Introduction

Precisely controlled dendritic arborization, spine morphogenesis, and neuronal migration ensure functional neural circuit assembly and neural network integrity during cortical development.1 Abnormal neurite patterning and disturbed neuronal positioning result in various MCD, including lissencephaly/pachygyria (LIS), polymicrogyria (PMG), and microcephaly.2,3,4,5 These disorders have been observed in individuals with neurodevelopmental disorders, such as seizures, intellectual disability, Rett syndrome, and autism spectrum disorder (ASD).6,7,8,9 Therefore, it is necessary to find the factors that regulate the growth and migration of neurons in the early stage of cortical development to understand the possible molecular mechanisms of the pathogenesis of MCD.

The KIF5C gene encodes heavy chain isoform 5C of kinesin, a microtubule-dependent motor protein that generates force during intraneuronal transport. It transports various cargos, including vesicles, membranous organelles, protein complexes, mRNAs, and chromosomes, over long distances along neuronal microtubules. These molecular motors are expected to play critical roles in several processes in the brain, including neuronal function, development, survival, and plasticity, by regulating transport within the axons, dendrites, and synapses of neurons.10,11,12,13

Previous studies have provided evidence that kinesin superfamily members play a fundamental role in early brain development. In vivo knockout of the Kif1a, Kif1b, Kif2a, Kif3a, Kif3b, Kif4a, Kif5a, Kif5b, or Kif5c genes result in various neurological phenotypes, including brain structural malformation, abnormal brain size, disturbed neuronal proliferation or apoptosis, altered synapse density, and even perinatal death due to severe neurological problems.13,14,15,16,17,18,19,20,21,22,23 The early embryonic death of Kif1b, kif2a, Kif3a/3b, and Kif5b knockout mice suggests an essential function in general developmental processes as well.16,17,18,23 Because genetic diagnostic technology has improved, an increasing number of clinical findings have revealed that KIF gene (KIF1A, KIF2A, KIF4A, KIF5A, KIF5C, KIF7, KIF11, and KIF14) deficiency is associated with various neurodevelopmental defects24,25,26,27,28,29,30,31,32,33,34,35,36,37 (Supplement Table I). With the widespread application of next-generation sequencing technology, several clinical studies have identified mutations in KIF5C in patients with MCD, suggesting its critical role in cortical lamination.27,28,38,39,40,41,42

In the present study, we investigated the effects of Kif5c knockdown on neurite branching, spine morphology, and cortical migration during early brain development. We found that Kif5c is required for proper dendritic branching and spine shaping. Kif5c knockdown causes a temporary delay in migration. Furthermore, the expression of a group of genes related to neuronal morphology and migration is altered after Kif5c knockdown in cortical neurons. These findings indicate that Kif5c is essential for cortical neuron growth and function in the early development stage and provide new insight into the neuronal pathology of Kif5c-related MCD in humans.

Result

Case series

Eleven cases of KIF5C mutation-related MCD have been reported since 2013 (Supplement Table II). According to the summary of clinical and genetic findings, we found that 100% of patients had malformations of cortical development and microcephaly, 5 patients were diagnosed with pachygyria and 2 with polymicrogyria. Five of six patients with EEG data presented seizures before 1 year of age, 4 patients presented generalized seizures, among which 2 presented intractable seizures. Six of six patients developed severe psychomotor and cognitive impairment and five of five patients presented abnormal behaviors, such as stereotypic hand movements and auto-mutilation. Additionally, nine of eleven patients carried p.E237V hot spot mutations, and the other 2 patients presented p.A268S and p.Y135C mutations. In addition to cortical malformations, some patients also exhibited an abnormal callosum, cerebellum, and increased ventricles. These findings elucidate that KIF5C might have a critical role in early human brain development.

Kif5c expression during cerebral cortex maturation

In vertebrates, the Kif5 gene family has 3 members, Kif5a, 5b, and 5c. Kif5b is widely expressed in all tissues, while Kif5a and Kif5c are mainly expressed in neurons.23 We reviewed publicly available single-cell data and found that KIF5C was predominantly expressed in the developing cerebral tissue in mice and humans (GSE156793 and GSE132044). We validated the Kif5c mRNA levels in the mouse cortical tissue at the early postnatal period using quantitative PCR. We observed increasing Kif5c expression after E14.5. The peak expression was detected between E16.5 and P0, and then the expression gradually decreased to a low level (Fig. 1a).

Fig. 1: Kif5c is expressed in neurons of the early developing cortex.
figure 1

a Quantitative analysis of the Kif5c mRNA expression in the mouse early cortical tissues (n = 3 pups at each time point) using Q-PCR. The expression level of Kif5c in P0 mice was considered as 1. Error bars indicate ± SEM. b Schematic diagram of the development of forebrain organoids that recapitulate the inside-out pattern of human cortical lamination. The pictures indicate the neuronal lamination of the human forebrain organoid (upper panel) and mouse cerebral cortex (lower panel). The green arrows show the direction of neuronal migration. c Quantitative analysis of the Kif5c mRNA expression level in forebrain organoids (n = 3 at each time point) using Q-PCR. The Kif5c expression level at DIV0 was considered as 1. Error bars indicate ± SEM. d Example pictures of forebrain organoids stained with SOX2 (green), CTIP2 (red), and CUX1 (red) antibodies. CTIP2 and CUX1 are the marks of the fifth and second layers of the cortical plate, respectively. Scale bar, 100 μm.

Full size image

We further examined the KIF5C expression pattern during the early development of the human brain. We used human induced pluripotent stem cells (iPSCs) to generate forebrain organoids to investigate KIF5C expression. The lamination of forebrain organoids mimics the inside-out pattern of the developing cortex in humans (Fig. 1b).43 KIF5C expression was significantly increased and reached the peak level at approximately Day 30, when V layer of the organoid was formed (Fig. 1c). These findings indicate a critical role of KIF5C in neuronal lamination. In summary, the expression results of humans and mice suggest that Kif5c is pivotal for developing cortical neurons.

Kif5c knockdown leads to abnormal cortical neuronal migration

We performed in utero electroporation to reduce Kif5c expression in mouse cortical neural progenitors to investigate its role in cortical neuron migration. First, we designed three candidate shRNA (fugw-H1-GFP) targeted Kif5c (shKif5c-#1, shKif5c-#2, and shKif5c-#3) and shDsred as control. Then we infected mouse cortical primary neurons with lentivirus containing those shRNAs. We found that the neurons expressing shKif5c-#2 showed a significant decrease in the Kif5c mRNA level (Fig. 2c). Therefore, shKif5c-#2 was used in subsequent experiments to reduce Kif5c in vivo and in vitro.

Fig. 2: Kif5c knockdown results in delayed neuron migration in vivo.
figure 2

a, b Example pictures of E18.5 mouse brain expressed the indicated plasmids using in utero electroporation at E13.5. Scale bar, 200 μm. c Quantitation of Kif5c mRNA expression levels relative to control (normalized to mouse Gapdh). Mouse primary cortical neurons were infected with lentivirus-mediated shRNAs at 0 DIV and harvested at 5 DIV. **P = 0.0003 for control vs. shKif5c-2, *P = 0.0226 for control vs. shKif5c-3, evaluated using Student’s t-test. Error bars indicate ± SEM. All data were obtained from three independent experiments. d Arrangement of cells expressed indicated shRNAs from the brain ependyma to the pia (a total of 13 pups from 3 dams under each condition were analyzed). Error bars indicate ± SEM. ****P < 0.0001 by multiple t-tests.

Full size image

We introduced shKif5c-#2 and shDsred into the embryonic mouse brain at E13.5, and the embryos were fixed and analyzed at E18.5. These constructs were injected into the lateral brain ventricle. Therefore, the shRNA was only expressed in GFP-labeled cortical ventricular zone (VZ) neurons, allowing us to trace neuronal migration from the VZ. At E18.5, most GFP-expressing neurons left the VZ and arrived at the intermediate zone (IZ) and the cortical plate (CP) in both groups. However, far more shDsred-expressing neurons reached the surface of the CP than shKif5c-expressing neurons (Fig. 2).

However, all GFP-labeled neurons of both groups arrived at the surface of the CP at P30, suggesting that early Kif5c deficiency leads to a short delay in cortical migration. These results suggest that Kif5c is necessary for the precise control of cortical neuron migration.

Kif5c knockdown leads to abnormal dendrite branching and spine development in vitro

Inhibition of kinesin heavy chain expression using antisense oligonucleotides leads to a significant reduction in total neurite length.44 Therefore, we knocked down Kif5c in primary neurons of the cortex to check whether Kif5c is involved in neuronal dendrite growth. Neurons were transfected with shDsred (control) and shKif5c-#2 by electroporation at 1 DIV, and then we performed anti-GFP immunostaining at 10 DIV. We found that neurons transfected with shKif5c-#2 had fewer primary dendrites, but no significant difference in the total dendrite length was found between the shKif5c-expressing neurons and the controls (Fig. 3). Thus, during early neuronal morphogenesis, Kif5c deficiency disturbs dendrite outgrowth in vitro.

Fig. 3: Kif5c regulates dendritic branching in vitro.
figure 3

a, b Example pictures of 10 DIV primary neurons transfected with the indicated constructs. Electroporation was performed at 1 DIV and immunolabeling was conducted at 10 DIV. Scale bar, 50 μm. c Quantitation of dendritic architecture in neurons collected at 10 DIV using Sholl analysis (n = 93 for control and n = 73 for shKif5c, Fiji, and ImageJ software). **P (10 μm) = 0.0005, ****P (20 μm) < 0.0001, **P (30 μm) = 0.0005, **P (30 μm) = 0.0005 using multiple t-tests. Error bars indicate ± SEM. d Quantitation of the total dendritic length of primary neurons transfected with the indicated plasmids. Error bars indicate ± SEM. P = 0.3569 by Student’s t-test.

Full size image

After in utero electroporation at E13.5, we collected the brain at P30 to analyze the density of dendritic spines with different morphologies. The thin spines in shKif5c-expressing neurons were denser than those in the control neurons. However, no significant difference in mushroom spines was found between the two groups (Fig. 4). The disturbed spine development might lead to increased neuronal excitability and induce seizures. In the above 6 patients with accessible EEG data, 83% (5/6) had seizures within 1 year of age. Based on these results, KIF5C deficiency might contribute to seizures by affecting early dendritic spine development.

Fig. 4: Kif5c knockdown alters the spine morphology.
figure 4

a, b Representative images of cortical layers II/III in the mouse brain electroporated with the indicated constructs (apical dendrites, P30, n = 3 mice per group). Scale bar, 4 μm. c Quantitation of the density of spines with different morphologies (a total of 4 brain slices in each group were randomly analyzed). *P = 0.0452 by Student’s t-test. Error bars indicate SEM.

Full size image

In addition, all GFP-labeled cells arrived at the surface of the CP in both the shKif5c and shDsred groups, indicating that kif5c deficiency would not lead to permanent termination of cortical migration. Taken together, Kif5c is required for neurite development and thus regulates the establishment of the functional neuronal circuits in the brain.

A neuronal development-associated gene set is differentially expressed between the shKif5c and control groups

We collected the total RNA from neurons transfected with lentivirus containing shKif5c vs. shDsred for RNA sequencing to define the transcriptomic differences between two groups. The mRNA levels of 204 genes were altered in Kif5c-deficient neurons compared with control neurons (Fig. 5a). Among these genes, 49 genes were upregulated, and 155 genes were downregulated (adjusted P-value < 0.01, log2(fold-change) > 1). Then, a Gene Ontology (GO) analysis of altered genes related to neuronal development was performed (Fig. 5b, c) according to the phenotypes observed in shKif5c-expressing neurons in vivo and in vitro. We further validated the alteration of 7 genes, including Kif5c, Itgb1, Nrp1, Nrp2, Gli3, and Pafah1b1 (Fig. 5d).

Fig. 5: GO analysis of differentially expressed genes in neurons after Kif5c knockdown.
figure 5

a Differentially expressed transcripts (highlighted in color, adjusted P-value < 0.01, log2(fold-change) > 1) between cortical neurons transfected with lentiviruses containing shKif5c and shDsred (control). b, c Gene Ontology (GO) analysis of altered genes related to neuronal development in cortical neurons after Kif5c knockdown compared with the control (b for upregulated and c for downregulated). Each subfigure shows a heat map clustered on rows and columns separately to visualize the genes enriched in each GO term. Colored boxes in the matrix represent differentially expressed genes (columns) belonging to GO terms (rows). Genes and GO terms were clustered in the existing binary matrix, and genes or GO terms with similar expression patterns were grouped (different groups were separated by black lines). Adjusted P-values (BH) for enriched GO terms are shown at the left of the corresponding term, with darker colors indicating greater significance. d Validation of altered genes using real-time PCR analysis. Data are presented as the means ± SEM; n = 3 experiments; two-tailed unpaired Student’s t-test, **P(Itgb1) = 0.0011, ****P(Nrp1) < 0.0001, ***P(Nrp2) = 0.0004, ****P(Gli3) < 0.0001, ****P(Neurog2) < 0.0001, ****P(Pafah1b1) < 0.0001, ***P(Kif5c) = 0.0007.

Full size image

Discussion

Only one Kif5 gene is expressed in the invertebrates Drosophila, Caenorhabditis elegant, and Aplysia, while 3 Kif5 variants are expressed in vertebrates, namely, Kif5a, Kif5b, and Kif5c.10 Kif5a and Kif5c are mainly expressed in neurons, and Kif5b is ubiquitously expressed. The expression of Kif5c in the cortex, hippocampus, cerebellum, and spinal cord is 3 to 5 times higher than that of Kif5a. These findings indicated that Kif5c might play a more important role in the function of the central nervous system (CNS). Although the three Kif5s have some functional redundancy,23 more recent findings have revealed their specific functions. For example, the transport of mitochondria in axons only depends on KIF5A in zebrafish.45 Knockdown of Kif5b and Kif5c leads to a significant reduction in the number of dendritic spines on mouse primary cortical neurons, while Kif5a deficiency does not affect spine development.46 Clinical findings showed that KIF5A deficits cause spastic paraplegia 10 (SPG 10, OMIM#602821) and neonatal intractable myoclonus (NEIMY, OMIM# 617235). Recently, several cases carrying rare KIF5C mutations have been reported. All of these patients presented microcephaly and malformation of cortical development (such as polymicrogyria, pachygyria, and cortical dysplasia), and some had experienced seizures beginning at an early age, suggesting its deficiency might lead to early neuronal development abnormalities.

We found that Kif5c is expressed at high levels during early brain development and Kif5c deficiency leads to abnormal neuronal morphology and radial migration. By performing an embryonic electroporation study, we found that Kif5c deficiency leads to an increased density of thin spines on cortical neurons during early brain development in vivo. Swarnkar et al. also reported a significant increase in the number of thin spines following Kif5c knockdown, but the number of mushroom spines was reduced.47 However, Zhao et al. observed reductions in the densities of both thin and mushroom spines following Kif5c knockdown in hippocampal neurons in vitro.46 Willemsen et al. found that KIF5C-E237K, the hot spot mutation, showed a significant reduction in distal neurite localization, which led to a significant decrease in the mEPSC amplitude.28 Based on these findings, Kif5c deficiency might cause disease by contributing to abnormal synaptic function. However, unlike the human brain, the mouse brain has no gyrus. Therefore, in addition to a smaller brain size of the mice with kif5c gene knockout, the nervous system does not show apparent malformations.23

A traditional claim is that proteins are synthesized in the neuronal soma and then conveyed to the terminals in membranous organelles or protein complexes. However, a group of mRNAs is delivered to neurite terminals to carry out local protein synthesis.48 KIF5C, which is involved in transporting cargos (including vesicles, membranous organelles, protein complexes, mRNAs, and chromosomes) from the minus end of microtubules to the plus end in neurites, is implicated in regulating local mRNA transport.49 In the present study, we harvested cortical neurons transfected with control or shkif5c plasmid. Then we found that Kif5c knockdown affects the mRNA levels of pivotal genes—Itgb1, Nrp1, Nrp2, Gli3, Neurog2, and Pafah1b1—related to neurite development and neuronal migration (Fig. 5d).

The cell surface receptors Itgb1, Nrp1, and Nrp2 have critical roles in regulating cortical neuronal migration during early brain development, and Nrp1 is also involved in neocortical dendrite development.50,51,52,53,54,55 Conditional knockout of Gli3 causes neuronal overmigration, resulting in an enlarged brain and abnormally folded structures.56 Neurog2 deficiency leads to cortical migration defects and changes in dendrite morphology.57,58 Pafah1b1 knockout studies revealed its essential function in the regulation of neuronal migration, progenitor cell proliferation, and synaptic E/I balance in the dentate gyrus.59,60,61,62,63 However, the mechanism by which kif5c deficiency alters the mRNA levels of these genes and whether it affects their local protein synthesis by modifying neurite transportation require further study.

Methods

Plasmids

The shKif5c (#1, #2, #3, and Dsred) was generated by inserting the annealed oligonucleotide primers into the FUGW-H1 construct (Addgene, 14883) using XbaI–BamHI cloning sites. Primers’ sequences are listed in Supplement Table III.

RNA extraction and reverse transcription

After anesthetized with pentobarbital sodium and sacrificed, we isolated the cortical area of the brain in pre-cooled PBS according to the map of The Mouse Brain in Stereotaxic Coordinates.64 We isolated total RNA from the cortical brain under manufacturer’s instructions of TRIzol™ Reagent (Invitrogen, 15596018) and then reversed the RNA by the Reverse Transcriptase M-MLV kit (TaKaRa, D2639B).

Quantitative real-time PCR

The mRNA expression level of indicated genes was measured through quantitative real-time PCR (Q-PCR) using SYBR green (Toyobo, QPK-201). The internal control is mouse Gapdh or Rpl191 or human GAPDH gene. The primer sequences are listed in Supplement Table III.

The culture and transfection of primary cortical neurons

Cortical tissues of E13.5 mouse embryos were minced and digested with 20 U/ml papain (Worthington, LS003126) to obtain dissociated neurons.65 Two neuronal transfection methods were used in our study. For assessing KD efficiency and transcriptomic experiments, we packaged the shRNAs into lentivirus for higher transfection efficiency. For the dendrite branching experiment, neurons and plasmids were mixed and resuspended in the Nucleofector Solution buffer and then electrotransfected using the “mouse, neuron, 0–005” process on the AMAXA nucleofector II device (AAD-1001S, Germany), in which the neuron viability was over 80%. In Lab-Tek II chamber slides (ThermoFisher Scientific, 154941), Cells were fed with Neurobasal medium (Gibco, 21103-049) supplemented with 2 mM Glutamax-I (Gibco, 35050-061) and 0.2% B27 (Gibco, 17504-044). We mixed the transfected neurons with untreated neurons in a ratio of 1:8 for sparse labeling.

In utero electroporation

In utero electroporation was performed using the previously reported methods.66 Firstly, E13.5 pregnant mice (C57BL/6J, RRID: IMSR_JAX: 000664) were intraperitoneally injected with 0.7% pentobarbital sodium (10 μl/g per mouse) to achieve deep anesthesia. Plasmids (shDsred or shKif5c, 2 µg/µl) with Fast Green (0.05%, Sigma) were injected into the ventricle using a glass micropipette by an experimenter who was blinded to the groups. In each embryo, only one side of the brain ventricle was injected with constructs. Plasmids were then delivered into the ventricular zone (VZ) cells by applying 5 electric pulses at 30 V for 50 ms at 1 s intervals using the Electro Square Porator (ECM 830, BTX, San Diego, CA). After closing the abdomen, mice were left to recover from anesthesia on a warm blanket. Pup’s brains were harvested at E18.5 or P30, then were fixed with 4% PFA and dehydrated in sucrose solutions.

Generation of forebrain organoids derived from human iPSCs

Forebrain organoids were produced following the previously published methods.67 For the first 4 days, human iPSCs were cultured with neural induction medium (NIM), which consists of DMEM/F12, 1xGlutaMAX, 1xN2, 1xNEAA, 2 μM dorsomorphin (Sigma-Aldrich), and 2 μM SB-431542 (Tocris). On days 5–6, the culture medium was half changed with differentiation medium (DM), which includes DMEM/F12, 1xGlutaMAX, 1xN2, 1xNEAA, 1 μM CHIR (Cellagentech), and 2 μM SB-431542. Then the organoids were embedded in Matrigel (BD Biosciences) and cultivated in cortical DM for another 7 days. On day 14, embedded organoids were gently separated from the Matrigel and cultured in DM supplemented with 1× B27 (Invitrogen), 1× 2-mercaptoethanol, and 2.5 μg/ml insulin (Sigma). The forebrain organoids were harvested on Days 15, 30, and 60 for subsequent experiments.

Immunohistochemical staining and image analysis

All the sample sections on slices were produced using the Leica freezing microtome (CM1950). The neurons, brain slices, and forebrain organoid slices were all fixed with 4% PFA and blocked at room temperature. The composition of the blocking buffer is slightly different. For cultured neurons, the blocking buffer is 3% BSA and 0.1% Triton X-100 in PBS. For brain and organoid slices, the buffer is 5% BSA and 0.3% Triton X-100 in PBS. Then the slices were incubated with indicated antibodies in a blocking buffer of 3% BSA and 0.1% Triton X-100. The following antibodies were used: anti-GFP (rabbit, Invitrogen, A11122): 1:1000, anti-CTIP2 (rat, Abcam, #ab18465): 1:300, anti-SOX2 (goat, R&D, AF2018): 1:500, and anti-CUX1 (mouse, Santa Cruz, sc_514008): 1:500. For immunohistochemical staining, organoid slices were stained with 2 cortical layer markers (CTIP2: Layer V and CUX1: Layer II) to examine developmental milestones.

All images were analyzed using ImageJ software (RRID: SCR_003070). For the dendrite branching experiments, the center of the cell soma was taken as the starting point when calculating the length of a primary dendrite. Slices labeling was blinded to the person who recorded the images to prevent selection bias.

Transcriptome sequencing and analysis

Sequencing libraries were generated using NEBNext® UltraTM RNA Library Prep Kit for Illumina® following the manufacturer’s recommendations. The clean data were obtained by removing reads containing adapter, ploy-N, and low-quality reads from raw data. Index of the reference genome was built and paired-end clean reads were aligned to the reference genome using Hisat2 v2.0.5. FeatureCounts v1.5.0-p3 was used to count the reads numbers mapped to each gene. And then Reads/Fragments Per Kilobase Million Reads (FPKM) of each gene was calculated based on the length of the gene and reads count mapped to this gene. Differential expression analysis of two groups was performed using the DESeq2 R package (1.16.1). Gene Ontology (GO) enrichment analysis of differentially expressed genes was implemented by the clusterProfiler R package. Significant differentially expressed genes were determined as those with adjusted P-values < 1e−4 and log2(fold-change) (FC) > 0.4.

Statistics

Comparisons between two groups were performed using a Two-tailed Student’s t-test, and multiple t-tests were used for the curves (dendrite branching and neuronal migration). GraphPad Prism 8 (www.graphpad.com) software was used for statistical analysis. Data are shown as the means ± SEM, and P-values are reported in the figure legend (statistical significance was determined at P < 0.05). The in vitro experiments were performed at least 3 times independently and at least 3 mice from 2 different litters were used. For migration experiments, we used one E13.5 mouse each time to minimize the bias caused by individual differences and repeated the operation three times independently. At E18.5, we collected all embryonic brains from three independent experiments, and the brains were selected under a fluorescence dissecting microscope (Nikon 80i, MBA75020, Japan). Only those with EGFP expression were retained for further immunofluorescence staining (Supplement Fig. I).

Study approval

All animal experiments were performed according to the Biomedical Research Ethics Committee guidelines of the Shanghai Institutes for Biological Science.