The microtubule end-binding proteins EB1 and Patronin modulate the spatiotemporal dynamics of myosin and pattern pulsed apical constriction

Changes in cell shape, position and number power morphogenetic movements. Apical constriction, a cell shape change characterized by reduction in apical surface area, powers tissue invagination, contraction and single cell delamination. Pioneering studies on Drosophila ventral furrow invagination have uncovered the mechanisms that remodel the actin cytoskeleton and generate cytoskeletal tension during apical constriction. Specifically, the regulation of GPCR signaling by transcription factors activates the small GTPase Rho, which governs actomyosin cytoskeleton dynamics (Kerridge et al., 2016; Kölsch et al., 2007; Martin et al., 2009; Sweeton et al., 1991). The differential regulation of RhoGTPase activity in the medial and circumapical actomyosin pools and biomechanical feedback pattern pulsed constrictions (Mason et al., 2013, 2016; Munjal et al., 2015). What governs the radially polarized distributions of the RhoGTPase activators and effectors in this context remains unclear.

Apical constriction also powers amnioserosa contraction during dorsal closure. Unlike ventral furrow invagination, in which ratchetted constrictions drive irreversible shape change, apical constriction in the amnioserosa is characterized by an early phase (phase I) during which un-ratchetted constrictions produce high-amplitude pulses of constriction and relaxation that generate little or no net reduction in apical area. Dampened pulsations and a net reduction in area accompany phase II. Motile, undirected, apicomedial actomyosin complexes exhibit cycles of nucleation, coalescence, movement and dispersal in phase I. Circumapical myosin enrichment, and non-motile and persistent apicomedial actomyosin networks accompany phase II (Blanchard et al., 2010; Dehapiot et al., 2020; Saravanan et al., 2013; Solon et al., 2009). A small fraction of amnioserosa cells constrict their apices rapidly and extrude/delaminate from the layer. Differential actomyosin contractility in the delaminating cell and its nearest neighbors, and pulse dampening in both populations accompany delamination (Meghana et al., 2011; Saravanan et al., 2013). Although the consequences of RhoGEF2 downregulation in the amnioserosa underscore the involvement of RhoGTPase signaling, the upstream regulators that activate the distinct constriction programs in the amnioserosa remain unclear (Azevedo et al., 2011). We have previously shown that subcellular laser ablation or Rho activation in single amnioserosa cells generates interfacial tension anisotropies that are sufficient to direct the movement of medial actomyosin complexes in the neighbors towards the interfaces with the perturbed cell (Meghana et al., 2011; Saravanan et al., 2013). How these anisotropies are generated and sensed to result in actomyosin complex movement remains poorly understood.

We have also previously uncovered the striking reorganization of the apical microtubule meshwork in amnioserosa cells during native and laser-induced cell delamination. This reorganization – early microtubule loss in the delaminating cell and the alignment and polarization of microtubules in the apical plane in its nearest neighbors – is associated with constriction of the delaminating cell and anisotropic shape transformations in the nearest neighbors. Disrupting microtubule integrity by inducing severing delayed ablation-induced cell delamination. Whether and how apical microtubule reorganization contributes to apical constriction in the amnioserosa remained unclear (Meghana et al., 2011).

The microtubule cytoskeleton performs essential cellular functions, including protein and organelle transport, force generation against cellular membranes, cellular resistance to compressive forces and modulation of cell signaling. These functions rely on spatiotemporally regulated microtubule interactions with distinct microtubule-associated proteins (Akhmanova and Kapitein, 2022; Brouhard and Rice, 2018; Gudimchuk and McIntosh, 2021). In plant and animal cells, microtubules enable cell shape maintenance, and microtubule organization is influenced by cell shape (Bidhendi et al., 2019; Chakrabortty et al., 2018; Gomez et al., 2016; Mirabet et al., 2018). Microtubule network integrity is necessary for morphogenesis that depends on apical constriction (Booth et al., 2014; Fernandes et al., 2014; Kasioulis et al., 2017; Lee et al., 2007) but whether and how the organization and dynamics of the microtubule cytoskeleton aids constriction remains poorly understood.

Here, we combine high-resolution live confocal microscopy with genetic perturbations that interfere with microtubule nucleation and growth to examine the influence of microtubule organization and dynamics on apical constriction in the amnioserosa. Our results uncover the importance of moving Patronin platforms and microtubule growth persistence in an intact microtubule meshwork for patterning cell constriction and for timely tissue contraction. They identify mechanistic links between actomyosin contractility and microtubule cytoskeleton dynamics.

The contraction of the amnioserosa by apical constriction occurs in two phases. High-amplitude pulses of constriction and relaxation with little net reduction in apical area in phase I are followed by dampened pulses and net isotropic reduction in phase II. Each phase is accompanied by the distinct localization and dynamics of apicomedial myosin (Blanchard et al., 2010; Saravanan et al., 2013; Solon et al., 2009). High-resolution, live confocal microscopy revealed that each myosin cycle in phase I (visualized using SqhGFP, which is GFP fused to the regulatory light chain of non-muscle myosin) begins with the nucleation (Fig. 1A1″) of an apicomedial myosin ‘seed’ that grows by coalescence and aggregation to form a larger ‘blob’ (Fig. 1A2″). This blob moves in the apical plane for a distance and constricts the apical membrane locally (Fig. 1A3″-A6″) before it dissolves and disappears (Fig. 1A7″). After a lag phase of ∼1 min minute, a new cycle commences. Coalescence and movement occur simultaneously in some cycles, and ∼40% (46/117) of myosin blobs do not exhibit substantial subcellular movement (Fig. S4). An average myosin cycle takes ∼2 min (115±17 s; n=40). The formation-dissolution cycle of myosin temporally correlates with the constriction and expansion phases of a single pulse, with the lowest apical cell surface area associated with the appearance of the most intense and tightly coalesced myosin blob. Phase II is associated with circumapical myosin enrichment and with non-motile and persistent apicomedial actomyosin networks (see also Blanchard et al., 2010; Dehapiot et al., 2020; Saravanan et al., 2013; Solon et al., 2009). What dictates the spatiotemporal dynamics of the myosin cycle and, in particular, what directs its movement remained unclear. This motivated us to examine the influence of the microtubule cytoskeleton dynamics on amnioserosa cell constriction and tissue contraction.

Tubulin antibodies labeled three microtubule pools in amnioserosa cells – an apical meshwork, lateral arrays and a basal mat – as reported in other epithelia (Bacallao et al., 1989; Bartolini and Gundersen, 2006; Toya et al., 2016). Unaligned, randomly oriented microtubules that showed no characteristic association with centrosomes formed the apical meshwork (Fig. S1A1-B3). The lateral array consisted of microtubules aligned parallel to and along the apicobasal axis. A randomly oriented array/carpet was found on the basal surface (Fig. S1A1-A4′).

Real-time analysis of microtubule organization using JupiterGFP, a GFP fusion protein that labels microtubules along its entire length (Karpova et al., 2006), revealed the dynamic reorganization of the apical meshwork in constricting cells. Buckling and length changes were observed in what appeared to be single microtubules. Microtubule collectives exhibited two configurations that correlated with the pulse cycle. Transient ring/aster/cage-like configurations were formed by a subset of short microtubules (Fig. S2A1-A6, Movie 1). Longer microtubules aligned perpendicularly to the axis of constriction and exhibited cycles of ‘bundling’, when they came in close proximity, and ‘splaying’, during which they dispersed and adopted random orientations (Fig. S2B1-C5 and Movie 1). Live imaging with JupiterGFP and ECadherinGFP revealed that both ring/aster/cage formation and ‘bundling’ were temporally correlated with the constriction phase of each pulse (Fig. S2A1-A6,C1-C5). Importantly, no visible changes in the mass of the apical microtubule meshwork accompanied constriction pulses. Additionally, the rings/asters/cages formed around apicomedial myosin complexes visualized using Sqh-mCherry (Fig. S2D1-D7″ and Movie 2).

To determine whether the observed microtubule reorganizations were polarized, and to establish the nature of their association with apicomedial myosin, we examined the dynamics of the microtubule end-binding proteins EB1 and Patronin that, respectively, facilitate and suppress dynamics at the microtubule plus and minus ends (Goodwin and Vale, 2010; Akhmanova and Steinmetz, 2015), using genomic EB1GFP and Ubiquitin promoter-driven PatroninGFP transgenes (Bulgakova et al., 2013; Wang et al., 2013). A diffuse cytosolic Patronin cloud/halo, which was visible when the cell was relaxed, condensed or closed in around a myosin seed to form a ring/cage as the seed grew and coalesced (in 86% of motile blobs, n=30 from 13 embryos). The Patronin foci then redistributed to form a crescent that trailed the myosin blob some distance away from it as it moved (in 70% of motile blobs, n=30 from 13 embryos). As the blob fragmented and the cell relaxed, Patronin foci dispersed to form a cloud again (Fig. 1A1-A7″,C1-C5, Fig.S3A1-A8″ and Movies 3 and 5). Patronin clouds also condensed around stationary myosin blobs (in 95% of stationary blobs, n=20 from 13 embryos, Fig. S4A1-A7”) but did not redistribute asymmetrically to form tails.

A subset of EB1 comets transiently reorganized around both motile and stationary apicomedial myosin blobs, forming a ring/cage/aster around them as the cell constricted (in 75% motile blobs, n=41 from 13 embryos; in 92% stationary blobs, n=26 from 13 embryos; Fig. 1B1-C5, Figs S3B1-B7″ and S4B1-B7″, Movies 4 and 5). This association, evident around a myosin seed, continued through its growth and coalescence. The comets trailed the myosin blobs as they moved (in 90% motile blobs, n=41 from 13 embryos). New comets became associated with the blobs during coalescence and movement. As the blob dissolved, the comets either retracted away from it or dispersed to become randomly oriented once again (Fig. 1B1-B7″, Fig. S3B1-B7″ and Movies 4 and 5). These results suggest that a subset of short, dynamic microtubules associate with the myosin blob to form what appears to be a predominantly ‘minus-end out’ ring/cage/aster during its coalescence, and then redistribute, to form an asymmetrically positioned tail behind it (Fig. 1D). Technical limitations do not allow us to rule out the possibility that microtubules form anti-parallel arrangements.

To further delineate the relationship between microtubule ends and apicomedial myosin complexes, we determined the speeds of EB1 comets, Patronin foci and myosin complexes. The speed of Patronin foci as they congregated/condensed (0.06 µm/s) was significantly lower than that of EB1 comets (0.27 µm/s) or motile myosin blobs (0.16 µm/s). These observations suggest that whereas the translocative movement of myosin may rely on EB1-dependent microtubule growth (Fig. 9B), the movement of Patronin foci must rely on mechanisms that may depend on actin polymerization or actomyosin contractility (∼0.1-0.3 µm/s, Cameron et al., 2000). They raise the possibility that reorganization of an intact microtubule meshwork by EB1-dependent growth from moving Patronin platforms might influence the organization or dynamics of myosin during pulsed apical constriction, and may also be influenced by it.

We first examined the requirement for microtubule growth on cell and tissue dynamics in the amnioserosa by driving the expression of a truncated version of EB1 (EB1-DN) that is known to function as a dominant negative and has been shown to reduce microtubule growth duration without affecting growth velocity in both mammalian cells and in Drosophila tissues (Komarova et al., 2009; Bulgakova et al., 2013). Amnioserosa cells expressing EB1-DN did not exhibit substantial alterations in microtubule mass or the gross architecture of the apical microtubule meshwork (visualized using JupiterGFP, Fig. S2E1-F7 and Movie 6), allowing us to specifically examine the effects of hindering microtubule growth dynamics.

We used ECadherinGFP to quantify the apical morphodynamics of amnioserosa cells upon EB1-DN expression (see Materials and Methods). A larger fraction of EB1-DN-expressing cells exhibited lower pulse amplitudes during phase I compared with controls (Fig. 2A). Although the mean area of EB1-DN expressing cells was marginally higher than that of control cells during phase I, their mean constriction rate was higher during early closure but comparable towards the end of closure (Fig. 2B,C). A higher proportion of EB1-DN-expressing cells exhibited steeper negative slopes compared with control cells (29% compared to <10%; Fig. S6A). These findings suggest that perturbing persistent microtubule growth changes the phase I cell dynamics that resemble dynamics associated with phase II.

The temporal dynamics of amnioserosa contraction were also affected. The area of the amnioserosa during early closure was larger and more variable in EB1-DN embryos compared with controls but was indistinguishable from wild type during late closure. Surprisingly, the rate of contraction during early closure was significantly higher in EB1-DN embryos compared with controls (Fig. 2E-H and Movie 7). These differences mirror the differences in cell area dynamics and suggest that differences in cell constriction dynamics might contribute to the differences in contraction dynamics. In addition, the frequency of cell delamination – a rare and stochastic cell behavior associated with rapid apical constriction and cell extrusion (Meghana et al., 2011; Muliyil et al., 2011) – was higher in EB1-DN embryos compared with controls (Fig. 2D). These results demonstrate that reducing microtubule growth persistence in an intact microtubule meshwork alters amnioserosa cell and tissue dynamics. Whether the observed increase in the constriction rate predisposes cells to delamination either cell autonomously or through its effects on tissue tension, or whether the increase in frequency of delamination contributes to the early increase in contraction rate (Toyama et al., 2008; Muliyil and Narasimha, 2014) remains unresolved.

The close spatial association between apicomedial myosin and the microtubule plus ends, and the influence of microtubule growth persistence on cell constriction prompted us to examine myosin dynamics in EB1-DN-expressing cells. In control embryos, apicomedial myosin exhibited the myosin cycle described previously (Figs 1A1-A7″, 3A1-A7 and Movie 8). In cells expressing EB1-DN, multiple punctae of apicomedial myosin failed to coalesce into a compact blob, and the circumapical pool of myosin was more prominent compared with controls (Fig. 3B1-D1). Although the total persistence time of medial myosin was not affected, myosin path length was significantly lower in EB1-DN cells (see Materials and Methods, Fig. 3E-G and Movie 8). These observations demonstrate the influence of microtubule growth persistence on myosin organization, distribution and movement.

The speed of motile apicomedial myosin complexes (0.16±0.04 µm/s; n=121 tracks from 4 embryos, Movie 8) was closer to the speed of EB1 comets (0.2±0.07 µm/s; n=101 punctae, n=4 embryos, Movie 4) and of microtubule growth (∼0.1 µm/s, Geisterfer et al., 2020) than the speed of transport along microtubules by motor proteins (which is significantly higher; Abraham et al., 2018). One possibility that is consistent with these speeds is that proteins associated with the growing ends of microtubules may regulate actomyosin contractility and myosin movement. The RhoGTPase activator RhoGEF2 has previously been shown to bind microtubule plus ends in an EB1-dependent manner (Rogers et al., 2004; de las Bayonas et al., 2019). In embryos carrying a genomically expressed RhoGEF2GFP fusion protein, three apical pools of RhoGEF2 – a circumapical pool, a punctate cytoplasmic pool and cytoplasmic/medial aggregates/nodes – were observed (Fig. 4A1-C2). The nodes, when visible, were closely associated with the actomyosin blob and with a subset of microtubules that appeared to be aligned towards it (Fig. 4B1-B4, Fig. S5A1-B4). Live-imaging revealed the pulsatile dynamics of the medial pool of RhoGEF2, characterized by the congregation and condensation of RhoGEF2 nodes upon cell constriction and its dispersal upon cell relaxation (Fig. 4A1-A5′, Movie 9), that was reminiscent of the dynamics of EB1 comets (Movies 4 and 5). The measured mean velocity of RhoGEF2 punctae (0.25±0.08 µm/s, n=96 punctae from five embryos) was also remarkably similar to the mean velocity of EB1 comets.

Simultaneous real-time imaging of RhoGEF2GFP and EB1 (using UAS EB1RFP expressed in the amnioserosa) identified moving comets that had both RhoGEF2 and EB1, albeit for a very short duration (4-14 s; Fig. S5C1-E3, Movie 9). These experiments, which are technically challenging due to low signal intensity and widespread cytosolic distribution of RhoGEF2, nonetheless raise the tempting possibility that RhoGEF2 movement and subcellular distribution may rely on its EB1-dependent association with growing microtubule ends. To determine whether the effect of perturbing persistent microtubule growth on myosin dynamics may rely on Rho activation by EB1-RhoGEF2 interactions, we first examined the distribution of RhoGEF2 in EB1-DN-expressing central amnioserosa cells during phase I. In contrast to control sibling embryos, amnioserosa cells expressing EB1-DN showed a significant increase in the circumapical pool of RhoGEF2 compared with the cytoplasmic pool. Although there was also a marginal increase in the cytoplasmic pool of RhoGEF2, apicomedial nodes were not discernible (Fig. 4C1-D). Overexpression of a full-length EB1 transgene (used in Fig. S5C1-E3) did not alter the relative distribution of RhoGEF2 between these pools (Fig. S5F,G) suggesting that perturbing microtubule growth and not merely the modulation of the number of RhoGEF-binding sites is responsible for the observed redistribution.

To determine whether the redistribution of RhoGEF2 affected Rho pathway activity, we examined the dynamics of Rho kinase/Rok (using a Sqh promoter-driven Rok-GFP fusion protein), the key effector of actomyosin contractility downstream of Rho, in EB1-DN-expressing cells. Rok exhibited pulsatile dynamics in wild-type cells that was reminiscent of the behavior of both EB1 comets and RhoGEF2 nodes (Movies 4, 9, 10). Discrete Rok punctae congregated to form a cluster when the cell constricted and dispersed as the cell relaxed. The brightest pixels of Rok-GFP in the cell resided in apicomedial clusters/nodes, with much less diffuse cytoplasmic signal compared with RhoGEF2, whereas the circumapical regions contained sparsely distributed punctae of Rok (Fig. 4E1-F7 and Movie 10). Substantially higher intensities of Rok-GFP were also found in the apicomedial pool compared with the circumapical pool in control embryos in fixed preparations (Fig. 4G1). In EB1-DN expressing cells, apicomedial clusters of Rok appeared more dispersed and less bright, whereas Rok in the circumapical pool was brighter and more continuous (Fig. 4G2 and Movie 10). These results demonstrate that EB1 maintains the apicomedial enrichment of Rok, and are consistent with its effects on the distribution of myosin. They uncover the requirement of RhoGEF2-EB1 interactions in regulating the spatiotemporal activation of RhoGTPase signaling during pulsed apical constriction. However, the measured velocities of congregating apicomedial Rok foci (0.06 µm/s) was lower than that of the moving myosin blob or EB1, suggesting that Rok clusters may not be directly associated with microtubule plus ends.

We envisaged that the local and dynamic nucleation and growth of a subset of microtubules from Patronin foci might enable the local and adaptive regulation of myosin dynamics in the amnioserosa, and modulate constriction and contraction dynamics. We therefore examined the consequences of downregulating Patronin using hypomorphic Patronin mutants (Nashchekin et al., 2016; Feng et al., 2019). Defects in germband retraction and dorsal closure, as well as compromised amnioserosa integrity were observed in approximately one-quarter of homozygous mutants (24%, n=260; Fig. S6C,D), underscoring the importance of Patronin for morphogenetic processes that depend on the amnioserosa.

We generated a recombinant chromosome containing the hypomorphic Patronin^EY05252 mutant allele and ECadherinGFP to allow quantitative morphodynamic analysis of the amnioserosa. As observed in EB1-DN cells, a significantly larger fraction of Patronin mutant amnioserosa cells exhibited lower pulse amplitudes compared with control cells (Fig. 5C). Like EB1-DN-expressing cells, Patronin mutant cells also exhibited mean areas and constriction rates that were marginally higher than that of control cells during phase I, but were comparable with control cells towards the end of closure (Figs 5A,B and 2B,C). However, the constriction patterns of Patronin mutant cells were qualitatively different from both control and EB1-DN-expressing cells. Significantly larger fractions of cells analyzed exhibited either a single steep negative slope across phase I and phase II (54% compared with 29% in EB1 DN and <10% in control cells, Fig. S6A,B) or aberrant patterns characterized by negative, positive and zero slopes (42% compared with <10% in EB1-DN or control cells, Fig. S6A,B). These findings suggest that Patronin-dependent processes pattern constriction pulses and shape dynamics both qualitatively and quantitatively.

The kinematics of amnioserosa contraction was also affected in Patronin mutants. In the 150 min before completion of closure, the rate of contraction of the amnioserosa was lower than in control embryos despite the smaller areas at the beginning (Fig. 5E,F, Movie 11). This was different from the higher rate of contraction during early closure observed in EB1-DN embryos compared with controls (Fig. 2E-H and Movie 7). Also in contrast to EB1-DN-expressing embryos, the frequency of cell delamination was not significantly different in Patronin mutants compared with controls (Fig. 5D). These results demonstrate the influence of Patronin-dependent nucleation or other processes in orchestrating cell and tissue dynamics during dorsal closure.

To determine whether Patronin might enable the local and dynamic nucleation of a subset of microtubules, we examined microtubule (re)organization in the Patronin mutant embryos in real time using JupiterGFP. As observed in EB1-DN cells, neither the integrity of the apical microtubule meshwork nor the microtubule mass was affected in Patronin mutant amnioserosa cells. In contrast to both wild-type and EB1-DN cells, the apical meshwork seemed temporally and spatially more invariant both with respect to the ability of individual microtubules to change their orientations and to their collective reorganization to form higher order configurations, including the cages/asters described above. Two predominant phenotypes were observed in Patronin mutant cells (Fig. 6A-C, Table 1 and Movie 12). In ∼75% of Patronin mutant embryos or cells, apical microtubules did not substantially change their alignment or position in the apicomedial region during a pulse over time scales larger than one constriction cycle in phase I. In a similar fraction of embryos and cells, the apical microtubules appeared to be anchored to the apical plasma membrane, sometimes producing complete or partial spoke wheel patterns, and the microtubule filaments were longer than controls. The majority of amnioserosa cells in an embryo exhibited these phenotypes, which were, however, not seen in all embryos, possibly on account of residual zygotic and maternal Patronin function. These findings hint at roles for Patronin in facilitating the dynamic (re)organization of the apical microtubule meshwork.

We next examined whether microtubule network reorganization influences the dynamics of apicomedial myosin using Patronin mutants carrying the SqhGFP transgene. In contrast to otherwise wild-type embryos, Patronin mutant cells had poorly coalesced apicomedial myosin blobs (Fig. 7A1-B7), the persistence times and path lengths of which were also significantly reduced compared with controls (Fig. 7E-G). In addition, myosin was enriched in the circumapical pool compared with control cells (Fig. 7A1-D1 and Movie 13). These results demonstrate that Patronin-dependent microtubule reorganization contributes to the subcellular organization and dynamics of apicomedial myosin.

Previous studies using microtubule-depolymerizing drugs or microtubule-severing proteins have demonstrated the requirement of microtubule network integrity for apical constriction (Lee et al., 2007; Booth et al., 2014). Colcemid-induced microtubule depolymerization depleted the apical microtubule meshwork, which appeared to be more sensitive to it than were the lateral arrays, in the amnioserosa cells of the majority of embryos examined (∼75%, n=57; Fig. S7A-C,G-H). In the remainder (∼20%, n=57), the apical meshwork appeared to be ‘stabilized’, with some embryos in this class showing recovery to dynamic reorganization (Fig. S7A1-D14,I1-I5). Whether the latter reflects incomplete exposure to colcemid or a different manifestation of its effects (Rozario et al., 2021) remains unclear. One-third (33%, n=57) of colcemid treated embryos, predominantly those with reduced microtubule mass, also exhibited tears in the amnioserosa. In addition, the curvature of the dome of the amnioserosa was affected, and apically expanded or extruding cells were often seen (Fig. S8). These severe defects, not found in EB1-DN or Patronin mutants, highlighted the importance of an intact microtubule meshwork on epithelial integrity and cell viability (also suggested by abnormal nuclear morphologies in colcemid treated amnioserosa cells, Fig. S7E1-I5) but precluded the analysis of cell and tissue dynamics.

Colcemid-treated amnioserosa cells with microtubule meshwork reduction exhibited qualitatively and quantitatively smaller apical cell areas compared with controls (Fig. S7A1-C7,G1-H5) in contrast to the slight increase in area observed in the end binding protein perturbations at similar stages. Also in contrast to the EB1-DN and Patronin mutant amnioserosa cells, apicomedial myosin appeared to be significantly reduced or diffusely distributed, independent of whether microtubules were depleted or stabilized (Fig. 8C1-E7′, Fig. S7E1-I5). Colcemid treatment, like perturbations to EB1 and Patronin, reduced cell pulsatility and myosin path length in cells with apicomedial myosin blobs (Fig. 8A,B,F-G). The absence of patterned pulses prevented a quantitative estimate of pulse amplitude. These findings reveal that microtubule meshwork depletion reduces apical cell area, pulsatility, and the localized enrichment of myosin to both apicomedial complexes and the junctional pool. Notably, in embryos that exhibited the ‘stabilized’ meshwork, the subsequent restoration of dynamic reorganization also restored pulsatility (Fig. S7D1-D14, Movie 14), further strengthening the requirement for meshwork reorganization for pulsed constriction.

Compromising microtubule meshwork integrity by overexpressing the severing protein Spastin also reduced pulse amplitude and myosin path length. In contrast, microtubules were longer and bundled in cells overexpressing Tau, and their pulse amplitude and myosin path length were, on average, similar to and higher, respectively, than in wild-type cells (Fig. S9D-F,H-J). Both Spastin and Tau overexpression resulted in a significant increase in the total persistence time of myosin (Fig. S9A1-C7,G). Although the effects on myosin upon Spastin overexpression were examined using SpastinGFP and SqhGFP, SpastinGFP formed cytosolic foci that were immotile, allowing us to distinguish them from SqhGFP foci, which were mobile but fragmented.

Collectively, our observations demonstrate that the effects of perturbing EB1 and Patronin on constriction and myosin dynamics are as strong as the effects of disrupting microtubule integrity using colcemid or Spastin, and establish the necessity of (re)organization and dynamics in an otherwise intact microtubule meshwork for tuning cell contractility during apical constriction.

Previous studies have highlighted the importance of an intact microtubule cytoskeleton for morphogenesis (Booth et al., 2014; Fernandes et al., 2014; Kasioulis et al., 2017; Lee et al., 2007). The work presented here demonstrates for the first time the importance of reorganization of an intact microtubule network for tissue contraction mediated by apical constriction. This reorganization relies on mobile platforms at both microtubule ends, the movements of which are independently regulated, and patterns cell constriction dynamics by spatiotemporally modulating actomyosin contractility. Our results suggest a model in which the rapid reorganization of the apical microtubule meshwork, facilitated by nucleation and organization ‘on-the-go’ by motile Patronin foci, enables the local regulation of EB1-dependent growth, and through it, the local and adaptive tuning of actomyosin contractility. This modulates force distribution along the radial axis. We propose that microtubule reorganization enables a self-organizing, mechanosensitive feedback loop that buffers the tissue against mechanical stresses. In this loop, the formation and movement of Patronin foci, presumably triggered by actomyosin anisotropies, spatially direct growth from the plus ends. This regulates local contractility and also generates cytoskeletal tension anisotropies. Increased local cytoskeletal tension might then trigger microtubule depolymerization (Fig. 9). Our work also uncovers a previously undescribed mechanism by which the microtubule cytoskeleton can influence morphogenesis driven by apical constriction, tissue contraction and epithelial fusion (Booth et al., 2014; Ko et al., 2019; Jankovics and Brunner, 2006).

Microtubule network integrity is thought to enable cell shape maintenance through its ability to resist compressive forces generated by the actomyosin cytoskeleton (Ingber et al., 2014). Colcemid induced microtubule destabilization compromised amnioserosa integrity and topology, reduced cell viability and apical cell area, and reduced myosin enrichment in both apicomedial and junctional pools, resulting in its diffuse distribution in the apical cytoplasm (Figs S7,S8). In contrast, apical areas of EB1-DN and Patronin mutant cells were initially larger, and myosin was redistributed to the junctional pool, resulting in higher constriction rates. However, all three perturbations reduced pulse amplitude and apicomedial myosin movement. These findings establish the importance of an intact but reorganizable microtubule meshwork for morphogenesis dependent on apical constriction. They also uncover significant, possibly also mechanistic differences between the effects of compromising microtubule integrity and microtubule reorganization, and identify a role for dynamic microtubule reorganization mediated by Patronin and EB1 in modulating the subcellular balance of forces by governing myosin distribution. What mechanisms account for the cell area reduction, and specifically whether contractility may be driven by the diffusely distributed apicomedial myosin observed upon colcemid treatment remains unclear. We suggest that the cytoskeletal balance of forces as well as protection against cell death cell may be mediated by confinement of actomyosin and apoptosis regulators by association with microtubule ends. Their release upon network destabilization possibly triggers a strong and pleotropic stress response (Arnette et al., 2016; Chang et al., 2008; de las Bayonas et al., 2019; Mollinedo and Gajate, 2003; Oropesa-Ávila et al., 2013). Alternatively, the reduction in cell area may be mediated by myosin-independent mechanisms.

Our results suggest a hierarchy in which Patronin operates on an intact microtubule cytoskeleton, upstream of EB1, to render the apical meshwork ‘reorganizable’ and to direct microtubule growth. We propose that the requirement of Patronin for the maintenance of amnioserosa integrity reflects the requirement for a reorganizable network to respond to and buffer mechanical stresses.

The microtubule reorganization we observe differs from recent studies that have demonstrated an interplay between the two cytoskeletal systems in two ways: first, microtubule reorganization occurs in the apical plane rather than along the apicobasal axis; second, the consequences, of the perturbations tested, on actomyosin and microtubule cytoskeleton are also different (Booth et al., 2014; Ko et al., 2019; Gillard et al., 2021).

Our results establish the role of microtubule ends as mobile platforms through which microtubules can be organized ‘on-the-go’ to spatially regulate signaling and force generation during pulsed apical constriction (Akhmanova and Steinmetz, 2015; Rogers et al., 2004; Verma and Maresca, 2019). They identify the importance of RhoGEF2-EB1 interactions in the spatial regulation of Rho activity and actomyosin contractility (Fig. 9). Microtubule plus ends have been previously shown to modulate actomyosin contractility through their association with RhoGEF2 and other Rho regulators (Verma and Maresca, 2019; Ito et al., 2017; Rafiq et al., 2019). RhoGEF2 and EB1 also colocalize in the ectoderm during Drosophila germ band extension (de las Bayonas et al., 2019). The comparable velocities and colocalization of RhoGEF2 and EB1 we observe in real time raises the possibility that RhoGEF2 tip-tracks. The transience of their association and the differences between the velocities of RhoGEF2 and Rok also suggests that RhoGEF2 might be released from microtubule ends at the apical cortex. Growing microtubule ends have been shown to explore the cytoplasm in a ‘search and capture’ mode, and are postulated to act as platforms that can concentrate proteins in the right conformation to enable more efficient protein-protein interactions than those occurring through diffusion in the cytoplasm. If the association of these proteins is itself coupled to microtubule dynamics, a rapid regulation of protein interactions in a spatially directed manner can be achieved (Galjart, 2010). It is tempting to speculate that EB1-RhoGEF2 interactions provide scaffolds on which RhoGTPase can be activated efficiently. An indication of this is the size and concentration of the apicomedial nodes of the Rho effector Rok. Whether and how the association of RhoGEF2 with the microtubule plus ends is modulated, and how its association with plus ends modifies Rho activity remain to be discovered. Microtubule disruption by nocodozale has been previously reported to aid release of plus end-bound RhoGEF-H1 and lead to the activation of both RhoA and RhoB to influence stress fiber formation in HeLa cells (Arnette et al., 2016; Chang et al., 2008). Our results reveal that perturbing microtubule growth can also modulate RhoGTPase activity (Fig. 9).

Our results identified an essential role for the minus end-binding protein Patronin in apical microtubule meshwork reorganization (Figs 6 and 9). Patronin functions as a non-centrosomal microtubule organizer that can nucleate microtubules or contribute to the formation of polarized microtubule arrays (Akhmanova and Steinmetz, 2015; Feng et al., 2019; Nashchekin et al., 2016; Takeda et al., 2018; Toya et al., 2016; Wang et al., 2013). It is presently unclear whether the ‘anchoring’ phenotypes we observe upon Patronin downregulation reflect the compensatory use of alternate organizing centers (Shtutman et al., 2008; Dong et al., 2017; Nagae et al., 2013) or whether they represent direct effects of the loss of Patronin from the microtubule minus end.

Our findings uncover the dependence of myosin distribution along the radial axis by microtubule growth and reorganization (Fig. 9A). The reversed radial polarity of Rho kinase (redistribution from the apicomedial pool to the circumapical pool) in EB1-DN cells could rely on the increased availability of RhoGEF2 for cortical localization and/or its inability to be concentrated to ‘nodes’ due to poor growth persistence. Alternatively, the two pools of active Rho could rely on different activators that are also differentially sensitive to microtubule dynamics (de las Bayonas et al., 2019).

Our experiments revealed two modes of meshwork reorganization during pulsed constriction: long, aligned microtubule bundles perpendicular to axis of maximum constriction, and rings/asters/cages and tails formed by ‘polarized’ short microtubules. We propose that the former provides a resistive network, analogous to the apical network described in the fly wing (Singh et al., 2018) and to the recently described persistent actin network in amnioserosa cells (Dehapiot et al., 2020), whereas the latter provides an adaptive mechanism that facilitates the local fine tuning of cell shape, dynamics and mechanics. In vitro reconstitution experiments and simulations have revealed that asters composed of short microtubules and nematic bundles composed of long extensible microtubules are observed in distinct phase spaces that are defined by ratios of microtubules to motor number or speed (Roostalu et al., 2018). The mechanisms that underlie the generation of the two microtubule configurations we observe in vivo, and the consequences of perturbing them individually will be interesting to investigate. Our results also raise the possibility that a symmetry-breaking ‘cage to tail’ or ‘ring to crescent’ transition might underlie movement of myosin blobs (Fig. 9B). Indeed, the vast majority of (but not all) motile blobs exhibited this transition and stationary blobs were invariably surrounded by microtubule rings/cages that persisted through their nucleation and growth but were not associated with tails. How this transition is regulated will be interesting to explore. It will also be interesting to determine whether the reorganization of the apical microtubule meshwork is regulated temporally and by post-translational modifications of tubulin. Indeed, microtubule acetylation increases during late closure (phase II, Fig. S1C), which is also associated with less motile and more persistent apicomedial actomyosin complexes.

Microtubules are capable of self-organization to form higher order network conformations in response to even weakly polarizing external and internal cues by mechanical anisotropy (Bidhendi et al., 2019; Gomez et al., 2016; Meghana et al., 2011; Mirabet et al., 2018). Our analysis suggests that actomyosin aggregates may break mechanical symmetry and direct microtubule nucleation and growth through the formation of Patronin foci. The work of Martin and colleagues (Ko et al., 2019) also revealed that the formation of apicomedial Patronin foci in ventral furrow cells depends on actomyosin contractility. Our results raise the possibility that the subsequent dispersal of EB1 and Patronin foci may also be influenced by the activation of contractility. Indeed, microtubule catastrophe occurs upon contact with the cortex (Akhmanova and Steinmetz, 2008). Together, these findings suggest that microtubules may (re)act, through reorganization, as first responders to anisotropies in tension. Whether and how microtubule growth and shrinkage are regulated by actomyosin contractility remains to be determined.

The following transgenic lines were obtained from the Bloomington Drosophila Stock Center: c381Gal4 (AS Gal4, for driving expression in the amnioserosa), JupiterGFP (to mark microtubules; Buszczak et al., 2007), Ubi::PatroninGFP (to mark the minus ends of microtubules; Wang et al., 2013), Sqh-mCherry and SqhGFP (to mark non muscle myosin II), sqh::RokGFP (Rho kinase tagged to GFP; Abreu-Blanco et al., 2014), RhoGEF2GFP (Sarov et al., 2016) and sqh::RhoGEF2GFP (RhoGEF2 tagged to GFP; Nakamura et al., 2017). The following transgenic lines were gifts from Nick Brown, University of Cambridge, UK: EB1 GFP (to mark microtubule plus ends) and UAS EB1-DNmCherry (to perturb the persistent growth of microtubules) (Bulgakova et al., 2013). UAS-EB1RFP (Mattie et al., 2010) was a kind gift from Melissa Rolls, Pennsylvania State University, USA. Ubi::ECadhGFP (to mark adherens junctions) was a kind gift from Tadashi Uemura, Kyoto University, Japan (Uemura et al., 1996). Patronin^EY05252/CyO (from the Bloomington Drosophila Stock Center) was used to downregulate the amount of Patronin, UAS Tau (from the Bloomington Drosophila Stock Center) was used to cause microtubule bundling and increase stability (Chatterjee et al., 2009), and UAS Spastin and UAS SpastinGFP were used to sever microtubules [Jankovics and Brunner, 2006; Sherwood et al., 2004; kind gifts from Kai Zinn (California Institute of Technology, Pasadena, CA, USA) and Damian Brunner (University of Zurich, Switzerland). Ubi::Spd2GFP was a kind gift from Jordan Raff, University of Oxford, UK (Dix and Raff, 2007). See Table S1 for list of stocks and Table S3 for a full list of genotypes analyzed.

For genetic perturbations and their controls, flies were allowed to lay for 4 h, and embryos were aged at 29°C for 8 h to enrich for stages of dorsal closure. For the analysis of cell and cytoskeletal dynamics in phase I, embryos were chosen that had completed germband retraction but had not yet initiated the formation of the anterior canthus. For the analysis of amnioserosa contraction and cell delamination, embryos were imaged until the completion of closure from the end of germband retraction, and tissue dynamics were analyzed in the 210-120 min before the completion of closure for amnioserosa contraction or from the formation of the posterior canthus until the end of closure for cell delamination.

For live imaging, embryos were dechorionated using 50% bleach for 2 min and washed with water. Embryos were then put on a 0.17 mm coverslip (Corning) on a thin film of Halocarbon oil 700 (Sigma) and imaged on an inverted microscope (Olympus FluoView1000 confocal microscope) on a 60×, 1.4NA objective with a digital zoom of 2.5 when only a few cells were being captured in the frame. Optical sections 0.5-1 µm apart were acquired at 1-10 frames/min depending on whether tissue, cell or subcellular dynamics was being imaged. Maximum intensity projections were made using the Image5D plug-in on ImageJ and assembled as a time series on ImageJ. For dual color imaging of GFP and Cherry, simultaneous excitation with 488 nm and 561 nm lasers was used. Images were prepared in Adobe Photoshop and figures were assembled using Adobe Illustrator. See Table S1 for software and algorithm details.

For fixed preparations, embryos were harvested and stained using standard protocols (Narasimha and Brown, 2006). The following primary antibodies were used: anti-GFP (rabbit 1:1000, Invitrogen), anti-EB1 (rat 1:200, Abcam), anti-pMLC Ser19 (rabbit 1:50, Cell Signaling Technology), anti-RFP (mouse 1:1000, Abcam), anti-ECadherin (rat 1:10; DSHB), anti-α tubulin (mouse 1:1000, Abcam) and anti-acetylated tubulin (mouse 1:500, Sigma). Alexa Fluor-conjugated secondary antibodies (Invitrogen) were used at 1:200 dilutions and Phalloidin Atto 647 (Sigma) was used at 1:100 dilution. DAPI was used to label nuclei. Embryos were then stored and mounted in Vectashield, and z stacks of confocal images with optical slices at 0.3 µm intervals were taken on an Olympus FluoView 1000 confocal microscope using a 60×1.4 NA objective. Maximum intensity projections were made using the Image5D plug-in on ImageJ. For dual or triple color imaging, sequential excitation with 488 nm, 561 and 633 nm lasers was used. Images were prepared in Adobe Photoshop and figures were assembled on Adobe Illustrator (Adobe Systems). See Table S1 for antibody and dye details.

Embryos were harvested from laying cages and dechorionated as described above. The dechorionated embryos were then transferred into a 100 µm cell strainer (BD Falcon, 252360) and treated with 1:10 Citrasolv in water for 10 min, with constant shaking for permeabilization, as previously described (Rand et al., 2010; Chung et al., 2017). The Citrasolv was washed off with three successive washes in 1× phosphate-buffered saline (PBS) and two washes in PBS with 0.05% Tween-20 (PBSTw). The permeabilized embryos were then immersed in colcemid (10 µg/ml; Roche) for 4 h in a capped microcentrifuge tube on a rotator. Subsequently, colcemid was removed by quickly dabbing the tube onto tissue paper. Embryos that were at the beginning of dorsal closure after colcemid treatment were handpicked using a paintbrush and put on imaging oil for live imaging. Only one early-stage dorsal closure embryo (soon after the end of germ band retraction) per set of treatments was imaged in each imaging session. For immunostaining, all treated embryos were put into fixative and only embryos at the beginning of dorsal closure were imaged. This ensured uniformity and reduced any variability that might arise due to differences in duration of treatment or the time interval between colcemid washout and imaging. The success of microtubule depolymerization was inferred from the reduced microtubule meshwork visualized by JupiterGFP (for live imaging) or α-tubulin (in fixed preparations). A small fraction (20%) of colcemid treated embryos did not exhibit a reduction in microtubule mass but instead showed microtubule stabilization (see Fig. S7D). Embryos that were only dechorionated (unpermeabilized or untreated) or were additionally permeabilized using CitraSolv and washed with PBS and PBSTw (permeabilized, untreated) were used as controls. See Table S1 for reagent details.

No manipulations other than level adjustments to stretch the intensity histogram (when necessary) were applied. Background subtraction was carried out on dual color movies in the images in Fig. 1A-B7″, Fig. S3B using ImageJ only to better visualize the movement of EB1 comets/Patronin foci. See below for details.

Apical areas of individual cells in phase I of dorsal closure were measured for a duration of 140 min before the closure, from maximum intensity projected, time-lapse images of ECadherinGFP. Central amnioserosa cells were segmented using Tissue Analyzer (Aigouy et al., 2016), the selected ROIs exported to Fiji (Schindelin et al., 2012) and the areas extracted. Cell area dynamics were analyzed by plotting the changes in apical area as a function of time (Microsoft Excel). Mean rates of change of area (±s.d.) and the linear regression was carried out using Origin Pro 2015. Similarly obtained cell area traces were also used to measure the amplitudes of pulses (Saravanan et al., 2013). Normalized pulse amplitude was calculated as (A_max/A₀)−(A_min/A₀), where A_max and A_min represent, respectively, the maximum and minimum apical cell areas during a pulse, and A₀ denotes the initial cell area. Pulse amplitude was calculated from pulses observed in cells in phase I over a 30 min interval and averaged over multiple cells from multiple embryos of the same genotype to obtain mean and spread.

The area of the amnioserosa was measured as the area of the ellipse bound by the leading edge/actin cable from 120 min before the completion of closure to the end of closure, from maximum intensity projected, time-lapse images of ECadherinGFP. The areas were extracted and plotted as mentioned above. The number of cell delamination events occurring in the time window from the formation of the posterior canthus formation to the end of dorsal closure were marked and counted using Fiji. The significance of differences between means of different genotypes was determined by performing tests mentioned in Table S2, using GraphPad (Prism Software). See Table S1 for software and algorithm details.

Apical cell areas extracted as described above were classified into four types based on constriction patterns. Type I traces exhibited a short duration (∼30 min) of pulsed constriction with no net reduction in area (phase I), followed by an increased rate of reduction (in the transition from phase I to phase II) and subsequently a steep reduction in area (phase II). Type II traces showed an extended period of pulsed constriction (∼75 min) with very little or no reduction in area (phase I), followed by a steep reduction in the area (phase II). Type III traces were characterized by a steep reduction in area with a single slope from the beginning (140 min before closure). Type IV traces exhibited very little net reduction in area during the latter part of the trace (phase II). For Patronin^EY05252 mutant embryos, cells were analyzed from the embryos that did not exhibit defects in germ band retraction or in the integrity of the amnioserosa, and also completed dorsal closure.

Cells in phase I were chosen as described above. Formation time of the myosin blob was defined as the time taken for the attainment of its largest size from the point of its initial detection. Dissolution time was defined as the time taken by the largest blob to disappear completely. Total persistence time was obtained by adding the formation and dissolution times. Myosin times from several myosin cycles from multiple central amnioserosa cells in phase I of dorsal closure from different embryos of the same genotype were pooled to obtain the mean myosin cycle times.

Path length was defined as the distance travelled by the coalesced myosin blob, the trajectory of which was tracked manually using an ImageJ plug-in MTrackJ (Meijering et al., 2012), from the time it was first detected until it disappeared. The plug-in also marks the traced tracks to enable the visualization of qualitative differences. The path lengths of several myosin cycles from multiple central amnioserosa cells in phase I from different embryos of the same genotype were pooled to obtain the mean path length for each genotype. The path length of each myosin blob was divided by its respective persistence time to obtain its speed, and the mean speed was determined from the trajectories of multiple myosin blobs. The significance of differences between the means obtained for different genotypes was determined using the Mann–Whitney test on GraphPad (Prism Software). See Table S1 for software and algorithm details.

Several myosin (sqhGFP) cycles from central amnioserosa cells of in phase I of dorsal closure were analyzed. The myosin blobs were first categorized as motile or stationary on the basis of their motility. Stationary blobs formed and dissolved at the same subcellular location, and exhibited very little or no movement. Motile blobs formed and dissolved at different subcellular locations, and exhibited movement in the apical plane. For a motile blob, the organization of microtubule ends was analyzed visually at two stages of the cycle: during coalescence and during its movement. Ring/cage/aster-like organization of the microtubules was scored on the basis of the radial arrangement of microtubules (JupiterGFP, EB1GFP and PatroninGFP) around the myosin blob. In contrast, tails were scored when the microtubule ends were arranged asymmetrically in a crescent around it. For stationary blobs, the organization of microtubule ends was characterized only during its coalescence. The percentage of blobs showing the major class of organization during coalescence, and movement of motile blobs and during stationary blob formation is presented in the main text. Background subtraction was carried out on dual-color movies in the images in Fig. 1A,B, Fig. S3B using ImageJ. For this, median blur of a radius that depended on the magnification was generated and subsequently subtracted from the original movie. This was carried out to enable the bright dots to stand out against the diffuse background signal to better visualize the movement of EB1 comets and/or Patronin foci. The images in Figs S3A and S4 are not background subtracted.

Central, phase I amnioserosa cells imaged on an Olympus FluoviewFV1000 using a 60×, 1.4 NA objective with 3 or 4× digital zoom were used to obtain spatiotemporally well resolved movies of Jupiter GFP. Several pulsed constriction cycles from approximately five or six central amnioserosa cells were visualized in each frame. Depending on the altered morphology and behavior of microtubule filaments in the apical plane in mutant embryos compared with controls, they were visually categorized into four non-exclusive classes.

(1) Increased microtubule length. As described in the text, both long and short microtubule filaments were seen in control cells. When the majority of microtubule filaments of a cell were long and had lengths longer than that of control cells, they were scored as having longer microtubules.

(2) Increased anchoring. Single microtubule filaments exhibit transient anchoring to the apical plasma membrane in control amnioserosa cells, where one end of these microtubule filaments was seen to interact transiently with the apical plasma membrane. Cells in which the majority of the microtubule filaments appeared to be associated with the apical membrane, and oriented perpendicularly to it, often in groups, for periods longer than the length of a constriction pulse were scored as increased anchoring. In its extreme form, this resulted in their ‘spoke wheel’ arrangement.

(3) Reduced microtubule mass. Cells in which the overall apical microtubule mass was visually less than that of control were scored as having fewer microtubules.

(4) Reduced microtubule reorganization. As described in the text, microtubules in control cells in phase 1 reorganized dynamically, changing their orientation in the apical plane during the course of a constriction pulse and also formed higher-order structures that included asters/cages. Cells in which the microtubule filaments did not show the dynamic reorientation of filaments or the formation of asters and cages observed in controls may also exhibit anchoring to the apical plasma membrane. Embryos in which the majority of cells exhibited a specific phenotype were scored in that category.

Movies of sqh::RokGFP, Ubi::PatroninGFP, genomic EB1GFP and sqh::RhoGEF2GFP were used to calculate the speed of Rok punctae, Patronin puctae, EB1 comets and RhoGEF2 comets, respectively. The distance traveled by the Rok punctae and Patronin punctae were calculated using Fiji plug-in MTrackJ by tracking individual puctae during their condensation in the constriction phase of a pulse. The distances traveled by the EB1 comets and RhoGEF2 comets were calculated manually using Fiji by tracking the tip of the comet in one direction until it disappeared or began to retract. The measured distances were divided by the time taken to travel them to obtain the speed. The significance of differences between the mean speeds was determined using statistical tests indicated in Table S2 in GraphPad Prism.

Rok intensity was measured from maximum intensity projections of apical slices of the central amnioserosa cells during phase I of dorsal closure from embryos carrying sqh::RokGFP that had been stained with an anti-GFP antibody. Straight lines were drawn from one interface of the cell to the opposite interface through the medial blob or the diffuse medial pool of Rok. Line intensity profiles were obtained using Fiji (Schindelin et al., 2012). Myosin intensity was measured from maximum intensity projections of apical slices of the central amnioserosa cells during phase I of dorsal closure from embryos carrying SqhGFP. Straight lines were drawn from one interface of the cell to the opposite interface through the medial blob at its most coalesced form. The intensity profile along the normalized length (from 0-1, with 0 and 1 being at the two membranes) was then plotted in Origin Pro.

For estimating RhoGEF2 intensities, apical projections of central amnioserosa cells of embryos expressing RhoGEF2GFP and stained with anti-GFP antibody were taken. For the circumapical pool, segmented lines of 1 µm width were used to trace the apical cell membrane using ECadherin as a marker. For the cytoplasmic pool, the polygon tool was used to make ROIs 1 µm internal to the apical plasma membrane. Both sets of ROIs were traced using Fiji (Schindelin et al., 2012), and used to measure the circumapical and cytoplasmic pool intensities of RhoGEF2, respectively. Intensity was normalized to area and plotted using Origin. No image manipulation was carried out for intensity quantification.

No statistical method was employed to predetermine the sample size. The sample size (n) represents data from multiple experiments (immunofluorescence stainings) and imaging sessions (live imaging), each containing one or many biological replicates (embryos, cells or myosin complexes) depending on the type of analysis performed (detailed in the sections above), as it was not possible to analyze a statistically significant number of embryos from a single imaging session. No data were excluded from the analysis. All measurements (individual data points) are presented in most of the graphs provided. Where a representative graph or image is shown in a figure, additional graphs/images are provided in Fig. S1-S9.

Box plots were made using BoxPlotR (Spitzer et al., 2014) and Origin Pro. The boxes show median±interquartile range. Thick black lines show the median, the plus mark shows the mean, and the box limits indicate 25th and 75th percentiles. Whiskers extend to 1.5 times the interquartile range from the 25th and 75th percentiles. Individual data points are represented by dots. The sample size (n) analyzed for the individual genotypes is mentioned in the graph. All statistical analysis was carried out using GraphPad Prism software.

See Table S2 for details on the statistical tests used to determine significant differences in values between the populations compared, alongside the numbers of the panels in the figures in which these comparisons have been made. See Table S1 for software and algorithm details.

We thank Nick Brown, Damian Brunner, Jordan Raff, Tadashi Uemura, Kai Zinn, Melissa Rolls, the Bloomington Drosophila Stock Center, the Vienna Drosophila Resource Center and the Developmental Studies Hybridoma Bank for reagents, Himanshu Bhat for the images in Fig. S1B, Richa Rikhy for advice on permeabilisation, members of M.N.’s lab for discussion and Sudeepa Nandi for comments on the manuscript. M.N. dedicates this work to the memory of Aditi Simha who inspired her thinking on the collective behavior of polar active filaments.