Abstracts

Rationale: Rare copy number variants (CNVs) are strongly implicated in the etiology of epilepsies, in particular large deletions at genomic hotspots. Previous studies either analyzed only one specific epilepsy type, were limited in cohort size or did not uniformly use the same technology to identify CNVs. These factors inhibited comparison across studies and it is today not clear to which degree CNVs contribute to specific epilepsy types. Here we present the largest CNV burden analysis including >11k epilepsy patients ascertained in the global acting consortium ‘Epi25’ and >7k controls. All samples have been systematically phenotyped, genotyped with the same technology and analyzed using the same bioinformatic pipeline. Methods: Samples were genotyped using the Illumina Global Screening Array-24 v1.0. After quality control, we analyzed a cohort of 11,051 European epilepsy cases and 7,278 ancestry-matched controls for burden of rare CNVs (< 1% frequency). CNV burden analyses were also performed for cases divided into five subgroups: Non-acquired focal epilepsy (NAFE, n=4,597), genetic generalized epilepsy (GGE, n=3,538), epileptic encephalopathy (EE, n=1,278), lesional focal epilepsy (LFE, n=1,238) and unclassified epilepsy (UE, n=400). Bonferroni correction was used to define significance thresholds. Results: Significant enrichment of deletions was observed for all epilepsies combined vs controls (odds ratio (OR) =1.93, 95%-CI=1.53-2.45, P=3.1×10^-8), with the strongest signal for enrichment coming from hotspot deletions in GGE (OR=2.69, 95%-CI=2.06-3.53, P=5.8×10^-13) and NAFE ( OR=1.63, 95%-CI=1.63-1.23, P=7.1×10^-4). The hotspot signal in GGE was mainly driven by only two CNVs, 15q13.3 (OR= 37.2, 95%-CI=5.87-1538.89, P=2.4×10^-8) and 16p13.11 (OR=21.72, 95%-CI=5.3-190.7, P=7.4×10^-9). Large deletions (>2Mb) were enriched in EE (OR=3.73, 95%-CI=2.1-6.07, P=9.5×10^-7), GGE (OR=1.32, 95%-CI=1.11-1.55, P=9.96×10^-4) and LFE patients (OR=3.13, 95%-CI=1.5-5.66, P=6.5×10^-4). Patients with UE were the only group, which showed significant enrichment for gene covering duplications (OR=1.55, 95%-CI=1.27-1.89, P=1.4×10^-5). No patient group showed genome-wide burden for noncoding CNVs. Exome-wide gene level analysis showed strongest enrichment for genes in hotspot regions at 15q13.2 and 15q13.3 in GGE cases (FAN1, OTUD7A both OR=23.8, 95%-CI=5.87-208, P=9.3×10^-10; KLF13, OR=22.8, 95%-CI=5.59-199, P=2.6×10^-9) and at 15q11, 15q12 and 15q13 in EE cases (GABRA5, GABRB3 both OR=inf, 95%-CI=12.84-inf, P=5.4×10^-9; ATP10A, OR=inf, 95%-CI=11.3-inf, P=3.6×10^-8). Even when excluding hotspots CNVs, significant enrichment remained for deletions covering genes under evolutionary selection (pLI>0.95) in GGE patients (OR=1.57, 95%-CI=1.28-1.9, P=1.0×10^-5). Conclusions: In summary, using our very large patient cohort we are able to discover novel CNV associations with epilepsy types. We provide evidence for deletion burden outside of known hotspot regions for GGE affecting highly constrained genes and implicate large deletions in the genetic architecture of LFE. Funding: NIH-supported study