DE-cadherin and Myosin II balance regulates furrow length for onset of polygon shape in syncytial Drosophila embryos

Spherical to polygonal cell shape transition is the first morphogenetic event that occurs in metazoan embryogenesis in the formation of epithelial-like cells (St Johnston and Ahringer, 2010). A polygonal array of epithelial-like cells containing a distinct apical and baso-lateral domain is formed upon cell divisions in embryogenesis. Mature epithelial cells are organized as a conserved hexagon-dominated polygonal array in various tissues in metazoans (Gibson et al., 2006). The polygon distribution in an epithelial tissue is determined by an interplay of junctional, cytoskeletal and physical factors, which together result in surface energy minimization (Classen et al., 2005; Farhadifar et al., 2007; Gibson et al., 2006; Sugimura and Ishihara, 2013).

Two fundamental forces that drive cell shape changes and, therefore, tissue morphogenesis are actomyosin-based contractions and cadherin-mediated cell–cell adhesion. These two forces play opposing roles, with adhesion facilitating expansion of the contact area between cells while actomyosin contractions work to reduce it (Heisenberg and Bellaïche, 2013; Maître and Heisenberg, 2013). A change in the balance of these two forces transforms the shapes of cells, as observed in two dimensions (Blankenship et al., 2006; Martin et al., 2009; Roh-Johnson et al., 2012). In addition, these forces also act along the third dimension of the lateral membrane (LM). The LM domain in mature columnar epithelial cells accounts for up to 60% of the total cell surface area (Tang, 2017). Fine tuning Myosin II (MyoII)-based contractility is important for regulating LM length as increase or decrease in MyoII activity along the LM results in shortening or expansion of the cell height, respectively (Widmann and Dahmann, 2009; Yu et al., 2016). An interplay of adhesion and contractility, therefore, also regulates LM length and stability along with epithelial cell shapes.

Here, we analyze the transition of plasma membrane shape from circular to polygonal during early Drosophila embryogenesis, and assess the relative contribution of adhesion and contractility in driving polygonal plasma membrane organization. Drosophila embryogenesis begins with nine nuclear division cycles (NC) taking place in the embryo interior, followed by migration of nuclei to embryo periphery. NC10–NC13 occur at the cortex without cytokinesis, where each nucleus is incompletely surrounded by the plasma membrane (PM), forming a syncytial blastoderm. NC11–NC13 result in cells with epithelial-like characteristics both in terms of polarity and polygonal PM organization (Holly et al., 2015; Mavrakis et al., 2009; McCartney et al., 2001; Zhang et al., 1996). Each NC begins with the formation of almost circular ‘membrane buds’ above the interphase nuclei, in the form of actin caps, which have cortical enrichment of actin in them. The intercap region, that is, the boundary between the actin caps is enriched in MyoII (Foe et al., 2000). The caps expand based on Arp2/3-dependent actin polymerization, pushing against the actomyosin borders. Depletion of Arp2/3 complex activity leads to small rounded caps, and depletion of MyoII activity leads to expanded caps (Cao et al., 2010; Sommi et al., 2011; Zhang et al., 2018). MyoII activity results in buckling forces on the cap borders, bending it into a nascent furrow membrane as the caps collide (Zhang et al., 2018). Cap expansion and furrow extension leads to a polygonal plasma membrane organization in each cortical syncytial cycle.

The furrow resembles an LM that extends perpendicular to the embryo surface between adjacent nuclei and contains junctional proteins (Foe and Alberts, 1983; Frescas et al., 2006; Harris and Peifer, 2004; Schejter and Wieschaus, 1993; Sullivan and Theurkauf, 1995). This furrow initiates during each cycle at interphase, reaches a maximum length at metaphase and then retracts during anaphase to telophase (Fig. S1A and Holly et al., 2015; Sherlekar and Rikhy, 2016; Silverman-Gavrila et al., 2008). Mutations in several actin regulatory proteins cause a decrease in the furrow length (Afshar et al., 2000; Stevenson et al., 2002). This property of extensive furrow or LM remodeling along with the availability of mutants that affect furrow length, allows an analysis of LM length in the initiation of polygonal architecture.

In our study, we focus on the step at which circular caps meet and initiate furrows that extend deeper into the embryos with simultaneous transition of the plasma membrane shape from circular to polygonal. We find that the onset of polygonal packing in the Drosophila syncytial blastoderm embryo correlates with extension of the furrow or LM increased in length from 4.75 to 5.75 μm, during each cortical syncytial division cycle. DE-cadherin (DE-cad; also known as Shotgun) and MyoII (comprising Zipper and Spaghetti squash in flies) levels at the furrow regulate the transition to polygonal shape. Depletion of DE-cad or an increase in MyoII leads to loss of polygon architecture formation.

Each nucleo-cytoplasmic domain of the syncytial Drosophila blastoderm embryo shows compartmentalization of molecules and organelles in the PM and cytoplasm (Frescas et al., 2006; Mavrakis et al., 2009). Hence, we refer to these domains as ‘syncytial cells’. The PM of syncytial cells is organized into polygons even before complete cell formation during cellularization (Holly et al., 2015; Mavrakis et al., 2009; McCartney et al., 2001; Zhang et al., 1996). To determine the temporal emergence of this polygonal architecture in the syncytial embryo, we imaged living embryos (Fig. S1A) expressing the GFP-tagged PH domain of GRP1 (tGPH). tGPH preferentially binds phosphatidylinositol (3,4,5)-trisphosphate (PIP3) in the PM (Britton et al., 2002), and labeled apical and lateral regions uniformly (Movie 1). We started with characterizing furrow length dynamics in the cortical syncytial cycles. The metaphase furrow length increased with each NC reaching ∼6 μm in NC11, 9 μm in NC12 and 12 μm in NC13 (Fig. 1A); these values were similar to those from previous reports (Holly et al., 2015). Within each NC, furrow length increased with respect to the time from interphase to metaphase (Fig. S1B). In NC13, there was an increase in furrow length with a slower phase until prophase and a faster phase from late prophase to metaphase (Fig. 1B). These phases have been recently found to occur due to the onset of zygotic gene expression (Xie and Blankenship, 2018).

We next tracked the change in shape of syncytial cell PM with respect to furrow length. Circular to polygon PM transition occurred during each cortical division cycle from NC11–NC14 (Movies 1,2). The PM organization changed from circular to polygonal during the syncytial division cycle, correlating with the increase in furrow length (Fig. 1C; Movies 1,2). To quantify this, we estimated the circularity (C) of the PM in syncytial cells, as a measure of the two-dimensional cell shape transition from circular to polygonal as the furrow changed lengths in NC13. Circularity decreased with time and with increase in the furrow length in NC11–NC13 (Fig. 1D; Fig. S1C, Movie 2). Circularity is equal to 1 for a perfect circle and less than 1 for angular shapes (Fig. S1D, Table S1; Xue and Sokac, 2016). Syncytial cells in interphase of NC13 with short furrows (approximately ≤4 μm), had an almost circular PM organization (C=0.954±0.015) and transformed into a polygonal shape (C=0.7–0.85, hexagons=0.832±0.044) in metaphase with longer furrow lengths (approximately ≥10 μm) (Fig. 1C,D; Table S1). This was also seen for NC11 and NC12 (Fig. S1C). Polygonality was seen in each syncytial cycle from NC11–NC13 when the furrows ingressed from 4.75 to 5.75 μm (Fig. 1D; Fig. S1C). It is interesting to note that this length range occurs at different syncytial cycle phases in NC11, NC12 and 1NC3. In NC11 polygonality arose just after nuclear envelope breakdown (NEB), a point before metaphase of the syncytial cycle. In NC12 and NC13, polygonality arose in prophase, much before NEB. Polygonality, therefore, arises at different stages in each syncytial division cycle. The furrow length range of 4.75–5.75 μm, at which polygonal transition occurred, however, remained unchanged and can therefore be taken as a testable parameter for a consistent change to polygonality in the syncytial Drosophila blastoderm embryo.

As the embryo progressed through NC11–NC14, the number of syncytial cells increased while the size of the polygon and edge length decreased. We, therefore, asked whether polygonality onset was correlated with an increase in crowding as a result of increased syncytial cell numbers. For this, we measured the circularity of syncytial cells across NC11–NC14 at two different furrow lengths, i.e. at a length below and above the range of 4.75–5.75 μm, irrespective of the cell cycle phase (Fig. 1E,F). The PM around syncytial cells remained circular in all NCs, when the furrow length was ∼4 μm (Fig. 1E,G). The PM shape was polygonal when the furrow length was ∼6.5 μm (Fig. 1F,G). Polygonality in the syncytial Drosophila embryo is, therefore, a function of furrow length, independently of number of syncytial cells or edge lengths (Fig. 1H).

Since there was a transition of PM shape from circular to polygonal when the furrow length increased beyond 5.75 μm, we assessed whether there was a change in the tension at the edges or junctions above this length. In order to do this, we performed laser ablations at the point of contact between syncytial cells below and above this length range (Fig. 2A–D,G). There was no significant difference in recoil velocity in edges of syncytial cells below 4 μm when compared to the control non-ablated neighboring edges (Fig. 2A,B,G). Interestingly, the recoil velocity was greater in the edges from furrows of ∼6.5 μm in length (Fig. 2C,D,G). Junctional tension was also probed at metaphase when the furrow length was the maximum, but no significant recoil was observed at this stage compared to that in control non-ablated neighboring junctions (Fig. 2E–G). This shows that edges of polygonal syncytial cells have higher tension than the points of contact between circular syncytial cells or even cells at metaphase. This could be either because of increased contractility or increased adhesion strength. Hence, the furrow length that supports polygonal shape transition is likely to have recruitment or reorganization of specific molecules to bring about this increase in tension at the PM.

Cellular force generation typically relies on molecular motors like MyoII, which can bind to and reorganize actin filaments in the cell. The cytoskeletal changes within a cell are then transmitted to other cells via cell–cell and cell–matrix adhesion molecules, like cadherin and integrins. Adhesive forces tend to expand the surface area of contact between cells while the actomyosin-based contractility opposes this expansion and tends to minimize the contact area. A balance of these two dominant forces results in an energy-minimized state that gives epithelial cells/tissues their characteristic shapes (Heisenberg and Bellaïche, 2013). Thus, to evaluate the role of key molecular players in achieving the shape transition during lateral furrow extension, we focused our attention to the roles of DE-cad in furrow membrane adhesion and MyoII in regulating contractility in the furrow membranes. DE-cad is present at the polygonal plasma membrane in syncytial embryos (Harris and Peifer, 2004; McCartney et al., 2001), and MyoII is recruited to the furrow in interphase and prophase and is lost from the furrow in metaphase of the syncytial cycle (Foe et al., 2000; Royou et al., 2002).

We further determined the relative dynamics of DE-cad and MyoII in a quantitative manner within one syncytial NC. We performed live imaging of embryos co-expressing the mCherry-tagged MyoII regulatory light chain Spaghetti squash (sqh-Sqh–mCherry) and GFP tagged DE-cad (ubi-DE-cad–GFP) and quantified their intensities on the PM from interphase to metaphase of NC13 (Fig. 3A–C). We observed that there was an increase in total DE-cad intensity throughout the NC until metaphase (Fig. 3C; Movie 3). Conversely, and as reported in literature (Foe et al., 2000; Royou et al., 2002) Sqh–mCherry intensity dropped consistently from interphase to metaphase (Fig. 3C; Movie 3). In addition, at metaphase, Sqh–mCherry was completely lost from the furrow membrane and became cytosolic, while DE-cad was retained on the PM (Fig. 3A–C; Movie 3). Sqh–mCherry was again recruited to the PM during late telophase at the sites of new furrow formation, followed by DE-cad accumulation (Fig. S2). Therefore, MyoII appears to be dispensable for furrow stability at the point of maximum extension at metaphase. This is also a point at which there is loss of tension in the membrane as revealed by the laser ablation experiments (Fig. 2E–G). We also evaluated the relative change in endogenous DE-cad and MyoII levels across the syncytial cycle. We found that DE-cad–GFP when expressed with the endogenous promoter from the endo-DE-cad–GFP (Huang et al., 2009) transgene, also increased from interphase to metaphase in a similar manner to what was seen with ubi-DE-cad–GFP. Endogenous MyoII distribution was estimated as a membrane to cytosol ratio by immunostaining against the heavy chain of MyoII, Zipper. Similar to sqh-Sqh–mCherry, Zipper was present on the membrane in interphase and was cytosolic in metaphase (Fig. S2B–E).

Next, we studied the mobility of both DE-cad and MyoII with ubi-DE-cad–GFP and sqh-Sqh–GFP at the furrow when the length was below and above the transition to polygonal shape by performing fluorescence recovery after photobleaching (FRAP). ubi-DE-cad–GFP mobility did not show significant differences in its mobile fraction before and after polygonal shape transition, suggesting that its organization does not change during this transition (Fig. 3D–G). However, the mobile fraction of Sqh–GFP was significantly reduced after polygonal shape transition, suggesting that the MyoII becomes more organized on the plasma membrane (Fig. 3H–K). Together with increase in junctional tension after polygonal shape transition, this suggests that, with increased DE-cad levels, MyoII becomes organized and immobilized on the junctions. In order to test how the forces generated by these two molecules regulate the furrow length and polygon shape, we studied mutants of DE-cad and MyoII activity regulators.

We assessed DE-cad mutant embryos for furrow length and circularity in the syncytial cycles. shotgun (shg) RNAi (shgⁱ) was maternally expressed in embryos (see Materials and Methods section for details). Embryos developing to syncytial stages had significantly reduced DE-cad staining (Fig. 4A–C). The MyoII levels were estimated by immunostaining for Zipper, the MyoII heavy chain subunit, and live imaging for Sqh–GFP. Dlg, a lateral domain polarity complex protein, was used to visualise the lateral furrow membranes. Zipper was enriched on the Dlg-marked membrane region as compared to the cytosol in shgⁱ embryos, although the levels were reduced as compared to controls (Fig. 4A′,B′,D). Live imaging with sqh-Sqh–GFP, on the other hand, did not show any significant difference in Sqh levels in shgⁱ embryos with respect to controls (Fig. 4E; Fig. S3A). We, therefore, conclude that DE-cad knockdown does not have a drastic effect on MyoII distribution.

Live imaging of shgⁱ embryos with tGPH showed defects during syncytial division cycles, which could be categorised into three classes: 22% embryos showed disruption of polygonal architecture, 17% showed polygonal organization with loose PM seen in the form of a spread tGPH signal, and 61% showed polygonal organization of the PM (Fig. 4F; Movies 4–6). This variation in architecture could be attributed to partial knockdown of DE-cad. The shgⁱ embryos that showed a disruption of polygonal architecture did not show formation of furrows. All categories of embryos were estimated for furrow length, and it was found to be significantly shorter than the control embryos across syncytial cycles 11–13 (Fig. 4G). There was a consistent increase in the circularity of the metaphase syncytial cell shapes of all NCs compared to that in control embryos (Fig. 4H). Circularity comparison at maximum furrow length in NC13 metaphase showed that shgⁱ embryos with furrows below the polygonal transition length range were circular as compared to those above the length range (Fig. 4I). The loss of furrows seen in shgⁱ is similar to that reported in a preliminary analysis of DE-cad mutants in syncytial stages (Wang et al., 2004).

Actin caps in interphase show expansion until they meet and stabilize their area. The cap expansion is promoted by Arp2/3-driven actin polymerization, while cap area restriction occurs through MyoII-driven contraction. As a result, a loss of MyoII activity leads to large caps while an increase in MyoII activity leads to small caps (Afshar et al., 2000; Cao et al., 2010; Stevenson et al., 2002; Zhang et al., 2018). In order to confirm the effect on contractility, we measured cap areas before and after the polygonal transition, as a functional readout of contractility at this stage, for all the perturbations with DE-cad and MyoII. The area of actin caps increased in control embryos at a length when polygonal transition was achieved. Consistent with the lack of effect on recruitment of MyoII, DE-cad loss did not affect cap expansion, and the cap area increased by a similar amount when above the polygonal transition length range to the increase seen in control embryos (Fig. S4A,B).

Taken together, these results show that DE-cad plays a significant role in keeping adjacent furrow membranes together for the formation of edges and appropriate extension in order to achieve onset of polygonality in the syncytial Drosophila embryo.

In order to change the balance of contractility and adhesion during shape transition, we depleted levels of activated MyoII by altering the levels of MyoII regulators. Rho-GTP is an important regulator of MyoII activity. RhoGEF2, a Rho GTP exchange factor (GEF), functions in furrow elongation in the syncytial cycles and during cellularization (Cao et al., 2008; Crest et al., 2012; Padash Barmchi et al., 2005). Rho-GTP activates ROCK (Drosophila Rok) and Rok, in turn, activates MyoII and deactivates the MyoII phosphatase through phosphorylation. MyoII is deactivated through its dephosphorylation by MyoII phosphatase and this is defective in mutants of MyoII-binding subunit (MBS), a subunit of the MyoII phosphatase (Mizuno et al., 1999; Mizuno et al., 2002; Tan et al., 2003; Zhang et al., 2018).

We analyzed the effect of reducing MyoII-based contractility on the onset of polygon formation. To this end, RhoGEF2 RNAi (RhoGEF2ⁱ) was maternally expressed in the embryos. RhoGEF2ⁱ has been previously shown to deplete RhoGEF2 protein from embryos and have a similar effect to loss of function mutants (Jiang and Harris, 2019; Sherlekar and Rikhy, 2016). In order to deplete active MyoII more directly, we also knocked down Rok by maternally expressing rok RNAi (rokⁱ) (see Materials and Methods for details). Both RhoGEF2ⁱ- and rokⁱ-expressing embryos showed ubi-DE-cad–GFP levels that were similar to those controls (Fig. 5A–D). As expected, RhoGEF2ⁱ and rokⁱ embryos showed cytosolic Zipper staining (Fig. 5A′–C′,E). In addition, RhoGEF2ⁱ also showed lowering of sqh-Sqh-mCherry as seen with live imaging (Fig. S3B,C).

Live imaging of RhoGEF2ⁱ and rokⁱ embryos with tGPH showed that the furrows are slightly but significantly shorter than in the controls. These embryos also showed furrows of variable lengths as reported previously (Fig. 5F,G; Movies 7,8; Sherlekar and Rikhy, 2016; Zhang et al., 2018). The variation in the furrow lengths was also partly greater in rokⁱ embryos because different regions of the embryos proceeded in different NCs, which reflects problems in nuclear axial expansion (Royou et al., 2002). The PM in RhoGEF2ⁱ and rokⁱ embryos was not taut unlike in controls, suggesting lowered tension in the cortex. We also assessed the actin cap area as an indicator of contractility in the cortex. The loss of contractility in RhoGEF2ⁱ was reflected in larger cap areas in interphase at the furrow length before polygonal shape transition, without a significant further expansion after polygonal shape transition. However, due to increased variation in cap areas in the rokⁱ embryos the difference was not found to be significant (Movie 8; Fig. S4A,B).

If the increase in contractility were balanced by increased adhesion between LMs (Figs 2,3) to form the polygonal architecture at an optimal LM length, then decreasing contractility would provide a paradigm to ask whether edge formation could occur at a lower LM length. An alteration in the extent of contractility with respect to adhesion would lead to change in circularity and we, therefore, measured circularity below and above the furrow length range for polygon formation, that is, at ∼4 µm and ∼6.5 µm, respectively in both RhoGEF2ⁱ- and rokⁱ-expressing embryos (Fig. 5H). Interestingly, the circularity of the syncytial PM at a furrow length below the polygonal transition range was decreased in RhoGEF2ⁱ- and rokⁱ-expressing embryos (Fig. 5H). This corresponded to the polygonal shape and was similar to what was achieved in control embryos at lengths above 5.75 µm. This shows that with a decrease in MyoII activity, polygonality is achieved at furrow length below that in controls or, in other words, the optimal length for circular to polygon transition shifted to a lower furrow length.

Thus, lowering MyoII-based contractility does not affect furrow initiation but allows edge formation to take place earlier and, thus, polygonality occurs at a shorter furrow length.

Next, we checked for alteration of syncytial cell PM morphology upon increased MyoII activity. To this end, we analyzed embryos maternally overexpressing RhoGEF2 (RhoGEF2-OE) or expressing an RNAi against MBS (mbsⁱ) (see Materials and Methods for details). The RhoGEF2-OE and mbsⁱ embryos showed a loss of DE-cad immunostaining (Fig. 6A–D). As expected, RhoGEF2-OE and mbsⁱ embryos showed a retention of Zipper on the apical plasma membrane in patches, even in metaphase, indicating an increase in MyoII activity in these mutants (Fig. 6A′–C′,E). The membrane here was marked by co-staining with Dlg. Interestingly, this Zipper increase corresponded to a complete loss of furrow extension, and circular cells with no polygonal organization, as seen with both phalloidin and Dlg staining (Fig. 6A–C,A′–C′). Further live imaging of RhoGEF2-OE embryos with DE-cad–GFP and Sqh–mCherry showed that some DE-cad–GFP remained on the membrane in between caps and was unable to organize into the junctions, possibly due to the inability of caps to expand and meet as a result of increased MyoII-based contractility (Fig. S3B,D).

Live imaging of RhoGEF2-OE and mbsⁱ embryos along with tGPH showed that syncytial cells were unable to transition into a polygonal state with very short furrow lengths, well below the length range where polygonal transition is seen in controls in NC11–NC13 (Fig. 6F,G; Movies 9,10). Plotting the circularity of the cells with different metaphase furrow lengths in RhoGEF2-OE embryos showed that short furrow lengths below the polygonal transition length range were abundant and these cells showed a high circularity. The RhoGEF2-OE embryos occasionally had furrows above 6 µm and, interestingly, they also showed an increased circularity compared to the control embryos (Fig. 6H). Increased contractility of the cortex was also reflected in cap area measurements, which showed smaller caps to begin with, as compared to control embryos, which did not undergo much further expansion (Fig. S4A,B). Thus, increased MyoII activity led to loss of cap expansion and the inability to initiate furrows. In addition, lack of furrow extension beyond the 4.75–5.75 µm length range gave rise to circular cells.

Our study proposes an LM length-based model for the initiation of polygonal epithelial-like architecture, in the earliest morphogenetic event of syncytial blastoderm development in Drosophila embryogenesis. The syncytial blastoderm shows cycling of cell shape from circular to polygonal coincident with the recruitment–removal cycle of MyoII onto the membrane in each NC. Because of this and a dynamic lateral furrow membrane, the syncytial blastoderm forms an effective model system to elucidate the relative contribution of key components of adhesion and contractility in the regulation of polygon shape transition. We show that ingression of furrows above an optimal length allows the coincident formation of edges between adjacent PM regions and the transition of syncytial cell shapes from circular to polygonal. The optimal lateral furrow length is achieved during each cortical syncytial division cycle from 11 to 13. The stage of the syncytial division at which this length is achieved varies for NC11, NC12 and NC13. Polygonal transition corresponds to the point at which the caps expand, touch each other and furrow extension happens. Cap expansion depends upon the centrosome movement in the syncytial cycle. Thus, it is likely that the polygonal plasma membrane transition is also coupled to the stage of cap expansion and centrosome movement in the syncytial cycle. It is also possible that polygonal PM organization occurs at a threshold LM length irrespective of the cell cycle phase, NC, cell crowding and cell size. The PM shape changes from circular to polygonal in NC11 for the first time, when the threshold LM length is also crossed for the first time during syncytial development. This threshold length is presumably determined by fine tuning the concentration and stabilization of MyoII relative to DE-cad in the LM. The threshold is likely to be the point at which MyoII-based contractility is balanced appropriately by DE-cad in the LM. This is supported by the fact that a decrease in MyoII activity allows the polygonality to be achieved earlier than for controls, whereas an increase in MyoII activity inhibits assembly of furrows and DE-cad is lost from the membrane, preventing the threshold from being achieved (Fig. 7).

Such a model also predicts that lowering adhesion may allow contractile forces to dominate and thus, phenocopy what is seen upon increased MyoII activity. Although we could not measure adhesion directly, we used DE-cad knockdown as a way to perturb adhesion. However, DE-cad knockdown resulted in increased circularity and disruption of polygonal architecture in only a small percentage of embryos, as compared to increasing MyoII activity, where circular PM organization was more prevalent. This may indicate either incomplete abrogation of the adhesion complex due to partial loss of DE-cad in the knockdown or the presence of other transmembrane adhesion molecules in the system at this stage that contribute to adhesive forces. Echinoid is one such cell adhesion molecule that cooperates with DE-cad to mediate adhesion in the wing epithelium. Loss of Echinoid is compensated for by the upregulation of DE-cad in the Echinoid mutant cells (Chang et al., 2011; Wei et al., 2005). It is possible that depletion of DE-cad is partly compensated for by the expression of Echinoid in the early embryo. Further studies testing the role of Echinoid in the syncytium may clarify this in the future.

Actin polymerization has a key role in furrow extension and loss of actin regulators like RhoGEF2, Diaphanous (Dia), Arp2/3 and SCAR decrease furrow lengths (Padash Barmchi et al., 2005; Sherlekar and Rikhy, 2016; Stevenson et al., 2002; Zallen et al., 2002). It is known that Arp2/3-based branched actin polymerization is essential for actin cap expansion in the syncytial embryo, while MyoII-based contractility tends to restrict it. Attachment of the growing actin network to the MyoII borders of the cap allows buckling of cap borders to initiate furrows (Zhang et al., 2018). However, loss of MyoII activity does not affect furrow initiation, because apposition of adjacent furrows is not perturbed in these mutants. To begin with, contractility dominates over adhesion in the actin cap, followed by an increase in DE-cad and a decrease in MyoII-levels that balance each other to reach the optimal membrane ingression. Reduction of contractility alone, allows adhesive forces to dominate and prematurely transition to a polygonal shape. Reduction of MyoII activity does not affect furrow initiation, but results in short furrows, although only marginally. Our data on the increase in tension in the plasma membrane in control embryos above the polygonal transition length range suggests that loss of MyoII activity is likely to destabilize the furrow membrane. Hence, a decrease in furrow length in these mutants could be a consequence of the loss of tension needed to stabilize polygonal architecture. Besides, RhoGEF2 also activates bundled actin polymerization by activating Dia (Grosshans et al., 2005; Kühn and Geyer, 2014; Padash Barmchi et al., 2005). Loss of furrows has been noted previously in embryos with a mutant dia⁵ allele in metaphase (Afshar et al., 2000). Thus, a decrease in furrow length, in this case, may also be due to reduction in Dia activation.

Changes in DE-cad and MyoII mobility or levels allows for plasticity of cell shape. A decrease in MyoII mobile fraction, as observed with FRAP, is usually indicative of increased pool of stable MyoII on the membrane resulting in increased tension (Munjal et al., 2015). An increase in the mobile fraction of DE-cad also corresponds to increased turnover that facilitates junctional remodeling during wing development (Classen et al., 2005; Iyer et al., 2019). In our study, we see a similar decrease in the mobile fraction of MyoII and a corresponding increase in tension after polygonal shape transition. DE-cad mobility, on the other hand, remains unchanged, but the levels of DE-cad on the lateral membrane increase in polygonal plasma membranes. E-cad is under constitutive tension because of the linkage to the underlying actomyosin cytoskeleton (Borghi et al., 2012). When subjected to forces, conformational modulation of α-catenin occurs at E-cad-based junctions, which recruits vinculin to promote junction stability and prevent junction disassembly (Borghi et al., 2012; Yonemura et al., 2010). Thus, even though E-cad mobility itself does not change drastically in our study or in the time scales across which polygonal transition is obtained, a change in the underlying MyoII organization can indirectly lead to increased tension on the junctions via E-cad. E-cad adhesion sites are also sites of active Rho signaling, as well as sites of actin polymerization via recruitment of actin nucleators like Arp2/3, formins and cortactin. This enables them to, in turn, contribute to the biogenesis and regulation of the underlying actomyosin (Lecuit and Yap, 2015). During junction maturation in mammalian cells, E-cad interaction with binding partners such as p120-catenin plays a crucial role in inhibiting Rho1 and active MyoII levels to decrease contractility along the LM in these cells (Yu et al., 2016). Therefore, one possibility is that increased DE-cad levels in the syncytial cycle of the Drosophila embryo allows for recruitment of embryo-specific cytoskeletal regulators that organize the underlying MyoII network causing increased tension during lateral furrow extension. The cytoskeletal regulators linking DE-cad adhesion and MyoII contractility remain to be identified in the syncytial blastoderm embryo. The increase in DE-cad levels may, thus, indirectly result in the change in tension to contribute to the transition of the cell shape from circular to polygonal. DE-cad depletion, on the other hand, could have an impact on adhesion of adjacent furrow membranes and organization of the underlying cytoskeletal network.

E-cad and MyoII enrichment occurs in a complementary manner in many instances. For example, the dorso-ventral borders have higher levels of DE-cad, while antero-posterior borders have higher MyoII during germ band extension in Drosophila embryogenesis. Loss of DE-cad with a simultaneous increase in medial and junctional MyoII levels results in loss of contacts between cells of the neuroectoderm, thus, giving rise to Drosophila neuroblast ingression (Lecuit and Yap, 2015; Simões et al., 2017). Shrinking of junctions to form rosettes and T1 to T2 transitions during cell intercalations also occur due to enrichment of MyoII (Bertet et al., 2004; Blankenship et al., 2006). Similarly, in our study, as the syncytial cells progress towards a taut and stable polygonal array from interphase to metaphase, the levels of DE-cad increase, while levels of MyoII decrease. The point of transition from circular to polygonal possibly represents the optimal levels of both molecules through their contribution to forces that can counterbalance each other and give rise to a stable LM and polygonal organization.

The concept of minimal LM attachment is particularly of interest when considering the reverse shape changes from polygonal to spherical that occur during epithelial-to-mesenchymal transition (EMT). EMT is an important step in cell migration during development and disease, involving dynamic modulations in cell–cell adhesion. For example, in Drosophila border cells, expansion of the sub-apical adherens junction complex containing DE-cad results in loss of detachment from the follicle cell epithelium (McDonald et al., 2008; Pinheiro and Montell, 2004). EMT involves a transition from strong junctional complexes, which favor static epithelial character, to having an increase in Rho GTPase activity, which favors cell motility (Lamouille et al., 2014). A fine balance of Rho GTPase activity at the adherens junctions results in either their stability or disassembly (Lamouille et al., 2014; Quiros and Nusrat, 2014). This is also seen in our study as both increasing and decreasing active RhoGTP levels impacts polygon formation. It is possible that a critical RhoGTP level regulates one state versus the other. An analysis of our optimal LM attachment model during these scenarios might reveal the point at which this balance tips off to either stability or disassembly. In conclusion, the lateral domain length could be an important parameter requiring alteration during epithelial remodeling in Drosophila development and in epithelia across different organisms.

Drosophila melanogaster stocks were raised in regular cornmeal agar at 25°C or 28°C. The stocks used in this study, their genotypes and their source are given in Table S2. Embryos obtained from Canton S flies or Canton S flies crossed to maternal α-tubulin Gal4-VP16 (mat-Gal4) or nanos-Gal4-VP16 (nos-Gal4) were used as control. Maternal driver line mat67;mat15 carrying maternal ɑ4 tubulin-Gal4-VP16, homozygous for chromosome II and III, was used for all RNAi and overexpression experiments except for shgⁱ and mbsⁱ for live imaging with tGPH. shgⁱ was crossed to a single chromosomal copy of nos-Gal4 and maintained at 18°C to lower the severity of phenotype and obtain fertilized eggs to perform experiments. F1 flies expressing shgⁱ with nos-Gal4 laid embryos that were arrested early in the pre-blastoderm stage of development when the cross was grown at 25 or 29°C and, hence, the experiments were performed at 18°C to allow for Gal4 dilution. The lethality of shgⁱ embryos was 100% (n=150) at 25°C and 29°C and 70% (n=200) at 18°C after 24 h. The shgⁱ-expressing embryos laid from the cross at 18°C gave an opportunity to test the effect of loss of DE-cad in the syncytial blastoderm embryo. RhoGEF2-OE embryos had 87% (n=200) lethality at 25°C after 24 h.

Embryos that were 0–2.5 h old were collected on sucrose agar plates, washed and dechorionated with 100% bleach for 1 min. A 1:1 mixture of 4% paraformaldehyde and heptane was used for embryo fixation for 20 min. Fixed embryos were then either hand-de-vitellinized (for phalloidin staining) or MeOH de-vitellinized, followed by three washes in 1× PBS with 0.3% Triton X-100 (1× PBST) and blocking in 2% BSA (in 1× PBST) (Sigma-Aldrich) for 1 h. Embryos were then incubated in primary antibody with appropriate dilution (see below) overnight, followed by three 1× PBST washes, and 1 h incubation in fluorescently coupled secondary antibodies (Molecular Probes) at 1:1000 dilution (Rothwell and Sullivan, 2007; Swedlow, 2011a,b). Hoechst 33258 (1:1000, Molecular Probes) was added for 5 min in 1× PBST to stain DNA. Finally, the embryos were washed three times in 1× PBST and mounted in Slow fade Gold antifade reagent (Molecular Probes). For Zipper and Dlg immunostainings, heat fixation was performed by adding dechorionated embryos in boiling 1× washing buffer (10× is 7% NaCl and 0.5% Triton X-100) and instantly adding ice-cold 1× washing buffer, followed by MeOH de-vitellinization.

The following antibodies were used in the study: rRat anti-DE-cadherin (DCAD2,RRID:AB_528120, DSHB), Mouse anti-Dlg (4F3,RRID:AB_528203, DSHB), Rabbit anti-Zipper was a gift from Thomas Jeffrey (Texas Tech University, TX). All the Alexa-Fluor-conjugated secondary antibodies and phalloidin, Hoechst 33258 (H-3569) and Slowfade Gold (S-36937) were purchased from Molecular Probes. The list of antibodies and dyes used in this study are given in Table S2.

Embryos that were 1–1.5 h old were collected and dechorionated with 100% bleach for 1 min, and mounted on coverslip-bottomed LabTek chambers. The chambers were filled with 1× PBS (Mavrakis et al., 2008) and imaged on either Zeiss Plan Apochromat 40×/ 1.4 NA oil objective or Leica SP8 40×/1.4 NA oil objective with a frame rate of 1.74s/frame and 2s/frame, respectively.

Live or fixed embryos were imaged using the Zeiss laser scanning confocal microscope LSM710, LSM780 (35 mW argon laser) and Leica laser scanning confocal microscope SP8 (60 mW Argon laser). The 40× objective with NA 1.4 was used to image living and fixed embryos. Laser power and gain were maintained with the range indicator mode, such that the 8-bit image acquired was within the 0–255 range. Averaging of 2 was used for both fixed and live imaging. Images were acquired with an optical section of 1.08 µm and 0.68 µm on the Zeiss and Leica microscopes, respectively.

tGPH- or DE-cad–GFP-expressing live embryos were used for visualizing the membrane for laser ablation experiments using the Leica TCS SP8 MP or Zeiss LSM780 microscopes. Regions of interest (ROIs) of size 0.4 µm in diameter were created at furrow edges before (as observed by assessing the circular membrane morphology at positions where adjacent membranes were touching each other) and after polygon transition and at metaphase. Ablations were performed with 800 nm multiphoton laser (3W MaiTai laser) at 80% power with ten iterations under 40× water immersion objective on Leica TCS SP8 MP or 30–45% laser power with 30 iterations under 63× oil immersion objective on Zeiss LSM780 MaiTai laser. The laser power and iterations were performed such that for a fixed iteration, the laser power was lowered until only photobleaching and recovery was observed with no physical damage. The power just above that required for photobleaching was then used for performing the ablation experiments. DE-cad always showed a small aggregation of GFP signal at the ablation ROI post ablation. Very high laser powers created a damage in the embryo and the vitelline membrane, and care was taken to use laser powers that cause ablation only at the ROI location in the cell. Aggregation of GFP signal in the DE-cad-GFP line was not seen in wing discs (data not shown) as documented previously (Farhadifar et al., 2007; Fernandez-Gonzalez et al., 2009). Please refer to the image analysis section for details on the estimation of recoil velocities from control and ablated regions.

Live embryos were imaged using Zeiss laser scanning confocal microscope LSM710. The 63× oil objective with NA1.4 was used to perform FRAP experiments on an ROI of size of 2 µm and 3 µm for ubi-DE-cad-GFP and sqh-Sqh-GFP, respectively. 488 nm laser was used at 100% with 80 iterations for photobleaching. Images were captured at zoom 2.5 and averaging of 2 with a frame rate of 970 ms/frame. The pinhole was opened to 160 µm. Three frames were captured before the photobleaching followed by time lapse imaging for 1 min. Please refer to the image analysis section for details on the estimation of the mobile fractions.

Sum intensity z projections of all z stacks with membrane signal for Sqh–mCherry and DE-cad–GFP were obtained using Fiji software (http://fiji.sc/wiki/index.php/Fiji) (Rueden et al., 2017) for each time point from interphase to metaphase. ROIs were drawn around six syncytial cells to get total intensity values for each time point using Fiji software. The total intensity at each time point was normalised to the maximum value and plotted as a ‘normalized intensity versus time’ graph.

The DE-cad–GFP signal was estimated from the ubi-DE-cad–GFP transgene, which expresses DE-cad at a higher level as compared to endogenous DE-cad. Hence, we also used endo-DE-cad-GFP to estimate the relative increase of DE-cad across the syncytial cycle. Although the endo-DE-cad-GFP signal was faint in the early embryos, it showed a similar trend of relative increase in DE-cadherin across the syncytial cycle as ubi-E-cad-GFP line. ubi-E-cadherin-GFP showed 1.3±0.2-fold and 2.0±0.4-fold enrichment (mean±s.d., with respect to before polygon transition) on polygon transition and at metaphase, respectively (Fig. 3C). On the other hand, endo-DE-cad–GFP showed 1.6±0.3- and 1.9±0.5-fold enrichment (with respect to before polygon transition) on polygon transition and at metaphase, respectively (n=24 syncytial cells, eight cells per embryo, three embryos) (Fig. S2B,C).

For estimating the change in relative distribution of Myosin II across the syncytial division cycle, we used live imaging of Sqh–mCherry as well as immunostaining with antibody against the Myosin II heavy chain subunit Zipper. We estimated the levels of protein at the membrane as a ratio to the corresponding cytosol and found that Sqh–mCherry and Zipper showed 1.7+0.04-fold and 1.6+0.09-fold enrichment on the membrane, respectively, at interphase. On the other hand, both showed a membrane to cytosol ratio of 1±0.008 at metaphase, suggesting that the Myosin leaves the membrane and becomes cytosolic at this point (Fig. S2D,E).

Circularity and the inverse of this are routine 2D measures to estimate cell shape transition to polygon shapes in epithelia in developmental contexts (Hoffmann et al., 2018; Thomas, 2004; Xue and Sokac, 2016). The organization of the plasma membrane as a circle is visible in the cross sections of the syncytial embryo, and the transition to polygon architecture is also observable with clarity. The circularity estimate allows a quantitative documentation of shape across several cells. The circularity estimate for different polygons seen in syncytial cells in NC13 is included as part of Table S2.

Metaphase furrow lengths were quantified from the orthogonal sections for different time points, and NCs were used for quantifying metaphase furrow lengths using the Zen Blue software. Approximately 6–8 furrows were measured per time point per embryo. These lengths were further confirmed by estimating the number of z stacks spanning the entire furrow length of syncytial cells in the field of view.

A distance versus time graph was plotted followed by linear fitting. The slopes of the vertex distance versus time graph represented the recoil velocities (Liang et al., 2016). The recoil velocities thus obtained were then plotted for comparison between the control and ablated edges before and after polygon transition, and at metaphase. In order to normalize for inherent differences of movement between controls at different stages of the syncytial cycle before and after polygon transition and at metaphase, all the points of the control and ablated edge of a given stage were divided by the average of the control of that stage.

All data are represented as mean±s.d. Statistical significance was determined using an unpaired two-tailed, Student's t-test to compare two means. One-way ANOVA with Dunnett's multiple comparison test was used when comparing three or more means together and comparing all points to a control point. One-way ANOVA with Tukey's multiple comparison test was used when comparing multiple means together and comparing all points with each other. Smoothing of circularity versus furrow length and DE-cad and Sqh intensity analysis curves were performed using a 2nd order, three-neighbor ‘Savistsky–Golay’ smoothing algorithm for better representation.

We thank Tony Harris and Adam C. Martin for providing us with fly lines. We thank Thomas Jeffrey for the Zipper antibody. We thank Girish Deshpande, L. S. Shashidhara, Girish Ratnaparkhi, and R.R. lab members for continuous discussions on the data. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. We also thank FlyBase (Thurmond et al., 2019), the DSHB for monoclonal antibodies, the IISER Pune Microscopy Facility for microscopy and IISER Pune Drosophila facility for help with rearing of flies. We thank ImageJ and Fiji (Rueden et al., 2017; Schindelin et al., 2012; Schneider et al., 2012) and GraphPad Prism for image analysis and data plotting.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.240168.reviewer-comments.pdf