Genetic risk for alzheimer disease is distinct from genetic risk for amyloid deposition

Abstract

Objective: Alzheimer disease (AD) is the most common form of dementia and is responsible for a huge and growing health care burden in the developed and developing world. The polygenic risk score (PRS) approach has shown 75 to 84% prediction accuracy of identifying individuals with AD risk.

Methods: In this study, we tested the prediction accuracy of AD, mild cognitive impairment (MCI), and amyloid deposition risks with PRS, including and excluding APOE genotypes in a large publicly available dataset with extensive phenotypic data, the Alzheimer's Disease Neuroimaging Initiative cohort. Among MCI individuals with amyloid-positive status, we examined PRS prediction accuracy in those who converted to AD. In addition, we divided polygenic risk score by biological pathways and tested them independently for distinguishing between AD, MCI, and amyloid deposition.

Results: We found that AD and MCI are predicted by both APOE genotype and PRS (area under the curve [AUC] = 0.82% and 68%, respectively). Amyloid deposition is predicted by APOE only (AUC = 79%). Further progression to AD of individuals with MCI and amyloid-positive status is predicted by PRS over and above APOE (AUC = 67%). In pathway-specific PRS analyses, the protein-lipid complex has the strongest association with AD and amyloid deposition even when genes in the APOE region were removed (p = 0.0055 and p = 0.0079, respectively).

Interpretation: The results showed different pattern of APOE contribution in PRS risk predictions of AD/MCI and amyloid deposition. Our study suggests that APOE mostly contributes to amyloid accumulation and the PRS affects risk of further conversion to AD. ANN NEUROL 2019;86:427-435.

Figure 1

Density plots of polygenic risk score (PRS) for Alzheimer disease (AD), mild cognitive impairment (MCI), and cognitively normal participants. Standardized individual PRS scores for 3 phenotypes are shown (AD, MCI, and controls).

Figure 2

Diagram of prediction of amyloid deposition and further prediction of conversion of mild cognitive impairment (MCI) individuals to Alzheimer disease (AD) in the sample that was first clinically diagnosed with MCI using APOE and AD polygenic risk score (PRS). PRS predictions were first made for individuals who had baseline diagnosis of MCI. APOE alone and the full PRS model were used to predict amyloid deposition. The same models were used to predict which MCI individuals would convert to AD versus those individuals who had an MCI diagnosis using the latest clinical diagnosis. AUC = area under the curve; PET = positron emission tomography.

Figure 3

Overlap between 3 pathways: (1) protein–lipid complex (P‐L‐C; 40 genes), (2) protein–lipid complex subunit organization (35 genes), and (3) reverse cholesterol transport (R‐C‐T; 17 genes).