Oligodendrocyte dysfunction due to Chd8 mutation gives rise to behavioral deficits in mice

Abstract

Mutations in the gene encoding the chromatin remodeler CHD8 are strongly associated with autism spectrum disorder (ASD). CHD8 haploinsufficiency also results in autistic phenotypes in humans and mice. Although myelination defects have been observed in individuals with ASD, whether oligodendrocyte dysfunction is responsible for autistic phenotypes has remained unknown. Here we show that reduced expression of CHD8 in oligodendrocytes gives rise to abnormal behavioral phenotypes in mice. CHD8 was found to regulate the expression of many myelination-related genes and to be required for oligodendrocyte maturation and myelination. Ablation of Chd8 specifically in oligodendrocytes of mice impaired myelination, slowed action potential propagation and resulted in behavioral deficits including increased social interaction and anxiety-like behavior, with similar effects being apparent in Chd8 heterozygous mutant mice. Our results thus indicate that CHD8 is essential for myelination and that dysfunction of oligodendrocytes as a result of CHD8 haploinsufficiency gives rise to several neuropsychiatric phenotypes.

Graphical Abstract

Introduction

Neural networks operate through propagation of action potentials, which is supported by myelination of axons by oligodendrocytes in the brain (1, 2). Long-range neuronal connectivity mediated by myelinated axons in white matter underlies many aspects of higher brain function (3). Emerging evidence suggests that myelination defects may be associated with a wide range of neuropsychiatric disorders such as schizophrenia, depression and autism spectrum disorder (ASD). For example, a reduced myelin content and myelin thickness along axons in white matter have been detected in individuals with ASD (4–7). Moreover, the speed of neuronal transmission is also reduced in a subset of ASD patients (8,9). Little is known, however, of whether myelination deficits are a consequence of neuronal deficits or whether they might represent a primary cause of ASD.

ASD is a highly prevalent neurodevelopmental disorder characterized by deficits in social interaction and communication as well as by restricted and repetitive behaviors. Recent exome sequencing studies of ASD patients have identified many mutations of genes encoding proteins related to synaptic function, transcriptional regulation or chromatin remodeling (10). Among these genes, encoding chromodomain helicase DNA-binding protein 8 (CHD8) is the most frequently mutated locus in ASD patients (11–14). Such patients with CHD8 mutations manifest characteristics including macrocephaly, distinct facial features, gastrointestinal complaints, cognitive deficits and anxiety (15). An understanding of how CHD8 mutation affects brain development and function is key to deciphering ASD pathogenesis.

CHD8 serves as an ATP-dependent chromatin remodeler and was originally identified as a negative regulator of the Wnt–β-catenin signaling pathway (16). Whereas homozygous deletion of Chd8 in mice results in early embryonic death (17, 18), heterozygous mutant mice manifest autistic-like phenotypes such as macrocephaly, increased anxiety-like behavior and altered social behavior (19–21) and cognitive deficits (22). In addition, knockdown of CHD8 in upper cortical layer neurons resulted in behavioral deficits in adult mice (23). CHD8 regulates the expression of ASD risk genes related to synaptic function and neurodevelopment, and Chd8 mutations in neural precursor cells or the brain result in dysregulation of these genes (19, 20, 23–26). Recent studies with oligodendrocyte-specific Chd8 homozygous knockout models (Olig1-Cre/Chd8^F/F mice) have shown that CHD8 plays a key role in oligodendrocyte development and central nervous system (CNS) myelination (27, 28). However, whether heterozygous mutation of Chd8 in oligodendrocytes affects behavioral phenotypes has remained unknown.

We have now examined the consequences of heterozygous mutation of Chd8 in oligodendrocyte lineage cells. Myelin thinning and a longer gap at nodes of Ranvier were observed in association with delayed nerve conduction in Chd8 heterozygous mutant mice. Importantly, oligodendrocyte-specific ablation of Chd8 not only recapitulated these structural and functional changes to myelin but also gave rise to behavioral characteristics including increased anxiety-like behavior, increased social interaction and a deficit in preference for social novelty. Our results thus open the possibility that dysfunction of oligodendrocytes may contribute to a specific subset of ASD.

Results

Expression of oligodendrocyte-specific genes is downregulated in the brain of Chd8^+/∆L mice and humans with ASD

To identify the cell types in the brain affected by CHD8 haploinsufficiency, we first compared the expression profiles of genes that are selectively transcribed in each cell type of the CNS (29, 30) (Supplementary Material, Table S1) of adult mice between CHD8-haploinsufficient (Chd8^+/∆L) and wild-type (WT, Chd8^+/+) mice. Gene set enrichment analysis (GSEA) revealed that, among the cell type-specific genes examined, the expression of oligodendrocyte-specific genes was downregulated in the brain of adult Chd8^+/∆L mice (Supplementary Material, Fig. S1A–D). The expression of genes specifically transcribed in pyramidal neurons or interneurons was also downregulated. These results suggested that oligodendrocytes might be affected by CHD8 haploinsufficiency in addition to neurons. Furthermore, the brain of ASD patients showed significant downregulation of the expression of oligodendrocyte-specific genes (31) (Supplementary Material, Fig. S1E–H). The expression of genes specifically active in astrocytes or microglia was upregulated in ASD patients but not in Chd8^+/∆L mice. Gene ontology (GO) analysis also revealed that genes whose expression was downregulated in the brain of ASD patients [false discovery rate (FDR) < 0.05] were enriched in those related to the myelin sheath (Supplementary Material, Fig. S1I). Consistent with these findings, defective myelin formation has been observed in a subset of ASD patients including those with CHD8 mutations (4, 6, 15, 32). Collectively, these observations revealing downregulation of the expression of oligodendrocyte-specific genes in both Chd8^+/∆L mice and ASD patients suggested that oligodendrocyte dysfunction might contribute to ASD pathogenesis.

Chd8 heterozygous mutant mice manifest myelination defects and slowed action potential transmission

A major function of oligodendrocytes is formation of the myelin sheath in the CNS (1, 2). Electron microscopy revealed that the myelin sheath of axons in the corpus callosum was thinner (Fig. 1A and B) and the g-ratio (axon diameter/total outer diameter of the myelinated fiber) higher (Fig. 1C) in Chd8^+/∆L mice than in WT mice at 9 weeks of age. We next examined the morphology of the node of Ranvier. Immunohistofluorescence analysis showed that the length of staining for the paranode marker Caspr did not differ between the two genotypes, whereas that for Nav1.6, a marker for the node of Ranvier, was greater in Chd8^+/∆L mice (Fig. 1D and E), indicative of nodal widening in the mutant animals. We confirmed this finding by electron microscopy, which also did not reveal peeling of myelin in Chd8^+/∆L mice (Fig. 1F and G). The structure of both paranode and juxtaparanode regions was maintained in the mutant mice (Supplementary Material, Fig. S2A). The abnormality in node length in Chd8^+/∆L mice might thus be attributable to myelin thinning and changes in the expression of molecules that form the node. Furthermore, whereas the number of mature myelin basic protein (MBP)-positive oligodendrocytes among differentiating cultures of oligodendrocyte precursor cells (OPCs) did not differ between genotypes, the expression of myelin-related genes was downregulated in these cell cultures derived from Chd8^+/∆L mice (Supplementary Material, Fig. S2B–D). Together, these results suggested that CHD8 haploinsufficiency results in a functional deficit in oligodendrocytes.

Figure 1

Chd8 heterozygous mutant mice manifest myelination defects and slower action potential transmission. (A) Electron microscopy of the corpus callosum of adult WT and Chd8^+/∆L mice at 9 weeks of age. Scale bars, 1 μm. (B, C) Myelin thickness and a scatter plot for the g-ratio and axon diameter (n = 5 mice, with a total of 450 axons examined, for each genotype), respectively, determined from images similar to those in (A). (D, E) Immunofluorescence staining of Caspr and Nav1.6 as well as the length of segments of Caspr and Nav1.6 staining, respectively, in the corpus callosum of adult mice at 9 weeks of age (n = 3 mice, with a total of 320 nodes or internode segments examined, per genotype). Scale bars, 5 μm. (F, G) Electron microscopy of nodes of Ranvier (N) as well as node length, respectively, in the corpus callosum of adult Chd8^+/+ and Chd8^+/∆L mice at 9 weeks of age (n = 5 mice, with a total of 42 and 35 nodes examined in WT and Chd8^+/∆L mice, respectively). Scale bars, 0.5 μm. Higher magnification views are shown in the lower panels. (H) Placement of stimulating (Stim.) and recording (Rec.) electrodes in the corpus callosum as well as representative overlaid traces of CAP recordings. Latency between stimulation and ‘onset’ or ‘peak’ was measured. (I) Latency and conduction velocity in the corpus callosum of WT and Chd8^+/∆L mice (n = 10 per genotype). Data are means ± SEM. ^*P < 0.05, ^**P < 0.01 (unpaired Student’s t-test).

Given that the myelin sheath is required for effective propagation of action potentials along axons, and hence for neuronal function (1, 2), we hypothesized that conduction velocity might be reduced as a result of the structural changes in myelin of Chd8 mutant mice. Electrophysiological analysis of compound action potentials (CAPs) in the corpus callosum (33) revealed that the latency of CAP transmission was increased and the conduction velocity decreased in Chd8^+/∆L mice compared with control mice (Fig. 1H and I). These results thus suggested that the myelination defect indeed affects neuronal function in Chd8^+/∆L mice.

Ablation of Chd8 in oligodendrocyte lineage cells impairs myelination and oligodendrocyte maturation

To investigate the role of CHD8 in oligodendrocyte lineage cells, we generated Olig1-Cre/Chd8_L^F/F mice by crossing mice homozygous for a floxed Chd8_L allele (Chd8_L^F/F mice) with mice that express Cre recombinase under the control of the mouse Olig1 promoter (Olig1-Cre mice) (19, 34). Effective deletion of Chd8 in oligodendrocyte lineage cells but not in other neural cell types was previously shown for Olig1-Cre/Chd8_L^F/F mice (28). We also confirmed that Chd8 floxed alleles were efficiently deleted by Cre in OPCs (Supplementary Material, Fig. S3A). Although Olig1-Cre/Chd8_L^F/F mice appeared normal at birth, most of them died before 3 weeks of age manifesting tremors and paralysis of the hind limbs (Supplementary Material, Fig. S3B and C). Whereas a normal optic nerve appears white as a result of a thick myelin sheath, that of Olig1-Cre/Chd8_L^F/F mice appeared semitransparent (Supplementary Material, Fig. S3D), indicative of defective myelin formation by CHD8-deficient oligodendrocytes. In addition, the expression level of MBP was greatly reduced in the whole brain, corpus callosum, cerebellum and spinal cord of Olig1-Cre/Chd8_L^F/F mice at postnatal day (P) 14 (Supplementary Material, Fig. S3E and F). Whereas a large number of myelinated axons was observed in the corpus callosum of control mice at P14 by electron microscopy, myelinated axons were rarely detected in the corpus callosum of Olig1-Cre/Chd8_L^F/F mice (Supplementary Material, Fig. S3G). We did not detect any molecular alterations in neurons or astrocytes in the brain of Olig1-Cre/Chd8_L^F/F mice (Supplementary Material, Fig. S3H–L). Consistent with the myelination defect, the number of oligodendrocytes was significantly reduced in the corpus callosum of Olig1-Cre/Chd8_L^F/F mice at P7 and P14 (Supplementary Material, Fig. S4A and B). In contrast, the number of NG2⁺ OPCs and their proliferation rate (as revealed by Ki67 positivity) were maintained in Olig1-Cre/Chd8_L^F/F mice (Supplementary Material, Fig. S4C–E). The frequency of cleaved caspase-3⁺ (apoptotic) cells in the corpus callosum also did not differ between control and Olig1-Cre/Chd8_L^F/F mice (Supplementary Material, Fig. S4F). These results thus indicated that CHD8 is essential for oligodendrocyte maturation and myelination in vivo.

Oligodendrocyte lineage-specific Chd8 heterozygous mutant mice manifest myelination defects and slowed action potential transmission

Similar to Chd8^+/∆L mice, Olig1-Cre/Chd8_L^+/F mice showed a higher g-ratio and a thinner myelin sheath in the corpus callosum (Fig. 2A–C). Immunohistofluorescence analysis revealed that the length of staining for Caspr did not differ between Olig1-Cre/Chd8_L^+/F and control mice, whereas that for Nav1.6 was greater in Olig1-Cre/Chd8_L^+/F mice (Fig. 2D), indicative of nodal widening in the mutant animals. Moreover, the latency of CAP transmission was increased, and the conduction velocity decreased in the corpus callosum of Olig1-Cre/Chd8_L^+/F mice compared with control mice (Fig. 2E). To investigate further whether Olig1-Cre–dependent Chd8 deletion might affect neuronal properties, we performed electrophysiological studies with pyramidal neurons in layers 2/3 of the prelimbic cortex, which are thought to be related to social behavior (35), in control and mutant (Chd8^+/∆L and Olig1-Cre/Chd8_L^+/F) mice. Neither the amplitude nor the frequency of spontaneous excitatory (sEPSC) or spontaneous inhibitory (sIPSC) postsynaptic currents differed between control and either Chd8^+/∆L (Supplementary Material, Fig. S5A and B) or Olig1-Cre/Chd8_L^+/F (Supplementary Material, Fig. S5C and D) mice. These results suggested that the excitatory and inhibitory presynaptic inputs as well as the postsynaptic properties of pyramidal neurons in the prelimbic cortex were nominally intact in Olig1-Cre/Chd8_L^+/F mice, highlighting the functional importance of the oligodendrocyte autonomous defects induced by Chd8 mutation.

Figure 2

Oligodendrocyte lineage-specific Chd8 heterozygous mutant mice manifest myelination defects and slower action potential transmission. (A) Electron microscopy of the corpus callosum of adult Olig1-Cre/Chd8^+/+ and Olig1-Cre/Chd8_L^+/F mice at 9 weeks of age. Scale bars, 1 μm. (B, C) Myelin thickness and a scatter plot for the g-ratio and axon diameter, respectively, determined from images similar to those in (A) (n = 5 mice, with a total of 450 axons examined, for each genotype). (D) Length of segments of Caspr and Nav1.6 staining in the corpus callosum of adult mice at 9 weeks of age (n = 3 mice, with a total of 320 nodes or internode segments examined, per genotype). (E) Latency and conduction velocity for CAPs in the corpus callosum of adult Olig1-Cre/Chd8^+/+ (n = 10) and Olig1-Cre/Chd8_L^+/F (n = 9) mice. Data are means ± SEM. ^*P < 0.05 (unpaired Student’s t-test).

CHD8 directly regulates expression of myelin-related genes by promoter binding

To examine whether CHD8 haploinsufficiency in oligodendrocyte lineage cells affects gene expression, we performed RNA-sequencing (RNA-seq) analysis of mouse brain (Supplementary Material, Table S2). This analysis revealed only a few genes with a log₂(fold change) of >0.3 or < −0.3 associated with a P value of <0.05 in adult Olig1-Cre/Chd8_L^+/F mice (Supplementary Material, Fig. S6A and B). GSEA revealed that the expression of oligodendrocyte-specific genes was downregulated in the brain of adult Olig1-Cre/Chd8_L^+/F mice, similar to Chd8^+/∆L mice (Fig. 3A and B; Supplementary Material, Fig. S6C and D). The fold changes in expression level of each gene in mutant mice compared with corresponding control mice were significantly related for Chd8^+/∆L and Olig1-Cre/Chd8_L^+/F mice based on the definition of oligodendrocyte-specific genes by Zeisel et al. (30) (Supplementary Material, Fig. S6E) or that by Cahoy et al. (29) (Supplementary Material, Fig. S6F). These results suggested that changes in oligodendrocyte-specific gene expression influence oligodendrocyte function in Olig1-Cre/Chd8_L^+/F mice.

Figure 3

CHD8 directly regulates the expression of oligodendrocyte-specific genes by binding to promoters. (A) GSEA for brain cell type-specific gene sets (30) in the brain of adult Olig1-Cre/Chd8_L^+/F mice compared with Olig1-Cre/Chd8^+/+ mice. NES, normalized enrichment score; NOM, nominal. (B) GSEA plot of differentially expressed genes among the list of oligodendrocyte-specific genes in (A) for the adult Olig1-Cre/Chd8_L^+/F mouse brain. (C) CHD8 target genes identified from our previous ChIP-seq data (19) for the adult mouse brain as a proportion of all genes and the indicated cell type-specific genes (30). P values were determined for comparisons between all genes and each set of cell type-specific genes (hypergeometric test). (D) ChIP and real-time PCR analysis of CHD8 binding to the transcription start site (TSS) regions of the indicated myelin-related genes or Cspg4 as a negative control in the adult mouse brain (n = 4 independent experiments). ChIP was performed with normal immunoglobulin G (IgG) as a control for the antibodies to CHD8 (anti-CHD8). Data in (D) are means ± SEM and were analyzed with the unpaired Student’s t-test. Significance levels for P and FDR q values are indicated with *, ** and *** for < 0.05, < 0.01 and <0.001, respectively. NS, not significant.

To gain insight into the molecular mechanisms by which CHD8 controls gene expression in oligodendrocytes, we reanalyzed our previous chromatin immunoprecipitation-sequencing (ChIP-seq) data sets for the adult mouse brain (19). Whereas 39.4% of all genes were associated with CHD8, this value increased to 55.8% for oligodendrocyte-specific genes (Fig. 3C). Given that the percentage of genes binding CHD8 was highest for oligodendrocyte-specific genes among cell type-specific genes, CHD8 appears to bind preferentially to oligodendrocyte-specific genes. Similar results were obtained with ChIP-seq data for the mouse frontal cortex (26) (Supplementary Material, Fig. S6G). We validated the binding of CHD8 to the promoter regions of oligodendrocyte-specific genes by ChIP followed by quantitative polymerase chain reaction (PCR) analysis (Fig. 3D). RNA-seq analysis revealed that the expression of 304 genes was reduced and that of 239 genes increased in the brain of P14 Olig1-Cre/Chd8_L^F/F mice compared with that of Olig1-Cre/Chd8^+/+ mice [log₂(fold change) of >0.3 or < −0.3 and P < 0.05] (Supplementary Material, Fig. S7A). GO analysis of the 304 downregulated genes revealed significant enrichment not only for ‘myelination’ and ‘myelin sheath’ but also for ‘membrane’, ‘lipid metabolic process’ and ‘sterol biosynthesis process’, all of which are related to myelin formation and oligodendrocyte morphology (Supplementary Material, Fig. S7B). Oligodendrocyte-specific genes were found to be significantly downregulated in the brain of Olig1-Cre/Chd8_L^F/F mice at P14 compared with that of Olig1-Cre/Chd8^+/+ mice (Supplementary Material, Fig. S7C). Moreover, overexpression of CHD8 increased luciferase reporter activity driven by the promoters of several myelin-related genes including those for myelin-associated glycoprotein, MBP and proteolipid protein 1 (PLP1) (Supplementary Material, Fig. S7D). Collectively, these results suggested that CHD8 likely regulates the expression of many myelin-related genes directly and thereby promotes myelin formation and oligodendrocyte maturation (Supplementary Material, Fig. S7E).

Oligodendrocyte lineage-specific Chd8 heterozygous mutant mice recapitulate abnormal behavioral phenotypes of Chd8^+/∆L mice

We then examined whether oligodendrocyte defects due to CHD8 haploinsufficiency in oligodendrocyte lineage cells might be primarily responsible for the macrocephaly, gastrointestinal defects and ASD-like behavioral phenotypes apparent in both Chd8^+/∆L mice and ASD patients with CHD8 mutations (15, 19–23). Body and brain weight did not differ between control and Olig1-Cre/Chd8_L^+/F mice (Fig. 4A and B). In addition, we detected no difference in intestine length or intestinal transit between genotypes (Supplementary Material, Fig. S8A and B). These results suggested that oligodendrocyte defects are not responsible for macrocephaly and gastrointestinal defects.

Figure 4

Oligodendrocyte lineage-specific Chd8 heterozygous mutant mice manifest anxiety-like behavior. (A) Body weight of adult male mice of the indicated genotypes at 9–14 weeks of age (n = 20 per genotype). Chd8^+/+ = 27.07 ± 0.43 g, Olig1-Cre/Chd8^+/+ = 27.22 ± 0.42 g, Chd8_L^+/F = 27.95 ± 0.30 g, Olig1-Cre/Chd8_L^+/F = 27.27 ± 0.44 g; F_(3,76) = 0.97, P = 0.4128 (one-way ANOVA). (B) Brain weight of adult male mice at 9 weeks of age (n = 8 per genotype). Chd8^+/+ = 467.30 ± 7.34 mg, Olig1-Cre/Chd8^+/+ = 464.55 ± 5.29 mg, Chd8_L^+/F = 467.91 ± 6.04 mg, Olig1-Cre/Chd8_L^+/F = 469.54 ± 6.49 mg; F_(3,28) = 0.11, P = 0.9550 (one-way ANOVA). (C–E) Total distance traveled (C), time spent in the central area (D) and vertical activity (E) for 10- to 14-week-old male mice in the open-field test (n = 20 per genotype). In (C): Chd8^+/+ = 106.69 ± 9.75 m, Olig1-Cre/Chd8^+/+ = 102.12 ± 10.80 m, Chd8_L^+/F = 102.13 ± 8.98 m, Olig1-Cre/Chd8_L^+/F = 77.73 ± 7.59 m; F_(3,76) = 1.97, P = 0.1251 (one-way ANOVA). In (D): Chd8^+/+ = 717.18 ± 122.00 s, Olig1-Cre/Chd8^+/+ = 702.85 ± 99.46 s, Chd8_L^+/F = 484.31 ± 49.50 s, Olig1-Cre/Chd8_L^+/F = 440.03 ± 93.40 s; F_(3,76) = 2.32, P = 0.0821. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.9996; Chd8^+/+ versus Chd8_L^+/F, P = 0.3120; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.1732; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 3681; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.2122; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.9875 (one-way ANOVA with Tukey’s post hoc analysis). In (E): Chd8^+/+ = 1136.75 ± 158.70, Olig1-Cre/Chd8^+/+ = 1242.80 ± 155.68, Chd8_L^+/F = 1062.60 ± 126.70, Olig1-Cre/Chd8_L^+/F = 663.35 ± 96.93; F_(3,76) = 3.42, P = 0.0214^*. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.9468; Chd8^+/+ versus Chd8_L^+/F, P = 0.9807; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0770; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.7881; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0190^*; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.1745 (one-way ANOVA with Tukey’s post hoc analysis). (F–H) Time spent in the light chamber (F), distance traveled in the light and dark chambers (G) and number of transitions between the light and dark chambers (H) for 10- to 14-week-old male mice in the light–dark transition test (n = 20 per genotype). In (F): Chd8^+/+ = 169.28 ± 10.69 s, Olig1-Cre/Chd8^+/+ = 190.6 ± 10.16 s, Chd8_L^+/F = 183.25 ± 11.56 s, Olig1-Cre/Chd8_L^+/F = 134.73 ± 9.40 s; F_(3,76) = 5.59, P = 0.0016^**. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.4796; Chd8^+/+ versus Chd8_L^+/F, P = 0.7820; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.1001; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.9598; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0018^**; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.0085^** (one-way ANOVA with Tukey’s post hoc analysis). For light chamber in (G): Chd8^+/+ = 7.35 ± 0.51 m, Olig1-Cre/Chd8^+/+ = 7.85 ± 0.42 m, Chd8_L^+/F = 7.34 ± 0.51 m, Olig1-Cre/Chd8_L^+/F = 5.41 ± 0.44 m; F_(3,76) = 5.21, P = 0.0025^**. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.8711; Chd8^+/+ versus Chd8_L^+/F, P = 1.0000; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0250^*; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.8660; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0026^**; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.0258^* (one-way ANOVA with Tukey’s post hoc analysis). For dark chamber in (G): Chd8^+/+ = 13.12 ± 0.52 m, Olig1-Cre/Chd8^+/+ = 12.21 ± 0.59 m, Chd8_L^+/F = 12.66 ± 0.54 m, Olig1-Cre/Chd8_L^+/F = 11.47 ± 0.70 m; F_(3,76) = 1.42, P = 0.2447 (one-way ANOVA). In (H): Chd8^+/+ = 30.50 ± 0.27, Olig1-Cre/Chd8^+/+ = 30.25 ± 0.21, Chd8_L^+/F = 27.60 ± 0.23, Olig1-Cre/Chd8_L^+/F = 21.70 ± 0.21; F_(3,76) = 3.12, P = 0.0309^*. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.9998; Chd8^+/+ versus Chd8_L^+/F, P = 0.8124; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0430^*; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.8500; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0521; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.2807 (one-way ANOVA with Tukey’s post hoc analysis). All data are means ± SEM. ^*P < 0.05, ^**P < 0.01.

To evaluate the behavioral characteristics of Olig1-Cre/Chd8_L^+/F mice, we performed a battery of behavioral tests (Supplementary Material, Table S3). Olig1-Cre/Chd8_L^+/F mice showed no apparent change in general health or motility (Supplementary Material, Fig. S8C–F). Total distance traveled and time spent in the center of the open field in the open-field test were not significantly changed in the mutant mice (Fig. 4C and D). Vertical activity in the open-field test differed significantly between Olig1-Cre/Chd8^+/+ and Olig1-Cre/Chd8_L^+/F mice and tended to differ between Chd8^+/+ and Olig1-Cre/Chd8_L^+/F mice, similar to our findings with Chd8^+/∆L mice (Fig. 4E). The T-maze forced alternation test revealed that Olig1-Cre/Chd8_L^+/F mice did not show any change in memory compared with control mice (Supplementary Material, Fig. S8G and H). The T-maze left–right discrimination test showed that the percentage of correct responses after directional reversal was similar in Olig1-Cre/Chd8_L^+/F and control mice (Supplementary Material, Fig. S8I), indicating that the mutant animals do not manifest the persistence apparent in Chd8^+/∆L mice (19). Furthermore, the acoustic startle response and prepulse inhibition (PPI) did not differ between control and Olig1-Cre/Chd8_L^+/F mice (Supplementary Material, Fig. S8J and K). Together, these results suggested that learning and memory as well as the acoustic startle response are not impaired in Olig1-Cre/Chd8_L^+/F mice.

We next performed the light–dark transition test to evaluate anxiety-like behavior. Anxiety is a symptom of ASD patients, including those with CHD8 mutations (15, 36). Many experimental models of ASD including CHD8-haploinsufficient mice also manifest anxiety-like behavior (19, 20, 37–39). In the light–dark transition test, the time spent in the light chamber was significantly decreased for Olig1-Cre/Chd8_L^+/F mice compared with control mice with the exception of Chd8^+/+ mice (Fig. 4F). Distance traveled in the light chamber was also significantly decreased in Olig1-Cre/Chd8_L^+/F mice (Fig. 4G), indicative of increased anxiety-like behavior and similar to findings with Chd8^+/∆L and other mouse models for the study of ASD. In addition, the number of transitions between the light and dark chambers was significantly or tended to be decreased for Olig1-Cre/Chd8_L^+/F mice compared with Chd8^+/+ mice and Olig1-Cre/Chd8^+/+ mice, respectively (Fig. 4H). These results were not due to group differences in motility (Fig. 4C), and they suggested that CHD8 haploinsufficiency in oligodendrocyte lineage cells gives rise to increased anxiety-like behavior.

We also performed behavioral tests to assess deficits in social interaction, a prominent characteristic of ASD (37, 38, 40). Whereas the number of social contacts did not differ among genotypes (Fig. 5A), duration per contact tended to be increased, and total contact time was significantly increased in Olig1-Cre/Chd8_L^+/F mice (Fig. 5B and C). We observed no difference in motility during the social interaction test among genotypes (Fig. 5D). We also performed the three-chamber social approach test to evaluate sociability and preference for social novelty (40, 41). In the three-chamber test of sociability, Olig1-Cre/Chd8_L^+/F mice and control animals showed a significant preference for a novel mouse (stranger 1) (Fig. 5E), as did Chd8^+/∆L mice (19, 20, 22). In the social novelty preference test, however, control (Chd8^+/+ and Chd8_L^+/F) mice showed a significant preference for a novel mouse (stranger 2) over a familiar mouse (stranger 1), whereas Olig1-Cre/Chd8_L^+/F mice did not (Fig. 5F). Olig1-Cre/Chd8_L^+/F mice thus showed a mild deficit in social novelty preference but not in sociability. Given that Olig1-Cre/Chd8^+/+ mice also did not show a significant preference for a novel mouse (stranger 2) over a familiar mouse (stranger 1), however, we are unable to formally exclude the possibility that Cre expression itself might affect this test. The olfactory habituation test showed that odor recognition did not differ between Olig1-Cre/Chd8_L^+/F and control mice (Supplementary Material, Fig. S9A), suggesting that the deficit in preference for social novelty in Olig1-Cre/Chd8_L^+/F mice was not the result of impaired odor recognition. Other behavioral tests—including the elevated plus-maze test, nest-building test and Porsolt forced swim test—did not reveal significant differences between Olig1-Cre/Chd8_L^+/F and control mice (Supplementary Material, Fig. S9B–E). Collectively, these results suggested that Olig1-Cre/Chd8_L^+/F mice recapitulate a subset of behavioral phenotypes of Chd8^+/∆L mice.

Figure 5

Oligodendrocyte lineage-specific Chd8 heterozygous mutant mice manifest a mild deficit in social behavior. (A–D) Number of contacts (A), mean duration of contacts (B), total duration of contacts (C) and total distance traveled (D) for the social interaction test in a novel environment (n = 10 per genotype). Two male mice of the same genotype at 11–15 weeks of age that had previously been housed in different cages were placed together in a box and allowed to explore freely for 10 min. In (A): Chd8^+/+ = 48.6 ± 4.1, Olig1-Cre/Chd8^+/+ = 52.5 ± 3.1, Chd8_L^+/F = 51.1 ± 3.9, Olig1-Cre/Chd8_L^+/F = 57.4 ± 3.2; F_(3,36) = 1.06, P = 0.3771 (one-way ANOVA). In (B): Chd8^+/+ = 1.51 ± 0.15 s, Olig1-Cre/Chd8^+/+ = 1.55 ± 0.12 s, Chd8_L^+/F = 1.76 ± 1.33 s, Olig1-Cre/Chd8_L^+/F = 2.42 ± 0.41 s; F_(3,36) = 2.90, P = 0.0484^*; Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.9995; Chd8^+/+ versus Chd8_L^+/F, P = 0.8908; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0618; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.9314; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0794; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.2521 (one-way ANOVA with Tukey’s post hoc analysis). In (C): Chd8^+/+ = 73.64 ± 9.82 s, Olig1-Cre/Chd8^+/+ = 79.94 ± 5.59 s, Chd8_L^+/F = 85.23 ± 6.28 s, Olig1-Cre/Chd8_L^+/F = 128.60 ± 14.5 s; F_(3,36) = 6.58, P = 0.0012^**. Chd8^+/+ versus Olig1-Cre/Chd8^+/+, P = 0.9676; Chd8^+/+ versus Chd8_L^+/F, P = 0.8339; Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0017^**; Olig1-Cre/Chd8^+/+ versus Chd8_L^+/F, P = 0.9804; Olig1-Cre/Chd8^+/+ versus Olig1-Cre/Chd8_L^+/F, P = 0.0060^**; Chd8_L^+/F versus Olig1-Cre/Chd8_L^+/F, P = 0.0164^* (one-way ANOVA with Tukey’s post hoc analysis). In (D): Chd8^+/+ = 2966.23 ± 225.66 cm, Olig1-Cre/Chd8^+/+ = 2752.15 ± 154.62 cm, Chd8_L^+/F = 2673.10 ± 126.35 cm, Olig1-Cre/Chd8_L^+/F = 2487.11 ± 140.594 cm; F_(3,36) = 1.42, P = 0.2516 (one-way ANOVA). (E and F) Time spent in each chamber during the first (E) and second (F) tasks for the sociability and social novelty preference test. In the first task, one of the side chambers contains a caged stranger mouse (stranger 1), and the other side chamber contains an empty cage. In the second task, the first stranger mouse (stranger 1) has become familiar to the test mouse, and the empty cage is replaced by a new caged stranger mouse (stranger 2). All the mice used in this test were 13–17 weeks of age (n = 20 per genotype). In (E): Chd8^+/+ empty side = 213.80 ± 13.29 s, center = 73.65 ± 5.07 s, stranger 1 side = 312.55 ± 12.85 s; empty side versus stranger 1 side, P = 0.0011^**, t₍₁₉₎ = 3.85. Olig1-Cre/Chd8^+/+ empty side = 175.70 ± 12.54 s, center = 65.05 ± 5.98 s, stranger 1 side = 359.25 ± 12.49 s; empty side versus stranger 1 side, P < 0.0001^***, t₍₁₉₎ = 7.55. Chd8_L^+/F empty side = 192.75 ± 12.39 s, center = 76.9 ± 6.50 s, stranger 1 side = 330.35 ± 14.09 s; empty side versus stranger 1 side, P < 0.0001^***, t₍₁₉₎ = 5.35. Olig1-Cre/Chd8_L^+/F empty side = 149.35 ± 11.63 s, center = 66.05 ± 7.35 s, stranger 1 side = 384.6 ± 13.94 s; empty side versus stranger 1 side, P < 0.0001^***, t₍₁₉₎ = 9.56 (paired Student’s t test). In (F): Chd8^+/+ stranger 1 side = 232.00 ± 11.25 s, center = 73.75 ± 4.06 s, stranger 2 side = 294.25 ± 12.05 s; stranger 1 side versus stranger 2 side, P = 0.0138^*, t₍₁₉₎ = 2.71. Olig1-Cre/Chd8^+/+ stranger 1 side = 236.30 ± 16.77 s, center = 75.85 ± 6.60 s, stranger 2 side = 287.85 ± 17.00 s; stranger 1 side versus stranger 2 side, P = 0.1361, t₍₁₉₎ = 1.56. Chd8_L^+/F stranger 1 side = 226.05 ± 12.81 s, center = 67.8 ± 4.92 s, stranger 2 side = 306.15 ± 13.23 s; stranger 1 side versus stranger 2 side, P = 0.0055^**, t₍₁₉₎ = 3.13. Olig1-Cre/Chd8_L^+/F stranger 1 side = 256.25 ± 16.51 s, center = 68.35 ± 7.16 s, stranger 2 side = 275.4 ± 13.29 s; stranger 1 side versus stranger 2 side, P = 0.5185, t₍₁₉₎ = 0.66 (paired Student’s t-test). (G) Comparison of phenotypes among whole-body Chd8 mutant mice including Chd8^+/∆L and Chd8^+/∆SL mice (this study) (19), Chd8^+/− mice (20), Chd8^+/del5 mice (22) and Chd8^+/− mice (21), as well as oligodendrocyte lineage (OL)-specific Olig1-Cre/Chd8_L^+/F mice (this study). All quantitative data are means ± SEM. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

Discussion

A key question in ASD research concerns the identity of the cell types that are primarily responsible for the behavioral features of this disorder. We have now shown that oligodendrocyte dysfunction is responsible for several behavioral characteristics in Chd8 heterozygous mutant mice. Oligodendrocyte-specific ablation of Chd8 resulted in myelin thinning, expansion of the gap at nodes of Ranvier and delayed nerve conduction, all of which were associated with behavioral deficits such as increased anxiety-like behavior, increased social interaction and loss of preference for social novelty.

CHD8 haploinsufficiency is a highly penetrant risk factor for ASD in humans (15). We have also previously shown that two independent lines of whole-body Chd8 heterozygous mutant mice (∆SL and ∆L) manifest abnormal behavioral phenotypes similar to those of human ASD (19) and two other groups have described similar findings (20, 21). On the other hand, another group reported distinct behavioral phenotypes of whole-body Chd8 heterozygous mutant mice (22). The behavioral phenotypes of these five Chd8 mutant mouse lines are summarized in Figure 5G. In the reciprocal social interaction test, altered social interaction such as increased contact time was apparent for ‘four’ lines of the mutant mice (19–21), whereas no such change was detected in the fifth (22). The differences in behavioral phenotypes among Chd8 mutant mouse models may be due to differences in genetic background or experimental conditions. For example, both sexes of the fifth line of mice were tested (22), whereas only one sex was tested for each of the other four lines, in the reciprocal social interaction test (19, 20). Although atypical, altered social interaction characterized by increased contact time has been observed in a subset of other mouse models for the study of ASD as well as in ASD patients (38, 42). Whereas individuals with ASD often show impaired social interaction, distinctive subtypes of interaction style such as aloofness, passivity and active-but-odd have been described (43). In the three-chamber test of sociability, all five lines of whole-body Chd8 heterozygous mutant mice showed normal behavior. In the social novelty preference test, however, the first three lines manifested a lack of preference for social novelty (19, 20), with the fourth and fifth lines not being subjected to this test (21, 22). A deficit in the preference for social novelty such as at the level of social recognition or social memory is apparent in many mouse models of ASD. For example, mice with a mutation in the neuroligin 3 gene (Nlgn3), a strong autism candidate gene, manifested a lack of preference for social novelty, but they did not show deficits in sociability (44). Anxiety-like behavior is also a symptom of ASD including that associated with CHD8 mutations (15, 36). The light–dark transition test or elevated plus-maze test revealed that whole-body Chd8 heterozygous mutant mice manifest increased anxiety-like behavior (19, 20). Furthermore, increased persistence (19), a lower acoustic startle response (19), increased PPI (19), enhanced acquired motor learning (20), excessive smelling of social cues (21) and learning and memory deficits (22) were observed in the various lines of these mutant mice. Overall, abnormal behavioral phenotypes have thus been detected in five independent Chd8 mutant mouse lines. In particular, increased anxiety-like behavior, increased social interaction, and a deficit in preference for social novelty were reproducibly observed in whole-body Chd8 heterozygous mutant mouse models generated by at least two independent groups. We now show that CHD8 haploinsufficiency only in oligodendrocyte lineage cells recapitulates these reproducible behavioral features.

CHD8 is a member of the CHD family of enzymes that belong to the SNF2 superfamily of ATP-dependent chromatin remodelers. Chromatin-remodeling complexes mediate dynamic regulation of chromatin architecture to allow access of transcriptional factors to DNA and thereby control gene expression. They have emerged as important regulators of cellular differentiation and development including the differentiation of oligodendrocytes (45, 46). The BRG1 component of the SWI/SNF chromatin-remodeling complex is thus required for the initiation of oligodendrocyte differentiation (47). In addition, the CHD family protein CHD7 interacts with Sox10 and thereby regulates the onset of myelination (48). Our results now show that CHD8 plays a key role in oligodendrocyte maturation and myelination. A recent study also found that CHD8 regulates the transcriptional program for oligodendrocyte lineage differentiation and that oligodendrocyte-specific Chd8 knockout mice manifest myelination defects (28), consistent with our present results. CHD8 might thus increase the accessibility of chromatin at myelin-related gene loci and thereby control oligodendrocyte development and myelination. Although this previous study (28) reported that Chd8 heterozygous knockout mice were phenotypically normal compared with WT mice, no data were presented in this regard. In our present study, the numbers of oligodendrocytes and OPCs did not differ between WT and Chd8 heterozygous mutant mice. Whereas the previous study (28) did not examine the structure of nodes of Ranvier, however, we found by electron microscopy that Chd8 heterozygous mutant mice have a reduced myelin thickness and an abnormal node structure. Our results thus suggest that CHD8 haploinsufficiency impairs formation of the myelin sheath.

Myelination by oligodendrocytes is required for proper connectivity during neurodevelopment (1, 2), and the development and maturation of white matter correlate with increased motor skills and cognitive functions (3). Myelination defects have been detected in a subset of ASD patients by magnetic resonance imaging or postmortem analysis (4, 6), and several mouse models of ASD also manifest downregulation of oligodendrocyte-specific gene expression and defective myelination (49, 50). However, it has remained unclear whether these deficits are a secondary consequence of neuronal deficits or whether they might represent, in at least some cases, a primary cause of ASD. Social isolation in mice has been shown to affect myelination in the prefrontal cortex, leading to alterations in sociability and working memory (51). Moreover, mice deficient in the myelin-related protein PLP1 were recently found to manifest mild myelin disruption as well as increased anxiety-like behavior and exploration of novel social odor (52), suggesting that mild myelin abnormalities might be a primary cause of behavioral changes. Our findings also support the notion that abnormal myelination in ASD patients can primarily contribute to some ASD symptoms.

Delayed action potential propagation as a result of defective myelination is thought to influence oscillation and timing within neural circuits (1, 2). A recent study found that alterations of functional connectivity in the medial prefrontal cortex, hippocampus, and amygdala were reproducibly detected in individuals with ASD from four independent cohort studies (53). It is thus likely that delayed action potential propagation due to defective myelination affects neural networks in the brain of Olig1-Cre/Chd8_L^+/F mice, which might give rise to the behavioral deficits in these animals.

Our results uncover a role for CHD8 in oligodendrocyte development, myelination, and ASD pathogenesis. Given that impaired myelination may contribute to behavioral deficits, promotion of myelination is a potential strategy for treatment of ASD-related behavioral abnormalities. Consistent with this notion, a recent study found that improvement of myelination or nerve conduction rescued the abnormal social behavior in a mouse model of a neurodevelopmental disorder (54). Overall, our data indicate that oligodendrocytes play a key role in ASD pathogenesis, a finding that may inform the development of new approaches to the treatment of specific ASD phenotypes.

Materials and Methods

Mice

Generation of Chd8^+/∆L and Chd8_L^F/F mice was described previously (19). A pair of loxP sites flanking exons 11–13 of Chd8 in Chd8_L^+/F mice was deleted by Cre recombinase to yield Chd8^+/∆L mice. Offspring were backcrossed onto the C57BL/6 J line for at least nine generations. Olig1-Cre transgenic mice were described previously (34). Chd8_L^F/F mice were crossed with Olig1-Cre heterozygous mice to produce Olig1-Cre/Chd8_L^F/F mice. All experiments were performed with male mice, with the exception of those examining mice younger than P28. Mice were genotyped by PCR analysis of genomic DNA with primers for Chd8_L (5′-CCCAAAAGACCAAATCAAACAAAC-3′, 5′-CCATAGGCTGAAGAACCGTAATTG-3′ and 5′-AGGCTTAGAAACCCGTCGAG-3′) and Cre (5′-AGGTTCGTTCACTCATGGA-3′ and 5′-TCGACCAGTTTAGTTACCC-3′). All animals were maintained under specific pathogen-free conditions, and all experiments were approved by the animal ethics committee of Kyushu University.

Antibodies

A rat monoclonal antibody to CHD8_L (anti-CHD8_L mAb) produced as described previously (19) was used for ChIP. Other antibodies included those to NeuN (MAB377, Millipore, Billerica, MA), to GFAP (131–17 719, Thermo Fisher Scientific, Waltham, MA), to CC1 (OP80, Millipore), to MBP (AB980, Millipore, or ab7349, Abcam, Cambridge, U.K.), to NG2 (AB5320, Millipore), to PDGFRα (558 774, BD Biosciences, San Jose, CA), to Nav1.6 (K87A/10, NeuroMab), to Kv1.2 (75–008, NeuroMab), to Caspr (ab34151, Abcam), to Ki67 (550 609, BD Biosciences) and to cleaved caspase-3 (9661, Cell Signaling Technology, Danvers, MA).

Immunofluorescence staining

Immunofluorescence staining of frozen sections was performed as described previously (55). The mouse brain was fixed overnight at 4°C with 4% paraformaldehyde in phosphate-buffered saline. Frozen sections (16 μm) were exposed to 5% bovine serum albumin and 0.1% Triton X-100 before incubation overnight at 4°C with primary antibodies. Immune complexes were detected with Alexa Fluor 488- or Alexa Fluor 546-conjugated goat secondary antibodies (Thermo Fisher Scientific). Sections were counterstained with 4′,6-diamidino-2-phenylindole. Images were acquired with a laser scanning confocal microscope (LSM700, Zeiss, Oberkochen, Germany) and ZEN imaging software (Zeiss). ImageJ software (NIH) was used for quantification of the number of stained cells. For measurement of node of Ranvier and paranode length in the corpus callosum, at least 100 nodes for each mouse were analyzed with ImageJ software.

Electron microscopy

Brain specimens were fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer containing 0.1 M sucrose and 3 mm CaCl₂ and were then embedded in resin (56). Ultrathin sections were viewed with an FEI Tecnai 20 transmission electron microscope. ImageJ software was used to measure the diameter of axons, the total outer diameter of myelinated fibers, the myelin thickness and the length of the node of Ranvier. The genotypes of mice were known before experiments involving electron microscopy. For image analysis, the electron microscopic images were quantified in a blinded manner.

Isolation of OPCs

Primary OPCs were isolated from the cortices of mice at P7 by immunopanning as previously described but with minor modifications (57). In brief, the tissue was incubated for 20 min at 37°C with 0.25% trypsin-EDTA and was then subjected to gentle dissociation after the addition of Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and DNase (10 mg/ml). The dissociated cells were incubated for 45 min at room temperature in dishes coated with antibodies to platelet-derived growth factor receptor α (PDGFRα), nonadherent cells were removed by washing, and OPCs bound to the dishes were collected by exposure to 0.25% trypsin-EDTA. Isolated OPCs were cultured at 37°C in OPC medium consisting of neurobasal medium supplemented with B-27 supplement (20 ml/l), 2 mm L-glutamine and nonessential amino acids (10 ml/l) (all obtained from Invitrogen, Waltham, MA). Mouse basic fibroblast growth factor (40 ng/ml, R&D Systems, Minneapolis, MN) and recombinant human PDGF-AA (20 ng/ml, PeproTech, Rocky Hill, NJ) were added to the medium to promote cell proliferation (58). Triiodothyronine (40 ng/ml, Sigma, St. Louis, MO) was added to the medium to induce differentiation.

Reverse transcription and real-time PCR analysis

Total RNA (1 μg) isolated from mouse brain tissue with the use of Isogen (Nippon Gene) was subjected to reverse transcription with a QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany), and the resulting cDNA was subjected to real-time PCR analysis with the use of SYBR Green PCR Master Mix (Thermo Fisher Scientific) and specific primers in a StepOnePlus Real-Time PCR System (Applied Biosystems, Waltham, MA). Data were normalized by the abundance of Rplp0 mRNA. PCR primers (sense and antisense, respectively) were as follows: Rplp0, 5′-GGACCCGAGAAGACCTCCTT-3′ and 5′-GCACATCACTCAGAATTTCAATGG-3′; Chd8_L, 5′-TCCCTTTTTGGTCATTGCTC-3′ and 5′-TTCAGCCTATGGGCTTCATC-3′; Mbp, 5′-ATCCAAGTACCTGGCCACAG-3′ and 5′-CCTGTCACCGCTAAAGAAGC-3′; Plp1, 5′-CTGGCTGAGGGCTTCTACAC-3′ and 5′-GACTGACAGGTGGTCCAGGT-3′; Cnp, 5′-CATCCTCAGGAGCAAAGGAG-3′ and 5′-GTACGCCTCGGAGAAGTCTG-3′; Mag, 5′-GGACCCCATCCTTACCATCT-3′ and 5′-CGGGTTGGATTTTACCACAC-3′; Mog, 5′-TTAGTTGGGGATGAAGCAGAGC-3′ and 5′-CATTTCGGTAGAGGTGAACCACTC-3′; Pdgfra, 5′-TATCCTCCCAAACGAGAATGAGA-3′ and 5′-GTGGTTGTAGTAGCAAGTGTACC-3′; Tubb3, 5′-ACTTGGAACCTGGAACCATGG-3′ and 5′-GGCCTGAATAGGTGTCCAAAGG-3′; Rbfox3, 5′-GCAGATGAAGCAGCACAGACAG-3′ and 5′-CGAACATTTGCCGCAGGTC-3′; Dcx, 5′-GGTCCTGACTGACATCACAGAAGC-3′ and 5′-GAGAAATCATCTTGAGCATAGCGG-3′; Gfap, 5′-GGGGCAAAAGCACCAAAGAAG-3′ and 5′-GGGACAACTTGTATTGTGAGCC-3′; and Slc1a3, 5′-CCTTGGATTTGCCCTCCGAC-3′ and 5′-GCCATTCCTGTGACGAGACTGG-3′.

Plasmids

Complementary DNA encoding mouse CHD8 was subcloned into pcDNA3 (Invitrogen). Mouse genomic DNA containing the promoter regions of Mag (positions −302 to +82 relative to the transcription start site), Mbp (−611 to +325) or Plp1 (−1049 to +91) was cloned into the pGL4-Basic vector (Promega, Madison, WI).

Cell culture and transfection

Neuro2A cells (ATCC) were cultured under an atmosphere of 5% CO₂ at 37°C in DMEM supplemented with 10% fetal bovine serum. Transfection of the cells with plasmid DNA was performed with the use of the Lipofectamine 3000 reagent (Thermo Fisher Scientific).

Luciferase assay

Neuro2A cells transfected for 48 h with an expression vector for mouse CHD8 together with pRL-TK (Promega) as an internal control and luciferase reporter plasmids containing promoter sequences of Mag, Mbp or Plp1 were lysed and assayed for luciferase activity with a Dual-Luciferase Reporter Assay System (Promega). Firefly luciferase activity was normalized by that of Renilla luciferase.

Gastrointestinal transit test

A gastrointestinal transit test was performed as described previously (19). Adult male mice at 9 weeks of age were given 200 μl of a charcoal marker (10% charcoal, 5% gum arabic) by oral gavage. The mice were killed 30 min later, and the intestine from the region of the pyloric sphincter to the ileocecal junction was removed for measurement (without stretching) of its length and the distance traveled by the marker.

General protocol for behavioral tests

Chd8 mutant or control mice were group-housed (four animals per cage) in a room with a 12 h light and 12 h dark cycle (lights on at 7:00 a.m.) and with access to food and water ad libitum. Behavioral tests were performed with male mice at 9–38 weeks of age between 9:00 a.m. and 6:00 p.m. as described previously (59–62), unless indicated otherwise. Each apparatus was cleaned with dilute sodium hypochlorite solution before testing of each animal in order to prevent bias due to olfactory cues. Behavioral tests included motor function tests (wire hang test and rotarod test), the hot-plate test, the open-field test, the light–dark transition test, the elevated plus-maze test, startle response and PPI tests, T-maze forced alternation and left–right discrimination tests, the social interaction test in a novel environment, the sociability and social novelty preference test, the nest-building test and the Porsolt forced swim test. Most of the tests were conducted with automated analysis systems as described below, although some behaviors in motor function tests, the hot-plate test and the nest-building test were measured through direct observation by well-trained experimenters. Given that the experimenters were not always blind to genotypes, the possibility of experimenter bias in behavioral measurements by direct observation cannot be completely excluded.

Motor function tests

A wire hang test apparatus (O’Hara & Co., Tokyo, Japan) was used to assess balance and grip strength. The apparatus consists of a box (21.5 × 22 × 23 cm) with a wire mesh grid (10 × 10 cm) on top that can be inverted. Mice at 9–16 weeks of age were placed individually on the wire mesh, which was then inverted, causing the animal to grip the wire. The latency to the mouse falling was recorded with a 60 s cutoff time. Motor coordination and balance were tested with the rotarod test. The mouse was placed on a rotating drum with a diameter of 3 cm (Accelerating Rotarod; UGO Basile, Varese, Italy), and the time that each animal was able to maintain its balance on the rod during its acceleration from 4 to 40 rpm over 5 min was measured. Statistical analysis was performed by one-way analysis of variance (ANOVA).

Hot-plate test

The hot-plate test was used to evaluate sensitivity to a painful stimulus. Mice at 11–15 weeks of age were placed on a hot plate at a temperature of 55.0 ± 0.3°C (Columbus Instruments, Columbus, OH), and the latency to the first hind paw response (foot shake or paw lick) was recorded with a cutoff time of 15 s. Statistical analysis was conducted by one-way ANOVA.

Open-field test

Each mouse at 10–14 weeks of age was placed in the corner of an open-field apparatus (40 × 40 × 30 cm; AccuScan Instruments, Columbus, OH), which was illuminated at 100 lux. The VersaMax system (AccuScan Instruments) with photobeam sensors was used for automatic storage of activity data. Total distance traveled, vertical activity (rearing, measured by counting the number of photobeam interruptions) and time spent in the central area (20 × 20 cm) were recorded over 120 min. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc analysis.

Light–dark transition test

The apparatus for the light–dark transition test consisted of a cage (21 × 42 × 25 cm) that was divided into two sections of equal size by a partition with a door (O’Hara & Co.). One chamber was made of white plastic and brightly illuminated (390 lux), whereas the other was black and dark (2 lux). Mice at 10–14 weeks of age were placed in the dark side and allowed to move freely between the two chambers with the door open for 10 min. The number of transitions between the two compartments, latency to first entry into the light chamber, distance traveled and time spent in each chamber were recorded with the use of ImageLD software. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc analysis.

Elevated plus-maze test

The apparatus consisted of two open arms (25 × 5 cm) and two enclosed arms of the same size with 15-cm-high transparent walls (O’Hara & Co.). The arms and central square were made of white plastic plates and were elevated to a height of 55 cm above the floor. The likelihood of animals falling from the apparatus was minimized by the presence of 3-mm-high plastic ledges on the open arms. Arms of the same type were arranged on opposite sides. Each mouse at 10–15 weeks of age was placed in the central square of the maze (5 × 5 cm) facing one of the closed arms, and its behavior was recorded over 10 min. The number of entries into and the time spent in the open and enclosed arms were measured with the use of ImageEP software. Statistical analysis was performed by one-way ANOVA.

Startle response and PPI tests

A startle reflex measurement system (O’Hara & Co.) was used to measure startle response and PPI. An acceleration sensor was used for automatic quantification of the startle response of each mouse. Each mouse at 15–19 weeks of age was placed in a Plexiglas cylinder and left undisturbed for 10 min. White noise (40 ms) was used as the startle stimulus for all trial types. The startle response was recorded for 400 ms (with measurement of the response every 1 ms) starting with the onset of the startle stimulus. The background noise level in the chamber was 70 dB. The peak startle amplitude recorded during the 140 ms sampling window was measured. A test session consisted of six trial types (two types for startle stimulus-only trials and four types for PPI trials). The intensity of the startle stimulus was 110 or 120 dB. The prepulse sound (74 or 78 dB) was presented 100 ms before the startle stimulus. Four combinations of prepulse and startle stimuli (74 and 110, 78 and 110, 74 and 120 and 78 and 120 dB) were applied. Six blocks of the six trial types were presented in pseudorandom order such that each trial type was presented once within a block. The average intertrial interval was 15 s (range, 10–20 s). Statistical analysis was conducted by one-way ANOVA with Tukey’s post hoc analysis.

T-maze forced alternation task

The forced alternation task was performed with an automatic T-maze (63) constructed of white plastic runways with walls 25 cm in height. The maze is partitioned into six areas by sliding doors that open downward. The stem of the T comprises area S2 (13 × 24 cm), and the arms of the T comprise areas A1 and A2 (11.5 × 20.5 cm). Areas P1 and P2 are connecting passageways from each arm (A1 or A2) to the start compartment (area S1). The end of each arm is equipped with a pellet dispenser for provision of a food reward. Pellet sensors automatically record pellet intake by the mouse. One week before pretraining, each mouse at 26–31 weeks of age was deprived of food until its body weight was reduced to 80% from 85% of the initial value. It was then fed a maintenance diet throughout the course of all T-maze experiments. Before the first trial, the animal was subjected to a 30 min habituation session, during which it was allowed to freely explore the T-maze with all doors open and both arms baited with food. Beginning on the day after habituation, the animal was subjected to pretraining every other day. With all the doors closed and a pellet deposited in the food tray, the mouse was placed in area A1. After it had consumed the pellet or after 5 min had elapsed without pellet consumption, the mouse was transferred to area A2, and the process was repeated. Such pretraining was repeated four times every other day and continued until the animal consumed >80% of the pellets provided during a day. Beginning on the day after completion of pretraining, the mouse was subjected to a forced alternation protocol for 8 days (one session consisting of 10 pairs of training trials per day, with a cutoff time of 50 min). For the first (sample) trial of each pair, the mouse was forced to choose one of the arms of the T-maze (A1 or A2) and received a reward at the end of the arm. After the mouse had consumed the pellet or it had stayed for >10 s without consuming the pellet, the door separating the arm (A1 or A2) and connecting passageway (P1 or P2) was opened to allow the mouse to return to the starting compartment (S1). After a 3 s delay, the mouse was given a free choice between the two T arms and was rewarded for choosing the arm that was not selected for the first trial of the pair. Choosing the incorrect arm resulted in no reward and confinement to the arm for 10 s. The location of the sample arm (left or right) was varied in a pseudorandom manner across trials with the use of a Gellermann schedule so that each mouse received equal numbers of left and right presentations. Various fixed extramaze cues surrounded the apparatus. On the 6th to 8th days, different delays (3, 10, 30 or 60 s) were applied between the forced- and free-choice trials of each pair. Data acquisition and analysis were performed automatically with ImageTM software. Statistical analysis was performed by one-way ANOVA.

T-maze left–right discrimination task

The left–right discrimination task was performed with an automatic T-maze (63) and after food deprivation as described above for the forced alternation task. On the day after completion of the forced alternation task, mice at 31–38 weeks of age were subjected to the left–right discrimination task for 18 days (one session consisting of 10 trials with a cutoff time of 50 min). The animal was able to freely choose either the right or left arm of the T-maze (A1 or A2), with the correct arm being randomly assigned to each mouse. If it chose the correct arm, it received a reward at the end of the arm. Selection of the incorrect arm resulted in no reward and confinement to the arm for 10 s. After the mouse had consumed the pellet or stayed in the arm for >30 s without consuming the pellet, the door that separated the arm (A1 or A2), and connecting passageway (P1 or P2) was opened to allow the mouse to return to the starting compartment (S1). On the 11th day, the correct arm was changed for reversal learning. A variety of fixed extramaze cues surrounded the apparatus. Data acquisition and analysis were performed automatically with ImageTM software. Statistical analysis was performed by one-way ANOVA.

Social interaction test in a novel environment

In the social interaction test, two mice at 11–15 weeks of age and of identical genotypes that had previously been housed in different cages were placed together in a box (40 × 40 × 30 cm) and allowed to explore freely for 10 min. Analysis was performed automatically with the use of ImageSI software. The total number of contacts, total contact duration, mean duration per contact, and total duration of active contacts were measured. Images were captured at a rate of three frames per second, and the distance traveled between two successive frames was determined for each mouse. If the two mice contacted each other and the distance traveled by either mouse was >10 cm, then the behavior was considered an active contact. Active contacts included sniffing and following behavior. Statistical analysis was performed by one-way ANOVA with Tukey’s post hoc analysis.

Sociability and social novelty preference test

The testing apparatus consisted of a rectangular, three-chambered box with a lid fitted with a video camera (O’Hara & Co.). Each chamber measured 20 × 40 × 22 cm, and the dividing walls were made from clear Plexiglas, with small rectangular openings (5 × 3 cm) allowing access into each chamber. An unfamiliar mouse (stranger 1) that had had no prior contact with the subject mouse was placed in one of the side chambers. The location of stranger 1 in the left- versus right-side chamber was systematically alternated between trials. The stranger mouse was enclosed in a small, round wire cage that allowed nose contact between the bars but prevented fighting, with the other side chamber containing an empty cage. Each cage was 11 cm in height, with a bottom diameter of 9 cm and vertical bars 0.5 cm apart. The subject mouse was first placed in the middle chamber and allowed to explore the entire social test box for 10 min. The amount of time spent around the cage and in each chamber was measured with the aid of the camera fitted on top of the box in order to quantify social preference for stranger 1. A second unfamiliar mouse (stranger 2) enclosed in an identical small wire cage was then placed in the chamber that had been empty during the first session. The test mouse thus had a choice between the first, already-investigated unfamiliar mouse (stranger 1) and the novel unfamiliar mouse (stranger 2). The amount of time spent around each cage and in each chamber during a second 10 min session was measured as before. All the mice used in this test were 13–17 weeks of age. Data acquisition and analysis were performed automatically with the use of ImageCSI software. Statistical analysis was conducted with the paired Student’s t-test.

Olfactory habituation–dishabituation test

The ability to discriminate nonsocial and social odors was measured with the olfactory habituation–dishabituation test performed with some modifications, as previously described (37). Mice at 19–21 weeks of age were transferred from their housing to a sound-attenuating room adjacent to a sound-attenuating testing room and were maintained singly in a plastic cage (250 × 182 × 139 mm) with a stainless steel lid and with fresh paper chips as bedding (Paper Clean, Japan SLC, Shizuoka, Japan) for 30 min before testing. During the pretest acclimation period, a clean, dry cotton swab was inserted through the cage lid. Each mouse was then transferred in the new cage to the testing room and was individually assessed for the time spent sniffing cotton-tipped swabs suspended from the lid of the cage. The swabs had been dipped in ultrapure water or in almond or banana extract (Golden Kelly Patent Flavor, Osaka, Japan) each diluted 1:10 in ultrapure water, or they had been wiped in a zigzag pattern across the floor of a plastic cage that contained either four unfamiliar male C57BL/6J mice at 23–25 weeks of age (social odor 1) or four unfamiliar male mice of the same strain but different (22–24 weeks of) age (social odor 2), immediately before the beginning of each trial. The odorant stimuli were presented for 2 min per trial for three consecutive trials each in the order: water, almond, banana, social odor 1 and social odor 2. The intertrial interval was ~1 min. Time spent sniffing the swab was measured manually by an experimenter with an event-recording program developed. Statistical analysis was performed with one-way ANOVA.

Nest-building test

Mice at 16–20 weeks of age were housed individually in cages containing paper chip bedding and one square of pressed cotton (Nestlet, Ancare, Bellmore, NY). No other nesting material was present. After 24 h, manipulation of the Nestlet and the constitution of the built nest were assessed according to a five-point scale as described previously (64): (1) Nestlet not noticeably touched; (2) Nestlet partially torn; (3) Nestlet mostly shredded but with no identifiable nest site; (4) an identifiable but flat nest; and (5) a (near) perfect nest. Statistical analysis was performed with the Kruskal–Wallis test.

Porsolt forced swim test

The testing apparatus consisted of four Plexiglas cylinders (20 cm in height and 10 cm in diameter) that were filled with dilute sodium hypochlorite solution at ~23°C up to a height of 7.5 cm. Mice at 16–20 weeks of age were placed in the cylinders, and immobility time was recorded over a 10 min test period. Images were captured at a rate of two frames per second. For each pair of successive frames, the area (number of pixels) within which the mouse moved was measured. When this area was below a certain threshold, the mouse was judged to be immobile. When the area equaled or exceeded the threshold, the mouse was considered to be moving. The optimal threshold was determined by adjustment based on the degree of immobility measured by human observation. Immobility lasting <2 s was not included in the analysis. Data acquisition and analysis were performed automatically with the use of ImageTS/PS software. Statistical analysis was performed by one-way ANOVA.

Image analysis for behavioral tests

The light–dark transition test, elevated plus-maze test, T-maze tests, social interaction test, sociability and social novelty preference test and Porsolt forced swim test were each performed with the use of an automated analysis system consisting of the testing apparatus, a video camera mounted on the apparatus and computer with image analysis program (ImageLD, ImageEP, ImageTM, ImageSI, ImageCSI, ImagePS). The programs are based on the public domain ImageJ program (developed by W. Rasband at the National Institute of Mental Health, Bethesda, MD) and were developed and modified for each test by T.M. ImageLD/EP/TM can be freely downloaded from the Mouse Phenotype Database (http://www.mouse-phenotype.org). The programs were used to automatically measure mouse behavior by analysis of video images [for details, see (63, 65–67)], which might help to minimize potential experimenter effects on behavioral results.

Chromatin immunoprecipitation

ChIP was performed essentially as described previously (68). Nuclear extracts of male mouse brain at 9 weeks of age were fixed with a final concentration of 0.5% formaldehyde, suspended in ChIP buffer [5 mm HEPES-KOH (pH 8.0), 200 mm KCl, 1 mm CaCl₂, 1.5 mm MgCl₂, 5% sucrose, 0.5% Nonidet P-40, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 mm phenylmethylsulfonyl fluoride], incubated for 10 min on ice, subjected to ultrasonic treatment with the use of a Diagenode Bioruptor and digested with micrococcal nuclease for 40 min at 30°C. After the addition of EDTA to a final concentration of 0.1 mm, each digested sample was centrifuged at 15 000×g for 10 min at 4°C, and the resulting supernatant was incubated with rotation overnight at 4°C with antibodies conjugated to magnetic beads. Bound proteins were eluted from the beads, and cross links were reversed by incubation overnight at 65°C with 1% SDS in Tris-EDTA buffer. After washing twice both with ChIP buffer and with Tris-EDTA buffer, DNA was purified with the use of NucleoSpin Gel and PCR Clean-Up (Takara Bio, Shiga, Japan) and subjected to real-time PCR analysis as described above. PCR primers (sense and antisense, respectively) were as follows: Cspg4 (used as a control), 5′-CCTTACAAGTCCAGGCACCCAG-3′ and 5′-TCAGTCCCTCACTCACCAGGAG-3′; Mbp, 5′-AGCAGCAGCCAGCACCAGTAAG-3′ and 5′-TCGCCTCTCAAGGAGTCAGATTC-3′; Plp1, 5′-GGGTGCGGTGTGTTTGGTAG-3′ and 5′-TAAATGGAGGGGTTCTTGAATCC-3′; Cnp, 5′-GCAAATGAGGACACCAGGTTCC-3′ and 5′-TCCAAAGACACTAAGACTCCCCC-3′; Mag, 5′-GCCTGGAGCTTTCAGAAAGATGA-3′ and 5′-CTGGGACTGGGCAGCTTGAT-3′; Mog, 5′-GTTTCGTTTCCTTTTTTCATCGTTC-3′ and 5′-GCCACCTCACTGACCCTCAAC-3′; Ugt8a, 5′-TTCCCTCACGCCCAGCCTCC-3′ and 5′-CGCTCCCTCACCTCGGACAG-3′; Myrf, 5′-ATCAACCTCCTGCTCTGGGATAG-3′ and 5′-CGAAGCCATTCACCTCGCTG-3′; and Mobp, 5′-CACCTGGCAACAGAATCTCACTC-3′ and 5′-CTCCCTCTCTCTCTCTTTTCCTGC-3′.

ChIP-seq analysis

Available ChIP-seq data obtained with antibodies to CHD8 (DRA003116 and GSM1381224) (19, 26) were reanalyzed as previously described (19). In brief, the reads were uniquely mapped to the mouse (mm9) genome with the use of BowTie software (version 2.2.5), and duplicated reads were removed with SAMtools (version 0.1.19-44 428cd). Markedly enriched regions of the genome were identified with the use of the MACS peak caller (version 2.1.0.20140616, with the option ‘-gsize mm-nomodel-extsize 160-broad-to-large-pvalue 1e-3’).

RNA-seq analysis

Total RNA was extracted from the whole brain of male mice at P14 and 9 weeks of age with the use of a TRIzol Plus RNA Purification Kit (Thermo Fisher Scientific). Messenger RNA (1 μg) purified from total RNA with the use of a NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs, Ipswich, MA) was used to prepare a cDNA library with the use of a NEBNext Ultra Directional RNA Library Prep Kit for Illumina (New England Biolabs) (68). The library was then sequenced with the use of a HiSeq 1500 system (Illumina). Quality of the raw sequencing data was checked with FastQC (version 0.11.5), and trimming of adapter sequences was performed with FASTX-Toolkit (version 0.0.14). The total amount of each transcript was calculated with the use of a series of programs including Top Hat (version 2.1.1) and Cufflinks (version 2.2.1). RNA-seq reads were mapped against the mouse (mm9) genome. RNA-seq data have been deposited in the NCBI sequence read archive under accession number PRJNA389641.

Gene set enrichment analysis

GSEA is a computational method that determines whether an a priori-defined set of genes shows statistically significant, concordant differences in expression between two biological states (69). The primary result of GSEA is the enrichment score (ES), which reflects the degree to which a gene set is overrepresented at the top or bottom of a ranked list of genes. The top portion of a GSEA plot (represented by a green line) shows the running ES for the gene set as the analysis moves down the ranked list. The score at the peak of the plot (the score farthest from 0) is the ES for the gene set. The normalized enrichment score represents this ES value normalized by the mean ES for all permutations of the data set and takes into account differences in gene set size. It thus represents the degree of expression skew. The bottom portion of the plot shows where the members of the gene set (indicated by black bars) appear in the ranked list of genes. If most of the black bars are gathered to the left side (or right side), then most of the genes in the gene set are upregulated in the sample named at the lower left (or lower right). GSEA was performed as described previously (69) with the use of GSEA version 3.0. The expression data for 43 ASD patients (11–67 years of age, with a median of 30 years; 32 males and 11 females) and 63 control subjects (15–60 years, with a median of 28 years; 55 males and 8 females) were obtained from a previous study (31). Gene sets specifically expressed in each cell type of the CNS were also obtained from previous studies (29, 30).

GO analysis

GO analysis was performed with the use of DAVID (70). A total of 558 genes significantly downregulated (FDR q value of < 0.05) in the ASD cortex was used for analysis (31). A total of 304 genes significantly downregulated [log₂(fold change) of < −0.3 and P < 0.05] in the brain of Olig1-Cre/Chd8_L^F/F mice was also analyzed.

In vivo electrophysiological analysis

Male mice at 9–16 weeks of age were anesthetized by intraperitoneal injection of a mixture of xylazine (10 mg/kg) and ketamine (60 mg/kg). Supplemental doses of the anesthetic (a maximum of one-half of the initial dose) were administered as needed to maintain an adequate level of anesthesia. Atropine sulfate (0.4 mg/kg) and dexamethasone (0.2 mg/kg) were injected intramuscularly to avoid the vagal response and brain edema. After tracheal intubation, mice were artificially ventilated with a rodent respirator (SN-480-7, Shinano). They were fixed in a stereotaxic frame and immobilized by intramuscular injection of pancuronium bromide (0.08 mg/kg) before the recordings. Rectal temperature was maintained above 35°C by warming during recording. For recording from the corpus callosum, a concentric electrode was inserted to stimulate commissural fibers of the right hemisphere (1.0 mm posterior and 1.5 mm lateral to the bregma and ~1.8 to 2.2 mm below the surface) (71). A tungsten electrode was inserted in commissural fibers of the left hemisphere (1.0 mm posterior and 1.5 mm lateral to the bregma and ~1.8 to 2.2 mm below the surface) to record CAPs evoked by stimulation. The intensity and pulse duration of stimulation were fixed at 70 μA and 100 μs, respectively. The data were stored in a computer with pCLAMP 10 software (Molecular Devices, San Jose, CA). The recorded volleys were positive–negative–positive triphasic peaks corresponding to current source–sink–source related to propagating axonal action potentials. We defined the initial, positive peak as ‘onset’ and the second, negative peak as ‘peak.’ The initial, positive peak was regarded as the time of impulse arrival through the fastest conducting fibers and the second, negative peak as the median. Conduction velocity was estimated from latency and electrode distance.

In vitro electrophysiological recordings

Male mice at 7–10 weeks of age were deeply anesthetized by intraperitoneal administration of pentobarbital (100 mg/kg). The brain was rapidly removed and immersed in an ice-cold solution containing 234 mm sucrose, 2.5 mm KCl, 10 mm MgCl₂, 0.5 mm CaCl₂, 25 mm NaHCO₃, 1.25 mm NaH₂PO₄, 0.5 mmmyo-inositol and 11 mm glucose. Coronal brain slices (300 μm) were cut with a vibratome (VT-1200S, Leica) and immersed in a physiological solution (125 mm NaCl, 2.5 mm KCl, 1 mm MgCl₂, 2 mm CaCl₂, 26 mm NaHCO₃, 1.25 mm NaH₂PO₄, 20 mm glucose, 5 mm sodium ascorbate, 3 mm sodium pyruvate, and 2 mm thiourea), which was continuously bubbled with a mixture of 95% O₂ and 5% CO₂. Ascorbate, pyruvate and thiourea were omitted during recordings. The whole-cell patch clamp was made at 31–32°C. Layer 2/3 pyramidal neurons in the prelimbic cortex were identified with a microscope (BX50WI, Olympus, Tokyo, Japan) fitted with a 40× water immersion objective (NA of 0.8, Olympus). The pipette solution for voltage clamp recording consisted of 108 mm cesium gluconate, 0.4 mm EGTA, 2.8 mm NaCl, 5 mm triethylammonium chloride, 4 mm ATP (Mg²⁺ salt), 0.3 mm GTP (Na⁺ salt), 10 mm disodium phosphocreatine, and 20 mm HEPES, with an osmolarity of 290 mosM. The membrane potential was not corrected for liquid junction potential. The series resistance of the recorded cells was <25 MΩ. Spontaneous and miniature excitatory or inhibitory postsynaptic currents (sEPSCs, sIPSCs, mEPSCs and mIPSCs) were recorded from the same neuron as inward currents at −70 mV and outward currents at 0 mV, respectively. The mEPSC and mIPSC events were recorded in the presence of physiological solution containing 0.5 μM tetrodotoxin (Alomone Labs, Jerusalem, Israel). Recordings were amplified with a Multiclamp 700A amplifier (Molecular Devices), digitized at 10 kHz with a Digidata 1322A apparatus (Molecular Devices) and collected with pCLAMP 8 software (Molecular Devices). Events were detected with a scaled template algorithm and analyzed with IGOR Pro (WaveMetrics) and Clampfit (Molecular Devices) software. The amplitude and frequency of events were determined for each neuron over a 3 min period of recording.

Statistical analysis

Electron microscopy, immunofluorescence staining of Caspr and Nav1.6/Kv1.2, electrophysiology, RNA-seq analysis and behavioral tests were performed with male mice. Immunofluorescence staining of GFAP and primary culture of OPCs were performed with male and female mice. The genotypes of mice were known before performance of cell-based experiments, electron microscopy and immunohistofluorescence analysis. For image analysis, cell and electron microscopy images were quantified in a blinded manner. No statistical method was used to predetermine sample sizes, which were similar to those generally adopted in the field. The number of mice subjected to each experiment is stated. Statistical analysis by the paired Student’s t-test, two-tailed one-sample t-test, hypergeometric test, the Kruskal–Wallis test or one-way ANOVA with Tukey’s post hoc analysis was performed with the use of SAS University Edition software and JMP software. The unpaired Student’s t-test or Welch’s t-test was performed with the use of Microsoft Excel software or GraphPad Prism 7.3 software. Pearson’s correlation was calculated with the use of JMP software. All data are means ± SEM. Significance levels for P and FDR q values are indicated with ^*, ^** and ^*** for <0.05, <0.01 and <0.001, respectively.

Acknowledgements

We thank Takayoshi Yamauchi, Yasuyuki Kita, Kayoko Tsunematsu, Shogo Nakayama and other laboratory members for the technical assistance and discussion; Ryo Ugawa, Kanako Ichikawa, Masato Tanaka, Emiko Koba, Tomomi Akinaga, Shiori Akutsu, Kaoru Isa, Yoshio Bando, Kaoru Takakusaki, Eiichi Hinoi, and Yasuyuki Ohkawa for the technical assistance and advice; and Akane Ohta for helping with the preparation of the manuscript. Computations were performed in part on the National Institute of Genetics supercomputer at National Institute of Genetics, Research Organization of Information and Systems. A.K. was supported by a fellowship from the Japan Society for the Promotion of Science (JSPS).

Conflict of Interest statement: None declared.

Funding

KAKENHI grants (18H05215, 19H05220); Grant-in-Aid for Scientific Research on Innovative Areas (Comprehensive Brain Science Network) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.