Modeling Neurodevelopmental Disorders and Epilepsy Caused by Loss of Function of kif2a in Zebrafish

Visual Abstract


Introduction
In humans, proper cortical development is highly dependent on the following three primary time-dependent phases: proliferation of the pool of progenitor cells, along with neuronal migration in the cortical plate, and, ultimately, their differentiation (Diaz and Gleeson, 2009;Sun et al., 2017;Buchsbaum and Cappello, 2019). Alterations in any of these series of orchestrated events may result in a disrupted neuronal circuitry and predispose an individual to the development of neurodevelopmental disorders including malformations of cortical development (MCDs; Leventer et al., 2008;Buchsbaum and Cappello, 2019). MCDs are increasingly recognized as a cause of developmental delay, intellectual disability and drug-resistant epilepsy (DRE), and can be caused by various underlying genetic or exogenous factors (Pang et al., 2008;Buchsbaum and Cappello, 2019). Recent advances in genetics have led to the discovery of a number of new genes associated with MCDs. Among these are kinesin superfamily genes (KIFs), encoding ATP-driven microtubule-associated proteins that use the energy of ATP hydrolysis to carry out mechanical work along microtubule tracks (Poirier et al., 2013;Ali and Yang, 2020). It has been reported that de novo mutations in kinesin family member 2a (KIF2A) represent a significant cause of lissencephaly, microcephaly, and DRE (Bahi-Buisson and Cavallin, 2016;Homma et al., 2003;Poirier et al., 2013;Guerrini and Dobyns, 2014;Tian et al., 2016;Cavallin et al., 2017;Broix et al., 2018). KIF2A belongs to the human kinesin-13 family (M-kinesins) consisting of KIF2A, KIF2B, KIF2C/MCAK, and KIF24 (Miki et al., 2005), and is highly abundant in developing neurons being involved in neuronal migration, axonal elongation, and pruning (Homma et al., 2003(Homma et al., , 2018Maor-Nof et al., 2013;Broix et al., 2018). Several studies demonstrated that multiple KIF2A mutations found in patients are located in or near the ATP binding site of the kinesin motor domain (residues p.S317, p.T320, and p.H321; Poirier et al., 2013;Guerrini and Dobyns, 2014;Cavallin et al., 2017), thus affecting ATP hydrolysis or microtubule binding, ultimately resulting in a nonfunctional kinesin (Poirier et al., 2013).
Since epilepsy is the most common clinical consequence of MCDs (Leventer et al., 1999(Leventer et al., , 2008Guerrini and Dobyns, 2014;Tian et al., 2016;Cavallin et al., 2017), antiepileptic drugs are used to control seizures; however, many patients remain unresponsive to the medication therapy (Represa, 2019). Hence, there is an unmet medical need to advance our understanding of the pathogenesis of MCDs and related epilepsies. In recent years, mouse KIF2A models have been developed, which has led to important insights concerning cortical defects and epilepsy (Homma et al., 2003(Homma et al., , 2018Gilet et al., 2020). Kif2a-null mice showed multiple neurodevelopmental abnormalities like early postnatal death, cortical lamination defects as a result of aberrant neuronal migration, and lateral cortical ventricle enlargement; however, no manifestation of epileptic seizures could be reported (Homma et al., 2003). Further characterization of the disease phenotype was allowed by a tamoxifen-inducible Kif2a conditional knock-out (KO) mouse model, demonstrating the role of KIF2A in the precise postnatal hippocampal wiring (Homma et al., 2018). Although neuronal proliferation or migration was not disrupted, mice displayed severe epilepsy, likely a cause of death at 6 weeks postbirth (Homma et al., 2018). Recently, a conditional knock-in (KI) mouse model was developed, bearing the human KIF2A p. H321D missense point variant identified in MCD patients (Gilet et al., 2020). KIF2A 1/H321D KI mice survived and displayed microcephaly and neuroanatomical anomalies like hippocampal structure abnormalities and cortical layer disorganization, resulting from abnormal neuronal migration and increased cell death (Gilet et al., 2020). Moreover, KIF2A-deficient animals showed behavioral deficits and susceptibility to epilepsy, correlating with the described human phenotype.
Over the years, the zebrafish has emerged as a promising model to study neurodevelopmental diseases and their pathologic processes in early development. The developing larval zebrafish brain presents numerous similar structures and cell types that are found in mammals (Kalueff et al., 2014). Moreover, their small size, high fecundity, ex utero development, and genetic amenability allow for higher-throughput screenings, improving CNS drug discovery (Stewart et al., 2015;Khan et al., 2017).
This study intends to advance our understanding of KIF2A-related MCDs by investigating the pathologic consequences of Kif2a loss of function in vivo in a novel zebrafish model. By means of neurobehavioral and electrophysiological assays, we demonstrated that a loss of Kif2a in zebrafish causes behavioral alterations like hypoactivity and habituation deficits, reduced head size, neuronal cell proliferation defects, apoptosis, and, importantly, increased pentylenetetrazole (PTZ)-induced seizure susceptibility. Despite these abnormalities, kif2a À/À larvae survived until adulthood and were fertile. Altogether, our results show that the kif2a KO model could ultimately be used in the search for novel and more effective medications that alleviate the clinical symptoms associated with MCDs.

Zebrafish husbandry
All zebrafish (Danio rerio) strains used in this study were maintained at 28 6 2°C on a 14:10 h light/dark cycle under standard aquaculture conditions in a UV-sterilized rack recirculating system equipped with a mechanical and biological filtration unit. Embryos were collected via natural spawning and immediately transferred to Danieau's medium (0.3Â Danieau's medium), containing 1.5 mM HEPES buffer pH 7.2, 17.4 mM NaCl, 0.21 mM KCl, 0.18 mM Ca(NO 3 ) 2 , 0.12 mM MgSO 4 , and 0.6 mM methylene blue. For the experiments, embryos and larvae were kept at 28.5°C with a photoperiod of 14 h. In all experiments, zebrafish larvae were treated in a humane way and with regard for alleviation of suffering. All zebrafish experiments were performed in accordance with the guidelines of and approval by the Ethical Committee of the University of Leuven (P023/ 2017 and P027/2019) and by the Belgian Federal Department of Public Health, Food Safety and Environment (LA1210199).

Generation of the kif2a CRISPR knock-out zebrafish line
A kif2a knock-out line was generated via the CRISPR/ Cas9 technique (Hruscha et al., 2013;Hwang et al., 2013). kif2a single-guide RNA (sgRNA) targeting exon 5 in the kif2a gene (59-CAGCCAGAATCAGCACCCCC-39) was designed using the CHOP CHOP web tool (https://chopchop.cbu.uib. no), and was further transcribed using the MEGAshortscript T7 Transcription Kit (Ambion) and purified with the MEGAclear Transcription Clean-Up Kit (Ambion). Cas9 (GeneArt CRISPR Nuclease mRNA) was purchased from Thermo Fisher Scientific. Single cell-stage, fertilized wild-type embryos of the AB line were injected with 100 pg of kif2a sgRNA and 150 pg of Cas9 mRNA (in 1 nl volume). The mutation at the target site was verified via T7 endonuclease assay. The remaining sgRNA/Cas9injected embryos were raised till adulthood, outcrossed with wild-type adults, and screened for indels by Sanger sequencing. F0 founder with germline transmission and a high rate of indels was selected to establish the knock-out line. F1 generation embryos of F0 founder were raised to adulthood, fin clipped, and sequenced. Individuals carrying 4 bp deletion of CCAG were identified and pooled together. Experiments were performed on embryos coming from homozygous and heterozygous F2 or F3 progeny.
To confirm the genotype of the larvae, a fin clip was placed in separate tubes with 50 ml of lysis buffer (100 mM Tris, 10 mM EDTA, 0.7 mM proteinase K, 0.2% Triton X-100 in Milli-Q water) to extract genomic DNA. Lysis was performed at 55°C for 3 h, followed by 10 min at 95°C. Lysed samples were genotyped by performing a PCR to amplify the genomic region spanning 4 bp deletion using KAPA HiFi HotStart Polymerase (Roche) and kif2a-specific primers. Successfully amplified PCR products were purified using QIAquick PCR Purification Kit (Qiagen), and the sequence was verified with Sanger sequencing (LGC Genomics). The genotypes of the individual larvae were analyzed using SeqMan software (DNASTAR Lasergene).

Survival assay
From 1 to 5 d postfertilization (dpf) zebrafish embryos were cultured in Danieau's medium in a 10 cm Petri dish. At 6 dpf, they were transferred to a specialized infant incubator system at 28.5°C on a 14:10 h light/dark cycle. They were fed three times a day with 1 dpf brine shrimp and dry granular food (SDS). Larvae were checked every day to a juvenile stage (30 dpf). Dead or moribund zebrafish larvae were registered, removed from the system, and frozen until further analysis. After a period of 30 d, larvae were genotyped as described previously, and survival curves were generated using a Kaplan-Meier estimate.

Whole-mount RNA in situ hybridization
The 973 bp coding sequence fragment of kif2a was amplified from cDNA of AB wild-type strain and cloned into a Zero Blunt TOPO PCR Cloning Kit (Thermo Fisher Scientific). Cloned DNA was linearized by XhoI and HindIII, then synthesized by SP6 RNA polymerase and T7 RNA polymerase using DIG RNA labeling kit (all Roche) for sense-and antisense DIG-labeled RNA probes, respectively.
Embryos were fixed in 4% PFA, then washed with 1Â PBS with Tween 20 (1Â PBST), sequentially washed with 100-25% methanol and stored in 100% methanol at À20°C until needed. On the first day of whole-mount RNA in situ hybridization, embryos were washed with 50-25% methanol, followed by 1Â PBST. After treatment with proteinase K (Sigma-Aldrich) according to the developing stages for permeabilization, embryos were fixed again with 4% PFA and washed by 1Â PBST. Embryos were hybridized in Hyb 1 solution with the kif2a RNA probes at 70°C overnight. On the second day, after serial washing with 2Â SSCT (sodium citrate buffer with 0.1% Tween 20) with 50% formamide, 2Â SSCT, and 0.2Â SSCT at 70°C, embryos were blocked with 5% horse serum (Sigma-Aldrich) and incubated with anti-digoxigenin-AP Fab fragments (Roche) overnight at 4°C. On the third day, embryos were developed with BCIP/NBT (nitroblue tetrazolium) substrate (Roche). Staining was developed and stopped before the background signals started to appear in the embryos hybridized with the sense RNA probe.

Western blotting
Zebrafish larvae of 1-7 dpf were pooled (15 embryos for 1 dpf and 10 larvae for 2-7 dpf; all viable at the time of collection) and lysed in RIPA buffer (Sigma-Aldrich). Following the BCA assay, 30 mg of total protein was loaded onto 4-12% Bis-Tris mini gels (Thermo Fisher Scientific), separated by electrophoresis according to their molecular weight and electroblotted on nitrocellulose membranes (Thermo Fisher Scientific). Membranes were blocked in Intercept blocking buffer (LI-COR) for an hour at room temperature (RT), then incubated overnight with anti-KIF2A polyclonal rabbit antibody (catalog # ab71160, Abcam) or anti-GAPDH polyclonal antibody (catalog #SAB2701826, Sigma-Aldrich) at 4°C and subsequently with a DyLight 800 4Â polyethylene glycol-conjugated secondary antibody (catalog #SA5-35571, Thermo Fisher Scientific) for 1 h at RT. Afterward, membranes were washed three times with 1Â TBST for 15 min. The proteins were visualized by a biomolecular imaging system (Typhoon NIR, GE Healthcare) and bands quantified with Image Studio Lite software.

Behavioral studies
Locomotor studies were performed as published previously (Scheldeman et al., 2017;Siekierska et al., 2019). Five to eight days postfertilization zebrafish larvae were individually arrayed in a 24-well plate containing 400 ml of Danieau's medium per well, followed by a 20 min habituation period in an automated tracking device (Zebrabox, Viewpoint). The average total locomotor activity was quantified during a 5 min light and 10 min dark period using ZebraLab software (Viewpoint) and expressed in actinteg units, defined as the sum of all image pixel changes detected during the time of the tracking experiment. Every day, before the experiment, larvae were replenished with 400 ml of fresh Danieau's medium. The results consist of data from quadruple experiments.
PTZ experiments were executed as published previously (Afrikanova et al., 2013). Larvae were individually arrayed in a 24-well plate containing 400 ml of Danieau's medium per well. Subsequently, an equal volume of 12 mM PTZ (Sigma-Aldrich) was added before recording to obtain a 6 mM final PTZ concentration. The average total locomotor activity was quantified during a 30 min period using ZebraLab software (Viewpoint) and expressed in actinteg units. The results consist of data from quadruple experiments.
For habituation assay, a previously published protocol was adapted (Wolman et al., 2011). Six days postfertilization larvae were individually arrayed in a 96-well plate containing 100 ml of Danieau's medium per well and equilibrated for 3 h in the uniformly illuminated automated tracking device (Zebrabox, Viewpoint). Subsequently, the larvae were subjected to a spaced training consisting of 120 dark flashes (DFs) with 15 s interstimulus intervals (ISIs) alternated with 10 min light periods. After a 30 min light pause, the larvae were exposed to the test assay consisting of 10 DFs with 60 s ISIs. Total movement was expressed in actinteg units. The analysis of the habituation assay was performed as previously described by Nelson et al. (2020). Spaced training results were presented as a percentage of habituation using a formula 1 -(average movement from last 30 DFs/average movement from first 30 DFs) * 100 per larva. Test results were presented as a habituation ratio of the average movement between 10 DFs from the test and the last 10 DFs from the spaced training. The results consist of data from three experiments.

Electrophysiological studies
Noninvasive local field recordings were performed as published previously (Scheldeman et al., 2017;Siekierska et al., 2019). A 5 dpf larva was immobilized in 2% lowmelting point agarose (Thermo Fisher Scientific), and the electric signal was recorded using a glass electrode that was placed on the skin of the head of a larva above the optic tectum, connected to a high-impedance amplifier filled with artificial CSF (124 mM NaCl, 2 mM KCl, 2 mM MgSO 4, 2 mM CaCl 2 , 1.25 mM KH 2 PO 4 , 26 mM NaHCO 3 , and 10 mM glucose). A grounding electrode and a reference electrode were placed in the recording space. The differential signal between the recording electrode and the reference electrode was amplified 10,000 times by a DAGAN 2400 Amplifier, band pass filtered at 0.3-300 Hz, and digitized at 2 kHz via a PCI-6251 interface (National Instruments) with WinEDR (John Dempster, University of Strathclyde, Glasgow, UK). Single recordings were performed for 10 min. Spontaneous electrical discharges were classified as positive events when their amplitude was at least three times the amplitude of the baseline and lasted for at least 100 ms. Analysis of the recordings was performed in a blinded manner using Clampfit 10.2 software (Molecular Devices).
PTZ experiments were executed as described previously (Afrikanova et al., 2013), larvae were arrayed in a 96-well plate containing 100 ml of Danieau's medium. Subsequently, an equal volume of 12 mM PTZ (Sigma-Aldrich) was added for 15 min before recording to obtain a 6 mM final PTZ concentration. Electrophysiological recordings were performed and analyzed as described above. Data were expressed as the mean 6 SD.

Head and body surface area measurements
For head area measurements, larvae were positioned in 3% methylcellulose. Lateral images of 3 and 5 dpf larvae were acquired with a microscope (model MZ 10F, Leica; with a DFC310 FX digital camera, Leica) and Leica Application Suite version 3.6 software. Measurements were performed manually in ImageJ software in a blind manner. Zebrafish body surface area was measured by encircling the total surface of a larva, starting from the anterior tip of the snout to the base of the posterior caudal fin. Head surface areas were measured by tracing the boundary of the surface of interest using reference spots on the head such as the otic vesicle or dorsal indentation just above the eye, at the level of the pineal gland, as previously published (Scheldeman et al., 2017;Siekierska et al., 2019). The absolute values of measured head surface area were normalized to the body surface area of the larvae.

Hematoxylin and eosin staining
The 5 dpf larvae were fixed in 4% PFA at 4°C overnight and kept in 70% ethanol. At least four larvae per genotype group were embedded in 1% agarose in 1Â TAE buffer. A mold, specifically designed to align zebrafish larvae, was used to produce agarose blocks with identical distributed wells of the same depth. Agarose blocks were gradually dehydrated in an enclosed automated tissue processor (Shandon Excelsior ES, Thermo Fisher Scientific) and subsequently embedded in paraffin. The heads of paraffin-embedded larvae were sectioned on an HM 325 manual rotary microtome (Thermo Fisher Scientific) at a thickness of 5 mm. The specimens were stained with hematoxylin and eosin (H&E) stain using Varistain Gemini ES Automated Slide Stainer (Thermo Fisher Scientific) according to laboratory protocols. The resulting sections were imaged at 20Â and 40Â magnification in SPOT 5.1 software (SPOT Imaging) by a SPOT-RT3 camera mounted on a Leica microscope. The brightness of the images was adjusted for the white background. The number of sections was counted per larva per genotype and expressed as the mean 6 SD. Additional counting of brain nuclei was performed to support size difference. Four equivalent sections were selected for each brain region per genotype, and hematoxylin-positive stained nuclei were counted using QuPath (0.2.3) software. The results were expressed as the average of hematoxylinpositive stained nuclei within each selected brain area.

Whole-mount BrdU labeling
Whole-mount BrdU staining was performed on 5 dpf larvae as described previously (Tiraboschi et al., 2020). Larvae were incubated in a 10 mM BrdU (Sigma-Aldrich): 1Â E3 with 15% DMSO solution for 20 min on ice. Embryos were placed at 28°C for 5 min and subsequently fixed for 2 h in 4% PFA at RT. Fixed larvae were washed three times for 5 min in 0.1% PBS-Tween at RT, incubated in 150 mM Tris-HCl, pH 9, for 15 min at RT, then at 70°C for 15 min. Larvae were washed three times for 5 min in 0.1% PBS-Tween and permeabilized in acetone for 30 min at À20°C. Afterward, larvae were incubated in 2N HCl for 1 h at RT to expose the BrdU epitope. Next, HCl was neutralized with 0.1 M borate buffer, pH 8.5, for 20 min at RT, after which larvae were washed 3Â for 5 min in 0.1% PBS-Tween and afterward, again for 15 min. Larvae were bathed overnight in blocking buffer at 4°C in 0.1% PBS-Tween/10% goat serum/0.8% Triton X-100/1% BSA on a rotary shaker. The day after, blocking buffer was exchanged for BrdU antibody (clone 3D4; dilution, 1:200; BD Biosciences) in 0.1% PBS-Tween/1% goat serum/0.8% Triton X-100/1% BSA for 72 h at 4°C. Further, larvae were washed 3Â for 5 min in 0.1% PBS-Tween and incubated for 5 h in 0.1% PBS-Tween/0.8% Triton X-100. Afterward, secondary antibody (dilution, 1:2000; Alexa Fluor 594 goat-antimouse, Abcam) and NucRed (dilution, 1:5000; Live 647 ReadyProbes Reagent, Thermo Fisher Scientific) were applied in 0.1% PBS-Tween/1% goat serum/ 0.8% Triton X-100/1% BSA for 72 h at 4°C. After 72 h, larvae were washed 3Â for 5 min in 0.1% PBS-Tween and incubated overnight in 0.1% PBS-Tween/0.8% Triton X-100. The day after, larvae were mounted on glass slides and dorsal z-stacks of the optic tectum were acquired using an LSM 780-SP Mai Tai HP DS confocal microscope (Zeiss) equipped with an LD LCI Plan Apo 25Â/0.8 objective. To quantify proliferative cells in the optic tectum using ImageJ, a nonspecific signal was removed by creating a mask on the nuclear dye channel. Then, a threshold was set in the BrdU channel to identify individual proliferative cells, and particle analysis was performed on a z-projection of this signal.

Caspase-3 immunohistochemistry
Immunohistochemical detection of cell death was conducted on 5-mm-thick deparaffinized and rehydrated sections corresponding to the sections stained with H&E, as published previously (Siekierska et al., 2019). Specimens were subjected to heat-induced antigen retrieval by incubation in 10 mM sodium citrate, pH 6.0, for 10 min at 98°C in a water bath, followed by a 30 min cool down. Subsequently, specimens were treated twice for 8 min with 3% hydrogen peroxide. The sections were blocked for 30 min in 5% normal goat serum in 1Â TBST at RT, and further incubated with primary antibody against active caspase-3 (1:500 dilution; catalog #ab13847, Abcam) for 1 h at RT. After rinsing 3Â for 15 min with 1Â TBST, an HRP-conjugated secondary antibody (1:200 dilution; catalog #111-035-003, Jackson ImmunoResearch) was applied for 1 h at RT. The slides were treated with DAB 1 /chromogen (DAKO) for 1 min at RT and rinsed with deionized water. As a nuclear counterstain, hematoxylin was added for 3 min and the specimens were subsequently rinsed in deionized water. After clearing in ethanol and histoclear, the slides were mounted. For each staining, a negative control was included by processing sections in the absence of the primary antibody. The images were taken at 20Â magnification in a SPOT 5.1 software (SPOT Imaging) by a SPOT-RT3 camera mounted on a Leica microscope. The brightness of the images was adjusted for the white background. The number of DAB positively stained nuclei in the whole brain was counted using QuPath (0.2.3) software per larva per genotype. The results were expressed as the percentage of apoptotic cells 6 SD (DAB positively stained nuclei over the total cell number) per larva per genotype.

Experimental design and statistical analysis
When comparing parametric data from two groups, an unpaired Student's t test was used. If the data were from more than two groups, a one-way ANOVA with a Dunnett's multiple-comparisons test were used (Table 1). Outliers were identified from datasets using the ROUT method (Q = 1%). The significance of all statistical comparisons was set at p , 0.05. All statistical comparisons were performed using GraphPad Prism (version 8.0.0) software.

Results
Mutation of the kif2a gene by CRISPR/Cas9 results in a stable homozygous mutant zebrafish line Zebrafish genome encodes a single KIF2A ortholog that shows a high degree of conservation between mouse and human proteins (specifically within the kinesin motor domain), with 84.66% identity and 89% similarity for human and zebrafish (Fig. 1a). To investigate KIF2A loss of function in vivo, a zebrafish knock-out model was generated by the CRISPR/Cas9 genome-editing technology (Hruscha et al., 2013;Hwang et al., 2013). We targeted the fifth exon of the zebrafish kif2a located in the globular N-terminal domain. A positive founder, transmitting a premature stop codon, was ultimately selected and used to obtain the F2 generation (Extended Data Fig. 1-1a,b). Genomic analysis confirmed the presence of the mutation in kif2a À/À larvae (Fig. 1b).
Visual inspection of the development of kif2a 1/1 and kif2a À/À larvae from 3 until 5 dpf was performed. No obvious macroscopic differences in appearance such as gross deformation, edema, necrosis, or impairment in touch response were detected, making kif2a À/À larvae morphologically inconspicuous when compared with kif2a 1/1 animals (Fig. 1d). The same trend was observed for adult mutants, which were all morphologically similar to their kif2a 1/1 and kif2a 1/À siblings (data not shown).

4-nt deletion C C G C A G C C A G A A T C A C
To understand the potential function of kif2a during development, its spatial expression patterns in kif2a 1/1 and kif2a À/À larvae were examined. The 8 h postfertilization (hpf) kif2a mRNA was found to be equally expressed in the whole embryo. At 23 hpf, the expression was limited to the head, and gradually became restricted to midbrain and hindbrain at 2 dpf. At 5 dpf, kif2a mRNA was observed only in the hindbrain (Fig. 2c). In kif2a À/À larvae, no mRNA expression was detected at any developmental stage examined (Fig. 2c). This suggests nonsense-mediated mRNA decay (NMD) because of the mutation resulting in a premature termination codon and thus the total absence of mRNA. Additionally, we also confirmed that there was an absence of protein in kif2a À/À larvae at 5 and 7 dpf, confirming the efficiency of the knock-out (Fig. 2d).
kif2a 2/2 larvae display neurobehavioral alterations like hypoactivity, habituation learning deficits, and increased locomotor behavior to PTZ To investigate the impact of Kif2a loss of function on behavioral activity, locomotor experiments were performed on 5-8 dpf kif2a 1/1 and kif2a À/À larvae subjected to a light/dark stimulus. kif2a À/À larvae showed significantly decreased swimming activity during both the light and dark phases from 6-8 dpf ( Moreover, since we observed hypoactivity in kif2a À/À larvae, we investigated whether they presented increased susceptibility to seizures by the administration of a chemical convulsant. kif2a À/À animals were incubated with a subthreshold dose of PTZ, a GABA A receptor antagonist that is widely used to induce acute seizures in zebrafish (Afrikanova et al., 2013). We found a significant increase in epileptiform activity in kif2a À/À larvae (33,431 6 12,315, n = 20, p = 0.0004, Student's t test; Fig. 3b) after PTZ incubation in comparison to their kif2a 1/1 siblings (19,807 6 12,194, n = 20), confirming increased seizure susceptibility.
Additionally, to evaluate whether kif2a À/À larvae deriving from heterozygous parents had a similar phenotype, we performed the same experiments on heterozygous continued normal macroscopic morphology of 3 and 5 dpf kif2a 1/1 and kif2a À/À larvae. Scale bar, 1 mm. The design of the CRISPR/Cas9 kif2a zebrafish knock-out line is illustrated in Extended Data Figure 1-1. qPCR analysis of kif2c expression levels showed no upregulation of kif2c during development in kif2a À/À larvae (Extended Data Fig. 1-1).  (Desikan and Barkovich, 2016;Represa, 2019), we also examined whether kif2a À/À larvae had difficulties in acquiring, storing, and recalling learned information by performing a habituation learning assay (Wolman et al., 2011). During the experiment, larvae were subjected to a spaced training consisting of multiple dark flashes (DFs) alternated with 10 min light periods and, after a pause, were further exposed to a test assay consisting of a block of DFs (Fig. 3c). kif2a 1/1 (47.82 6 31.70, n = 87) and kif2a À/À (51.41 6 21.77, n = 88, p = 0.3833, Student's t test; Fig. 3d) larvae showed an adaptation profile similar to DFs during the habituation, as quantified by calculation of the habituation percentage. Strikingly, kif2a À/À larvae (1.723 6 0.7958, n = 30, p = 0.0053, Student's t test; Fig. 3e) demonstrated memory deficits as they were not able to respond to DFs in the same way as their kif2a 1/1 siblings (1.212 6 0.4738, n = 27) during the training, after a period of inactivation.  Figure 2. Expression and localization pattern of kif2a in the developing zebrafish. a, qPCR analysis of kif2a levels in kif2a 1/1 larvae normalized to actin and represented as the fold change expression to 0 hpf. Values are reported as the mean 6 SEM of three separate experiments. Significant values are noted as ****p 0.0001 and *p 0.05. b, Relative quantification of Kif2a protein expression in kif2a 1/1 larvae of 1-7 dpf normalized to Gapdh and represented as the fold change expression to 1 dpf. Values are reported as the mean 6 SEM of three separate experiments. Significant values are noted as ***p 0.001, **p 0.01, and *p 0.05. Below the graph is a representative Western blot image of Kif2a protein expression levels in kif2a 1/1 larvae of 1-7 dpf. c, Spatiotemporal expression patterns of kif2a by whole-mount RNA in situ hybridization at 8 hpf, 23 hpf, 2 dpf, and 5 dpf. B, Brain; H, hindbrain; M, midbrain; MBH, midbrain-hindbrain boundary; Som, somite; R, retina. Scale bar, 200 mm. d, Representative Western blot image of Kif2a protein expression levels comparing kif2a 1/1 with kif2a À/À larvae at 5 and 7 dpf.
These results demonstrate that kif2a deficiency results in disability in habituation, suggesting memory decay and learning disabilities.
Correspondingly, to characterize the phenotype of kif2a À/À larvae deriving from heterozygous parents, this experiment was performed on heterozygous offspring (Extended Data Fig. 3-1b-d). Both kif2a 1/À (percentage of habituation: 42.39 6 35.83%, n = 66; habituation ratio: 1.161 6 0.6025, n = 30) and kif2a À/À larvae  . kif2a À/À larvae exhibit locomotor abnormalities, increased locomotor behavior to PTZ and habituation learning disabilities. a, Average total movement of kif2a 1/1 and kif2a À/À larvae from 5 to 8 dpf expressed in actinteg units. Values are reported as the mean 6 SD of four separate experiments. Significant values are noted as ****p 0.0001, ***p 0.001, **p 0.01, and *p 0.05. b, Average total movement of 5 dpf kif2a 1/1 and kif2a À/À larvae after 6 mM PTZ immersion for 30 min. Values are reported as the mean 6 SD of four separate experiments. Significant values are noted as ***p 0.001. c, The habituation learning assay performed on 6 dpf kif2a 1/1 and kif2a À/À larvae was composed of different protocols: starting with a 3 h incubation period in the light (indicated in yellow), followed by a habituation training consisting of four periods with 120 DFs with a 15 s ISI (regions with gray and yellow stripes), alternated by 10 min of light, a period of 30 min pause in the light preceding the actual test, and a test consisting of 10 DFs with a 60 s ISI (region with gray and yellow stripes). d, Percentage habituation of kif2a 1/1 and kif2a À/À larvae to DFs during the spaced training. Values are reported as the mean 6 SD of three separate experiments. e, Habituation ratio of average movement of kif2a 1/1 and kif2a À/À larvae to DFs during the test. Values are reported as the mean 6 SD of three separate experiments. Significant values are noted as **p 0.01. No locomotor abnormalities and habituation learning disabilities were observed in kif2a À/À zebrafish larvae from 1/incrosses (Extended Data Fig. 3-1). from homozygous adults is needed to obtain a fullblown phenotype.
Moreover, enlarged ventricles and cerebellar anomalies have been described as common features in Kif2a-null mice (Homma et al., 2003). To investigate whether microcephaly was caused by dysmorphologies in the brain and whether enlarged ventricles and cerebellar abnormalities were among the characteristics of our model, histologic examination of 5 dpf zebrafish brain tissue sections was performed (Fig. 5e,. Although zebrafish kif2a À/À larvae did not present enlarged ventricles relative to the brain tissue, we found significantly fewer brain sections for kif2a À/À larvae (75.5 6 1.915, n = 4; p , 0.0001, Student's t test) from forebrain to hindbrain (Fig. 5e,f, diagram) compared with their kif2a 1/1 siblings (85.0 6 0.0, n = 4).

Discussion
Although mutations in tubulin and kinesin genes have been described in various syndromes associated with brain malformations and DRE (Bahi-Buisson and Cavallin, 2016;Poirier et al., 2013;Guerrini and Dobyns, 2014;Tian et al., 2016;Cavallin et al., 2017), and considerable preclinical/clinical knowledge has been obtained (Homma et al., 2003(Homma et al., , 2018Broix et al., 2018), the majority of patients with MCDs and associated DRE remain unresponsive to current pharmacological therapies (Papayannis et al., 2012;Represa, 2019). Therefore, a greater understanding of the mechanisms underlying MCDs is essential to developing more effective therapeutic agents against MCDs. Although mice are common choices for modeling human diseases, zebrafish offer a number of unique advantages of in vitro scalability in a vertebrate model, which is continued Figure 5. kif2a larvae present microcephaly that is associated with defects in neurologic development. a-d, Comparison of the individual measurements for body area (a, b) and head size (c, d) for kif2a 1/1 and kif2a À/À larvae at 3 and 5 dpf. Data are represented as the mean 6 SD. Significant values are noted as *p 0.05. e, Histologic assessment of 5 dpf kif2a À/À and kif2a 1/1 larval brains. Six brain regions (from forebrain to hindbrain) were selected per genotype as indicated in the diagram. f, Comparison of the number of brain sections from forebrain to hindbrain (e) between kif2a 1/1 and kif2a À/À larvae. Data are represented as the mean 6 SD. Significant values are noted as ****p 0.0001. g, h, Coronal sections stained with H&E imaged at 20Â and 40Â (indicated by orange square) magnification. DT, Dorsal thalamus; Hi, intermediate hypothalamus; lfb, lateral forebrain bundle; PG, preglomerular complex; PrT, pretectum; Teg, midbrain tegmentum; TeO, tectum opticum. Bar graphs compare hematoxylin-positive stained nuclei between kif2a 1/1 and kif2a À/À larvae. Data are represented as the mean 6 SD. Significant values are noted as ****p 0.0001 and **p 0.01. A significant neuronal loss was observed in specific brain regions of kif2a À/À zebrafish larvae (Extended Data Fig. 5-1).
The aim of the present study was to advance our understanding of KIF2A-related MCDs and DRE by investigating the pathologic consequences of Kif2a loss of function in a novel zebrafish model. By targeting an early exon of kif2a sequence with CRISPR/Cas9, we generated an outof-frame deletion leading to a premature stop codon, thus blocking downstream translation of Kif2a essential structural domains required for its activity, resulting in a total lack of Kif2a protein in our model. Correspondingly, missense variants described in patients with KIF2A mutations led to nonfunctional KIF2A protein (Poirier et al., 2013).
Our generated kif2a knock-out model possessed several similarities to KIF2A mouse models that recapitulated many features of the human phenotype. In particular, kif2a À/À zebrafish larvae had reduced head size compared with their kif2a 1/1 siblings. This neuroanatomical anomaly was supported by neuronal loss in the brain because of a decrease in cell proliferation and increased levels of apoptotic cells in the brains of kif2a À/À mutants. Correspondingly, Gilet et al. (2020) found that abnormal neuronal migration, large-scale apoptosis, and increased p53 levels during early stages in the entire mouse cortex were likely responsible for the observed microcephaly in KIF2A 1/H321D mice. Programmed cell death was shown to be increased in many mouse models of microcephaly (Poulton et al., 2011;Little and Dwyer, 2019), and, in contrast, a reduction of apoptosis leads to increased brain size (Kuida et al., 1998). Additionally, previous studies have already reported the correlation between KIF2A inactivation and induced apoptosis in cancer cells (Zhang et al., 2016Li et al., 2019). The silencing of KIF2A by siRNA resulted in a decrease of PI3 kinase/Akt expression and downstream proteins, a key signaling pathway involved in the regulation of cell proliferation and growth, and ultimately in the initiation of apoptosis (Wang et al., 2014;Zhang et al., 2020). In addition, the neuroprotective role of the PI3 kinase/Akt signaling pathway in zebrafish has already been elucidated, emphasizing its function in the attainment of normal brain size during zebrafish embryogenesis (Chen et al., 2017).
Moreover, Kif2a-cKO mice (Homma et al., 2018) and KIF2A 1/H321D mice (Gilet et al., 2020) presented with behavioral deficiencies and susceptibility to epilepsy. Similarly, kif2a À/À larvae showed habituation learning deficits as they were not able to respond to the trained stimulus during the learning period, suggesting inefficient storage of information and memory capacity. These data are in line with memory decay and learning problems, hallmarks of developmental delay and intellectual disability observed in human subjects presenting with KIF2A mutations (Poirier et al., 2013;Tian et al., 2016;Cavallin et al., 2017). We believe that defects in neuronal cell proliferation and increased apoptosis in kif2a À/À mutants may have direct effects on key structures in the brain playing a crucial role in memory. Furthermore, DRE was one of the disease symptoms among patients with mutations in KIF2A (Poirier et al., 2013;Tian et al., 2016;Cavallin et al., 2017). In contrast to the spontaneous epileptiform events observed in Kif2a-cKO mice during postnatal week 5 (Homma et al., 2018), kif2a mutant larvae only exhibited epileptiform events when incubated with a subthreshold dose of a convulsant, PTZ. These findings were supported by the study from Gilet et al. (2020), where KIF2A 1/H321D mice also presented seizure susceptibility after PTZ administration.
Albeit our kif2a zebrafish model recapitulated many phenotypic hallmarks identified in mouse models and human patients with KIF2A mutations, neuroanatomical brain abnormalities described in Kif2a mice models (Homma et al., 2003;Gilet et al., 2020) could not be addressed in our model. This could possibly be because of the fact that the brain ventricular system in mammals is different from that of teleosts (Korzh, 2018), and, although that dorsal pallium likely contains structures homologous to the mammalian hippocampus (Cheng et al., 2014;Saleem and Kannan, 2018), a precise functional mapping of these brain territories is still largely missing. Also, in contrast to the hyperactivity documented in KIF2A mouse models (Homma et al., 2018;Gilet et al., 2020), we observed a reduced locomotor activity in kif2a À/À larvae. Although epilepsy is often linked to increased locomotor activity and thus hyperexcitability, various genetic epilepsy zebrafish models have already reported hypoactivity (Scheldeman et al., 2017;Swaminathan et al., 2018;Prentzell et al., 2021). More particularly, some genetic epilepsy mouse models showed hypoactivity as well (Gu et al., 2019), whereas patients are often hyperactive (Pelc et al., 2008). Although light-dark transitions mainly result in hyperactivity (Basnet et al., 2019) because of a boost in the stress or anxiety levels in zebrafish larvae, locomotor behavior strongly depends on precise CNS development and proper functioning of the brain pathways. Our results show that a failure of any of these factors may be responsible for the observed decrease in locomotor activity (Bilotta, 2000;MacPhail et al., 2009;Vignet et al., 2013;Basnet et al., 2019).
Given the fact that Kif2a KO mice died within a day of birth (Homma et al., 2003) and Kif2a-cKO mice died by continued Figure 6. Neuronal cell proliferation defects and increased apoptosis in kif2a À/À larvae. a, Cell proliferation was measured in the optic tectum of 5 dpf kif2a 1/1 and kif2a À/À larvae as indicated in the diagram. b, Comparison of the number of proliferating cells in the optic tectum of kif2a 1/1 and kif2a À/À larvae showing BrdU-positive cells in dorsal z-stacks. Data are represented as the mean 6 SD. Significant values are noted as *p 0.1. c, Representative z-stacks of optic tecta with BrdU and nuclei staining of kif2a 1/1 and kif2a À/À larvae at 5 dpf. d, Histologic assessment of 5 dpf kif2a À/À and kif2a 1/1 larval brains. Six brain regions (from forebrain to hindbrain) were selected per genotype as indicated in the diagram. e, Total amount of apoptotic cells in the brain of kif2a 1/1 and kif2a À/À larvae. Data are represented as the mean 6 SD. Significant values are noted as **p 0.01. f, g, Coronal sections stained against active caspase-3, imaged at 20Â magnification. h, Percentage of apoptotic cells in the hindbrain region of kif2a 1/1 and kif2a À/À larvae. Data are represented as the mean 6 SD. Significant values are noted as *p 0.05. the end of week 6 (Homma et al., 2018), and because kif2a was selectively expressed in the developing brain during embryogenesis, we hypothesized that our (maternal zygotic) kif2a KO zebrafish larvae would die prematurely. Interestingly, kif2a À/À zebrafish larvae survived until adulthood and were fertile, and thus its offspring, hence excluding that survival was essentially regulated by compensatory effects of maternal kif2a transcripts or Kif2a protein. Possibly, some genetic compensation mechanisms could be involved; however, we did not observe any upregulation originating from a direct kinesin-13 family member. We therefore speculate that additional adaptive compensation by other genes may apply; however, such investigations would require the use of transcriptomics, which could provide an elaborate gene ontology enrichment map of upregulated or downregulated genes. Nonetheless, it is also possible that Kif2a function in zebrafish is less crucial than in mice and not necessarily compensated per se. In addition, we observed that kif2a À/À zebrafish obtained from heterozygous parents had no behavioral phenotype compared with kif2a À/À zebrafish derived from homozygous fish. Given the substantial role of maternal mRNA in zebrafish oocytes, maternal kif2a might continue to perform its function in early larval development (even after zygotic gene expression initiation), leaving kif2a À/À larvae from heterozygous parents unaffected (Pelegri, 2003).
Together, we demonstrated that our kif2a À/À zebrafish model mimicked well the important aspects of the phenotype observed in mice models being reminiscent of the phenotype in patients with KIF2A mutations. Our results highlight the value of our novel kif2a KO model in understanding of the pathologic consequences of KIF2A loss of function. Importantly, since our model does not require tedious genotyping procedures, it could possibly be deployed in medium-to high-throughput MCD/DRE-relevant drug screens and hence used as a high-throughput screening model in a future search for novel and more effective therapies.