Non-coding Regulatory Variants in ASD (Autism Spectrum Disorders) Disrupt CTCF Domains and Shape Cell-Type–Specific Neurodevelopmental Landscapes Revealed by Single-Cell Analyses and Cortical Organoids

Background Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder with substantial genetic heterogeneity. While coding variants have been extensively studied, non-coding regulatory variants—which constitute approximately 99% of the human genome—remain largely underexplored. Their functional impact, particularly at the single-cell level and in organoid models, requires further characterization to better understand how these variants contribute to ASD pathogenesis. Methods We performed targeted sequencing of 85,394 active cis-regulatory elements (cCREs) from ENCODE v2 in 600 samples (200 trios) previously negative for pathogenic coding variants. Using the Sei deep learning framework, we predicted regulatory impact scores for de novo and ultra-rare inherited variants. Gene mapping of regulatory variants was done using the T Gene tool which assigns target genes by integrating expression data. Differential gene expression was assessed using single-nucleus RNA-seq data from ASD and control brains (n = 64 individuals, ~ 600,000) from Wembley et al, 2023 and temporal gene expression dynamics was characterized in human cortical organoids scRNA seq data across four developmental stages (23 days to 6 months) from the Cortical Organoid Atlas (Broad Institute). Results We identified 164 de novo variants in 103 probands and 53,023 ultra-rare inherited variants in 199 probands. Following Sei-based prioritization, 47 de novo variants in 44 probands and 13,258 inherited variants in 198 probands showed significant regulatory impact score > 1.1. De novo variants were enriched in promoter-like sequences (PLS) (p = 0.02), with 53.19% affecting CTCF-mediated chromatin boundaries. Inherited variants showed significant enrichment for CTCF-binding loss (p = 1.55×10⁻⁸) and demonstrated maternal transmission bias (50.85% vs 49.15%, p = 9.97×10⁻⁵). Single-cell analysis of the target predicted genes revealed distinct cellular architectures for both gene groups. De novo genes ( POGZ, ROCK2, NFIB ) showed focused dysregulation in oligodendrocytes and specific neuronal populations, while inherited genes ( SYNJ2, UBC, CLU, DLC1 ) exhibited broader effects across multiple cell types. Cortical organoid profiling demonstrated that de novo genes display progressive developmental activation with precise spatiotemporal specificity— POGZ and ROCK2 exhibited bimodal expression peaks at 1 and 6 months in outer radial glia while NFIB showed linear upregulation across progenitor populations. In contrast, inherited genes maintained constitutive expression throughout development, with the exception of CLU, which showed regulated upregulation at later stages. Remarkably, de novo regulatory genes (POGZ, NFIB) are previously identified autism risk genes in large scale sequencing studies, providing validation of dosage sensitivity mechanisms. Conclusions Our analysis of regulatory variants in ASD indicates that non-coding mutations significantly contribute to ASD risk by disrupting CTCF domains, leading to alterations in cell type–specific gene regulation. Most de novo regulatory mutations predominantly target genes whose expression is dynamic, varying across different cell types and neurodevelopmental stages, whereas inherited regulatory variants affect constitutively expressed genes across multiple cell types, with stable expression throughout development with the exception of CLU. We also found that genes affected by de novo regulatory variants substantially overlap with those harboring pathogenic coding mutations identified in large-scale sequencing studies. These