Cdk5 and its substrates, Dcx and p27^kip1, regulate cytoplasmic dilation formation and nuclear elongation in migrating neurons

During organogenesis, newly generated cells from the progenitors are allocated on the basis of their subtypes or properties to construct sophisticated tissue structures. Because tissue stem cells and progenitors are spatially restricted in many organs, including developing brains, differentiating cells need to migrate to their final destinations. Cell migration patterns can be classified according to cell type. Epithelial cells exhibit a collective cell migration, whereas fibroblasts display a mesenchymal single cell migration (Kawauchi, 2012). Nevertheless, these two different types of migrating cells share some common features. Migrating (or leading) cells extend filopodia and lamellipodia, and form a new adhesion to the extracellular matrix, followed by nuclear forward movement. These cellular events are controlled by Rho family small GTPases, such as Rac1, Cdc42 and RhoA (Nobes and Hall, 1995).

By contrast, migrating neurons undergo unique morphological changes. They extend a thick neurite, called a leading process, and form a ‘cytoplasmic dilation’ (also known as a ‘swelling’) at the proximal region of the leading process (Bellion et al., 2005; Schaar and McConnell, 2005). Subsequently, the nucleus elongates and moves into the dilation. The cytoplasmic dilation is a distinctive subcellular domain, observed only in migrating neurons, not in other migrating cells such as fibroblasts and neutrophils (Schaar and McConnell, 2005). Therefore, it is believed that a neuron-specific molecule may control the formation of cytoplasmic dilations. However, such a molecule has not been identified, although a recent study revealed that Rac1 and its interacting protein POSH (Sh3rf1 – Mouse Genome Informatics) are required for dilation formation (Yang et al., 2012).

Cyclin-dependent kinase 5 (Cdk5) is an unconventional CDK protein, because it is predominantly activated in post-mitotic neurons rather than in dividing neural progenitors (Tsai et al., 1993; Kawauchi et al., 2013). In addition, unlike conventional CDKs, Cdk5 activity is not dependent on cyclins, but is controlled by its specific activators, p35 or p39 (Hisanaga and Saito, 2003). Consistent with its high activity in neurons, Cdk5 controls the early phases of neuronal migration in the developing brains mainly through the regulation of microtubules and actin cytoskeleton (Niethammer et al., 2000; Tanaka et al., 2004b; Kawauchi et al., 2006).

In the developing cerebral cortex, neurons exhibit multi-step modes of neuronal migration, of which the locomotion mode is a major one as it covers most of the neuronal path (Rakic, 1972; Nadarajah and Parnavelas, 2002). However, it is difficult to analyze the locomotion mode without dealing with the secondary defects derived from the early phase of migration, because inhibition of many molecules, including Cdk5 and Rac1, leads to defects in the early phase of migration (Ayala et al., 2007; Kawauchi and Hoshino, 2008; Govek et al., 2011). We recently established a novel ex vivo chemical inhibitor assay, which allows us to directly analyze the locomotion mode of neuronal migration (Nishimura et al., 2010). Using this novel technique, we showed that rottlerin, widely used as a specific inhibitor for protein kinase C delta (PKCδ), suppresses the nuclear movement of the locomoting neurons partly via the inhibition of the activity of a novel downstream target molecule, c-jun N-terminal kinase (Jnk) (Nishimura et al., 2010).

In this study, we show that rottlerin treatment disturbs the cytoplasmic dilation formation and nuclear elongation of migrating neurons. Interestingly, however, inhibition of Jnk has little effects on dilation formation. We found that Cdk5/p35 is a novel downstream target of rottlerin, and that Cdk5 is required for the formation of cytoplasmic dilation and nuclear morphological changes, possibly through the regulation of microtubule organization and an endocytic pathway. These observations indicate that Cdk5, which is predominantly activated in post-mitotic neurons, characterizes the neuron-specific migration pattern through regulation of cytoplasmic dilation formation and nuclear elongation.

Our ex vivo chemical inhibitor screens previously found that treatment with rottlerin strongly suppresses the locomotion mode of neuronal migration in slice-cultured cerebral cortices, partly through the inhibition of a novel downstream target, Jnk, but not through inhibition of PKCδ (Nishimura et al., 2010). This led us to investigate the effects of rottlerin on cytoplasmic dilation formation and nuclear morphological changes in the cortical slices. Mouse cerebral cortices were electroporated in vivo with EGFP- and nuclear localizing signal (NLS)-fused DsRed-expressing vectors at embryonic day 14 (E14), and the electroporated brains were subjected to slice culture at E16. Time-lapse analyses showed that the locomoting neurons formed a cytoplasmic dilation at the proximal region of the leading process and subsequently exhibited nuclear elongation to enter the dilation in the control cortical slices (Fig. 1A; supplementary material Fig. S1A,C and Movie 1) (Bellion et al., 2005; Schaar and McConnell, 2005; Tsai et al., 2007). By contrast, in the rottlerin-treated cortical slices, cytoplasmic dilation was greatly perturbed in the locomoting neurons (Fig. 1A; supplementary material Fig. S1B,C and Movie 2). The ratio of cells displaying a cytoplasmic dilation was dramatically reduced 3 h after rottlerin treatment (Fig. 1B). At the same time-point, the nuclei in the rottlerin-treated cells appeared round in shape, whereas many control cells exhibited elongated nuclear morphologies (Fig. 1C). The ratio of length to width of the nuclei was significantly reduced in the rottlerin-treated cortical slices at 3 and 8 h after rottlerin treatment, compared with that of control (Fig. 1D).

Rottlerin has generally been used as a specific inhibitor for PKCδ (Gschwendt et al., 1994; Kang et al., 2013). However, our previous study indicated that, in contrast to rottrelin treatment, knockdown of PKCδ does not affect the locomotion mode of neuronal migration, and that rottlerin also inhibits Jnk activity, which is required for the locomotion (Nishimura et al., 2010). Therefore, we next examined the involvement of Jnk in the formation of the cytoplasmic dilation.

We constructed a short hairpin RNA (shRNA)-based knockdown vector for Jnk1-coding sequence (Jnk1-sh788), and confirmed that this shRNA efficiently reduced endogenous Jnk1 protein levels in primary cortical neurons 2 days after transfection (supplementary material Fig. S2A). We then electroporated Jnk1-sh788 into mouse embryonic cerebral cortices at E14 and the electroporated brains were subjected to slice culture at E16. The Jnk1 knockdown resulted in delayed nuclear movements and abnormal leading process morphologies in locomoting neurons (Fig. 1E and supplementary material Fig. S3E), as previously reported (Kawauchi et al., 2003; Nishimura et al., 2010). To our surprise, however, these locomoting neurons formed a cytoplasmic dilation to a similar extent to that in control, although their morphologies were rough and irregular in parts (Fig. 1F). By contrast, their nuclear elongation was disrupted, which was rescued by co-expression of human wt-Jnk1 (Fig. 1G). In our ex vivo chemical inhibitor assay, a Jnk inhibitor, SP600125, also showed similar results to Jnk1 knockdown (supplementary material Fig. S3A-D). These results suggest that Jnk is required for the nuclear elongation and forward movement but is dispensable for dilation formation in the locomoting neurons. It further suggests that other rottlerin-target molecule(s) may be involved in the formation of the cytoplasmic dilation.

We subsequently searched for a novel molecular pathway suppressed by rottlerin treatment, and found that rottlerin decreased the phosphorylation of focal adhesion kinase (FAK) at Ser732 (Fig. 2A). It is known that the Ser732 residue on FAK is specifically phosphorylated by Cdk5, because Ser732 phosphorylation disappears in the Cdk5-knockout cerebral cortex (Xie et al., 2003). Furthermore, protein levels of a Cdk5-specific activator, p35, which is also phosphorylated by Cdk5 (Asada et al., 2012), are reduced in the rottlerin-treated cortical neurons (Fig. 2A). By contrast, treatment with bisindolylmaleimide I (BIM), a chemical inhibitor for all PKC proteins, including PKCδ, did not affect either Ser732 phosphorylation of FAK or p35 protein levels, suggesting that rottlerin inhibits a molecule upstream of Cdk5/p35, independently of PKCδ suppression. These data suggest that rottlerin indirectly decreases Cdk5 activity. Our previous data showed that inhibition of Cdk5 does not decrease Jnk activity (Kawauchi et al., 2005), but suppresses the locomotion mode of neuronal migration (Nishimura et al., 2010), suggesting that Cdk5 may have a function in locomoting neurons in a Jnk-independent manner.

To elucidate the role of Cdk5 in locomotion, we applied a CDK inhibitor, roscovitine (Meijer et al., 1997), in our ex vivo chemical inhibitor assay. Cytoplasmic dilation formation in neurons in the roscovitine-treated cortical slices was severely affected (Fig. 2B; supplementary material Fig. S4 and Movie 3). The ratio of the cells with dilations was significantly reduced 3 h after roscovitine treatment (Fig. 2C), and nuclei exhibited round morphologies (Fig. 2D,E). More-detailed analyses revealed that control locomoting neurons exhibited a saltatory movement, as previously reported (Edmondson and Hatten, 1987; Schaar and McConnell, 2005); after the peak of the migration speed per unit, the sphericity of the nucleus was increased, i.e. the nucleus became rounded (Fig. 3A). By contrast, although the migration speed of the neurons in the roscovitine-treated cortical slices remained constantly low, the nuclear sphericity increased continuously from 30 min after roscovitine treatment (Fig. 3A-C). The nuclear sphericity gradually increased as nuclear movement abated (Fig. 3A), but even 30 min after roscovitine treatment, the rounding of the nucleus was significantly higher than that of control (Fig. 3B), suggesting that at least in part, Cdk5 actively regulates nuclear morphological changes. This is consistent with a previous report that Cdk5-mediated phosphorylation of Ser732 on FAK regulates the nuclear morphological changes (Xie et al., 2003).

It has been reported that roscovitine is a reliable specific inhibitor for CDK proteins (Meijer et al., 1997; Bain et al., 2003,, 2007), but from these experiments it is unclear whether the effects of roscovitine are cell-autonomous or not. To test this and confirm our inhibitor experiments, we performed knockdown experiments using Cdk5-shRNA-expressing vectors (Cdk5-sh250) (Kawauchi et al., 2006) (supplementary material Fig. S2B). Because Cdk5 also has important roles in the early phase of neuronal migration (Kawauchi et al., 2006; Ohshima et al., 2007), most Cdk5-knockdown neurons are unable to transition into the locomotion mode, but a small fraction enter the cortical plate exhibiting locomoting morphologies (Fig. 4A,F). These Cdk5-knockdown locomoting neurons showed delayed migration as we have previously reported (Nishimura et al., 2010), suggesting that the Cdk5 activity is at least partially suppressed in these neurons even after the transition to the locomotion mode (Fig. 4A). Therefore, we analyzed the morphologies of these Cdk5-knockdown locomoting neurons. Consistent with the inhibitor experiments, the ratio of the cells with a cytoplasmic dilation was dramatically reduced and nuclear elongation was significantly disturbed in the Cdk5-sh250-electroporated cortical slices; these phenotypes were rescued by the co-expression of human wt-Cdk5 (Fig. 4A-D). The area occupied by leading processes 10 µm from the cell soma, which corresponds to the dilation, was decreased in the Cdk5, but not Jnk1, knockdown neurons (Fig. 4E). These data suggest that Cdk5 cell-autonomously regulates the formation of cytoplasmic dilation and nuclear morphological changes.

To confirm these results in vivo, we examined the involvement of Cdk5 in dilation formation in fixed cortical sections. Cdk5-sh250 was electroporated into embryonic cerebral cortex at E14, and the electroporated brains were harvested and examined at E17. A small fraction of Cdk5-knockdown neurons exhibited locomoting morphologies, but very little cytoplasmic dilation was observed (Fig. 4F,G). Taken together, these results indicate that Cdk5, but not Jnk, is essential for formation of the cytoplasmic dilation.

We next examined the events downstream of Cdk5 that might regulate dilation formation in locomoting neurons. Previous electron microscopy analyses showed that a dilation/swelling contains Golgi apparatus, clathrin-coated pits and abundant microtubules elongating from a centrosome (Bellion et al., 2005; Schaar and McConnell, 2005; Shieh et al., 2011). Cdk5-knockdown or roscovitine-treated primary cortical neurons exhibited abnormal microtubule orientations surrounding the nuclei (Fig. 5A; supplementary material Figs S5A,B and S6). Roscovitine seemed to perturb Golgi localization in primary cortical neurons. Furthermore, the distance between the electroporated Golgi-localized EYFP and the nucleus was decreased in the locomoting neurons in the roscovitine-treated cortical slices (supplementary material Fig. S5C,D). The distance between the GM130-positive cis-Golgi and nucleus was also decreased in fixed cortical sections, possibly owing to lower motility of the Golgi in the Cdk5-knockdown locomoting neurons (Fig. 5B and supplementary material Fig. S5E).

We next examined the involvement of Cdk5 in clathrin-mediated endocytic pathways by using a transferrin (Tf) uptake assay. It is known that Tf that is internalized via clathrin-mediated endocytosis is transported to Rab5-positive early endosomes and subsequently to Rab11-positive perinuclear recycling endosomes (Ullrich et al., 1996; Trischler et al., 1999). Primary cortical neurons were transfected with Cdk5-sh250 or control vectors, and treated with Alexa594-conjugated Tf (Tf-594) for 10 or 30 min. By 10 min after the treatment, Tf-594-mediated signals displayed dot-like patterns in both control and Cdk5-knockdown neurons (Fig. 5C, upper panels). While Tf-594 was transported to perinuclear regions 30 min after Tf-594 treatment in control neurons, as previously reported (Kawauchi et al., 2010), Tf-594-mediated signals remained as dot-like patterns 30 min after treatment in Cdk5-knockdown neurons (Fig. 5C, lower panels). The ratio of cells with perinuclear accumulation was significantly decreased in Cdk5-knockdown neurons (Fig. 5D). The dot-like staining in Cdk5-knockdown neurons 30 min after Tf-594 treatment was partially colocalized with transfected EGFP-Rab5, a marker for early endosomes, suggesting that Cdk5 regulates the endocytic pathway at early endosomes (Fig. 5E).

These results were supported by inhibitor experiments. Primary cortical neurons, transfected with EGFP-Rab5 as a marker for early endosomes, were treated with roscovitine or control solvent for 60 min, and additionally treated with Tf-594 for 10 or 30 min. Similar to the knockdown experiments, roscovitine treatment suppressed the Tf-594 transport (supplementary material Fig. S7). These results suggest that suppression of Cdk5 eventually results in inhibition of the endocytic trafficking pathway from clathrin-mediated endocytosis to the perinuclear endosomes in embryonic cortical neurons.

We have previously reported that Rab5-dependent endocytosis, comprising almost all clathrin-mediated endocytosis, and Rab11-dependent recycling are essential for neuronal migration (Kawauchi et al., 2010), but the roles of endocytic pathways in the cytoplasmic dilation formation and nuclear elongation remained unclear. Because in vivo electroporation-mediated inhibition of Dynamin results in cell death 3 days after electroporation (Kawauchi et al., 2010), we used an ex vivo chemical inhibitor assay to bypass the abnormalities at the early phase of neuronal migration.

Dynasore is a specific inhibitor for dynamin 1 and dynamin 2, both of which are essential for many types of endocytosis, including clathrin-mediated endocytosis (Macia et al., 2006). When cortical slices were treated with 80 µM of dynasore, we found that the locomotion mode of neuronal migration was strongly inhibited (data not shown). However, at this concentration, dynasore has been reported to inhibit a mitochondrial dynamin-related protein, Drp1, as well as endocytic dynamin 1 and dynamin 2 (Macia et al., 2006). Therefore, in this study, we used a lower concentration of 40 µM of dynasore to suppress endocytic trafficking.

Cerebral cortices were electroporated with EGFP- and NLS-fused DsRed-expressing vectors at E14, and the electroporated brains were subjected to slice culture at E16. Treatment with dynasore delayed the nuclear movement in the locomoting neurons (Fig. 6A,B), as previously reported (Wilson et al., 2010), and also decreased both the ratio of cells with a cytoplasmic dilation and the ratio of length to width of the nuclei (Fig. 6C-E; supplementary material Movie 4). Furthermore, knockdown-mediated inhibition of Rab5 also disturbed dilation formation and nuclear elongation, both of which were rescued by co-expression of human wt-Rab5 (Fig. 6F-H). These data suggest that endocytic pathways are required for the formation of cytoplasmic dilation and nuclear elongation during the locomotion mode of neuronal migration.

To examine the involvement of microtubule organization in dilation formation, cortical slices were treated with 100 nM nocodazole, an inhibitor of microtubule polymerization. Nocodazole treatment resulted in decrease of the ratio of cells with dilation and with elongated nucleus, suggesting that microtubule organization as well as endocytic trafficking are required for dilation formation and nuclear elongation during the locomotion mode of migration (Fig. 7).

Because F-actin is concentrated at the proximal region of the leading process in migrating cerebellar granule neurons (Solecki et al., 2009), we analyzed actin dynamics using time-lapse imaging of the electroporated EGFP-actin. In control locomoting neurons, strong EGFP-actin signals were observed above the nucleus or proximal region of the leading process before nuclear movement (supplementary material Fig. S8), consistent with a previous report (Solecki et al., 2009). By contrast, strong EGFP-actin accumulation was hardly detected in the Cdk5-knockdown neurons. However, both control and Cdk5-knockdown neurons sometimes extended lamellipodia-like structures (supplementary material Fig. S8). Therefore, Cdk5 suppression did not completely inhibit actin dynamics in the locomoting neurons. We next tried to inhibit actin dynamics by treatment with cytochalasin D or latrunculin B, both of which inhibit actin polymerization. However, because both drugs induced disruption of cortical slice tissues (supplementary material Movies 5 and 6), we could not analyze the effect of actin dynamics on dilation formation.

Cdk5 is known to phosphorylate many substrate molecules (Kawauchi, 2014). One of them, Doublecortin (Dcx), is involved in microtubule polymerization, membrane trafficking and actin organization (Francis et al., 1999; Gleeson et al., 1999; Friocourt et al., 2005; Tsukada et al., 2005; Moores et al., 2006; Yap et al., 2012; Fu et al., 2013). Another, p27^kip1 (Cdkn1b – Mouse Genome Informatics) controls microtubule and actin cytoskeletal organization (Kawauchi et al., 2006; Godin et al., 2012). As previously reported (Tanaka et al., 2004b; Kawauchi et al., 2006), Cdk5 knockdown decreased the phosphorylation of Dcx and p27^kip1 at Ser297 and Ser10, respectively, and total protein levels of p27^kip1 (supplementary material Fig. S2D,E).

Because Cdk5-mediated phosphorylation is known to negatively regulate Dcx, we overexpressed Dcx. Overexpression of wt-Dcx decreased the ratio of cells with dilation and with elongated nucleus in the locomoting neurons in cortical slices and fixed cortical sections (supplementary material Fig. S9). Importantly, knockdown of Dcx partially rescued the defects in dilation formation and nuclear elongation in the Cdk5-knockdown locomoting neurons (supplementary material Fig. S9).

Knockdown of Dcx alone also disturbed dilation formation and nuclear elongation (Fig. 8), suggesting that proper regulation of Dcx activity is important for the locomotion mode of migration. These defects were rescued by the co-expression of wt-Dcx (Fig. 8), and therefore not due to off-targeting effects. In our experimental conditions, the majority of Dcx-knockdown neurons entered the cortical plate, but their migration speeds were decreased (supplementary material Fig. S10), which is consistent with previous observations in Dcx-knockout mice (Pramparo et al., 2010).

Knockdown of p27^kip1 suppressed dilation formation and nuclear elongation in the locomoting neurons in cortical slices and fixed cortical sections; these defects were rescued by the co-expression of human wt-p27^kip1, suggesting that p27^kip1 also plays some roles in the locomotion mode of neuronal migration (Fig. 8). However, overexpression of a Ser10-phospho-mimic mutant of p27^kip1 (p27-S10D) did not rescue the Cdk5-knockdown phenotypes, probably because Cdk5 regulates many downstream events (supplementary material Figs S9 and S11).

Neuronal migration is a fundamental cellular event towards constructing a functional brain, and defects in neuronal migration result in various neurological disorders, including lissencephaly (Gleeson and Walsh, 2000; Kawauchi and Hoshino, 2008). In the developing cerebral cortex, the majority of migrating neurons exhibit the locomoting morphology. They extend a leading process and form a cytoplasmic dilation at the proximal region of the leading process. The cytoplasmic dilation is a migrating neuron-specific structure and thought to be essential for the forward movement (Bellion et al., 2005; Schaar and McConnell, 2005). Therefore, it determines the character of the unique migration pattern of neurons. In this study, we show that Cdk5, a molecule predominantly activated in neurons, is important for the neuron-specific migrating features: the formation of the cytoplasmic dilation and nuclear elongation.

Cdk5 is known to control dynein-dependent motor activity mainly through the regulation of the Ndel1-Lis1 complex (Ayala et al., 2007). However, knockdown of dynein heavy chain or Lis1 does not affect the cytoplasmic dilation formation, although inhibition of either protein perturbs nuclear forward movement (Tsai et al., 2007), suggesting that another cellular event is involved in the Cdk5-mediated dilation formation. Although Dcx and Lis1, causative gene products of lissencephaly, share some common functions in the dynein-mediated regulation of centrosome positioning (Tanaka et al., 2004a), Dcx plays other roles in microtubule polymerization and endocytic trafficking (Francis et al., 1999; Gleeson et al., 1999; Friocourt et al., 2005; Moores et al., 2006; Yap et al., 2012), which are shown here to regulate dilation formation. These dynein-independent functions of Dcx may have important roles in dilation formation. It is consistent with a previous report that deficiency of Diap1 and Diap3 (previously mDia1 and mDia3) disturbs the centrosomal positioning, but not swelling/dilation formation in cortical interneurons (Shinohara et al., 2012), suggesting that the regulation of dynein activity and its mediated centrosomal positioning may be distinct from dilation formation.

In addition to Dcx, p27^kip1 has been shown to regulate dilation formation. Interestingly, suppression of p27^kip1 in cortical interneurons promotes swelling/dilation formation (Godin et al., 2012), suggesting that regulation of swelling/dilation formation in future excitatory and inhibitory neurons require different mechanisms. The distance between the swelling/dilation and nucleus of interneurons is longer than that of excitatory neurons (Bellion et al., 2005). Interestingly, knockdown of p27^kip1 slightly increased the ‘interneuron-type’ of swelling/dilation (data not shown). Conversely, the deficiency of Diap1 and Diap3, which are involved in the migration of cortical interneurons but not of excitatory projection neurons, results in less separation of the cell soma from the swelling/dilation (Shinohara et al., 2012). Thus, p27^kip1 and Diap proteins might regulate the neuronal subtype-specific migration pattern.

A recent report indicates that Rac1 and POSH, a Rac1 interacting protein, are required for dilation formation (Yang et al., 2012). The relationship between Cdk5 and Rac1 activity has been reported (Nikolic et al., 1998; Xin et al., 2004), and Posh contains several proline-directed Ser/Thr residues, which are potential phosphorylation targets of Cdk5. Therefore, Cdk5 might control the formation of a cytoplasmic dilation in cooperation with Rac1 and POSH.

Nuclear elongation during the locomotion mode of neuronal migration is a process by which the nucleus enters the dilation, implicating nuclear elongation as an important step for the forward movement (Bellion et al., 2005; Schaar and McConnell, 2005). Interestingly, unlike cytoplasmic dilation formation, Jnk and Cdk5 are required for nuclear elongation and forward movement. The decrease of migration speed in the Jnk1-knockdown neurons is consistent with that observed in mice with knockouts of Jnk upstream kinases (Mkk4, Mkk7, Dlk or Mekk4) or electroporated with dominant-negative Jnk1 (Kawauchi et al., 2003; Sarkisian et al., 2006; Wang et al., 2007; Yamasaki et al., 2011). Both Jnk and Cdk5 are known to control microtubule stability and organization (Kawauchi et al., 2003, 2005; Gdalyahu et al., 2004), implying that Jnk and Cdk5 may cooperatively control nuclear elongation via microtubule regulation. Interestingly, nuclear Jnk has a role in neuronal migration (Westerlund et al., 2011), implying that Jnk might directly regulate nuclear morphological changes.

The relationship between Cdk5 and endocytic pathways is still controversial. Endocytic pathways consist of various trafficking pathways that originate from endocytosis, such as endocytosis, recycling and lysosomal degradation pathways (Stenmark, 2009). At the presynapse, Cdk5 negatively regulates endocytosis via phosphorylation of dynamin 1 and amphiphysin 1 (Floyd et al., 2001; Tomizawa et al., 2003), although inhibition of Cdk5 is reported to disturb the phosphorylation/dephosphorylation cycle of dynamin 1 and amphiphysin 1, resulting in a defect in endocytosis at the presynapse (Tan et al., 2003; Evans and Cousin, 2007). By contrast, Cdk5 promotes the endocytosis of NMDA receptors at postsynaptic sites (Zhang et al., 2008). Cdk5 also regulates the N-cadherin-mediated cell adhesion (Kwon et al., 2000) and a Rab11-dependent recycling pathway (Takano et al., 2012), both of which are important for neuronal migration (Kawauchi et al., 2010). Therefore, whether the total output of Cdk5 function promotes or suppresses endocytic pathways may be dependent on cell type or on local environment. Our data show that Cdk5 inhibition disturbs the trafficking of transferrin at Rab5-positive early endosomes, suggesting that Cdk5 coordinates the early steps of endocytic trafficking in embryonic cortical neurons. Alternatively, Cdk5 may indirectly control endocytic pathways via the regulation of microtubule organization.

We have previously reported that Rab5-dependent endocytosis is required for several steps of neuronal migration, including the radial fiber-dependent locomotion mode (Kawauchi et al., 2010; Shikanai et al., 2011). However, in a series of these experiments, Rab5-shRNAs were expressed prior to the transition to the locomotion mode. In this study, our ex vivo chemical inhibitor experiments clearly show that blocking endocytosis disturbs the locomotion mode of neuronal migration. Together with our previous results (Kawauchi et al., 2010) and a report from Susan McConnell's group (Shieh et al., 2011), the present study provides a model in which, under the control of Cdk5, N-cadherin is internalized at the cytoplasmic dilation in a clathrin- and Rab5-dependent manner and recycled back to the plasma membrane, a cycle that is essential for the radial fiber-dependent locomotion mode of neuronal migration.

Plasmids were prepared using the EndoFree plasmid purification kit (Qiagen). EGFP-wt-Rab5 (human) (Rosenfeld et al., 2001), mouse wt-Dcx without 3′UTR, EYFP-Golgi (Clontech) and DsRed-monomer-Golgi (Clontech) cDNAs were inserted into pCAG=MCS2 (Kawauchi et al., 2005) to generate CAG=EGFP-wt-Rab5, CAG=wt-Dcx, CAG=EYFP-Golgi and CAG=DsRed-monomer-Golgi. CAG=EGFP, CAG=nuclear localizing signal (NLS)-fused DsRed, CAG=EGFP-actin, CAG=wt-Cdk5 (human), CAG=wt-Jnk1 (human), CAG=wt-Rab5 (human), CAG=p27-S10D (human), CAG=wt-p27 (human), Cdk5-sh250 (shCK5B), Rab5-sh108 and p27-shRNA (sh27A) have been described previously (Kawauchi et al., 2003,, 2006,, 2010; Nishimura et al., 2010). To construct shRNA-expressing vectors, oligonucleotides targeting the Jnk1-coding sequence (Jnk1-sh788: 5′-GGCCTAAATACGCTGGATA-3′) or a control scrambled sequence (sh-scr2: 5′-CGTCGGATAATACTAATCT-3′), and their complementary sequences were inserted into the mU6pro vector (Yu et al., 2002). All contain a nine-base hairpin loop sequence (5′-TTCAAGAGA-3′). These sequences were designed based on information from shRNA sequence analyses (B-Bridge International). Dcx-shRNA targeting the Dcx 3′UTR sequence (5′-GCTCAAGTGACCAACAAGGCTAT-3′) was constructed as described previously (Ramos et al., 2006).

Primary antibodies used in this study were anti-Jnk (Cell Signaling, 9252, 1/1000), anti-Cdk5 (Santa Cruz, sc-6247, 1/100), anti-β-tubulin (Sigma, T5201, 1/5000 for immunoblot, 1/400 for immunocytochemistry), anti-S732-phosphorylated FAK (Sigma, F7803, 1/500), anti-FAK (Santa Cruz, sc-932, 1/500), anti-p35 (Santa Cruz, sc-820, 1/500), anti-S10-phosphorylated p27 (Zymed, 34-6300, 1/100), anti-p27 (BD Biosciences, 610241, 1/1000), anti-S297-phosphorylated Dcx (Cell Signaling, 4605, 1/100), anti-Dcx (Abcam, ab18723, 1/500), anti-EGFP (Molecular Probes, A6455, 1/1500), anti-Glu-tubulin (Chemicon, AB3201, 1/100), anti-p115 (BD Biosciences, 612260, 1/100) and anti-GM130 (Cell Signaling, 2296, 1/100). Roscovitine, nocodazole, cytochalasin D and latrunculin B were purchased from Sigma, bisindolylmaleimide I (BIM) and rottlerin from Calbiochem, dynasore from Santa Cruz, SP600125 from BIOMOL, 4′,6-diamidino-2-phenylindole dihydrochloride solution (DAPI) from Wako and Alexa594-conjugated transferrin from Molecular Probes.

Pregnant ICR mice were obtained from SLC Japan. Animals were handled in accordance with guidelines established by Keio University, Aichi Human Service Center and NCNP. In utero electroporation was performed on E14 embryos, as described previously (Kawauchi et al., 2003). Cortical slice culture was performed as described previously (Nishimura et al., 2010). Briefly, the electroporated brains at E16 were cut into 300 µm coronal slices with a microtome (Leica), and cortical slices were cultured under confocal laser scanning time-lapse microscopy, FV1000 (Olympus) or TCL-SP2 (Leica). The following inhibitors were added to culture media; control DMSO (0.1%), roscovitine (100 µM), rottlerin (5 µM), SP600125 (50 µM), dynasore (40 µM), nocodazole (100 nM), cytochalasin D (10 µM) and latrunculin B (10 µM).

Primary culture of embryonic cerebral cortices, transfection and immunoblot analysis were performed as described previously (Kawauchi et al., 2003,, 2006,, 2010). For immunohistochemistry, frozen cortical sections were treated with HistoVT-One (Nacalai) for 20 min at 70°C, and other steps were performed as described previously (Kawauchi et al., 2010).

E15 cerebral cortices were dissociated and cultured for 2 days. Primary cultured neurons were incubated in Opti-MEM media (Invitrogen) for 30 min at 37°C and treated with 20 µg/ml Alexa594-conjugated transferrin in Opti-MEM media. After incubation for 5-15 min on ice, neurons were incubated for 10 or 30 min at 37°C.

For immunocytochemistry, cortical neurons were fixed with 4% PFA in PBS for 20 min, permeabilized with 10% goat serum in PBS (GS-PBS) containing 0.15% Triton X-100 for 5 min and blocked with GS-PBS for 30 min at room temperature. Subsequently, the cultured neurons were incubated with diluted primary antibodies in GS-PBT (GS-PBS containing 0.1% Tween20) at 4°C overnight. After three washes in PBS, neurons were treated with Alexa488- or Alexa555-conjugated secondary antibodies (Molecular Probes) diluted in PBS for 60 min at room temperature, followed by three washes in PBS. Fluorescent images were obtained by TCL-SL or TCL-SP5 laser scanning confocal microscopy (Leica).

Migration speed was analyzed by DippMotion 2D software (Ditect). Dilation was defined as a swollen leading process (more than one-third of the nuclear diameter) at the proximal region (5 µm above the pial edge of the nucleus) with a narrowing between the swollen region and nucleus. Dilation area was defined as the maximum area of leading processes in 10 µm distances from the cell soma during 10 h time-lapse observations. Nuclear sphericity was analyzed using Imaris software (Bitplane). Sphericity was defined by Wadell in 1935 (Wadell, 1935) as a measure criterion of ‘how spherical an object is’ and calculated as the ratio of the surface area of a sphere with the same volume as the given particle to the surface area of the particle.

Data are presented as mean±s.e.m. Statistical significance was calculated using two-tailed Student's t-test or Welch's t-test. P<0.05 was considered statistically significant.

We thank Drs Brian J. Knoll (University of Houston), Jun-ichi Miyazaki (Osaka University) and David L. Turner (University of Michigan) for providing plasmids. The authors also thank Drs Ruth T. Yu (Salk Institute), Michisuke Yuzaki (Keio University) and Satoshi Miyashita (NCNP) for proofreading of the manuscript, providing the experimental environments and supporting time-lapse experiments, respectively.