Alzheimer's Disease Sequencing Project: Age Extremes x APOE Genotype Sampling for Genetic Discovery
Alan E. Renton1, Edoardo Marcora1, Tushar R. Bhangale2, Yi Zhao3, Brian Fulton-Howard1, Amanda Kuzma3, Victoria M. Fernandez4, Carlos Cruchaga4, Richard P. Mayeux5, Alzheimer's Disease Sequencing Project Protective Variant Working Group, Sudha Seshadri6, Adam C. Naj3, Gerard D. Schellenberg3, Timothy W. Behrens2, Robert R. Graham2, Lindsay A. Farrer6, Joshua C. Bis7, Alison M. Goate1
1Icahn School of Medicine at Mount Sinai, New York, NY; 2Genentech Inc, South San Francisco, CA; 3University of Pennsylvania, Philadelphia, PA; 4Washington University, St. Louis, MO; 5Columbia University, New York, NY; 6Boston University, Boston, MA; 7University of Washington, Seattle, WA
The APOE locus harbors the strongest known genetic risk factor for sporadic Alzheimer's disease (AD). However, clinical onset is delayed or absent in some APOE risk allele carriers, possibly due to protective genetic variants. Simulations suggest that sampling phenotypic extremes can increase power in rare variant association analyses.
To identify novel protective and risk AD loci, we undertook discovery case-control association among non-Hispanic whites using age extremes sampling in two strata defined by APOE genotype:  APOE4 extremes (ε4/ε4 or ε3/ε4 AD cases with age at onset [AAO] ≤ 65 years vs. controls with age at last assessment [ALA] ≥ 75 years for ε4/ε4 subjects or ≥ 80 years for ε3/ε4 subjects) from the Alzheimer's Disease Sequencing Project (ADSP) and Genentech sequencing datasets; and  APOE33 extremes (ε3/ε3 AD cases with AAO ≤ 70 years vs. ε3/ε3 controls with ALA ≥ 85 years) from ADSP. Replication was performed according to the same criteria in non-Hispanic white participants from the Alzheimer's Disease Genetics Consortium using GWAS datasets imputed to the Haplotype Reference Consortium panel. We undertook covariate-adjusted SKAT-O gene-based aggregation analyses.
We detected borderline study-wide significance for EIF2B3 in the APOE4 stratum. Homozygous EIF2B3 mutations cause vanishing white matter leukoencephalopathy, which is phenotypically similar to Nasu-Hakola disease caused by homozygous TREM2 mutations. Our analyses in the APOE33 stratum confirmed a known association for TREM2 and found a novel study-wide significant association for HOXD4. Discovery signals were not replicated, suggesting larger sample sizes are needed to leverage phenotypic extremes sampling in this disorder.