Extremely rare CNVs contributing to Alzheimer disease risk: a case-control association analysis of exome sequencing data from 22,319 individuals

Rare coding single nucleotide variants (SNV) and short insertions or deletions (indels) contribute to Alzheimer disease (AD) genetic risk, from pathogenic variants in autosomal dominant genes to risk factors with diverse effects. In contrast, copy number variants (CNV) have been scarcely studied, with the exception of a few autosomal dominant examples, such as APP gene duplications. We took advantage from a large case-control dataset of 22,319 exomes (4,150 early-onset AD (EOAD, onset [≤]65 years), 8,519 late onset AD (LOAD) and 9,650 controls) to detect CNVs. We first identified 17 causative CNVs in autosomal dominant genes (9 novel: 7 APP and 2 MAPT duplications). After exclusion of carriers of these, we performed an original two-step analysis: (i) a protein-coding genome-wide analysis at the transcript level using a gene dosage strategy (EOAD versus controls) and then (ii) an integrated loss-of-function (LOF) analysis gathering short truncating variants with CNV-deletions in genes prioritized in (i) and in a list of known AD risk genes. We identified AD association with dosage of 20 genes at 4 different loci with a false discovery rate (FDR) below 10%, including the chr22q11.21 central region (FDR=0.0386), a region in linkage disequilibrium with the APOE locus on chr19 (FDR=0.0271), and two single-gene loci, namely FADS6 (FDR=0.0271), and ADI1 (FDR=0.0916). Replication in an independent dataset made of genotyping array data from 2,780 EOAD cases, 15,222 LOAD cases and 273,979 controls was consistent with the results obtained in the discovery dataset. The integrated LOF analysis helped narrowing the region of interest to the SCARF2-KLHL22-MED15 region at the 22q11.21 locus. In addition, the integrated LOF analysis highlighted rare deletions in the known AD-risk genes ABCA1 and ABCA7 that represented 10% (3/30) and 8.6% (10/115) of LOF alleles of these genes, respectively, as well as 4 TYROBP deletions. Finally, we identify CTSB LOF alleles as candidate rare AD risk factors (p=0.0089). In conclusion, our results show that carriers of a deletion of FADS6 or a 22q11.21 deletion encompassing the SCARF2-KLHL22-MED15 region, including some patients with DiGeorge syndrome, may have a higher risk of developing AD. CNVs represent a source of genomic variation that can contribute to AD etiology in new genes but also in GWAS defined-genes.