Prickle is known to be involved in planar cell polarity, including convergent extension and cell migration; however, the detailed mechanism by which Prickle regulates cellular functions is not well understood. Here, we show that Prickle1 regulates front-rear polarization and migration of gastric cancer MKN1 cells. Prickle1 preferentially accumulated at the cell retraction site in close proximity to paxillin at focal adhesions. Prickle1 dynamics correlated with those of paxillin during focal adhesion disassembly. Furthermore, Prickle1 was required for focal adhesion disassembly. CLASPs (of which there are two isoforms, CLASP1 and CLASP2, in mammals) and LL5β (also known as PHLDB2) have been reported to form a complex at cell edges and to control microtubule-dependent focal adhesion disassembly. Prickle1 was associated with CLASPs and LL5β, and was required for the LL5β-dependent accumulation of CLASPs at the cell edge. Knockdown of CLASPs and LL5β suppressed Prickle1-dependent cell polarization and migration. Prickle1 localized to the membrane through its farnesyl moiety, and the membrane localization was necessary for Prickle1 to regulate migration, to bind to CLASPs and LL5β, and to promote microtubule targeting of focal adhesions. Taken together, these results suggest that Prickle1 promotes focal adhesion disassembly during the retraction processes of cell polarization and migration.
Prickle was first identified as a protein that regulates planar cell polarity (PCP) in Drosophila imaginal discs (Gubb et al., 1999). Loss of Prickle results in phenotypes affecting the stereotypical arrangement of sensory bristles and cellular hairs on the wing, abdomen and thorax, as well as ommatidia in the eye. These are similar to phenotypes resulting from loss of dishevelled (Dsh) and frizzled (fz), which are also known to encode core PCP proteins (Devenport, 2014; Veeman et al., 2003a; Yang and Mlodzik, 2015). In addition to these genes, flamingo (Fmi), Strabismus (Stbm) and Diego (Dgo) are required for acquisition of epithelial polarity in the Drosophila eye, wing and epidermis. At the cellular level, asymmetric localization of core PCP proteins at the apical cortex is important to establish PCP (Peng and Axelrod, 2012; Strutt, 2002). Fz, Dsh and Dgo concentrate on the one face of a cell, and Stbm and Prickle1 concentrate on the opposite; Fmi is present at both. Stbm is required for the recruitment of Prickle1 at proximal cell ends in the pupa wing and promotes the degradation of excess Prickle1 to maintain its asymmetrical localization (Strutt et al., 2013). Arf1 and adaptor protein-1 (AP-1) are required for the planar polarized enrichment of core PCP proteins along the proximal–distal axis (Carvajal-Gonzalez et al., 2015).
These core PCP proteins are conserved in vertebrates, functioning as regulators of cell morphology and behavior – in processes such as polarized cilia localization, sensory hair polarization, body hair orientation, neural tube closure and long bone cartilage elongation – during tissue organization (Goodrich and Strutt, 2011; Gray et al., 2011; Simons and Mlodzik, 2008; Singh and Mlodzik, 2012; Zallen, 2007). The specific function of core PCP proteins in vertebrates is to regulate the convergent extension movement during gastrulation that results in anterior–posterior elongation of the body axis (Gray et al., 2011; Heisenberg et al., 2000). Furthermore, loss-of-function Wnt5 mutations or Prickle1 knockdown in zebrafish as well as Wnt5a knockdown in Xenopus impair convergent extension and axis elongation in embryos (Kilian et al., 2003; Veeman et al., 2003b), suggesting that Wnt5 and Prickle1 cooperatively regulate the vertebrate PCP pathway.
Among the four Prickle (Prickle1 to Prickle4) proteins found in mice and humans, Prickle1 and Prickle2 have been relatively well studied. Prickle1 and Prickle2 have three conserved LIM domains, a PET domain and a C-terminal farnesylation site (Maurer-Stroh et al., 2007; Sweede et al., 2008). Both Prickle1 and Prickle2 are expressed mainly in neuronal cells during mouse embryogenesis (here, the mouse genes have been referred to as mpk1 and mpk2, respectively) (Okuda et al., 2007; Tissir and Goffinet, 2006). Knockout mouse studies have revealed that mpk1 is required for the maintenance and establishment of epiblast apical and basolateral polarity (Tao et al., 2009). Knockdown and overexpression of Prickle1 reduces and induces, respectively, neurite outgrowth of mouse neuroblastoma cells (Fujimura et al., 2009; Okuda et al., 2007), suggesting that Prickle1 might be involved in neuron polarization; however, these cells do not show the polarity characteristics of neurons. Prickle2 localizes to the postsynaptic density of asymmetric synapses in the adult mouse brain and forms a complex with PSD-95 and NMDA receptors (Hida et al., 2011). Furthermore, loss-of-function Prickle mutations in flies or Prickle1 mutations in mice and humans result in epileptic phenotypes (Bassuk et al., 2008; Tao et al., 2011). Prickle organizes microtubule polarity, thereby influencing axon growth in Drosophila neurons (Ehaideb et al., 2014). However, it is not clear whether these Prickle functions are associated with Wnt signaling.
Prickle1 is also expressed in non-neural tissues, such as in the developing limbs of mice (Bekman and Henrique, 2002). Disrupting Prickle1 function in mice results in altered gene expression in skeletal condensates and is associated with an altered apoptosis pattern in digit tips and interdigital membranes, resulting in shorter and wider bones with disrupted chondrocyte polarity (Yang et al., 2013). Therefore, Prickle1 might have previously unknown functions in regulating polarity and migration in non-neuronal mammalian cells. Xenopus Prickle2 in multiciliated epithelial cells controls the asymmetric localization of a subset of core PCP family proteins (Butler and Wallingford, 2015). The underlying mechanism of Prickle-mediated non-neuronal polarity and migration, however, remains to be defined.
Cell migration is indispensable for animal development and tissue remodeling. Cell migration is dependent on traction forces generated within cells. Focal adhesions, which consist of integrins and adaptor proteins, connect the cytoskeleton to the extracellular matrix and transmit intracellular forces to the cell exterior (Parsons et al., 2010). These forces promote focal adhesion assembly, resulting in mature focal adhesions (Gardel et al., 2010). For cells to polarize and to migrate, focal adhesions must be disassembled. Microtubules play an essential role in focal adhesion disassembly during migration (Stehbens and Wittmann, 2012). Microtubule plus-end-tracking proteins (+TIPs) bind to the growing microtubule plus end (Schuyler and Pellman, 2001). CLIP-associating proteins (CLASPs, of which there are two isoforms CLASP1 and CLASP2 in mammals) play an essential role in the regulation of microtubule dynamics (Akhmanova et al., 2001; Akhmanova and Steinmetz, 2008) and tether microtubules to focal adhesions, facilitating focal adhesion disassembly (Stehbens et al., 2014). LL5β (also known as PHLDB2) localizes to the cell membrane through its pleckstrin homology (PH) domain, and directly binds to CLASPs, resulting in the recruitment of CLASP-bound microtubules to focal adhesions (Dowler et al., 2000; Lansbergen et al., 2006). Here, we demonstrate that Prickle1 is involved in focal adhesion disassembly in cooperation with the CLASP–LL5β complex, thereby regulating cell polarity and migration.
Prickle1 localizes adjacent to focal adhesions and promotes cell migration
Because we have previously used MKN1 gastric cancer cells as a model to analyze cell migration and focal adhesion turnover (Kurayoshi et al., 2006; Yamamoto et al., 2009), morphological changes of MKN1 gastric cancer cells were monitored using time-lapse analyses. MKN1 cells formed lamellipodia randomly until 30 min after cell seeding, after which cells began to polarize. Polarization was completed between 60 and 70 min post-seeding, at which point lamellipodia consistently formed at one location (Fig. 1A). One hour after seeding, approximately 30% of MKN1 cells showed a polarized morphology in which actin and stress fibers accumulated at the periphery of the cells. Caveolin then accumulated in a linear manner opposite to the site of actin accumulation (Fig. S1A). Actin and caveolin are considered to mark front protrusion (leading) and rear retraction (trailing) edges, respectively (Gardel et al., 2010; Navarro et al., 2004). Thus, these cells had a single front protrusion and single rear site. This morphological type of cell was designated as being polarized here. The remaining 70% of the MKN1 cells exhibited pleomorphic shapes, and those cells showed multiple protrusions and retraction sites. Caveolin localized to punctate structures in the central region of the cells (Fig. S1A). In this study these cells were designated as non-polarized cells.
Knockdown of Prickle1 with small interfering RNA (siRNA) #1, which targets the 3′-untranslated region of PRICKLE1 mRNA, decreased the ratio of polarized to non-polarized cells, whereas stable expression of hemagglutinin (HA)-tagged Prickle1 increased this ratio (Fig. 1B,C; Fig. S1B,C). Prickle1 siRNA #2, which targets the open reading frame of PRICKLE1 mRNA, also suppressed cell polarization to a similar level as Prickle1 siRNA #1 (Fig. S1C). Furthermore, stable expression of HA–Prickle1 rescued the phenotype induced by Prickle1 siRNA #1, indicating that siRNAs against Prickle1 did not induce off-target effects (Fig. 1C).
Random migration of single cells was analyzed with time-lapse imaging (Pankov et al., 2005). All trajectory paths of at least 60 randomly selected cells are shown together; the beginning of each cell path was placed at the intersection of the x and y axes, and black dots indicate the ends of the cell paths (Fig. 1D). Euclidean distance (linear distance between start and end position) was calculated as an indicator of cell migration activity. Knockdown of Prickle1 by siRNA #1 decreased MKN1 cell migration (43.6% decrease in mean Euclidean distance); however, HA–Prickle1-expressing cells migrated faster than control cells (Fig. 1D) (106.2% increased in mean Euclidean distance). The accumulated distance (total distance of trajectory path between start and end position) was also decreased and increased by Prickle1 knockdown and HA–Prickle1 expression, respectively; however, there was no clear difference in directionality (Fig. 1D). Prickle1 siRNA #2 also suppressed cell migration to a similar level as Prickle1 siRNA #1 (Fig. S1D). HA–Prickle1 expression rescued the decrease in migration induced with Prickle1 knockdown using siRNA #1 (Fig. 1D) (144.1% increase in mean Euclidean distance). Taken together, these data suggest that Prickle1 is involved in the polarization and migration of MKN1 cells.
To understand the role of Prickle1 in cell migration, subcellular localization of Prickle1 was examined. In the cell retraction site of non-polarized cells, where caveolin had accumulated, HA–Prickle1 was observed as small punctate structures and accumulated in localized areas adjacent to paxillin in focal adhesions (Fig. 1E). HA–Prickle1 was rarely observed to be near focal adhesions in the protrusion site, where F-actin was enriched, cortactin was present and lamellipodia were formed (Fig. 1E; Fig. S1E). In polarized cells, HA–Prickle1 accumulated in a linear manner along the retraction site and partially localized close to focal adhesions (Fig. S1F). Close localization of Prickle1 to focal adhesions in the retraction site was more evident in non-polarized cells, probably because focal adhesion structures were highly dynamic and transient in polarized cells. HA–Prickle1 localization adjacent to focal adhesions was observed to a similar extent in HeLaS3 cervical cancer cells (Fig. S1G). Endogenous Prickle1 was immunohistochemically observed adjacent to focal adhesions in U251-MG malignant glioblastoma cells, which expressed higher levels of Prickle1 as compared to MKN1 cells (Fig. 1F).
Mouse Prickle1 (mpk1) and Prickle2 (mpk2) are expressed in embryos at an early developmental stage, and knockout of the mpk1 gene results in early embryonic lethality (Tao et al., 2009). In contrast, mpk2-null mice do not show a strong phenotype (K. Minegishi, H. Hamada et al., Graduate School for Frontier Biosciences, Osaka University, Osaka, Japan, personal communications). To confirm the role of Prickle1 in cell migration, a pair of mpk1+/−;mpk2−/− mice were crossed, and mouse embryonic fibroblasts (MEFs) were prepared from the resulting embryos. mpk1+/−;mpk2−/− MEFs and mpk1−/−;mpk2−/− MEFs had reduced Euclidean distances (31.2% and 34.9% decrease, respectively, expressed as mean Euclidean distances for comparison) as compared with mpk1+/+;mpk2−/− MEFs (Fig. S2A). When these MEFs were subjected to a wound healing assay, mpk1+/−;mpk2−/− MEFs and mpk1−/−;mpk2−/− MEFs showed reduced migration as compared to mpk1+/+;mpk2−/− MEFs (Fig. S2B). HA–Prickle1 also localized close to paxillin in MEFs, similar to the localization pattern seen in MKN1 and HeLaS3 cells (Fig. S2C). These data demonstrate that the role of Prickle1 in polarized cell migration is not limited to MKN1 cells.
Previous studies have revealed that Wnt5a, and van gogh 1 and van gogh 2 (VANGL1 and VANGL2, respectively; vertebrate homologs of strabismus) are genetically linked to PRICKLE1 in PCP phenotypes (Kilian et al., 2003; Veeman et al., 2003b). Wnt5a, Vangl1 and Vangl2 were knocked down using siRNAs to determine whether they play a role in Prickle1 in localization in MKN1 cells (Fig. S3A). Additionally, MKN1 cells were treated with IWP2 to block the secretion of endogenous Wnt proteins, resulting in inhibition of Dvl2 (one of the mammalian homologs of Dsh) phosphorylation (Chen et al., 2009) (Fig. S3B). Prickle1 localization adjacent to focal adhesions was unaffected by Wnt5a knockdown, Vangl1 and Vangl2 knockdown, or IWP2 treatment (Fig. S3C,D). Thus, Prickle1 localization close to focal adhesions and Prickle1-induced migration might be independent of Wnt signaling and the PCP pathway.
It has been demonstrated that in Drosophila Prickle and Dsh mutually inhibit localization and function of each other (Jenny et al., 2005; Tree et al., 2002), and that in mice Smurf ubiquitin ligases bind to Dvl and degrade Prickle proteins, thereby controlling convergent extension and cochlea hair cell polarity (Narimatsu et al., 2009). Knockdown of all Dvl proteins (Dvl1, Dvl2 and Dvl3) neither affected the close localization of HA–Prickle1 to focal adhesions nor inhibited Prickle1-dependent migration in HA–Prickle1-expressing MKN1 cells, although knockdown of the three Dvl proteins suppressed migration of control MKN1 cells (Fig. S3E,F). Because it is well known that Dvl regulates the cytoskeleton through Rho and Rac, it is reasonable to assume that Dvl is involved in basal levels of cell migration activity. Transiently expressed FLAG–Dvl2 showed small punctate structures, and they did not affect the localization of HA–Prickle1 (Fig. S3G). Thus, it is likely that Dvl proteins are not involved in the Prickle1 functions observed in this study.
Prickle1 promotes focal adhesion turnover
Next, the role of Prickle1 in focal adhesion turnover was interrogated. GFP–paxillin was expressed in MKN1 cells, and its dynamics were visualized in red, green, blue and magenta at 0, 15, 30 and 45 min after the start of imaging respectively (Fig. 2A). The images showing the basal surface of the cell were superimposed, and dynamic adhesions were defined as areas containing more than one color (Yamana et al., 2006). In Prickle1-depleted MKN1 cells, the focal adhesion size was enlarged, and areas of paxillin localization are shown in white, indicating that focal adhesion dynamics were decreased (Fig. 2A). In contrast, focal adhesion dynamics were increased in HA–Prickle1-expressing MKN1 cells (Fig. 2A). These data indicate that Prickle1 plays a positive role in focal adhesion turnover. mCherry–Prickle1 was localized to projection-like structures at the cell periphery; these structures readily protruded and retracted during the observation period. A kymograph assay revealed that mCherry–Prickle1 accumulated at cell retraction sites but not to protrusion sites (Fig. 2B; Movie 1), suggesting that Prickle1 is involved in the retraction process at the cell periphery.
In the cell periphery of non-polarized cells, GFP–paxillin-positive focal adhesions gradually turned over for up to about 40 min after the start of the retraction process (Fig. 2C). mCherry–Prickle1 showed a similar turnover time course (Fig. 2C), suggesting that the appearance and loss of Prickle1 is associated with those of paxillin at the retraction sites. In comparison with fixed cells, mCherry–Prickle1 in live cells gave a rather smeared appearance and partially colocalized with GFP–paxillin.
It has been previously shown that microtubules are guided to substrate adhesion sites in order to induce contact release and cell edge retraction (Kaverina et al., 1999). Therefore, the effects of the microtubule depolymerizing agent nocodazole on Prickle1 localization and dynamics were investigated. Nocodazole treatment stabilized and enlarged sites of GFP–paxillin, which was not lost from sites until about 105 min after treatment, and mCherry–Prickle1 sites also enlarged and persisted to a similar time as GFP–paxillin (Fig. 2D), suggesting that microtubule disassembly extends the lifetime of Prickle1 and focal adhesions, probably owing to the inhibition of their degradation, and that intact microtubules are not required for Prickle1 assembly but are necessary for Prickle1 disassembly. Rho, Rho kinase and myosin-mediated cell contractility are required for the maintenance of mature focal adhesions (Katoh et al., 2001; Rottner et al., 1999). Treatment with the Rho-kinase inhibitor Y-27632 led to the rapid loss of GFP–paxillin and mCherry–Prickle1 (Fig. 2E). Y-27632 treatment after nocodazole treatment also induced disassembly of GFP–paxillin and the associated mCherry–Prickle1 (Fig. 2D). These results suggest that Prickle1 assembly depends on focal adhesion structures and that intact microtubules are not required for Prickle1 disassembly when actomyosin contraction has been inhibited. Thus, the data indicate that Prickle1 dynamics are associated with focal adhesion turnover.
Prickle1 forms a complex with CLASPs and LL5β
As shown in Fig. 1E, HA–Prickle1 closely localized but did not completely overlap with paxillin. Prickle1 localization seemed to be similar to that of CLASPs and LL5β (Lansbergen et al., 2006). In addition, focal-adhesion-associated localization of CLASPs and LL5β has been shown to depend on focal adhesion structures, similar to Prickle1 localization (Stehbens et al., 2014). CLASPs have been shown to associate with the plus end of growing microtubules and to attach to the cell cortex through the interaction with LL5β, and LL5β-mediated CLASP recruitment facilitates focal adhesion turnover (Lansbergen et al., 2006; Stehbens et al., 2014). HA–Prickle1, endogenous LL5β and endogenous CLASP1 colocalized, and these proteins localized adjacent to GFP–paxillin in focal adhesions (Fig. 3A). Consistent with these results, HA-Prickle1 formed a complex with GFP–CLASP1, GFP–CLASP2 and GFP–LL5β in HEK293T kidney epithelial cells (Fig. 3B). HA–Prickle1 did not form a complex with GFP–LL5α (Fig. 3B). In addition, endogenous Prickle1 bound to GFP–LL5β, and endogenous CLASP1 bound to HA–Prickle1 (Fig. S4A). Deletion mutant analyses showed that the PET-LIM domain of Prickle1 (amino acid residues 1–313) was necessary and sufficient for the interaction between Prickle1 and LL5β or CLASP1 in HEK293T cells (Fig. S4B).
In control MKN1 cells, endogenous CLASP1 and LL5β localized close to paxillin in distal areas of the cell cortex (Fig. 3C). In Prickle1-depleted MKN1 cells, CLASP1 was distributed evenly in distal and sub-distal areas, although LL5β still localized in distal areas, and colocalization of LL5β and CLASP1 was lost (Fig. 3C). In MEFs and MKN1 cells, CLASP1 and LL5β colocalized adjacent to focal adhesions. In mpk1−/−;mpk2−/− MEFs, LL5β accumulated close to focal adhesions in distal areas of the cell cortex. CLASP1 was diffusely distributed in distal and sub-distal areas (Fig. S4C). Knockdown of CLASP1 and CLASP2, or LL5β in MKN1 cells did not affect paxillin localization at focal adhesions (Fig. 3D; Fig. S4D). Knockdown of LL5β, but not of CLASP1 and CLASP2, impaired close localization of HA–Prickle1 to paxillin (Fig. 3D). Therefore, localization of Prickle1 adjacent focal adhesions might play an important role in the formation of a complex between LL5β and CLASPs, and might be required for cortical CLASP localization. Furthermore, knockdown of CLASP1 and CLASP2 or LL5β reduced HA–Prickle1-dependent cell polarization and migration (Fig. 3E,F). Thus, Prickle1 could be involved in focal adhesion turnover through its interaction with the CLASP–LL5β complex.
Proper membrane localization of Prickle1 is necessary for its function
Mouse Prickle1 is modified with a farnesyl moiety at cysteine residue 829 (Cys829) (Fig. 4A). Farnesylation modification is necessary for Prickle localization to cell membranes as well as its function in Drosophila melanogaster wings (Strutt et al., 2013). We hypothesized that Prickle1 farnesylation might play a role in its cortical localization to focal adhesions as well as in Prickle1-mediated regulation of cell migration and polarization. Cell fractionation analysis revealed that HA–Prickle1 is mainly present in the membrane fraction; however, when Cys829 was mutated to serine (HA–Prickle1-C829S), Prickle1 was found in the cytosolic fraction (Fig. 4B). Consistent with these results, immunocytochemical analysis showed that HA–Prickle1-C829S was distributed diffusely throughout the cytosol and was not localized close to focal adhesions (Fig. 4C). HA–Prickle1-C829S expression did not promote cell migration or polarization (Fig. 4D,E).
To determine whether membrane localization is sufficient for Prickle1 activity, a myristic-acid-binding site (Zhou et al., 1994) was fused to the N-terminal region of HA–Prickle1-C829S to generate Myr–HA–Prickle1-C829S, which was then stably expressed in MKN1 cells. Similar to HA–Prickle1, Myr–HA–Prickle1-C829S was present in the membrane fraction and localized to the entire cell periphery; however, Myr–HA–Prickle1-C829S did not localize to the vicinity of focal adhesions (Fig. 4B,C). Furthermore, Myr–HA–Prickle1-C829S did not promote cell migration or polarization (Fig. 4D,E). Neither HA–Prickle1-C829S nor Myr–HA–Prickle1-C829S associated with GFP–CLASP1 or GFP–LL5β (Fig. 4F). These results suggest that proper membrane localization of Prickle1 adjacent to focal adhesions through C-terminal farnesylation is necessary for cell polarization and migration, as well as its ability to form a complex with CLASPs and LL5β.
Proper membrane localization of Prickle1 is necessary for microtubule targeting to focal adhesions
As shown in Fig. 2C, Prickle1 dynamics were closely correlated with paxillin dynamics. Prickle1 overexpression promoted GFP–paxillin disassembly (Fig. 5A). Microtubules that have been targeted to focal contact sites promote focal adhesion disassembly in living fibroblasts (Kaverina et al., 1999). In non-polarized MKN1 cells, microtubules and focal adhesions were visualized using RFP–tubulin and GFP–paxillin, respectively; a single focal-adhesion-targeting event was defined as the overlap of the microtubule tip with paxillin (Fig. 5B; Movie 2). The frequency of targeting events was assessed within a 20-min period. In control MKN1 cells, the majority of focal adhesions were targeted an average of 9.7 times (Fig. 5C; Fig. S4E). Enforced expression of HA–Prickle1 increased the frequency of targeting events to an average of 14.1 times, whereas that of HA–Prickle1–C829S and Myr–HA–Prickle1-C829S did not affect the frequency (Fig. 5C; Fig. S4E). Prickle1 knockdown reduced the frequency of targeting to an average of 7.5 times from 11.3 times in control; HA–Prickle1 expression rescued this phenotype, as demonstrated by an increase in the number of targeting events to an average of 14.2 times, but Prickle1 mutants only partially rescued the Prickle1 knockdown phenotypes (Fig. 5D; Fig. S4F).
Paxillin is known to be ubiquitylated and degraded during mesodermal cell migration in Xenopus laevis gastrulation (Iioka et al., 2007). It has also been shown that Prickle1 is degraded by the proteasome through the action of an SCF E3 ubiquitin ligase in Drosophila pupal wings (Strutt et al., 2013). Indeed, HA–Prickle1 was rapidly (within 2 h) degraded in MKN1 cells in the presence of the protein synthesis inhibitor cycloheximide; when MKN1 cells were treated with the proteasome inhibitor MG132 in addition to cycloheximide, Prickle1 degradation was suppressed (Fig. S4G). Nocodazole treatment suppressed HA–Prickle1 degradation (Fig. S4H), suggesting that intact microtubules are required for Prickle1 degradation as well as for Prickle1 disassembly.
Overexpression of HA–Prickle1-C829S or Myr–HA–Prickle1-C829S did not affect GFP–paxillin disassembly (see Fig. 5A). These Prickle1 mutants were also degradation resistant (Fig. S4H). Taken together, these data indicate that proper membrane localization of Prickle1 through farnesylation is necessary for the ubiquitin-dependent Prickle1 degradation; thus, Prickle1 degradation, focal adhesion disassembly and cell migration might be functionally linked.
Prickle1 mediates EGF-dependent cell signaling and migration
It has been shown that EGF induces focal adhesion disassembly (Xie et al., 1998). Indeed, EGF promoted paxillin disassembly, and Prickle1 knockdown using siRNA #1 suppressed EGF-mediated paxillin disassembly in MKN1 cells (Fig. 6A). EGF also increased the ratio of polarized cells and promoted MKN1 cell polarization (Fig. 6B,C). Knockdown of Prickle1 using siRNA #1 suppressed EGF-dependent polarization and migration; these phenotypes were rescued by HA–Prickle1 expression (Fig. 6C,D). EGF activated EGF receptor (EGFR), AKT and JNK1 and JNK2 in MKN1 cells, as indicated by their phosphorylation in response to EGF treatment. Prickle1 knockdown using siRNA #1 had no effect on EGFR and AKT phosphorylation, but inhibited JNK1 and JNK2 phosphorylation (Fig. 6E). This suggests that Prickle1 acts downstream of EGFR and AKT, but upstream of JNK1 and JNK2.
Consistent with these results, HA–Prickle1 expression in HEK293T cells activated JNK1 and JNK2 (Fig. 6F). The JNK inhibitor SP600125 suppressed HA–Prickle1-induced polarization, but the inhibitor of p38 MAPKs SB203580 and the MEK inhibitor U0126 did not affect HA–Prickle1-induced polarization (Fig. 6G). Consistent with these results, SP600125 suppressed EGF-dependent polarization (Fig. 6H).
Interestingly, HA–Prickle1-C829S and Myr–HA–Prickle1-C829S also activated JNK1 and JNK2 to a similar extent as HA–Prickle1 (Fig. 6F). However, HA–Prickle1-C829S and Myr–HA–Prickle1-C829S did not rescue the effects of Prickle1 knockdown on of EGF-induced cell polarity (Fig. 6I). These data suggest that proper Prickle1 membrane localization is not necessary for JNK activation but is required for EGF-dependent cell polarization. Thus, these gain- and loss-of-function experiments suggest that EGF-dependent JNK activation through Prickle1 is involved in focal adhesion turnover and cell polarization.
Migrating cells continuously form and disassemble cell–substrate adhesions, not only at the leading edge but also at the center and the trailing edge (Broussard et al., 2008). Once formed, focal adhesions must be released for directional movement in a process termed adhesion turnover (Webb et al., 2002). The coordinated asymmetry of assembly and disassembly is necessary for directional migration, and microtubules play a key role in asymmetric adhesion dynamics (Kaverina et al., 1999). Microtubule targeting to focal adhesions occurs behind the leading edge and at the trailing edge, resulting in focal adhesion disassembly. We found that Prickle1 is located with CLASPs and LL5β adjacent to focal adhesions in the retracting sites, that it promotes front–rear polarization and cell migration, and that microtubules are targeted to focal adhesions through the tertiary complex of Prickle1, CLASP and LL5β. Taken together with the observations that alternative Prickle protein isoforms control the orientation of microtubule network (Olofsson et al., 2014), we suggest Prickle1 is involved in cell migration through microtubule dynamics.
These findings are not confined to MKN1 cells because Prickle1 was observed in punctae close to paxillin in HeLaS3, MEFs and U251 cells. Prickle1 knockout also suppressed MEF migration. The effects of homozygous knockout were similar to those of heterozygous knockout. The exact reason is not known currently, but one possibility is that MEF migration is less dependent on Prickle1 than MKN1 cell migration, and that Prickle1 haploinsufficiency results in the inhibition of migration to a similar level to Prickle1 homozygous knockout. Although Prickle1 is considered to be one of the proteins that regulates the PCP pathway (Devenport, 2014; Gray et al., 2011; Singh and Mlodzik, 2012; Veeman et al., 2003a), our results show that Prickle1 localization close to focal adhesions does not depend on Dvls, Wnt or Vangl1 and Vangl2. Therefore, the regulation of focal adhesion turnover would be a new function of Prickle1, independent of the Wnt and PCP pathway.
Peripheral membrane localization of Prickle1 through farnesylation was dependent on focal adhesion structure and LL5β localization. Focal adhesions recruit CLASPs independently of microtubules through the interaction of CLASPs and LL5β (Lansbergen et al., 2006). This recruitment was dependent on proper membrane localization of Prickle1, as evidenced by the inability of the farnesylation-deficient Prickle1 mutant (Prickle1-C829S) to promote complex assembly and microtubule targeting. Furthermore, N-terminal myristolyation through the addition of a myristic-acid-binding site to Prickle1-C829S (Myr–Prickle1-C829S) did not rescue these functions, indicating that the membrane localization of Prickle1 was not sufficient to form a complex with CLASPs and LL5β, or to target microtubules to focal adhesions. This suggests that C-terminal farnesylation-dependent Prickle1 membrane localization is substantively different from N-terminal myristoylation-dependent localization. Farnesylation-mediated Prickle1 membrane localization might lead Prickle1 to form a tight complex with LL5β to recruit CLASPs, thereby making a tertiary complex of Prickle1, CLASPs and LL5β; alternatively, farnesylation-mediated Prickle1 membrane localization might recruit additional proteins that regulate focal adhesion turnover. LL5β localizes close to focal adhesions, probably through the binding to phosphatidylinositol (3,4,5)-trisphosphate and integrins (Stehbens and Wittmann, 2012), and its localization might not depend on Prickle1. By contrast, CLASP localization close to focal adhesions depends on the interaction with LL5β (Lansbergen et al., 2006) and Prickle1 (this study). Therefore, Prickle1 would support the binding between LL5β and CLASP1 adjacent to focal adhesions, leading to the promotion of microtubule targeting.
Prickle1 localization to the correct membrane position was also required for its degradation, as evidenced by the observed degradation resistance of Prickle1-C829S and Myr–Prickle1-C829S. It has been previously reported that Prickle1 is degraded in a ubiquitin-dependent manner by a Cullin E3 ligase in Drosophila (Cho et al., 2015; Strutt et al., 2013) and by Smurf in mammals (Narimatsu et al., 2009). Strabismus (the invertebrate homolog of vangl proteins) promotes the recruitment of farnesylated Prickle to the membrane, as well as its degradation, in Drosophila (Strutt et al., 2013). Precisely regulated Prickle protein levels establish polarity by modulating internalization and removal of Strabismus and Flamingo (Cho et al., 2015). Our data indicate that LL5β could perform a similar function to Strabismus in mammals. Prickle1 degradation might contribute to the maintenance of the asymmetric distribution of PCP proteins by controlling the amount of Prickle1 protein present at focal adhesions. Our results suggest that Prickle1 degradation is associated with focal adhesion disassembly. Microtubule disassembly stabilized Prickle1, and degradation-resistant Prickle1 mutants were unable to rescue microtubule targeting to focal adhesions in Prickle1-depleted cells. Therefore, proper Prickle1 membrane localization might play an important role in microtubule targeting to focal adhesions and in Prickle1 degradation, leading to focal adhesion disassembly.
It is well established that growth factor receptor signaling and integrin signaling merge on focal adhesions to regulate cellular proliferation, adhesion and migration through the activation of both Src-family kinases and integrin-linked kinases (Dedhar et al., 1999; Parsons and Parsons, 1997). It has also been shown that integrins co-operate with EGFR in several cell types to regulate multiple signaling pathways (Streuli and Akhtar, 2009). In addition, the interaction between Fz PCP signaling and the transcription factor Fos, which acts downstream of EGFR and JNK signaling, have been shown to regulate the Drosophila photoreceptor cell fates and ommatidial polarity (Weber et al., 2008). Consistent with these previous studies, our results revealed that Prickle1 mediates EGF signaling and EGF-dependent polarization and migration. Proper membrane localization of Prickle1 through farnesylation was also required for EGF-dependent polarization. This supports the hypothesis that signaling molecules that regulate EGF-dependent cellular functions are localized together at focal adhesions. JNK has been shown to phosphorylate paxillin and to regulate cell migration (Huang et al., 2003). In the present study, Prickle1 led to JNK activation downstream of EGFR. Therefore, Prickle1-dependent JNK activation might be important for EGF-dependent cell polarization. However, Prickle1-C829S and Myr–Prickle1-C829S were still able to activate JNK. This indicates that Prickle1-dependent JNK activation is not sufficient for cell polarization and migration. Proper membrane localization of Prickle1 therefore must act through additional pathways to control cellular functions.
It has been reported that Prickle1 activates AKT to regulate focal adhesion turnover and promotes breast cancer cell migration through the interaction with the mammalian target of rapamycin complex 2 (mTORC2) (Daulat et al., 2016). Upregulation of Prickle1 in basal breast cancers is correlated with poor prognosis, suggesting that Prickle1 has tumor progressive functions in vivo. To confirm the involvement of Prickle1 in cancer cell migration in vivo, we examined migration of MKN1 cells in skin xenografts using two-photon intravital microscopy. Consistent with previous observations that the majority of cancer cells, other than melanoma or leukemia cells, are immotile in vivo (Clark and Vignjevic, 2015), MKN1 cells were immobile for several hours in skin xenografts (data not shown). In vivo roles of the Prickle1–CLASP–LL5β module in development and tumorigenesis need to be examined in future experiments.
In conclusion, we have identified new functions of Prickle1 in focal adhesion turnover and mammalian cell migration. These functions are dependent on its localization to the retraction site through C-terminal farnesylation. The proper trafficking of Prickle1 to the cell surface is involved in its association with CLASPs and LL5β. Similarly, its proper localization increases the frequency of contact between microtubules and focal adhesions, leading to focal adhesion turnover through Prickle1 degradation. These new functions of Prickle1 could be important for cell polarization and migration.
MATERIALS AND METHODS
Cell culture and transfection
MKN1 gastric cancer cells were kindly provided by Dr Wataru Yasui (Hiroshima University, Hiroshima, Japan). MKN1 cells were grown in RPMI1640 supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (Yokozaki, 2000). HEK293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM):Ham's F12 (1:1) supplemented with 10% FBS, 0.065 g/l penicillin and 0.1 g/l streptomycin. U251-MG cells were maintained in Eagle's minimal essential medium supplemented with 10% FBS, 1% non-essential amino acids (NEAA) and 1 mM sodium pyruvate (NaP). For live-cell imaging, cells were plated on fibronectin-coated glass-bottomed dishes.
MKN1 cells stably expressing HA–Prickle1 and mutants were generated by co-transfecting cells with pPGK-neo and pCGN-HA-Prickle1, pCGN-HA-Prickle1-C829S or pCGN-Myr-HA-Prickle1-C829S. Cells were selected and maintained in medium containing 400 μg/ml G418. To transiently express proteins, cells were transfected with plasmids using Lipofectamine LTX or Lipofectamine 2000 (Life Technologies, Thermo Fisher Scientific) according to the manufacturer's protocol.
For the preparation of MEFs, a pair of mpk1+/−;mpk2−/− mutant mice were crossed, and MEFs were prepared using a 3T3 protocol (Todaro and Green, 1963) from the resulting embryos at E13.5 that were obtained from Drs Katsura Minegishi and Hiroshi Hamada (Graduate School for Frontier Biosciences, Osaka University, Osaka, Japan), and maintained in DMEM supplemented with 10% FBS.
Cells that had been grown on fibronectin-coated glass coverslips were fixed for 10 min at room temperature in phosphate-buffered saline (PBS) containing 4% (w/v) paraformaldehyde. Immunocytochemistry was performed according to a previous paper (Matsumoto et al., 2010). Images were taken with a LSM510, LSM710 or LSM880 confocal microscope (Carl Zeiss, Jena, Germany).
Polarized cells were defined as cells that had formed a single lamellipodium (protrusion) at the leading edge and in which caveolin had localized in a linear manner to the single retraction site, or as cells in which caveolin had localized to the single retraction site in a linear manner but that had no apparent lamellipodium formation. Non-polarized cells were defined as cells that had formed multiple lamellipodia (protrusions) and multiple finger-like retraction sites surrounding the cell periphery.
Cell migration assays
Cells were seeded for 30 min. Time-lapse imaging was performed using an IX81-ZDC microscope (Olympus, Tokyo, Japan). For EGF treatment, 25 ng/ml of EGF was added to growth medium. For inhibitor treatment, inhibitors were added to growth medium 2 h before cell seeding. Images were captured every 5 min for 10 h, and movies were exported using MetaMorph software (Molecular Devices). Cell tracking and data analysis was done using the Manual Tracking and Chemotaxis Tool in ImageJ software [National Institutes of Health (NIH)].
To perform wound healing assays, MEFs were plated onto collagen-coated coverslips. The monolayer of MEFs was then manually scratched with a plastic pipette tip. Cells were washed with PBS three times, then the wounded cell monolayers were allowed to heal for 2, 4, 6 and 8 h in RPMI1640 medium containing 10% FBS (Kobayashi et al., 2006). The wound size was measured to determine the distance traveled by MEFs using AxioVision 22.214.171.124 (Carl Zeiss).
Knockdown of protein expression with siRNAs
siRNA target sequences used in this study are described in Table S2. Cells were transfected with 40 nM siRNA using Lipofectamine RNAiMAX (Life Technologies, Thermo Fisher Scientific) according to the manufacturer's instructions.
For rescue experiments, we expressed HA–Prickle1 (wild type) in MKN1 cells that had been transfected with Prickle1 siRNA #1 targeting the 3′-untranslated region of Prickle1. In the case of Prickle1-knockdown experiments, Prickle1 siRNA #1 was used in all experiments, and Prickle siRNA #2 was used in Fig. 1B and Fig. S1C,D. Transfection efficiency of siRNA was almost 100% in MKN1 cells when it was assessed with Cy3-labeled siRNA (Takara, Tokyo, Japan).
Focal adhesion turnover assay
In Fig. 2A, the images at 0, 15, 30 and 45 min are depicted in red, green, blue and magenta, respectively, and were superimposed. Non-dynamic adhesions were defined as spots appearing in white after superimposition; dynamic adhesions were defined as spots that did not appear in white after superimposition. Analyses were performed according to the previous paper (Matsumoto et al., 2010; Yamana et al., 2006).
In Fig. 2C–E, to determine paxillin and Prickle1 turnover, time-lapse images of MKN1 cells were acquired for 150 min at 3-min intervals. Regions of interest were defined as areas surrounding individual focal adhesions. Where indicated, cells were treated with 10 µM nocodazole and/or 20 µM Y-27632. For turnover dynamics of GFP–paxillin focal adhesions and surrounding mCherry–Prickle1, fluorescence intensity profiles were measured using ImageJ software (NIH) and plotted as a function of time, and were normalized to the maximum fluorescence intensity. For Fig. 2C, normalized intensity data were aligned based on points of maximum intensity for GFP–paxillin (n=20 focal adhesions in Fig. 2C; n=4 focal adhesions in Fig. 2D,E; from one to three cells).
In Fig. 5A, to observe and quantify focal adhesion disassembly, time-lapse images of MKN1 cells were acquired for 5 h at 5-min intervals. Regions of interest were defined as areas surrounding individual focal adhesions. Focal adhesion disassembly time was calculated as the time between when the focal adhesions reached maximum size (≥4 µm) and when focal adhesions had completely disappeared. Quantification focused on those focal adhesions at cell retraction sites.
Imaging of focal adhesion turnover was performed using time-lapse fluorescence spinning-disc microscopy with an Observer.Z1 inverted microscope equipped with a Yokogawa confocal scanner unit CSU-W1 (Yokogawa Electric Corporation, Tokyo, Japan).
Quantification of the distribution of CLASP1 and LL5β in MKN1 cells and MEFs
To examine CLASP1 and LL5β accumulation at distal and sub-distal regions of cells, the mean fluorescence intensity in a 5 nm×5 nm area at the distal (cell edge) and the sub-distal (center between cell edge and nucleus) regions was measured using ZEN software (Carl Zeiss) (n≥17 regions from five cells).
Microtubule and focal adhesion targeting assay
In Fig. 5B, RFP–tubulin was stably expressed in MKN1 cells and MKN1 cells expressing HA–Prickle1, HA–Prickle1-C829S or Myr–HA–Prickle1-C829S cells. Where indicated, cells were transfected with control siRNA or siRNA against Prickle1 together with GFP–paxillin, and then visualized by using spinning-disc microscopy. Images were taken at 30-s intervals for 30 min. A single targeting event was defined as an overlap between microtubules and paxillin. Only focal adhesions at the retraction sites were monitored, and at least 35 focal adhesions were observed and quantified for each cell line and treatment.
Time-lapse imaging was performed for 2 h at 1-min intervals using an IX81-ZDC microscope (Olympus). Kymographs were constructed using MetaMorph software (Molecular Devices).
Materials and chemicals
GFP–CLASP1, GFP–CLASP2, GFP–LL5α, GFP–LL5β, and RFP–tubulin expression vectors were kindly provided by Dr Yuko Mimori-Kiyosue (RIKEN Center for Life Science Technologies, Kobe, Japan). pPGK-neo was provided by Dr Shinji Takada (National Institutes of Natural Sciences, Okazaki, Japan).
Details on primary antibodies used in this study are described in Table S1. Other materials were purchased from commercial sources.
Standard recombinant DNA techniques were used to construct pCGN-HA-Prickle1, pCGN-HA-Prickle1(1-313), pCGN-HA-Prickle1(308-832) and pCDNA/FLAG-Dvl2. pCGN-HA-Prickle1-C829S was generated by introducing a point mutation at amino acid 829 to change the cysteine codon (TGT) into a serine codon (TCT). pCGN-Myr-HA-Prickle1-C829S was generated by inserting the Src myristoylation sequence (5′-GGGAGCAGCAAGAGCAAGC-CCAAGGACCCCAGCCAGCGCGCC-3′; amino acid sequence, GSSKSKPKDPSQRA) between the HA tag and the first ATG sequence of pCGN-HA-Prickle1-C829S.
Protein expression using lentivirus and adenovirus
To generate lentiviruses, lentiviral expression vectors were transfected into HEK293T (Lenti-X 293T) cells with the packaging vectors pCAG-HIV-gp and pCMV-VSV-G-RSV-Rev using FuGENE HD (Roche Applied Science, Basel, Switzerland). To generate MKN1 cells expressing RFP–tubulin, parental cells (5×104 cells/well in a 12-well plate) were transduced with conditioned medium containing lentiviral particles and 10 µg/ml polybrene. The cells were then centrifuged at 1200 g for 1 h, replated and incubated for an additional 24 h.
To generate adenoviruses, the pAd/CMV/GFP-paxillin plasmid was linearized with PacI, phenol-chloroform extracted and then transfected into 293A cells (5×105 cells/well in a 6-well plate) using Lipofectamine 2000 (Life Technologies, Thermo Fisher Scientific). At 36 h after transfection, cells were replated in a 10-cm dish and further incubated until ready for harvest (typically 7 to 10 days after transfection).
MKN1 cells expressing HA–Prickle1 (wild type) or its mutants were suspended in 200 µl of homogenization buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl and 1 mM dithiothreitol) containing 2 µg/ml leupeptin, 2 µg/ml aprotinin and 1 mM PMSF. These suspensions were homogenized using the sonic homogenizer (Ultrasonic homogenizer VP-5S, TAITEC, Saitama, Japan). After total homogenate had been centrifuged at 100,000 g for 30 min, the supernatant was reserved as the cytosol fraction. The precipitate was extracted in 200 µl of Laemmli sample buffer after washing with PBS. These samples were used as the membrane fraction. Aliquots (20 µl each) were probed for HA, transferrin receptor (membrane marker) and β-tubulin (cytosol marker).
mRNA and protein analysis
Quantitative reverse-transcription PCR (RT-PCR) was performed using a StepOne Real-Time PCR system (Applied Biosystems, Life Technologies, Thermo Fisher Scientific). Forward and reverse primers used are described in Table S3. Western blot data shown in the figures is representative of at least three independent experiments.
All experiments were performed at least three times, and data are expressed as mean or mean±s.d. Statistical analyses were performed using a paired Student's t-test. For experiments with more than two conditions, we employed ANOVA test with Bonferroni or Dunnett correction. P-values less than 0.05 were considered statistically significant. Quantification of protein expression by using western blotting was performed using densitometry analysis with ImageJ software (NIH). Protein signals are expressed as relative area and intensity.
The authors would like to thank the Center of Medical Research and Education, Graduate School of Medicine, Osaka University for providing CSU-W1 microscopy system. The authors would also like to thank Drs Yuko Mimori-Kiyosue, Wataru Yasui, Shinji Takada, Katsura Minegishi and Hiroshi Hamada for donating cells, plasmids and Prickle-knockout mice.
B.C.L. and S.M. designed experiments, performed cell experiments and wrote the manuscript. H.Y. performed biochemical analyses of Prickle1 subcellular localization. H.M., J.K. and M.I. performed in vivo imaging analysis. A.K. designed experiments and wrote the manuscript.
This work was supported by Grants-in-Aid for Scientific Research (Japan Society for the Promotion of Science) to A.K. (2013–2015) [grant number 25250018] and S.M. (2013–2014) [grant number 25860211]; by Scientific Research on Innovative Areas (Ministry of Education, Culture, Sports, Science, and Technology) to A.K. (2012–2016) [grant number 23112004]; and by grants from the Uehara Memorial Foundation (2014).
The authors declare no competing or financial interests.