Abstract
During neuronal migration, forces generated by cytoplasmic dynein yank on microtubules extending from the centrosome into the leading process and move the nucleus along microtubules that extend behind the centrosome. Scaffolds, such as radial glia, guide neuronal migration outward from the ventricles, but little is known about the internal machinery that ensures that the soma migrates along its proper path rather than moving backward or off the path. Here we report that depletion of KIFC1, a minus-end-directed kinesin called HSET in humans, causes neurons to migrate off their appropriate path, suggesting that this molecular motor is what ensures fidelity of the trajectory of migration. For these studies, we used rat migratory neurons in vitro and developing mouse brain in vivo, together with RNA interference and ectopic expression of mutant forms of KIFC1. We found that crosslinking of microtubules into a nonsliding mode by KIFC1 is necessary for dynein-driven forces to achieve sufficient traction to thrust the soma forward. Asymmetric bouts of microtubule sliding driven by KIFC1 thereby enable the soma to tilt in one direction or another, thus providing midcourse corrections that keep the neuron on its appropriate trajectory. KIFC1-driven sliding of microtubules further assists neurons in remaining on their appropriate path by allowing the nucleus to rotate directionally as it moves, which is consistent with how we found that KIFC1 contributes to interkinetic nuclear migration at an earlier stage of neuronal development.
SIGNIFICANCE STATEMENT Resolving the mechanisms of neuronal migration is medically important because many developmental disorders of the brain involve flaws in neuronal migration and because deployment of newly born neurons may be important in the adult for cognition and memory. Drugs that inhibit KIFC1 are candidates for chemotherapy and therefore should be used with caution if they are allowed to enter the brain.
Introduction
The formation of the mammalian neocortex is a complex process requiring regulated neurogenesis and postmitotic neuronal migration (Northcutt and Kaas, 1995; Solecki et al., 2004; Lodato and Arlotta, 2015; Arai and Taverna, 2017). Neural progenitors divide and give rise to newborn neurons that migrate from the ventricular zone (VZ) of the neocortex to their destinations in the brain (Kriegstein and Alvarez-Buylla, 2009; Fietz and Huttner, 2011; Cooper, 2013). Before neurogenesis, neural progenitors oscillate their nuclei in synchrony with cell cycle progression in a process called interkinetic nuclear migration (INM) (Taverna and Huttner, 2010; J. W. Tsai et al., 2010; Kosodo, 2012; Lee and Norden, 2013). During INM, the soma moves in a defined trajectory, first basally away and then apically toward the VZ, to maximize the number of cell divisions possible within the limited apical surface of the ventricle.
The newborn neuron consists of an elongated leading process, a soma, and a trailing process. The trajectory of migration is stereotyped, with neurons moving radially without veering in directions markedly off the radial path. Once the neuron reaches its destination, migration ceases, the soma becomes stationary, and the neuron extends an axon and dendrites (Schaar and McConnell, 2005; Dantas et al., 2016; Bertipaglia et al., 2018; Goncalves et al., 2019). While some aspects of these various steps are influenced by external cues, such as radial glial tracks that guide neuronal migration (Campbell and Gotz, 2002; Marin et al., 2010), a great deal is dependent on the machinery within the neuron itself.
Microtubules are the fibers within the leading process that yank on the centrosome to pull the soma forward during neuronal migration. They are also the fibers along which the nucleus moves during both INM and neuronal migration. A model has emerged from decades of work in which the most important forces for these movements are generated by cytoplasmic dynein (L. H. Tsai and Gleeson, 2005; Kuijpers and Hoogenraad, 2011; Buchsbaum and Cappello, 2019), with actomyosin providing additional forces pushing the soma as the cell moves along (Kholmanskikh et al., 2003; Solecki et al., 2009; Inoue et al., 2019). While it is generally assumed that all functionally relevant microtubules in migratory neurons are attached to the centrosome (J. W. Tsai et al., 2007), the presence of centrosome-unattached microtubules in migratory neurons and their progenitors has been established (Schaar and McConnell, 2005; Umeshima et al., 2007). We previously showed that sliding of centrosome-unattached microtubules finetunes the trajectory of neuronal migration (Rao et al., 2016). Unresolved, however, was the molecular motor that regulates this sliding.
Various kinesins are important during neuronal development, many of which are repurposed from their roles in mitosis (Sharp et al., 1997; Ferhat et al., 1998; Baas, 1999; Ahmad et al., 2000; Baas and Ahmad, 2001; He et al., 2005; Myers and Baas, 2007; Falnikar et al., 2011, 2013; Lin et al., 2012; del Castillo et al., 2015; McNeely et al., 2017). Of these, KIFC1 (called HSET in humans) is the most highly expressed in neurons (Silverman et al., 2010). KIFC1 is a member of the kinesin-14 subfamily, the only group that moves toward minus ends of microtubules. KIFC1 can drive microtubule sliding but can also crosslink microtubules in a manner that prohibits sliding (Cai et al., 2009; Braun et al., 2013; Ems-McClung et al., 2020). We recently reported that these two modes of KIFC1, which are temporally and spatially orchestrated during the cell cycle (Cai et al., 2009; Ems-McClung et al., 2020), are also used in organizing microtubules of the axon (Muralidharan and Baas, 2019). Here, we posit that KIFC1, via these two modes, regulates the sliding of microtubules needed to ensure fidelity of the migratory journey. To test this, we used rat migratory neurons in vitro, developing mouse brain in vivo, RNA interference, and mutant KIFC1 constructs. As part of these studies, we investigated the roles of KIFC1 in INM as well as nuclear rotation that occurs during neuronal migration.
Materials and Methods
Experimental model and subject details
All primary neuronal cultures used in this study were generated from Sprague Dawley rats purchased from Charles River Laboratories. For in utero electroporation experiments, CD-1 mice were used, also purchased from Charles River Laboratories. All experiments on Sprague Dawley rats or CD-1 mice were approved by Drexel University Institutional Animal Care and Use Committee and conformed to National Institutes of Health standards. For primary neuronal cultures, fetuses/pups of both sexes were used. Cerebellar migratory neurons acquired from P1-P4 Sprague Dawley rat pups were cultured as previously described (Falnikar et al., 2011, 2013). RFL-6 rat fibroblasts were cultured with Ham's F12K medium, 80%; FBS, 20%.
siRNA-based depletion of KIFC1
Control scrambled siRNA (Sigma, SIC001) or KIFC1 siRNA (Sigma, SASI_Rn01_00080300, SASI_Rn01_00080301, SASI_Rn01_00080302, SASI_Rn01_00080304) was introduced into neurons by nucleofection (Nucleofector 2b, Lonza) before plating. Cells were cultured for 36-60 h (to allow for siRNA-targeted proteins to be depleted) on 35-mm-diameter plastic culture dishes (Thermo Fisher Scientific, 171099) coated with 1 mg/ml poly-L-lysine (Sigma, P2636-25MG), and then replated on glass-bottomed dishes (Cellvis, #D35-14-1.5-N) coated with 1 mg/ml poly-L-lysine. Cells were replated either as dissociated cells (to study the behavior of individual cells) or as aggregates (to assess the behavior of large numbers of cells). In the case of the aggregates, after 36 h of siRNA treatment, the transfected neurons were suspended in 5 ml of media in 50 ml tubes overnight in the incubator and then plated on coated glass-bottomed dishes to allow intact aggregates to settle and attach. In a limited number of experiments on RFL-6 rat fibroblasts, control scrambled siRNA or KIFC1 siRNA was introduced by nucleofection (Nucleofector 2b, Lonza) before plating. The cells were then cultured for 48 h and fixed with 100% ice-cold methanol for immunostaining analysis.
Pharmacologic studies
For pharmacologic inhibition of cytoplasmic dynein, cells were treated with 50 μm Ciliobrevin D (Firestone et al., 2012) (Millipore Sigma, 250401) for 1 h, after which cells were immediately fixed in 100% ice-cold methanol.
KIFC1 constructs
Full-length nonmutated human KIFC1 (termed WT-HSET) was obtained as a gift from Claire Walczak, as were the HSET-GFP; HSET-N593K mutant (“crosslinking-only mutant”) (Cai et al., 2009), which can crosslink microtubules but not slide them, and a motor stalk construct, which is a nonfunctional truncated mutant. A rigor mutant construct was also used, prepared as previously described (Muralidharan and Baas, 2019). HSET-Flag (EX-Z7417-M39) and Empty Flag (EX-NEG-M39) constructs were purchased from Genocoepia. We used WT-HSET gifted to us by Claire Walczak to clone mScarlet-WT-HSET, which was used for rescue experiments in vivo. We also used WT-HSET GFP, crosslinking-only mutant, rigor mutant, and HSET-Flag for some of the in vitro rescue experiments. The siRNA sequences chosen to target rat KIFC1 do not affect human KIFC1/HSET, hence permitting us to use HSET for rescue experiments. shRNA constructs used for in vivo experiments designed to target mouse KIFC1 and scrambled control sequences cloned were cloned into pSCV2-Venus plasmid (a gift from Franck Polleux) (Hand and Polleux, 2011). The target sequences used for KIFC1 and scrambled control were GTGCTCTGCTATTGGTACA and ACTACCGTTGTTATAGGTG, respectively.
Immunocytochemistry
For almost all experiments, neuron cultures were coextracted and fixed for 12 min in a solution containing 4% PFA, 1× PHEM buffer (PIPES, HEPES, EDTA, MgCl2), 0.2% glutaraldehyde, and 0.1% Triton X-100. Glutaraldehyde was then quenched by two treatments for 15 min each with 2 mg/ml sodium borohydride (Sigma-Aldrich, 452882-25G). Cultures were then blocked for 1 h with normal goat serum (Jackson ImmunoResearch Laboratories, #005-000-121), followed by incubation with primary antibodies overnight at 4°C and then secondary antibodies for 2 h at room temperature. For nuclear membrane morphology experiments, cells were fixed with 100% ice-cold methanol (Thermo Fisher Scientific, A412SK-4) for 15 min and then treated with 0.1% Triton X-100, after which immunostaining was performed as described above. Primary antibodies were rabbit anti-βIII-tubulin, mouse anti-βIII-tubulin, mouse anti-Lamin B1, and mouse anti-Lamin B2. Appropriate secondary antibodies purchased from Invitrogen were used. Secondary antibodies were goat anti-rabbit AlexaFluor-488, goat anti-mouse AlexaFluor-488, goat anti-rabbit AlexaFluor-555, and goat anti-mouse AlexaFluor-555. All antibody concentrations and catalog numbers are provided in Table 1.
Immunohistochemistry
For almost all tissue staining, mounted mouse brain sections were first treated with 0.3% Triton X-100 for 1 h. Then sections were quenched using a mixture of hydrogen peroxide (Thermo Fisher Scientific, H324500) in 100% ice-cold methanol for 1 h. Sections were then blocked for 1 h with normal goat serum (Jackson ImmunoResearch Laboratories, #005-000-121), followed by incubation with primary antibodies overnight at 4°C and then secondary antibodies for 2 h at room temperature. For all BrdU experiments, tissue was first treated with 0.3% Triton X-100 for 1 h followed by 1N HCl for another 90 min. Then quenching was done twice for 15 min each with 2 mg/ml sodium borohydride. Quenching and antibody steps were followed as described above. Primary antibodies were against the following: Cyclin D1, TBR2, PAX6, Nestin, NeuN, BrdU, Ki67, and histone-H3 (Phospho S10). Appropriate secondary antibodies were Cy3-AffiniPure Goat anti-Rat IgG, goat anti-rabbit AlexaFluor-647, and goat anti-mouse AlexaFluor-647. DAPI (Thermo Fisher Scientific, D1306) was used at 1:20,000 concentration to stain the nuclei. All images were acquired as Z-stack acquisitions of the tissue using Leica True Confocal System SP8 with 63× oil objective. All images were analyzed using ImageJ and prepared for publication using Adobe Photoshop.
Western blotting
Protein lysates were extracted from transfected cultures using Pierce RIPA Buffer (Thermo Fisher Scientific, 89901) supplemented with protease inhibitor cocktail (Thermo Fisher Scientific, 78430) and phosphatase inhibitor cocktail (Sigma, #P0044-1ML). Total protein concentration in the lysates was measured using the BCA protein assay (Thermo Fisher Scientific, 23227); 40 μg of protein was loaded per lane on 4%-15% gradient SDS-PAGE gels (Bio-Rad #456-1084), after which transfer was conducted onto PVDF membranes (Bio-Rad, #162-0177) at 4°C for 16-18 h at 30 V. Membranes were blocked for 1 h in 5% fat-free milk (Lab Scientific, #M0841) in 1× TBS-T (10× TBS Bio-Rad #170-6435), and then incubated in primary antibodies rabbit anti-KIFC1 and mouse anti-GAPDH overnight at 4°C, followed by peroxidase secondary antibodies for 2 h at room temperature. Blots were visualized using a mix of Super Signal West Pico kit (Thermo Fisher Scientific, #34580) and Super Signal West Femto kit (Thermo Fisher Scientific, #34094).
Microtubule sliding assays
After 60 h of siRNA treatment, cerebellar migratory neurons were replated after transfection with tdEos-tubulin onto glass-bottomed culture dishes coated with poly-D/L-lysine (0.1 mg/ml) and laminin (Thermo Fisher Scientific, #23017015). Sixteen hours after replating, dishes were placed in a live imaging workstation. Two bleached zones were made in the green channel: one zone at the soma and the other in the middle region of the leading process. Bleaches zones were made using the Andor Mosaic system, and images were captured every 60 s for ∼10 cycles in the green and red channels using a Zeiss AxioObserver Z1 equipped with a 63× Plan Apochromat objective (1.46 numerical aperture) and EMCCD. Decay over time was assayed of the red fluorescence as a readout of the sliding of long microtubules out of the photoconverted zone to the nonphotoconverted zone. The appearance of red fluorescence in the nonphotoconverted zone of the leading process flanking the photoconverted zone was also assayed as further evidence of the sliding of long microtubules.
End-binding protein (EB3) comet assay
EB3-GFP (EB3, which marks the plus ends of microtubules as they assemble) and hcent2-RFP (centrin, a centrosomal protein marker) plasmids were cotransfected into neurons before replating. Live imaging was conducted on an environmentally controlled microscope stage, with images acquired at 3 s intervals for a total of 120 frames using the Carl Zeiss 100×/1.46 α Plan-Apochromat objective. Directionality of the comets was manually counted with hcent2-RFP as a reference position and graphed out as percentages.
In utero electroporation
In utero electroporation was performed as previously described (Cornell et al., 2016; X. Liu et al., 2021). In brief, pregnant dams were anesthetized and the uterus was exposed. KIFC1 shRNA plasmid (1 μg) was then injected into the lateral ventricle of E13.5 or E15.5 embryos. Embryo heads were then held with forceps-style electrodes with the anode directed toward the cortex and electric pulses were applied (three pulses of 33 V with 50 ms intervals for E13.5 embryos and 35V for E15.5 embryos) using a CUY21 electroporator by Napa GENE. For BrdU-labeling experiments, mouse embryos were electroporated at E13.5 and BrdU (100 μg/g) was injected at E14.5. Animals were killed at indicated time points following BrdU injection. Brains were then harvested after 30 min, 3 h, 6 h, and 24 h and were fixed in 4% PFA and submerged in 15% sucrose overnight and in 30% sucrose subsequently; 20 μm cryosections were made, and the tissue was processed for immunohistochemistry.
For tracing the trajectory of cells in vivo, brain slices for live imaging were performed as previously described (Cornell et al., 2016). Briefly, in utero electroporation was prepared at E15.5 and embryos developed undisturbed until E17.5. Brains were then removed and placed in ice-cold ACSF. Brains were embedded in 4% low-melting agarose and slices were cut with a 300 μm thickness in ice-cold ACSF using a VTS1000 vibratome (Leica Microsystems). Slices were incubated in DMEM/F-12 imaging media without phenol red supplemented with 10% FBS for at least 1 h at 37°C, 5% CO2 for recovery. Slices were transferred to a 35 mm dish and submerged in neutralized rat tail collagen I (Invitrogen). Collagen was solidified for 30 min at 37°C, 5% CO2. Slices were then covered with imaging media. Time-lapse live imaging was performed using an upright confocal laser scanning microscope (TCS SP2 VIS/405, Leica Microsystems) with a 20× HCX APO L water-dipping objective (NA 0.5). During imaging, slices were cultured in the imaging media and kept at 37°C with 95% air/5% CO2 in a stage top chamber incubator (DH-40iL, Warner Instruments). Confocal Z-stack images were taken every 10 min for an average of 23 h.
For all in vivo rescue experiments, WT-HSET tagged with mScarlet was used. WT-HSET was electroporated at the same time as the shRNA. Rescued cells were identified as cells that express WT-HSET (mScarlet+) and KIFC1 shRNA (Venus-YFP+) for all analyses. Rescued cells were compared with the KIFC1-shRNA group to assess rescue of the relevant phenotype.
In vivo image analysis
All labeling indexes were calculated as percentages of the positive cell Venus-YFP+ divided by total number of Venus-YFP+ cells. For BrdU chase analysis, VZ was divided into 6 bins and the percentage of BrdU+ Venus-YFP+ cells was calculated. For neuronal migration analysis, the brain section was divided into 5 bins and the percentages of Venus-YFP+ cells were calculated to estimate the level of migration. For the aggregate migration assay, similar-size relatively circular aggregates were quantified. The midpoint of the aggregate was identified, and three concentric circles, each 20 µm bigger, were drawn around the aggregate and considered as bins. The intensity of DAPI and the percentage of DAPI+ cells were calculated and represented as a percentage. For nuclear rotation, all nuclear rotation trajectory tracing was performed using TrackMate plugins in Fiji (National Institutes of Health). For live image analysis, Z projections were made at each time point and compiled into a movie using Fiji. The Manual Tracking Plug-in was used to track the trajectory of individual somas during the imaging. Polar histograms were created using MATLAB to quantify the average direction of soma movement throughout imaging. Ninety degrees represents the cortical plate (CP), and 270° represents the VZ/subventricular zone (SVZ). Graphs were represented as polar histograms and percentages. Speed and displacement were calculated by exporting the data points from the tracks of individual somas.
In vitro image analysis
A complete nuclear rotation was defined as rotation of HP1β spots of >10° over a period of at least 1 min. For one rotation of the cell, trajectories of three HP1β-RFP spots were tracked in three-dimensional time-lapse images using TrackMate (Tinevez et al., 2017). Principal component analysis was performed on the three-dimensional spot data of all three trajectories, to predict the fitted plane. The axis of rotation was calculated by averaging the vectors of the three fitted planes. The angle between the rotation axis and the direction of migration was calculated in the range of 0°-90°. The direction of migration was set to 0°. For nuclear trajectory, the nucleus was followed for a period of 60 min and the analysis of trajectory behavior was performed similarly. For nuclear morphology, line scan intensity measurements were performed. A tracing line across the nucleus toward the leading process was drawn. Background intensities were subtracted, and data were presented as arbitrary fluorescence units. Microtubule sliding assays were analyzed using publicly available, open-source software, Fiji. The bleached zones (soma and segment of the leading process) were traced to measure the intensity decay. The length of the leading process was traced for intensity to measure the sliding of microtubules into the leading process. All Western blots were analyzed using Image Lab software.
For all in vitro rescue experiments, WT-HSET tagged with GFP/Flag was used. WT-HSET was ectopically expressed at the same time as the siRNA. Rescued cells were compared with the KIFC1-siRNA group to assess rescue of the relevant phenotype.
Experimental design and statistical analysis
Statistical analyses were conducted, and graphs constructed using Excel (Microsoft) or GraphPad (Prism version 7.0) unless otherwise stated. Data are shown as mean ± SEM unless stated otherwise. As for violin plots, box plots always outline distribution of the data. The middle line indicates the median of the data. The violin always extends from the 25th to 75th percentiles. All polar graph histograms and polar graphs were generated using Matplotlib: a 2D graphics environment. In each experiment, three independent repeats were performed. For in vivo experiments, all analysis was done from at least three embryos from three different independent timed pregnant mice. For statistical analysis, the mean difference was significant if p < 0.05. Specific statistical analyses are indicated in the figure legends. All data were checked for normality using the Shapiro–Wilk test. Parametric data were assessed using the t test to compare means and nonparametric data used the Mann–Whitney U test to compare medians. Multiple group comparison was performed by one-way ANOVA followed by Bonferroni post hoc analyses in the case of parametric data and Kruskal–Wallis test in the case of nonparametric data.
Results
KIFC1 is necessary for nuclear migration toward the VZ during INM
Before the peak of neurogenesis, neural progenitors (stem cells) undergo predominantly symmetric proliferative divisions to amplify the progenitor pool (Taverna and Huttner, 2010; Cooper, 2013). Coordinately with oscillatory movements called INM, cells undergo different stages of the cell cycle, that is, the cells move from the VZ by basal migration (G1 phase), reach the SVZ (S phase), move back into VZ by apical migration (G2 phase), and finally undergo mitosis at the VZ (Xie et al., 2007; Kriegstein and Alvarez-Buylla, 2009; Kosodo, 2012; Y. T. Yang et al., 2012). It has been established that cytoplasmic dynein is responsible for nuclear movement toward the apical surface while Kinesin-3 is responsible for basally directed nuclear movement (J. W. Tsai et al., 2010). These oscillatory movements happen so that the nucleus can reorient itself and translocate from and to the VZ. To determine whether and how KIFC1 contributes to nuclear migration during INM in cortical progenitors, we introduced KIFC1 shRNA or scrambled shRNA into cortical progenitors at the VZ in the developing mouse brain. shRNA efficacy was tested in neural stem cells cultured from E14.5 mouse brains and analyzed using Western blotting, which revealed the depletion of endogenous KIFC1 (Fig. 1A). To achieve this in vivo, we used in utero electroporation on E14.5 mouse brains. The shRNA was tagged with Venus-YFP to allow visualization of the transfected cells; 24 h after introducing the shRNA, we traced the progenitors by labeling them in S-phase using BrdU and investigated their positions after 30 min, 3 h, or 6 h by immunohistochemistry. Nuclei that had incorporated BrdU began to migrate toward the VZ after 30 min (Fig. 1B,C). Quantification of control shRNA cells colabeled with BrdU at 3 and 6 h indicated that most of these cells had reached the VZ by 6 h (Fig. 1D,E). Further quantification of KIFC1 shRNA-treated cells revealed that the nuclei of these progenitors had migrated comparatively less, with many of the nuclei remaining in the upper half of the VZ 6 h after injection with BrdU. Finally, we investigated whether the migration defect is only apical (downward to the VZ) and not basal (upward away from VZ). Basally migrating cells are distinguished by G1 and S phase of the cell cycle; therefore, we analyzed the percentage of basally migrating cells away from VZ by immunostaining for cyclin-D1 (G1 phase cells) (Fig. 1F,G) and BrdU (S phase cells) (Fig. 1H,I) and did not detect any difference between the control and KIFC1-depleted groups.
To investigate whether the apical INM defects associated with KIFC1 depletion might affect the progenitor pool, we conducted in utero electroporation similarly, where the progenitor cells were transfected with scrambled or KIFC1 shRNA in mouse embryos at E13.5. We then observed the number of Ki67+ (a proliferating cell marker) Venus-YFP+ cells at E15.5. Quantification revealed that the percentage of proliferating cells was lower in the KIFC1-depleted group (Fig. 2A,B). When nuclear movement toward the VZ is impaired, cells may enter mitosis prematurely because their nuclei no longer migrate back to the VZ. To assess this, we analyzed the position of transfected mitotic nuclei (PHH3+) relative to the VZ. We found that the overall distance between Venus-YFP+ PHH3+ nuclei and the ventricular surface was significantly greater compared with control (Fig. 2C,D). If these cells are entering mitosis prematurely away from VZ, they also would exit more readily, so we analyzed the number of overall Venus-YFP+ PHH3+; and true to this prediction, there was a smaller number of KIFC1-depleted cells undergoing mitosis (Fig. 2E). To further validate that the impaired migration apical migration (toward VZ) was not because of off-target effects of shRNA, we performed rescue experiments by expressing the WT human isoform of KIFC1 (termed WT-HSET) in vivo. WT-HSET was tagged with mScarlet and electroporated along with KIFC1-shRNA for visualization of HSET-expressing cells that also were depleted of endogenous KIFC1. Quantification revealed that Ki67+ mScarlet+ Venus-YFP+ cells in the rescue group were significantly higher in number compared with the number of Ki67+ Venus-YFP+ cells in the KIFC1-depleted group (Fig. 2A,B). Similarly, we also assessed the of number and distance from VZ, in both KIFC1-depleted and WT-HSET rescue group. Quantification revealed that Venus-YFP+ PHH3+ Scarlet+ cells from the rescue group were significantly higher than Venus-YFP+ PHH3+ from the KIFC1-depleted group (Fig. 2D). Finally, measuring the distance of the Venus-YFP+ PHH3+ Scarlet+ cells of the rescue group from the VZ from revealed that these cells were much closer than Venus-YFP+ PHH3+ from the KIFC1-depleted group (Fig. 2E). Collectively, these results indicate that WT-HSET is able to rescue the apical migration defects caused by KIFC1 shRNA, therefore leading us to conclude that these effects are specifically because of KIFC1 depletion and not off targeted effect of the shRNA.
To further assess the conclusion that the impairment in nuclear movement influenced progenitors to prematurely differentiate, we administered BrdU to label S-phase cells at E14.5 and analyzed progenitor proliferation at E15.5. We then quantified cell cycle exit index, which indicates the fraction of cells that were proliferating at E14.5 but exited the cell cycle at E15.5. This index was significantly increased, further supporting the idea that progenitors depleted of KIFC1 exited the cell cycle more rapidly when they are unable to migrate toward the VZ (Fig. 2F,G). Additionally, to establish that impairment in apical migration by KIFC1 during INM influenced neither mitosis nor duration of the cell cycle, we administered a short 30 min BrdU pulse at E15.5. In 30 min, mitosis is generally achieved in mammalian cells and the cell will move into S phase cells where the BrdU will be incorporated in the nuclei. After the incorporation of BrdU, we quantified the percentage of cycling proliferating progenitors Venus-YFP+ Ki67+ BrdU+ in S phase in the VZ (cell cycle length index). If there is an issue during mitosis, the cycle length index will differ between both the groups, but we did not observe any significant change in the cell cycle length index between both the groups (Fig. 3A,B). This further supports a key role for KIFC1 in apical migration while not being essential for mitosis of cortical progenitors. Because of the defect in apical migration, we did observe fewer apical progenitors (PAX6+ Venus-YFP+) at VZ (Fig. 3C,D) and far more basal progenitors (Tbr2+ Venus-YFP+) in the SVZ (Fig. 3E,F). Additionally, we observed an increase in the percentage of newborn neurons (NeuN+ Venus-YFP+) in the intermediate zone (IZ) and CP compared with in the VZ, suggesting that KIFC1 depletion led to premature differentiation of basal progenitors into neurons (Fig. 3G–I). Together, these data support a role for KIFC1 in apical migration during INM.
KIFC1 depletion disrupts trajectory of postmitotic neuronal migration
Since the neurons were born prematurely, we wanted to see whether KIFC1-depleted neurons reached the CP earlier than control neurons. Therefore, we introduced shRNA into cells using in utero electroporation at E16.5 and then waited for the pups to be born, after which the newborn mouse pups were immediately killed at postnatal day 0, and neuronal migration was analyzed. In the control group, the Venus-YFP+ neurons migrated from the VZ efficiently toward the CP. Contrary to the prediction of an early arrival at the CP, Venus-YFP+ neurons from the KIFC1-depleted group did not reach the CP but were spread throughout the cortex (Fig. 4A,B). Our cortical progenitor data suggest that, although they were prematurely born in the KIFC1-depleted group, the neurons nevertheless migrated far less efficiently to the CP. To further validate these findings, we rescued the migration defect by expressing the WT-HSET-mScarlet in vivo at similar time points as mentioned. Most mScarlet+ Venus-YFP+ cells were observed in the top one-third of the cortex toward the CP compared with Venus-YFP+ cells from the KIFC1-depleted group that were spread throughout the cortex. Therefore, collectively, the data indicate that, when KIFC1 was specifically depleted, migration of the soma was impaired in both neuronal progenitors during INM and in neurons during radial migration.
Next, we observed the migratory trajectory of these electroporated neurons to better appreciate and understand the role of KIFC1 in migration. For this, we performed live imaging of brain slices from the electroporated embryos and traced the trajectory of these cells for a period of 24 h in layer 2/3. Unlike control neurons, which migrate directly toward the CP, KIFC1-depleted neurons frequently changed their directions (Fig. 4C). We plotted these trajectories as a polar histogram to help us understand their behavior; the histogram revealed that a higher percentage of KIFC1-depleted neurons moved short distances and away from the CP and migrated sideways (IZ) and backward (SVZ/VZ) (Fig. 4D,E). We also calculated the speed of KIFC1-depleted neurons and found it to be significantly slower than that of control counterparts (Fig. 4F). Additionally, we observed a higher percentage of stationary than migratory cells in the KIFC1-depleted groups (Fig. 4G). These data suggest that, in the absence of KIFC1, the cell loses its ability to directionally migrate and maintain its appropriate course.
To elucidate the mechanistic role of KIFC1 in regulating migration trajectory, we cultured migratory neurons from P0-P4 rat pups in vitro. We used rat instead of mouse for the in vitro work because our previous in vitro studies on mitotic motors in migrating neurons were all done on rat (Baas and Ahmad, 2001; Muralidharan and Baas, 2019). We first subjected these neurons to KIFC1 siRNA treatment and plated aggregates of these neurons on laminin-coated dishes and allowed them to migrate outward from the aggregate to recapitulate migratory defects in culture. After 24 h, the cells were fixed, and outward migration was determined by measuring the number of DAPI+ cells moving away from the center of the aggregate. Analysis revealed that there was a deficit in the outward migration of KIFC1-depleted neurons compared with controls and the leading process was longer (Fig. 4H–J). Additionally, live imaging of these cultured aggregates revealed that KIFC1-depleted neurons had a significantly slower speed (Fig. 4K).
We then tested the capacity to rescue the migration phenotype in vitro using WT-HSET, as well as three different mutants of KIFC1 in a simple migration assay. WT-HSET and all three mutants were characterized in our previously published work on the axon (Muralidharan and Baas, 2019). The mutants include the following: (1) the crosslinking-only mutant that bears a mutation in the motor domain such that it cannot hydrolyze ATP but can crosslink microtubules; and (2) the rigor-mutant, which behaves similarly to the crosslinking-only mutant, except that once the microtubules are crosslinked, they are only the WT-HSET rescued the impaired migration phenotype in the aggregate assay, with none of the mutants showing even a partial rescue (Fig. 4J).
KIFC1 is indispensable for directional rotation of the nucleus
Given that both INM and neuronal migration require directional rotation of the nucleus, we wondered whether KIFC1 might also play a role in this process during neuronal migration. To explore this, we transfected the KIFC1-depleted neurons in vitro with fluorescently tagged HP1β, a nuclear histone protein classically used to trace nucleokinesis behavior. The tagged protein appears as fluorescent spots in the nucleus that can be used as fiduciary marks to assess nuclear rotation (Kengaku, 2018; Wu et al., 2018). As expected, control neurons efficiently migrated forward, and their nucleus translocated along with the soma as it moved. In contrast, KIFC1-depleted neurons were notably less efficient in migrating forward, and the nucleus correspondingly advanced less efficiently, with some nuclei of KIFC1-depleted neurons moving forward and backward (Fig. 5A,B). We also rescued the less efficient migration of the nucleus by ectopically expressing the human isoform of KIFC1, WT-HSET (Fig. 5A,B). Throughout migration, transient peaking at the leading edge of the nucleus was observed during somal translocation, which corresponds to the pull of the dynein-driven forces; this phenomenon was amplified in the case of KIFC1-depleted neurons. Whereas control cells showed a single transient peaking from the nucleus in the direction of neuronal migration (Fig. 5A), KIFC1-depleted cells frequently showed much larger outpocketings, usually bulbous in shape, that were only rarely observed in control neurons. After expression of WT-HSET, nucleus of migrating neurons no longer exhibited prolonged large outpocketings (Fig. 5C–E). For quantification, we calculated outpocketing index, which represents the ratio of cells that exhibited these outpocketings from 100 cells sampled from three randomly selected regions, as well as the time period over which the outpocketings lasted, which we refer to as the outpocketing duration.
Additionally, we observed that the nucleus of control neurons would often make a rotation forward before the soma would migrate, whereas the nuclei of KIFC1-depleted neurons did not rotate. Therefore, we tracked three HP1β spots defined as anchor points for a period of 1 min in three-dimensional time-lapse images using TrackMate (Fig. 6A). Principal component analysis was used on the three-dimensional trajectories of the spots to calculate the fitted plane which predicts the axis of rotation and the angle of rotation (Fig. 6B,B′). The angle between the rotation axis and the direction of migration was calculated in the range of 0-90°. The direction of migration (the leading process) was set to 0°. Consistent with previous published studies, the axis of rotation in control nuclei was almost perpendicular to the leading process (Kengaku, 2018; Wu et al., 2018; Nakazawa and Kengaku, 2020). However, after KIFC1 depletion, neurons had a smaller angle between the rotation axis and leading process, indicating that the nucleus did not directionally rotate in the manner of the control (Fig. 6C). To investigate the properties of KIFC1 relevant to these observations, we depleted the endogenous KIFC1 and observed that WT-HSET was able to rescue the phenotype, but the KIFC1 crosslinking-only mutant was only able to partially restore the rotation axis (Fig. 6C). These results indicate that both the motor properties and the microtubule crosslinking properties of KIFC1 are required to perform its function in nuclear rotation.
From there, we tracked and traced the HP1β spots for a period of an hour, and the trajectory of these spots was displayed on a three-dimensional graph. Control nuclei directionally steered smoothly as the cell moved forward with complete nuclear rotations, whereas the nuclei of KIFC1-depleted neurons made incomplete rotations, or these nuclei were observed to be pulled in different directions (Fig. 6D,E), resulting in shorter rotation and displacement (Fig. 6F,G). Analyses of the aspect ratio of the nucleus of KIFC1-depleted neurons revealed a higher ratio, which suggests that the nucleus was more elongated (Fig. 6H). We further validated these finding by rescuing the directional rotation defect by expressing WT-HSET (Fig. 6F,G). Together, these data further fortify the conclusion that the nucleus of control neurons rotates toward the direction of migration, but this is not the case with KIFC1-depleted neurons.
No peaking or outpocketing of the nucleus was observed under conditions of dynein (heavy chain) depletion by RNAi (or inhibition by ciliobrevin; data not shown), supporting our contention from earlier that these phenomena are a manifestation of the dynein-driven forces that translocate the nucleus. Dynein depletion inhibited nuclear rotation, but unlike the case with KIFC1 depletion, dynein-depleted neurons were completely stationary as were their nuclei (Fig. 6E–H). These results indicate that KIFC1 is not rotating the nucleus in a fashion independent of its movement by cytoplasmic dynein.
Loss of directional migration stresses the nucleus
Previous studies have shown that nuclear migration by cytoplasmic dynein is aided by its interaction with the LINC (Linker of Nucleoskeleton and Cytoskeleton) complex. The LINC complex constitutes physical bridges between the nuclear lamina and the microtubule, thereby anchoring the nucleus to the microtubule cage allowing the nucleus to be oriented toward the direction of migration. We suspect that the orientation of rotation enabled by KIFC1 would functionally protect the nucleus from undue stress under the pulling forces of cytoplasmic dynein. To visualize if disrupted nuclear rotation had an impact on nuclear membrane integrity, we immunostained for lamin-B proteins (B1 and B2), which are intermediate filament proteins found in the nuclear envelope. We conducted line scan intensity analysis with a line drawn across the nucleus toward the leading process. The line indicates the fluorescence intensity at the nuclear rim and the nucleoplasm across the nucleus as a measure of membrane integrity. The KIFC1-depleted neurons displayed asymmetric distribution of nuclear lamina at the rim where lamins B1 and B2 were increasingly present at the leading edge of the nucleus facing the leading process, with overall decreasing intensity of lamins B1 and B2, compared with their control counterparts. Moreover, regions of chromatin not surrounded by nuclear lamina were observed all over the nucleus in an uneven asymmetric distribution in KIFC1-depleted cells, indicating that cytoplasmic dynein exerts forces all around the nucleus. Additionally, ectopic expression of HSET-GFP restored the asymmetric distribution of nuclear lamina seen at the rim of the nucleus (Fig. 7A–D). Moreover, we also quantified that the number of neurons depleted with KIFC1 siRNA showed more nuclei with asymmetric distribution of lamin B2 were more than their control counterparts (Fig. 7E). As discussed earlier, the aspect ratio (ratio of nucleus width by nucleus length) of migrating neurons indicated that KIFC1-depleted cells were more elongated (Fig. 6H). These results together indicate that, when the nucleus of the cell moves with no specific orientation, the nucleus is susceptible to being abnormally elongated. Treatment with Ciliobrevin D (Fig. 7E–J) partially rescued the asymmetric distribution of lamina along the rim of the nucleus, supporting the conclusion that dynein's translocation of the nucleus along microtubules imposes the stress on the nucleus that is mitigated by KIFC1.
KIFC1 regulates the forces of microtubule sliding, thereby orienting directional migration
Our analysis on the directionality of nucleokinesis shows that KIFC1 plays a key role in maintaining the rotation axis in reference to the leading process. On this basis, we posit that the neuron can move forward when KIFC1 crosslinks microtubules on both sides of the neuron, enabling the dynein forces to achieve enough traction to propel the soma forward. However, when KIFC1 does not crosslink the microtubules, the microtubules surrounding the nucleus are no longer able to support the directionality and tilting of the soma because the microtubules now have too great of a propensity to slide. This prevents cytoplasmic dynein from achieving enough traction to directionally propel the soma while not mitigating its forces on the nucleus, thus leading to stress and deformation of the nuclear membrane. We therefore speculate that, if the KIFC1 crosslinking is undone on only one side of the neuron, the neuron should tilt its trajectory away from that side. To assess this, we investigated microtubule organization and sliding in neurons depleted of KIFC1. First, we visualized the plus-ends of growing microtubules by transfecting GFP-tagged EB3, which marks the plus ends of growing microtubules. The cells were also transfected with RFP-tagged hcent2 (centrin), which marks the centrosome. We observed EB3-GFP comets in the cell soma in reference to the centrosome marked by hcent2-RFP. In control neurons, we saw ∼80% of comets emanate from the front of the soma (i.e., location of the centrosome) and these comets moved toward the trailing process, whereas the rest moved toward the leading process. This ratio was disrupted in KIFC1-depleted neuron (Fig. 8A,B), indicating that KIFC1 plays a significant role in the normal organization of perinuclear microtubules, presumably by regulating their sliding by other motors.
To further investigate this, we expressed tdEos tubulin ectopically in cultured migratory neurons. The photo-convertible tag termed tdEos fluoresces in the green channel until bleached in the green channel, after which it fluoresces in the red channel. To visualize the amount of sliding regulated at the soma by KIFC1, we photoconverted the tubulin population in the soma and assessed intensities of the photoconverted tubulin as an indicator of the level of microtubule sliding. The rate of signal decay of the converted tubulin was measured in the red channel, the idea being that more sliding of microtubules would result in a greater decay of fluorescence signal at the photoconverted zone and emergence of signal in the regions flanking the converted zone (Fig. 8C). True to our prediction, KIFC1-depleted neurons showed higher red tubulin signal down the length of the leading process, indicating that microtubules from the soma had slid into the leading process (Fig. 8D,E). To test whether KIFC1 regulates microtubule sliding only at the soma and not in leading process, two distinct regions of the cell will be converted: one in the soma and another in the middle one-third of the leading process. Once converted, dual-channel live imaging of the neurons was performed. In KIFC1-depleted neurons, red and green images merged more notably than in controls, with a greater amount of microtubule movement in the initial region of the leading process. This was indicated by development of yellow in the regions flanking the soma but not in the borders of the region converted in the middle of the leading process (Fig. 8F). Corroborating the emergence of yellow only at the region flanking the soma in KIFC1-depleted neurons, we observed decay in the photoconverted tubulin fluorescence signal only at the soma and not at the leading process (Fig. 8F–H). We ectopically expressed the crosslinking-only mutant or WT-HSET individually to rescue the excessive sliding we observed in the soma, but we were only partially able to rescue the phenotype using the crosslinking-only mutant but were able to efficiently rescue the excessive sliding using the human isoform HSET. Together, these results indicate that KIFC1's crosslinking and sliding properties are both necessary to regulate appropriate sliding in the soma (Fig. 8G,H).
Discussion
While cytoplasmic dynein is the primary workhorse for generating forces that slide and organize microtubules in neurons, also participating are kinesins classically considered mitosis-specific that continue to be expressed after the final mitotic division of the neuroblast (Schaar and McConnell, 2005; Silverman et al., 2010; Cooper, 2013). These kinesins play important roles in organizing the polarity patterns of axonal and dendritic microtubules, regulating the transport of short microtubules in these processes, and imposing forces on longer microtubules critical for determining whether axons grow or retract and for their navigation toward their targets (Baas, 1999; Baas et al., 2006; Nadar et al., 2008; Rao et al., 2017). Even before the elaboration of these processes from a stationary soma, kinesins play critical roles in regulating INM and neuronal migration. For example, kinesin-6 subfamily members KIF23 and KIF20B are important for paring down the multipolar morphology of newborn neurons into the classic bipolar morphology of a migrating neuron, and they also affect INM (Falnikar et al., 2013; Janisch et al., 2013). Kinesins 5 and 12, on the other hand, act as brakes on the sliding of microtubules during neuronal migration (Myers and Baas, 2007; M. Liu et al., 2010). Specifically, when kinesin 5 or 12 is inhibited or depleted, the neuron takes on a longer leading process and the cell migrates faster because there is greater sliding of microtubules and less opposition to cytoplasmic dynein (Falnikar et al., 2011, 2013). While kinesins 5 and 12 have not been explored in INM, a common regulator of them called TPX2 has been shown to be important for maintaining the polarity of neural progenitors and promoting apical nuclear migration during INM, possibly by regulating microtubule sliding (Kosodo et al., 2011). We previously showed that, when microtubule sliding is pharmacologically inhibited throughout the migratory neuron, the leading process becomes shorter, microtubules buckle, and the neuron deviates from its normal path (Rao et al., 2016). On this basis, we posited that a small degree of microtubule sliding is important during neuronal migration, and the same is probably true of INM.
Previous studies established that, while most microtubules in the migratory neuron are attached to the centrosome, small populations of centrosome-unattached microtubules are also present (Umeshima et al., 2007; Kriegstein and Alvarez-Buylla, 2009). Some of the centrosome-unattached microtubules are short, and transit anterogradely down the leading process. Longer centrosome-unattached microtubules extend their plus ends into the leading process and their minus ends behind the centrosome. In the soma, these microtubules intermingle with the microtubules of the nuclear cage (Umeshima et al., 2007; Wu and Kengaku, 2018; Wu et al., 2018), thus allowing for microtubule-microtubule sliding. While the length of the leading process may be determined in part by the activity of mitotic motors within the leading process, the trajectory of the cell is most likely because of sliding of these microtubules in the soma.
In a recent study on the axon, we found that KIFC1 has the capacity to slide microtubules but also the capacity to crosslink microtubules in a manner that prohibits their sliding by other motors (Muralidharan and Baas, 2019). These properties, which are consistent with the known properties of KIFC1 in cell-free motility assays (Hentrich and Surrey, 2010; Braun et al., 2017; Norris et al., 2018), are unlike any other motor that we had studied thus far in neurons. Kinesins 5 and 12 can act as brakes on sliding because they are such slow motors that they impose a limit on how fast any other motor can slide the microtubules (Falnikar et al., 2011, 2013). On the other hand, KIFC1 is a true “static” crosslinker because it becomes immobilized on the microtubules in an ATP-independent fashion, thus linking them together into a superstructure that resists sliding (Cai et al., 2009; Weaver et al., 2015). When KIFC1 was depleted or inhibited in axon-bearing neurons in our previous studies, there were significant changes in the morphology of the axon, the polarity orientation of its microtubules, and the ratio of short microtubules moving in each direction (Muralidharan and Baas, 2019). Both the microtubule sliding and crosslinking properties of KIFC1 contributed to these effects, as indicated by the phenotype not being rescuable without restoring both. Here we found that depletion of KIFC1 from neurons undergoing neuronal migration also results in phenotypes dependent on both the sliding and crosslinking properties of KIFC1. We found, both in vivo and in vitro, that depletion of KIFC1 causes the migrating soma to lose its appropriate trajectory and even to collapse back on itself. This phenotype, like the axonal phenotype of KIFC1 inhibition or depletion, is partially, but not completely, rescued by a KIFC1 mutant that can crosslink but cannot slide microtubules. Based on results presented here, we conclude that a sufficiently KIFC1-crosslinked microtubule array is required for dynein-driven forces to achieve enough traction to thrust the cell forward. As suggested by our live imaging data, small bouts of microtubule sliding enabled by switching KIFC1's mode from crosslinking to sliding in specific locales within the soma would be enough to enable the neuron to tilt directionally, thus providing midcourse corrections in the trajectory of neuronal migration (Fig. 9).
Other data presented here suggest that regions of microtubule sliding regulated by KIFC1 support the trajectory of migration in a second way, by helping the nucleus to rotate directionally and in the direction of migration as it is moved along by cytoplasmic dynein (J. W. Tsai et al., 2007). This function repurposes a similar role our data indicate for KIFC1 in INM and is consistent with data from other cell types showing interactions between KIFC1 and membrane proteins (W. X. Yang et al., 2006). In the absence of KIFC1's activity, the nucleus does not orient itself to the direction of migration and displays atypical prolonged outpocketings because of the pulling forces of cytoplasmic dynein. These atypical outpocketings reflect the stress on the nucleus that puts nuclear lamina at risk and are consistent with a loss of appropriate directionality in the absence of KIFC1's trajectory corrections. As to how exactly the microtubule-microtubule sliding plays into this role for KIFC1, we speculate that the nucleus would not be able to rotate properly if cytoplasmic dynein were moving it along two adjacent long immobile microtubules. However, inserting short sliding microtubules into the machine would allow for the needed maneuverability (Fig. 9). Regulation of nuclear position, nuclear movement, nuclear shaping, and cell migration by KIFC1 is not unique to migrating neurons but also seen in skeletal muscle differentiation (Gache et al., 2017), spermatogenesis (Ma et al., 2017a,b), and cancer metastasis (Ogden et al., 2017). These various studies emphasize KIFC1's role in regulating neuronal trajectory by directing the nucleus.
In conclusion, the present results indicate that KIFC1, a member of the minus-end-directed kinesin-14 subfamily of kinesins, has a unique role to play in neuronal migration and INM, different from that of cytoplasmic dynein or any other molecular motor studied to date. Fortifying the importance of the unique crosslinking abilities of KIFC1, other studies have shown that KIFC3, a closely related kinesin-14 family member that cannot crosslink microtubules in the same fashion as KIFC1, does not overlap in these functions, as evidenced by the lack of a neuronal migration phenotype when knocked down (Cao et al., 2020). Unlike other mitotic kinesins, expression levels of KIFC1 remain high in neurons into adulthood (Silverman et al., 2010), suggesting that KIFC1's specialized ability to alternately crosslink and slide microtubules is used for various purposes throughout the life of the neuron. For this reason, caution is due in using KIFC1 drugs for cancer treatment, especially if the drugs can enter the brain.
Footnotes
This work was supported by National Institute of Neurological Disorders and Stroke Grants R01 NS28785 and R01 NS118117 and National Institute on Aging Grant R21AG068597 to P.W.B.; and National Institute of Neurological Disorders and Stroke Grant R01 NS096098 to K.T-o., H.M. and A.P. were supported by a Dean's Fellowship for Excellence in Collaborative or Themed Research from Drexel University. S.A.B. was supported by National Institute of Child Health & Human Development predoctoral fellowship F31 HD103405. We thank Dr. Liang Qiang, Dr. Wenqian Yu, and Seyma Akarsu for advice with experimental design.
The authors declare no competing financial interests.
- Correspondence should be addressed to Peter W. Baas at pwb22{at}drexel.edu