UC San Diego

Rare copy number variants (CNVs) have been strongly implicated in autism (ASD) and schizophrenia (SCZ), but well-powered genome-wide studies of rare variants have not been carried out across multiple major psychiatric disorders. Here we perform a genome-wide association of CNVs across schizophrenia (SCZ), autism spectrum disorder (ASD), bipolar disorder (BD), major depressive disorder (MDD), post-traumatic stress disorder (PTSD), attention-deficit/hyperactivity disorder (ADHD) and in the combined cross-disorder cohort (XD) (N=537,466).

CNV calling was performed using a consistent ensemble pipeline that enables the combining of datasets and disorders. CNV burden and CNV-GWAS were used to characterize genetic associations genome-wide. Associations were tested for deletion (DEL) and duplication (DUP) separately, controlling for ancestry, genotyping platform, and cohort.

CNV burden analysis revealed that rare CNVs contribute to all 6 disorders but with effects that differ by disorder. Genome-wide association analyses of CNV across disorders found 36 significant associations at 21 independent loci. For all disorders, alleles span the full range of frequencies, but disorders differ in their distribution of effect sizes. Effects were strongest for ASD and were invariably positive. Rare variants in SCZ, BD, and MDD were a mixture of positive and negative effects. Effect sizes in MDD were comparatively weak. Many CNVs were associated with multiple disorders but not to the same extent. Some were predominantly associated with ASD (15q11-13 Dup), some were most strongly associated with SCZ (16p11Dup, 22q11Del), and other CNVs were weakly associated with several disorders but reached genome-wide significance only in the combined XD cohort (ASTN2, DLG2). Without exception, all associations occurred in genomic regions that are prone to high rates of structural mutation. 15 were hot spots for non-allelic homologous recombination (NAHR) and 6 were in common fragile sites (CFS) where chromosome breakage occurs within large neural genes (ASTN2, DLG2, DLGAP2/CSMD1, IMMP2L, NRXN1, SHANK3). Several large neural genes also intersect with topologically associating domain (TAD) boundaries, but associations were driven by protein coding deletions, and were not attributable to non-coding variants. An enrichment of CNV associations near TAD boundaries in this case may not be attributable to cis-regulatory effects of rare variants, but instead attributable to the intrinsic genomic instability of these regions that gives rise to deletions.

Genome-wide analysis of rare CNVs across major psychiatric disorders identifies many risk loci including both positive and negative associations and novel gene associations. A comparative analysis highlights considerable genetic overlap between disorders but also distinguishable differences in allelic architecture and considerable phenotypic specificity to CNV associations.