Low-frequency variant imputation identifies rare variant candidate loci in a genome-wide association study of late-onset Alzheimer disease. K. L. Hamilton1, B. W. Kunkle1, A. C. Naj2, W. R. Perry1, A. Partch2, O. Valladeres2, L. S. Wang2, G. Jun3, J. Chung3, M. A. Schmidt1, G. W. Beecham1, E. R. Martin1, R. P. Mayeux4, J. L. Haines5, L. A. Farrer3, G. D. Schellenberg2, The Alzheimer's Disease Genetics Consortium 1) Hussman Institute for Human Genomics, Miller School of Medicine, University of Miami, Miami, FL; 2) Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; 3) School of Medicine, Boston University, Boston, MA; 4) Taub Institute of Research on Alzheimer's Disease, Columbia University, New York, NY; 5) Department of Epidemiology & Biostatistics, Case Western Reserve University, Cleveland, OH.

   Recent Genome-wide Association Studies (GWAS) have identified 19 susceptibility loci, in addition to APOE, for Late-onset Alzheimer Disease (LOAD). The Alzheimer Disease Genetics Consortium (ADGC) conducted a rare variant focused GWAS using 31 datasets (including 16 new datasets), increasing our sample to 14,459 cases and 14,556 controls. Two datasets are family-based and thus potentially enriched for rare variants. Using IMPUTE2, all data were imputed using a 1000 Genomes 2012 reference set of over 37 million variants, many of which are low-frequency variants and indels. Association analysis was conducted adjusting for age, gender and population substructure. Individual datasets were analyzed with SNPTest using the score test (recommended for low-frequency variants), and within-study results were meta-analyzed in METAL. The imputation increased the number of high quality analyzable variants by 54.2% over the previous analyses done in collaboration with the International Genomics of Alzheimer Project (IGAP) (~15,500,000 variants vs ~7,100,000 variants). Approximately 9 million and 6.5 million of the total analyzable variants had a minor allele frequency (MAF) <0.05 and <0.01, respectively. 102 low-frequency variants (defined as MAF0.02) demonstrated genome-wide significance (P510-8), with 65 of these variants being located in the APOE region. While other previously associated common loci produced significant signals, none showed significance for low-frequency variants. 37 variants with MAF0.02 in 9 novel loci demonstrated genome-wide significance (5 are family dataset specific), with one of the loci including the previously reported TREM2 gene (P=1.64x10-8; MAF=0.002). Other significant loci include the actin-related gene, ACTN1 (P=3.92x10-11; MAF=0.001), which has been previously associated with LOAD related changes in hippocampal gene expression, and LRFN5 (P=4.40x10-13; MAF=0.002), a cell adhesion molecule that promotes neurite outgrowth and synapse maturation in hippocampal neurons. An additional 37 variants at 17 loci were nominally associated (P=3.92x10-6), including a variant in LUZP2, a gene previously implicated in LOAD in Amish through linkage and association analyses. Using an imputation set with a large amount of rare variation, we identified several novel rare variant candidate loci for LOAD, giving support to the hypothesis that rare and low-frequency variant imputation can identify novel associations with disease.

You may contact the first author (during and after the meeting) at