Nesprins and opposing microtubule motors generate a point force that drives directional nuclear motion in migrating neurons

During mammalian brain development, newly born neurons migrate from their birthplace in the germinal layers to their final destination in the emerging cortices and nuclei, where they are eventually integrated into functional neural circuits. Migration of postmitotic neurons is an essential step in brain development, and its failure is associated with various brain disorders (Manzini and Walsh, 2011; Cooper, 2013; Moon and Wynshaw-Boris, 2013). Migrating neurons typically exhibit a bipolar shape with a long, thick leading process extending from the cell soma toward its destination, followed by a thin trailing process on the opposite site of the soma. The nucleus in the cell soma is translocated into the preceding leading process through the crowded neural tissue by the force generated by actin, microtubules and related motor proteins (Vallee et al., 2009; Trivedi and Solecki, 2011). The actomyosin contractile force plays a major role in driving nuclear translocation, exerting either a pulling force in the leading process or a pushing force in the cell rear (Bellion et al., 2005; Schaar and McConnell, 2005; Solecki et al., 2009; He et al., 2010; Martini and Valdeolmillos, 2010; Shinohara et al., 2012). Perinuclear microtubules emanating from the leading process and their associated cytoplasmic dynein complex are also indispensable for directional motion of the nucleus (Shu et al., 2004; Tanaka et al., 2004; Tsai et al., 2007; Umeshima et al., 2007). These cytoskeletons and associated motors are anchored to the protein complexes embedded in the nuclear membrane, including the linker of nucleoskeleton and cytoskeleton (LINC) complex and the nucleopore complex, by which they transduce the force to the nucleus (Crisp et al., 2006; Zhang et al., 2009; Hu et al., 2013). However, the precise interplay of actomyosin and microtubules during nuclear translocation as well as the site and extent of the force generated by these cytoskeletons remain to be elucidated.

Here, we observed the dynamic motion of the nucleus in migrating cerebellar granule cells (CGCs) at a high spatiotemporal resolution using spinning-disk confocal microscopy, in order to enhance our understanding of the mechanism and dynamics of nuclear migration in neurons. We found that the nucleus in migrating neurons displays a highly dynamic motion, including fast rotation in the soma along the direction of its translocation. We demonstrate that the nuclear rotation and some mode of nuclear translocation are driven by the motor activity of cytoplasmic dynein complex and kinesin-1, which is transmitted to the nucleus via the LINC complex. By contrast, actomyosin appeared not to be involved in nuclear rotation, suggesting that actin and microtubules act somewhat independently and control different modes of nuclear dynamics during neuronal migration.

Tangential migration of the CGCs can be reconstituted in a reaggregate culture (Umeshima et al., 2007). At 1-2 days in vitro (DIV), each CGC extended a long leading process and migrated out of the cell aggregate at an average somal speed of 12.2±1.5 µm/h. In differential interference contrast time-lapse imaging, we observed that the nucleus sometimes rotated during migration, as determined by the relative position of the nucleoli (Fig. S1A). The nucleoli appeared to rotate more than 30° in 1 min. To perform more detailed image analysis, we labeled heterochromatin as a position marker in the nucleus by transfecting heterochromatin protein 1 beta (HP1β)-mCherry (HP1β is also known as Cbx1) together with a GFP cell-volume marker (Dialynas et al., 2006). Consistent with previous reports, heterochromatin formed focal aggregates in the nucleoplasm and nuclear periphery in the CGCs. Time-lapse imaging with spinning-disk confocal microscopy at 15 s intervals revealed that the nucleus is highly dynamic during neuronal migration, exhibiting frequent deformation from a round to an irregular oval shape (Movie 1, left). In addition, the heterochromatin spots exhibited intermittent rotation in the cell soma of highly motile neurons at 1-2 DIV (Fig. 1A,B). The nucleus spun clockwise or counterclockwise, and it occasionally switched rotation direction. By contrast, the nucleus in mature postmigratory neurons at 6 DIV was rather static and did not show notable rotation (Fig. 1B, Fig. S1B,C; Movie 1, right).

To determine whether only the nucleoplasmic compartments including heterochromatin or the entire nucleus including the nuclear envelope rotates, we visualized the nuclear envelope by transfecting mCherry-lamin B1. mCherry-lamin B1 was unevenly distributed on the nuclear membrane and allowed us to track rotation at a comparable speed to heterochromatin, suggesting that the nuclear interior and nuclear envelope rotated together as a unit (Fig. S1D, Movie 2).

We next examined whether other organelles associated with the nucleus rotate simultaneously. Cultured CGCs were co-transfected with either AcGFP-Golgi or centrin 2 (Cetn2)-GFP together with HP1β-mCherry to visualize the dynamics of the Golgi apparatus and the centrosome, respectively (Wu et al., 2015). The Golgi stacks juxtaposed to the nucleus in the cell soma (Fig. 1C). Unexpectedly, the Golgi was rather static during migration and stayed in the same position even when the nucleus underwent active rotation (Movie 3, left). Likewise, the centrosome did not follow nuclear spinning even though it moved dynamically in the cell soma during migration as previously reported (Umeshima et al., 2007) (Fig. 1C; Movie 3, right). These data suggest that the nuclear rotation is not driven by global cytoplasmic streaming but by a force that is applied specifically to the nucleus.

To verify whether the nuclear rotation is an artifact of isolated, free-moving neurons on a glass surface, we observed CGCs migrating in neural tissue in an organotypic slice of the cerebellum (Umeshima et al., 2007). HP1β-mCherry and GFP were electroporated in the primary fissures of postnatal day (P) 8 mice cerebella, and cerebellar slices were prepared at P10 when CGCs undergo tangential and radial migration. Time-lapse imaging using confocal microscopy revealed that the nucleus was drastically more deformed and underwent frequent rotation during both tangential and radial migration in brain tissue (Fig. 1D, Movie 4). This indicates that the nuclear rotation is a natural event in migrating neurons in the developing cerebellum.

As nuclear rotation was observed only in migratory neurons, we hypothesized that the rotation was driven by the same force as that propelling nuclear translocation. Consistent with this hypothesis, nuclear rotation was frequently observed immediately before or after saltatory movement of the nucleus (Fig. 2A,B). Cytoplasmic dilations were observed at the proximal leading process, as previously reported (Movie 1) (Tsai et al., 2007; Trivedi et al., 2017), while nuclear rotation was also seen in those lacking a typical dilation.

The driving force for nuclear movement must be applied to multiple points on the round surface of the nucleus, and displacement must be driven when the net force acts on the center of mass. On the other hand, biased forces must exert a torque and drive the nuclear rotation around an axis orthogonal to the direction of nuclear translocation (Fig. 2C). To prove our hypothesis, we determined the rotation axis and its angle to the direction of migration. The motion of the HP1β-mCherry spots was traced during rotation by three-dimensional particle tracking. The positions of the heterochromatin spots in each nucleus remained fixed relative to one another so that we were able to track their trajectories in real (xyz) space (Fig. S1C). In all cases examined, the rotation axis was nearly orthogonal to the leading process, which directs migration (mean angle to the leading process, 75.1°; Fig. 2D,E). On the other hand, unlike the flat nucleus in fibroblasts, which primarily rotates in two dimensions parallel to the substrate plane (Levy and Holzbaur, 2008), the oval nucleus in CGCs rotated in three dimensions with some bias toward the substrate plane (mean angle to the substrate plane, 57.5°; Fig. 2E, Fig. S1E). There was no bias, but frequent switches, between clockwise or counterclockwise rotation (Fig. 2E, Movie 5). The mean velocity of nuclear rotation was 47.1±3.4°/min, which is one order faster than those observed in other cell types, such as migrating fibroblasts (<10°/min) and developing muscle cells (<6°/min) (Capers, 1960; Paddock and Albrecht-Buehler, 1986; Ji et al., 2007; Levy and Holzbaur, 2008; Brosig et al., 2010; Wilson and Holzbaur, 2012). These results are consistent with our hypothesis that nuclear rotation is not a random event but is driven by a strong directional force towards the orientation of nuclear translocation.

We found that the nuclei actively rotated when they were spherical, whereas the non-rotating nuclei were more elongated in the direction of migration (Fig. 2F). Notably, moving nuclei were dynamically sharpened at their front, implying a point force applied to the anterior of the nuclear envelope (Fig. 2G, Movies 5 and 6). Thus, the tensile force of migration appears to cause nuclear strain energy that is converted to either anterior deformation, displacement, or rotation, depending on the nucleus shape and stress points. We thus utilized nuclear rotation as a new readout of the cytoskeletal forces driving nuclear migration.

It has been shown that the LINC complex proteins nesprin-1 and -2 (Syne1 and Syne2) are directly involved in nuclear translocation by connecting the cytoskeleton and related motor proteins to the nuclear envelope (Zhang et al., 2009). Nesprins interact via the KASH domain with SUN proteins, which span the inner nuclear envelope and anchor to the nuclear lamina (Starr and Fischer, 2005). To examine whether nuclear rotation is driven by a cytoskeletal force mediated by the LINC complex, we transfected CGCs with a deletion mutant of nesprin-1 lacking the N-terminal cytoplasmic region (mCherry-KASH1), which is reported to inhibit the activity of endogenous nesprin-1/2 (Stewart-Hutchinson et al., 2008). Time-lapse imaging showed that overexpression of mCherry-KASH1 strongly inhibited nuclear rotation and, to a lesser extent, nuclear translocation (Fig. 3A-C, Movie 7). As a negative control, we also expressed mCherry-KASH1ext, which has additional C-terminal amino acids that do not bind to the cytoskeleton or interfere with the endogenous LINC complex (Stewart-Hutchinson et al., 2008). Overexpression of mCherry-KASH1ext did not affect nuclear rotation or translocation, indicating that inhibition of the LINC complex specifically interferes with rotation and translocation (Fig. 3B,C, Movie 7).

Nesprin-1/2 have been shown to interact with actin and the microtubule motors dynein and kinesin (Gerlitz and Bustin, 2011; Rajgor and Shanahan, 2013; Chang et al., 2015). To further identify the cytoskeletal systems involved in nuclear rotation, we examined the effect of cytoskeleton-disrupting agents. Treatment with the actin depolymerizer cytochalasin B (20 μM) immediately disrupted the actin cytoskeleton, as revealed by LifeAct-2xGFP transfected into migrating CGCs (Movie 8, left). Surprisingly, the nucleus actively rotated even after complete disruption of actin filaments, whereas nuclear translocation was strongly inhibited (Fig. 4A,B). We also confirmed that the nucleus stopped forward movement but continued spinning intermittently after treatment with blebbistatin (50 μM), a potent inhibitor of non-muscle myosin II (Fig. 4A,B). These results indicate that a contractile force of actomyosin is required for nuclear translocation, but not for rotation.

We next tested the effect of the microtubule depolymerizer nocodazole. Treatment with nocodazole (1 μM) disrupted microtubule filaments, as visualized by DCX-GFP expression (Fig. S2; Movie 8, right). Many of the nocodazole-treated cells sprouted exuberant protrusions and strongly agitated the cell soma, consistent with previous studies showing that nocodazole increases actin-based cell contractility via activation of the RhoA-ROCK-MLC pathway (Chang et al., 2008). Despite the increased cell motility, nuclear spinning was significantly attenuated after disrupting microtubule filaments with nocodazole (Fig. 4A). The directional translocation of the nucleus, in the forward direction, was also perturbed (Fig. 4B). We also tested the effect of the microtubule stabilizer taxol (5 μM) and found similar inhibition of nuclear translocation and rotation (Fig. 4A,B). These results suggest that nuclear rotation is associated with nuclear translocation driven by the microtubule-dependent force.

We further visualized microtubule dynamics during nuclear spinning by transfecting CGCs with DCX-GFP and HP1β-mCherry. We observed thick bundles of microtubules extending from the leading process toward the anterior surface of the nucleus, as reported previously (Umeshima et al., 2007). These microtubules were highly dynamic during neuronal migration, and some filaments appeared to closely associate with the nucleus and follow its rotation, supporting the idea that microtubules are directly associated with the rotating nucleus (Fig. 4C, Movie 9).

It is widely accepted that cytoplasmic dynein is the predominant motor mediating the microtubule-dependent force during nucleokinesis in neuronal migration (Cooper, 2013). We investigated the involvement of cytoplasmic dynein in nuclear rotation by molecular perturbation of the two dynein adaptors Lis1 (Pafah1b1) and dynactin, which were previously shown to function in parallel to recruit dynein to the nucleus (Vallee et al., 2012). It has previously been shown that an N-terminal fragment of Lis1 (Lis1N), which has a dominant inhibitory effect on Lis1-dynein function, inhibits nuclear migration in neurons (Tsai et al., 2005; Umeshima et al., 2007). Overexpression of Lis1N significantly inhibited nuclear rotation in CGCs (Fig. 5A; Movie 10, left). Dynactin has been implicated in most dynein-mediated functions (Vallee et al., 2012). Dynactin function can be inhibited by overexpressing the first coiled-coil domain of the dynactin subunit p150^glued (p150-CC1; also known as Dctn1) (Quintyne et al., 1999). CGCs transfected with p150-CC1 were altered in shape, with aberrant filopodia. Nuclear spinning was strongly inhibited in these cells (Fig. 5A; Movie 10, right). These results suggest that dynein is involved in the force that generates nuclear rotation.

Nesprin-2 has been shown to interact with kinesin light chain (KLC), which forms the kinesin-1 complex with the kinesin heavy chain Kif5 (Schneider et al., 2011; Wilson and Holzbaur, 2015). It is known that CGCs express three isoforms of Kif5 (Kif5A, B and C) and at least three isoforms of KLC (KLC1, 2 and 4), which share high similarity and are likely to complement each other's function (Brain Transcriptome Database, http://www.cdtdb.neuroinf.jp). Double immunofluorescence revealed that Kif5B is highly expressed in the leading process and cell body of migrating CGCs in a partially overlapping, but non-identical, pattern with dynein (Fig. S3A). Strong signals were observed on the nuclear surface in cells permeabilized in a stringent buffer, suggesting that kinesin-1, as well as dynein, bound to the nuclear envelope in CGCs (Fig. S3B,C). The association of Kif5B with the nuclear envelope was further confirmed in a cell fractionation assay (Fig. S3D). Overexpression of mCherry-KASH1 abolished the Kif5B signals on the nuclear surface, suggesting that kinesin-1 binds to the nucleus via nesprin-1/2 (Fig. S3E,F). We therefore used a truncated mutant of Kif5B lacking the N-terminal motor domain (Kif5B-tail) as a dominant-negative form to inhibit all three types of Kif5 (Coy et al., 1999; Konishi and Setou, 2009). Moderate GFP-Kif5B-tail overexpression indeed decreased the frequency of nuclear rotation in CGCs with little obvious change in cell polarity and shape (Fig. 5A, Fig. S4; Movie 11, left). ShRNA knockdown of Kif5B, the most abundant Kif5 subtype in the brain (Kanai et al., 2000), also suppressed nuclear rotation (Fig. 5B). To further verify the function of kinesin-1 in nuclear dynamics, we examined the effect of a dominant-negative mutant of KLC1 deleted of its heptad repeats that encompass the Kif5-binding domain (GFP-KLC1-TPR). KLC1-TPR is supposed to mask nesprins and to inhibit the binding of the kinesin-1 motor complex to the nucleus. As the immunofluorescence signal of KLC was not clear in CGCs, we used C2C12 cells and confirmed that GFP-KLC1-TPR accumulated on the nuclear surface and delocalized endogenous KLC1 to the cytoplasm (Fig. S5). Overexpression of GFP-KLC1-TPR reduced the rotation frequency of the nucleus, supporting the proposal that kinesin-1 drives nuclear motion by binding to nesprins (Fig. 5A; Movie 11, right). Molecular perturbation of dynein or kinesin-1 also suppressed nuclear translocation, consistent with the hypothesis that nuclear rotation is driven by the cytoskeletal force steering nuclear translocation (Fig. 5C,D). Together, these data strongly suggest that dynein and kinesin-1 mediate the microtubule-dependent forces that operate during nuclear rotation and translocation in CGCs.

The kinesin-binding domain of the cytoplasmic stretch of nesprin has been identified (Schneider et al., 2011). To investigate whether kinesin is sufficient for driving nuclear rotation, we overexpressed a nesprin-2 mutant deleted of its cytoplasmic region except for the four C-terminal spectrin repeats, which includes the kinesin-binding motif (nesp2-SR). We confirmed that GFP-nesp2-SR expressed in HEK293T cells bound to kinesin-1 but not dynein (Fig. S6). Overexpression of GFP-nesp2-SR induced a striking change in shape of the nucleus, with a constant sharp peak in the forward direction, which is seen only temporarily in normal cells (Fig. 6A, compare with Fig. 2G). However, rotation and translocation were downregulated in nesp2-SR-overexpressing cells, suggesting that binding of kinesin alone is insufficient to induce nuclear motion (Fig. 6B,C, Movie 12).

It is thought that most microtubules in migrating neurons originally emanate from the centrosome in front of the nucleus and align with their plus-end to the nucleus (Tanaka et al., 2004; Tsai et al., 2007). If the microtubules interact with the nucleus exclusively at their plus-ends, kinesins would be incapable of generating the plus-end motor activity. To explore the microtubule arrangement in migrating CGCs, we monitored the plus-ends of growing microtubules by transfecting GFP-tagged EB3 (Mapre3). We observed comet-like streaks of GFP-EB3 in the cell soma moving across the nucleus. In addition to the streaks, which emanate from the cell front toward the trailing process as previously reported (Tsai et al., 2007), a significant number of comets move toward the leading process (78.6% microtubules moving backward versus 21.4% microtubules moving forward, n=437 streaks from 17 cells; Fig. 6D). This is consistent with recent studies that identified a considerable fraction of non-centrosomal microtubules in migrating CGCs (Rao et al., 2016). These parallel and antiparallel microtubules were not segregated in distinct areas but were mixed in the perinuclear region (Fig. 6D, Movie 13).

Finally, we investigated whether kinesin-1 is required for the migration of CGCs in vivo, as kinesin motor activity has never been implicated in nucleokinesis in postmitotic neurons. GFP-Kif5B-tail or control GFP was electroporated into immature granule neurons in the external granular layer (EGL) of the P8 mouse cerebellum. Injected brains were fixed and sectioned at P11 to assess neuronal migration from the EGL to the internal granular layer (IGL). The large majority of GFP-transfected CGCs finished migration in the IGL (73.3%), except for a small number of cells that remained in the EGL (5.1%) or molecular layer (ML) (10.0%) (Fig. 7A,B). By contrast, transfection with GFP-Kif5B-tail significantly increased the number of CGCs that remained in the EGL (16.8%) and ML (28.0%), at the expense of those reaching the IGL (37.6%). Inhibition of Kif5B by shRNA knockdown also significantly retarded CGC migration in vivo (Fig. 7C,D). These results indicate that kinesin-1 is involved in the generation of cytoskeletal force driving nuclear rotation and translocation in migrating CGCs during cerebellum development.

In this study, a fine spatiotemporal analysis of nuclear behavior reveals highly dynamic movements of the nucleus, including intermittent fast rotation, in newly born CGCs. In contrast to fibroblast nuclear migration driven by a direct link between nesprins and actin (Luxton et al., 2010; Bergert et al., 2015), our results implicate the interaction of nesprins and microtubules in the nuclear rotation of migrating CGCs. While microtubules in migrating neurons are highly dynamic, they essentially align parallel to the force vector of rotation along the direction of the leading process (Fig. 4C, Fig. 6D, Movie 13) (Tsai et al., 2007; Umeshima et al., 2007). We observed dynamic peaking of the front of the nucleus during migration, which is enhanced by the excess interaction of nesprin and kinesin-1 (Movies 5 and 12). These observations support the idea that microtubule motors are directly anchored to the LINC complex, applying a point force to the nuclear envelope and triggering nuclear deformation and rotation. Whereas nuclear rotation is solely dependent on microtubule motors, nuclear translocation requires both actomyosin and microtubule motors. Previous studies have demonstrated that the actomyosin in the leading process generates the traction force pulling the nucleus in CGCs (He et al., 2010; Jiang et al., 2015), and this actomyosin tethers perinuclear microtubules and transmits the contractile force to the microtubules (Trivedi et al., 2017). We thus surmise that the point force of microtubules and associated motors are responsible for steering the nucleus during rotation and deformation, while the strong contractile force of actomyosin synergistically drives nuclear translocation by acting on a large area of the nucleus and/or regulating microtubule positioning in CGCs.

It is widely accepted that minus-end motor activity of dynein plays a major role in nuclear migration in postmitotic neurons (Cooper, 2013). Perinuclear microtubules are thought to orient uniformly with their minus-ends toward the cell front, excluding a major role of the plus-end-directed kinesin in forward nuclear motion. One exception is bi-directional interkinetic nuclear migration in cortical neural stem cells, which is driven apically by dynein and basally by kinesin-3 (Tsai et al., 2010; Carabalona et al., 2016). Here we demonstrate that perinuclear microtubules are of mixed polarity and both dynein and kinesin-1 are involved in directional nuclear motion in the postmitotic neuron. It has been shown that kinesin indirectly regulates nuclear migration by recruiting dynein to the nuclear surface (Yamada et al., 2010), by regulating neuronal polarity (Falnikar et al., 2013), or by promoting transport and secretion of neurotrophins (Carabalona et al., 2016). By contrast, our data suggest that kinesin motor activity drives nuclear motion directly through interaction with the LINC complex. How do the opposing motors cooperatively drive unidirectional nuclear motion? It is unlikely that dynein and kinesin are segregated by microtubules of opposite polarity and always generate the same vector force. We observed frequent switches of rotation direction, suggesting a dynamic interaction of the opposing motors with nesprins driving both forward and backward movement of the nucleus. Consistently, overexpression of nesprin with only the kinesin-binding domain induces persistent nuclear deformation while inhibiting rotation and translocation. Given that microtubule polarity is strongly biased toward those with the minus-end to the cell front, dynein on these dominant microtubules is considered to be the major player in nuclear translocation, while kinesin may enhance nuclear rotation and smooth translocation (Fig. S7).

We observed that the nucleus shows remarkable deformation during migration in dense neural tissues in organotypic slice cultures. Such nuclear deformation is seen in various cell migration events, such as hematopoiesis, metastasis and inflammation, and is due to local compression of the cell during passage through narrow regions of interstitial tissue or basement membrane (Wolf et al., 2007, 2013; Shin et al., 2013). Recent studies have indicated that kinesin-1 and cytoplasmic dynein are anchored to the LINC complex and synergistically regulate the nuclear migration through narrow constrictions in C. elegans embryos (Fridolfsson and Starr, 2010; Bone et al., 2016). Nuclear rotation is likely to be a consequence of unbalanced point forces exerted by the microtubule motors that pull the nucleus through constrictions. Dynamic changes in the cytoskeletal arrangement caused by the rotation might sometimes optimize the positioning of the nucleus and the cytoskeletons, and thus enable smooth translocation in the crowded neural tissue. Alternatively, rotation may release the excess force and thereby protect the nucleus from damage due to extreme deformation toward the narrow constrictions (Denais et al., 2016; Raab et al., 2016).

Given the apparent diversity of cytoskeletal organization among neuronal species, the nuclear dynamics and regulatory mechanism must be analyzed comparatively in other neurons to obtain a better understanding of the physiology and pathology of nucleokinesis in the brain.

ICR mice were obtained from Japan SLC. All experiments involving mice were approved by the Animal Experimentation Committee of Kyoto University, and were performed in accordance with its guidelines.

pAcGFP1-Golgi was obtained from Clontech. pCAG-EGFP, pCentrin2-EGFP and pCAG-DsRed2-Nuc were constructed as previously described (Umeshima et al., 2007). pCAG-LifeAct-2xGFP was created by insertion of the LifeAct sequence fused with tandemly duplicated EGFP into pCAGGS vector. HP1β-mCherry and HP1β-GFP were created by cloning heterochromatin protein 1 beta from a mouse brain cDNA library fused with mCherry or EGFP at the C-terminus. For mCherry-KASH1, nesprin-1 cDNA (8726-8799 aa) was amplified and fused with mCherry at the N-terminus. mCherry-KASH1ext was created by fusing an additional C-terminal stretch (VDGTAGPGSTGSR peptide derived from pEGFP-C1) into mCherry-KASH1. mCherry-lamin B1, DCX-GFP, and GFP-EB3 were created by insertion of lamin B1, doublecortin and EB3 cDNAs cloned from a mouse brain cDNA library into pCAGGS. EGFP or mCherry sequence was inserted at the N- or C-terminus of respective cDNAs. GFP-Lis1N (1-87 aa of Lis1), GFP-p150-CC1 (214-548 aa of p150^Glued), GFP-Kif5B-tail (808-963 aa of Kif5B), GFP-KLC1-TPR (187-499 aa of KLC1), and GFP-nesp2-SR (6231-6874 aa of nesprin-2) were amplified and fused with EGFP at the N-terminus. For RNAi experiments, target sequences for mouse Kif5B 5′-GCATATTTCTTATTAATGTAA-3′ (shKif5B#1) and 5′-GGATAAAGATATTGCTATTAC-3′ (shKif5B#2) were cloned into the pBAsi-mU6 vector (TAKARA). The promotor and shRNA sequence regions were excised and inserted into pCAG-EGFP.

Reaggregate cultures of CGCs were prepared as previously described with a few modifications (Kawaji et al., 2004). Cerebella from P4-6 ICR mice of either sex were dissociated and incubated in a cocktail containing 4-8 µg plasmid DNA and 6 µl Lipofectamine 2000 (Invitrogen) for 6-8 h to make reaggregates. The resulting reaggregates were plated on glass-based dishes coated with poly-D-lysine and laminin and maintained in Basal Medium Eagle (BME) with 26.4 mM glucose, 25 mM sodium bicarbonate, 1% bovine serum albumin (BSA), 1× N-2 supplement (Thermo Fisher Scientific) and penicillin-streptomycin at 37°C/5% CO₂.

Migrating CGCs in reaggregated cultures were immunostained for kinesin, dynein or tubulin and imaged as described in the supplementary Materials and Methods. Nuclear and cytosolic fractions from dissociated cerebella of P4 ICR mice were prepared for western analysis as described in the supplementary Materials and Methods.

Reaggregate cultures were observed with an incubation imaging system (LCV100, Olympus) through a 20× objective (N.A. 0.7), or with a spinning-disk confocal microscope (CV1000, Yokogawa) through a 20× objective (N.A. 0.75) and a 100× oil-immersion objective (N.A. 1.3) at 37°C with 5% CO₂ flow. For nuclear rotation analysis, time-lapse images were obtained every 15 s for 1 h to trace the trajectories of HP1β spots. For translocation analysis, images were obtained every 5 min for 5 h to measure the cumulative distance of directional travel of the nucleus. For drug treatments, half of the culture medium was replaced with pre-warmed fresh medium containing blebbistatin (50 μM; Sigma), cytochalasin B (20 μM; Sigma), taxol (5 μM; Sigma) or nocodazole (1 μM; Sigma) dissolved in DMSO. For imaging of migration in organotypic slice cultures, cerebella were dissected and embedded in 3% agarose 2 days after electroporation, and sectioned into 300 μm-thick coronal slices with a vibratome. Slices placed on Millicell-CM (Millipore) were mounted in collagen gel and soaked in medium (60% BME, 25% Earle's Balanced Salt, 15% horse serum, 3 mM L-glutamine, 1 mM sodium pyruvate, 31 mM glucose, 21.4 mM sodium bicarbonate, 1× N-2 supplement). The tissue was kept in an incubator chamber attached to an upright microscope stage (BX61WI, Olympus) at 37°C with 85% O₂/5% CO₂ flow. Images were obtained with a laser-scanning confocal microscope (FV1000, Olympus) every 15 or 30 s through a 60× water-immersion objective (N.A. 1.1).

P7 or P8 mice were anesthetized on ice and injected with plasmid DNAs diluted in Tris-EDTA buffer (pH 8.0) with a 33-gauge needle connected to the anode. Tweezer-type electrodes connected to the cathode were placed on the occipital regions. Six electric pulses of 70 mV for 50 ms duration were applied with 150 ms intervals using a CUY21 (Nepagene). The pups were revived at 37°C and returned to the litter. Electroporated mouse brains were isolated at P10 or P11 and fixed with 4% PFA in PBS. The cerebella were then sectioned into 100 μm-thick sagittal slices with a vibratome and stained with DAPI. Images were acquired with the FV1000 through a 20× objective (N.A. 0.75).

C2C12 cells were cultured, induced to myogenic differentiation, immunostained for KLC1 and GFP, and imaged as described in the supplementary Materials and Methods.

NIH3T3 (RCB1862, RIKEN BRC) cells were maintained at 37°C/5% CO₂ in D-MEM with 10% fetal bovine serum and penicillin-streptomycin. shRNA constructs were transfected with Nucleofector II (Lonza) program U-030. After 2.5 days of transfection, cells were harvested in lysis buffer [50 mM Tris-HCl pH 7.4, 75 mM sodium chloride, 2.5 mM magnesium chloride, 1 mM EDTA, 0.1% Triton X-100, 1× protease inhibitor cocktail (Thermo Fisher)]. Lysates were subjected to SDS-PAGE and western blotting with the following antibodies (see also supplementary Materials and Methods): rabbit anti-Kif5B (1:1000; ab167429, Abcam), HRP-conjugated anti-rabbit IgG (1:10,000; 170-6515, Bio-Rad) and HRP-conjugated mouse anti-β-actin (1:10,000; sc-47778, Santa Cruz). For HEK293T cell culture and western blotting, see the supplementary Materials and Methods.

Nuclear rotation was defined as rotation of HP1β spots of more than 15° over a period of at least 1 min. For one rotation event, trajectories of several HP1β-mCherry spots were tracked in three-dimensional time-lapse images using TrackMate (Tinevez et al., 2016). Principal component analysis was performed on the three-dimensional data of each trajectory to determine the best fitting plane. The axis of rotation was calculated by averaging the normal vectors of each fitted plane. The angle between the rotation axis and the direction of migration was calculated in the range of 0 to 90°. The z-component of the direction of migration was set 0.

The duration of nuclear rotation was compared using the Kolmogorov–Smirnov test. In vivo migration was compared using the chi-squared test. P<0.05 was considered statistically significant. Nuclear translocation was compared using a left-tailed t-test with P<0.03 considered significant.

We thank Professor M. Kaneko (Osaka University) and his lab members for invaluable discussions; Mr F. Ishidate for technical assistance; and Professor D. Packwood for critical reading of the manuscript.