Full-length isoform transcriptome of the developing human brain provides further insights into autism

Abstract

Alternative splicing plays an important role in brain development, but its global contribution to human neurodevelopmental diseases (NDDs) requires further investigation. Here we examine the relationships between splicing isoform expression in the brain and de novo loss-of-function mutations from individuals with NDDs. We analyze the full-length isoform transcriptome of the developing human brain and observe differentially expressed isoforms and isoform co-expression modules undetectable by gene-level analyses. These isoforms are enriched in loss-of-function mutations and microexons, are co-expressed with a unique set of partners, and have higher prenatal expression. We experimentally test the effect of splice-site mutations and demonstrate exon skipping in five NDD risk genes, including SCN2A, DYRK1A, and BTRC. Our results suggest that the splice site mutation in BTRC reduces translational efficiency, likely affecting Wnt signaling through impaired degradation of β-catenin. We propose that functional effects of mutations should be investigated at the isoform- rather than gene-level resolution.

Keywords: alternative splicing; autism risk gene; autism spectrum disorder; co-expression module; human brain development; isoform transcriptome; neurodevelopmental disease; protein interaction network; splice-site mutations; splicing isoform expression.

Figure 1.. Differential gene and isoform expression analyses

(A) Number of significant DEGs and DEIs in the adjacent brain developmental periods and in prenatal versus postnatal periods. Isoform identifiers were summarized to gene identifiers for simplicity of comparison. Shaded areas represent identifiers shared between gene and isoform datasets, whereas unshaded bars represent genes (red) or isoforms (turquoise) unique to each dataset. (B) Effect size (absolute log₂ fold change) distribution of DEGs (red) and DEIs (turquoise) of combined data (top) and per developmental period (bottom). Average absolute effect sizes for genes and isoforms are marked by corresponding colored vertical lines, and differences were tested using two-sample t tests (*FDR < 0.05). (C) Enrichment of cell types and literature-curated gene sets among genes and isoforms unique to each dataset (unshaded sets from a panel). Fisher’s exact test was used to calculate p values. (D) Gene Ontology (GO) enrichment of DEGs and DEIs unique to each dataset (unshaded sets from a panel). Three adjacent periods are shown as examples (P04/P05, P07/P08, and P08/P09). DEIs are enriched in nervous system-related processes.

Figure 2.. Analyses of isoforms affected by rare de novo ASD LoF variants

(A) Mean expression of isoforms affected by case rare de novo ASD LoF variants (affected by ASD LoF) is significantly higher in prenatal periods compared with those affected by control LoF mutations or to non-affected isoforms. (B) Proportion of protein-coding isoforms of high-risk ASD genes from Satterstrom et al. (2020), uniquely differentially expressed at isoform level, affected (red) or not affected (blue) by rare de novo ASD LoF variants. (C) Ward hierarchical clustering of isoforms from (B) based on average expression values across developmental periods. (D) Expression profiles of affected and non-affected isoforms of four ASD risk genes across development, demonstrating higher prenatal expression of some affected isoforms. (E) Schematic of alternatively regulated micro-exons (top panel), proportion of all brain-expressed genes with alternatively regulated microexons (bottom left), and proportion of all brain-expressed isoforms with alternatively regulated microexons (bottom right). *p ≤ 0.1, **p ≤ 0.05, ***p ≤ 0.01.

Figure 3.. Gene and isoform co-expression analyses

(A) Association of gene and isoform co-expression modules clustered by module eigengene with developmental periods (top). Linear regression beta coefficients were calculated using linear mixed-effects models. Module enrichment in cell type and literature curated gene sets (bottom) was calculated using Fisher’s exact test. (B) Module eigengene expression profiles across brain development for modules most significantly associated with each cell type: astrocytes, iM25; oligodendrocytes, iM6; microglia, iM36; NPCs, iM10; excitatory neurons, iM2; interneurons, iM17. (C) GO functional enrichment analyses of gM1/iM1 and iM30 modules significantly affected by case ASD LoF mutations. (D) Gene (top panel) and isoform (bottom panel) co-expression modules affected by case and control ASD LoF mutations. Normalized effect rate per module is shown. Significance was calculated by permutation test (1,000 permutations, *FDR ≤ 0.05). (E) Gene-level and isoform-level co-expressed protein interaction networks for the KMT2A gene from the gM1 and iM1 turquoise modules. Only edges in the top 10% of expression PCCs that are also supported by gene-level protein interactions are retained.

Figure 4.. Functional effect of the de novo splice site mutations from individuals with NDDs

(A–D) Minigene assays demonstrate the effect of splice-site mutations in four genes: SCN2A (A), DYRK1A (B), DLG2 (C), and CELF2 (D). A schematic of the cloned minigenes, the expected splicing patterns, and the effects of the mutations are shown below the gel image. Numbers denote base pairs. M, molecular marker; E, exon. (E) Expression profiles across brain development of the brain-expressed isoforms transcribed by these four genes, annotated with module memberships; highly overlapping expression profiles are unlabeled for readability.

Figure 5.. The de novo autism splice-site mutation causes exon skipping in BTRC isoforms and reduces their translational efficiency

(A) The exon structure of three splicing isoforms of the BTRC gene showing positions of the cloned abridged introns and the splice-site mutation; numbers denote base pairs. (B) Minigene assays demonstrate exon 4 skipping as a result of the splice-site mutation. The assays show the results of RT-PCR performed using total RNA from HeLa cells transfected with BTRC minigene constructs; numbers denote base pairs. (C) Splicing assays with full-length constructs carrying abridged introns, confirm exon skipping observed in the minigene assays. (D) Immunoblot (IB) of whole-cell lysates of HeLa cells transfected with different BTRC minigene constructs and an empty vector, as indicated. Membranes were probed to observe BTRC overexpression and investigate expression of p-β-catenin, Cul1, and SKP1. β-Actin was used as a loading control. IP was performed with the antibody recognizing the V5 tag, and proteins were detected by IB with p-β-catenin, Cul1, SKP1, and V5 antibodies. The splice-site mutation causes reduced translational efficiency of the BTRC_1Mut and BTRC_2Mut mutant isoforms compared with their WT counterparts. A schematic of the Skp1-Cul1-BTRC ubiquitin protein ligase complex is shown at the bottom. (E) Quantification of protein pull-down with V5 IP using ImageJ software. The band intensity values were normalized to WT expression levels. Error bars represent 95% CI based on 3 independent experiments. On average, a 40% reduction of BTRC protein expression is observed as a result of a mutation. Consequently, a reduction of the corresponding BTRC binding partners (p-β-catenin, Cul1, and SKP1) is also observed. (F) Expression profiles of brain-expressed BTRC isoforms show higher expression of ASD-affected BTRC-001 and BTRC-002. Numbers denote base pairs (A–C) or kilodaltons (D). *p < 0.05, **p ≤ 0.01, ***p ≤ 0.001.