Heterogeneous epigenetic variation converges on splicing dysregulation in opioid addiction

Disease heterogeneity presents a major challenge for genetic and epigenetic dissection of complex traits such as opioid use disorder (OUD). Here, we leverage individual heterogeneity to identify epigenetic variation linked to opioid overdose by profiling H3K27ac ChIP-seq in nucleus accumbens (NAc) from 91 overdose cases and accidental death controls. Machine learning models built on combinations of peaks distinguished cases from controls with high accuracy (AUROC = 0.95) and were enriched for OUD genetic risk. We defined individual-specific H3K27ac changes as Variant Enhancer Loci (VEL) and observed four-fold greater convergence on shared target genes in NAc than in prefrontal cortex. Convergent VEL target genes were enriched in independent OUD cohorts, and RNA splicing emerged as the top-enriched ontology. Knockdown of the splicing regulator CELF5 in iPSC-derived medium spiny neurons disrupted extracellular matrix organization genes. Collectively, this study provides new insights into opioid addiction and a conceptual framework for dissecting heterogeneous disorders.