COMMON STRUCTURAL VARIANTS DRIVE TOP GWAS SIGNALS IN AUTISM AND RELATED NEURODEVELOPMENTAL DISORDERS

Background Genome-wide association studies (GWAS) have identified dozens of loci for autism and related neurodevelopmental disorders (NDDs), yet pinpointing causal variants remains challenging. Standard fine-mapping assumes the causal variant is a SNP present in reference panels, but common structural variants (SVs)—deletions, duplications, inversions—are largely absent from these efforts. While rare SVs show robust NDD associations, whether common SVs contribute substantially to genetic architecture, and whether SNP signals tag underlying causal SVs at top GWAS loci, remains unknown. Methods We developed a population-scale SV imputation panel from 15x Oxford Nanopore long-read whole-genome sequencing of 834 multi-ancestry 1000 Genomes subjects, enabling imputation of > 42k common SVs (MAF > 0.01) from cis-window tag SNPs, including > 20k deletions, > 21k insertions/duplications, 115 inversions, and 517 translocations. High-confidence SVs (R2 > 0.8) were imputed into GWAS summary statistics for autism, ADHD, and a transdiagnostic NDD factor. We assessed LD (R2≥0.8) between index SNPs and neighboring SVs, and performed GCTA-COJO conditional analyses to determine whether locus-level signals were attributable to SNP-tagged SVs. Candidate SVs were then functionally annotated for gene impact, chromatin states, brain regulatory elements, and 3D genomic organization. Results Common SVs explained 4 of 41 (9.7%) GWAS loci across autism, ADHD, and the NDD factor. The top autism locus on chromosome 20 harbored a genome-wide significant (GWS) ∼3kb intronic deletion within XRN2, which was also GWS for ADHD and the NDD factor. Conditioning on this SV attenuated the lead SNP signal > 10⁸-fold, collapsing three distinct SNP associations into a single pleiotropic deletion. Three additional loci harbored SV associations: an upstream deletion near MEF2C (ADHD/NDD factor) and intronic insertions within FOXP2 and ST3GAL3 for ADHD. These SVs show regulatory convergence—XRN2 and ST3GAL3 SVs colocalize with RNA-binding protein motifs and repetitive elements; MEF2C breakpoints intersect SVA_E retrotransposons; the FOXP2 insertion region is haploinsufficiency-sensitive. Finally, XRN2, MEF2C, and FOXP2 are all constrained transcriptional regulators with peak expression during early brain development. Discussion Common SVs represent a previously hidden component of NDD genetic architecture. Nearly 10% of GWAS signals, including top loci, likely reflect underlying SVs rather than SNP associations. As long-read sequencing panels grow, this SNP-SV framework will be critical for overcoming limitations of conventional fine-mapping and revealing functional biology of neurodevelopmental disorders.