Transcriptome analysis reveals dysregulation of innate immune response genes and neuronal activity-dependent genes in autism

Abstract

Recent studies of genomic variation associated with autism have suggested the existence of extreme heterogeneity. Large-scale transcriptomics should complement these results to identify core molecular pathways underlying autism. Here we report results from a large-scale RNA sequencing effort, utilizing region-matched autism and control brains to identify neuronal and microglial genes robustly dysregulated in autism cortical brain. Remarkably, we note that a gene expression module corresponding to M2-activation states in microglia is negatively correlated with a differentially expressed neuronal module, implicating dysregulated microglial responses in concert with altered neuronal activity-dependent genes in autism brains. These observations provide pathways and candidate genes that highlight the interplay between innate immunity and neuronal activity in the aetiology of autism.

Figure 1. Single-gene expression analysis identifies two transcriptome-wide significantly differentially expressed genes between autism and control brains.

(a) Manhattan plot for 13,262 expressed genes. The threshold for transcriptome-wide significance was calculated based on 400 permutations (P<4.76 × 10⁻⁷) and is indicated by the dotted grey line. (b,c) Boxplot of gene expression in 57 controls (grey) and 47 cases (red), indicating a 1.2-fold increase for MAL and a 0.6-fold decrease for C11orf30 in cases relative to controls.

Figure 2. Weighted gene correlation network analysis identifies 12 co-expression modules.

(a) Dendogram of 12 co-expressed modules, with major cell type/function enrichment noted (Supplementary Data 12,13a–k). Mod12 was not significantly enriched for any cell type or GO terms. (b) Disease association of each module, represented by each module’s first principal component (eigengene). Three co-expression modules are associated with autism with nominal significance (black), with mod5 significant after multi-test correction (P<9.64 × 10⁻⁴; Supplementary Data 11). A positive (+) sign indicates upregulated gene expression in autism cases. (c–f) Enrichment analysis for gene lists compiled from the literature (Supplementary Data 7,12). (c) Individual dysregulated co-expression modules in autism brains identified by Voineagu et al. are captured by multiple co-expression modules in the current study, allowing for refinement of the signal associated with autism. (d) Genes with known common and rare de novo variants associated with autism are enriched only for mod2, which does not show differential expression. (e) FMRP targets identified by Darnell et al. are enriched in neuronal co-expression modules. (f) FMRP targets were split into fetal and adult/adolescent expression patterns, and are captured by different co-expression modules. Red, P<0.05; grey, P>0.05.

Figure 3. Gene co-expression module mod5 is associated with autism.

(a) The module eigengene of mod5 is upregulated in autism cases (red) compared with controls (black). (b) Heatmap of mod5 co-expression for 759 genes, stratified by disease status, showing greater co-expression between cases (bottom left) compared with controls (upper right). (c) Mod5 is significantly enriched for genes associated with M2-microglial cell states (Supplementary Data 12). (d) Mod5 is significantly enriched for GO terms related to immune response (Supplementary Data 13e).