The tumor suppressor Lgl1 forms discrete complexes with NMII-A and Par6α–aPKCζ that are affected by Lgl1 phosphorylation

Cell migration is a highly integrated multistep process involving coordinated cytoskeletal remodeling (Lauffenburger and Horwitz, 1996; Mitchison and Cramer, 1996). The basic steps involved are: polarizaton of the cell; extention of an actin-rich protrusion from the leading edge; establishment of new adhesion sites in the new protrusion; pulling the cell body forward in the direction of the new attachment sites; and, finally, detachment of the rear of the cell and tail retraction (Lauffenburger and Horwitz, 1996; Mitchison and Cramer, 1996; Ridley et al., 2003). Non-muscle myosin II (NMII), an actin-based motor protein, plays a key role in cell migration through its effects on cell polarity, adhesion, lamellar protrusion and rear retraction (Conti and Adelstein, 2008; Jay et al., 1995; Verkhovsky et al., 1999; Vicente-Manzanares et al., 2009).

NMII is a hexamer composed of two heavy chains of ∼200 kDa and two pairs of essential and regulatory light chains (Vicente-Manzanares et al., 2009). The NMII heavy chains consist of an N-terminal globular head domain and a tail domain. The tail domain includes an α-helical coiled-coil-forming rod and a non-helical tailpiece on the C-terminus. The α-helical coiled-coil domain is responsible for assembly of NMII monomers into filaments, which are the functional structures required for NMII activity (Dulyaninova et al., 2005; Even-Faitelson and Ravid, 2006; Murakami et al., 2000; Ronen and Ravid, 2009; Rosenberg and Ravid, 2006). Vertebrates have three different NMII isoforms, NMII-A, NMII-B and NMII-C that differ in their heavy chain sequences, in the kinetics of their ATPase activity, in their subcellular localization and in their functions in cell polarization during migration (Golomb et al., 2004; Katsuragawa et al., 1989; Kolega, 1998; Shohet et al., 1989; Simons et al., 1991; Vicente-Manzanares et al., 2009). In migrating cells NMII-A is dynamic and assembles actomyosin bundles in protrusions (Cai et al., 2006; Vicente-Manzanares et al., 2007), whereas NMII-B incorporates into preformed actin bundles, remains stationary, and defines the center and rear of the migrating cell (Vicente-Manzanares et al., 2008).

Recently we showed that NMII-A and the mammalian homolog of the tumor suppressor Lethal (2) giant larvae (Lgl; also known as L(2)gl) play a central role in the polarization of migrating cells (Dahan et al., 2012). Lgl is a tumor suppressor protein essential for the development of polarized epithelia, for cell polarity associated with asymmetric cell division of neuroblasts during fly development and for the correct polarization of migrating cells (Bilder et al., 2000; Bilder and Perrimon, 2000; Dahan et al., 2012; Ohshiro et al., 2000; Peng etal., 2000). Lgl is composed of two domains, the N-terminal region folds into two β-propellers. These provide docking platforms for simultaneous interactions with a number of proteins. The C-terminal domain is an Lgl-family-specific domain containing several conserved serine residues which are targeted for phosphorylation by atypical protein kinase C isoform ζ (aPKCζ) (Betschinger etal., 2003; Plant etal., 2003; Vasioukhin, 2006).

Mammalian cells express two Lgl homologs, Lgl1 and Lgl2 (also known as Llgl and Llgl2). Lgl1 is expressed almost ubiquitously, whereas Lgl2 is expressed in a tissue-specific manner (Klezovitch etal., 2004; Sripathy etal., 2011). In cultured cell lines and in mouse brain, Par6, aPKC and Lgl form a multiprotein polarity complex in which Lgl is targeted for phosphorylation by aPKCζ (Betschinger etal., 2003; Plant etal., 2003). This phosphorylation is important for the correct polarization of embryonic fibroblasts in response to wounding (Plant etal., 2003) and for centrosome reorientation in astrocytes (Etienne-Manneville, 2008).

Biochemical and genetic analyses suggest that the Drosophila Lgl is a component of the cytoskeleton that interacts with NMII and that this interaction is regulated by the phosphorylation of Lgl (Betschinger etal., 2005; Kalmes etal., 1996; Strand etal., 1994). Recently we showed that in mammalian cells Lgl1 interacts directly with NMII-A through a domain containing the phosphorylation sites for aPKCζ (Dahan etal., 2012). Lgl1 interaction with NMII-A inhibits NMII-A filament assembly in vitro, excludes NMII-A from the leading edge of the cell and regulates the morphology of focal adhesions (Dahan etal., 2012). We proposed that the interaction between Lgl1 and NMII-A is electrostatic and that phosphorylation of Lgl1 decreases the positive charge of the interacting domain of Lgl1, thus preventing it from binding to NMII-A (Dahan etal., 2012). This hypothesis fits previous findings that phosphorylation of Drosophila Lgl dissociates it from the cytoskeleton (Betschinger etal., 2005). The experiments reported here reveal that phosphorylation of Lgl1 by aPKCζ affects its interaction with NMII-A both in vitro and in vivo. We show that Lgl1 forms two distinct complexes, Lgl1–NMIIA and Lgl1–Par6α–aPKCζ and that NMII-A and aPKCζ compete for binding to Lgl1. Furthermore, we demonstrate that Lgl1 phosphorylation affects its cellular localization and the organization of the acto-NMII cytoskeleton.

Studies in Drosophila indicate that Lgl is associated with NMII (Strand etal., 1994), and that phosphorylation of Drosophila Lgl possibly dissociates it from the cytoskeleton (Betschinger etal., 2005). In human Lgl2, serine residues 649, 653 and 660 are important for its phosphorylation in vitro, and serine 653 in human Lgl2 and the corresponding serine residue 659 in human Lgl1 are phosphorylated in vivo (Yamanaka etal., 2003). We recently showed that mammalian Lgl1 interacts directly with NMII-A, inhibiting its ability to assemble into filament in vitro (Dahan etal., 2012). To investigate whether phosphorylation of Lgl1 by aPKCζ affects its interaction with NMII-A, we created phosphomimetic, as well as unphosphorylatable, mouse Lgl1 C-terminal fragments involving serine residues 658, 662 and 669, corresponding to serines 649, 653 and 660 in the human Lgl1. These phospho-mutant Lgl1 C-terminal fragments were fused to maltose binding protein, MBP–Lgl1-C^Asp and MBP–Lgl1-C^Ala, respectively (Fig. 1A). These Lgl1 C-terminal phospho-mutant protein fragments were incubated with a NMII-A C-terminal fragment (NMII-A Rod,</emph> Fig. 1B) and subjected to a pull-down assay. MBP–Lgl1-C^ΔMBD (Fig. 1A), which lacks the NMII-A-binding site (Dahan etal., 2012), was used as a negative control. MBP–Lgl1-C^WT and MBP–Lgl1-C^Ala clearly co-precipitated with NMII-A Rod (Fig. 2A). In contrast, MBP–Lgl1-C^Asp, like MBP–Lgl1-C^ΔMBD, did not co-precipitate with NMII-A Rod (Fig. 2A). These results indicate that Lgl1 phosphorylation by aPKCζ plays an important role in the interaction between Lgl1 and NMII-A Rod in vitro.

To test the effect of Lgl1 phosphorylation on its interaction with NMII-A in vivo, we examined the ability of Lgl1 C-terminal phospho-mutant protein fragments to bind to endogenous NMII-A in a cell extract obtained from the NIH 3T3 cell line (Fig. 2B). Similar to the in vitro results, only traces of endogenous NMII-A co-precipitated with MBP–Lgl1-C^Asp, whereas much more co-precipitated with MBP–Lgl1-C^WT and MBP–Lgl1-C^Ala (Fig. 2B).

To further study the effect of Lgl1 phosphorylation on its interaction with NMII-A in vivo, we transfected HEK293T cells with full-length Lgl1 tagged with GFP (GFP–Lgl1^WT, GFP–Lgl1^Asp and GFP–Lgl1^Ala) and performed a co-immunoprecipitation assay using endogenous NMII-A. GFP–Lgl1^WT and GFP–Lgl1^Ala but not GFP–Lgl1^Asp co-immunoprecipitated with NMII-A (Fig. 2C). These results indicate that phosphorylation of Lgl1 prevents its interaction with NMII-A both in vitro and in vivo.

Because Lgl1 binding to NMII-A Rod inhibits its ability to assemble into filaments (Dahan etal., 2012), we examined the effect of MBP–Lgl1-C phospho-mutants on NMII-A Rod filament assembly. To do this we induced NMII-A Rod filament assembly in the presence of MBP–Lgl1-C^WT, MBP–Lgl1-C^Asp MBP–Lgl1-C^Ala and MBP–Lgl1-C^ΔMBD. Without the addition of Lgl1, 73% of NMII-A Rod assembled into filaments. However, in the presence of MBP–Lgl1-C^WT or MBP–Lgl1-C^Ala, NMII-A Rod filament assembly decreased to 48% and 38%, respectively (supplementary material Fig. S1). Addition of MBP–Lgl1-C^Asp or MBP–Lgl1-C^ΔMBD had only a minor effect, with 69% and 70% of NMII-A Rods assembled into filaments, respectively (supplementary material Fig. S1). Thus, the phosphorylation state of Lgl1 affects the degree of NMII-A filament inhibition.

We recently demonstrated that Lgl1 regulates the polarity of migrating cells by its cellular localization and by controlling the assembly state of NMII-A (Dahan etal., 2012). Because Lgl1 phosphorylation regulates its interaction with NMII-A as well as its inhibitory effect on NMII-A (Fig. 2 and supplementary material Fig. S1), we tested the effect of Lgl1 phosphorylation on its cellular localization. aPKCζ^−/− primary embryo fibroblasts (Leitges etal., 2001) and control cells were transfected with GFP-Lgl1^WT. Control cells were also transfected with GFP-Lgl1^Ala or GFP-Lgl1^Asp. These cells were subjected to a scratch wound assay to produce polarized migrating cells, and the cellular localization of the GFP-tagged proteins was determined (Fig. 3A,B and supplementary material Fig. S2). In control cells GFP–Lgl1^WT localized mainly at the leading edge of the cell (Fig. 3A and supplementary material Fig. S2), whereas in aPKCζ^−/− cells GFP–Lgl1^WT was confined to the cell cortex, forming a shell-like structure around the cell (Fig. 3B and supplementary material Fig. S2). GFP–Lgl1^Ala behaved similarly to GFP–Lgl1^WT in aPKCζ^−/− cells, forming a sphere around the cell cortex (Fig. 3B and supplementary material Fig. S2). The effects of the absence of aPKCζ on the localization of GFP–Lgl1^WT thus resembled those observed with GFP–Lgl1^Ala.

In contrast, GFP–Lgl1^Asp expressed by control cells was completely diffuse, with the protein distributed throughout the cytoplasm. F-actin staining showed that GFP–Lgl1^Asp did not reach the lamellipodium (Fig. 3A and supplementary material Fig. S2). These observations were confirmed by quantification of the fluorescence intensity of GFP–Lgl1 phospho-mutants from the leading edge of the cell towards its center (Fig. 3C,D). The fluorescence intensity patterns of GFP–Lgl1^WT and F-actin were similar with maximum fluorescence at the cell edge (Fig. 3C). The fluorescence intensity of GFP–Lgl1^Ala in control cells, as well as GFP–Lgl1^WT in aPKCζ^−/− cells, was higher than that of GFP–Lgl1^WT in control cells, demonstrating the concentration of this protein at the cell cortex (Fig. 3C,D). In contrast, GFP–Lgl1^Asp fluorescence intensity was much lower than that of GFP–Lgl1^WT, further demonstrating the diffusible behavior of this protein (Fig. 3C). These results thus suggest that phosphorylation of Lgl1 by aPKCζ plays an important role in the cellular localization of Lgl1. Interestingly, in cells expressing GFP–Lgl1^Ala, or in aPKCζ^−/− cells expressing GFP–Lgl1^WT, F-actin formed a shell-like structure similar to Lgl1^Ala, and in cells expressing GFP–Lgl1^Asp, F-actin and GFP–Lgl1^Aap did not localized to the lamellipodium, indicating that the state of Lgl1 phosphorylation might also affect the cellular localization of F-actin.

To further study the role of Lgl1 phosphorylation on its cellular localization, we determined the amount of GFP–Lgl1 phospho-mutants associated with the cytoskeleton using a Triton X-100 solubility assay. A higher percentage of GFP–Lgl1^WT was associated with the cytoskeleton in the absence of aPKCζ than in control cells (36% versus 27%; Fig. 4A). Similarly, expression of the unphosphorylated form of Lgl1, GFP–Lgl1^Ala in control cells resulted in higher cytoskeletal association than with GFP–Lgl1^WT (34% versus 27%, Fig. 4A). The phosphomimetic form of Lgl1, GFP–Lgl1^Asp, showed less association with the cytoskeleton than GFP–Lgl1^WT (20% versus 27%, Fig. 4A). Together, these results indicate that Lgl1 phosphorylation by aPKCζ regulates its interaction with the cytoskeleton, thus affecting its cellular localization. These findings are consistent with previous reports that Drosophila Lgl dissociates from the cytoskeleton on phosphorylation by aPKC (Betschinger etal., 2005).

Because Lgl1 affects the cellular localization of NMII-A (Dahan etal., 2012), we studied the effect of the expression of phospho-Lgl1 mutants on the association of NMII-A with the cytoskeleton. Overexpression of GFP–Lgl1^WT reduced NMII-A association with the cytoskeleton compared to control cells (32.6% versus 41.6%, Fig. 4B). Expression of GFP–Lgl1^Ala in control cells resulted in a further decrease in the association of NMII-A with the cytoskeleton (23.5% versus 41.6%, Fig. 4B). These results indicate that Lgl1 phosphorylation affects the solubility properties of NMII-A, possibly by regulating NMII-A filament assembly.

Lgl1 forms multiprotein polarity complex with Par6α and aPKC (Plant etal., 2003). To begin understanding the effect of Lgl1 phosphorylation on the regulation of NMII-A filament assembly, we tested whether Lgl1 exists in a complex with Par6α, aPKCζ and NMII-A. HEK293T cells were co-transfected with GFP-Lgl1^WT and HA-Par6α and subjected to co-immunoprecipitation assays using anti-NMII-A, anti-GFP anti-HA or anti-aPKCζ-specific antibodies. Lgl1 immunoprecipitates contained Par6α and aPKCζ as well as NMII-A (Fig. 5A,B). However, Par6α immunoprecipitates contained Lgl1 and aPKCζ but not NMII-A (Fig. 5A,B). Similarly, NMII-A immunoprecipitates contained Lgl1 but not Par6α or aPKCζ (Fig. 5A,B). Thus, Lgl1 forms two distinct complexes in vivo, Lgl1–NMII-A and Lgl1–Par6α–aPKCζ.

To study the effect of Lgl1 phosphorylation on its interaction with Par6α and aPKCζ, we examined the ability of Lgl1 C-terminal phospho-mutant protein fragments to bind to endogenous aPKCζ obtained from NIH 3T3 a cell extract, and to expressed Par6α obtained from the HEK293T cell line transfected with HA-Par6α. Only traces of endogenous aPKCζ and HA–Par6α co-precipitated with MBP–Lgl1-C^Asp, whereas MBP–Lgl1-C^WT and MBP–Lgl1-C^Ala co-precipitated with endogenous aPKCζ and HA–Par6α (Fig. 5C,D). These results indicate that Lgl1 phosphorylation by aPKCζ inhibits its interaction with aPKCζ and Par6α, as well as NMII-A (Fig. 5E).

Because Lgl1 forms two distinct complexes it is possible that Lgl1 does not bind to aPKCζ and NMII-A concurrently. Hence, we tested whether aPKCζ and NMII-A bind to the same site on Lgl1. Lgl1 C-terminal fragments with different deletions (Fig. 1A) were incubated with endogenous aPKCζ in a cell extract obtained from NIH 3T3 cells, and subjected to a pull-down assay. A Lgl1 fragment containing residues 615–1036 [MBP–Lgl1-C^(615–1036)] bound to endogenous aPKCζ, whereas a Lgl1 fragment containing residues 678–1036 [MBP–Lgl1-C^(678–1036)] did not (Fig. 6A). This indicated that the domain of Lgl1 interacting with aPKCζ lay within residues 615–678. This region contains the Lgl1 domain mediating its interactions with NMII-A (residues 645–677) (Dahan etal., 2012). To determine whether this domain also mediates the interaction of Lgl1 with aPKCζ, we tested the ability of MBP–Lgl1-C^ΔMBD, which lacks the NMII-A-binding site, to bind endogenous aPKCζ. As shown in Fig. 6A, MBP–Lgl1-C^ΔMBD did not bind the endogenous aPKCζ, indicating that aPKCζ and NMII-A bind to Lgl1 in the same domain.

We next tested whether aPKCζ competes with NMII-A for interaction with Lgl1. Proteins that belong to the PKC family are regulated by auto-inhibition through their pseudosubstrate domain, with the pseudosubstrate site binding to the substrate-binding pocket in the kinase domain (Rosse etal., 2010). We therefore used a recombinant catalytic domain of aPKCζ (aPKCζ-CAT). Preincubation of MBP–Lgl1-C^WT with NMII-A Rod at a molar ratio of 1∶1 or 1∶5 completely inhibited the subsequent binding of aPKCζ-CAT (Fig. 6B). Similarly, preincubation of MBP–Lgl1-C^WT with aPKCζ-CAT at a molar ration of 1∶1 or 1∶5 inhibited the binding of NMII-A Rod to MBP–Lgl1-C^WT. These results indicate that aPKCζ and NMII-A show competitive binding to Lgl1 amino acid residues 645–677 (Fig. 6C).

Finding that Lgl1 formed a complex with Par6α and aPKCζ led us to examine the localization properties of these polarity proteins in polarized migrating cells. Control cells were transfected with GFP-Lgl1^WT or HA-Par6α and subjected to a scratch wound assay to achieve polarized migrating cells. The cellular localizations of GFP–Lgl1^WT, HA–Par6α, endogenous aPKCζ and NMII-A were determined. In migrating polarized cells, GFP–Lgl1^WT or HA–Par6α colocalized with aPKCζ at the leading edge of the cell (Fig. 7A,B; supplementary material Figs S3 and S4). NMII-A was present in the lamellum but absent from the lamellipodium (Fig. 7A,B; supplementary material Figs S3 and S4) as previously described (Gupton and Waterman-Storer, 2006; Kolega, 1998; Vicente-Manzanares etal., 2007). These observations were confirmed by quantification of the fluorescence intensity of GFP–Lgl1^WT, HA–Par6α, aPKCζ and NMII-A from the leading edge of the cell towards its center (Fig. 5C,D). The fluorescence intensity of GFP–Lgl1 or HA–Par6α, and aPKCζ revealed a similar pattern, with maximum fluorescence at the cell edge (Fig. 7C,D). In contrast very little NMII-A was present at the cell edge but rose gradually toward the cell center (Fig. 7C,D). These results confirmed the presence of Lgl1 and Par6α at the leading edge of the cell and their colocalization with aPKCζ, as well as the absence of NMII-A in this region. These results are in agreement with a previous study showing that Lgl1 colocalized with aPKC in migrating murine fibroblast cells (Plant etal., 2003).

We recently showed that NMII-A and the tumor suppressor Lgl1 play a central role in the polarization of migrating cells (Dahan etal., 2012). In mammalian cells Lgl1 interacts directly with NMII-A through a negatively charged region on the NMII-A molecule, inhibiting its ability to assemble into a filament. The binding domain of Lgl1 for NMII-A is positively charged and contains the phosphorylation sites for aPKCζ (Dahan etal., 2012). Therefore, we proposed that the interaction between Lgl1 and NMII-A is electrostatic and that phosphorylation of Lgl1 decreases the positive charge of the Lgl1 interacting domain. This would prevent its binding to NMII-A and thereby regulate Lgl1–NMII-A interaction (Dahan etal., 2012). This hypothesis is supported by previous findings that phosphorylation of Drosophila Lgl dissociates it from the cytoskeleton (Betschinger etal., 2005). Here we confirmed this hypothesis and provided new insights into the mechanism by which Lgl1–NMII-A interaction is regulated by the phosphorylation of Lgl1 by aPKCζ.

Lgl1 and Lgl1^Ala but not Lgl1^Asp bind to NMII-A both in vitro and in vivo. These results fit with our findings that in polarized migrating cells the state of Lgl1 phosphorylation determines its localization properties and its association with the cytoskeleton. The phosphorylatable form of Lgl1 localized to the leading edge of the cell, whereas the phosphomimetic form of Lgl1 (Lgl1^Asp) diffused throughout the cell and was completely absent from the leading edge. In contrast, the unphosphorylatable form of Lgl1 (Lgl1^Ala) was mainly restricted to the cell cortex. These cellular localization properties were also reflected by the degree of association of the various Lgl1 phospho-mutants with the cytoskeleton. These results indicate that phosphorylation of Lgl1 plays a crucial role in its association with the cytoskeleton, as well as in its cellular localization.

Lgl1 phosphorylation also affects the association of NMII-A with the cytoskeleton. Expression of the phosphorylatable form of Lgl1 decreased the amount of NMII-A associated with the cytoskeleton. Expression of the unphosphorylatable form of Lgl1 decreased this further. Lgl1 binding to NMII-A inhibits filament assembly (Dahan etal., 2012) and Lgl1 phosphorylation decreased the interaction between these proteins. We therefore propose that the effect of a decrease in the amount of NMII-A associated with the cytoskeleton because of the expression of Lgl1 or Lgl1^Ala mirrors the increased amounts of non-filamentous NMII-A. The reduction in the amount of NMII-A associated with the cytoskeleton as a result of the expression of Lgl1 and Lgl1^Ala is possibly due to the partial phosphorylation of the expressed Lgl1 by endogenous aPKCζ.

aPKCζ is part of the polarity complex Par6α–Par3 involved in establishing the apical-basal polarity of epithelial cells (Bilder, 2004; Gao and Macara, 2004; Knust and Bossinger, 2002). A direct interaction between basolateral Lgl and the apical aPKC–Par6 complex has been demonstrated in Drosophila and mammalian epithelial cells (Betschinger etal., 2003; Hutterer etal., 2004; Plant etal., 2003; Yamanaka etal., 2003). Drosophila Lgl phosphorylation by aPKC is required for the exclusion of Lgl from the apical region in epithelial cells (Betschinger etal., 2003; Hutterer etal., 2004; Müsch etal., 2002; Yamanaka etal., 2003). In a genetic study in Drosophila, reduction in aPKC levels suppressed the development of Lgl mutant phenotypes, such as cell polarity defects and tumorigenesis (Rolls etal., 2003). Further, studies using Drosophila or Xenopus embryo indicated that mutual inhibition between apical aPKC and basolateral Lgl is important in maintaining epithelial membrane polarity (Chalmers et al., 2005; Hutterer etal., 2004). These observations suggest that an antagonistic interaction of Lgl with apical Par3–aPKC–Par6 complex is important for the development of polarized membrane domains in epithelial cells. Here we have shown that aPKCζ is necessary for proper cellular localization of Lgl1. The cellular localization of Lgl1 in aPKCζ^−/− cells was aberrant, similar to that of unphosphorylatable Lgl1, indicating the important role of aPKCζ in the regulation of Lgl1.

Finally, our findings indicate that in vivo Lgl1 exists in two distinct complexes, Lgl1–NMII-A and Lgl1–Par6α–aPKCζ that are affected by the state of Lgl1 phosphorylation. The formation of two discrete complexes is explained by our finding that NMII-A and aPKCζ compete to bind to the same region on Lgl1. This behavior may ensure the cellular localization of the two complexes to different cellular compartments, thereby establishing front–rear polarization in migrating cells. Indeed, the complex Lgl1–Par6α–aPKCζ resides in the leading edge, a region that is not occupied by NMII-A. We propose that Lgl1 in the leading edge is in the unphosphorylated state and therefore forms a complex with Par6α–aPKCζ in this region. It is possible that non-filamentous NMII-A forms a complex with unphosphorylated Lgl1 at the lamellipodium, preventing NMII-A from assembling into filaments, thus allowing polymerization of F-actin in that region. In contrast, filamentous NMII-A that is not bound to Lgl1 is found in the lamellum where, along with F-actin, it forms the stress fibers required for detachment of a migrating cell. It is plausible that we did not detect the Lgl1–NMII-A in the lamellipodium because of the formation of Lgl1 and NMII-A-monomer complexes that are too small to detect.

Based on all the results presented here, we propose a model for the role of Lgl1–NMIIA and Lgl1–Par6α–aPKCζ in establishing front–rear polarization in migrating cells (Fig. 8). In migrating polarized cells Lgl1 resides at the leading edge of the cell in a complex with Par6α–aPKCζ, and it is this complex that defines the leading edge. In the lamellipodium Lgl1 binds to NMII-A but not to aPKCζ, inhibiting NMII-A filament assembly. These events allow the cell to polymerize F-actin to move the cell forward. According to our model, Lgl1 is absent from the rear part of the cell, allowing NMII-A to assemble into filaments to enable cell retraction.

The proteins used for this study were human NMII-A, mouse Lgl1 and mouse-Par6α, accession numbers NM_002473, NM_005964, NM_008502 and NM_001047436, respectively.

Control and aPKCζ^−/− primary embryo fibroblasts were kindly provided by Dr Jorge Moscat (Sanford-Burnham Medical Research Institute, La Jolla). Control and aPKCζ^−/− primary embryo fibroblasts, mouse embryonic fibroblast NIH 3T3 and HEK293T cell lines were maintained in high glucose DMEM supplemented with 2 mM L-glutamine, 10% FCS and antibiotics (100 U/ml penicillin, 100 mg/ml streptomycin and 1∶100 Biomyc3 anti-mycoplasma antibiotic solution; Biological Industries, Beit HaEmek, Israel). Cells were grown at 37°C in a humidified atmosphere of 5% CO₂ and 95% air.

Antibodies specific for the C-terminal region of mouse NMII-A were a kind gift from Dr R. S. Adelstein (NIH/NHLBI). Antibodies specific for the C-terminal region of human NMII-A were generated in rabbits according to the method of Phillips etal. (Phillips etal., 1995). Recombinant GFP antibodies were prepared in rabbits as described previously (Rosenberg and Ravid, 2006). Specific antibodies for the C- and N-terminal domains of Lgl1 were generated in rabbit using the peptides DTTLDTTGDVTVEDVKD and DDYRCGKALGPVESLQ, respectively (Dahan etal., 2012). Anti-mouse monoclonal β-actin antibodies were purchased from Sigma-Aldrich and anti-rat monoclonal HA antibodies from Roche Diagnostics Corporation (Basel, Switzerland). Anti-mouse monoclonal aPKCζ antibodies were purchased from Santa Cruz Biotechnology.

Mouse Lgl1 and pMBP-3 (Sheffield etal., 1999) were kindly provided by Dr T. Pawson (Mount Sinai Hospital, Toronto, Canada), and Dr P. Sheffield (University of Virginia, Charlottesville, Virginia, USA), respectively. pMBP-Lgl1-C, pMBP-Lgl1(678–1036), pMBP-Lgl1(615–1036) and pMBP-Lgl1-CΔ(645–677) were created as described previously (Dahan etal., 2012). To create pGST-Lgl1, Lgl1 in a pCMV5 vector was digested with BamHI and the fragment was inserted into a pGEX-3X vector that was digested with BamHI. GST–Lgl1^Ala was created by replacing the conserved serine residues at positions 658, 662 and 669 in the Lgl1 protein with alanine residues using the QuikChange site-directed mutagenesis kit (Stratagene) with the following primer 5′-CGGGTGAAGGCCCTCAAGAAGGCACTGAGACAGGCATTCCGGC-3′. The mutations were verified by DNA sequencing (Center for Genomic Analysis, The Hebrew University, Jerusalem). GST-Lgl1^Asp was created with a series of mutagenesis reactions by substituting the conserved serine residues at positions 658, 662 and 669 in the Lgl1 protein with aspartate residues using the QuikChange Site-directed mutagenesis kit (Stratagene) with the following primers 5′-GAAGTCACTGAGACAGGATTTCCGGCGAATCCGCAAGAG-3′ for substitution of the serine residue in position 669 and 5′-CCACTGTCACGGGTGAAGGACCTCAAGAAGGACCTGAGACAGG-3′ for substitution of serine residues in positions 658 and 662. The mutations were confirmed by DNA sequencing (Center for Genomic Analysis, The Hebrew University, Jerusalem). To create pGFP-Lgl1, GFP-Lgl1^Ala and GFP-Lgl1^Asp, pGST-Lgl1 were digested with BamHI and the fragment was ligated into pGFP-C2 that was digested with BamHI. To create pMBP-Lgl1^Ala and pMBP-Lgl1^Asp the vectors GFP-Lgl1^Ala and GFP-Lgl1^Asp were digested with BamHI and the fragments were inserted into pMBP-3 digested with BamHI. To create pMBP-Lgl1-C^Ala and pMBP-Lgl1-C^Asp, the vectors pMBP-Lgl1^Ala and pMBP-Lgl1^Asp were digested with KpnI and the fragments were inserted into different pMBP-3 plasmids digested with KpnI. Par6α was kindly donated by Dr T. Pawson. To create pGFP-Par6α, Par6α in the pCDN3 vector was digested with Cfr42I and EcoRI and the fragment was inserted into pGFP-2 digested with Cfr42I and EcoRI. HA fused to Par6α was created by removing the GFP tag from GFP-Par6α using NheI and HindIII and the following synthetic primers coding for the HA tag 5′-CTAGCATGGATTACCCATACGATGTTCCAGATTACGCTCGA-3′ were subsequently ligated into the gap. NMII-A Rod fragment in pET21 that encodes for 480 residues from the NMII-A heavy chain C-terminus (SwissProt P35579, amino acids 1480–1960) with the addition of two amino acids on the N-terminus was prepared as previously described (Ronen and Ravid, 2009). Histidine-tagged aPKCζ-CAT was kindly provided by Dr Yehiel Zick (Weizmann Institute, Rehovot, Israel).

MBP-Lgl1-C constructs were transformed into Escherichia coli BL21-CodonPlus(DE3)-RIL, obtained from Dr Tsafi Danieli, (Hebrew University of Jerusalem, Israel) and the bacteria were grown in 100 ml lysogeny broth (LB) with 50 µg/ml ampicillin at 37°C to and OD₆₀₀ = 0.5. 0.1 mM IPTG was then added and the bacteria were grown for an additional 3 hours at 25°C. The bacteria were pelleted at 12,000 g at 4°C for 20 minutes and the pellets frozen at −20°C. The bacterial pellet was dissolved in 4 ml MBP buffer [20 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 150 mM NaCl, 1 mM dithiothreitol (DTT) and 0.1 mM phenylmethylsulfonyl fluoride (PMSF)] and 1 mg/ml lysozyme and incubated on ice for 15 minutes. The lysate was sonicated for 5× 10 seconds (Misonix Microson-Ultrasonic cell disruptor) and centrifuged at 16,000 g at 4°C for 20 minutes. 500 µl of 50% amylose beads slurry (New England BioLabs) was washed twice with MBP buffer, added to the bacterial lysate and incubated at 4°C for 2 hours on a rotator, followed by three washes with MBP buffer. Expression and purification of NMII-A Rod was carried out according to Straussman etal. (Straussman etal., 2007). Histidine-tagged aPKCζ-CAT was expressed and purified according to the manufacturer's instructions (Thermo Scientific, Rockford, IL). The concentrations of the MBP–Lgl1-C proteins bound to amylase beads, NMII-A Rod and aPKCζ-CAT were determined by comparing the densitometry of the band of a sample of each protein on Coomassie-Blue-stained polyacrylamide gels to known amounts of BSA samples on the same gel.

3 µg MBP–Lgl1-C phospho-mutant protein fragments immobilized on amylose beads in 500 µl CHAPS buffer was incubated with 1 µg NMII-A Rod at 4°C for 2 hours on a rotator. Bound proteins were washed three times in CHAPS buffer and resolved by 10% SDS-PAGE. For competition assays, MBP–Lgl1-C^WT immobilized on amylose beads in 500 µl CHAPS buffer was incubated with NMII-A Rod fragment (1∶1 and 1∶5 molar ratios) at 4°C for 2 hours. After washing with CHAPS buffer, aPKCζ-CAT (1∶1 and 1∶5 molar ratios) was added and incubated at 4°C for 2 hours on a rotator. For the reciprocal experiment aPKCζ-CAT was preincubated with immobilized MBP–Lgl1-C^WT. After washing with CHAPS buffer, NMII-A Rod was added. Bound proteins were resolved by SDS-PAGE analysis and then immunoblotted with anti-Lgl1, anti-NMII-A and anti-aPKCζ.

Maltose (0.1 M) was added to 1.2 ml MBP–Lgl1 C-terminal phospho-mutant protein fragments immobilized on amylose beads, which were then incubated on a rotator for 1 hour at 4°C and centrifuged at 6000 g at 4°C for 2 minutes. The supernatants were dialyzed against MBP buffer containing 1 mM DTT and 0.1 mM PMSF. Protein concentrations were determined as described in Materials and Methods. MBP–Lgl1-C fusion proteins at 3–6 µM were added to monomeric NMII-A Rod in 20 mM Tris-HCl (pH 7.5), 600 mM NaCl, 5 mM EDTA, 1 mM DTT, dialyzed against 10 mM phosphate buffer, pH 7.5, 2 mM MgCl₂ and 150 mM NaCl for 4 hours at 4°C. The protein mix was centrifuged in a TL-100 ultracentrifuge (Beckman Coulter, Fullerton, CA) at 100,000 g, at 4°C for 1 hour, and the supernatant and pellet were separated on 10% SDS-PAGE. The gels were stained with Coomassie Brilliant Blue, scanned, and quantified using the densitometry program Fujifilm ImageGauge V.

NIH 3T3 cells were grown on 60 mm plates to 60–80% confluency and lysed with ice-cold CHAPS buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, 5 mM EGTA, 1% CHAPS, 1 mM DTT) and protease inhibitor cocktail (Sigma-Aldrich). Lysates were transferred to Eppendorf tubes, incubated on ice for 15 minutes and centrifuged at 16,000 g at 4°C for 15 minutes. The supernatants were incubated at 4°C with MBP–Lgl1-C phospho-mutant protein fragments bound to amylase beads, followed by three washes with CHAPS buffer. Samples were separated on 7% SDS-PAGE, and western blotting was performed using anti-mouse NMII-A, anti-aPKCζ and anti-Lgl1 antibodies.

pGFP-Lgl1 phospho-mutants or pGFP-Lgl1^WT and HA-Par6α were co-transfected into an ∼70% confluent HEK293T cell line on 100-mm plates using polyethylenimine (PEI) transfection reagent (Sigma-Aldrich). Forty-eight hours after transfection the cells were washed three times with ice-cold PBS and lysed with 1 ml NP-40 buffer (10 mM Tris-HCl, pH 8, 1 mM EDTA, 100 mM NaCl, 1 mM DTT and 0.5% NP-40) containing protease inhibitor cocktail (Sigma-Aldrich). Cells were then sonicated, incubated for 15 minutes on ice, and centrifuged at 25,000 g at 4°C for 15 minutes. 50 µl of the supernatants were used for input control, the remainder of the supernatants were transferred to fresh tubes containing protein A agarose beads (Pierce, Rockford, IL) coupled to anti-human NMII-A or anti-GFP antibodies and protein G agarose beads (Pierce, Rockford, IL) coupled to anti-HA antibodies, rotated at 4°C for 2 hours, followed by four washes with NP-40 buffer. Samples were dissolved in SDS-PAGE sample buffer, separated using 7% SDS-PAGE and subjected to western blotting using anti-human NMII-A, anti-GFP, anti-aPKCζ and anti-HA antibodies.

Control cells (6×10⁵) were transfected with GFP-Lgl1^WT, GFP-Lgl1^Ala or GFP-Lgl1^Asp or HA-Par6α and 6×10⁵ aPKCζ^−/− cells were transfected with GFP-Lgl1^WT using linear polyethylenimine (L-PEI) transfection reagent (PolyScience). Cells were then seeded on coverslips coated with 27 µg/ml collagen type I (Sigma-Aldrich). At 48 hours post-transfection three parallel scratches were made with a small pipette tip, and the cells were washed twice in PBS to remove cell debris. High-glucose DMEM medium was then added, and cells were incubated for 7 hours, after which they were washed three times with PBS and fixed for 20 minutes in 1.5 ml 3.7% formaldehyde in PBS. aPKCζ^−/− and control cell lines without transfection were treated in the same way and were immunostained with anti-mouse NMII-A antibody and secondary antibody conjugated to Cy5. aPKCζ was stained with anti-mouse monoclonal aPKCζ antibodies, HA–Par6α was stained with anti-rat monoclonal HA antibodies. F-actin was stained with Rhodamine–phalloidin as described previously (Ronen and Ravid, 2009). Cells were viewed using confocal microscopy.

aPKCζ^−/− (6×10⁵) and control cells (8×10⁵) were transfected with GFP–Lgl1 phospho-mutant proteins using linear polyethylenimine (L-PEI) transfection reagent (PolyScience). Both transfected and untransfected aPKCζ^−/− and control cells were seeded on 30 mm dishes, washed twice with 1 ml PBS and lysed by adding 200 µl PEM buffer (100 mM PIPES pH 6.9, 1 mM MgCl₂, 1 mM EGTA) with 1% Triton-X-100 and a protease inhibitor mix (Sigma). This was followed by incubation on ice for 5 minutes. The Triton-soluble fractions were collected from the plates and centrifuged for 5 minutes at 16,000 g to remove the remnants of the insoluble fraction. 100 µl of supernatant was removed to fresh tubes, 25 µl of 5× SDS-PAGE sample buffer was added, and the tubes were boiled for 5 minutes at 100°C. The insoluble fraction was washed once with 300 ml PEM buffer, then 120 µl SDS-PAGE sample buffer was added to the plates and the insoluble fraction was collected and boiled for 5 minutes at 100°C. After separation on 8% SDS-PAGE, western blot analysis was performed as described above using anti-mouse NMII-A and anti-Lgl1 antibodies. The western blots were developed using the EZ-ECL Chemiluminescence Detection kit (Biological Industries) and the intensities of the bands were analyzed using Fujifilm ImageGauge V software. In the final calculations of the percentage of the proteins in the soluble fractions, the amount of the proteins in the Triton-soluble fraction was corrected by a factor of two and the intensity of proteins in the Triton-soluble fraction was divided by the sum of the intensities of the proteins in the Triton-soluble and -insoluble fractions.

We thank Tony Pawson (Mount Sinai Hospital, Toronto, Canada) for Lgl1 and Par constructs and Jorge Moscat (Sanford-Burnham Medical Research Institute, La Jolla) for aPKCζ knockout MEF cells. S.R. holds the Dr Daniel G. Miller Chair in Cancer Research.