Molecular networks linked by Moesin drive remodeling of the cell cortex during mitosis

A universal feature of animal cells undergoing mitosis is the series of transformations in their shape necessary to generate two identical daughter cells. Mitotic cell shape remodeling relies on a precise coupling of cortical actomyosin forces with the plasma membrane. At mitosis entry, increased hydrostatic pressure and isotropic cortical contractility drive the characteristic rounding of prometaphase cells (Matzke et al., 2001; Maddox and Burridge, 2003; Carreno et al., 2008; Kunda et al., 2008; Stewart et al., 2011). Subsequently, asymmetry in cortical tensions leads to polar relaxation and equatorial contraction, which contribute to anaphase cell elongation and to cytokinesis (Hickson et al., 2006; Surcel et al., 2010; Sedzinski et al., 2011). Although mitotic stages were originally described more than one century ago (Flemming, 1882), the molecular networks that modify the cortex to drive transformations in the shape of dividing cells remain to be identified.

We and others have shown that Moesin (Moe) plays essential roles in the regulation of cell shape during mitosis in Drosophila melanogaster (Carreno et al., 2008; Kunda et al., 2008). Moe is the sole Drosophila member of the ERM (Ezrin, Radixin, and Moesin) family of cytoskeletal regulators, which allow, in a signal-dependent manner, bridging of the actin cytoskeleton to the plasma membrane (Fehon et al., 2010). A flexible α-helical linker separates an N-terminal (FERM [4.1 and ERM]) domain from a C-terminal domain (C-ERMAD), which interact with the plasma membrane and with F-actin, respectively. ERM proteins are regulated by a conformational change: in their dormant cytoplasmic state, interaction between the FERM and the C-ERMAD domains masks the two binding surfaces. In response to various signals, ERM proteins open and provide a bridge between actin filaments and the plasma membrane. Activation of ERM proteins involves both the binding of the FERM domain to phosphoinositol 4,5-bisphosphate (PI(4,5)P₂) and the phosphorylation of a conserved threonine residue (T559 in Drosophila Moe) located in the C-ERMAD moiety. Although phosphorylation is a hallmark of ERM activation, interaction with PI(4,5)P₂ has emerged as playing important roles in their regulation (Coscoy et al., 2002; Hao et al., 2009; Roch et al., 2010). Current models state that PI(4,5)P₂ favors conformational opening and that phosphorylation further stabilizes this open active form at the cell cortex (Fehon et al., 2010).

ERM function and proper regulation is required during cell division in both flies (Carreno et al., 2008; Kunda et al., 2008; Cheng et al., 2011) and mammals (Luxenburg et al., 2011). In Drosophila, depletion of Moe or Slik, the kinase necessary to phosphorylate Moe on T559, leads to similar mitotic defects: mutant cells are unable to round at mitosis entry, and their cortex is continuously deformed throughout cell division. Importantly, these cortical defects impinge on additional aspects of mitosis, including spindle morphogenesis and positioning, as well as chromosome segregation (Carreno et al., 2008; Kunda et al., 2008). Therefore, unraveling ERM function and regulation at the cortex represents a key step to better understanding how cell shape transformations are coordinated during division.

Using functional approaches and high-resolution live imaging in Drosophila cultured cells, we show here that the regulated activity of Moe orchestrates changes in tension applied at the cortex and thereby, controls cell shape transformations at the successive steps of cell division. Through systematic screenings of candidate regulators, we identify two networks that collectively provide a spatiotemporal control of Moe activity. The first one relies on Pp1-87B, a phosphatase that counteracts activity of the Slik kinase to restrict high Moe function to early mitosis. Then, the PI(4)P 5-kinase Skittles and PI(3,4,5)P₃ phosphatase Pten further refine the pattern of activated Moe through the local production of PI(4,5)P₂, which is required for both Moe cortical recruitment and phosphorylation. Integration of these two regulatory networks provides a cell cycle–regulated burst of isotropic Moe activation at the cortex, which is required for cell rounding at G2/M transition. Subsequently, the concomitant equatorial enrichment and polar diminution of Moe activity after the anaphase onset synchronizes equatorial contractions with polar relaxation to allow cell elongation and cytokinesis.

As deduced from the pattern of phosphorylated Moe (P-Moe) in fixed samples (Carreno et al., 2008), the location of activated Moe parallels the sites of cortical contractions during mitosis. To gain insight into the role and the regulation of Moe activity at the cell cortex throughout the cell cycle, we examined dynamics of a functional GFP-tagged Moe (Roch et al., 2010) stably expressed in Drosophila S2 cells. Time-lapse microscopy confirmed that Moe localization is tightly regulated during the cell cycle. Although mostly cytoplasmic in interphase, Moe-GFP was recruited to the cell cortex upon mitosis entry (Fig. 1 A). Then, it was redistributed into the cytoplasm after furrow ingression, before complete abscission. Throughout mitosis, the spatial pattern of Moe association with the cortex directly correlated with cell morphology. In prophase, metaphase, and early anaphase, when cells are spherical, Moe-GFP was isotropically associated with the cortex. When cells started to elongate in anaphase B, Moe-GFP was progressively lost from the poles and accumulated at the equator (Fig. 1, B and C; and Video 1).

This dynamic Moe localization likely requires precise regulation of its activity. We tested this hypothesis by substituting endogenous Moe with a constitutively active phosphomimetic mutant, Moe-TD-GFP. To this aim, cells stably expressing Moe-TD-GFP were treated by a double-strand RNA (dsRNA) specifically targeting the 3′ untranslated region (UTR) of endogenous Moe RNA. Under these conditions, hereafter referred to as Moe-TD cells, Moe-TD-GFP accumulated abnormally at the cortex in interphase and remained associated with the polar cortex during cell elongation (Fig. 1, D, F, and G). Although Moe depletion triggered a low rate of cytokinesis failure (Carreno et al., 2008), substituting endogenous Moe by Moe-TD-GFP impaired cytokinesis in 50% of the cells (Fig. 1 E). Because in many cell types completion of cytokinesis relies on proper elongation in anaphase B (Rappaport, 1971), we evaluated the geometry of Moe-TD cells when progressing through anaphase. Indeed, most Moe-TD cells displayed defective elongation, a phenotype associated to cytokinesis failure (Fig. 1 E). It has been shown that anaphase cell elongation requires concomitant equatorial actomyosin contractions and polar relaxation (Hickson et al., 2006). Because activated Moe increases cortical rigidity (Kunda et al., 2008), we hypothesized that Moe overactivation at the polar cortex triggers excessive stiffness by increasing F-actin association and, consequently, blocks cell elongation. Consistently, we measured an increase in F-actin associated with the polar cortex versus the equatorial region in Moe-TD cells (Fig. 1 H). These results indicate that reduction of Moe activity at the polar cortex contributes to cell elongation during anaphase B and thus, to cytokinesis.

To further investigate the role of Moe in anaphase cell elongation, we imaged cortical dynamics in dividing S2 cells. Since pioneering works (Prothero and Spencer, 1968), several studies have shown that short-lived blebs form at each pole during anaphase elongation (Porter et al., 1973; Fishkind et al., 1991; Burton and Taylor, 1997; Boucrot and Kirchhausen, 2007). These polar blebs have been recently shown to be necessary to release the cortical tension and cytoplasmic pressure created by cytokinetic ring furrowing (Sedzinski et al., 2011). High-resolution analyses of living Moe-GFP cells indicated that Moe was not initially associated with the bleb cortex during its expansion and was recruited at the bleb rim just before retraction. In normal conditions, blebs were rapidly retracted, within a couple of minutes (Figs. 2 A and S1 and Video 2), and thus, displayed a small size (Fig. 2 B). The absence of Moe activity leads to abnormally large cytoplasmic bulges deforming the cortex of dividing cells as observed in fixed samples (Carreno et al., 2008), suggesting that it disrupts cortical organization. We thus analyzed Moe-GFP dynamics in living cells during polar blebbing and anaphase relaxation in the absence of Slik kinase to preclude Moe activation. Slik depletion decreased the cortical association of Moe-GFP and prevented its recruitment to the bleb cortex (Fig. 2 A and Video 3). Furthermore, similarly to Moe depletion, disrupting Moe activation through Slik knockdown promoted unregulated growth of blebs as well as their delayed retraction, leading to an apparently exaggerated relaxation of the anaphase cortex (Figs. 2 B and S1).

To further explore how Moe controls normal blebbing during polar relaxation in anaphase, we imaged the actomyosin cortex using stable cell lines that expresses the calponin homology (CH) domain of Utrophin fused to GFP to probe actin (Burkel et al., 2007) or the myosin regulatory light chain spaghetti squash (sqh)-GFP to visualize activated myosin (Rogers et al., 2004). In control cells, actin and myosin show similar dynamic association with the bleb cortex. Although they were not associated with bleb membranes when blebs expand, they were recruited to the bleb cortex when they retract in a manner that parallels the recruitment of Moe (Figs. 2 C and S1 and Video 4). In cells depleted of Moe or Slik, actin and myosin were no longer recruited uniformly to the bleb cortex (Figs. 2 C and S1 and Video 5). Secondary blebs then formed, ultimately leading to multiple blebs on blebs and triggering the abnormal mitotic cytoplasmic bulges observed in fixed samples (Carreno et al., 2008).

Together, these results show that cell division involves a precise regulation of Moe activity to control polar relaxation via mitotic blebbing. Defective Moe cortical recruitment promotes abnormal, unregulated polar blebs and thus, excessive polar relaxation. On the contrary, overactivation of Moe at the polar cortex impairs anaphase relaxation and cell elongation, leading to cytokinesis failure.

Having shown that the anisotropic distribution of Moe in anaphase contributes to proper cell division, we then sought to identify regulators of Moe activity. Although Moe activation relies on the activity of Slik kinase (Figs. 3, A and B; and S2 A; Hipfner et al., 2004; Carreno et al., 2008; Kunda et al., 2008), treatment with calyculin A (a PP1 and PP2A phosphatase inhibitor) increased Moe phosphorylation (Fig. 3 A), revealing the existence of a counteracting phosphatase activity.

To identify the phosphatase that inactivates Moe, we individually knocked down each of the Drosophila PP1 and PP2A catalytic subunits (Chen et al., 2007) and examined levels of P-Moe by Western blotting. We found that the depletion of Pp1-87B increased Moe phosphorylation (Fig. 3 A). To further identify at which stage of the cell cycle Pp1-87B inactivates Moe, we assessed P-Moe levels by immunofluorescence. We observed that Pp1-87B depletion promoted a threefold increase in Moe phosphorylation at the cortex in interphase (Figs. 3 B and S2 A). Nevertheless, Pp1-87B depletion did not increase P-Moe levels in metaphase, suggesting that Pp1-87B acts to down-regulate Moe at the cortex in interphase. Accordingly, time-lapse analyses showed that after cell division, Moe-GFP was abnormally associated with the cortex of interphase Pp1-87B–depleted cells (Fig. 3 C). In addition, Moe-GFP stayed also aberrantly associated with the polar cortex during anaphase elongation in Pp1-87B–depleted cells, similar to the distribution of the phosphomimetic Moe-TD mutant form (Fig. 1 D). As observed in Moe-TD cells, Pp1-87B depletion triggered frequent cytokinesis failure (21%, n = 554), further supporting that Moe down-regulation at the polar cortex is required to achieve proper cell division.

To evaluate how Pp1-87B controls Moe activity during the cell cycle, we examined Pp1-87B distribution using a functional GFP fusion (Fig. S2 C). Although mostly cytoplasmic in interphase, GFP–Pp1-87B was redistributed to the mitotic spindle upon mitosis entry (Figs. 3 D and S2 D). Similarly to Pp1-γ, its human orthologue (Trinkle-Mulcahy et al., 2006), GFP–Pp1-87B accumulated on chromosomes, close to the polar cortex in anaphase (Figs. 3 D and S2 D); then, it concentrated to the intercellular bridge where Moe is normally inactivated before abscission (Fig. 1 A). Therefore, dynamics of Pp1-87B distribution inversely correlates with sites of Moe activity. In contrast, Slik associated with the whole-cell cortex in interphase and metaphase and accumulated at the cleavage furrow after the anaphase onset (Fig. 3 D). These data show that a balance of Slik kinase and Pp1-87B phosphatase activities coordinates Moe activation with the cell cycle, restricting high levels of active cortical Moe to early mitosis.

Despite alteration of Moe cortical levels, neither Slik nor Pp1-87B depletion completely abolished the equatorial enrichment of Moe-GFP when cells progress through anaphase (Figs. 2 A and 3 C). Furthermore, the reciprocal phospho mutant forms of Moe, Moe-TD-GFP (Fig. 1 D) and Moe-TA-GFP (Fig. S3), remained enriched at the equatorial cortex in anaphase. These findings suggest that a phosphorylation-independent mechanism regulates the cortical distribution of Moe after the anaphase onset. Binding of the FERM domain to PI(4,5)P₂ plays a key role in Moe function in vivo (Roch et al., 2010), opening the possibility that PI(4,5)P₂ could contribute to the regulation of Moe activity/distribution during cell division.

As a first test of this hypothesis, we examined the distribution of the Moe FERM domain fused to GFP (FERM^Moe-GFP) in live cells. Although FERM^Moe-GFP isotropically associated with the cortex from interphase to metaphase, it was then enriched at the equatorial cortex in anaphase B and telophase (Fig. 4 A). Reciprocally, a full-length Moe mutant (Moe-KN-GFP) unable to bind to PI(4,5)P₂ (Roch et al., 2010) was not associated with the cortex throughout mitosis (Fig. 4 A). Therefore, PI(4,5)P₂ binding is required for Moe enrichment at the anaphase equator.

We then examined the localization of PI(4,5)P₂ in a S2 cell line that stably expresses GFP-Tubby, a specific probe for this phosphoinositide (Szentpetery et al., 2009; Ben El Kadhi et al., 2011). As observed for Moe-GFP (and FERM^Moe-GFP), GFP-Tubby cleared from polar regions and concentrated at the equator when cells started to elongate in anaphase (Fig. 4, B and C). These results support the idea that PI(4,5)P₂ contributes to the local control of Moe activation during cell division.

To directly test this conclusion, we used a system that rapidly reduces PI(4,5)P₂ cortical levels (Varnai et al., 2006), through a rapamycin-inducible recruitment of a type IV phosphoinositide 5-phosphatase domain to the plasma membrane (Fig. 4 D). In this assay, rapamycin induces the heterodimerization between a membrane-targeted FKBP-rapamycin–binding domain fragment of mammalian target of rapamycin and a 5-phosphatase domain fused to FKBP12 (Varnai et al., 2006), individually expressed from two constructs, hereafter referred to as rapamycin constructs (RCs). Although addition of rapamycin to control cells did not modify the localization of GFP-Tubby (Figs. 4 E and S4 A), rapamycin promoted its rapid translocation from the membrane to the cytosol in RC cells (Figs. 4 E and S4 B [for quantification] and Video 6). These results validate the efficiency of the RC system in S2 cells and the specificity of the GFP-Tubby probe for PI(4,5)P₂. We then explored whether reduction of PI(4,5)P₂ affects the distribution of wild-type or TD Moe-GFP. Moe-TD displayed a strong cortical association in interphase, which was lost after rapamycin treatment in RC cells (Fig. 4 E). Furthermore, when rapamycin was added on RC cells in metaphase, Moe-GFP rapidly delocalized from the cortex to the cytoplasm (Figs. 4 E and S4 C [for quantification] and Video 7). Similarly to Moe dsRNA depletion, this rapamycin-inducible release of Moe from cortical membranes triggered excessive polar relaxation, with unregulated blebbing during cell elongation (Fig. 4 F and Video 8). Together, these results support that PI(4,5)P₂ is necessary to ensure mitotic cortical stability, at least in part, through regulation of Moe cortical distribution and activation.

To further investigate how PI(4,5)P₂ regulates mitotic cortical stability and Moe activation, we individually depleted each of the Drosophila kinases (10), phosphatases (16), and phospholipases (5) predicted to directly or indirectly control PI(4,5)P₂ metabolism (Figs. 5 A and S5), and we assessed the consequences of these depletions on the mitotic cortex by time-lapse microscopy. Among the 31 enzymes we have tested, depletion of three proteins induced high cortical defects, including excessive polar relaxation and unregulated blebs during anaphase (Figs. 5 B and S5). Two of them, Skittles and Pten directly produce PI(4,5)P₂ by phosphorylating PI(4)P and dephosphorylating PI(3,4,5)P₃, respectively. The third enzyme, encoded by CG10260 and predicted to produce PI(4)P from phosphatidylinositol, was required to a lesser extent for mitotic cortical stability (Fig. 5 B). To quantify the respective contribution of these enzymes to PI(4,5)P₂ production, we measured cellular phosphoinositides using radiolabeling assays (Payrastre, 2004). PI(4,5)P₂ pools were decreased by 15, 31, and 48% after depletion of CG10260, Pten, or Skittles, respectively (Fig. 5 C). Consistently, although the cortical association of GFP-Tubby was only slightly reduced by depletion of CG10260 (not depicted), the absence of Pten or Skittles further prevented the cortical recruitment of this biosensor, showing that these enzymes generate PI(4,5)P₂ at the cortex (Fig. 5 D).

Having shown that Skittles and Pten play major roles in PI(4,5)P₂ production in S2 cells, we first examined whether their localization could account for PI(4,5)P₂ enrichment at the equator in anaphase. Comparable with PI(4,5)P₂ dynamics, Pten-GFP and Skittles-GFP uniformly associated with the cortex in prometaphase. Both enzymes then progressively accumulated at the equator, as cells underwent anaphase elongation (Fig. 6, A and B). Because Pten metabolizes PI(3,4,5)P₃, we analyzed the distribution of this PI(4,5)P₂ precursor during cell division using the Grp1-pleckstrin homology (PH)-GFP–specific PI(3,4,5)P₃ probe (Gray et al., 1999). In control cells, Grp1-PH-GFP was weakly associated with the cortex of mitotic cells, albeit enriched at the cleavage furrow in anaphase (Fig. 6 A). Pten depletion increased the cortical levels of Grp1-PH-GFP, with accumulation at the anaphase equator, where Pten normally localizes. This indicates that Pten locally dephosphorylates PI(3,4,5)P₃ to control PI(4,5)P₂ cortical levels during mitosis. Accordingly, depletion of Pten provoked the formation of abnormal long-lived polar blebs, a phenotype reminiscent of that seen after rapamycin-induced reduction of PI(4,5)P₂. The absence of Skittles further decreased cortical PI(4,5)P₂ levels during mitosis, as deduced from its impact on GFP-Tubby distribution (Fig. 6 B). The depletion of Skittles also enhanced unregulated polar blebbing in anaphase (Fig. 6 B), mimicking the defects observed in cells with reduced PI(4,5)P₂ levels by the RC experiment (Fig. 4 F) or those lacking Moe activity (Fig. 2, A and C).

We then tested whether Skittles or Pten was required for Moe phosphorylation during mitosis, using FACS analysis of cells costained for P-Moe and the mitotic marker phospho–histone H3. When compared with control conditions, Pten depletion caused a slight decrease in P-Moe levels (27.7% of mitotic cells negative for P-Moe staining vs. 17% in controls). These defects were enhanced when Skittles was depleted, with 81.7% of mitotic cells displaying reduced levels of Moe phosphorylation (Fig. 6 C).

Accordingly, Skittles depletion impaired the association of Moe-GFP with the mitotic cortex (compare Fig. 6 D and Video 9 with Fig. 1 A). This lack of Moe recruitment at the cortex of polar blebs correlated with defects in actin organization (Fig. 6 E and Video 10) and triggered the “blebs on blebs” phenotype and unregulated bleb growth (5.57 ± 1.58–µm length, n = 26 vs. 3.25 ± 0.44–µm length, n = 40 in controls; P = 1.5 × 10⁻⁸) observed after Moe or Slik depletion (Fig. 2, A and C).

These results show that Skittles and Pten contribute to enrich PI(4,5)P₂ at the equator, when dividing cells progress through anaphase. Thus, Skittles and Pten spatially mediate Moe cortical activation to promote efficient bleb dynamics necessary to control anaphase cell elongation.

These findings unravel how, by integrating two regulatory networks, Moe activity provides a spatiotemporal framework to control cell shape transformations during division (Fig. 7). We find that the increase in cortical rigidity that drives cell shape remodeling at the interphase/mitosis transition involves a Pp1-87B/Slik molecular switch that timely regulates Moe phosphorylation. We further identify PI(4,5)P₂ as a spatial cue that controls Moe distribution at the cortex. This latter aspect coordinates the spatial balance in cortical stiffness/contractility that is required for anaphase cell elongation and cytokinesis. We propose that the concerted action of these two regulatory networks ensures the proper series of mitotic cell shape transformations required for the fidelity of cell division (Fig. 7).

A global increase in cortical actomyosin forces generate cell rounding at mitosis entry (Maddox and Burridge, 2003; Stewart et al., 2011). These forces are transmitted to the plasma membrane through the activation of ERM proteins (Carreno et al., 2008; Kunda et al., 2008; Luxenburg et al., 2011). At mitosis exit, both cortical contractions and ERM activity must be down-regulated to allow cells to go back to their interphase shape. In Drosophila cells, the Slik kinase was shown to activate Moe at mitosis entry (Carreno et al., 2008; Kunda et al., 2008). Here, we identify the Pp1-87B phosphatase as essential for Moe inactivation after cytokinesis and in interphase.

Although Slik homogenously associates with the cell cortex in both interphase and early mitosis, Pp1-87B is cytoplasmic in interphase and relocalizes to the spindle in pro/metaphase (Fig. 7 B). An attractive model would be that together with a “constitutive” cortical association of the Slik activator in interphase and pro/metaphase, intracellular redistribution of the Pp1-87B inhibitor represents an efficient way to restrict high levels of Moe phosphorylation to mitosis entry. During anaphase, Pp1-87B concentrates near the chromosomes migrating toward the polar cortex, whereas Slik accumulates at the cleavage furrow. In this model, redistribution of both Pp1-87B and Slik after the anaphase onset contributes to enrich Moe at the equator and to decrease it at poles. Finally, relocalization of Pp1-87B in the cytoplasm after cytokinesis would contribute to relax the cortex for the next interphase by maintaining low Moe activity (Kunda et al., 2008; Théry and Bornens, 2008). A growing number of evidence supports that Pp1 phosphatases play important roles in the temporal control of cell division (De Wulf et al., 2009). Pp1-87B being required for mitotic spindle morphogenesis (Axton et al., 1990; Chen et al., 2007), this phosphatase could contribute to synchronize cell shape control operated through Moe regulation to chromosome segregation. Although additional investigations will be required to unravel how the activity and distribution of Pp1-87B and Slik are regulated, these results indicate that the Slik/Pp1-87B switch represents an important control of Moe activity during the cell cycle.

Our results show that local levels of PI(4,5)P₂ provide an additional mechanism to regulate Moe function at the cortex of dividing cells. Several studies have established a role of PI(4,5)P₂ in the localization of ERM proteins in polarized processes of differentiated cells (Fievet et al., 2004; Hao et al., 2009; Roch et al., 2010). Here, we provide evidence that during mitosis, PI(4,5)P₂-rich membrane domains act as a spatial cue that regulates both Moe distribution and activation at the cortex.

The distribution of PI(4,5)P₂ at the plasma membrane is tightly regulated during mitosis. As in mammalian cells (Emoto et al., 2005; Field et al., 2005), we find that PI(4,5)P₂ is actively enriched at the equator of anaphase Drosophila S2 cells, suggesting that equatorial accumulation of PI(4,5)P₂ is a feature shared by most animal cells. Although a previous study did not detect PI(4,5)P₂ enrichment at the cleavage furrow of Drosophila spermatocytes (Wong et al., 2005), whether this is caused by an intrinsic difference between mitosis and meiosis or by experimental limitations in vivo remains to be established. However, how this dynamic localization is regulated remained unknown. Here, we show that the equatorial enrichment of PI(4,5)P₂ relies, at least in part, on the enzymatic activity of Skittles and Pten. During cytokinesis, the equatorial accumulation of PI(4,5)P₂ plays a role in cleavage furrow formation and ingression, through controlling the activity and/or recruitment of several components of the contractile ring (Janetopoulos and Devreotes, 2006; Ben El Kadhi et al., 2011). PI(4,5)P₂ hydrolysis is also necessary for maintaining cleavage furrow stability and efficient cytokinesis (Emoto et al., 2005; Field et al., 2005; Wong et al., 2005). Our findings extend the functional repertoire of PI(4,5)P₂ during mitosis to the control of local properties of the mitotic cortex, which are required for polar relaxation and cell elongation. Through functional screenings, we identify novel regulators of cell division among the entire set of enzymes implicated in phosphoinositide biosynthesis. Two main pathways regulate PI(4,5)P₂ levels in mitotic cells, and their alterations provoke similar cortical disorganization. The first pathway involves the Pten tumor suppressor, a PI(3,4,5)P₃ 3-phosphatase. Pten was shown to accumulate at the septum of dividing yeast cells (Mitra et al., 2004), as well as at the cleavage furrow in Dictyostelium discoideum (Janetopoulos et al., 2005). Our results of living Drosophila cells show a progressive delocalization of Pten from the polar cortex to the equator after anaphase onset (Fig. 7 B), suggesting that Pten dynamics rely on mechanisms conserved throughout evolution. Furthermore, depletion of Pten leads to a significant enrichment of PI(3,4,5)P₃ at the cortex, especially at the cleavage furrow. These results show that Pten uses PI(3,4,5)P₃ to spatially control PI(4,5)P₂ levels at the mitotic cortex.

The second pathway relies on Skittles, a PI(4)P 5-kinase that plays a major role in regulating the levels and localization of PI(4,5)P₂ during mitosis. Skittles switches from an isotropic cortical distribution in pro/metaphase to equatorial enrichment after the anaphase onset (Fig. 7 B). Depletion of Skittles results in a phenotype similar to the mitotic cortical defects observed after inducible PI(4,5)P₂ hydrolysis. We also find that CG10260, a phosphoinositide 4-kinase, contributes to the organization of the mitotic cortex. Genetics screens have identified a role for phosphoinositide 4-kinases in the division of budding and fission yeast (Audhya et al., 2000; Hama et al., 2000; Desautels et al., 2001) as well as for cytokinesis of male spermatocytes in flies (Brill et al., 2000). CG10260 is involved in PI(4)P synthesis, the major substrate of Skittles to produce PI(4,5)P₂. Together, these data show that Skittles acts as a key regulator of PI(4,5)P₂ levels and Moe activation at the mitotic cortex. Interestingly, Skittles is required for Moe activation in Drosophila oocytes (Gervais et al., 2008), suggesting that this enzyme plays a broad role in the regulation of ERM proteins.

An important question is how Skittles and Pten are enriched at the equator in anaphase. It has been reported that activated RhoA stimulates a PI(4)P 5-kinase activity and promotes PI(4,5)P₂ synthesis in mammalian cells (Chong et al., 1994). During anaphase, activated RhoA localizes at the equatorial cortex (Yoshida et al., 2009), where it could recruit and/or activates Skittles to promote PI(4,5)P₂ production. This anisotropy in PI(4,5)P₂ distribution might be in turn reinforced by the localized activity of Pten, whose membrane association is itself dependent on PI(4,5)P₂ (unpublished data; Campbell et al., 2003; Rahdar et al., 2009). Together, the activity of Skittles and Pten could therefore provide a feed-forward regulatory loop of local PI(4,5)P₂ levels at the cortex of dividing cells.

The metaphase/anaphase transition is characterized by a break in cortical symmetry, with concomitant relaxation of the polar cortex and contraction of the equator. We find that the anisotropic distribution of Moe participates in coordinating this differential in cortical tension. Overactivation of Moe impairs cell elongation and causes cytokinesis failure, suggesting that the polar cortex is too rigid for cell division. Accumulation of F-actin at the cleavage furrow can be attributed, at least in part, to a cortical flow of F-actin filaments from polar regions to the equator (Chen et al., 2008). Overactivation of Moe at the poles could block this actin cortical flow, through an excessive bridging of the actin cytoskeleton with the plasma membrane, leading to an abnormal stiffness of the polar cortex. Therefore, redistribution of activated Moe from the polar cortex to the equator participates in polar relaxation, anaphase cell elongation, and cytokinesis fidelity.

Contraction of the equatorial actomyosin ring increases the cytoplasmic pressure exerted on the plasma membrane. Relaxation of the polar cortex is thus required to dissipate this extra pressure by increasing the cellular volume, a process that was proposed to involve short-lived polar blebs (Prothero and Spencer, 1968; Brugués et al., 2010; Sedzinski et al., 2011). These polar blebs were recently found to play important roles during cell division. Perturbation of their dynamics triggers anaphase spindle rocking (Rankin and Wordeman, 2010) and destabilization of cleavage furrow positioning (Sedzinski et al., 2011). Although recent studies have addressed how cortical blebs are regulated in interphase (Charras et al., 2005, 2006), our understanding of the signalization that controls dynamics of cortical blebs in mitosis has poorly progressed since pioneering studies. Our results show that a transient recruitment of Moe at the mitotic bleb membrane is required for efficient polar bleb retraction, as are the functions of the Moe positive regulators Slik, Skittles, and Pten. Active Moe contributes to cortical bleb organization because alteration of Moe function (or regulation) disrupts actin organization and efficient bleb retraction. This leads to disorganization of the mitotic cortex, characterized by giant blebs that continue growing in an unregulated manner. Therefore, although a global decrease in Moe activity at the polar cortex contributes to cell elongation and cytokinesis, transient and local association of Moe at the rim of polar blebs is important for their retraction. If the binding of Moe to PI(4,5)P₂ is required at both the equator and bleb membrane, the influence of the Slik kinase on Moe activation appears different between these two regions of the anaphase cortex. Although Slik depletion abolishes Moe recruitment to polar blebs, remnants of cortical Moe are still visible at the equator, likely as a result of high PI(4,5)P₂ levels at the furrow.

Although these mechanisms synergistically contribute to the cortical contractility at the equator, they also allow cortical relaxation at the polar cortex through control of transient anaphase blebs. We propose that this dual mechanism of Moe regulation is exploited by animal cells to ensure proper cell division.

Pten cDNA obtained from V. Archambault (Institute for Research in Immunology and Cancer, Montréal, Québec, Canada; IP16020; Drosophila Genomics Resource Center) was amplified by PCR (Phusion; New England Biolabs, Inc.) and fused to GFP in the pAc5.1 vector (Invitrogen). RCs (PM-FRB-monomeric RFP [mRFP] and mRFP-FKBP-5-phosphatase domain) obtained from T. Balla (National Institutes of Heath, Bethesda, MD; Varnai et al., 2006) and GFP-Utrophin-CH obtained from W.M. Bement (University of Wisconsin, Madison, WI; Burkel et al., 2007) were subcloned in the pAc5.1 vector (Invitrogen). pUAS-Skittles-GFP was obtained from A. Guichet (Institut Jacques Monod, Unité Mixte de Recherche 7592, Paris, France; Gervais et al., 2008), pUAS-Slik-GFP was obtained from D. Hipfner (Institut de Recherches Cliniques de Montréal, Montréal, Québec, Canada; Hipfner et al., 2004), and pUAS-GFP-Tubby and pTubulin-Grp1-PH-GFP were obtained from A.A. Kiger (University of California, San Diego, San Diego, CA). Other DNA constructs were previously described in Carreno et al. (2008).

dsRNA primers used in this study were as follows: Skittles_1 forward, 5′-TAATACGACT­CACTATAGGGA­GAACTGATGAG­CACCATTAAGCG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAGAATGTTTC­AGTGGAAATGTGG-3′; Skittles_2 forward, 5′-TAATACGACT­CACTATAGGGA­GACCCCCTCCA­AGCGCAAAACC­TCG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAGCGTGGACG­TGTAGCTGGGCG-3′; Pten forward, 5′-TAATACGACT­CACTATAGGGA­GACAAACGCAA­ACAGCCTAATAGC-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GACCACAGTTG­AATTTTGGAGAGG-3′; Pten_1 forward, 5′-TAATACGACT­CACTATAGGGA­GATATCCAGCA­CCGGATAAACTA-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GACAAGGTTTT­CAGTCTATCTGG-3′; Pten_2 forward, 5′-TAATACGACT­CACTATAGGGA­GACATCAATAC­CAGTTTCCGGCG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GATCGAAGTAT­TTTTTGAATTGG-3′; CG10260_1 forward, 5′-TAATACGACT­CACTATAGGGA­GACGATTCGGT­GAACGAGCTGCG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GATCGGCGAAG­TCCTTCAAGCGC-3′; CG10260_2 forward, 5′-TAATACGACT­CACTATAGGGA­GAGGTCTGTCG­CTGCTGCAGGGC-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GACCCGGCACT­ATAAGATCCGGC-3′; Pp1-87B forward, 5′-TAATACGACT­CACTATAGGGA­GATCGGAAATT­AAAGATTTCGAGG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAAAGATCTTC­TCGTCGACAATGG-3′; Pp1-87B 5′UTR forward, 5′-TAATACGACT­CACTATAGGGA­GACGGCAGTGT­GGCAACATCAG­CTAGAAGC-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAGTTTGCGTG­CGAAAGTGTGG­ATCTGG-3′; Pp1-87B 3′UTR forward, 5′-TAATACGACT­CACTATAGGGA­GATATCACACA­ACTGCAGCACC­ACG-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GACTTGCTCTA­TACTCTTAAGGCC-3′; CG10260 forward, 5′-TAATACGACT­CACTATAGGGA­GAGGAGTATCC­CAATCAGATTCCC-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAATCAGTATG­GGCTTGCTGACC-3′; and Slik forward, 5′-TAATACGACT­CACTATAGGGA­GAACTTTGTCA­AAAAGGGTAAGGC-3′, and reverse, 5′-TAATACGACT­CACTATAGGGA­GAACCTCACTT­TCATCCAGTTTGC-3′.

S2 Drosophila cells were cultured under standard conditions. We generated stable cell lines expressing wild-type Moe-GFP, Moe-TA-GFP, Moe-TD-GFP, FERM^Moe-GFP, Moe-KN-GFP, GFP-Utrophin-CH, GFP-Tubby, and Grp1-PH-GFP in combination with Tubulin-mCherry to identify mitotic stages. We used FACS sorting (FACSAria; BD) to select cell lines expressing moderate levels of the different GFP fusions. The distribution of Pp1-87B–GFP, Skittles-GFP, Pten-GFP, and Slik-GFP was analyzed after transient transfection (FuGENE HD; Roche) of the corresponding constructs. Cells stably expressing Moe-GFP or GFP-Tubby were transfected by the RC constructs 36 h before rapamycin addition (final concentration of 100 nM).

Phosphatase subunits and phosphoinositide enzymes were identified by annotation of the fly genome complemented by a search for individual orthologues of mammalian counterparts. Production of dsRNA was previously described in Carreno et al. (2008). dsRNA were synthesized using DNA templates from the Drosophila dsRNA library (Thermo Fisher Scientific). Each positive hit was confirmed by the use of at least one additional dsRNA, targeting an independent region of the gene of interest to avoid off targets.

Cells were cultured for 4 h on glass coverslips, fixed in 4% formaldehyde (or 10% TCA for P-Moe), and processed for immunostaining (Carreno et al., 2008). We used anti–P-Moe (Carreno et al., 2008) at 1:100, anti–α-tubulin coupled to FITC (F2168; Sigma-Aldrich) at 1:50, and Texas red goat anti–rabbit (Invitrogen) at 1:200. Texas red phalloidin (Invitrogen) was used at 1:100 for F-actin staining. Fixed imaging was performed at room temperature using mounting medium (Vectashield; Vector Laboratories) on a microscope (DeltaVision; Applied Precision) using softWoRx software (Applied Precision) equipped with a camera (CoolSNAP HQ2; Photometrics) at 1 × 1 binning and 60× Plan Apochromat (1.42 NA) objective.

For live imaging, cell lines were cultivated in glass-bottom plates and multichannel images were acquired at 27°C with a microscope (DeltaVision) using softWoRx software equipped with a camera (CoolSNAP HQ2) at 2 × 2 binning and 60× Plan Apochromat (1.42 NA) objective or a wide-field microscope (DMIRE2; Leica) using Metamorph software (Molecular Devices) equipped with a camera (CoolSNAP HQ2) at 2 × 2 binning and 63× Plan Apochromat (1.32 NA). Images were acquired every 15 or 7 min for broad analyses of cell division and every 7 or 20 s for fast tracking of bleb dynamics. Images were analyzed using Metamorph, softWoRx Explorer, and ImageJ (National Institutes of Health) packages. Images were deconvoluted using Scientific Volume Imaging or softWoRx and processed with Photoshop (Adobe).

S2 cells labeled with [³²P]phosphate were processed for lipid extraction and separation by TLC using CHCl₃/CH₃COCH₃/CH₃OH/CH₃COOH/H₂O (80:30:26:24:14 vol/vol) as a solvent. The TLC plate (silica gel 60 with concentrating zone; Merck) had been treated with 1% potassium-oxalate in 2 mM EDTA/methanol (vol/vol) for 15 min and activated at 100°C for 1 h. Spots corresponding to phosphoinositol phosphate and PI(4,5)P₂ were visualized with a PhosphoImager scanner (GE Healthcare), identified by comparison with the migration of authentic standards, scraped off separately, and deacylated by adding 1 ml methylamine reagent composed of 26.8% (vol/vol) of 40% methylamine, 45.7% (vol/vol) methanol, 11.4% (vol/vol) n-butanol, and 16% (vol/vol) H₂O to the silica powder. After incubation at 53°C for 50 min, the methylamine reagent was completely evaporated under a nitrogen stream at 37°C. The samples were resuspended in 1.2 ml H₂O and separated by HPLC using a column (Partisphere 5 SAX; Whatman) 4.6 × 125 mm with guard cartridge anion exchanger units (Whatman) and a gradient of 1 M (NH₄)2HPO₄, pH 3.8, and bidistilled water as previously described (Payrastre, 2004).

For each condition (controls, dsRNA, or 50 nM Calyculin A–treated cells), 10⁶ cells were washed with 1 ml PBS NaF/Na₃VO₄/β-glycerophosphate (50 mM/1 mM/40 mM). After centrifugation (5 min at 4°C), pellets were resuspended in 100 µl Laemmli buffer at 95°C supplemented with antiphosphatases and antiprotease cocktail (Roche). Membranes were blocked with TBS Tween/BSA/NaF/Na₃VO₄/β-glycerophosphate (0.1%/2%/50 mM/5 mM/40 mM) and then incubated with an antibody against Moe (1:25,000) or P-Moe (1:200).

Cells were fixed for 4 h using ethanol 75% and permeabilized by 0.02% saponin. Cells were labeled using anti–P-Moe (1:100), antiphospho–histone H3 (Ser10) 647 conjugate (1:100; Cell Signaling Technology), and Alexa Fluor 488 goat anti–rabbit (1:1,000; Invitrogen) antibodies. Cells were then analyzed by FACS (LSR II; BD).

Fig. S1 shows localization of sqh-GFP during mitotic blebbing and quantification of mitotic bleb dynamics in control and Slik dsRNA conditions. Fig. S2 shows characterization of Pp1-87B function during cell division. Fig. S3 shows localization of Moe-TA-GFP during cell division. Fig. S4 shows quantification of the RC experiments shown in Fig. 4. Fig. S5 shows the results of the functional screen for enzymes implicated in phosphoinositide metabolism shown in Fig. 5. Video 1 shows the dynamic localization of Moe-GFP during division of a control cell over a 25-min period with a 1-min interval (as shown in Fig. 1 B). Video 2 shows the dynamic localization of Moe-GFP during division of a control cell over a 675-s period with a 7-s interval (as shown in Fig. 2 A). Video 3 shows the dynamic localization of Moe-GFP during division of a Slik-depleted cell over a 500-s period with a 20-s interval (as shown in Fig. 2 A). Video 4 shows the dynamic localization of Utrophin-CH-GFP (as an F-actin probe) during division of a control cell over a 337.5-s period with a 2.5-s interval (as shown in Fig. 2 C). Video 5 shows the dynamic localization of Utrophin-CH-GFP (as an F-actin probe) during division of a Slik-depleted cell over a 225-s period with a 2.5-s interval (as shown in Fig. 2 C). Video 6 shows the dynamic localization of GFP-Tubby (as a PI(4,5)P₂ probe) in interphase RC cells before and after addition of rapamycin over a 450-s period with a 15-s interval (as shown in Fig. 4 E). Video 7 shows the dynamic localization of Moe-GFP of a metaphase RC cell before and after addition of rapamycin over a 300-s period with a 10-s interval (as shown in Fig. 4 E). Video 8 shows the dynamic localization of Moe-GFP of an ana/telophase RC cell 435 s after addition of rapamycin over a 603-s period with a 9-s interval (as shown in Fig. 4 F). Video 9 shows the dynamic localization of Moe-GFP during division of a Skittles-depleted cell over a 90-min period with a 10-min interval (as shown in Fig. 6 D). Video 10 shows the dynamic localization of Utrophin-CH-GFP (as an F-actin probe) during division of a Skittles-depleted cell over a 217.5-s period with a 7.5-s interval (as shown in Fig. 6 E).

We are grateful to A.A. Kiger, G. Hickson, T. Balla, W.M. Bement, A. Guichet, V. Archambault, D.R. Hipfner, and P. Roux for antibodies, cell lines, DNA clones, and reagents. We thank C. Charbonneau, P. Maddox, and A.S. Maddox (Institut de Recherche en Immunologie et en Cancérologie) and A. Leru, B. Ronsin, and P. Cochard (Centre de Biologie du Développement and Plate-Forme Infrastructures Biologie Santé et Agronomie d’Imagerie Cellulaire de Toulouse) for help with microscopy, Danièle Gagné for her help in cell sorting, other laboratory members for their continuous help, and A.S. Maddox, S. Tournier, J.C. Labbé, A. Merdes, and Y. Gachet for critical reading and comments on the manuscript.

This work was supported by the Canadian Institutes for Health Research (grant MOP-89877) to S. Carreno and from the Association pour la Recherche contre le Cancer (grants 3832 and 1111), Fondation pour la Recherche Medicale (Equipe 2005), and Agence Nationale de la Recherche Blanc to F. Payre (Netoshape) and to B. Payrastre (Phosphoinopatho and Myotubularinopathies). C. Roubinet was supported by fellowships from Ministère de la Recherche et de l’Education Supérieure, fondation Lavoisier, European Molecular Biology Organization, and Association pour la Recherche contre le Cancer. J.F. Dorn was supported by he Canadian Cancer Society (grants 018450 and 019162) and the Terry Fox Foundation (awarded to P. Maddox and A.S. Maddox, respectively) and is a fellow of the Swiss National Science Foundation. S. Carreno is a Fonds de la Recherche en Santé du Québec Junior fellow. Institut de Recherche en Immunologie et en Cancérologie is supported in part by the Canadian Center of Excellence in Commercialization and Research, the Canada Foundation for Innovation, and the Fonds de Recherche en Santé du Québec.