Distinct, dosage-sensitive requirements for the autism-associated factor CHD8 during cortical development

Background CHD8 haploinsufficiency causes autism and macrocephaly with high penetrance in the human population. Chd8 heterozygous mice exhibit relatively subtle brain overgrowth and little gene expression changes in the embryonic neocortex. The purpose of this study was to generate new, sub-haploinsufficient Chd8 mouse models to allow us to identify and study the functions of CHD8 during embryonic cortical development. Methods To examine the possibility that certain phenotypes may only appear at sub-heterozygous Chd8 levels in the mouse, we created an allelic series of Chd8-deficient mice to reduce CHD8 protein levels to approximately 35% (mild hypomorph), 10% (severe hypomorph) and 0% (neural-specific conditional knockout) of wildtype levels. We used RNA sequencing to compare transcriptional dysregulation, structural MRI and brain weight to investigate effects on brain size, and cell proliferation, differentiation and apoptosis markers in immunostaining assays to quantify changes in neural progenitor fate. Results Mild Chd8 hypomorphs displayed significant postnatal lethality, with surviving animals exhibiting more pronounced brain hyperplasia than heterozygotes. Over 2000 genes were dysregulated in mild hypomorphs, including autism-associated neurodevelopmental and cell cycle genes. We identify increased proliferation of non-ventricular zone TBR2+ intermediate progenitors as one potential cause of brain hyperplasia in these mutants. Severe Chd8 hypomorphs displayed even greater transcriptional dysregulation, including evidence for p53 pathway upregulation. In contrast to mild hypomorphs, these mice displayed reduced brain size and increased apoptosis in the embryonic neocortex. Homozygous, conditional deletion of Chd8 in early neuronal progenitors resulted in pronounced brain hypoplasia, partly caused by p53 target gene derepression and apoptosis in the embryonic neocortex. Limitations Our findings identify an important role for the autism-associated factor CHD8 in controlling the proliferation of intermediate progenitors in the mouse neocortex. We propose that CHD8 has a similar function in human brain development, but studies on human cells are required to confirm this. Because many of our mouse mutants with reduced CHD8 function die shortly after birth, it is not possible to fully determine to what extent reduced CHD8 function results in autism-associated behaviours in mice. Conclusions Together, these findings identify important, dosage-sensitive functions for CHD8 in p53 pathway repression, neurodevelopmental gene expression and neural progenitor fate in the embryonic neocortex. We conclude that brain development is acutely sensitive to reduced CHD8 expression and that the varying sensitivities of different progenitor populations and cellular processes to CHD8 dosage result in non-linear effects on gene transcription and brain growth. Shaun Hurley, Conor Mohan and Philipp Suetterlin have contributed equally to this work.

CHD8 encodes a member of the ATP-dependent CHD chromatin remodelling family of proteins [8] and was initially identified as a direct repressor of β-catenin and p53 target genes [9][10][11][12]. Early embryonic lethality of homozygous Chd8 deletion in the mouse is associated with p53-mediated apoptosis, consistent with its role as a transcriptional repressor of p53 target genes [11]. By contrast, CHD8 is typically recruited to promoters enriched for transcriptionally-permissive chromatin marks in neural progenitors, suggesting a role for CHD8 in transcriptional activation [13,14]. Indeed, ASD-associated genes were downregulated in neural progenitor cells upon CHD8 knockdown [13,14]. Evidence for mild brain overgrowth, reminiscent of the macrocephaly observed in patients with CHD8 mutations, has been reported in several different Chd8 heterozygous mouse models [15][16][17][18].
To explore the transcriptional dysregulation that may underlie abnormal brain development in heterozygous mice, gene expression has been investigated at different stages of brain development. These studies have revealed subtle gene expression changes in Chd8 +/− mice during embryonic development [15,[17][18][19]. By contrast, in vitro studies on neural progenitor cells have identified more substantial transcriptional dysregulation arising from CHD8 knock-down. For instance, Sugathan et al. observed 1756 differentially expressed genes (DEGs) upon CHD8 knock-down in human iPSC-derived neural progenitor cells [14]. Gene expression changes in Chd8 +/− mice also appear to be strongly influenced by developmental stage, with more extensive transcriptional changes observed at peri-and postnatal stages [17,18].
The striking phenotypes associated with human CHD8 mutations and pronounced gene expression changes in neural progenitor cell lines upon CHD8 knockdown, contrast with the mild brain and embryonic transcriptional abnormalities observed in Chd8 heterozygous mice. The only study so far to report convincing ASD-like behavioural phenotypes associated with Chd8 deficiency, involved Chd8 knock-down to ~ 20% of wildtype protein levels by selective in utero electroporation of neuronal progenitors contributing to upper layer neurons [20]. Together, these findings suggest that CHD8 haploinsufficiency may have more pronounced effects on human brain development, or that some human ASD-associated mutations may reduce CHD8 function by more than 50%.
To examine the possibility that certain phenotypes may only appear at sub-heterozygous Chd8 levels in the mouse, we created an allelic series of Chd8-deficient mice to reduce CHD8 protein gradually to approximately 35% (mild hypomorph), 10% (severe hypomorph) and 0% (conditional knockout) of wildtype levels. Non-monotonic effects on brain growth were observed, with mild hypomorphs exhibiting increased brain size, and severe hypomorphs and conditional knockout mice smaller brain size. We found that increased brain size in mild hypomorphs was associated with increased proliferation of TBR2+ intermediate progenitors. As this cell type contributes to increased human brain expansion during evolution [21], this finding suggests that CHD8 may have more pronounced effects on human brain development. We conclude that reducing CHD8 function below 50% has disproportionally large, and non-linear effects on gene expression and brain development.

Genotyping of mice
Genomic DNA was extracted for genotyping from ear notches (or yolk sac for embryos aged E14.5 and below) using Proteinase K digestion or the HotSHOT method [25]. Genotyping reactions were then performed for the presence of Chd8 wildtype, neo, null or floxed alleles, p53 wildtype or floxed alleles, as well as the presence of Cre. Thermal cycles for all genotyping reactions were as follows: 94 °C, 5 min; 35X (94 °C, 30 s; 58 °C, 30 s; 72 °C, 30 s); 72 °C, 5 min. Primer pairs to amplify a sequence distinguishing between Chd8 flox , Chd8 neo and wildtype alleles ('Chd8flox' primers, 212 bp and 275 bp product for mutant or wildtype, respectively), to detect the presence of the Chd8 null allele ('Chd8null' primers, 395 bp), to distinguish between p53 floxed and wildtype alleles ('p53flox' primers, 390 bps and 270 bps, respectively) and primers to amplify a specific Cre sequence ('Cre' primers, 390 bp product) were used as listed in Additional file 1: Table S5.

RNA extraction and qRT-PCR analysis
Cortical RNA was extracted by lysing cortices in 600 µl Trizol (Life Technologies). After purification, RNA was DNase treated using the Direct-zol RNA MiniPrep kit (Zymo Research) according to the manufacturer's recommendations. cDNA was synthesised for qRT-PCR experiments using 50 ng RNA from 4 biological replicates per condition with the Precision nanoScript 2 Reverse Transcription Kit (PrimerDesign Ltd.) according to the manufacturer's instructions. qRT-PCRs were performed on a Stratagene Mx3000p (Agilent Technologies) using Pre-cisionPlus-MX 2 × qPCR Mastermix with SYBR green (PrimerDesign Ltd.) and primers against Atm, Atr, Trp53, Cdkn1a, Ccng1, Mdm2, Chd8, Pmaip1, Kdm5b, Zcwpw1 and Stxbp1. Relative expression levels were calculated using the 2 −∆∆CT method and Gapdh and Ywhaz were used as endogenous control genes.

Western blot
Telencephalic vesicles were dissected from E12.5 embryos and whole cell protein prepared by lysing in 8 M urea, 1% CHAPS, 50 mM Tris (pH 7.9) lysis buffer containing protease inhibitors (PMSF, Pepstatin A, Leupeptin, Aprotinin; Roche) and a phosphatase inhibitor cocktail (Sigma). After rotating at 4 °C for 30 min, DNA was removed from lysates by centrifugation. Supernatant was transferred to a fresh tube and stored at -80 °C. Protein loading samples were made by diluting samples in Laemmli buffer containing 10% β-mercaptoethanol, followed by boiling at 100 °C for 10 min. Samples were loaded (10 µg total protein per lane) onto a Mini-PRO-TEAN pre-cast gel (Bio-Rad) and resolved using gel electrophoresis. Protein was transferred to a nitrocellulose membrane (Bio-Rad), which was then blocked in 5% nonfat milk powder (Bio-Rad) and 1% bovine serum albumin (BSA, Sigma) in TBS with 0.1% Tween-20 (TBST) for one hour at room temperature. Where β-actin was used as a loading control, the membrane was then cut in two: the higher molecular weight section was incubated with anti-CHD8 primary antibody (rabbit anti-CHD8 N-terminal, A301-224A, Bethyl Laboratories, 1/2000) and the lower molecular weight section incubated with antiβ-actin antibody (rabbit anti-β-actin, ab8227, Abcam, 1/4000); both antibodies in 3% non-fat milk powder and 1% BSA in TBST overnight at 4 °C. After washing, membrane was incubated with HRP-conjugated secondary antibody (Millipore) for one hour at room temperature. HRP was detected with Clarity ECL reagent (Bio-Rad) and the membrane imaged using a Bio-Rad ChemiDoc system. Where GAPDH was used as a loading control, the uncut membrane was washed in TBST after detection of CHD8 protein and incubated overnight at 4 °C in 0.05% sodium azide in PBS, before washing and incubation with anti-GAPDH primary antibody (rabbit anti-GAPDH, ab8245, Abcam, 1/40,000) overnight at 4 °C. Membrane was probed with HRP-conjugate and imaged as before. Raw proteins levels were quantified using Bio-Rad ImageLab software. All replicates on each given blot were first normalised to their respective GAPDH loading controls. Where the same samples were run across multiple blots their normalised values were averaged across all blots and this value used for statistical analysis. Normalised values for all samples were compared by ANOVA to identify statistically significant differences between the different genotypes.

Structural MRI
Mice were terminally anaesthetised and intracardially perfused as described previously [18]. Voxelwise comparisons were made between mutants and all wildtypes taken from both the Chd8 +/− and Chd8 neo/neo batches. As wildtype brain sizes differed slightly between the two groups due to different age at analysis, Chd8 neo/neo data were first normalised (beta-corrected) to wildtypes in the Chd8 +/− batch before analysis. Voxelwise comparisons were then made between mutants and all wildtypes, and multiple comparisons in this study were controlled for using the False Discovery Rate [26].

Behavioural assessments
Mice for behavioural testing were maintained as described previously [18]. Housing and test rooms were kept at constant temperature (21 °C) and humidity (45%) and maintained under a regular light/dark schedule with lights on from 07:30 to 19:30 h (light = 270 lx). All mice used in behavioural assessments were housed and tested at the same facility at King's College London.
One week before performing the rotarod test, mice were singly-housed to avoid any potential confounds from social and aggressive behaviour hierarchies, which could influence the controlled assessment of social behaviours [27]. Sawdust was changed every other week but never on the day before, or the day of testing. The enrichment (nesting material and house) was changed less regularly to minimize the disruption to the animals. For all social tests, conspecific mice were housed in a separate room to the test mice to ensure the conspecifics were unfamiliar to the test mice. Test mice were never exposed to the same conspecific during testing to ensure novelty.
Behavioural experiments were conducted between 08:30 and 18:30 in sound-proofed rooms under standard room lighting unless stated otherwise. Behaviours were recorded using a camera positioned above the test arenas and movement of each mouse tracked using EthoVision (Noldus Information Technologies, bv Wageningen, The Netherlands). Social investigation, olfaction and grooming were scored manually using MATLAB 2016b (The MathWorks, Inc., Natick, Massachusetts, USA). After each individual trial of a specific test, boli and urine were removed from the test arena which was cleaned with 1% Anistel ® solution (high level surface disinfectant, Trisel Solution Ltd, Cambridgeshire, UK) to remove any odours. Experimenters were blind to the genotype of the animals both during the testing and subsequent scoring of the recorded behaviours.

RNA sequencing
For RNA-sequencing at E10.5, total RNA from 2 embryos was pooled for each biological replicate (n = 3 per condition). No pooling was performed at E12.5 (n = 3 per condition). mRNA was isolated and reverse transcribed into cDNA. cDNA was end-repaired, adaptor-ligated and A-tailed. Paired-end sequencing (75 bp read length) was performed on the Illumina HiSeq 4000 platform. Quality of the raw sequencing data was checked using FastQC version 0.11.2 and trimming of adaptor sequences was performed using Trim Galore! version 0.4.1 [28]. Reads were aligned to the mouse genome (GRCm38.p4) using Tophat version 2.1.0 and aligned reads were counted using FeatureCounts version 1.5.0 [29,30]. Differential expression testing was performed using DESeq2 version 1.10.1, as previously described [31]. Gene ontology analysis and functional classification was performed using DAVID with all detected DEGs below a 0.05 FDR [32]. For heatmaps, data were transformed with a variance stabilising transformation, scaled and clustered with the Ward.d2 method using maximum distance, and plotted with the R package pheatmap version 1.0.8. The R package ggplot2 version 2.1.0 was used to generate volcano plots and DESeq2 was used to generate normalised read count plots for individual genes. The list of ASD associated genes used for overlap with the hypomorph DEGs was obtained from the SFARI Human Gene database (https ://gene.sfari .org/autdb /HG_Home.do). RNA-seq data have been deposited into GEO, accession number GSE121381.

Tissue collection and processing
Embryos were collected and brains dissected from the skulls in ice-cold PBS for E18.5 embryos. Wholemount pictures were taken on a Nikon SMZ1500 stereomicroscope equipped with a Nikon DS-Fi1 camera head, followed by fixation in 4% PFA for 24 h at 4 °C. For BrdU experiments, pregnant mothers (day 13 or 15 of gestation) were injected with 40 mg/kg BrdU in 0.9% saline 1 h prior to embryo collection. After fixing, embryos were dehydrated and paraffin embedded. Paraffin blocks were then cut into 10 µm (cKO and cKO; p53-het embryos) or 5 µm (Chd8 +/− , Chd8 neo/neo and Chd8 neo/− embryos) thick coronal sections and mounted on slides.

X-gal staining
E9.5 embryos were collected and dissected in ice-cold PBS and fixed in 4% PFA for 10 min. Following three washes in PBS (5 min each), embryos were incubated in X-Gal staining solution (10 mM TRIS-HCL, pH 7.3, 0.005% Na-deoxycholate, 0.01% IGEPAL, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 2 mM MgCl 2 , 0.8 mg/ml X-Gal, in PBS) at room temperature until adequate signal was observed. Reactions were stopped by washing in PBS (3 × 5 min) followed by post-fixation in 4% PFA for 1 h. Control embryos never showed any staining.

Immunohistochemistry and immunofluorescence
Coronal brain sections were re-hydrated using standard protocols. Antigen retrieval was conducted by heating slides in 10 mM Sodium Citrate solution (pH6) for 20 min and cooled on ice. For non-fluorescence immunohistochemistry, endogenous peroxidases were blocked by incubating in 3% H 2 O 2 and 10% MeOH in PBS for 15 min. Sections were then washed in 0.2% Triton X-100 (Sigma-Aldrich) in PBS (PBT2) for 5 min and blocked using 10% heat-inactivated normal goat serum (GS) and 2% gelatin in PBT2 for 1 h. Sections were incubated in 5% GS in PBT2 containing primary antibody overnight at 4 °C. The following antibodies were used: mouse anti-BrdU (BD Biosciences, 1/100), rabbit anti-phosphohistone 3B (Cell Signaling, 1/100), mouse anti-phosphohistone 3B Slides were mounted onto coverslips using DPX (Sigma-Aldrich). Images were acquired on a Nikon 80i microscope equipped with a Hamamatsu C4742 CCD or Nikon 5 M pixel Nikon DS digital cameras. Images were processed using Adobe Photoshop and Illustrator.

Fluorescence-activated cell sorting (FACS) and qRT-PCR
Pregnant females were sacrificed by cervical dislocation on embryonic day 14.5 (E14.5) and embryos were dissected out and placed in ice cold Hibernate-E solution supplemented with B27. The brain was dissected out from each embryo and the cortex removed. Cortices were maintained in ice-cold Hibernate-E/B27 whilst embryos were genotyped, after which cortices with corresponding genotypes were pooled together in digest solution (0.25% Trypsin, 0.01% DNase, 10 mM HEPES in HBSS) preheated to 37 °C. Cortices were incubated in digest solution for 10 min and washed in HBSS + 0.01% DNase to inactivate trypsin. Single-cell suspension was achieved through mechanical dissociation by gentle pipetting. Cells were washed and resuspended with PBS, incubated on ice and in the dark for 30 min with 1 µl of the near-IR fluorescent reactive dye from the LIVE/DEAD fixable dead cell stain kit (Ther-moFisher). Cells were washed once with PBS, strained into fixing solution (1.6% PFA and 0.1% saponin in molecular-grade PBS) supplemented with 1% RNasin Plus RNAse inhibitor and incubated on ice for 30 min. Cells were washed twice with wash solution (0.2% BSA, 0.1% saponin, 0.1% RNasin Plus in molecular-grade PBS) before antibody staining. The following primary antibodies were used: phycoerythrin (PE)-conjugated mouse anti-PAX6 (sc-81649, Santa Cruz Biotechnology) and chicken anti-TBR2 (AB15894, Merck Millipore). Cells were suspended in staining solution (1% BSA, 0.1% saponin, 1% RNasin Plus in moleculargrade PBS) containing primary antibodies at a 1:1000 dilution and incubated on ice for 30 min. Cells were washed twice with wash solution then suspended in staining solution containing anti-chicken AlexaFluor 488 secondary antibody (A-11039, ThermoFisher) at a 1:1000 dilution. Cells were washed twice with wash solution then strained and resuspended in sorting buffer (3% BSA and 2 mM EDTA in molecular grade PBS). Fluorescence-activated cell sorting (FACS) was performed using a FACSAria III machine (BD). Samples stained individually for TBR2 and PAX6 were used as compensation controls. Debris was excluded from samples by gating on a forward and side scatter area plot and doublets excluded by gating on a plot of side scatter area by side scatter width. The LIVE/ DEAD marker was detected in the APC-Cy7 channel. Any cells showing greater fluorescence intensity than an unstained control sample were considered dead and excluded. The unstained sample was also used to set a double negative gate on a PE by FITC fluorescence intensity plot. The PAX6+ gate was placed based on increased PE fluorescence and the TBR2+ gate placed based on increased FITC fluorescence. Selection gates were placed towards the extreme ends of the cell population to ensure the sorting of single positive cells only. Cells falling within the PAX6+ or TBR2+ gates were sorted and collected in sorting buffer. Cells were then pelleted, snap frozen in liquid nitrogen and stored at −80 °C. RNA was extracted from sorted cell samples using the RecoverAll Total Nucleic Acid isolation kit (Life technologies) using the RNA isolation protocol with an adapted protease treatment for 3 h at 50 °C to maximise RNA yield. RNA concentration was determined with a NanoDrop 2000 (ThermoFisher). Typically, samples yielded approximately 120 ng of RNA which was stored at −80 °C. Reverse transcription was performed using the Precision nanoScript2 Reverse Transcription Kit (Primerdesign) according to the manufacturer's instructions using random nonamer primers. 60 ng of RNA was reverse transcribed for all samples and resulting complementary DNA (cDNA) stored at −20 °C. qPCRs were performed using Luna Universal qPCR Master Mix (New England Biolabs). Samples of 3.75 ng cDNA were run in triplicate for each primer pair using a Lightcycler 480 (Roche) and fold expression changes were calculated using the 2 −ΔΔCT method. Primer sequences are given in Additional file 1: Table S5.

General
Data are reported as Mean ± SEM and graphs show all individual data points where feasible. Significant p values are reported in the results section and figure legends provide details of relevant statistical parameters, including group sizes. Statistical analyses were performed either with SPSS (Version 22, IBM, Armonk, USA) or Graph-Pad Prism (Version 6, GraphPad Software, La Jolla, California, USA). All analyses were performed blind to genotype.

RNAseq
Processing of raw data and differential expression testing is described in the methods section. Multiple comparisons were controlled for using an FDR < 0.05. Exact p values and FDR adjusted p values for all differentially expressed genes are listed in Additional file 3: Tables S2  and Additional file 5: Table S4.

Cell marker quantification Proliferation
Proliferation was quantified by counting either ventricular or non-ventricular phosphohistone 3B-positive cells and normalising cell counts to the length of ventricular surface. Cells were counted in sections located between A-P positions 219-236 of the E14.5 Allen brain reference atlas for coronal sections (https ://devel oping mouse .brain -map.org/exper iment /thumb nails /10007 4513?image _type=hp_yello w) and from the boundary with the eminences up to the "apex" of the neocortex in the M-L axis. Normalised counts were averaged across both hemispheres for at least three sections to give the number of phosphohistone 3B-positive cells per μm of ventricular surface in the dorsal cortex of each embryo. To determine the molecular identity of non-ventricular PH3B+ cells at E14.5, sections were co-stained for TBR2 and the number of TBR2/PH3B double-positive cells counted and normalised to ventricular surface length.

Apoptosis
Cleaved Caspase 3 (CC3) positive cells were counted in the E14.5 neocortex and normalised to ventricular zone length in each hemisphere as described above. Counts were averaged across both hemispheres and across a minimum of three sections per biological replicate. In E12.5 cKO embryos, CC3-positive cells were counted in 50 μm × 50 μm boxes. Three boxes were counted for both inner (ventricular side) and outer (pial side) regions of the dorsal cortex to generate an average number of CC3-positive cells per μm 2 for both inner and outer cortical regions, which were then averaged to provide the overall mean of CC3-positive cells per μm 2 . These were calculated for both sides of the brain individually in at least two sections per biological replicate. Apoptosis in the ventral cortex was quantified by counting CC3+ cells either in three 0.1 mm × 0.1 mm boxes in both lateral ganglionic eminences and three 0.15 mm × 0.15 mm boxes in both medial ganglionic eminences (WT and cHet), or three 0.1 mm × 0.1 mm boxes placed at equivalent positions across the ventral cortex (cKO and cKO-p53Het) and counts averaged for each section. These were calculated for both sides of the brain individually in at least two sections per biological replicate.

Data availability
The RNAseq raw data and read counts were deposited at the Gene Expression Omnibus (GEO) archive under the Accession Number GSE121381.
Chd8 neo/neo mice exhibited a significant reduction in postnatal survival ( Table 1). As CHD8 is expressed in multiple tissues during development [35], this postnatal lethality is likely a consequence of congenital defects affecting essential organs.
Mice were born at expected Mendelian frequencies from Chd8 neo/+ intercrosses. Percentage survival is shown for the different genotypes at P0, P14 and P35. Note the significantly reduced percentage of homozygous Chd8 neo/ neo mutants obtained at P14 and P35 (***p < 0.001, Chisquare test) in two independent groups of mice.
High resolution structural MRI revealed volumetric increases in a number of brain regions in Chd8 neo/ neo mice compared to wildtype littermates (Fig. 1d). This phenotype was more pronounced than in Chd8 +/− mice (Fig. 1d), with total brain volume increased by 4.5% in Chd8 neo/neo mice, compared to the 2.7% increase in Chd8 +/− mice (Fig. 1e). As our Chd8 +/− data were from males only, we also calculated the increase in brain volume increase in male Chd8 neo/neo mice only, to exclude the possibility that female Chd8 neo/neo mice were responsible for the observed increase in brain volume. The average brain volume of Chd8 neo/neo male mice (n = 8) were increased by 5.4% compared to their male wildtype littermates (n = 14), compared to the 2.7% increase in volume of male Chd8 +/− mice (n = 12) compared to their male wildtype littermates (n = 12). Several regions that showed evidence of overgrowth in Chd8 +/− mice (Fig. 1e, Table 2) demonstrated robust increases in volume in Chd8 neo/neo mice, including the frontal, cingulate and entorhinal cortices and the hippocampus (Fig. 1e, Table 2, Additional file 2: Table S1).
Chd8 neo/neo mice demonstrate mild behavioural anomalies similar to those seen in Chd8 +/− mice To explore whether behavioural phenotypes may be exacerbated in the Chd8 neo/neo mild hypomorphs compared to Chd8 +/− mice, we assessed socio-communicative, repetitive, anxiety and motor behaviours. In total, three batches of Chd8 neo/neo mice were assessed. As no batch effects and no sexually dimorphic behaviours were observed, data from all three experiments and both sexes were combined for analysis. Chd8 hypomorphs displayed normal social approach and investigation behaviours in the three-chamber (Fig. 2a) and reciprocal social investigation tests (Fig. 2b). Relative to wildtype littermates, hypomorph pups emitted normal numbers of ultrasonic vocalisations (USVs) on separation from the nest, suggesting no deficits in socio-communicative abilities (Fig. 2c). Furthermore, Chd8 neo/neo mice demonstrated a normal capacity for habituation and dishabituation to both social (sex-matched urine) and non-social (banana) odours (Fig. 2d). We did not observe any differences in repetitive self-grooming behaviour between hypomorphs and wildtypes (Fig. 2e). In the marble burying test, a task also designed to measure repetitive behaviours in mice, Chd8 neo/neo mice buried fewer marbles than controls (Fig. 2f ). Interestingly, Chd8 hypomorphs showed increased aversion towards the centre of the open field arena (Fig. 2g, h), suggesting that reduced Chd8 expression may predispose to anxiety. In addition, these mice demonstrated decreased locomotor activity in the open field arena, even when controlling for their anxiety phenotype (Fig. 2i). This hypo-active phenotype was corroborated by homecage running wheel tests (Fig. 2j) and may account for the reduced marble burying observed in these mice. However, the hypo-activity did not appear to be due to motor defects, as these mice showed motor performance comparable to controls in the rotarod test (Fig. 2k).
In summary, Chd8 neo/neo mice showed no evidence of socio-communicative deficits or repetitive behaviours typically associated with ASD, suggesting that the reason for a lack of robust ASD-like behaviours in Chd8deficient mouse models is unlikely to be an insufficient reduction in Chd8 expression. The most robust, reproducible phenotype was locomotor hypoactivity, which was present in both our mild hypomorphic and Chd8 +/− mice [18], and also Chd8 +/− mice reported by Platt et al. [16]. Intriguingly, whereas our Chd8 +/− mice did not show clear anxiety phenotypes [18], the Chd8 hypomorphs did, suggesting that this particular phenotype may be enhanced by a reduction in Chd8 levels.

Gradual reductions in CHD8 levels result in progressively more pronounced gene expression changes
To understand how sub-heterozygous Chd8 levels affect gene expression in the embryonic neocortex, RNA-sequencing (RNA-seq) was performed from E12.5 neocortices. Data from heterozygous (Chd8 +/− ), mild (Chd8 neo/neo ) and severe (Chd8 neo/− ) hypomorphs, together with their respective wildtype littermate controls were included for analysis.
Comparing DEGs in Chd8 neo/neo and Chd8 neo/− samples identified 1389 genes common to both datasets (Fig. 3f ), all of which were changed in the same direction. ASD-associated genes were highly enriched in the DEGs from both Chd8 neo/neo (145 genes, p = 1.32 × 10 −9 , OR = 1.83, Fisher's exact test for count data, Fig. 3f, Additional file 3: Table S2) and Chd8 neo/− embryos (152 genes, p = 5.532 × 10 −7 , OR = 1.62, Fisher's exact test for count data, Additional file 3: Table 2). Nearly half (46%) of these ASD-associated genes were common to Chd8 neo/ Fig. 2 Behavioural assessment of mild Chd8 hypomorphic mice. a-k Behavioural assessments of a cohort of adult Chd8 neo/neo (neo/neo, n = 24; 14 males, 10 females, a, b, d, e, g, i or n = 15; 10 males, 5 females, f, k or n = 11; 6 males, 5 females, j) and Chd8 +/+ (+/+, n = 41; 26 males, 15 females, a, b, d, e, g, i or n = 31, 21 males, 10 females, f, k or n = 12; 6 males, 6 females, j) and pup Chd8 neo/neo (neo/neo, n = 10; 4 males, 6 females, c) and Chd8 +/+ (+/+, n = 12; 6 males, 6 females, c) mice. Chd8 +/+ animals are illustrated using filled circles, dotted lines with grey bars and Chd8 neo/neo animals with open triangles, solid lines and open bars. Apart from USVs (c), all other tests were conducted on young adults 6-14 weeks of age. a Duration, in seconds, spent in each chamber of the three-chamber sociability test. All mice spent a significantly higher proportion of time in the chamber with the age-and sex-matched stranger con-specific mouse compared to the other chambers. Mean ± SEM; *** < 0.001 (between-subjects ANOVA with student's t test as post-hoc analysis). b Duration, in seconds, of social investigation over a three-minute period. Social investigation was defined as the combined total duration of head, body and anogenital sniffing of a conspecific mouse. Mean ± SEM (between-subjects ANOVA). c The mean number of ultrasonic vocalisations per minute on indicated postnatal days. Mean ± SEM (repeated-measures ANOVA). d Graph demonstrating the performance in the olfactory habituation/dishabituation test. Mean ± SEM (repeated-measures ANOVA). e The duration, in seconds, mice spent self-grooming during the 10-min self-grooming test. Mean ± SEM (between-subjects ANOVA). f The average number of marbles buried, out of a maximum of 12, within a 30-min period period. Mean ± SEM; * < 0.05, ** < 0.01 (repeated-measures ANOVA with student's t test as post-hoc analysis). g Representative ethovision tracks of a Chd8 neo/neo (neo/neo) and Chd8 +/+ (+ / +) animal plotting their movements during the 10-min open field task. h The percentage of time spent in the centre of the open field arena during the 10-min test. Mean ± SEM; **p < 0.01 (between-subjects ANOVA). i The total distance travelled in the outer part of the open field arena over a 10-min time-period. Mean ± SEM; *** < 0.001 (between-subjects ANOVA). j The total activity counts per 12 h period on running wheels in the homecage during 7 days of dark-phase recording. Mean ± SEM; *p < 0.05 (repeated-measures ANOVA). k The mean latency of mice to fall from the rotarod. Mean ± SEM (repeated-measures ANOVA) neo and Chd8 neo/− mice (Fig. 3f ). The majority (89% and 88%, respectively) of ASD-associated DEGs were downregulated (orange and red fractions in Fig. 3g, Additional file 3: Table S2), supporting the idea of CHD8 as an important positive regulator of neurodevelopmental, autism-associated genes [13,14].
To determine if CHD8 functions as a positive or negative regulator of gene expression, we identified genes most likely directly regulated by CHD8, using ChIP-seq data from Cotney et al. [13]. From the Cotney dataset, a consensus set of gene promoters with CHD8 recruitment in E17.5 embryonic mouse brain, human midfetal cortex, and neural progenitor contexts was defined. This gene list was compared to the list of differentially expressed genes in Chd8 hypomorphs. This analysis revealed that the majority of DEGs were not identified in the consensus CHD8 ChIP-seq dataset (blue and orange fractions in Fig. 3g), suggesting that most transcriptional changes are induced by indirect effects. This is consistent with previous observations and the prevailing view that CHD8 regulates the expression of many other chromatin and epigenetic modifiers [13,14]. DEGs with CHD8 localisation to their promoters in the Cotney data (red and purple fractions in Fig. 3g), were present in both upand down regulated gene sets, suggesting that CHD8 can function as both an activator and repressor of these genes during cortical development.
To identify potential transcriptional co-regulators and DNA-binding factors that may cooperate with CHD8 during embryonic cortical development, Gene Set Enrichment Analysis was performed using the "ENCODE and ChEA Consensus TFs from ChIP-X" database in Enrichr [36]. This analysis revealed an over-representation of E2F (E2F4, E2F6 and E2F1) targets in the upregulated genes (Additional file 1: Fig. S4, Additional file 3: Table S2). E2Fs compose a family of transcription factors with important roles in DNA replication, cell cycle progression and proliferation. CHD8 has been previously shown to be involved in E2F-dependent transcriptional activation, and is necessary for recruitment of the "activator" E2F transcription factors E2F1 and E2F3 to G1/S transition promoters [37]. Our findings suggest that CHD8 functions as a repressor of E2F-regulated genes in the developing cortex and implicate increased progenitor proliferation as a potential mechanism for the brain hyperplasia in these mice.
For downregulated genes, an over-representation of targets of REST (RE1-Silencing Transcription factor) and the Polycomb component Suz12 was seen (Additional file 1: Fig. S4, Additional file 3: Table S2). As Suz12 is essential for the activity and stability of the PRC2 complex, we asked if any of the DEGs are marked by the PRC2-repressive modification H3K27me3 in normal neural progenitor cells [38]. The majority of DEGs that were marked by H3K27me3 in neural progenitors were downregulated in hypomorphic mice (pink fractions in Fig. 3h). Thus, some genes positively regulated by CHD8 are Polycomb targets in neural progenitors. Although further mechanistic studies will be required to test this, these findings are consistent with CHD8 functioning as an antagonist of Polycomb repression as expected of a Trithorax family member. REST is a master regulator of neurodevelopment, has been shown to directly interact with CHD8, and is abnormally activated in Chd8 haploinsufficient mouse brain [15]. Overlapping DEGs with REST ChIP-seq data [39], we found REST target genes predominantly amongst downregulated genes (orange and green fractions in Fig. 3h), supporting the notion that aberrant REST activation in Chd8-deficient embryonic brain may contribute to gene repression. Furthermore, 40% of the downregulated REST target genes are also marked by H3K27me3 in neural progenitor cells (green fractions in Fig. 3h), implying roles for both REST  Table S3 and Polycomb in the repression of these genes in Chd8 hypomorphs.
To provide further insights into the biological processes affected by these gene expression changes, gene ontology analyses were performed. These identified a significant enrichment of cell cycle, DNA replication and repair genes in the upregulated genes in hypomorphs (Fig. 3i, Additional file 4: Table S3). Neurodevelopmental gene categories were enriched in the downregulated gene sets (Fig. 3i, Additional file 4: Table S3).

Chd8 deficiency increases proliferation of cortical progenitors outside of the ventricular zone
To explore whether the increased expression of cell cycle and DNA replication genes in hypomorphs is indeed associated with increased progenitor proliferation, we immunolabelled progenitors in the G2/M phase of mitosis in coronal brain sections with an antibody against phosphohistone-3B (PH3B). No difference in the number of mitotic progenitors was observed in E12.5 Chd8 +/− or Chd8 neo/neo neocortices, compared to wildtype littermates (Additional file 1: Fig. S5A-C). To determine if differences may arise later, we performed the same experiment at E14.5 (Additional file 1: Fig. S5D, E). Again, no significant differences were seen in the numbers of proliferating ventricular progenitors (Additional file 1: Fig. S5F). Intriguingly, a significant increase in the proliferation of non-ventricular (or basal) progenitors was observed in Chd8 neo/neo neocortices (Additional file 1: Fig. S5G). No significant differences were observed in Chd8 +/− or Chd8 neo/− embryos Additional file 1: Fig. S5F, G), suggesting that the abnormal expansion of basal cortical progenitors only occurred within a restricted window of reduced CHD8 expression.
To determine the identity of the non-ventricular progenitors that proliferated more in mild hypomorphs, we stained serial sections with antibodies to TBR2, a marker for intermediate progenitors, and PH3B (Fig. 4a). Counting these proliferating TBR2+ progenitors revealed a significant increase in the proliferation of these cells in Chd8 neo/neo embryos, compared to controls (Fig. 4b). No significant changes in the number of proliferating TBR2+ progenitors was observed in other genotypes (Fig. 4b).
Taken together, these experiments showed that TBR2+ intermediate progenitors accounted for the increased non-ventricular proliferation observed in Chd8 neo/neo embryos.
Our data suggests that TBR2+ intermediate progenitors might be particularly sensitive to reduced Chd8 expression. To find if Chd8 deficiency directly impacts gene regulation in these cells, we sorted TBR2+ and PAX6+ progenitors from E14.5 Chd8 +/+ , Chd8 +/− and Chd8 neo/neo neocortices by FACS and compared gene expression by qRT-PCR. We selected two genes identified in our RNA-seq analysis that are upregulated in Chd8 neo/neo mice, Pim1, a gene functionally linked to progenitor cell proliferation [40] and Axin2, a direct transcriptional target and feedback antagonists of WNT-β-catenin signalling [41]. Both Pim1 and Axin2 were significantly upregulated in Chd8-deficient TBR2+, but not PAX6+ progenitors, with Axin2 showing a clear Chd8 dosage-sensitive effect (Additional file 1: Fig. S6B, C). Together, these findings show that Chd8 deficiency impacts the expression of specific genes linked to neural progenitor proliferation and fate specifically in TBR2+ intermediate progenitors. The gene expression changes in Chd8 +/− embryos may not be sufficient to increase the proliferation of these progenitors, or may have consequences at later stages of development.
Next, we sought to identify the potential causes of cortical hypoplasia in severe hypomorphs (Additional file 1: Fig. S2E, G). We did not detect any reduction in cortical progenitor proliferation at E14.5 in these mutants (Fig. 4b). Comparing gene ontology analyses between mild and severe hypopmorphs (Additional file 1: Figs. S7 and S8; Additional file 4: Table S3), revealed a slight increase in the number of p53-regulated genes and ribosomal genes like Rpl26 that can augment p53 mRNA translation [42] in severe hypomorphs, raising the possibility that progenitors may be more prone to apoptosis in these mice. Indeed, cleaved caspase 3 (CC3) immunostaining revealed a small increase in the numbers of apoptotic cells in Chd8 neo/− embryos compared to Chd8 neo/neo and control embryos (Fig. 4c-f ), suggesting that increased cell death may be responsible for the brain hypoplasia in the severe hypomorphs. Thus, reducing CHD8 levels below 15% of wildtype, appeared to reveal another threshold below which p53-regulated genes become derepressed with an apparent impact on cell survival.  5 Conditional deletion of Chd8 from the embryonic neuroepithelium results in severe hypoplasia of the telencephalon and neocortex. a Schematic representation of the wildtype (wt) mouse Chd8 gene (Chd8 wt ), targeting construct for homologous recombination in embryonic stem cells, the Chd8 targeted allele (Chd8 neo ), the Chd8 conditional allele after Flp-mediated excision of the neomycin resistance cassette (Chd8 flox ) and the Chd8 null allele produced by Cre-mediated deletion of exon 3. Boxes represent exons, with exon 1 (Ex1) to 6 and 13 shown and exon 3 filled in black. The blue box represents a neomycin resistance cassette (neo), red triangles represent loxP sites and tan triangles frt sites. The long 9.45 kb (5′) homology arm is indicated in red and the short 4.4 kb (3′) homology arm in green in the targeting construct. b X-gal staining of a Sox1-Cre; R26R embryo at E9.5 (left); and immunostaining for CHD8 protein on Chd8 flox/flox (Ctrl) and conditional knockout Sox1-Cre; Chd8 flox/flox (cKO) E10.5 neural tube (right). Scale bar = 50 μm. c, d Wholemount images of E18.5 brains of a representative Ctrl and cKO embryo, anterior is to the right. Scale bar = 1 mm. e, f Wholemount images of embryonic day 12.5 heads, anterior to the right. Scale bar = 1 mm. g, h Wholemount images of E10.5 embryos, anterior to the right. c′-h′) Cresyl violet-stained frontal sections through brains as indicated in c-h above. Scale bars = 500 μm (c′, d′) and 200 μm (e′-h′). The following subcortical structures are labelled in Ctrl (c′) and cKO (d′) at E18.5: NCtx: Neocortex, mhN: medial habenular nucleus, vmt: ventral medial thalamic nucleus, amN: amygdaloid nucleus. Images are representative of at least 3 independent samples CHD8 expression is essential for repression of p53 target genes in early embryonic neocortex As the brain phenotype and effects on p53-regulated genes and apoptosis in severe hypomorphs were subtle, we decided to explore the consequences of complete CHD8 loss (0%), by conditional deletion of Chd8 in early neural progenitors. Sox1-cre-mediated deletion (Fig. 5a, b, Additional file 1: Fig. S9) of loxP-flanked (flox) exon 3 results in an early frameshift and termination of translation at amino acid 419, which is predicted to produce a protein that lacks all functional domains and results in a Chd8-null allele as shown previously in Chd8 +/− mice [18].
The pan-neuronal, conditional deletion of Chd8 by Sox1-cre (Fig. 5b, Additional file 1: Fig. S9) resulted in pronounced brain hypoplasia in homozygous conditional knockout (cKO) embryos, compared to controls (Ctrl, Chd8 flox/flox (Chd8 f/f )) ( Fig. 5c, d). Neocortical hypoplasia, with the maintenance of some subcortical brain structures was evident upon histological analysis of E18.5 embryos (Fig. 5c′, d′). To identify the origin of these defects, cKO embryos were examined at earlier stages of development. Telencephalic hypoplasia with markedly thinner neuroepithelium was evident in E12.5 cKO embryos when compared to controls (Fig. 5e, f′). Examination of E10.5 cKO embryos showed telencephalic vesicles of near-normal size with neuroepithelia that were slightly thinner than controls (Fig. 5g-h′), suggesting that CHD8 is essential for expansion of the pallium from early embryonic development.
To identify the potential causes of this striking phenotype, we performed RNA-seq at the onset of the phenotype. RNA-seq analysis identified 2032 DEGs in E10.5 cKO telencephalic vesicles compared to controls (Fig. 6a, Additional file 5: Table S4). KEGG pathway mapping of all dysregulated DEGs identified the p53 pathway as the most significantly affected (Fig. 6b, Additional file 1: Fig. S10). Interestingly, GO analysis identified cell cycle as the most dysregulated bioprocess (Additional file 5: Table S4), with a slight majority of genes within this category downregulated (60 out of 111). Quantitative RT-PCR (qRT-PCR) confirmed significant upregulation of multiple p53-regulated genes (Fig. 6c). Furthermore, genes normally upstream of p53, Atr and Atm, and Trp53 (the gene encoding p53 itself ) were not affected by Chd8 deletion (Fig. 6c), consistent with a role for CHD8 in directly repressing p53 target genes [10].
To ask if increased p53 pathway activity was responsible for the cKO phenotype, we reduced p53 gene expression in the neuronal lineage to test if this can reduce the severity or incidence of phenotypic abnormalities. Neocortical hypoplasia was partially rescued in Sox1-Cre; Chd8 f/f ; Trp53 f/+ (conditional knockout p53 heterozygous, cKO-p53het) embryos (Fig. 6d), providing strong genetic evidence that this phenotype was caused by elevated p53 signaling. A substantial increase in apoptosis was observed in the cKO embryos (Fig. 6e, f, h, i). The number of apoptotic cells was significantly reduced in the pallium of cKO-p53het embryos, compared to cKO embryos (Fig. 6f-i). We also noted the presence of certain cell cycle inhibitors amongst genes upregulated in the cKO (e.g. p21/CDKN1A) (Fig. 6c). Therefore, we investigated neural progenitor proliferation. Quantification of PH3B+ cells in the ventricular zone of the neocortex confirmed a strong reduction in cell proliferation in cKO embryos (Fig. 6j, j′, k, k′, m). Cell proliferation was slightly increased in cKO-p53het embryos compared to cKO embryos (Fig. 6l, l′, m), consistent with only a partial rescue of neocortical size in these animals (Fig. 6d). The visualisation of proliferating BrdU+ cells in the same sections appear to support this conclusion (Fig. 6j-l′).
Taken together, these data identify CHD8 as an essential repressor of p53 pathway activation during neocortical development. CHD8 loss leads to increased apoptosis, reduced neural progenitor proliferation and precocious cell differentiation during early embryonic development, resulting in severe neocortical hypoplasia by the end of gestation.

Discussion
Human genetic studies have identified heterozygous, likely gene disrupting mutations in CHD8 as a possible cause of ASD and macrocephaly. Chd8 heterozygous mice have been generated by several groups, but these mice were found to exhibit relatively subtle brain overgrowth [15][16][17]19]. Observations of relatively small transcriptional changes in the mid-gestation Chd8 +/− mouse brain appeared at odds with the many genes dysregulated upon Chd8 knockdown in progenitor cell lines and after in utero electroporation [13,14,20]. Together, these studies led us to explore whether different sensitivities to reduced CHD8 dosage might account for some of these inconsistencies. A comparison of brain size, gene expression and neural progenitor fate in a mouse Chd8 allelic series yielded several key findings: (1) A small additional reduction in Chd8 expression in mild hypomorphs compromised the capacity of neural progenitor cells to maintain stable expression of over 2000 genes in the mid-embryonic neocortex, which included over 140 ASD-associated genes; (2) CHD8 has a key role in limiting the expansion of TBR2+ intermediate progenitors, a population particularly important for human cortical development; (3) A gradual reduction in CHD8 expression can have non-linear effects on gene expression and brain growth (Fig. 7). In addition to the precipitous gene expression changes in mild hypomorphs, we observed brain hyperplasia in Chd8 +/− and Chd8 neo/neo mice, but brain hypoplasia in severe hypomorphs and conditional knockouts (Fig. 7). Together, these findings indicate that CHD8 levels need to be tightly regulated during development and that the interpretation of experimental manipulations that involve Chd8 knock-down should consider these non-linear, threshold effects.

CHD8 as a phenotypic capacitor
It has been posited previously that ASD-associated chromatin remodelling factors may act as phenotypic capacitors, buffering against perturbations to normal development in order to maintain stable phenotypes [44]. Heterozygosity for a capacitor is predicted to result in a loss of robustness, such that the cells are more susceptible to additional genetic and non-genetic risk factors.
Our findings that over 2200 genes, many of which are ASD risk factors, became dysregulated by a small additional decrease in CHD8 dosage below 50%, supports the idea that some neurodevelopmental genes and processes may be close to a critical threshold in the Chd8 heterozygous neocortex and therefore easily perturbed by small additional changes. It will be important to determine if specific ASD-associated CHD8 mutations in humans reduce CHD8 function by more than 50% by dominant negative mechanisms, as our findings would predict these mutations to be significantly more pathogenic than purely haploinsufficient mutations. It is also important to consider that the C57BL/6 genetic background has been used for all Chd8 heterozygous mouse studies so far. This background may be protective and more robust in the context of Chd8 haploinsufficiency, such that different, or more severe phenotypes may emerge on different genetic backgrounds. We report a significant effect of sub-haploinsufficient levels of Chd8 expression on the proliferation of intermediate progenitors in the E14.5 embryonic cortex. It will be of interest to determine if other cell types relevant to CHD8 function in the brain, like oligodendrocytes [45], exhibit similar Chd8 gene dosage sensitive responses.

CHD8 regulates the proliferation of non-ventricular cortical progenitors
Our findings raise the possibility that brain development in human and mouse differ in sensitivity to a similar reduction in CHD8 function, such that CHD8 heterozygosity may cause more pronounced changes to brain growth and transcriptional regulation in the developing human brain. In this regard, fundamental differences in mouse and human brain development may result in CHD8 haploinsufficiency having more pronounced effects on human brain development. Comparative studies of gyrencephalic and lissencephalic animals have identified important differences in the capacity of nonventricular progenitors to expand and subsequently contribute to cortical expansion. This population of progenitors consists of outer radial glia cells (oRGs), and TBR2-expressing intermediate progenitor cells (IPs) [46]. In humans, oRGs are located in an expanded outersubventricular zone (oSVZ) and are capable of asymmetric divisions that generate an oRG daughter cell, which maintains the pool of non-ventricular progenitors, and an IP daughter cell that can undergo transit-amplifying divisions to expand and generate neuronal progeny [47]. By contrast, mouse oRGs primarily undergo self-renewing, neurogenic divisions and populate a non-ventricular region lacking the distinct, expanded cytoarchitecture of the oSVZ typically seen in gyrencephalic species [48]. Furthermore, mouse IPs likely possess a more limited capacity for self-renewal, as a majority of mouse IP divisions appear to generate two neuronal daughter cells [49,50]. Therefore, if CHD8 has an especially crucial role in regulating the expansion of TBR2+ progenitors also in the human brain, then it is possible that CHD8 deficiency in these cells could result in more pronounced phenotypes with regard to cortical over-growth and circuit disruption in humans. The lack of exacerbated ASD-linked behavioural phenotypes in Chd8 neo/neo mice is consistent with this notion.
Interestingly, we also note that Bernier et al. previously identified an enrichment for CHD8 expression in areas outside the ventricular zone in human mid-fetal cortex [6], further supporting the idea that CHD8 may have an important role in regulating expansion of these cells. An analysis of neural progenitor subsets in CHD8-deficient human brain organoids should be a viable way to test this hypothesis.

CHD8 is an essential repressor of p53 in neural progenitors
One of the most striking non-monotonic effects of Chd8 efficiency reported here is brain hypoplasia in pan-neuronal Chd8 cKO mice, partly as a result of de-repression of the p53 pathway in early neuroepithelial cells. This discovery identifies CHD8 as a critical repressor of p53 target gene activation in neural progenitors. Our findings Fig. 7 The non-monotonic relationship between CHD8 protein levels, gene expression and brain size. The effects of gradual reductions in CHD8 protein levels to ~ 50% (heterozygous), ~ 35% (mild hypomorph), ~ 10% (severe hypomorph) and 0% (conditional knockout) on the transcription of neurodevelopmental, cell cycle and p53-regulated genes and brain size are depicted. CHD8 appears to function primarily as a positive regulator of neurodevelopmental genes via recruitment to H3K4me3-modified (yellow ball) histones (gray spool), presumably via enabling the recruitment of key transcription factors (TF). A sharp reduction in the expression of many of these genes (arrow) is only observed in E12.5 neocortex when CHD8 levels are reduced to below a threshold less than haploinsufficient levels. CHD8 appears to repress E2F-regulated cell cycle genes in this context, with significant induction only becoming evident at sub-haploinsufficient levels, although low expression increases (grey arrow) likely drives subtle increases in proliferation in the heterozygous state. Cell cycle genes are dysregulated in the opposite direction in the cKO, suggestive of non-monotonic effects (blue arrow). CHD8 can interact with p53 and histone 1 (H1), leading to stable heterochromatin formation and repression of p53 target genes. A few p53-regulated genes become activated in hypomorphic mice (grey arrow), but the majority remains fully repressed with de-repression only becoming evident upon complete CHD8 loss. Note the different CHD8 thresholds for different groups of genes (broken lines) and the non-monotonic effects on gene expression and over-all brain size suggest that very low levels of CHD8 are sufficient to repress p53 target genes and maintain neural progenitor self-renewal. One could speculate that the CHD8dependent recruitment of histone H1 to p53 target genes to initiate a cooperative process of chromatin compaction [10], may require lower levels of CHD8 than another process that is dependent upon the constitutive recruitment of RNA polymerase [51] or other co-activating factors by CHD8 (Fig. 7). Our gene expression and apoptosis data suggest that CHD8 protein levels in Chd8 neo/− embryos were close to this critical threshold. Interestingly, Cotney et al. also reported p53 signaling as one the most dysregulated pathways upon CHD8 knock-down in human neural stem cells [13]. However, other studies in human cell lines have not demonstrated the same changes, including an in vitro knock-down of CHD8 to 20-25% of control levels in human SK-N-SH neural progenitor cells [52]. Intriguingly, CHD8, and its family member CHD7, also maintains the survival of oligodendrocyte precursors by inhibiting the p53 pathway, although this appears to be primarily mediated via direct repression of the p53 gene [53]. Together, these findings suggest that transcriptional responses to reduced CHD8 levels are highly contextdependent and may help shed light on reports that Chd8 knock-down in utero led to reduced proliferation and enhanced differentiation of neural progenitors [20].

Behavioural phenotypes in Chd8-deficient mice
A number of Chd8 +/− mouse models have been reported. Given the strong association between CHD8 mutations and ASD in humans, one might have expected these mice to exhibit robust, ASD-like behaviours. However, that is not the case, a finding not that surprising if one takes into account the pronounced differences in brain size and behaviour between these species. Nevertheless, we tested whether mice with more pronounced brain and gene expression phenotypes exhibit more pronounced behavioural phenotypes. With the possible exception of anxiety, the behavioural phenotypes of Chd8 +/− and hypomorphic mice were remarkably similar, including robust, reproducible hypo-activity [18].

Limitations
As with all other Chd8 mouse studies so far, our models were on a C57BL/6 genetic background. Different, or more severe phenotypes may emerge on different genetic backgrounds. Given the significant postnatal lethality of Chd8 hypomorphic mice, we cannot exclude the possibility that the reason for not observing a particularly prominent exacerbation of behavioural phenotypes compared to heterozygous mice, may be that only mildly affected, surviving mice could be behaviourally phenotyped. Furthermore, unlike Chd8 +/− mice, all Chd8 neo/neo mice were born to Chd8 neo/+ mothers, and the possibility that differences in maternal care could account for subtle behavioural differences cannot be ruled out. Recent work have reported sexually dimorphic effects in another Chd8 model [19]. We did not observe significant behavioural differences between males and females in our studies, and the structural, molecular and developmental experiments reported here were performed with mixed sex samples. We therefore cannot rule out sex-specific effects in gene expression and perhaps even developmental phenotypes. Our findings identify an important role for the autism-associated factor CHD8 in controlling the proliferation of intermediate progenitors in the mouse neocortex. Our analysis of TBR2+ intermediate progenitors show a robust increase in the proliferation of these cells in Chd8 hypomorphs, but not Chd8 +/− mice (Fig. 4), despite the latter showing significant alterations in gene expression (Additional file 1: Fig. S6). We cannot at this stage rule out the possibility that the dynamics of neural progenitor proliferation and differentiation differ in heterozygous and hypomorphic mice, and that intermediate progenitors over-proliferate also in Chd8 +/− mice, but at a different developmental stage. A comprehensive, timed analysis of TBR2+ progenitor fate will be required to test this possibility. We propose that CHD8 also regulates intermediate progenitor proliferation in human brain development, as indeed suggested by CHD8 expression studies [6], but studies on human cells are required to confirm this. Finally, we show that CHD8 represses p53-regulated genes and that p53 pathway hyperactivation in neuroepithelial cells that lack CHD8 is at least in part responsible for cellular phenotypes that can lead to cortical hypoplasia. These findings do not exclude the contribution of other pathways and cellular mechanisms to the hypoplasia phenotype.

Conclusion
In conclusion, our analysis of an allelic series of Chd8deficient mice has identified clear non-monotonic effects on gene expression and brain growth (Fig. 7). Recognition of the differing sensitivities of important cellular processes to CHD8 dosage and how small differences in CHD8 levels may lead to disproportionally large differences in phenotype is an important step in understanding the context-specific transcriptional roles of CHD8 in brain development.
Additional file 1: Supplementary information.