Activated ezrin controls MISP levels to ensure correct NuMA polarization and spindle orientation

Correct positioning of the mitotic spindle axis within the cell is a fundamental process in development and stem cell division (Gönczy, 2002; Siller and Doe, 2009). In symmetrically dividing cells, precise spindle orientation and positioning ensures equal distribution of cellular components. In contrast, in asymmetrically dividing cells, accurate orientation and placement of the mitotic spindle away from the center of the cell results in cell fate diversity (Horvitz and Herskowitz, 1992; Gönczy, 2002; Ahringer, 2003; Grill and Hyman, 2005). In most epithelia, cells divide symmetrically and orient their mitotic spindle parallel to the apical–basal surface, ensuring expansion of the epithelial sheet with side-by-side growing daughter cells (Fleming et al., 2007). Any misregulation in spindle orientation can result in disorganized tissue morphology due to cell multi-layering, which could be associated with the earliest cancer developments (McCaffrey and Macara, 2011; Pease and Tirnauer, 2011).

The precise spindle position and orientation in the cell is achieved by signaling pathways generating pulling and pushing forces on the spindle, both externally or internally (Gönczy, 2002; Grill and Hyman, 2005; Théry et al., 2005; Fink et al., 2011). The longest established player in spindle orientation is the conserved ternary complex, composed of Gα_i, Leu-Gly-Asn repeat-enriched protein (LGN, also known as GPSM2) and nuclear mitotic apparatus (NuMA, also known as NUMA1) (Du et al., 2001; Du and Macara, 2004). Whereas Gα_i localizes at the whole cortex, LGN and NuMA are found in a crescent-shaped localization at the cortex, facing one or both spindle poles. Thereby, the crescent localization of LGN and NuMA determines the position of force concentration. Through direct binding, NuMA recruits the force generator minus-end-directed motor protein complex of dynein and dynactin to the cortex, where it generates pulling forces by binding to astral microtubules (MTs) (Toyoshima and Nishida, 2007; di Pietro et al., 2016). Through the polarized localization of LGN, NuMA and dynein–dynactin, the forces are generated at opposite sides of the cell ensuring optimal spindle orientation/positioning and equal cell division. In epithelial cells, LGN and NuMA specifically localize to the lateral sides, allowing planar cell division (Zheng et al., 2010; Peyre et al., 2011). Besides a few recently identified proteins that regulate polarized distribution of LGN and NuMA, like Afadin and aPKC, the regulatory mechanism that restricts LGN and NuMA localization still has to be identified (Hao et al., 2010; Zheng et al., 2010; Carminati et al., 2016).

Recently published data reveal that the ezrin/radixin/moesin (ERM) protein family acts as an additional factor in mitotic spindle orientation and positioning (Machicoane et al., 2014). More precisely, activated cortical ERM proteins in mitosis are involved in promoting polarized, cortical association of LGN and NuMA. However, the direct molecular link between ERM and the ternary complex is not known. The ERM proteins, namely ezrin, radixin and moesin, are evolutionarily highly conserved proteins, and are expressed differently in various tissues (Fehon et al., 2010). They exert crucial functions in cell migration, as well as cell invasion, by directly linking the actin cytoskeleton to the plasma membrane (Algrain et al., 1993; Clucas and Valderrama, 2014). ERM proteins exist in two conformational states, an inactive, closed conformation with the FERM and C-terminal tail domain (C-ERMAD) forming an intramolecular interaction, and an active, open confirmation where the two domains are dissociated (Bretscher et al., 1995, 1997; Gary and Bretscher, 1995). The active form is produced by phosphatidylinositol-4,5-bisphosphate (PIP₂) binding to the FERM domain and a subsequent membrane-dependent phosphorylation of the highly conserved C-terminally located threonine residue at position 567, 564 and 558 for ezrin, radixin and moesin, respectively (Nakamura et al., 1995; Niggli et al., 1995; Hirao et al., 1996; Matsui et al., 1998; Simons et al., 1998; Bretscher et al., 2002; Fievet et al., 2004; Pelaseyed et al., 2017). This phosphorylation in epithelial cells is mainly catalyzed by lymphocyte-oriented kinase (LOK, also known as STK10) and Ste20-like kinase (SLK) (hereafter SLK/LOK) (Kunda et al., 2008; Machicoane et al., 2014).

The recently identified unstructured protein mitotic interactor and substrate of Plk1 (MISP) is strongly associated with actin and is highly phosphorylated during mitosis by Cdk1 and Plk1 (Maier et al., 2013; Zhu et al., 2013; Kumeta et al., 2014). MISP localizes at the cell cortex throughout the cell cycle and contributes to proper cortical distribution of p150^glued (also known as DCTN1), astral MT stability and correct mitotic spindle placement (Zhu et al., 2013). However, the exact mechanism of how cortical proteins collectively contribute to the regulation of mitotic spindle orientation is not fully understood.

Here, we report that the ERM family member ezrin acts upstream of MISP in controlling mitotic spindle orientation and positioning. Our results indicate that MISP is a direct binding partner of ezrin in its open active state. We uncover that, in mitotic cells, SLK/LOK-activated ezrin is required to prevent excessive accumulation of MISP at the cell cortex. Aberrant MISP levels abolish cortical polarization of NuMA and impairs astral MT dynamics, which in turn results in spindle misorientation. Our data suggest that the protein levels of MISP at the cortex need to be tightly controlled to ensure NuMA polarization and proper spindle orientation. Thus, we provide novel mechanistic insights into how ezrin governs NuMA-regulated spindle orientation by monitoring cortical levels of the actin-binding protein MISP.

To understand the mechanism by which MISP regulates spindle orientation, we aimed at identifying MISP-interacting proteins in both asynchronous and mitotic cells. By using several approaches, including co-immunoprecipitations (co-IPs) and BioID (Roux et al., 2012), we identified the human ERM protein family, consisting of ezrin, radixin and moesin, as such interacting proteins (Fig. S1A).

We verified the interaction between MISP and the ERM proteins ezrin, radixin and moesin, respectively, after co-expression in HEK293T cells (Fig. 1A; Fig. S1B). Since ezrin is the most ubiquitous ERM protein in epithelial cells, we focused on the interplay between MISP and ezrin, and could observe that MISP also interacted with ezrin in endogenous co-IPs from HeLa cells (Fig. 1B). To find out whether MISP and ezrin directly bind to each other, we performed in vitro pulldown experiments with recombinant proteins. Surprisingly, we observed no interaction between MISP and wild-type ezrin (Fig. 1C; Fig. S1C). As reported previously, ezrin exists in two different conformations, a closed (inactive) and an open (active) form that is exerted by T567 phosphorylation and PIP₂ binding (Fehon et al., 2010; Pelaseyed et al., 2017). To determine whether the interaction between MISP and ezrin is dependent on phosphorylation, or whether a specific conformation of ezrin facilitates MISP binding, we analyzed the ability of MISP to bind to the ezrin mutants T567D (mimicking the phosphorylated form) and T567A (a non-phosphorylatable mutant). Additionally, we tested MISP binding to ezrin R579A, a mutant unable to self-interact and to bind to actin (Saleh et al., 2009). Interestingly, in contrast to what we found with wild-type ezrin, we identified a strong direct interaction between MISP and ezrin T567D and a moderate interaction between MISP and ezrin R579A (Fig. 1C; Fig. S1C). No interaction was observed between MISP and ezrin T567A. These data demonstrate that MISP and ezrin are direct interaction partners, that this interaction is independent of actin and that MISP binds to ezrin in its open conformation.

To analyze whether MISP and ezrin function together in mitosis, we performed colocalization studies via confocal immunofluorescence in both polarized and non-polarized cells. In comparison to what is seen in single-cell models, studies in polarized cells reflect a more physiological context by recapitulating many aspects and characteristics of epithelial cells as for example cell–cell contacts and spindle orientation (Debnath and Brugge, 2005; McCaffrey and Macara, 2011). For studies in polarized cells, we used Caco-2 BBe cells in an ‘on-top’ of Matrigel approach to create polarized spherical cysts (Ivanov et al., 2008; Lujan et al., 2016). It has been previously shown that these cysts are characterized by a central lumen surrounded by a monolayer of polarized cells with basolateral and apical sides (O'Brien et al., 2002; Debnath and Brugge, 2005). In cysts, MISP mainly localized at the apical internal cortex facing the central lumen and only weakly at the basolateral sides (Fig. 1D; Fig. S1D), very much like the previous reported actin and ezrin localization (Bretscher et al., 2002; Ivanov et al., 2008; Jaffe et al., 2008). Importantly, our studies show that MISP is strongly associated with actin in polarized cells (Fig. S1D), which is similar to what was described in non-polarized cells (Maier et al., 2013; Zhu et al., 2013; Kumeta et al., 2014). Interestingly, MISP and ezrin partially colocalize in polarized cells at the apical cortex (Fig. 1D). In addition, we confirmed that MISP and ezrin partially colocalize in mitotic HeLa cells (Fig. 1E; Fig. S1E). Collectively, our results show that MISP and ezrin partially colocalize in both polarized and non-polarized cells.

It was previously shown that loss of Drosophila moesin, the only encoded ERM protein in D. melanogaster, as well as simultaneous downregulation of all three ERM proteins in human cells, leads to spindle misorientation (Carreno et al., 2008; Kunda et al., 2008; Nakajima et al., 2013; Machicoane et al., 2014). To find out whether all three ERM proteins are equally involved in the regulation of spindle orientation, we used HeLa cells growing on fibronectin-coated coverslips. Cells were treated with two different sets of siRNAs directed against each single ERM protein or against all ERMs simultaneously, and the spindle angle in respect to the substratum was measured (Toyoshima and Nishida, 2007) (Fig. S2A). We observed that the siRNA-mediated downregulation of ezrin, but not downregulation of radixin and moesin, resulted in a misorientation of the mitotic spindle (Fig. S2B,C). To extend our studies to spindle positioning, we analyzed the position of the mitotic spindle within the cell by calculating the difference in the pole-to-cortex distance from each pole (Fig. S2A) (Kern et al., 2016). Similarly, we could measure a significant increase in the pole-to-cortex difference in ezrin-silenced cells (Fig. S2D). Both mitotic spindle orientation and position defects caused by ezrin loss could be rescued by providing an siRNA-resistant ezrin version simultaneously to ezrin_1 siRNA (Fig. S2C,D). Thus, our results suggest that ezrin is the only ERM in human cells guiding correct spindle orientation and positioning.

The advantage of studies in polarized cells is that they imitate many features of in vivo epithelial cells (Debnath and Brugge, 2005; McCaffrey and Macara, 2011). Especially, mitotic spindle orientation is a crucial process in polarized cell organization that is necessary to create properly arranged cell layers with ordered cell–cell contacts (Bergstralh and St Johnston, 2014). To assess the role of both MISP and ezrin in cyst organization, we used Caco-2 BBe cells treated with siRNA in the ‘on-top’ of Matrigel approach. Indeed, absence of MISP or ezrin resulted in an abnormal cystogenesis (Fig. 2A; Fig. S2E). In the control situation, 90% of the formed cysts display a single lumen (control siRNA, 90.1±1%, mean±s.d.). In contrast, ablation of MISP or ezrin led to a significant reduction in the number of cysts with single lumen (MISP_1 siRNA, 69.4±4.4%; MISP_2 siRNA, 66.3±2.7%; ezrin_1 siRNA, 65.8±4.5%; ezrin_2 siRNA, 59.1±5.4%) and to a simultaneous increase of cysts with multiple lumens (Fig. 2A). As formation of multiple lumens in cysts is often caused by spindle misorientation (Jaffe et al., 2008; Hao et al., 2010; Rodriguez-Fraticelli et al., 2010), and as both MISP and ezrin function in spindle orientation, we investigated whether the spindle is misoriented in cysts after loss of MISP or ezrin. In fact, the mitotic spindle was misoriented after siRNA-mediated downregulation of MISP, from 70.8±16.9° in control cells, to 59.7±21.3° (MISP_1 siRNA) and 55.1±22.1° (MISP_2 siRNA), respectively (Fig. 2B; Fig. S2E). In addition, siRNA-mediated downregulation of ezrin in Caco-2 BBe cysts led to a broad distribution of the mitotic spindle angle (ezrin_1 siRNA, 58.5±21.7°; ezrin_2 siRNA, 52.4±23.8°). These results indicate that MISP and ezrin are important for correct mitotic spindle orientation in both non-polarized and polarized cells, and, in the latter case, this ensures organized cystogenesis with a monolayer of cells surrounding a single lumen.

Since both MISP and ezrin are required for proper spindle orientation and positioning, we analyzed whether MISP and ezrin are acting in the same or in different pathways to regulate spindle orientation. We performed siRNA silencing experiments to study either the effect of MISP ablation on phosphorylated ERM (pERM) proteins, since cortical ERMs are highly phosphorylated in mitosis by the SLK kinase (Machicoane et al., 2014), or the loss of ERM proteins and its impact on MISP. To show this, we measured the cortical amount of MISP and pERM proteins in relation to their cytoplasmic pool using fluorescence microscopy (Fig. 3A; Fig. S3A–C). Interestingly, upon ezrin siRNA treatment alone, or upon the simultaneous loss of all ERMs, cortical MISP levels were significantly increased (Fig. 3A; Fig. S3B). In contrast, siRNA-mediated downregulation of radixin or moesin did not alter the cortical-to-cytoplasmic MISP ratio, indicating that ezrin function is required for correct MISP localization at the cortex (Fig. 3A, Fig. S3A). Vice versa, downregulation of MISP did not cause a significant change in cortical pERM levels (Fig. S3C) suggesting that MISP acts downstream of ezrin. The enhanced cortical MISP levels after ezrin loss could be rescued upon co-expression of a GFP–ezrin siRNA-resistant construct (Fig. 3B, Fig. S3D). To find out whether ezrin also regulates cortical MISP levels in polarized cells, we used Caco-2 BBe spherical cysts treated with ezrin siRNAs. After downregulation of ezrin, we observed a significant increase in apical MISP levels (Fig. 3C). Overall, we propose that the cortical levels of MISP are specifically regulated by the single ERM protein ezrin. However, an involvement of the other two ERM proteins (radixin and moesin) cannot be completely ruled out.

Since the phosphorylation of ezrin at T567 is important for its full activation, we asked whether activation of ezrin is required to regulate cortical MISP levels. Indeed, after siRNA-mediated downregulation of ezrin, only expression of the phosphomimicking GFP–ezrin T567D was able to rescue the enrichment in cortical MISP levels (Fig. 4A, Fig. S3D). In contrast, the non-phosphorylated GFP–ezrin T567A mutant could not reduce cortical MISP levels. Hence, we assume that the phosphorylation state of ezrin might be crucial for control of MISP.

Additionally, we tested whether a specific domain of ezrin is responsible for MISP regulation. We generated different fragments of ezrin comprising either the FERM domain, which directly binds to PIP₂ at the plasma membrane, or the C-ERMAD domain, including the highly conserved threonine 567 and the actin-binding domain at the very C-terminus (Fehon et al., 2010). Interestingly, we found that it was not the dominant-negative FERM domain (Allenspach et al., 2001; Shiue et al., 2005) but rather the C-ERMAD fragment of ezrin (ezrin 297–586) that was capable of reducing the elevated cortical-to-cytoplasmic MISP ratio (Fig. 4A). Furthermore, we observed that ezrin PIP-T567D (ezrin PIP has the four point mutations K253N and K254N, K262N and K263N) (Barrett et al., 2000), which is active, because it mimics the T567 phosphorylation, but is unable to bind to PIPs, is still able to reduce the cortical MISP levels (Fig. 4A), indicating that PIP₂ binding is not essential for the ability of ezrin to control MISP levels. Interaction studies in HeLa cells revealed that MISP is indeed binding to C-ERMAD of ezrin (Fig. S3E). Additionally, MISP interacted with the N-terminally located FERM domain even though this interaction does not seem to be required for ezrin-mediated control of MISP levels.

To further confirm that the activity state of ezrin monitors cortical MISP levels, we used a recently described ezrin inhibitor (NSC668394) (Bulut et al., 2012), which inhibits T567 phosphorylation as well as the actin binding of ezrin (Pore et al., 2015). In fact, upon treatment of cells with NSC668394 we observed an increase in the cortical-to-cytoplasmic MISP ratio (Fig. 4B). In epithelial cells, ezrin is mainly phosphorylated at T567 by SLK/LOK (Viswanatha et al., 2012; Machicoane et al., 2014). Inhibition of ezrin activation through T567 phosphorylation upon siRNA-mediated downregulation of SLK or LOK resulted in a significant increase of MISP levels at the cortex (Fig. 4C; Fig. S3F). This finding suggests that activation of ezrin is crucial to prevent aberrant accumulation of cortical MISP. To address the involvement of ezrin in the dynamics of MISP at the cell cortex, fluorescence recovery after photobleaching (FRAP) experiments were performed with GFP–MISP-expressing HeLa cells treated with control or ezrin siRNAs. While the halftime of GFP–MISP recovery at the cortex in control cells was τ_1/2=4.9±1.9 s, the halftime decreased significantly to τ_1/2=3.8±1 s upon loss of ezrin (Fig. S3G,H). Similarly, the mobile fraction of MISP after ezrin loss slightly increased, pointing to a higher MISP dynamics, presumably due to loss of the stabilizing factor ezrin (Fig. S3G,H). Taken together, our results indicate that activation of ezrin by SLK/LOK-dependent phosphorylation is critical for appropriate localization of MISP at the cortex.

Our data imply that MISP and ezrin are direct interaction partners (Fig. 1C), and, at the same time, ezrin depletion leads to an increase of MISP levels at the cortex. An explanation for these results could be that MISP and ezrin bind to each other, but simultaneously compete for actin-binding sites at the cortex. To address this hypothesis, we first tested the actin-binding behavior of MISP upon ezrin depletion, and after cells were treated with Latrunculin B to depolymerize the actin cytoskeleton. As shown in Fig. 4D and Fig. S3I, Latrunculin B disrupted the actin cytoskeleton and reduced MISP levels at the cortex both in ezrin siRNA-treated and in wild-type cells. We conclude that ezrin loss increases MISP binding to actin. To further confirm the hypothesis that MISP and ezrin compete for actin binding in vitro, we performed an actin co-sedimentation assay with a combination of actin and MISP, actin and ezrin, or all three proteins. Interestingly, the ability of ezrin to bind actin was reduced in the presence of MISP (Fig. 4E). Similarly, MISP bound less strongly to actin in the presence of ezrin (Fig. 4E). Overall, our results indicate that MISP and ezrin selectively interact, but at the same time seem to compete for actin-binding sites.

Our data imply that activated ezrin interacts with and regulates the localization of MISP at the cell cortex to ensure correct mitotic spindle orientation. Next, we aimed at analyzing the impact of elevated cortical MISP levels on mitotic spindle orientation. If ezrin forms a complex with MISP as prerequisite for spindle orientation, ectopic expression of MISP should lead to spindle misorientation due to unbalanced ezrin–MISP levels. We found that overexpression of MISP leads to increased cortical MISP levels (Fig. S4A) that consequently result in a significant increase in the mitotic spindle angle (Flag, 4.4±3.5°; Flag–MISP, 8.0±5.7°; Fig. 5A). We additionally analyzed the position of the spindle within the cell by measuring the pole-to-cortex distance. MISP overexpression induced a significant imbalance in spindle positioning, as indicated by an increased pole-to-cortex difference (Fig. 5B). Overexpression of MISP, however, had no impact on pERM levels (Fig. S4B). Based on these results, we propose that elevated MISP levels at the cortex are the cause for the observed spindle orientation defects. In order to challenge this hypothesis, we induced spindle orientation and positioning defects by siRNA-mediated downregulation of ezrin (40 nM) and simultaneously treated these cells with a small amount of MISP siRNA (10 nM). The collective siRNA treatment resulted in a cortical:cytoplasmic MISP ratio that was similar to that seen in untreated cells (Fig. 5C). Importantly, through the simultaneous downregulation of ezrin and MISP, we were able to reduce the defects in orientation and positioning of the mitotic spindle (Fig. 5D,E). Interestingly, partial ablation of MISP was also able to rescue single lumen formation in cysts and to correct spindle orientation in ezrin-depleted cysts (Fig. 5F,G). From these data, we conclude that MISP and ezrin function in the same pathway, and that ezrin acts upstream of MISP to ensure proper mitotic spindle orientation and positioning.

Spindle orientation defects caused by MISP depletion lead to a shortening of astral MTs (Zhu et al., 2013) and thereby to a loss of the spindle-to-cortex connection. Thus, we analyzed whether the overexpression of MISP has similar effects on astral MTs. Interestingly, we observed that increased cortical MISP levels do not result in shortening of astral MTs, as observed in response to MISP depletion but instead in their elongation and buckling (Fig. 6A). In accordance with this pronounced effect on astral MT length, we also noticed a modest increase in astral MT intensity (Fig. 6A). Similarly, loss of ezrin led to an increase in both the length of astral MTs and their brightness (Fig. S4C). To test the dynamic properties of astral MTs after MISP overexpression, we performed short-term live-cell imaging of Tomato–EB3 (EB3 is also known as MAPRE3) comets to visualize the growing ends (+Tips) of MTs (Sironi et al., 2011). In these experiments, the length of Tomato–EB3 comets displayed in kymographs represents the MT stability, whereas the slope represents the growth speed of MTs (Bieling et al., 2007; Sironi et al., 2011). Interestingly, the EB3-tracking experiments revealed that astral MTs have a higher stability in MISP-overexpressing cells (Fig. 6B). Their overall number and growth rate, however, remained unaltered (Fig. S4D).

Since we observed a misplacement of the mitotic spindle within the cell after MISP overexpression, it seems likely that the increased stability of astral MTs produces unbalanced forces that keep the spindle close to one side of the cell. To determine whether the elongated and stabilized astral MTs cause spindle misplacement, we treated cells with a low dose of nocodazole to suppress MT dynamics by causing depolymerization of astral MTs, and thereby preventing an unbalanced MT–cortex connection (Kern et al., 2016). Intriguingly, nocodazole treatment restored the spindle orientation and positioning defects after overexpression of MISP and after loss of ezrin (Fig. 6C; Fig. S4E). These results indicate that increased cortical MISP levels misorient and misposition the mitotic spindle because of unbalanced forces generated by a defective cortex-to-spindle association through astral MTs that are both elongated and stabilized.

Besides astral MTs, the ternary complex protein NuMA is a key player in mitotic spindle orientation. In prometaphase/metaphase cells, NuMA localizes to the spindle poles and polarizes at the cortex in a crescent-shaped localization (Du and Macara, 2004; Woodard et al., 2010; Kiyomitsu and Cheeseman, 2012; Kotak et al., 2012). Through direct binding to dynein–dynactin, NuMA links the astral MTs to the cortex, thus ensuring optimal mitotic spindle orientation and positioning (Gaglio et al., 1995; Merdes et al., 1996; Du et al., 2002). On the basis of unbalanced MT–cortex attachments, we anticipated that NuMA localization would be disturbed after an increase in cortical MISP levels. To test this, we analyzed the cortical NuMA localization after overexpression of MISP. In control cells, more than 80% of cells exhibit a polar/crescent localization of NuMA (Fig. 7A). In contrast, after MISP overexpression, the extent of polar NuMA localization drops significantly. Concurrently, we observed a 2-fold increase in cells with lost cortical NuMA localization (Fig. 7A). However, we did not observe an abnormal LGN localization in response to MISP overexpression (Fig. S4F). Next, we analyzed whether ezrin and MISP collaborate to regulate cortical NuMA polarization, similar to what was as shown for spindle orientation and positioning (Fig. 5). As shown in Fig. 7B, we found that double ablation of both ezrin and MISP restored crescent NuMA localization at the cortex.

Loss of cortical NuMA leads to a reduced cortical association of the dynein–dynactin complex subunit p150^glued, whereas increased cortical association of NuMA results in an increased binding of p150^glued (Kotak et al., 2013). Since previous results indicated that MISP influences the cortical distribution of p150^glued (Zhu et al., 2013), we hypothesized that overexpression of MISP might also alter the localization of p150^glued. Therefore, we analyzed the cortical association of p150^glued after MISP overexpression. Indeed, we identified a reduced binding of p150^glued to the cell cortex, similar to what we found for NuMA (Fig. 7C). Taken together, we find that enhanced cortical MISP levels disturb cortical NuMA polarization as well as cortical p150^glued binding.

The integrity of MT dynamics is important for correct NuMA localization at the cortex (Woodard et al., 2010). However, it is not clear whether NuMA binding to the cortex is influenced by astral MTs (Woodard et al., 2010; Seldin et al., 2013). To test whether MISP directly controls NuMA and p150^glued localization or whether changes in cortical NuMA distribution are caused by altered dynamics of astral MTs leading to elongated, buckled astral MTs, we treated the cells with a low dose of nocodazole after Flag–MISP overexpression and analyzed cortical NuMA localization. High levels of MISP perturbed polarized localization of NuMA, even in the absence of astral MTs (Fig. S4G). Hence, disrupting the astral MTs by the addition of a low dose of nocodazole did not rescue the cortical NuMA association, suggesting NuMA and p150^glued are acting upstream of astral MTs. To conclude, these data show that increased cortical MISP levels disturb cortical NuMA and p150^glued localization, which influences the stability and dynamic of astral MTs.

Our results described above suggest that the ezrin–MISP complex acts in a spindle orientation pathway upstream of NuMA and p150^glued. Since loss of MISP disturbed spindle orientation in polarized and non-polarized cells (Fig. 2) (Zhu et al., 2013), we analyzed whether MISP ablation initiates an abnormal NuMA localization. Interestingly, loss of MISP resulted in an excess cortical NuMA localization distributed at the whole cortex (number of cells for ‘all’, 30.9±11.5%) in contrast to the control (number of cells for ‘all’, 11.0±3.1%) (Fig. 7D; Fig. S5A). It was previously shown that NuMA localization to spindle poles and the cell cortex is regulated by opposing Cdk1 and PP2CA activities (Kotak et al., 2013) where Cdk1 phosphorylates NuMA at T2055. Phosphorylated NuMA localizes to the spindle poles whereas non-phosphorylated NuMA localizes to the cell cortex. Therefore, we addressed whether we could redirect NuMA localization to the spindle poles in a MISP downregulation background by simultaneously overexpressing Cdk1 and thereby enhancing NuMA phosphorylation. In fact, analyzing cortical NuMA distribution shows that MISP depletion together with Cdk1 overexpression rescues NuMA localization (Fig. S5B).

Cortical localization of NuMA depends on LGN, phosphatidylinositol 4-monophosphate (PIP) and PIP₂ (hereafter PIP/PIP₂) in vivo (Kiyomitsu and Cheeseman, 2013; Seldin et al., 2013; Kotak et al., 2014; Zheng et al., 2014). As changes in MISP levels did not influence the cortical LGN polarization (Fig. S4F), we asked whether NuMA binding to PIP/PIP₂ is increased in response to MISP downregulation. In control cells, depletion of PIP/PIP₂ by ionomycin and Ca²⁺ (Hammond et al., 2012) resulted in a slight increase in the number of cells that lost cortical NuMA (Fig. 7E). Interestingly, in MISP-depleted cells, the enhanced NuMA localization at the cortex could be reduced by PIP/PIP₂ depletion (Fig. 7E). Thus, MISP might prevent NuMA binding to the plasma membrane.

Our previous work showed that MISP is phosphorylated by Cdk1 (Zhu et al., 2013). Therefore, we asked whether Cdk1 phosphorylation of MISP might be required to regulate cortical NuMA localization. We treated the cells with MISP siRNA and simultaneously overexpressed different MISP phospho-mutants. The analysis of cortical NuMA levels in prometaphase/metaphase HeLa cells showed that MISP phosphorylated by Cdk1 is required to regulate cortical NuMA distribution (Fig. 7F; Fig. S5C). Whereas wild-type MISP and the form mimicking phosphorylation by Cdk1 (MISP 9DC) can mediate crescent NuMA localization, MISP that cannot be phosphorylated by Cdk1 (MISP 9AC) is not able to rescue abnormal NuMA localization. From these data, we conclude that MISP affects mitotic spindle orientation and positioning by controlling cortical polarized NuMA distribution in a Cdk1-dependent manner.

In this study, we describe how MISP collaborates with the NuMA force generator and the ERM proteins in regulating spindle orientation. We show that MISP and ezrin in its open confirmation directly interact (Fig. 1). We propose that SLK/LOK-phosphorylated, and thereby activated, ezrin forms a complex with MISP at the cell cortex (Fig. 7G). Thereby, ezrin controls MISP levels at the cortex to prevent an accumulation of MISP bound to actin (Figs 3 and 4). These tightly controlled cortical MISP levels are required for correct spindle orientation, since they ensure polarized NuMA localization, proper cortical dynein–dynactin association, dynamic astral MTs and, consequently, balanced forces exerted on the mitotic spindle (Figs 6 and 7). Loss of the MISP–ezrin complex by either ablation of one of these proteins or by interfering with the equal MISP:ezrin ratio may lead to altered NuMA and dynein–dynactin localization, and abnormal spindle orientation (Fig. 7G). Hence, we find that a complex of activated ezrin and MISP may ensure proper PIP/PIP₂ binding to NuMA and p150^glued localization at the cortex, which is necessary for balanced cortex–spindle forces and, in turn, proper spindle orientation.

How could an accumulation of MISP in the absence of ezrin be induced at the cortex where both proteins are supposed to interact? One possibility is that MISP and ezrin compete for actin-binding sites at the cortex. As shown by actin co-sedimentation assays (Fig. 4E), we observed that less ezrin bound to actin when MISP was present and vice versa. These data were supported by experiments where the actin cytoskeleton was depolymerized by Latrunculin B treatment. We found that both in ezrin-depleted and control cells MISP levels at the cortex were reduced (Fig. 4D; Fig. S3I). It is therefore conceivable that both proteins interact but simultaneously can compete for actin binding.

We found that accumulation of MISP at the cortex affects NuMA polarization and leads to elongated and buckled astral MTs (Figs 6 and 7). Additionally, we observed a Cdk1-dependent enrichment of NuMA at the cortex in the absence of MISP (Fig. 7). We conclude that tightly controlled MISP levels at the cortex seem to be critical for proper crescent-like cortical NuMA association, and hence for opposing polarized forces governing equal cell division. An accumulation of MISP might block PIP/PIP₂-binding areas of NuMA at the cortex, whereby a loss of MISP could expose these additional binding sites for NuMA. Similarly, impaired NuMA association to the cortex was recently observed after inhibition of ezrin activation by SLK downregulation (Machicoane et al., 2014). Furthermore, in polarized cells MISP strongly localizes at the apical side (Fig. 1D; Fig. S1D), supporting a regulatory function in controlling cortical NuMA distribution, as previously described for other apically localizing proteins (Hao et al., 2010; Zheng et al., 2010; Carminati et al., 2016).

It is possible that ERMs act directly on astral MTs. In fact, the ERM family member moesin has been shown to bind and stabilize MTs at the cortex in Drosophila S2 cells, where moesin is the only ERM family member (Solinet et al., 2013). However a similar function for ezrin in mammalian cells has not been described. Our data show that loss of MISP destabilizes astral MTs (Zhu et al., 2013). Although MISP does not seem to bind MTs or the mitotic spindle directly, our results suggest that MISP is able to modulate MT dynamics, as Fig. 6, which shows enhanced astral MT stability and increased astral MT length. While MISP loss induces an excess of NuMA at the cortex, NuMA enrichment results in an accumulation of dynein–dynactin at the cortex (Merdes et al., 1996; Kotak et al., 2012, 2013; Kiyomitsu and Cheeseman, 2013; Seldin et al., 2013). Increased cortical dynein–dynactin might induce higher levels of MT catastrophe and shrinkage, resulting in a loss of MTs that leads to spindle misorientation and mispositioning (Kotak et al., 2012; Laan et al., 2012). On the other hand, elevated cortical MISP levels led to a loss of NuMA at the cortex and, presumably, to a loss of dynein–dynactin. This could lead to stabilization of MTs, resulting in continous growth and buckling, as we have observed (Fig. 6) (Laan et al., 2012; Kotak et al., 2013). Taken together, defective spindle orientation and positioning induced by uncontrolled MISP levels might be the consequence of an unbalanced spindle-to-cortex connection initiated by disturbed dynein–dynactin levels and impaired MT stability.

MISP is expressed predominantly in adherent cancer cells (Maier et al., 2013; Zhu et al., 2013). In samples of pancreatic, colorectal and stomach cancer from patients, MISP and ezrin strongly overexpressed (data available from https://www.proteinatlas.org/, v16.1; Uhlen et al., 2015). Interestingly, it has been shown that the localization of ezrin, its phosphorylation state and its expression profile correlates with its tumorigenic potential (Arpin et al., 2012). Therefore, tumor formation induced by ezrin overexpression might also rely on MISP expression. In the future, it will be of considerable interest to address whether blocking MISP accumulation in tumor cells could serve as a novel approach in cancer therapy.

HeLa (ATCC CCL-2), HeLa Flp-In-T-Rex (Marius Lemberg, ZMBH, Heidelberg, Germany), HeLa EGFP-CenpA Tomato-EB3 (Andrew McAinsh, Warwick Medical School, Warwick, UK) and HEK293T (ACC 635, DSMZ, Braunschweig, Germany) cells were cultured in DMEM containing 1 g/l and 4.5 g/l glucose, respectively, supplemented with 10% (v/v) fetal bovine serum (FBS; Sigma-Aldrich), 1% (v/v) PenStrep (Sigma-Aldrich) (complete medium for for HeLa and HEK293Tcells) at 37°C in 5% CO₂. Caco-2 BBe (C2BBe1) cells (Maja Köhn, EMBL, Heidelberg, Germany) (Lujan et al., 2016) were cultured in 4.5 g/l glucose Dulbecco's modified Eagle's medium (DMEM) containing 10% (v/v) FBS, 1% (v/v) PenStrep and 0.01 mg/ml holo-transferrin (AppliChem) (complete medium for Caco-2 BBe cells). By using the T-Rex system, a HeLa Flp-In-T-Rex GFP-MISP and HeLa Flp-In-T-Rex MISP-BirA* stable cell lines were generated according to the manufacturer's instructions (Life Technologies). GFP–MISP or MISP–BirA* expression was induced with 1 µg/ml doxycycline. Cell lines were periodically tested for contaminations, and HeLa and HEK293T cells authenticated by Multiplexion, Heidelberg, Germany.

To enrich the mitotic population, cells were synchronized with a single thymidine block by adding 3 mM thymidine. Cells were released after 24 h incubation by washing five times with PBS pre-warmed to 37°C. Finally, cells were cultured in complete medium for 8 to 10 h until they reached mitosis. Nocodazole was either used with a final concentration of 20 nM for 1 h (low-dose treatment) (Kern et al., 2016) or 333 nM overnight to enrich for the mitotic population. Taxol (Paclitaxel) was used in a final concentration of 100 nM overnight to enrich for the mitotic population. ezrin inhibitor (NSC668394) was used in a final concentration of 10 µM for 6 h. Ionomycin was used at a final concentration of 10 µM, together with 1 mM CaCl₂, for 15 min.

siRNA transfections were undertaken by using Lipofectamine 2000 (for HeLa) or Lipofectamine RNAiMAX (for Caco-2 BBe; Invitrogen) according to the manufacturer's instructions. For downregulation experiments, cells were transfected once (HeLa cells) or twice (Caco-2 BBe cyst formation) with 10 to 40 nM siRNA and analyzed 48 to 72 h after transfection. For rescue experiments, 1 µg plasmid and 40 nM siRNA were transfected simultaneously using Lipofectamine 2000. The following siRNA sequences were used: GL2 (firefly luciferase, control), 5′-CGUACGCGGAAUACUUCGAtt-3′; MISP_1, 5′-GUGUCCAAGUUGUGGAUGATT-3′; MISP_2, 5′-GGGAGGACAAGGAGAUGAAGACCUA; ezrin_1, 5′-UCUGUUUCCAGCUGUUGCCtt-3′; ezrin_2, 5′-GCGCAAGGAGGAUGAAGUUtt-3′; radixin_1, 5′-AUGUUCUUCAUGCCAGUUCtt-3′; radixin_2, 5′-UACAUUUCUAGAUCUUGUGtt-3′; moesin_1, 5′-AAUGGUAUCAGGCUUGCGAtt-3′; moesin_2, 5′-UUAGACUGGACAGCAUACGtt-3′; SLK, 5′-GCAGAAACAGACUAUCGAAtt-3′; LOK, 5′-GAAGAGCAUCGGAACCAGAtt-3′.

For IPs, HeLa or HEK293T cells were transfected for 24 h with 7.5 to 25 µg plasmid DNA per 15 cm dish using polyethylenimine (Polysciences, Inc.) at a final concentration of 5 µg/ml.

To grow Caco-2 in cysts, cells were seeded on top of a Matrigel-coated (Corning) eight-well coverglass chamber (Nalge Nunc) (Lujan et al., 2016). Therefore, chambers were coated with 10% (v/v) Matrigel in serum-free medium for 1 h at room temperature. Cells were trypsinized into a single-cell suspension of 5×10⁴ cells/ml in complete medium supplemented with 5% (v/v) Matrigel. Cysts were grown for 4 days.

The rabbit anti-MISP antibody was as described previously (Zhu et al., 2013) and used at a final concentration of 1 µg/ml. This antibody was used in immunofluorescence (IF) and for western blotting (WB) unless otherwise specified. A rabbit polyclonal anti-GFP antibody was raised against GFP tag, which was expressed in E. coli (Innovagen). The purification of the antibody was performed by using GFP protein immobilized on CNBr-activated Sepharose. Rabbit anti-GFP antibody was used at a final concentration of 1 µg/ml. Mouse anti-Flag clone M2 (Cat#F3165 Clone F3165; 1:2000), mouse anti-α-tubulin (Cat#T5168; 1:5000) and mouse anti-vinculin (Cat#V9131 Clone hVIN-1; 1:2000) were obtained from Sigma-Aldrich. Rabbit anti-pericentrin (Cat#ab4448; 1:3000) was obtained from Abcam and mouse anti-Myc (9E10; sc-40; 1:500) were from Santa Cruz Biotechnology. Mouse anti-penta-His (Cat#34660; 1:1000) was purchased from Qiagen and mouse anti-actin (Ab-1; Cat#CP01 Clone JLA20; 1:5000) from Calbiochem. Mouse anti-ezrin (Cat#MA5-13862 Clone 3C12; WB, 1:1000; IF, 1:400) obtained from Thermo Scientific was used for immunofluorescence and western blotting, if not mentioned otherwise. Rabbit anti-ERM (Cat#3142; WB, 1:1000; IF, 1:200), rabbit anti-ezrin (Cat#3145; WB, 1:1000; IF, 1:400) and rabbit anti-pERM (Cat#3141L; WB, 1:100; IF, 1:400) antibodies were purchased from Cell Signaling. Mouse anti-SLK (G-9; sc-515493; 1:1000) and mouse anti-LOK (D-6; sc-398083; 1:1000) antibodies were obtained from Santa Cruz Biotechnology. Rabbit anti-NuMA was a kind gift from Duane A. Compton (Biochemistry and Cell Biology, Geisel School of Medicine at Dartmouth, Hanover, NH) (1:2000). Rabbit anti-LGN was nicely provided by Fumio Matsuzaki (Laboratory for Cell Asymmetry, RIKEN Center for Developmental Biology, Kobe, Japan) (Konno et al., 2008). If possible, antibodies were validated by RNAi experiments. Streptavidin-conjugated horseradish peroxidase (HRP; Cat#19534-050; 1:1000) was purchased from Invitrogen. Secondary antibodies for western blotting were horseradish peroxidase-conjugated donkey anti-rabbit-IgG (Jackson Laboratories; 1:5000–10,000) and goat anti-mouse-IgG (Novus; 1:5000–10,000). Secondary antibodies for immunofluorescence were goat or donkey anti-mouse-IgG, anti-rabbit-IgG antibodies coupled to Alexa Fluor 488, Alexa Fluor 568, Alexa Fluor 594 or Alexa Fluor 633 (Molecular Probes; 1:500-1000). SirActin (CY-SC001; 1:500) was purchased from Spirochrome, Cytoskeleton.

Ezrin, radixin and moesin cDNA was obtained from Genomics and Proteomics Core Facility or Stefan Wiemann, DKFZ, Heidelberg, Germany and amplified by PCR. Ezrin and ezrin fragments were cloned into pEGFP-C3, pEGFP-N1 and pET22b vectors via XhoI, SalI; HindIII, XhoI; EcoRI, HindIII cutting sites, respectively. Radixin and moesin were cloned into pCMV-3Tag2A via HindIII, EcoRI cutting sites. pEGFP-C3-NuMA was a gift from Andreas Merdes (Centre de Biologie Intégrative, Toulouse, France). pCMV-3Tag1A-MISP WT, 7AP, 9AC and pGEX-4T3-MISP were described previously (Zhu et al., 2013). MISP 9DC mutations were introduced as described previously (Seyfang and Jin, 2004) or via PCR-based site directed mutagenesis. For pCDNA5-BirA*-MISP, MISP was PCR amplified and cloned into pCDNA5-FRT (Invitrogen) via the BamHI and KpnI sites. Ezrin mutations were introduced either via a QuikChange Lightning kit (Cat#210514) from Agilent Technologies or PCR-based site directed mutagenesis.

All recombinant proteins were expressed in E. coli BL21-Rosetta. Bacteria were grown in LB medium at 37°C and protein expression was induced with 0.4 mM IPTG overnight at 16°C. N-terminally GST-tagged MISP was natively purified by single-step affinity chromatography using glutathione agarose CL-4B beads (Sigma-Aldrich) according to the manufacturer's instructions. C-terminally His-tagged ezrin versions were natively purified by a single-step affinity chromatography using Ni-NTA–Sepharose according to the manufacturer’s instructions (Qiagen).

For western blotting, cell lysates were prepared with NP-40 lysis buffer [40 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% (v/v) NP-40, 5 mM EDTA, 10 mM β-glycerolphosphate, 5 mM NaF, 1 mM DTT, 0.1 mM Na₃VO₄ and protease inhibitors]. Flag-M2 immunoprecipitation was performed using Flag-M2 affinity beads (Sigma-Aldrich). For GFP immunoprecipitation, the in-house made antibody was added to the cell lysates and collected with protein A–Sepharose (GE Healthcare). GFP traps were performed with GFP traps made in-house by covalently coupling size-exclusion chromatography-purified GFP nanobody (Kubala et al., 2010) (expressed in E. coli BL21-Rosetta) to Sepharose beads. For Flag-, GFP- or endogenous immunoprecipitations, 1 to 5 mg lysates were prepared in NP-40 buffer whereas for GFP traps, lysates were prepared in immunoprecipitation buffer [75 mM HEPES pH 7.4, 1.5 mM EGTA pH 7.5, 1.5 mM EDTA, 150 mM KCl, 10% (v/v) glycerol, 1% (v/v) Triton X-100]. For all immunoprecipitations, lysates were briefly sonified (Bioruptor, diagenode) and pre-cleared by incubation with CL4B–Sepharose beads for 0.5 to 2 h at 4°C on a rotating wheel. Cleared lysates were either incubated with corresponding antibodies or rabbit- or mouse-IgG (Santa Cruz Biotechnology) as a control for endogenous IPs, with Flag-M2 beads overnight at 4°C or with GFP traps for 2 to 4 h at 4°C. For endogenous immunoprecipitation, the antibodies were collected by adding protein A– or protein G–Sepharose (GE Healthcare) and additionally rotated at 4°C for 30 min. After extensive washing with increasing salt concentration up to 300 mM, the proteins were eluted with 4× SDS-Buffer for 15 min at room temperature. The cleared eluate was boiled for 5 min and analyzed by SDS-PAGE and western blotting.

For in vitro pulldown experiments, GST–MISP was incubated overnight at 4°C with His-tagged ezrin variants in NP-40 buffer, and GST–Sepharose beads were used to collect GST–MISP proteins. On the next day after washing, the proteins were eluted with 4× SDS buffer for 15 min at room temperature. The cleared eluate was boiled for 5 min and analyzed by SDS-PAGE and western blotting.

Immunoreactive signals were detected with Immobilon Western Chemiluminescent HRP substrate (Millipore).

The actin co-sedimentation assay was performed by following the instructions of the Non-Muscle Actin Binding Protein Spin-Down Biochem kit (BK013; Cytoskeleton). Actin polymerization was induced according to the protocol and incubated for 1 h at room temperature with either GST–MISP, His–ezrin wild-type or with both proteins together. The samples were centrifuged in a TLA-55 at 10,000 g for 1.5 h at 24°C. The pellet and the supernatant were analyzed by SDS-PAGE. Band intensities were quantified by using Fiji (Schindelin et al., 2012).

For identification of MISP interaction partners, an BirA–MISP proximity pulldown was performed as described previously (Roux et al., 2012). The HeLa Flp-In-T-Rex MISP-BirA* cell line was generated according to the manufacturer's instructions (Life Technologies). BirA*–MISP expression was induced with doxycycline (1 μg/ml, 24 h) and treated with biotin (50 μM, 24 h). The BioID pulldown was performed as described previously (Roux et al., 2013) using streptavidin–Sepharose beads from GE Healthcare.

Elution fractions of BioID and Flag immunoprecipitation were resolved by SDS-PAGE and co-precipitating proteins were detected in-gel by staining with Colloidal Coomassie. Analysis was performed by the DKFZ protein analysis facility (Heidelberg). Gel lanes were cut into slices and trypsin digested. Tryptic peptide mixtures were analyzed by nanoLC ESI-MS/MS using a nano Acquity UPLC system (Waters GmbH) coupled online to an LTQ Orbitrap XL mass spectrometer (Thermo Scientific). Data were acquired by scan cycles of one FTMS scan with a resolution of 60,000 at m/z 400 and a range from 300 to 2000 m/z in parallel with six MS/MS scans in the ion trap of the most abundant precursor ions. The mgf-files generated by Xcalibur software (Thermo Scientific) were used for database searches with the MASCOT search engine (Matrix Science; version 2.4) against the SwissProt database (version SwissProt 2014_07) with taxonomy set as human. Peptide mass tolerance for database searches was set to 5 ppm or 7 ppm, and fragment mass tolerance to 0.4 Da. Carbamidomethylation of cysteine was set as fixed modification. Variable modifications included oxidation of methionine and deamidation of asparagine and glutamine. One missed cleavage site in case of incomplete trypsin hydrolysis was allowed. Furthermore, proteins were considered as identified if at least two unique peptides had an individual ion score exceeding the MASCOT identity threshold. Identified proteins were summarized in an Excel table.

For immunofluorescence, cells grown on coverslips were fixed with 4% (w/v) PFA in PBS at room temperature for 10 min (for MISP/ezrin staining), with 10% (w/v) trichloroacetic acid (TCA) in H₂O at 4°C for 20 min (for NuMA, LGN and pERM staining), or at 37°C with 0.5% (v/v) glutaraldehyde and 1% (v/v) Triton X-100 in cytoskeleton buffer (10 mM MES, 150 mM NaCl, 5 mM EGTA, 5 mM MgCl₂ and 5 mM glucose) for 1 min followed by 1% (v/v) glutaraldehyde in cytoskeleton buffer for 10 min or with methanol at −20°C for 8 min (for visualization of MTs). For p150^glued staining, cells were pre-extracted with PHEM buffer [60 mM Pipes pH 6.9, 25 mM Hepes, 10 mM EGTA and 2 mM MgCl₂ supplemented with 0.5% (v/v) Triton X-100] for 1 min and fixed with −20°C MeOH for 8 min. Afterwards, cells were permeabilized with PBS with 0.5% (v/v) Triton X-100 (PBST-0.5%) for 10 min and blocked with 3% (w/v) BSA in PBS with 0.2% (v/v) Triton X-100 (PBST-0.2%) for 30 to 60 min at room temperature. Cells were incubated with primary antibodies diluted in 3% (w/v) BSA in PBST-0.2% for 1 to 3 h, then washed with PBS and incubated with secondary antibodies in 3% (w/v) BSA in PBST-0.2% (v/v) for 1 to 2 h at room temperature. After washing, DNA was stained with 2 µg/ml Hoechst 33258 (Invitrogen) in PBS for 5 min at room temperature followed by mounting the coverslips onto glass slides with Mowiol (Calbiochem). Cysts were washed three times with PBS and fixed with 4% (w/v) PFA in PBS for 30 min at room temperature. Afterwards, cells were quenched with 0.1 M glycine in PBS for 30 min at room temperature, followed by permeabilization with PBST-0.5% for 10 min and blocking with 10% (v/v) FBS in PBST-0.2%. Primary antibodies were incubated overnight at 4°C diluted in 10% (v/v) FBS in PBST-0.2%. After washing, secondary antibody was diluted in 10% (v/v) FBS in PBST-0.2% and incubated for 1 to 2 h at room temperature. After washing, DNA was stained by incubating cysts with Hoechst 33258-containing PBS for 10 min. Cells were imaged either with an upright motorized Zeiss LSM 700 confocal microscope or an inverted motorized Zeiss LSM 710, with a 63×/1.4 or 40×/1.4 oil DIC III objective, and containing laser diodes 405-5, 488, 555 and 633. Images were analyzed with Fiji (Schindelin et al., 2012).

Cortical MISP intensity was quantified in mitotic cells. A maximum projection of three planes located in the middle of the cell was used to measure the integrated density using Fiji software (Schindelin et al., 2012). Background was subtracted and the ratio of cortical to cytoplasmic area was calculated, which were specifically selected using a macro. In cysts, the apical MISP portion was selected automatically via thresholding, and the intensity level quantified. The whole cysts were selected manually, quantified and the lumen portion subtracted.

For measuring astral MT intensity, cells were fixed with glutaraldehyde and stained with antibodies directed against α-tubulin and pericentrin. Intensities of astral MTs (α-tubulin signal) were measured for each pole separately. For this, the z-plane with the highest spindle pole intensity was selected. A circle around the whole cell (total) as well as a circle around the spindle excluding the chosen spindle pole with corresponding astral MTs (inner) was drawn. The astral MT intensity (I) was calculated after background subtraction from the selected plane as well as one plane below and above using Fiji (Schindelin et al., 2012): I_{astral MTs}=(I_total−I_inner)/I_inner.

Spindle orientation and position was quantified in cells seeded on fibronectin-coated glass coverslips. For this, coverslips were incubated with 50 µg/ml fibronectin (Santa Cruz Biotechnology) in PBS for 30 min at 37°C. Cells grown on fibronectin-coated coverslips were fixed with glutaraldehyde or methanol, and stained for α-tubulin and pericentrin. Metaphase cells were imaged on a confocal microscope with a z-stack step of 0.3 µm. The x, y and z coordinates were measured via Fiji (Schindelin et al., 2012), and the difference in angle relative to the substratum calculated. For spindle position, a maximum projection was produced and the distance from each pole to the closest cortex in line with the spindle measured. In cysts, the spindle angle was calculated by measuring the x, y and z coordinates of the poles and the middle point of the cyst.

A HeLa cell line stably expressing Tomato–EB3 and EGFP–CenpA was transfected with Flag or Flag–MISP using Lipofectamine 2000 at 48 h prior to imaging. At 1 day before microscopy, cells were transferred to a µ-Slide 8-well glass-bottom ibidi chamber (ibidi GmbH). Mitotic cells were imaged on a Zeiss motorized inverted Observer.Z1 using a 63×/1.4 oil Pln Apo DICII objective at 37°C and 5% CO₂. The microscope is equipped with a mercury arc burner HXP 120 and a grayscale CCD camera AxioCamMRm. A Zeiss Apotome optical sectioning device with structured illumination was used during z-stack imaging for near-confocal images. To analyze EB3 tracks, a single plane of a mitotic cell (mitotic stage judged by the CENP-A signal) was imaged over 2 min with a time interval of 2 s. A kymograph was produced by Fiji in which the length (stability) as well as the slope (time rate) was measured (Sironi et al., 2011; Schindelin et al., 2012).

After induction of GFP–MISP expression in HeLa cells, cells were transfected with relevant siRNAs using Lipofectamine 2000 at 72 h prior to imaging. At 1 day before microscopy, cells were transferred to a µ-Slide 8-well ibidi chamber. Mitotic cells were imaged on a Leica TCS SP5II confocal microscope with a Leica PL APO 63×/1.4 oil objective at 37°C and 5% CO₂. FRAP was performed with the help of the FRAP wizard of the Leica software. Before bleaching the cell, pre-bleach images were taken with a 488 nm argon laser line intensity of 5%. Using the fly mode of the wizard, a region of 1.5×3.5 µm at the cortex was selected and bleached over several cycles with a maximum laser intensity of the 405, 458, 476 and 488 nm lasers. The recovery of the signal was measured using the 488 nm argon laser with an intensity of 5%. Intensities were normalized to the pre-bleached signals.

Each experiment was repeated independently at least three times. All statistical analysis was performed with GraphPad Prism 6.0. The results shown in the figures represent the mean±s.d., if not indicated otherwise. Either Student's t-test (two-tailed distribution) or two-way analysis of variance (ANOVA) was performed to calculate the P value that is indicated with asterisks or as ns (not significant). Statistical details of each experiment can be found in the figure legends. No statistical method was used to predetermine sample size.

We thank M. Lemberg, A. McAinsh, D. Compton, F. Matsuzaki, A. Merdes and M. Köhn for providing reagents and cell lines. We acknowledge the members of the DKFZ Microscopy and Mass Spectrometry Core Facility for providing equipment and excellent technical assistance. The members of our lab are thanked for comments and critical reading of the manuscript.