Center, Pernile Rørth, Denise Montell and Shigeo Hayashi for fly stocks; and Thomas Schwarz, Andreas Wodarz and Pernile Rørth for antibody reagents.