TBR1 regulates autism risk genes in the developing neocortex

Abstract

Exome sequencing studies have identified multiple genes harboring de novo loss-of-function (LoF) variants in individuals with autism spectrum disorders (ASD), including TBR1, a master regulator of cortical development. We performed ChIP-seq for TBR1 during mouse cortical neurogenesis and show that TBR1-bound regions are enriched adjacent to ASD genes. ASD genes were also enriched among genes that are differentially expressed in Tbr1 knockouts, which together with the ChIP-seq data, suggests direct transcriptional regulation. Of the nine ASD genes examined, seven were misexpressed in the cortices of Tbr1 knockout mice, including six with increased expression in the deep cortical layers. ASD genes with adjacent cortical TBR1 ChIP-seq peaks also showed unusually low levels of LoF mutations in a reference human population and among Icelanders. We then leveraged TBR1 binding to identify an appealing subset of candidate ASD genes. Our findings highlight a TBR1-regulated network of ASD genes in the developing neocortex that are relatively intolerant to LoF mutations, indicating that these genes may play critical roles in normal cortical development.

Figure 1.

TBR1 binds near high-confidence ASD genes. (A) Regulatory domains of Grin2b with eight adjacent TBR1 ChIP-seq peaks and Auts2 with 22 adjacent TBR1 ChIP-seq peaks. (B) Significance of the number of TBR1 ChIP-seq peaks adjacent to each high-confidence ASD gene set given the total number of peaks and size of the genomic regions used to associate peaks with their adjacent genes (the negative logarithm of the GREAT binomial P-value; x-axis) compared to the significance of the number of high-confidence ASD genes with an adjacent TBR1 peak given the total number of genes with an adjacent TBR1 peak (negative logarithm of the GREAT hypergeometric P-value; y-axis). Enrichment compared to E14.5 neocortex EP300 ChIP-seq, E15.5 neocortex SATB2 ChIP-seq, and 28 ENCODE ChIP-seq sets including tissues at different developmental time-points and primary cell lines. Dashed gray lines represent P = 0.05 significance level.

Figure 2.

TBR1 is necessary for ASD gene expression in specific cortical lamina. Radioactive in situ hybridization (RISH) of high-confidence genes at E15.5 (A) and P0 (B) in Tbr1^+/+ and Tbr1^−/− cortices reveal expression differences. RISH expression (lines in shades of red) corresponds to normalized grain counts (see Methods), and significance was determined using the two-sided t-test (n = 3–6 sections encompassing three samples per genotype) (see Methods). E15.5 cortical plate (red), P0 upper layers (brown), and P0 deep layers (pink). Upper layers correspond to layers 2–5 and deep layers correspond to layer 6 (see Methods). Error bars represent SD. (*) P-value < 0.05; (**) P-value < 0.01; (***) P-value < 0.001.

Figure 3.

Probable ASD genes that are TBR1 targets are more depleted for ExAC LoF mutations and biallelic LoF mutations in Icelanders. (A) Box plots depicting the distributions of fraction LoF scores for each gene from the ExAC reference population (y-axis) for merged ASD gene lists (x-axis). Probable ASD genes with adjacent TBR1 ChIP-seq peaks in the developing cortex have lower fraction LoF scores than those without an adjacent TBR1 peak. A pseudocount of one LoF allele for 121,412 sampled alleles, the maximum number sampled at any locus, was included for each gene for visualization purposes. Significance was determined using the one-sided two-sample Wilcoxon test. (B) The fraction of genes in each gene list with biallelic mutations in a study of Icelandic individuals (Sulem et al. 2015). Significance was determined using the one-sided Fisher's exact test. (*) P-value < 0.05; (***) P-value < 0.001.