Persistent cortical excitatory neuron dysregulation in adult Chd8 haploinsufficient mice

Abstract

CHD8 mutations cause autism spectrum disorder, cognitive deficits, and macrocephaly. Chd8 ^+/- mouse models exhibit macrocephaly and transcriptional pathology, with inconsistent findings regarding neurogenesis, neuron function, and behavior. Via stereology and single nucleus transcriptomics (snRNA-seq), we found increased Chd8 ^+/- cortical volume was not explained by increase in neuron number. Differential expression (DE) was present across cortical cell types, with excitatory neurons exhibiting high DE burden and shared and subclass-specific DE signatures. Bulk RNA-seq DE of constitutive Chd8 ^+/- and conditional Camk2a-Cre Chd8 ^+/- mice identified shared transcriptional pathology. DE in synaptosomal versus nuclear mRNA identified overlapping DEGs, but also significant differences and exaggerated synaptosomal changes. Building on DE findings implicating glutamatergic neurons, we found Chd8 ^+/- mice exhibited altered excitatory neuron spine density and dynamics, decreased GCaMP activity correlation, and sleep perturbation. Thus, Chd8 haploinsufficiency causes lasting excitatory neuron dysfunction, perturbs RNA regulation beyond transcription, and impacts neuronal properties, cortical microcircuits, and behavior.

Figure 1.. Increased cortical volume in Chd8^+/− mice without a corresponding increase in neuron number.

(A) Schematic overview of study investigating the role of Chd8 haploinsufficiency in mature mouse cortex. (B–C) Representative coronal sections across the rostro-caudal extent of the cerebral cortex used for unbiased stereological analysis. (B) Cortical boundaries as determined for volume measurements. (C) Cellular organization visualized using Nissl staining (top) and NeuN immunostaining (bottom) in Chd8^+/+ and Chd8^+/− mice. (D–E) Stereological profiles across the Z-plane showing (D) cortical volume estimates based on the Cavalieri method and (E) total neuron number as assessed by unbiased stereology in the same rostro-caudal regions. Data are stratified by sex and genotype. Error bars represent SEM. Statistical comparisons were performed using two-way ANOVA (Cortical volume: P=0.032 unstratified, P=0.009 males; P=0.61 females; Neuron Count: P=0.76, unstratified; P=0.37, males; P=0.30, females Student’s t-test, N = 8 per sex/genotype).

Figure 2:. Cell type-specific pathology in adult Chd8^+/− mouse cortex.

A) Schematic depicting experimental design for RNA-seq and snRNA-seq experiments. B) UMAP of 23,920 cortical nuclei from P60 Chd8^+/− and WT littermates with cell types annotated. C) Clustering of Chd8^+/− (HET) and WT nuclei (top) and male and female nuclei (bottom) in UMAP space. D) Cell type identification dot plot representing average expression of cell type-specific marker genes in each cell type. Tree (bottom) shows 3 levels of cell type identification (Cell_Type_1–3). E) Cell type proportion comparison between Chd8^+/− (HET) and WT nuclei, revealing a significant decrease in mutant L6 IT Glut nuclei (*p = 0.027) and increase in mutant OPCs (*p = 0.048) (Student’s t-test). F) DEG burden across cell types (Burden = DE genes / expressed genes * 100). Significant DE burden was found in L2/3 IT (*p = 0.00001), L4/5 IT (*p = 0.00001), L5 IT (*p = 0.00001), L5 ET (*p = 0.0045), L6 CT Glut (*p = 0.0004), and VLMC (*p = 0.0009) nuclei (permutation test, n= 10,000). G) Heatmap showing −log10(pval) of top 25 unique upregulated (top) and downregulated (bottom) bulk and pseudobulk DEGs in each cell type (p < 0.1).

Figure 3:. hdWGCNA reveals differentially expressed glutamatergic cell coexpression modules.

A) 10 glutamatergic modules identified via hdWGCNA B) Number of genes per module. C) Relative expression of glutamatergic modules across glutamatergic neuron subtypes (module score = average expression of all module genes compared to expression-matched background genes; DEMs identified via comparison of median pseudo-bulk logFC of module genes to that of the grey module via t-test; DEMs (*p < 0.05) in at least one population are labeled in yellow. D) DEG burden (calculated as in Fig. 2F) of each DEM in each cell type E) Heatmap of relative median pseudo-bulk logFC of module genes for glutamatergic cell types scaled by p-value. White and black boxes in C-E indicate the cell types in which each corresponding module was DE. F) Representative significant GO terms for Biological Processes enriched in glutamatergic modules. (*p < 0.05, **p < 0.01, ***p < 0.001). G) Box plots for 4 representative DEMs showing logFCs of module genes in each glutamatergic subtype. Non-opaque boxes indicate cell types where the module shown was significantly differentially expressed (P-value < 0.05). H) Feature plots (left) showing relative expression of representative genes from DEMs of interest in UMAP space. Dot plots (right) show median sample-level expression of the same genes. Black boxes indicate cell types where the corresponding module is differentially expressed, or where the direction of change in module expression is consistent across batches.

Figure 4:. Shared DE in constitutive Chd8^+/− and Camk2a-CRE conditional Chd8^+/flox mice, with divergent impacts on nuclear and synaptosomal mRNA and protein.

A) Barplot of DEG counts at P < 0.05 and FDR < 0.1 across bulk RNA-seq experiments (constitutive Chd8^+/−, Camk2a-CRE conditional Chd8^+/flox, and nuclear and synaptosomal preparations from the Chd8^+/− line). B) Chd8 mRNA is down-regulated in all but the nuclear preparation (P-values from EdgeR). C-E) Concordance plots comparing bulk RNA-seq logFC of DEGs detected in either dataset. Number of DEGs at different criteria listed above, R value for best fit line for FDR < 0.1 is listed on the plot. For P < 0.05, local density is plotted instead of individual data points due to large number of genes. For FDR < 0.1, individual datapoints are shown. The count of P < 0.05 DEGs in each quadrant listed in red. Higher data points in the upper right and lower left indicate logFC directional concordance between experiments. C) Concordance between constitutive and conditional datasets. D) Concordance for nuclear (top) and synaptosomal (bottom) versus whole cortex. E) Reduced concordance between nuclear and synapsomal datasets. F-G) Selected enriched GO terms for upregulated F) and downregulated G) genes across experiments. H) Downregulation of Dlg2 in synaptosomal RNA. I-J) Western blot (I) validating decreased synaptosomal PSD93 protein (J) in Chd8^+/− cortex (N = 8 Chd8^+/+, 10 Chd8^+/−, P-value from Student’s t-test).

Figure 5:. Delayed spinogenesis during Chd8^+/− development.

A) Dendritic segment imaged over 7-days in somatosensory cortex of adolescent WT and Chd8^+/− mice. Arrowheads: eliminated spines; arrows: formed spines; asterisks: filopodia. Scale bar = 2 μm. B, D) Spine density in adolescent (b) and adult (D) Chd8^+/− or WT littermates. C, E) The percentage of spines formed or eliminated over 7d is higher in adolescent (D) and adult (E) Chd8^+/−than WT littermates, mean, SEM., unpaired t-test. **p<0.01.

Figure 6:. Disruption of coordinated activity in Chd8^+/− PFC microcircuits.

A. Mice injected with Syn-GCamp6f and activity recorded from acute slices after 4–6 weeks. B, wide field and GFP at 4x (left). 10X of individual neurons and z-projection of standard deviation across frames (right). C, z-scored fluorescence traces from neurons in a single slice. D, Raster of events. E, Cumulative probability function showing overlap in fraction of frames active for neurons from WT (blue) and Chd8^+/− (red) slices. F, Activity of each cell as a function of depth from cortical surface. Cells binned by distance from surface (gray boxes) and mean correlations within and between bins calculated. G, H, 8×8 matrix of mean correlation coefficient calculated for cells within each pair of bins, averaged for WT (G) or Chd8^+/− (H) slices. I, cumulative distribution function (CDF) of mean correlation across 544 pooled WT (blue) neurons and 818 pooled Chd8^+/− (red) neurons for all bins. J, Mean correlation coefficient of WT and Chd8^+/− slices (p < 0.05, t-test). K, CDF of the mean value of the subset of bins in which neuron 1 was < 300 microns from cortical surface. L, Mean correlation coefficient of superficial neurons in WT and HET slices (p < 0.01, t-test).

Figure 7:. Perturbed sleep patterns in Chd8^+/− mice.

A) Schematic of 24hr EEG/EMG sleep analysis. B) Example EEG/EMG traces showing activity during REM, NREM, and Wake state. B’) Spectrogram of the EEG signal in B C-D) Summary of time spent in each state during each hour of the 24hr monitoring period for wild type (C) and Chd8^+/− (D) groups. E) Summary of mean total time spent in each state separated by light and dark stages. F) Summary of SD of time spent in each state. Mean and SD shown, significance form unpaired t-test with Welch’s correction. See text for details.