To address the role of phosphatases in border cell migration, we carried out a small-scale genetic screen to inhibit selected serine-threonine phosphatases that are expressed during oogenesis using RNAi as well as a protein inhibitor that targets Pp1 catalytic subunits (Table 1; Miskei et al., 2011; Bennett et al., 2003). We drove expression of RNAi and the inhibitor using c306-GAL4, an early anterior follicle cell driver expressed at high levels in border cells and polar cells (Figure 1—figure supplement 1A). Inhibition of Pp4-19C (one RNAi line) and Pp1c, through overexpression of Nuclear inhibitor of Protein phosphatase 1 (NiPp1), significantly disrupted border cell migration (Table 1). NiPp1 is an endogenous protein that when overexpressed, effectively and specifically blocks Pp1 catalytic subunit activity in vivo (Bennett et al., 2003; Winkler et al., 2015; Parker et al., 2002; Van Eynde et al., 1995). Pp1 and associated complexes are important phosphatase regulators of many cellular processes. Moreover, females expressing NiPp1 driven by c306-GAL4 did not produce adult progeny when crossed to wild-type males, consistent with infertility and suggesting a role for Pp1 in normal oogenesis (Figure 1—figure supplement 2A). Here we focused on further elucidating the function of Pp1 in border cells.

We used two GAL4 drivers to assess phenotypes, c306-GAL4 to determine early broad function of Pp1 in border cells and polar cells and slbo-GAL4 for later more restricted function in just border cells (Figure 1—figure supplement 1). Expression of NiPp1 strongly disrupted both the ability of border cells to organize into a cohesive cluster and to migrate successfully (Figure 1G–J). Unlike control border cells, most NiPp1-expressing border cells failed to reach the oocyte by stage 10 (98%; Figure 1I). Importantly, NiPp1-expressing border cells were no longer found in one cohesive cluster. Instead, individual cells and smaller groups split off from the main cluster (Figure 1H). Whereas control border cells migrated as a single cohesive unit (‘one part’), NiPp1-expressing border cells split into two to three (50%), or more (40%), parts (Figure 1H,J). Migration and cluster cohesion defects were observed when NiPp1 was expressed early in both border cells and the central polar cells (c306-GAL4; Figure 1I,J; Figure 1—figure supplement 2B) or later in just border cells (slbo-GAL4; Figure 1—figure supplement 2C–G). Polar cells, through JAK/STAT signaling, recruit border cells to form a migratory cluster, and anchor border cells to the cluster (Cai et al., 2014; Ghiglione et al., 2002; Silver and Montell, 2001). Therefore, we tested the function of Pp1 in polar cells. We observed no defects in cohesion or migration when NiPp1 was expressed only in polar cells (upd-GAL4; Figure 1—figure supplement 2C,H–K). Fragmentation of clusters, however, was stronger when NiPp1 was driven by c306-GAL4 rather than slbo-GAL4 (compare Figure 1J to Figure 1—figure supplement 2G), possibly due to earlier and higher expression of c306-GAL4 (Figure 1—figure supplement 1; Silver and Montell, 2001). Although polar cells are normally located at the center of the border cell cluster and maintain overall cluster organization (Cai et al., 2014; Niewiadomska et al., 1999), individual NiPp1-expressing border cells could completely separate from polar cells as well as the other border cells (Figure 1—figure supplement 2L–N). Finally, NiPp1 border cells appeared rounder than normal, indicating that individual cell shapes were altered (see below). Together, these results demonstrate that NiPp1 expression in border cells, but not polar cells alone, disrupts collective migration, cluster organization and adhesion.

Because very few border cells reached the oocyte, we investigated whether NiPp1-expressing border cells were correctly specified and functional. We first examined the expression of the transcription factor Slbo, the fly C/EBP homolog, which is required for border cell specification in response to JAK/STAT signaling (Silver and Montell, 2001; Montell et al., 1992). NiPp1-expressing border cells generally expressed Slbo, similarly to control cells (Figure 1—figure supplement 3A–B’; 30/33 border cells expressed Slbo, n = 6 egg chambers). Proper specification through JAK/STAT signaling restricts the number of follicle cells that become migrating border cells (Silver and Montell, 2001; Starz-Gaiano et al., 2008). When NiPp1 expression was driven by c306-GAL4, the total number of cells in the cluster (border cells and polar cells) was slightly increased to a mean of seven NiPp1 cells compared to six control cells per cluster (Figure 1—figure supplement 3C; n = 27 egg chambers for each genotype). This modest increase in cells per cluster is far fewer than what is observed upon ectopic activation of JAK/STAT (Silver and Montell, 2001; Starz-Gaiano et al., 2008), suggesting that NiPp1 does not greatly impact the specification or recruitment of border cells. Thus, NiPp1 prevents properly specified border cells from staying together and completing migration.

To determine where and when NiPp1-expressing border cells stopped migrating and dissociated from the cluster, we examined border cell clusters using live time-lapse imaging (Prasad and Montell, 2007; Dai and Montell, 2016). Both control and NiPp1 border cells delaminated from the surrounding epithelium and began their migration as a group (Figure 1K–L’’; Videos 1, 2, 3, 4). NiPp1 border cells separated into multiple sub-clusters or single cells at various points during migration, particularly after moving between the nurse cells (Videos 2, 3, 4). NiPp1 border cells typically migrated as small groups but also could arrange themselves into co-linear chains (Video 3). A few NiPp1 border cells reached the oocyte, although considerably later than control border cells. Indeed, NiPp1-expressing border cells migrated more slowly overall compared to control border cell clusters (~0.35 µm/min NiPp1 versus ~0.65 µm/min control; Figure 1M). Individual NiPp1 border cells also moved at variable speeds, with lagging border cells sometimes pushing ahead of the nominal leading cell (Video 2). Labeling with a cortical cell membrane marker, PLCδ-PH-EGFP (slbo-GAL4 >UAS-PLCδ-PH-EGFP), allowed us to determine that some NiPp1 border cells completely disrupted their cell-cell contacts, whereas other border cells remained in contact (Video 5). Finally, single border cells that broke off from the cluster were frequently left behind and stopped moving forward, appearing to get ‘stuck’ between nurse cells (Videos 2, 3, 4). Taken together, these data show that NiPp1 disrupts the ability of border cells to maintain a collective mode of migration, and leads to border cells now moving as single cells or small groups with slower speed that typically fail to reach the oocyte.

NiPp1 is a specific inhibitor of Pp1c activity in vitro as well as in vivo (Winkler et al., 2015; Parker et al., 2002; Van Eynde et al., 1995). Drosophila has four Pp1c subunit genes (Dombrádi et al., 1993; Dombrádi et al., 1990), whereas humans have three genes (Lin et al., 1999). Pp1α−96A, Flapwing (Flw), and Pp1-87B transcripts are each expressed at moderate-to-high levels in the adult ovary, whereas Pp1-13C RNA is mainly detected in adult males (http://flybase.org/) (Graveley et al., 2011). We examined the localization of Pp1α−96A using a genomic fosmid transgene in which the open reading frame of Pp1α−96A is driven by its endogenous genomic regulatory regions and C-terminally tagged with GFP (‘Pp1α−96A-GFP’) (Sarov et al., 2016). Pp1α−96A-GFP was detected in the cytoplasm, with higher levels at the cortical membranes of border cells, follicle cells, the oocyte, and nurse cells (Figure 2A–C). Endogenous Flw, as visualized using a functional in-frame YFP protein trap (Yamamoto et al., 2013) (‘Flw-YFP’), was also expressed ubiquitously during the stages in which border cells migrate (Figure 2D–F). Specifically, Flw-YFP was enriched at the cell cortex and cytoplasm of all cells, including border cells. Due to a lack of specific reagents, we were unable to determine whether Pp1-87B or Pp1-13C proteins are present in border cells. Therefore, at least two Pp1c subunit proteins are expressed in border cells throughout their migration.

We next determined whether NiPp1 specifically inhibited Pp1c activity in border cells. Overexpression of each of the four Drosophila Pp1c subunits individually did not impair border cell migration (Figure 2—figure supplement 1A–D). When co-expressed with NiPp1, two of the catalytic subunits, Pp1α−96A and Pp1-87B, strongly suppressed the migration defects caused by NiPp1, with 90% (NiPp1 + Pp1α−96A) and 75% (NiPp1 + Pp1-87B) of border cells now reaching the oocyte compared to 40% with NiPp1 alone (NiPp1 + RFP; Figure 2G; Figure 2—figure supplement 1F–H). Co-expression of Pp1α−96A and Pp1-87B partially suppressed the NiPp1-induced cluster fragmentation, leading to 55% (NiPp1 + Pp1α−96A) and 65% (NiPp1 + Pp1-87B) of border cell clusters now found intact compared to ~10% with NiPP1 alone (NiPp1 + RFP; Figure 2H; Figure 2—figure supplement 1F–H). Flw and Pp1-13C only mildly suppressed the NiPp1-induced cluster splitting and migration defects (Figure 2G,H; Figure 2—figure supplement 1I,J). The observed phenotypic suppressions were likely due to titration of NiPp1 inhibitory activity by excess Pp1c protein, in agreement with previous studies in Drosophila (Bennett et al., 2003; Parker et al., 2002). Partial suppression could be due to levels of overexpressed Pp1c or effectiveness of the respective Pp1c to titrate NiPp1 in border cells. Co-expression of a human Pp1c homolog (‘hPPP1CC’) fully suppressed the NiPp1-induced phenotypes and did not disrupt migration when expressed on its own (Figure 2G,H; Figure 2—figure supplement 1E,K). hPPP1CC has high homology to Pp1-87B (93% identical, 96% similar), Pp1α−96A (89% identical, 94% similar), and Pp1-13C (91% identical, 95% similar), with lower homology to Flw (84% identical, 91% similar), although further analysis through the DIOPT DrosophilaRNAi Screening Center Integrative Ortholog Prediction Tool) database suggests higher homology to Pp1-87B and Pp1α−96A (http://flybase.org/) (Hu et al., 2011). The suppression by multiple Pp1 proteins and full suppression by hPPP1CC suggests that Pp1 catalytic subunit genes have overlapping functions in border cells.

To better understand how NiPp1 inhibits Pp1 activity in border cells, we next analyzed the subcellular localization of Flw-YFP and Pp1α−96A-GFP when NiPp1 was co-expressed. Expression of HA-tagged NiPp1 alone was itself predominantly nuclear, with low expression in the cytoplasm (Figure 2—figure supplement 2A–A’’). Pp1α −96A-GFP and Flw-YFP normally localize to the cortical membrane and cytoplasm of border cells (Figure 2A–F). Upon co-expression with NiPp1, however, Flw-YFP and Pp1α −96A-GFP were now primarily localized to border cell nuclei along with NiPp1 (HA-tagged NiPp1; Figure 2—figure supplement 2B–C”). These results suggest that ectopic NiPp1, in addition to directly inhibiting Pp1c activity also sequesters PP1 catalytic subunits in the nucleus (Winkler et al., 2015; Parker et al., 2002; Trinkle-Mulcahy et al., 2001).

To determine whether Pp1 catalytic activity itself is required for border cell migration, we next downregulated the Pp1c genes by driving the respective UAS-RNAi lines with c306-GAL4 (Figure 3A–D). RNAi lines that target 3 of the four catalytic subunits (Pp1α−96A, Pp1-87B, and Pp1-13C) strongly disrupted border cell migration (Figure 3B–E). The majority of Pp1c RNAi border cells either did not migrate (‘no migration’) or stopped along the migration pathway (‘incomplete migration’; Figure 3E). Pp1α−96A-RNAi in particular, caused a significant fraction of border cells to fail to migrate at all, likely due to a failure to delaminate from the epithelium (~15%; Figure 3E). Knockdown of Pp1c genes also caused ≥50% of border cell clusters to dissociate into multiple sub-clusters and single cells (Figure 3B–D,F). Using live time-lapse imaging, we confirmed that decreased levels of Pp1α−96A, Pp1-87B, and Pp1-13C by RNAi altered border cell migration and caused cells to split from the main cluster (Figure 3G; Videos 6, 7, 8, 9 and 10). Some Pp1α−96A-RNAi border cells did not delaminate from the epithelium during the course of imaging (Figure 3—figure supplement 1A; Video 8). Multiple flw RNAi lines (see Materials and methods) did not impair migration or cluster cohesion when expressed in border cell clusters. However, RNAi does not always fully knock down gene function in cells (Mohr et al., 2014). As complete loss of flw is homozygous lethal, we generated border cells that were mosaic mutant for the strong loss of function allele flw^FP41 (Sun et al., 2011). Mosaic flw^FP41 border cell clusters were typically composed of a mixture of wild-type and mutant cells that frequently fell apart, with ~90% splitting into two or more parts (Figure 3H–I; Figure 3—figure supplement 1B–B”). In egg chambers with flw mutant border cells, 40% of border cell sub-clusters did not delaminate or migrate at all (‘no migration’) whereas 20% partially migrated but did not reach the oocyte (Figure 3H–H”,J; Figure 3—figure supplement 1B–B”). NiPp1 expression results in more severe phenotypes than RNAi knockdown, or loss, of individual Pp1c genes, at least with respect to migration and cluster cohesion, suggesting that Pp1c subunits have both distinct and overlapping functions. In particular, Pp1α−96A and Flw appear to function in border cell delamination, whereas all four subunits likely promote migration and cluster cohesion.

One of the strongest effects of decreased Pp1c activity was the dissociation of border cells from the cluster. In many cell collectives, cadherins critically mediate the attachment of individual cells to each other during migration, although other cell-cell adhesion proteins can also contribute (Friedl and Mayor, 2017; Collins and Nelson, 2015). The cadherin-catenin complex members E-Cadherin (Drosophila Shotgun; Shg), β-Catenin (Drosophila Armadillo; Arm) and α-Catenin are all required for border cell migration (Cai et al., 2014; Niewiadomska et al., 1999; Sarpal et al., 2012; Desai et al., 2013). E-Cadherin, in particular, is required for traction of border cells upon the nurse cell substrate, for producing overall front-rear polarity within the cluster, and for attachment of border cells to the central polar cells (Cai et al., 2014; Niewiadomska et al., 1999). Complete loss of cadherin-catenin complex members in border cells prevents any movement between nurse cells (Niewiadomska et al., 1999; Sarpal et al., 2012; Desai et al., 2013). This has precluded a definitive analysis of whether all, or some, complex members promote adherence of border cells to the polar cells and/or to other border cells.

To determine whether adhesion of border cells to the cluster requires a functional cadherin-catenin complex, we used c306-GAL4 to drive RNAi for each gene in all cells of the cluster (Figure 1—figure supplement 2B). Multiple non-overlapping RNAi lines for E-Cadherin, β-Catenin, and α-Catenin each reduced the respective endogenous protein levels and disrupted border cell migration, in agreement with previous results that used mutant alleles (Figure 4A–E,G,I; Figure 4—figure supplement 1A–H’; Video 11; Niewiadomska et al., 1999; Sarpal et al., 2012; Pacquelet and Rørth, 2005). Importantly, RNAi knockdown for each of the cadherin-catenin complex genes, driven by c306-GAL4, resulted in significant fragmentation of the border cell cluster compared to controls. E-Cadherin (40–50%) and β-Catenin (55–80%) RNAi lines exhibiting stronger, while α-Catenin RNAi lines exhibited milder (~20–30%), cluster fragmentation (Figure 4A–D,F,H,J; Video 11). Dissociated RNAi border cells could localize to the side of the egg chamber (Figure 4B,D), although others remained on the normal central migration pathway (Figure 4C,D). While α-Catenin RNAi knockdown in polar cells alone (upd-GAL4) caused border cell cluster splitting and migration defects, this effect was significantly milder than the effects of α-Catenin knockdown in both polar cells and border cells using c306-GAL4 (compare Figure 4I,J to Figure 4—figure supplement 1I,J). These results indicate that the cadherin-catenin complex keeps border cells attached to each other and to the polar cells, which in turn maintains a cohesive cluster.

We next wanted to determine whether Pp1 regulated these adhesion proteins in border cells. We analyzed the levels and localization of E-Cadherin and β-Catenin at cell-cell contacts in NiPp1-expressing border cell clusters that were still intact or loosely connected (Figure 4K–P). In wild-type clusters, E-Cadherin and β-Catenin are highly enriched at cell contacts between border cells (BC-BC) and between border cells and polar cells (BC-PC; Figure 4K–K”,M–M”). NiPp1-expressing border cell clusters exhibited reduced levels of E-Cadherin and β-Catenin at most BC-BC contacts (Figure 4L–L”,N–N”). Pp1-inhibited polar cells generally retained E-Cadherin and β-Catenin, which was higher compared to border cells (Figure 4L–L”,N–N”). We quantified the relative levels of E-Cadherin (Figure 4O) and β-Catenin (Figure 4P) at BC-BC contacts in control versus NiPp1 clusters, normalized to the levels of those proteins at nurse cell-nurse cell junctions. Both E-Cadherin and β-Catenin were reduced by almost half compared to matched controls. These data together suggest that Pp1 activity regulates cadherin-catenin proteins at cell-cell contacts, which contributes to adhesion of border cells within the cluster.

Border cells with impaired Pp1 activity migrated significantly slower than control clusters (Figures 1M and 3G), suggesting that border cell motility was altered. Migrating cells form actin-rich protrusions at the front, or leading edge, which help anchor cells to the migratory substrate and provide traction for forward movement (Ridley, 2011; Caswell and Zech, 2018). In collectives, protrusive leader cells also help sense the environment to facilitate directional migration (Mayor and Etienne-Manneville, 2016). Border cells typically form one or two major protrusions at the cluster front (Prasad and Montell, 2007; Poukkula et al., 2011; Wang et al., 2010; Figure 5A–A””,C; Figure 5—figure supplement 1A; Video 6). Pp1-inhibited border cells (Pp1c RNAi) still extended forward-directed protrusions (Figure 5A–C; Videos 7, 8, 9 and 10). Additionally, the numbers, lifetimes, lengths and areas of side- and back-directed protrusions were not generally increased in Pp1-inhibited border cell clusters compared to control (Figure 5C–F; Figure 5—figure supplement 1B,C). However, the number of protrusions produced at the front of the cluster was reduced in Pp1 RNAi border cells (range of 0.5–0.85 mean protrusions per frame, all genotypes) compared to control (1.0 mean protrusions per frame; Figure 5C). Additionally, the lifetimes of Pp1 RNAi forward-directed protrusions were reduced (Figure 5D). Control protrusions at the cluster front had a lifetime of ~18 min, whereas Pp1-inhibited front protrusions persisted for 5–10 min. These short-lived Pp1 RNAi protrusions were also reduced in length, from a third to half the size of control front-directed protrusions (Figure 5E; Figure 5—figure supplement 1B). Further, Pp1-inhibited front protrusions were smaller, with a mean area of ~10–20 µm² compared to the control mean of ~40 µm² (Figure 5F; Figure 5—figure supplement 1C). Thus, Pp1 activity promotes normal protrusion dynamics, including the number, lifetime and size of front-directed protrusions.

The majority of NiPp1 and Pp1c RNAi border cells followed the normal migratory pathway down the center of the egg chamber between nurse cells, even when cells broke off from the main cluster (Figure 1H,L–L” and 3B–D; Videos 2, 3, 4, 5 and 7, 8, 9, 10). Moreover, in Pp1 RNAi border cells, front-directed protrusions still formed though with altered dynamics. These observations together suggest that Pp1 activity is not required for directional chemotactic migration. To further test this idea, we made use of a Förster Resonance Energy Transfer (FRET) activity reporter for the small GTPase Rac. Normally, high Rac-FRET activity occurs at the cluster front during early migration in response to guidance signals from the oocyte, and correlates with protrusion extension (Figure 5—figure supplement 1D; Wang et al., 2010). Under conditions of PP1-inhibition, the most severely affected clusters fall apart, sometimes on different focal planes, making it difficult to interpret Rac-FRET signal. We therefore measured global Rac-FRET only in those NiPp1-expressing border cell clusters that remained intact. We detected elevated Rac-FRET activity in NiPp1 border cells similar to control, indicating that Rac activity was largely preserved although with slightly elevated levels (Figure 5—figure supplement 1D,E). In sum, these data indicate that Pp1 activity influences protrusion dynamics and cell motility, but does not appear to be critical for directional orientation of the cluster to the oocyte.

Migrating cells, including cell collectives, change shape to facilitate their movement through complex tissue environments (Te Boekhorst et al., 2016). Some cells maintain a single morphology, such as an elongated mesenchymal or rounded amoeboid shape, throughout migration, whereas other cells interconvert from one shape to another as they migrate. The border cell cluster overall is rounded, although individual border cells within the group appear slightly elongated (Figure 6A,A’; Videos 1 and 6; Aranjuez et al., 2016). However, NiPp1 border cells, whether present in small groups or as single cells, were visibly rounder than control border cells (Figure 1H,L–L”; Videos 1–4). We observed similar cell rounding when the Pp1c genes were knocked down by RNAi, although some border cells appeared more noticeably round than others (Figures 3B–D and 5B–B””; Videos 7–10). To quantify these altered cell shapes, we expressed the membrane marker PLCδ-PH-EGFP to visualize individual cells within the cluster and measured ‘circularity’, which indicates how well a shape approaches that of a perfect circle (1.0; Figure 6A–C). Control border cells overall were slightly elongated with a mean of ~0.7, although the circularity of individual cells varied substantially (range of ~0.4 to 0.95), suggesting that border cells undergo dynamic shape changes during migration (Figure 6C). In contrast, NiPp1 border cells were rounder, with a mean of ~0.9, and exhibited less variation than control (range of ~0.7 to 1.0; Figure 6C).

The rounder cell shapes suggested that Pp1 inhibition alters the cortical cytoskeleton of the border cells. Wild-type border cells exhibit a marked enrichment of F-actin at the cluster periphery, whereas lower levels are detected inside the cluster at contacts between border cells (Figure 6D,D’,F; Video 12; Lucas et al., 2013; Wang et al., 2018). Upon Pp1 inhibition, F-actin now accumulated around each individual border cell, especially at BC-BC membrane contacts, rather than just being enriched at outer cluster surfaces (Figure 6E,E’,G; Video 13). Similarly, Myo-II as visualized by GFP-tagged Spaghetti Squash (Sqh-GFP), the Drosophila homolog of the myosin regulatory light chain (MRLC), is highly dynamic and normally concentrates in enriched foci at the outer periphery of live border cell clusters both during early (Figure 6—figure supplement 1A–A””’; Video 14) and later stages of migration (Figure 6H–H””’; Video 16; Aranjuez et al., 2016; Combedazou et al., 2017; Majumder et al., 2012). In NiPp1 border cells, however, Sqh-GFP was now present at cortical cell membranes in dynamic foci surrounding each border cell (or sub-cluster) rather than at the entire cluster periphery, both during early migration (Figure 6—figure supplement 1B–B””’; Video 15) and at mid-migration stages (Figure 6I–I””’; Video 17). Thus, inhibition of Pp1 converts collectively polarized F-actin and Myo-II to that characteristic of single migrating cells. As a result, individual border cells now have enriched and dynamic actomyosin localization consistent with elevated cortical contractility in single cells rather than at the collective level.

Rok and other kinases phosphorylate the Myo-II regulatory light chain Sqh (Vicente-Manzanares et al., 2009). This leads to fully activated Myo-II, which then forms bipolar filaments, binds to F-actin, and promotes cell contractility. Given the altered distribution of Sqh-GFP when Pp1 was inhibited, we next analyzed the levels and distribution of active Myo-II. We used an antibody that recognizes phosphorylated Sqh at the conserved Ser-21 (mammalian MRLC Ser-19; ‘p-Sqh’) (Majumder et al., 2012). Control border cells exhibited p-Sqh signal primarily at the cluster periphery (‘BC-NC’ contacts; Figure 7A–A’’). This pattern of p-Sqh closely resembles the pattern of Sqh-GFP in live wild-type border cells (Figure 6H–H””’; Aranjuez et al., 2016; Majumder et al., 2012; Zeledon et al., 2019; Plutoni et al., 2019). NiPp1 border cells, however, had high levels of p-Sqh distributed throughout the cluster including at internal BC-BC contacts (Figure 7B–B’’), similar to Sqh-GFP in live NiPp1 border cells (Figure 6I–I””’). We measured the relative ratio of p-Sqh fluorescence intensity at BC-NC contacts versus BC-BC contacts in control and NiPp1 border cell clusters (Figure 7C). Control border cells had a higher p-Sqh ratio than NiPp1, indicating less p-Sqh signal at BC-BC contacts. These data support the idea that Pp1 inhibition elevates Myo-II activation within single border cells and at BC-BC contacts.

Myo-II undergoes cycles of activation and inactivation via phosphorylation and dephosphorylation, respectively, to generate dynamic cellular contraction in vivo (Vicente-Manzanares et al., 2009). We previously showed that waves of dynamic Myo-II maintain the collective morphology of border cells to facilitate movement through the egg chamber (Aranjuez et al., 2016). The myosin phosphatase complex consists of a Pp1c subunit and a specific regulatory subunit, the myosin binding subunit (Mbs; also called myosin phosphatase-targeting subunit [MYPT]), which together dephosphorylate Sqh and inactivate Myo-II (Grassie et al., 2011). Previously, we found that Mbs was required for border cell cluster delamination from the epithelium and cell shape (Aranjuez et al., 2016; Majumder et al., 2012), although cluster cohesion had not been explicitly assessed. We therefore wanted to determine whether myosin phosphatase contributed to the above-described Pp1 functions in cell shape, cluster cohesion and migration. First, we confirmed that Mbs transcript and protein were expressed in border cells throughout migration (Figure 7—figure supplement 1A–F). Mbs protein colocalized with Pp1c subunits near border cell membranes and in the cytoplasm (Figure 7—figure supplement 1G–J). In general, Mbs colocalized more extensively with Flw-YFP than with Pp1α−96A-GFP (Figure 7—figure supplement 1G–J).

Next, we analyzed the functions of Mbs in border cells. Border cells deficient for Mbs (Mbs-RNAi) were rounder than control border cells, exhibited incomplete migration (~30%), and dissociated from the cluster (60%) along the migration pathway (Figure 7D–H). The phenotypes observed with Mbs-RNAi were generally milder than those observed with Pp1-inhibition (either NiPp1 or Pp1c-RNAi; compare to Figures 1I,J and 3E,F). This could be due to incomplete knockdown by Mbs-RNAi, although we observed significant decreases in the levels of endogenous Mbs (Figure 7—figure supplement 1K–L”). Alternatively, myosin phosphatase, through a complex of Mbs/Pp1c, could be one of multiple Pp1 complexes required for border cell cluster migration and cohesion (see Discussion). Nonetheless, these findings indicate that myosin phosphatase, a specific Pp1 complex, helps promote the normal cell morphology and collective cohesion of border cells, in addition to facilitating the successful migration of the border cells.

RhoA activates Rho-associated kinase (Rok), thus leading to activation of Myo-II (Vicente-Manzanares et al., 2009). We and others previously found that expression of constitutively-activated RhoA (Drosophila Rho1) causes markedly rounder border cells and alters the distribution of F-actin and Myo-II at cell-cell contacts between border cells (Aranjuez et al., 2016; Combedazou et al., 2017). We therefore investigated whether Pp1 regulated RhoA activity in migrating border cells. We used a FRET construct that was recently shown to specifically report RhoA activity in ovarian follicle cells (Qin et al., 2017). Inhibition of Pp1 by NiPp1 moderately increased the overall levels of Rho-FRET in intact border cell clusters compared to control border cells (Figure 7—figure supplement 2A–C). These data suggest a general upregulation of the RhoA pathway upon Pp1 inhibition.

During cellular morphogenesis, the cadherin-catenin complex and actomyosin contractility can interact to influence cell-cell junction stability (Mège and Ishiyama, 2017; Yap et al., 2018; Priya et al., 2015; Ratheesh et al., 2012; le Duc et al., 2010). Given the effects of Pp1 inhibition on the cadherin-catenin complex, F-actin, and Myo-II, we asked whether the observed Pp1-dependent phenotypes were secondarily due to decreased adhesion and/or altered actomyosin contractility. Knockdown of E-cadherin or β-catenin by RNAi decreased the enrichment of F-actin and p-Sqh at the cluster periphery compared to controls (Figure 6—figure supplement 2A–C’, E–G,I–K). This is in agreement with a recent study that observed decreased cortical Myo-II in live E-cadherin-RNAi border cells (Mishra et al., 2019). Despite this decrease in F-actin, migrating live α-Catenin RNAi border cells, while slower than control, extended protrusions with normal dynamics (Figure 5—figure supplement 1F–J). Interestingly, F-actin was also no longer enriched at the cluster periphery of Sqh-RNAi border cells (Figure 6—figure supplement 2D,D’,H). Thus, F-actin enrichment at the cluster periphery requires both cadherin-catenin and Myo-II. Moreover, the cadherin-catenin complex promotes enriched activated Myo-II at the outer cluster.

Next, we asked if Myo-II was required for cadherin-catenin enrichment at border cell-border cell junctions. Sqh-RNAi border cells had normal levels of E-cadherin (Figure 6—figure supplement 3A–B’’, E) and normal to slightly higher levels of β-catenin (Figure 6—figure supplement 3C–D’’, F). Knockdown of Sqh did not disrupt distribution of E-cadherin or β-catenin at border cell-border cell contacts (Figure 6—figure supplement 3A–D’’). These data suggest that Myo-II is not a major regulator of the cadherin-catenin complex in border cells. The phenotypes observed with RNAi-mediated knockdown of the cadherin-catenin complex and Sqh are in contrast to those observed with Pp1 inhibition (e.g. Figures 5–7). These results are consistent with a more direct role for Pp1 activity in controlling collective versus single cell dynamics of actomyosin and cadherin-catenin in border cells.

Crosses were generally set up at 25°C unless otherwise indicated. The tub-GAL80^ts (‘tsGAL80’) transgene (McGuire et al., 2004) was included in many crosses to suppress GAL4-UAS expression during earlier stages of development; these crosses were set up at 18°−22 °C to turn on tsGAL80. For c306-GAL4, c306-GAL4-tsGal80, slbo-GAL4, or upd-GAL4 tsGAL80 crosses, flies were incubated at 29°C for ≥14 hr prior to dissection to produce optimal GAL4-UAS transgene expression. c306-GAL4 is expressed early and more broadly in border cells, polar cells, and terminal (anterior and posterior) follicle cells (Figure 1—figure supplement 1A; Figure 1—figure supplement 2B; Silver and Montell, 2001). During oogenesis, slbo-GAL4 turns on later than c306-GAL4, and is expressed in border cells but not polar cells, as well as a few anterior and posterior follicle cells at stage 9 (Figure 1—figure supplement 1B; Figure 1—figure supplement 2C,D; Silver and Montell, 2001; Rørth et al., 1998). upd-GAL4 is restricted to polar cells at all stages of oogenesis (Figure 1—figure supplement 2C,H; Cai et al., 2014). Mosaic mutant clones of flw were generated using the FLP-FRT system (Xu and Rubin, 1993). The flw^FP41 FRT 19A line was crossed to ubi-mRFP.nls hsFLP FRT19A; the resulting progeny were heat shocked for 1 hr at 37°C, two times a day for 3 d, followed by 3 d at 25°C prior to fattening and dissection. Mutant clones were identified by loss of nuclear RFP signal from ubi-mRFP.nls.

The following Drosophila strains (with indicated stock numbers) were obtained from the Bloomington Drosophila Stock Center (BDSC, Bloomington, IN, USA): c306-GAL4 (3743), UAS-NiPp1.HA (23711), UAS-Pp1-87B.HA (24098), UAS-Pp1-13C.HA (23701), UAS-Pp1α−96A.HA (23700), UAS-hPPP1CC (64394), UAS-mCD8-ChRFP (27392), UAS-mCherry RNAi (35785), UAS-Pp2B-14D RNAi (25929, 40872), UAS-mts RNAi (27723, 38337, 57034, 60342), UAS-Pp4-19C RNAi (27726, 38372, 57823), UAS-CanA-14F RNAi (38966), UAS-PpD3 RNAi (57307), UAS-PpV RNAi (57765), UAS-CanA1 RNAi (25850), UAS-CG11597 RNAi (57047, 61988), UAS-rgdC RNAi (60076), UAS-Flw RNAi (38336), UAS-β-Catenin RNAi JF01252 (31305), flw^FP41 FRT 19A (51338), ubi-mRFP.nls hsFLP FRT19A (31418), UAS-PLCδ-PH-EGFP (‘membrane GFP’; 39693), UAS-GFP.nls (4776).

The following Drosophila strains (with indicated stock numbers) were obtained from the Vienna Drosophila Resource Center (VDRC, Vienna, Austria): UAS-Pp1α−96A RNAi (v27673), UAS-Pp1-87B RNAi (v35024), UAS-Pp1-13C RNAi (v29058), UAS-Flw RNAi (v29622, v104677), UAS-Mbs RNAi (v105762), UAS-Pp2B-14D RNAi (v46873), UAS-Pp4-19c RNAi (25317), UAS-E-Cadherin RNAi (v27082, v103962), UAS-β-Catenin RNAi (v107344), UAS-α-Catenin RNAi (v20123, v107298), UAS-Sqh RNAi (v7916), fTRG Pp1α −96A (v318084), fTRG Sqh (v318484).

Other Drosophila strains used in this study were: slbo-GAL4, slbo-GAL4 UAS-mCD8-GFP, upd-GAL4;; tsGAL80, and slbo-LifeAct-GFP line 2M (from D. Montell, University of California, Santa Barbara, Santa Barbara, CA, USA), flw^CPTI002264 protein trap (line 115284, Kyoto Stock Center, Kyoto, Japan), UAS-mCherry-Jupiter (from C. Doe, University of Oregon, Eugene, OR, USA), UAS-Rac FRET (Wang et al., 2010), UAS-Rho FRET/CyO; UAS-Rho FRET/TM6B (Qin et al., 2017), and UAS-Flw.HA (FlyORF) (Bischof et al., 2013). The c306-GAL4 tsGAL80 (Aranjuez et al., 2016) and c306-GAL4 tsGAL80/FM6; UAS-NiPp1.HA/TM3 Ser stocks were created in our lab.

Fertility was determined according to established methods (Tootle and Spradling, 2008). Briefly, four c306-GAL4 tsGAL80/FM6; Sco/CyO (control) or c306-GAL4 tsGAL80/FM6; UAS-NiPP1/TM3 Ser (experimental) females were outcrossed to four w¹¹¹⁸ males. The flies were allowed to mate for 2 days followed by a 24 hr egg lay at 30 °C on fresh food medium supplemented with yeast. Adults were then removed and the progeny allowed to develop in the vial at 25 °C; the food was periodically monitored to avoid drying out. Scoring of eclosed adult progeny from each vial was performed 16–20 d after egg laying and reported as the average progeny per female.

Fly ovaries from 3- to 5-d-old females were dissected in Schneider’s Drosophila Medium (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Seradigm FBS; VWR, Radnor, PA, USA). Ovaries were kept whole or dissected into individual egg chambers, followed by fixation for 10 min using 4% methanol-free formaldehyde (Polysciences, Warrington, PA, USA) in 0.1 M potassium phosphate buffer, pH 7.4, or in 1X Phosphate Buffered Saline (PBS). Washes and antibody incubations were performed in ‘NP40 block’ (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5% NP40, 5 mg/ml bovine serum albumin [BSA]). For α-Catenin immunostaining, dissected egg chambers were fixed for 20 min in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in potassium phosphate buffer, pH 7.4, followed by a separate blocking step for 30 min (2% BSA in 1x PBS) prior to each antibody incubation. For p-Sqh antibody staining, ovaries were fixed for 5 min in 8% methanol-free formaldehyde. For the F-actin staining in Figure 6, the entire dissection procedure was performed in less than 10 min to preserve F-actin structures, followed by fixation in the presence of Phalloidin at 1:400 dilution; after washing off the fix, the egg chambers were incubated in Phalloidin at 1:400 for 2 h (Spracklen et al., 2014).

The following primary antibodies from the Developmental Studies Hybridoma Bank (DSHB, University of Iowa, Iowa City, IA, USA) were used at the indicated concentrations: rat anti-E-Cadherin 1:10 (DCAD2), mouse anti-Fasciclin III 1:10 (FasIII; 7G10), mouse anti-Arm (β-Catenin) 1:75 (N2-7A1), concentrated rat anti-α-Catenin 1:1000 (DCAT1), mouse anti-Eyes Absent 1:100 (eya10H6), mouse anti-Lamin Dm0 1:10 (ADL67.10), and mouse anti-Singed 1:25 (Sn7C). Additional primary antibodies used were: rabbit anti-Phospho-Myosin Light Chain 2 (Ser19) 1:10 (#3671, Cell Science Technology, Danvers, MA, USA), rat anti-HA 1:1000 (11867423001, Millipore Sigma, Burlington, MA, USA), rabbit anti-Mbs 1:200 (from C. Tan, University of Missouri, Columbia, MO, USA); rabbit anti-GFP polyclonal 1:1000-1:2000 (A-11122, Thermo Fisher Scientific), chicken anti-GFP polyclonal 1:1000 (ab13970, Abcam, Cambridge, MA, USA), rabbit anti-PPP1R8 (NiPP1) polyclonal 1:100 (HPA027452, Millipore Sigma), rat anti-Slbo 1:2000 (from P. Rørth, Institute of Molecular and Cell Biology, Singapore). Alexa Fluor 488, 568, or 647 secondary antibodies (Thermo Fisher Scientific) were used at 1:400 dilution. Alexa Fluor Phalloidin (488 or 568; Thermo Fisher Scientific) and Phalloidin–Atto 647N (Millipore Sigma) were used at 1:400 dilution. 4’,6-Diamidino-2-phenylindole (DAPI, Millipore Sigma) was used at 0.05 µg/ml. Egg chambers were mounted on slides with Aqua-Poly/Mount (Polysciences) or FluorSave Reagent (Millipore Sigma) for imaging.

Images of fixed egg chambers were acquired with an upright Zeiss AxioImager Z1 microscope and Apotome.2 optical sectioning, or on a Zeiss LSM 880 confocal microscope with or without Airyscan (KSU College of Veterinary Medicine Confocal Core), using either a 20 × 0.75 numerical aperture (NA) or 40 × 1.3 NA oil-immersion objective.

Live time-lapse imaging was performed as described (Prasad and Montell, 2007; Dai and Montell, 2016). Briefly, ovarioles were dissected in room-temperature sterile live imaging media (Schneider’s Drosophila Medium, pH 6.95, with 15–20% FBS). Fresh live imaging media, supplemented with 0.2 µg/ml bovine insulin (Cell Applications, San Diego, CA, USA), was added to the sample prior to mounting on a lumox dish 50 (94.6077.410; Sarstedt, Newton, NC, USA). Time-lapse videos were generally acquired at intervals of 2–3 min for 3–6 hr using a 20 × Plan Apochromat 0.75 NA objective, a Zeiss Colibri LED light source, and a Zeiss Axiocam 503 mono camera. The LED light intensity was experimentally adjusted to maximize fluorescence signal and to minimize phototoxicity of the live sample. Live time-lapse Sqh-GFP imaging was performed on a Zeiss LSM 880 confocal, as described (Dai and Montell, 2016), with a 40 × 1.2 NA water-immersion objective using an interval of 1 min for up to 20 min total time and a laser setting of 1.5%. Imaging gain and other acquisition parameters were the same, except that the range of z-stacks varied slightly depending on the sample. In some cases, multiple z-stacks were acquired and merged in Zeiss AxioVision, Zeiss ZEN 2, or FIJI (Schindelin et al., 2012) to produce a single, in-focus time-lapse video.

FRET images (Rac FRET, Rho FRET) of live cultured egg chambers were acquired with a Zeiss LSM710 microscope essentially as described (Wang et al., 2010). A 40 × 1.3 NA oil inverted objective was used to capture single high-resolution stationary images. A 458 nm laser was used to excite the sample. CFP and YFP emission signals were collected through channel I (470–510 nm) and channel II (525–600 nm), respectively. The CFP and YFP channels were acquired simultaneously for most experiments. Sequential acquisition of CFP and YFP channels was tested but produced the same result as simultaneous acquisition.

Image measurements and editing were performed using Zeiss ZEN or FIJI (Schindelin et al., 2012). Analyses of live border cell migration time-lapse videos was performed using Zeiss ZEN software. The migration speed was calculated from the duration of border cell movement. Protrusion quantification was performed as described (Sawant et al., 2018). Briefly, a circle was drawn around the cell cluster, and extensions greater than 1.5 µm outside the circle were defined as protrusions (Figure 5—figure supplement 1A). Protrusions were classified as directed to the front (0°−45° and 0°−315°), side (45°−135°and 225°−315°), or back (135°−225°), based on their positions within the cluster. The first 1 hr of each video was used for protrusion quantification.

To determine the number of cells per cluster, egg chambers were stained for the nuclear envelope marker Lamin, the DNA stain DAPI, and the cell membrane marker E-Cadherin. Only clusters that had delaminated, moved forward, and had any detectable E-Cadherin were imaged. This allowed confidence that the scored cells were border cells. Acquisition of z-stacks that encompassed the entire cluster (border cells and polar cells) were defined by nuclear Lamin signal. This was followed by manual counting of the nuclei from the resulting images.

The circularity of border cells was measured in FIJI. Individual border cells were outlined manually based on the PLCδ-PH-GFP signal using the ‘Freehand Selections’ tool. Within the ‘Set Measurements’ analysis tool, ‘shape descriptors’ was selected, followed by the ‘Measure’ function, which provided a measurement of circularity. A value of 1.0 indicates a perfect circle, whereas 0.0 represents an extremely elongated shape.

Measurements of E-Cadherin and β-Catenin intensity at cell–cell junctions were performed on egg chambers that were stained using identical conditions. Samples were imaged with a 40 × 1.3 NA oil objective. Identical confocal laser settings were used for each channel and a full z-stack of the cluster was produced. Images were then subjected to 3D reconstruction through the ‘3D Project’ function in FIJI. Border cell-border cell (BC-BC) contacts and nurse cell-nurse cell (NC-NC) contacts were manually identified, a line (width set as 6) drawn, and mean fluorescence intensity across the line was obtained using the ‘measure’ tool. A ratio of BC-BC intensity versus NC-NC intensity was calculated to normalize protein levels.

To measure colocalization between Mbs and Flw, or Mbs and Pp1α−96A, the ‘RGB Profiler’ FIJI plugin was used. After converting the image to RGB, a line was drawn across the whole border cell cluster to generate the image intensity plot. The localization patterns of F-actin and Mbs with Pp1α−96A-GFP and Flw-YFP were measured through the ‘Analyze >Plot Profile’ function in FIJI. A line was drawn across the border cells and polar cells and the pixel intensity value was obtained across the line. The values for each channel were normalized to the highest pixel value, and a scatter plot showing F-actin and DAPI was generated in Microsoft Excel.

For Rho-FRET and Rac-FRET, the CFP and YFP images were first processed in ImageJ. A background region of interest was subtracted from the original image. The YFP images were registered to CFP images using the TurboReg plugin. The Gaussian smooth filter was then applied to both channels. The YFP image was thresholded and converted to a binary mask with the background set to zero. The final ratio image was generated in MATLAB, during which only the unmasked pixels were calculated as described (Wang et al., 2010).

Figures were assembled in Adobe Photoshop CC. Illustrations were created in Affinity Designer (Serif, Nottingham, United Kingdom). Videos were assembled in Zeiss AxioVision 4.8, Zeiss ZEN 2, or FIJI. Graphs and statistical tests were performed using GraphPad Prism 7 or Prism 8 (GraphPad Software, San Diego, CA, USA). The statistical tests and p values are listed in the figure legends.