PI(4,5)P₂ forms dynamic cortical structures and directs actin distribution as well as polarity in Caenorhabditis elegans embryos

Asymmetric division generates cellular diversity and is prevalent throughout development. During intrinsic asymmetric division, cell polarity is first established and then coupled with spindle positioning, resulting in proper cleavage and partitioning of contents to daughter cells. The evolutionarily conserved partitioning-defective (PAR) proteins are crucial for cell polarity and asymmetric division (reviewed by Goldstein and Macara, 2007; Gönczy, 2008; Knoblich, 2010). In contrast to the wealth of knowledge regarding PAR proteins and interacting components, the involvement of lipid plasma membrane components in these processes is less understood, in particular in developing systems.

The early Caenorhabditis elegans embryo has proven instrumental for uncovering mechanisms governing asymmetric division (reviewed by Hoege and Hyman, 2013; Pacquelet, 2017; Rose and Gönczy, 2014). Shortly after fertilization, the embryo surface exhibits uniform contractions of the cortical actomyosin network located underneath the plasma membrane. These contractions are driven by non-muscle myosin 2 (NMY-2), activated by the Rho GTPase RHO-1 and its guanine nucleotide exchange factor (GEF) ECT-2 (Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006). Thereafter, sperm-derived centrosomes induce the local disappearance of cortical ECT-2 in their vicinity, thereby determining the future embryo posterior. This results in local RHO-1 inactivation and cortical flows away from this region, towards the future embryo anterior (Bienkowska and Cowan, 2012; Cowan and Hyman, 2004; Motegi and Sugimoto, 2006). These flows promote PAR polarity establishment, whereby PAR-3, PAR-6 and atypical protein kinase C (PCK-3) are segregated to the anterior side, whereas PAR-1, PAR-2 and lethal giant larvae-like 1 (LGL-1) occupy the expanding posterior cortical domain (reviewed by Hoege and Hyman, 2013; Pacquelet, 2017; Rose and Gönczy, 2014). The Rho GTPase CDC-42 is also segregated to the anterior, where it stabilizes the actomyosin network and promotes PAR-6 cortical association (Kumfer et al., 2010; Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006).

PAR polarity can also be established in C. elegans zygotes through a partially redundant pathway, whereby microtubules nucleated from centrosomes protect PAR-2 from PKC-3-mediated phosphorylation, allowing PAR-2 association with phospholipids at the embryo posterior (Motegi et al., 2011; Zonies et al., 2010). In other systems, homologs of PAR proteins and interacting components also associate with phospholipids, as exemplified by Drosophila DmPar3 binding to phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂, hereafter PIP₂] and phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃, hereafter PIP₃] (Krahn et al., 2010). Furthermore, human Cdc42 binds to PIP₂ (Johnson et al., 2012). Overall, phospholipids can bind PARs and interacting proteins in some contexts, but their subcellular distribution and potential function in an asymmetrically dividing system, such as the C. elegans zygote, remain unclear.

PIP₂ is one of the most abundant of seven phosphorylated phosphatidylinositols and is present mostly in the inner leaflet of the plasma membrane (reviewed by De Craene et al., 2017). PIP₂ is mainly generated from phosphatidylinositol 4-phosphate (PI4P) by Type I PI(4)P5-kinases (PIP5K1), and can be further phosphorylated to PIP₃ by phosphatidylinositol 3-kinases (PI3K). Conversely, 5-phosphatases, including OCRL and synaptojanin, can dephosphorylate PIP₂. In systems from Saccharomyces cerevisiae to Homo sapiens, PIP₂ helps link the F-actin cortical network to the plasma membrane and promotes F-actin assembly and reorganization by stimulating, together with Cdc42, WASP family proteins that activate the actin nucleator Arp2/3 (reviewed by Brown, 2015; De Craene et al., 2017; Di Paolo and De Camilli, 2006; McLaughlin et al., 2002; Wu et al., 2014; Yin and Janmey, 2003; Zhang et al., 2012). Whether PIP₂ regulates cortical actomyosin network organization in the C. elegans embryo is not known.

The subcellular distribution of PIP₂ in the C. elegans zygote is also unclear. In other systems, PIP₂ can distribute unevenly in the plasma membrane, for instance accumulating in macrodomains in nascent phagosomes and membrane ruffles or at the leading edge of motile cells, which all exhibit curved membranes that are sites of actin reorganization (Chierico et al., 2015; McLaughlin et al., 2002; Zhang et al., 2012). Accordingly, PIP₂ can stimulate actin polymerization in curved but not flat model membranes (Gallop et al., 2013). Interestingly, PIP₂ patches assemble at the leading edge of neuronal PC12 cells before F-actin accumulates, but their formation also depends on F-actin (Golub and Caroni, 2005; Golub and Pico, 2005). Moreover, F-actin enrichment is thought to drive clustering of PIP₂-containing macrodomains, which in turn further regulates actin polymerization and branching (reviewed by Chichili and Rodgers, 2009). Overall, PIP₂ and F-actin polymerization exhibit reciprocal positive feedback regulation in several systems.

The single C. elegans PIP5K1 is PPK-1, which can synthesize PIP₂ from PI4P in vitro and in vivo (Weinkove et al., 2008). Overexpression of PPK-1 in developing neurons increases the level of PIP₂ and results in extended filopodial-like structures, probably through changes in the actin cytoskeleton (Weinkove et al., 2008). In the somatic gonad, PPK-1 is important for F-actin organization and gonad contractility (Xu et al., 2007). Moreover, PPK-1 is enriched on the posterior cortex of one-cell embryos and has been implicated in the regulation of spindle positioning, but not of cell polarity (Panbianco et al., 2008). However, in other systems, PIP₂ can modulate polarity by recruiting PAR proteins and reorganizing the actin cytoskeleton. Thus, in the Drosophila follicular epithelium, PIP₂ recruits DmPar3 to the apical plasma membrane to maintain apical–basal polarity (Claret et al., 2014). Moreover, PIP₂ might mediate interactions between PAR proteins, the actomyosin network and the plasma membrane in the fly oocyte (Gervais et al., 2008), as well as regulate apical constriction in the fly embryo (Guglielmi et al., 2015). Motile cells, such as mammalian neutrophils or Dictyostelium discoideum, also rely on PIP₂, together with PIP₃, for actin network reorganization during migration (reviewed by Wu et al., 2014). To summarize, in many systems, PIP₂ is essential for proper F-actin organization and cell polarization, but whether this applies to C. elegans embryos is unknown. In this study we investigated the cortical distribution and function of the plasma membrane lipid component PIP₂ in the worm zygote.

While monitoring the distribution of components involved in asymmetric division of the C. elegans zygote with confocal spinning disk microscopy, we discovered that the PIP₂ biomarker GFP::PH^PLC1δ1 (Audhya et al., 2005) formed distinct and dynamic structures at the cell cortex (Fig. 1A-J; Fig. S1A; Movie 1). The pleckstrin homology (PH) domain of mammalian phospholipase C1δ1 (PLC1δ1) binds to PIP₂ in vitro and in cells with high specificity and affinity (Garcia et al., 1995; Lemmon et al., 1995; Várnai and Balla, 1998). Moreover, in vertebrate cells, the lateral diffusion of GFP-PH^PLC1δ1 resembles that of PIP₂ (Golebiewska et al., 2008; Hammond et al., 2009), and membrane domains formed by fluorescently labeled PIP₂ overlap with those monitored by GFP-PH^PLC1δ1 (Chierico et al., 2015). Therefore, GFP-PH^PLC1δ1 is well suited to monitor PIP₂ dynamics in live cells.

Initially, when the cortical actomyosin network was contractile throughout the early C. elegans embryo, PIP₂ was present weakly and evenly on the cell cortex (data not shown). Thereafter, when the actomyosin network began to move towards the anterior at the onset of polarity establishment, striking elongated cortical structures enriched in PIP₂ became apparent, primarily on the anterior side of the embryo (Fig. 1A,B,K; Fig. S1A, top). Such PIP₂ cortical structures had an initial average area of ∼2.5 µm² and elongated as the zygote progressed through the cell cycle (Fig. 1L,M). All elongated PIP₂ cortical structures moved anteriorly during polarity establishment, distributing in a clearly polarized manner at pseudocleavage, when they covered ∼15% of the anterior cortical surface (Fig. 1C,D, arrow, K). During the subsequent centration/rotation stage, PIP₂ cortical structures decreased in size (Fig. 1E,F, arrowhead), before disappearing almost completely, with some remaining small foci by the time of nuclear envelope breakdown (NEDB) (Fig. 1G,H, arrowheads). A few elongated cortical structures reappeared during cytokinesis, primarily in the anterior (Fig. 1I,J). Unlike the structures visible when imaging the cortical plane, discrete PIP₂ entities were barely detectable as slight thickenings of the plasma membrane in the embryo middle plane (Fig. S1A, arrowhead), probably explaining why they were not noted previously (Audhya et al., 2005; Blanchoud et al., 2010; Panbianco et al., 2008).

We aimed to verify the cortical distribution revealed by GFP::PH^PLC1δ1 using fluorescently labeled PIP₂, delivering Bodipy-FL-PIP₂ to embryos the eggshells of which had been permeabilized using perm-1(RNAi) (Carvalho et al., 2011) (Fig. S1B-D). We found that Bodipy-FL-PIP₂ provided to otherwise wild-type embryos distributed in dynamic polarized cortical structures akin to those observed with mCherry::PH^PLC1δ1 (Fig. S1B,C; Movie 2). Moreover, delivering Bodipy-FL-PIP₂ to embryos expressing mCherry::PH^PLC1δ1 established that the two components overlapped (Fig. S1D). Overall, we concluded that GFP::PH^PLC1δ1 is a faithful marker of PIP₂ in one-cell C. elegans embryos, where it is present in dynamic and polarized structures at the plasma membrane.

We aimed to determine what regulates the polarized distribution of PIP₂ cortical structures, which is particularly apparent during pseudocleavage. We found that PIP₂ cortical structures did not overlap with GFP::PAR-2, which marks the posterior cortical domain (Fig. 2A), but did overlap with GFP::PAR-6, which marks the anterior polarity domain (Fig. 2B; Movie 3). Moreover, we observed that PIP₂ cortical structures overlapped with elongated GFP::PAR-6 cortical structures (Fig. 2B, arrow), but not with GFP::PAR-6 foci (Fig. 2B, arrowhead), which are two distinct cortical populations of GFP::PAR-6 (Beers and Kemphues, 2006; Robin et al., 2014; Rodriguez et al., 2017; Wang et al., 2017).

We next tested whether the polarized distribution of PIP₂ cortical structures depended on A–P polarity cues. Compared with the control condition, we found that, upon par-3(RNAi), PIP₂ cortical structures distributed more uniformly (compare Fig. 2C,D with 2E,F, and Fig. S2A,B with S2C,D), except for the very posterior of the embryo (Fig. S2C; Fig. 2E, Pseudocleavage), consistent with the known slight posterior clearing of the actomyosin network upon PAR-3 inactivation (Kirby et al., 1990; Munro et al., 2004). Moreover, we found that, upon par-2(RNAi), PIP₂ cortical structures first moved anteriorly (Fig. 2G,H, pseudocleavage, Fig. S2E), but then distributed in a more uniform manner (Fig. 2G,H, Mitosis, Fig. S2F), in line with PAR-2 being dispensable for polarity establishment, but essential for polarity maintenance (Cuenca et al., 2003; Hao et al., 2006; Munro et al., 2004). Furthermore, depletion of PAR-2 or PAR-3 in embryos expressing mNG::PH^PLC1δ1 and Lifeact::mKate-2 established that the boundary of the two domains along the A–P axis was similar (Fig. S2A-I, Movie 4). This probably also explains why PIP₂ cortical structures do not occupy mutually exclusive territories in par-3(RNAi) and par-2(RNAi) embryos, because their distribution coincided with the essentially uniform presence of cortical F-actin. Together, these findings established that the asymmetric distribution of PIP₂ cortical structures is regulated by PAR-dependent A–P polarity cues.

Given that PIP₂ and F-actin are interdependent in many systems, we tested whether these two components overlapped at the cell cortex of C. elegans embryos. As shown in Fig. 3A and Movie 5, we found a partial overlap of Lifeact::mKate-2, which monitors F-actin (Reymann et al., 2016), and of PIP₂ cortical structures marked by mNeonGreen::PH^PLC1δ1 (mNG::PH^PLC1δ1). By contrast, we detected no substantial overlap between mCherry::PH^PLC1δ1 and GFP::NMY-2 (Fig. 3B; Movie 6). Strikingly, we also found that PIP₂ cortical structures marked by mCherry::PH^PLC1δ1 fully colocalized with GFP::ECT-2, GFP::RHO-1 and GFP::CDC-42 (Fig. 3C-E; Movie 7).

We tested whether RHO-1 or CDC-42 are needed for PIP₂ cortical structures using partial RNAi-mediated depletion, since full depletion leads to sterility (Schonegg and Hyman, 2006). Upon rho-1(RNAi), we found that PIP₂ structures formed normally but distributed symmetrically, consistent with the requirement of RHO-1 for A–P polarity (Fig. S3A,C,J,K,M,N). Thereafter, PIP₂ cortical structures became polarized later than normal (Fig. S3B,D,G,H), probably through the PAR-2-dependent pathway (Motegi et al., 2011; Zonies et al., 2010). PIP₂ cortical structures also formed upon cdc-42(RNAi), with minor and variable shape and size alterations compared with the control (Fig. S3E,F,L,O). The PIP₂ structures first segregated towards the anterior in cdc-42(RNAi) embryos, but then extended posteriorly (Fig. S3E,F,I), consistent with CDC-42 being required for polarity maintenance (Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006).

Overall, we concluded that RHO-1 and CDC-42 are dispensable for PIP₂ cortical structure formation and, importantly, that such structures colocalize partially with F-actin, as well as completely with ECT-2, RHO-1 and CDC-42.

Live imaging of control, par-3(RNAi) and par-2(RNAi) embryos expressing Lifeact::mKate-2 and mNG::PH^PLC1δ1 suggested that movements of PIP₂ cortical structures and the F-actin network are coupled (Fig. S2A-I, Movie 5). To investigate this potential coupling quantitatively, we used particle image velocimetry (PIV) to simultaneously analyze cortical flows of Lifeact::mKate-2 and mNG::PH^PLC1δ1 during polarity establishment (Fig. 4A-C) (Thielicke and Stamhuis, 2014). This analysis revealed highly correlated local flow velocities and directions at each time point (Fig. 4B,C; Fig. S4A,B). Analogous findings were made when comparing mCherry::PH^PLC1δ1 and GFP::CDC-42 (Fig. S4C). By contrast, no strong correlation was found between mCherry::PH^PLC1δ1 and the caveolin marker CAV-1::GFP (Fig. S4C), which might mark lipid rafts (Kurzchalia and Ward, 2003; Kurzchalia et al., 1999; Merris et al., 2003). Together, these results demonstrated that movements of PIP₂ cortical structures and the F-actin network are coupled.

We next addressed whether the movements of PIP₂ cortical structures and of F-actin are either synchronous or exhibit a time shift, which could suggest that one component leads the other. Close examination of embryos expressing Lifeact::mKate-2 and mNG::PH^PLC1δ1 suggested that cortical PIP₂ structures are followed by polymerizing F-actin (Fig. 4E). To address this possibility quantitatively, we cross-correlated time-shifted images of Lifeact::mKate-2 and mNG::PH^PLC1δ1, and determined that maximal overlap between the two signals occurred when mNG::PH^PLC1δ1 was ∼9.3±1.5 s ahead of Lifeact::mKate-2 (Fig. 4D). Overall, we concluded that PIP₂ cortical structures move together with, but slightly ahead of, polymerizing F-actin filaments.

Given that PIP₂ cortical structures are followed by F-actin filaments, we hypothesized that actin polymerization might push PIP₂ cortical structures. Compatible with this possibility, we found that the velocity of PIP₂ cortical structures was ∼0.17±0.03 µm/s (Fig. S4D,E), in the range of actin polymerization-driven motility in other systems (Brangbour et al., 2011; Cameron et al., 2000; Carlsson, 2010; Mogilner and Oster, 1996). To test this possibility, we impaired actin polymerization by partially depleting the profilin PFN-1, which binds and functions together with the Formin Homology protein CYK-1 to promote polymer assembly (Severson et al., 2002; Velarde et al., 2007). A disruption of F-actin filaments monitored by Lifeact::mKate-2 was observed in pfn-1(RNAi) embryos (Fig. S5A, top, Movie 9). Importantly, we also found that the overall velocity of PIP₂ cortical structures was reduced to ∼0.06±0.04 µm/s in pfn-1(RNAi) embryos (Fig. S5B). Moreover, whereas control embryos exhibited fast and directed movement (Fig. S5C, Movie 8), in half of the pfn-1(RNAi) embryos (n=4/8), which were the least affected, given that cytokinesis was still occurring, PIP₂ cortical structures moved in a somewhat directed manner, albeit at lower velocities (Fig. S5,D). In more strongly affected pfn-1(RNAi) embryos (n=4/8), in which cytokinesis did not occur, PIP₂ cortical structures moved only very locally and in seemingly random directions with occasional jumps (Fig. S5E, Movie 10). Together, these findings indicated that PIP₂ cortical structure movements are driven by actin polymerization.

Given the tight coupling between cortical PIP₂ structures and F-actin, we investigated whether the actomyosin network regulates not only the movement, but also the formation of PIP₂ structures. We found that, in nmy-2(RNAi) embryos, in which actomyosin network contractility is abolished (Munro et al., 2004), PIP₂ cortical structures were present (Fig. 4F,G), but more elongated than usual (Fig. S6A,B). Moreover, PIP₂ cortical structures were distributed symmetrically in nmy-2(RNAi) embryos (Fig. 4F,G), as expected from the known requirement of NMY-2 in A–P polarity (Guo and Kemphues, 1996). Therefore, formation of PIP₂ cortical structures does not depend on a contractile actomyosin cortex. By contrast, we found that few PIP₂ cortical structures form in act-1(RNAi) embryos (Fig. 4G,H,I; Fig. S6C-F). Moreover, acute impairment of F-actin through treatment of perm-1(RNAi) embryos with cytochalasin D led to the disappearance of PIP₂ cortical structures (Fig. S6G,H). However, PIP₂ cortical structures remained upon depletion of the α-tubulin TBA-2 (Fig. 4J,K), indicating that they form independently of the microtubule cytoskeleton. Overall, we concluded that the formation of PIP₂ cortical structures depends on F-actin.

We set out to address whether, conversely, PIP₂ regulates F-actin organization. If so, then changing the level of PIP₂ should alter actomyosin network organization. We tested this prediction first by depleting PIP₂. To this end, we activated phospholipase C, an enzyme that cleaves PIP₂, by delivering ionomycin and Ca²⁺ into perm-1(RNAi) embryos (Fig. S7A) (Hammond et al., 2012; Várnai and Balla, 1998). Cleavage of PIP₂ at the plasma membrane was monitored by the gradual loss of GFP::PH^PLC1δ1 plasma membrane signal, which enabled us to determine the time t_1/2 when half of the initial GFP::PH^PLC1δ1 plasma membrane fluorescence signal disappeared in the embryo middle plane (Fig. S7B,C). We found that PIP₂ removal following ionomycin/Ca²⁺ treatment during pseudocleavage led to a rapid change in embryo shape on the anterior side, coincident with altered F-actin organization (compare Fig. 5A and Fig. 5B, Fig. S7E). An analogous F-actin alteration was observed following ionomycin/Ca²⁺ treatment during mitosis (Fig. 5C; Movie 11). Furthermore, we noted that ionomycin/Ca²⁺-treated embryos exhibited variable spindle positioning compared with the control condition (Fig. S8A-F).

To test whether alterations in F-actin organization cause the embryo shape change following ionomycin/Ca²⁺ treatment, we added latrunculin A to such embryos, thus depolymerizing the F-actin network. As shown in Fig. S7F and Movie 12, we found that this resulted in normally shaped embryos. Therefore, shape changes following PIP₂ removal are F-actin dependent. Although we cannot formally rule out that ionomycin/Ca²⁺ caused these phenotypes for another reason, these results taken together indicated that PIP₂ is crucial for proper F-actin distribution and, thus, the shape of the C. elegans zygote.

We sought to further examine the relationship between PIP₂ and F-actin by increasing the level of PIP₂. We investigated whether this could be achieved by altering individual enzymes from the PIP₂ biosynthetic pathway using RNAi or mutant animals (Fig. S7A, Table S1). Using both RNAi and an age-1(m333) null mutant, we first targeted AGE-1, the catalytic subunit of the sole C. elegans Class I PI3K, which can generate PIP₃ from PIP₂. We found that age-1(RNAi) embryos did not exhibit PIP₂ alterations (Table S1). Likewise, whereas the progeny of age-1 null mutants arrest as dauer larvae (Larsen et al., 1995; Morris et al., 1996), we found no change in PIP₂ in the corresponding embryos (Table S1), perhaps because compensatory mechanisms operate to keep the PIP₂ level constant (Ayyadevara et al., 2009; Shmookler Reis et al., 2012). Furthermore, we did not find another single RNAi or mutant condition with altered PIP₂ (Table S1), possibly because of redundancy among enzymes in PIP₂ biosynthesis. With this in mind, we jointly inactivated ocrl-1 and unc-26, for the following reasons. OCRL-1 is an inositol 5-phosphatase that hydrolyzes PIP₂ to PI4P, and its depletion leads to increased PIP₂ on C. elegans phagosomes (Cheng et al., 2015). Moreover, UNC-26 is homologous to Synaptojanin, a polyphosphoinositide phosphatase that also hydrolyzes PIP₂ to PI4P, and the impairment of which results in vesicle trafficking defects and cytoskeletal abnormalities in the worm nervous system (Charest et al., 1990; Harris et al., 2000).

We jointly depleted the function of these two PIP₂ phosphatases, using RNAi for ocrl-1 and a mutant for unc-26. By comparing cortical mCherry-PH^PLC1δ1 mean intensity values, we found that ocrl-1(RNAi) unc-26(s1710) embryos exhibited an increased overall level of PIP₂ (Fig. S9A,B). Importantly, this led to drastic alterations in PIP₂ monitored by GFP::PH^PLC1δ1 or mCherry-PH^PLC1δ1 (compare Fig. 5D and 5E,F top; Fig. S9C-H). First, in addition to motile PIP₂ structures, we found immotile PIP₂ clusters residing between the eggshell and the plasma membrane (Fig. 5E,H,K, arrows; Fig. S9I, arrowhead). Second, motile PIP₂ structures did not disappear as readily after pseudocleavage as they normally do (Fig. 5H,I, compare with 5G; Fig. S9D). Third, we found that motile PIP₂ structures exhibited altered distribution in all ocrl-1(RNAi) unc-26(s1710) embryos (Fig. 5E,F; Fig. S9F,H). In some cases (hereafter referred to as class I embryos, n=43/72), anteriorly directed movements of PIP₂ cortical structures did not stop at pseudocleavage, but instead continued until the end of mitosis, resulting in a small anterior PIP₂ domain [compare Fig. 5G,J (top) with Fig. 5H (top),K; Movie 13]. In class II ocrl-1(RNAi) unc-26(s1710) embryos (n=29/72), weak anteriorly directed movement of cortical PIP₂ was initiated, but PIP₂ structures then became distributed throughout the cortex by the end of the first cell cycle, except at the very posterior (Fig. 5I,L). Whereas a clear cytokinesis furrow formed in all class 1 embryos, this happened in only 14/29 class 2 embryos; this subset exhibited the most pronounced anteriorly directed movements. Overall, we concluded that depletion of PIP₂ 5-phosphatases was less pronounced in class 1 than in class 2 embryos, with the severe phenotype in the latter perhaps reflecting an impact on multiple processes. Interestingly, imaging ocrl-1(RNAi) unc-26(s1710) embryos expressing Lifeact::mKate-2 (class 1 n=3/7, class 2 n=4/7) or Lifeact::mKate-2 and GFP::PH^PLC1δ1 (class 1 n=3/7, class 2 n=4/7) revealed that F-actin distributed as PIP₂ cortical structures (Fig. 5D-L). We also noted that spindle positioning was more variable in both class 1 and class 2 ocrl-1(RNAi) unc-26(s1710) embryos than in the control condition (Fig. S8G-N). These findings established that an increase in PIP₂, as achieved in class 1 embryos, leads to sustained cortical flows towards the anterior side. Moreover, although there might also be indirect effects of PIP₂ 5-phosphatase depletion on phosphoinositides other than PIP₂, these results further indicated that PIP₂ regulates actin cytoskeletal organization in one-cell C. elegans embryos.

We sought to investigate whether such regulation of PIP₂ is exerted through the promotion of actin polymerization. We reasoned that, if this were the case, then enhancing actin polymerization in a different manner might mimic the unc-26(s1710)/ocrl1(RNAi) phenotype. Therefore, we treated perm-1(RNAi) embryos with jasplakinolide, which promotes actin polymerization (Fig. S10). Upon such treatment, large actin clumps formed, sometimes leading to detachment of the actin cytoskeleton from the plasma membrane (n=7/17; data not shown). However, the actin cytoskeleton moved exaggeratedly towards the anterior in most embryos treated with jasplakinolide during pseudocleavage (Fig. S10B,D,H; n=10/17), as in class 1 unc-26(s1710)/ocrl1(RNAi) embryos. To further test the hypothesis that PIP₂ regulation is exerted through actin polymerization promotion, we treated permeabilized unc-26(s1710)/ocrl1(RNAi) embryos with low concentrations of cytochalasin D (250-500 nM) (Fig. S10A-F). We reasoned that this should ameliorate the unc-26(s1710)/ocrl1(RNAi) phenotype if it is caused by increased actin polymerization. This expectation was verified: whereas DMSO-treated perm-1(RNAi)/unc-26(s1710)/ocrl1(RNAi) embryos usually exhibited class 1 or class 2 phenotypes (n=6 and n=1, respectively, with one additional embryo exhibiting no apparent phenotype), all cytochalasin D-treated perm-1(RNAi)/unc-26(s1710)/ocrl1(RNAi) embryos polarized normally (Fig. S10A,C,E; n=6). Together, these experiments established that PIP₂ regulates F-actin polymerization.

The regulation of actin polymerization and F-actin organization by PIP₂ could conceivably be mediated through the actin regulators ECT-2, RHO-1 and CDC-42, with which PIP₂ cortical structures coincide (Fig. 3). We explored this possibility for RHO-1 and CDC-42 by depleting PIP₂ using ionomycin/Ca²⁺. This resulted in the loss of both cortical GFP::RHO-1 (n=5) and GFP::CDC-42 (n=7) if PIP₂ depletion occurred before pseudocleavage (Fig. S11A-D). Moreover, PIP₂ depletion after pseudocleavage resulted in cortical GFP::RHO-1 loss (n=2), but generally not in that of GFP::CDC-42 structures (Fig. S11E, n=5/7).

Given the above, altering PIP₂ cortical structure distribution might likewise change that of RHO-1 and CDC-42. To test this possibility, we examined GFP::RHO-1 and GFP::CDC-42 distribution in unc-26(s1710)/ocrl1(RNAi) embryos, in which PIP₂ cortical structure organization is altered (Fig. 5). We found that GFP::RHO-1 and GFP::CDC-42 distributions mirrored that of mCherry::PH^PLC1δ1 in such embryos (Fig. S11F-K). Overall, these results indicated that PIP₂ is needed for RHO-1 cortical structures early and late during the first cell cycle, and also for CDC-42 cortical structures, primarily until the pseudocleavage stage.

It is well known that the actomyosin network is essential for A–P polarity in C. elegans zygotes (Guo and Kemphues, 1996; Hill and Strome, 1990; Munro et al., 2004; Strome and Wood, 1983). Given that a proper level of PIP₂ is essential for correct actomyosin network organization, we tested whether it is also important for A–P polarity. We investigated the impact of excess PIP₂ on polarity using ocrl-1(RNAi) unc-26(s1710) embryos expressing mCherry::PH^PLC1δ1 and GFP::PAR-2 or GFP::PAR-6 (Fig. 6A-L). We found that the distribution of GFP::PAR-2 and GFP::PAR-6 domains changed from early on in development, consistent with alterations in motile PIP₂ structures and F-actin. Thus, for GFP::PAR-6, either a small domain formed on the very anterior (Fig. 6B,E; class 1, n=5/9; Movie 14) or the fusion protein remained present over the entire cortex, except on the very posterior (Fig. 6C,F class 2; n=4/9). As expected, GFP::PAR-2 distributed reciprocally, either expanding drastically towards the anterior (Fig. 6H,K; class 1, n=12/22; Movie 15) or remaining restricted to the very posterior (Fig. 6I,L; class 2, n=10/22). Moreover, we observed that spindle positioning was affected in a variable manner in ocrl-1(RNAi) unc-26(s1710) embryos (Fig. S8G-N), as anticipated from altered A–P polarity and potentially other consequences of excess PIP₂. Overall, these results indicated that a correct level of PIP₂ is needed for proper F-actin network localization and, as a consequence, appropriate PAR polarity and accurate spindle positioning.

In the above experiments, the level of PIP₂ was in excess from the beginning of development, such that the impact on polarity might reflect a role either strictly during the establishment phase or during both the establishment and maintenance phases. We reasoned that one could test specifically a potential role for PIP₂ in polarity maintenance by adding ionomycin/Ca²⁺ during pseudocleavage, after polarity establishment, to perm-1(RNAi) embryos expressing mCherry::PH^PLC1δ1 and GFP::PAR-2. We found that GFP::PAR-2 expanded slowly towards the anterior, starting ∼3 min after t_1/2 (Fig. 6M-O), with the pseudocleavage furrow initially moving anteriorly, and then either disappearing (Fig. 6N, Movie 16; n=6/14), or remaining at the very anterior (Movie 17; n=8/14). Likewise, GFP::PAR-2 expanded slowly towards the anterior when t_1/2 occurred at nuclear envelope breakdown (NEBD) (Fig. 6O, Movie 18). Together, these results indicated that an appropriate level of PIP₂ is also essential for proper PAR polarity during the maintenance phase.

In principle, PIP₂ could alter PAR polarity during the maintenance phase either through an impact on F-actin organization or via an actin-independent role. Unlike its well-known role during polarity establishment, a potential role of F-actin in polarity maintenance is somewhat controversial (Goehring et al., 2011; Hill and Strome, 1990; Liu et al., 2010; Severson and Bowerman, 2003). Considering our findings related to changes in the PIP₂ level, we directly tested whether F-actin plays a role in polarity maintenance. We first added cytochalasin D to perm-1(RNAi) embryos expressing Lifeact::mKate-2 and GFP::PAR-2 after polarity establishment (Fig. 7A,B). Consistent with previous studies (Goehring et al., 2011; Hill and Strome, 1990), we found that cytochalasin D addition at this stage did not significantly alter GFP::PAR-2 distribution (Fig. 7A,B, Movie 19). However, we also found that cytochalasin D did not fully disrupt F-actin, because clumps of Lifeact::mKate-2 remained in the embryo anterior (Fig. 7B). Thus, we turned to inhibiting F-actin polymerization using latrunculin A, which resulted in complete F-actin depletion (Fig. 7C; n=12; Movie 20). We observed membrane invaginations that removed GFP::PAR-2 from the cortex into cytoplasmic aggregates (Fig. 7C, arrowhead), as previously reported (Goehring et al., 2011; Redemann et al., 2010). Importantly, we monitored changes in GFP::PAR-2 distribution as a function not of drug addition time, as done previously (Goehring et al., 2011), but of the time at which half of the Lifeact::mKate-2 fluorescence had disappeared from the membrane. In doing so, we found a decrease of the GFP::PAR-2 domain after t_1/2 in all embryos analyzed (Fig. 7C, bottom; n=12) that was highly correlated with Lifeact::mKate-2 disappearance (Fig. S9J), demonstrating that F-actin is crucial for PAR polarity maintenance.

Overall, we found that an appropriate level of PIP₂ is essential for the correct sizing of PAR domains, through the reorganization of F-actin, during not only polarity establishment, but also polarity maintenance.

In this work, we demonstrated that PIP₂ forms cortical structures in the one-cell C. elegans embryo. We showed that the formation and movement of these structures depend on F-actin and, reciprocally, that PIP₂ regulates F-actin organization, revealing an interdependence of these two components in the worm zygote (Fig. 7D). Moreover, probably through its impact on the actin cytoskeleton, PIP₂ is also needed for the correct sizing of PAR domains, demonstrating that a plasma membrane lipid component participates in A–P polarity in the C. elegans embryo.

The distribution and dynamics of PIP₂ at the plasma membrane of early C. elegans embryos were not clear before this work, primarily because only the middle embryo plane was analyzed in past investigations. Here, we assayed subcellular distributions at the cortex, where the function of PAR proteins and components that are crucial for asymmetric division is exerted. Thus, we discovered that PIP₂ is present in dynamic polarized cortical structures marked by the PIP₂ biomarker GFP::PH^PLC1δ1 and Bodipy-FL-PIP₂, consistent with recent observations mentioning non-uniform GFP::PH^PLC1δ1 distribution (Rodriguez et al., 2017; Wang et al., 2017). Although patches of plasma membrane enriched in PIP₂ have been observed elsewhere (Chierico et al., 2015; Golub and Caroni, 2005; McLaughlin et al., 2002; Zhang et al., 2012), the stereotyped progression through the first cell cycle of the large worm zygote enabled us to probe such cortical structures with unprecedented resolution. Why were these structures not observed previously? On top of not being noticeable when imaging the middle plane of the embryo, other plausible reasons are that PIP₂ cortical structures appear only transiently during the cell cycle and that they are not preserved upon fixation (data not shown).

How do PIP₂ cortical structures assemble? Two reasons lead us to propose that PIP₂ cortical structures might form by redistributing existing PIP₂ rather than by de novo synthesis through PIP5K1. First, PIP₂ in other systems has been suggested to diffuse much faster than it is synthesized (McLaughlin et al., 2002), such that potential local synthesis is unlikely to dictate restricted PIP₂ localization. Second, PPK-1, the sole PIP5K1 in C. elegans, is enriched in the posterior of the embryo (Panbianco et al., 2008), seemingly not where most PIP₂ cortical structures reside. However, localization of PIP5K1 to PIP₂ cortical structures might have been missed by analyzing fixed embryos and imaging the middle plane. Interestingly, we also found that PIP₂ cortical structures form and move independently of actomyosin contractions powered by NMY-2. Nevertheless, PIP₂ cortical structures might form at membrane protrusions or ruffles, which would be consistent with PIP₂ stimulating F-actin polymerization in curved but not flat membranes (Gallop et al., 2013), and accumulating in membrane ruffles, nascent phagosomes and the leading edge of motile cells (reviewed by McLaughlin et al., 2002; Zhang et al., 2012). In other systems, increased PIP₂ induces actin polymerization around membrane vesicles, generating actin comets that propel vesicles forward (Ma et al., 1998; Rozelle et al., 2000). We also found occasional vesicles and connected local F-actin accumulation upon increasing PIP₂, although no actin comets were observed (data not shown).

In summary, we propose that PIP₂ cortical structures form through the redistribution of existing PIP₂ at the plasma membrane in the C. elegans zygote, perhaps preferentially at membrane protrusions or ruffles.

PIP₂ and F-actin exhibit a reciprocal relationship in a number of systems, and we show here that this is also the case in C. elegans. We found that PIP₂ cortical structures and F-actin movements are coupled, with PIP₂ structures moving slightly ahead, at velocities compatible with actin polymerization driving these movements. Moreover, impairing the profilin PFN-1, which is essential for microfilament assembly (Severson et al., 2002; Velarde et al., 2007), reveals that PIP₂ structure movements are actin driven. This leads us to propose that actin polymerization pushes PIP₂ cortical structures, reminiscently of other actin-dependent motility processes, such as that of Listeria monocytogenes (reviewed by Mogilner and Oster, 1996). While being pushed ahead of F-actin in C. elegans, PIP₂ structures might recruit factors promoting actin polymerization and branching, such as ECT-1, RHO-1 and CDC-42, as in other systems (reviewed by Chichili and Rodgers, 2009). Intriguingly, a biosensor for active CDC-42 distributes on the cortex like PIP₂ does (Cheng et al., 2015; Kumfer et al., 2010), indicating that CDC-42 at PIP₂ structures is active. By contrast, the distribution of a biosensor for active RhoA overlaps with that of NMY-2 (Reymann et al., 2016; Tse et al., 2012). Given that we showed here that PIP₂ cortical structures did not overlap with NMY-2, whereas they overlapped with GFP::RHO-1, the bulk of RHO-1 associated with PIP₂ cortical structures might be inactive. Alternatively, given that RHO-1 colocalizes with its activating GEF ECT-2, this RhoA biosensor might not detect all active RHO-1 species. Furthermore, it is interesting that non-muscle myosin 2 contributes to actin network disassembly in fish keratinocytes (Wilson et al., 2010). Extrapolating from this observation, it is tempting to speculate that PIP₂, by promoting F-actin assembly, and NMY-2, by promoting F-actin disassembly, in addition to powering network contractility, might together ensure proper F-actin dynamics in C. elegans embryos.

PAR proteins are also distributed unevenly within their domain. Thus, cortical PAR-6 exists in two populations, one diffuse that depends on CDC-42 and one punctate that colocalizes with PAR-3 (Beers and Kemphues, 2006; Robin et al., 2014). Moreover, PAR-3 forms clusters that are crucial for proper polarity, and the assembly of which depends on PCK-3, CDC-42, as well as actomyosin contractility (Rodriguez et al., 2017; Wang et al., 2017). Moreover, the putative CDC-42 GAP CHIN-1 forms clusters on the posterior, and the transport of these clusters stabilizes the A–P boundary during polarity maintenance (Sailer et al., 2015). Intriguingly, we found that PIP₂ cortical structures colocalized within the more diffuse cortical PAR-6 population, which lacks PAR-3 (Beers and Kemphues, 2006; Robin et al., 2014; Rodriguez et al., 2017; Wang et al., 2017). Moreover, we established that increasing PIP₂ augmented the segregation of both PAR-6 populations to the anterior. Given that clustering of PAR-3 depends on actomyosin reorganization (Rodriguez et al., 2017; Wang et al., 2017), we propose that increasing the level of PIP₂ might reorganize the actin cytoskeleton in a way that promotes PAR-3 clustering, thereby aiding segregation.

We showed that an appropriate PIP₂ level is essential for proper polarity establishment and maintenance. When increasing the level of PIP₂, the continued movement of PAR domains towards the anterior until the end of mitosis altered their relative size. This is reminiscent of changes in PAR domain sizing that occur upon RGA-3/4 depletion (Schonegg et al., 2007). However, whereas rga-3/4(RNAi) embryos exhibit a hypercontractile actomyosin network, embryos with increased PIP₂ level do not. Thus, we propose that actomyosin contractility regulated by NMY-2 and F-actin organization regulated by PIP₂ contribute in concert to the correct movements of the actomyosin network and, therefore, proper sizing of PAR polarity domains.

The actomyosin network plays a well-established role during polarity establishment, whereas its role during polarity maintenance has been debated (Goehring et al., 2011; Hill and Strome, 1990; Liu et al., 2010; Severson and Bowerman, 2003). Our results, together with those of others, indicated that the actomyosin network regulates PAR domains in two ways. First, when the actomyosin network moves anteriorly during the establishment phase, PAR domains alter their distribution accordingly. This relationship was clear before this work, and we reached an identical conclusion here. Second, actin has been suggested to play merely a passive role during the maintenance phase in preventing cortical PAR-2 removal through membrane invaginations driven by microtubules (Goehring et al., 2011). We showed here that this can lead to the near-total disappearance of cortical GFP::PAR-2, further emphasizing the importance of actin also during the maintenance phase.

Overall, our results in C. elegans are consistent with the role of PIP₂ in F-actin reorganization and polarity in other organisms. Previous work in C. elegans showed that depletion of CSNK-1, which negatively regulates PPK-1 localization, although perturbing spindle positioning, does not impact polarity (Panbianco et al., 2008), perhaps because PPK-1 plays only a minor role in regulating the level of PIP₂ in the zygote. By contrast, we established here that alterations in the level of PIP₂ impair polarity establishment and maintenance during the first asymmetric division of C. elegans embryos.

Nematodes were maintained at 24°C using standard protocols (Brenner, 1974). The following strains were used: GFP::PH^PLC1δ1 (OD58, unc-119(ed3) III; ltIs38[pie-1p::GFP::PH(PLC1delta1)+unc-119(+)]) (Audhya et al., 2005); mCherry::PH^PLC1δ1 (OD70, unc-119(ed3) III; ltIs44[pie-1p::mCherry::PH(PLC1delta1)+unc-119(+)]V) (Audhya et al., 2005); GFP::PAR-2 (TH129) and GFP::PAR-6 (TH110) (Schonegg et al., 2007); GFP::NMY-2 (LP162, nmy-2(cp13[nmy-2::gfp+LoxP]) I) (Dickinson et al., 2013); CAV-1::GFP (RT688, unc-119(ed3) III; pwIs281[CAV-1::GFP, unc-119(+)]) (Sato et al., 2006); mNeonGreen::PH^PLC1δ1 (LP274, cpIs45[Pmex-5::mNeonGreen::PLCδ-PH::tbb-2 3′UTR+unc-119(+)] II; unc-119(ed3) III), mKate-2::PH^PLC1δ1 (LP307, cpIs54[Pmex-5::mKate2::PLCδ-PH(A735T)::tbb-2 3′UTR+unc-119(+)] II; unc-119(ed3) III) and mCherry::PH^PLC1δ1 (LP308, cpIs55[Pmex-5::mCherry-C1::PLCδ-PH::tbb-2 3′UTR+unc-119(+)] II; unc-119(ed3) III) (Heppert et al., 2016); Lifeact::mKate-2 (strain SWG001) (Reymann et al., 2016); GFP::RHO-1 (SA115, unc-119(ed3) III; tjIs1[pie-1::GFP::rho-1+unc-119(+)]) (Motegi et al., 2006); GFP::CDC-42 (SA131, unc-119(ed3) III; tjIs6[pie-1p::GFP::cdc-42+unc-119(+)]) (Motegi and Sugimoto, 2006); GFP::ECT-2 (SA125, unc-119(ed3) III; tjIs4[pie-1::GFP::ect-2+unc-119(+)]) (Motegi and Sugimoto, 2006); unc-26(s1710) (EG3027, unc-26(s1710) IV) (Charest et al., 1990); and age-1(m333), (DR722, age-1(m333)/mnC1 dpy-10(e128) unc-52(e444) II) (Riddle, 1988). Crosses were performed at 20°C to generate lines that were homozygous for all transgenes, which were then maintained at 24°C. For GFP::RHO-1 and mCherry::PH^PLC1δ1 (only in Fig. 3D), as well as GFP::PH^PLC1δ1 and Lifeact::mKate-2, worm lines were crossed and F1 progeny heterozygous for both transgenes were imaged.

RNAi-mediated depletion was performed essentially as described elsewhere (Kamath et al., 2001), using bacterial feeding strains from the Ahringer (Kamath et al., 2003) or Vidal libraries (Rual et al., 2004) (gift from Jean-François Rual and Marc Vidal). RNAi for par-2 (Ahringer), par-3 (Ahringer), nmy-2 (Ahringer), act-1 (Vidal), tba-2 (Vidal), rho-1 (Vidal), cdc-42 (Vidal) and ocrl-1 (Ahringer) was performed by feeding L3-L4 animals with bacteria expressing the corresponding dsRNA at 24°C for 16-26 h. RNAi for pfn-1 (Vidal) was performed by feeding L4 and young adults with bacteria expressing dsRNA at 24°C for 72-96 h and imaging embryos of their offspring. PERM-1 is a sugar-modifying enzyme essential for forming the permeability barrier of the eggshell (Carvalho et al., 2011; Olson et al., 2012); RNAi for perm-1 (Ahringer) was performed by feeding L4 and young adults with bacteria expressing dsRNA at 20°C for 12-18 h. Double RNAi for ocrl-1 and perm-1 was performed by mixing bacteria expressing ocrl-1 dsRNA and perm-1 dsRNA in a 1:1 ratio and then feeding them to L3 and L4 young adults for 24-30 h at 24°C. The effectiveness of depletion was assessed phenotypically as follows: par-2(RNAi) and par-3(RNAi): symmetric spindle positioning and equal cell division; nmy-2(RNAi) and act-1(RNAi): absence of cortical ruffles, symmetric spindle positioning, no cytokinesis; tba-2(RNAi): defective pronuclear meeting, no centration/rotation, no spindle assembly, misplaced cytokinesis furrow; ocrl-1(RNAi) [in combination with unc-26(s1710)]: immotile PIP2 structures, class 1 or class 2 phenotype (Fig. 5E,F,H,I; Fig. 6B,C,H,I), altered centrosome and spindle positioning (Fig. S8H-K); perm-1(RNAi): successful action of added drug; pfn-1(RNAi): defective cortical F-actin network; cdc-42(RNAi): partial loss of polarity during the maintenance phase; rho-1(RNAi): absence of cortical ruffles.

Gravid hermaphrodites were dissected in osmotically balanced blastomere culture medium (Shelton and Bowerman, 1996) and the resulting embryos mounted on a 2% agarose pad. DIC time lapse microscopy (Fig. 1A,C,E,G,I) was performed at 25°C±1°C with a 100× (NA 1.25 Achrostigmat) objective and standard DIC optics on a Zeiss Axioskop 2 microscope. All other images were acquired at 23°C using an inverted Olympus IX 81 microscope equipped with a Yokogawa spinning disk CSU-W1 with a 63× (NA 1.42U PLAN S APO) objective and a 16-bit PCO Edge sCMOS camera. Images were obtained using 488-nm and 561-nm solid-state lasers with an exposure time of 400 ms and a laser power of 20-60%. For cortical imaging, three planes at the cell cortex (each 0.25 μm apart) were acquired. Cell cycle stages were determined using transmission light microscopy, imaging the middle plane in parallel (data not shown).

Cortical images of GFP::PH^PLC1δ1 used for quantification were processed as follows: the three cortical planes were z-projected using average intensity projection, after which a median filter of 1 pixel was applied. The background of the entire image was subtracted using the measured mean background in each frame. Signal intensity decay because of photobleaching was corrected with the Fiji plugin ‘bleach correction’ using the exponential fitting method (ImageJ Plugin CorrectBleach V2.0.2. Zenodo; https://zenodo.org/record/30769#.WvQPLKQvwuU). The entire cortical region was segmented by applying a binary automated histogram-based threshold, followed by iterated morphological operations. Cortical structures were segmented by applying a binary intensity threshold, calculated by fitting the pixel intensity histogram with a Gaussian function and setting the threshold at 4 σ from the Gaussian peak. The boundary between anterior and posterior domains was determined manually during pseudocleavage formation. Upon depletion of NMY-2 and ACT-1, whereby no pseudocleavage furrow forms, the boundary was set during mitosis. The fraction of the total area in each half covered by the segmented PIP₂ cortical structures was determined in each case.

Curves of normalized cortical structures sizes were fitted with a sigmoidal model and synchronized, setting the sigmoid inflection point, which corresponded typically to the time of centration/rotation, as time t=0 s. Curves of normalized cortical structures sizes were aligned manually for act-1(RNAi) and unc-26(s1710) ocrl-1(RNAi) embryos using the clear landmark provided by NEDB as a reference, because a sigmoid function could not be fitted with the PH markers in these cases. Given that the time separating centration/rotation from NEDB is typically 150 s, t=0 was set at –150 s before NEDB for act-1(RNAi) and unc-26(s1710) ocrl-1(RNAi) embryos.

using the MATLAB image processing function ‘regionprops’. The Elongation Index was then normalized by a factor of 1/π, such that a square of 2×2 pixels has an Elongation Index of 1.

For assessing the A–P boundary of the segmented PIP₂ cortical structures (GFP::PH^PLC1δ1) and of F-actin (Lifeact::mKate-2) (Fig. S2), we automatically determined the extent of the PIP₂ and F-actin domains relative to the whole embryo length using a Matlab script. For Lifeact::mKate2, a histogram-based method was used to keep only the 95% brightest pixels.

Cortical images obtained by live confocal spinning disk imaging shown in the figures were processed as follows: the three cortical planes were z-projected using maximum intensity projection, then a median filter of 1 was applied. The gray value fluorescence intensity of some transgenes (GFP::PH^PLC1δ1 as heterozygote, GFP::PAR-2, GFP::PAR-6, Lifeact::mKate-2, mCherry::PH^PLC1δ1 and mNeonGreen-PH^PLC1δ1) was slightly variable, probably resulting from variable expression and/or folding of the fluorescent fusion protein. Therefore, the brightness and contrast of images resulting from embryos expressing these transgenes was adjusted accordingly. Such variability was especially pronounced in unc-26(s1710) ocrl-1(RNAi) embryos. Missing edges or corners in the following figures and movies result from one or multiple rotations to position the anterior of the embryo to the left: Fig. 4A, Fig. 5A,B,F,I, Fig. 6M,O, Fig. S2A,E,F, Fig. S3C,D, Fig. S5A, Fig. S9I, Fig. S10D, Fig. S11D, Movies 5, 9, 11, 12, 20. To compare the intensity of mCherry-PH^PLC1δ1 in control embryos and unc-26(s1710) ocrl-1(RNAi) embryos, three cortical planes, acquired as described above, were z-projected by summing the intensity of all slices. The resulting mean intensities were then computed as follows. First, an Otsu threshold was used to retrieve the brightest elements, retaining only the biggest blob, which corresponded to the embryo. Values outside the embryo were averaged to obtain the mean background intensity, which was subtracted from the embryo pixel intensities. Thereafter, embryo pixel values were averaged to obtain the mean pixel intensity value.

BODIPY FL phosphatidylinositol 4,5-bisphosphate (Echelon Bioscience, C-45F6) (final concentration 10 µM) was delivered to perm-1(RNAi) embryos by adding it to the buffer in which gravid worms were dissected. Embryos were then mounted on a 2% agarose pad using petroleum jelly as a spacer to decrease the pressure exerted on the fragile perm-1(RNAi) embryos.

For particle image velocimetry (PIV) analysis, cortical image sequences of mNeonGreen::PH^PLC1δ1 and Lifeact::mKate-2 were prepared by performing a maximum intensity z-projection of a stack of two planes (0.25 µm apart) and applying a median filter of 1 pixel. PIP₂ cortical structures and the F-actin network were then segmented as follows: the embryo was first extracted from the background using a histogram-based automated threshold, keeping only blobs of a size superior to one-third of the biggest blob. The resulting binary images were deemed to be the embryo area. We applied a morphological erosion to the mNeonGreen::PH^PLC1δ1 movies with a large structuring element (a disk 30 pixels in radius) to calculate the average value of the pixels not corresponding to PIP₂ cortical structures; the PIP₂ cortical structures were then segmented as the pixels of intensity higher than the computed average value multiplied by a scaling factor determined empirically (1.7). The extraction of the F-actin network was achieved by determining a histogram-based automated threshold on the morphological top-hat of the F-actin image. F-actin filaments and PIP₂ cortical structures were segmented before PIV analysis to ensure that only flow fields in the region of interest were measured.

PIV was then performed to measure cortical flows using the MATLAB based PIVlab toolbox (Thielicke and Stamhuis, 2014), which splits each image of a movie into a regular grid, for which the size of grid cells is given by the user. The position of each grid cell in the next image is estimated by finding the maximum normalized cell-to-cell cross-correlation of equivalent sizes in a geometrical neighborhood called the ‘interrogation area’. Such PIV analysis was applied to mNeonGreen::PH^PLC1δ1 and Lifeact::mKate-2 separately, after segmentation of the corresponding cortical structures. The choice of the grid cell sizes and interrogation areas resulted from a balance between two criteria: smaller cells allow one to compute displacements with high spatial resolution, but excessively small cells do not contain enough information to be reliably correlated with other cells; thus, the estimation of displacements of bigger cells is more reliable, but computed with lower resolution. We found empirically 32×32 pixels for cell sizes, and 64×64 pixels for interrogation areas, to be a good compromise.

to the location correlation as a function of the time shift. We calculated the best a, b and t₀ parameters using a least-squares method, and input the standard deviations of the correlations to create a weight matrix used during the adjustment. The results were the mean time shift t₀=9.3 s and the standard deviation sigma_t₀=1.5 s.

To track PIP₂ structures (Fig. 4E, Fig. S4D,E), embryos expressing mNG-PH^PLC1δ1 were imaged with an exposure time of 50 ms, a laser power of 60% and a 70-ms frame rate. PIP₂ structures were tracked manually on maximum intensity z-projection of the images containing the moving PIP₂ structures of interest. The length of the track was obtained by re-slicing it using the Fiji plugin ‘Reslice’. The velocity was calculated from the corresponding number of time points and track length.

Spindle positioning tracks were generated by manually tracking spindle poles from NEBD to cleavage furrow formation using a MATLAB script that also computed the distance from the first to the last tracked point, providing the corresponding x and y coordinates and the maximum velocity in µm/s. Tracks were automatically placed into ellipses fitted around the embryo. The end positions of the spindle poles were determined at cytokinesis onset.

The eggshell was permeabilized by performing perm-1(RNAi) as described above. Gravid hermaphrodites were dissected in a cell culture dish with a glass bottom, and the resulting embryos imaged with an inverted confocal spinning disk microscope (see above). Drugs were added under the microscope while imaging to precisely control the timing of drug addition. The following drugs and concentrations in solution were utilized: 30 µM ionomycin (Calbiochem, 407950), 3-5 mM CaCl₂ (Sigma-Aldrich, C5080), 20 µM cytochalasin D (AppliChem, 22144-77-0) and 12.5 µM latrunculin A (Sigma-Aldrich, 76343-93-6). For control movies, dimethyl sulfoxide (DMSO) at a concentration equivalent to the final DMSO concentration in the drug solutions was added to the buffer before dissection. For the drug delivery experiments shown in Fig. S10, gravid hermaphrodites were dissected in either 250 nM or 500 nM cytochalasin D (AppliChem, 22144-77-0) or 2.5 µM jasplakinolide (AdipoGen, AG-CN2-0037). Embryos were then imaged, acquiring 17 z-planes 0.5 µm apart starting every 30 s. For the experiments reported in Figs 5 and 6, whenever embryos were dissected in the drug-containing solution, we ensured that the action of the drug (as monitored by t_1/2) took place >6 min after polarity was established.

Successful drug action was determined for each embryo by the disappearance of the PH^PLC1δ1 fluorescence signal from the plasma membrane (ionomycin/Ca²⁺ and latrunculin A) and of Lifeact::mKate-2 from the cell cortex (cytochalasin D and latrunculin A). The time between drug addition and drug action was variable, probably because of variations in eggshell permeability upon perm-1(RNAi). Therefore, as a comparable reference time between embryos, we determined the time t_1/2 (t inflection) when half of fluorescence at the plasma membrane has disappeared, as follows: the total cortical region of the embryo was segmented by applying a binary automated histogram-based threshold; fluorescent values at a distance of 20 pixels from the edge were measured, and their mean fluorescence values plotted over time; the inflection point of a fitted sigmoid function was then determined as t_1/2.

After addition of ionomycin/Ca²⁺ to embryos expressing mCherry::PH^PLC1δ1/GFP::RHO-1 or mCherry-PH^PLC1δ1/GFP::CDC-42, the disappearance of the mCherry-PH^PLC1δ1 signal from the plasma membrane was imaged over time, whereas GFP::RHO-1 and GFP::CDC-42 were not imaged continuously to prevent photobleaching. Instead, only one end point image was taken by acquiring nine z-planes that were 0.5 µm apart.

The software packages JMP 13.2.0 (SAS Institute GmbH) and MATLAB 2016 were used to perform statistical analyses. Normal distribution of the data was tested using the Shapiro–Wilk test. Unless stated otherwise, statistical analysis was performed using nonparametric tests assuming non-normal distribution using the Wilcoxon Rank Sum test/Mann–Whitney test for two-group comparisons, and the Kruskal–Wallis test with a pairwise Wilcoxon Rank Sum test as a post-hoc test for three-group comparisons. In Fig. 4B, unpaired t-test was utilized to test the probability that two independent velocity fields result in the observed angle distribution. Values of P≤0.05 were considered statistically significant.

We thank Sachin Kotak and Kalyani Thyagarajan for initial observations of PIP₂ cortical structures, Olivier Burri (BioImaging and Optics Platform, BIOP, School of Life Sciences, EPFL) for help in developing the script for preprocessing of embryos with ImageJ, as well as the BIOP at large for microscopy support. We thank Aitana Lebrand (Neves) for writing the Matlab script used for tracking spindle poles. For strains, we thank Jon Audyha, Daniel Dickinson, Bob Goldstein, Barth Grant, Stephan Grill, Anthony Hyman, Karen Oegema and Anne-Cécile Reymann, as well as the Caenorhabditis Genetics Center (CGC), which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). We are also grateful to Alexandra Bezler, Radek Jankele and Kerstin Klinkert for comments on the manuscript.