RARE CODING VARIANTS IN 195,257 INDIVIDUALS REVEAL 29 NOVEL BIPOLAR DISORDER GENES AND STRUCTURAL MISSENSE HOTSPOTS IN PROTEIN SPACE

Bipolar disorder (BD) is a complex psychiatric disorder characterized by recurrent mood disturbances with significant variations in disease severity and treatment response. Despite its substantial impact on individuals, the genetic factors contributing to BD remain incompletely understood. Here, we present an updated effort from the Bipolar Exome (BipEx) consortium. We increased the effective sample size by 5-fold relative to BipEx 1.0 across diverse populations to investigate the role of protein-truncating variants (PTVs), copy number variants (CNVs), and damaging missense variants in the pathogenesis of BD. Our study cohort comprised 195,257 individuals, including 45,479 cases and 149,778 controls from ancestrally diverse populations. Whole exome sequencing was performed to capture protein-coding regions, followed by robust variant calling and stringent quality control. We analyzed ultra-rare variants with a minor allele count < 5, focusing on PTVs and damaging missense variants. CNVs were identified using GATK-gCNV, focusing on rare exonic deletions and duplications spanning more than three exons. To characterize structural convergence of missense variation, we mapped variant positions to protein 3D structures using AlphaFold and identified significant clustering in spatial neighborhoods across the proteome. We identified 29 FDR-significant genes, including AKAP11, SHANK1, ATP2B2, KDM5B, and DOP1A, associated with increased burden of PTVs and damaging missense variants. These genes cluster into early and late neurodevelopmental expression patterns and recapitulate clinically relevant pathways, including those associated with clozapine response. Using regional missense constraint metrics, we observed damaging missense variants in BD cases—but not controls—within a highly constrained region of ATP2B2. We also identified significant 3D missense clustering in ATP2A2 and G3BP2, implicating structurally focused hotspots as contributing to BD risk. CNV analysis revealed BPNT2 as the strongest deletion-associated gene (OR: 9.04, P = 2.67 × 10⁻⁹), a known lithium target, and BIRC6 as the most significant duplication hit (OR: 5.04, P = 2.49 × 10⁻¹⁴). We also found evidence for a potential reciprocal CNV locus at 15q11.2, where the duplication increases BD risk while the deletion is depleted in cases. By leveraging data from nearly 200,000 individuals, BipEx 2.0 provides a comprehensive view of rare genetic variation in BD, emphasizing contributions from PTVs, CNVs, and damaging missense variants. The identification of missense-convergent 3D protein hotspots, including in ATP2A2 and G3BP2, highlights the value of integrating structural biology into psychiatric gene discovery. These results illuminate convergent biological mechanisms across variant classes and point to new directions for mechanistic investigation.