Effect of polygenic risk scores for Alzheimer's disease on brain structure, cerebrospinal fluid, and cognition

Abstract

Background: Apolipoprotein E (APOE) is the most studied gene in relation to, and has the strongest genetic association with, sporadic Alzheimer’s disease (AD). The APOE epsilon 4 (ɛ4) risk variant is found in approximately 14% of the general population and in approximately 37% of the AD population, globally. This indicates that the presence of ɛ4 is not required for the development of sporadic AD. In fact, sporadic AD is considered a polygenic disease with thousands of causative variants. Therefore, there is a need to investigate other genes and associated variants, and the concurrent effect of these on sporadic AD. Polygenic risk scores (PRSs) are ideal for this purpose as they consider multiple genetic variants/single nucleotide polymorphisms (SNPs) that have genome-wide significance, simultaneously. However, PRSs are difficult to calculate since numerous factors must be taken into consideration, and there is no optimal procedure. Nonetheless, PRSs can improve understanding of genotype-to-phenotype associations and risk predictions. Aim: The aim of the current research is to investigate whether PRSs for AD can predict cross-sectional regional grey matter volume (Experiment One), cross-sectional cerebrospinal fluid (CSF) biomarkers (Experiment Two Part A), cross-sectional cognitive markers (Experiment Two Part B), longitudinal CSF biomarkers (Experiment Three Part A), and/or longitudinal cognition (Experiment Three Part B) across the continuum of sporadic AD. Methodology: Participants from the Alzheimer's Disease Neuroimaging Initiative (n = 738; mean age in years: 73.94 ± 7.58) were analysed using various binomial hierarchical logistic regression models, and multiple hierarchical linear regression models, that controlled for numerous demographic and clinical variables. Participants were stratified by several variables to assess risk of AD in different groups. Three PRSs were constructed: PRSwithAPOE, PRSwithoutAPOE, and APOEonlyPRS. SNPs relating to APOE were from the entire APOE region, and three statistical thresholds were calculated for each of the three PRSs. Results: Overall, PRSwithAPOE and APOEonlyPRS were associated consistently with outcome measures. Generally, PRSwithAPOE and APOEonlyPRS were: associated negatively with bilateral hippocampus (right (R) > left (L)), bilateral amygdala (R > L), bilateral entorhinal area, and R middle occipital gyrus; associated positively with CSF amyloid-beta 42 (Aβ42) more than they were with CSF phosphorylated-tau181 (p-tau181), cross-sectionally; associated negatively with memory and visuospatial cognition, cross-sectionally; associated with CSF p-tau181 more than they were with CSF Aβ42, longitudinally; associated positively with longitudinal cognitive decline. Discussion: The results highlight that stratification by amyloid positivity improves risk prediction over stratification by clinical diagnosis. The findings also concur to previous literature indicating that CSF amyloid decreases early on in AD after which it reaches a plateau, whereas changes in tau continue throughout the disease. The current research presents a novel, systematic, and thorough approach to evaluate risk of sporadic AD, including: a multi-modal approach that assesses outcome measures that are used to diagnose individuals with AD clinically (i.e., brain structure, CSF biomarkers, and symptomology); the use of both cross-sectional and longitudinal study designs; a multi-stratification approach whereby participants are divided in several ways to assess risk in different groups – for some groups, findings have been reported for the first time (to knowledge). Conclusion: This work highlights the added value of non-ɛ4-APOE SNPs and non-APOE SNPs (i.e., both PRSwithAPOE and APOEonlyPRS) in detecting associations with AD-specific cross-sectional and longitudinal outcome measures. In parallel with existing literature, an ideal threshold has not been determined in the current research. However, it may be that certain thresholds identify associations better according to the outcome measure at hand. Additionally, the current research demonstrates the clinical relevance of PRSs for AD and that these may support diagnosis and prognosis of AD. It is recommended that future investigations assess PRSwithAPOE and APOEonlyPRS with longitudinal CSF biomarkers, in independent samples, and in ethnically diverse populations.