Parallel in vivo analysis of large-effect autism genes implicates cortical neurogenesis and estrogen in risk and resilience

Abstract

Gene Ontology analyses of autism spectrum disorders (ASD) risk genes have repeatedly highlighted synaptic function and transcriptional regulation as key points of convergence. However, these analyses rely on incomplete knowledge of gene function across brain development. Here we leverage Xenopus tropicalis to study in vivo ten genes with the strongest statistical evidence for association with ASD. All genes are expressed in developing telencephalon at time points mapping to human mid-prenatal development, and mutations lead to an increase in the ratio of neural progenitor cells to maturing neurons, supporting previous in silico systems biological findings implicating cortical neurons in ASD vulnerability, but expanding the range of convergent functions to include neurogenesis. Systematic chemical screening identifies that estrogen, via Sonic hedgehog signaling, rescues this convergent phenotype in Xenopus and human models of brain development, suggesting a resilience factor that may mitigate a range of ASD genetic risks.

Keywords: CRISPR; Xenopus tropicalis; autism spectrum disorders; brain development; convergent; estrogen; genetics; neural progenitor cells; neurogenesis; sonic hedgehog.

Figure 1.. Top 10 ASD risk genes are all first expressed during telencephalic neurogenesis.

A) Whole-mount RNA in situ hybridization for top 10 ASD risk genes over development in X. tropicalis. Note co-expression in stage 40 near the first ventricle (insets with red arrows). See Fig. S1 for comparison with marker gene expression patterns. B) Mapping X. tropicalis brain RNA-Seq profiles during tadpole development to human BrainSpan RNA-Seq data by principal component analysis (PCA). Stage 40, when ASD genes are first all expressed in the telencephalon, maps closest to human early mid-prenatal development by plotting principal component (PC) 1, which tracks with developmental age. Bands indicate the mean PC1 value for each human stage plus a 95% confidence interval. Xenopus samples are shown as points with three biological replicates. “pcw” stands for post-conception weeks. See Fig. S1C–E for full PCA and SEA.

Figure 2.. Top ten ASD risk genes impact telencephalon size.

A) Unilateral mutants made by injecting Cas9 protein, a sgRNA for an ASD risk gene, and dye (red) into one cell of two-cell stage X. tropicalis embryos. Telencephalon (tel). B) Control CRISPR (right side) targeting pigmentation gene slc45a2 has a symmetric brain (β-tubulin stain). Telencephalon region for each half is outlined by a dotted line. Mutating nrxn1 (C) increased telencephalon size, while mutating syngap1 (D), pogz (E), or dyrk1a (F) decreased it. (G) Human DYRK1A plasmid injection rescues dyrk1a CRISPR. H) Telencephalon size quantification by targeted gene. Controls (blue) and ASD risk gene CRISPRs (red). Measurements normalized by within-animal control side. p values from nonparametric Mann-Whitney rank sum tests, compared to slc45a2 CRISPR. “n.s.,” not significant, indicates p > 0.05. (*) p < 0.05, (**) p < 0.01, (***) p < 0.001, and (****) p < 0.0001. See Fig. S2 for mutational efficiencies and control CRISPR images, and Fig. S3 for other brain region measurements. I) Human gene rescue quantification, nonparametric Mann-Whitney test.

Figure 3.. Top ten ASD risk genes impact NPC maturation.

Mutating adnp (A) or syngap1 (B) reduced ventricle size (green). Telencephalon and ventricle are outlined by a dotted line. C) NPC area quantification from PCNA staining by targeted gene (blue are controls and red are ASD risk genes). Measurements are normalized by the within-animal control side. D) NPC to differentiated neurons quantification by gene. E) CRISPRi against ASD risk genes in human neural progenitor cells causes an increase in the proportion of KI67⁺ proliferative NPCs. p values are from nonparametric Mann-Whitney rank sum tests where “n.s.,” not significant, indicates p > 0.05. (*) p < 0.05, (**) p < 0.01, (***) p < 0.001, and (****) p < 0.0001.

Figure 4.. Expression and functional convergence of ASD risk genes in human NPCs.

A) Overview of analysis. We integrate the top 102 high confidence ASD (hcASD) risk genes (Satterstrom et al. 2020) with the PCNet human interactome reference database (Huang et al. 2018) and BrainSpan layer-specific prenatal human frontal neocortex microarray gene expression data (Miller et al. 2014) to construct layer-specific hcASD interaction networks. We then identify the brain layer(s) with the strongest convergence of hcASD molecular interactions. B) hcASD genes are more connected than expected by chance in PCNet, as measured by the number of direct PCNet interactions among hcASD genes (p < 0.001), total number of hcASD genes that are connected to at least one other hcASD gene (p < 0.001), and number of interactions with hcASD genes and any other gene in PCNet (p = 0.009). Red line indicates the observed value, grey histogram shows the null distribution of 1,000 permutations. C) PCNet interactions have higher expression correlation in BrainSpan layer-specific expression data than non-PCNet interactions (p = 4.78 x 10⁻⁵). 7 BrainSpan layers assessed (VZ, SVZi, SVZo, IZ, SP, CPi, CPo), p-value obtained by paired sample T-test. D) The SVZi-specific hcASD interaction network has the highest average connectivity per interaction. E) SVZi has significantly higher hcASD interaction network connectivity than expected by chance (p < 0.0001). Red line indicates the observed connectivity, grey histogram shows the null distribution from 10,000 permutations. Abbreviations: VZ, ventricular zone. SVZi, inner subventricular zone. SVZo, outer subventricular zone. IZ, intermediate zone. SP, subplate. CPi, inner cortical plate. CPo, outer cortical plate.

Figure 5.. Drug screen identifies estrogen as a suppressor of the convergent ASD phenotype.

A) DYRK1A inhibitor alone (1.25 μM harmine) increases the ratio of NPCs to neurons (red dashed line) compared to control (DMSO, blue dashed line). See Fig. S4 for harmine validation. Each point is the mean ratio following treatment with a 10 μM NCI oncology set VIII drug and 1.25 μM harmine. Several estrogen pathway drugs modified the phenotype (change greater than one standard deviation (St. Dev.)), including estramustine (pro-estrogen), fulvestrant (estrogen receptor modulator), and raloxifene (aromatase inhibitor). B-D) 10 μM 17-β-estradiol suppresses the convergent ASD phenotype generated by 1.25 μM harmine treatment. E) Quantification of B-D. p value is from Mann-Whitney test.

Figure 6.. Estrogen signaling is required for telencephalon development.

A–C) Whole-mount RNA in situ hybridization on stage 40 X. tropicalis embryos highlights expression of estrogen receptors ɑ (ERɑ/esr1, A), β (ERβ/esr2, B), and aromatase (cyp19a1, C) in the telencephalon (red arrows). D–G) Unilateral loss of estrogen pathway components reduces telencephalon size autonomously (β-tubulin stain). G) Quantification of telencephalon size by condition (normalized by contralateral control). p values are from nonparametric Mann-Whitney rank sum tests where “n.s.,” not significant, indicates p > 0.05. (*) p < 0.05, (**) p < 0.01, (***) p < 0.001, and (****) p < 0.0001.

Figure 7.. Estrogen signaling inhibits Sonic hedgehog signaling.

A) Differentially expressed (DEX) genes (fold change > 2) from RNA sequencing of dissected brains following 10 μM 17-β-estradiol treatment (blue) or 5 μM cyclopamine treatment (yellow), showing significant (p < 0.0001, hypergeometric test) overlap of DEX genes (green) in the same direction. Treatment with 10 μM 17-β-estradiol (C) or 5 μM cyclopamine (D) causes a marked reduction in the ratio of NPCs (PCNA, green) to neurons (vGLUT1, magenta) and midline defects in the telencephalon. E) Quantification of B-D, Mann-Whitney test, **** is p < 0.0001. Treatment with 20 μM 17-β-estradiol (G) reduces SHH target gene patched1 expression compared to control DMSO (F). Positive control 10 μM cyclopamine (H) reduces patched1 expression, while 50 μM aromatase inhibitor increases it (I).

Figure 8.. The role of estrogen is conserved in models of human brain development.

A) Quantification of normalized percent Ki67⁺ human NPCs following CRISPRi and treatment with DMSO (control) or 5 μM 17-β-estradiol. Statistical comparisons are nonparametric Mann-Whitney rank sum tests, * indicates p < 0.05. B-E) 5 week old human iPSC-derived cortical organoids treated with DMSO (B), 20 μM 17-β-estradiol (C), 2 μM harmine (D), or both (E) for 7 days, stained for DAPI (blue, nuclei) and Ki67 (green, proliferating cells). Percent of Ki67+ cells (F). p value is from a Mann-Whitney test. G) Log2 fold change expression for differentially expressed genes from 17-β-estradiol treatment in Xenopus brain versus in human postconception week (GW) 6–10 primary neurons. Blue and orange genes are those that change in the same direction in both species. Black genes are those that change in opposite directions. Grey dots are those that change in Xenopus but not in human neurons.