Combating Genetic Heterogeneity for Polygenic Prediction of Susceptibility to Brain β-Amyloid Deposition: Beyond .

Background And Objectives: The (apolipoprotein E) ε4 allele is the strongest known genetic risk factor for sporadic Alzheimer disease (AD) and for brain amyloidosis, an early marker of disease pathophysiology. However, ε4 is present in only 25% of the general population and is by itself inadequate for explaining susceptibility to amyloid accumulation or AD diagnosis. Existing studies have been limited by potential confounding due to inclusion of individuals carrying ε4 or ε2 (which has a modest protective association). We hypothesized that genome-wide association study (GWAS) and genetic risk score (GRS) analyses in ε3/ε3 individuals would uniquely identify novel predictors of β-amyloid pathology in older adults.

Methods: We analyzed data from the Mayo Clinic Study of Aging (MCSA), Alzheimer's Disease Neuroimaging Initiative (ADNI), and Rush Religious Orders Study and Memory and Aging Project. Frequency of ε3/ε3 in those samples ranged from 48% to 61%. A GWAS was performed across 1,496 individuals with amyloid PET to identify candidate variants for GRS generation. Postmortem neuropathologic data (N = 710) were used to refine the variant list to capture high-likelihood true associations. An independent sample (N = 641) with plasma p-tau data was used for validation.

Results: The GWAS identified previously implicated (e.g., and ) and novel potential associations with amyloid PET burden. A non- GRS of top variants was strongly associated with amyloid PET levels in the MCSA ( = 4.34 × 10, β = 5.88) and ADNI ( = 1.87 × 10, β = 12.1) cohorts. In both cohorts, this non- amyloid GRS outperformed a comparator GRS (based on variants associated with clinically diagnosed AD dementia risk) in explaining phenotypic variation. The non- amyloid GRS was also associated with postmortem neuropathologic β-amyloid and neurofibrillary tangle burden and in an independent sample was associated with plasma p-tau concentrations (a robust indicator of cerebral amyloidosis).

Discussion: Our non- amyloid GRS, which appropriately includes variants associated with amyloid deposition in ɛ4/ɛ2 noncarriers, may advance personalized prediction of genetic susceptibility to β-amyloid accumulation within the large segment of the population that is ε3/ε3. This may have future implications for risk modification, trial enrollment, and treatment selection.