Alzheimer's disease polygenic risk's association with all-cause dementia through the plasma metabolome in the UK Biobank study

Abstract

Dementia, with Alzheimer's Disease (AD) accounting for 60-70% of its occurrences, is a multifactorial disorder marked by cognitive and functional decline. Polygenic risk scores (PRS) stratify genetic risk for complex diseases, including AD. Integrating PRS with metabolomics offers a pathway to better understand AD etiology and identify biomarkers for early detection. We explored the association between AD PRS, metabolomics, and dementia risk using data from the UK Biobank. The analysis emphasizes sex-specific associations and the mediating role of metabolites between AD PRS and dementia risk. The study utilized data from 205,219 UK Biobank participants aged ≥ 50 years at baseline. Dementia outcomes were derived using ICD-10 codes. AD PRS were computed using genome-wide association study data, while metabolomic data included 249 biomarkers measured via Nuclear Magnetic Resonance. Time-to-event analyses (Cox proportional hazards models) and generalized structural equation modeling assessed associations and mediation effects, adjusting for age, sex, and genetic principal components. Higher AD PRS was associated with increased risks of all-cause dementia (HR: 1.75, 95% CI: 1.70-1.79, P < 0.001) and AD (HR: 2.02, 95% CI: 1.95-2.09, P < 0.001), with stronger effects in women for both outcomes when considering the main AD PRS which was related to APOE4 status. Metabolomic analyses identified lipid-related markers as key mediators. LDL phospholipid content was the metabolite marker with the strongest positive relationship with AD genetic risk (effect size, b = + 0.11), while HDL phospholipids percentage showed a similarly strong inverse association (effect size b = -0.09). No metabolomic markers were significantly associated with another version of AD PRS that was less correlated with APOE4 status. Mediation analyses applied to the main AD PRS revealed modest effects, with LDL-related metabolomic components partially mediating the effect of genetic risk on dementia incidence, through a protective type of mediation (< -2%). Nevertheless, most genetic risk operated independently of metabolites. AD PRS is significantly associated with dementia risk, with sex differences and metabolomic pathways providing further insights. Lipid metabolism, particularly LDL-centric measures, emerged as potential mediators. This integrative approach highlights the utility of combining genetic and metabolomic data to identify biomarkers and potential targets for early intervention in AD and dementia.

Keywords: Alzheimer’s disease; Dementia; Lipid metabolism; Metabolomics; Polygenic risk scores; UK Biobank.

Fig. 1

Alzheimer’s Disease polygenic risk tertiles in relation to all-cause dementia, before and after stratification by sex: UK biobank 2006–2023 (a) Main AD PRS (b) Newly derived AD PRS. Abbreviations: chi2 = chi-square test; CI = confidence interval; KM = Kaplan Meier; Pr = P-values; T1 = First tertile (lowest AD polygenic risk); T2 = Second teritle (medium AD polygenic risk); T3 = Uppermost tertile (Highest AD polygenic risk); UK = United Kingdom. Note: Chi2 refers to a log-rank test. Age of follow-up is truncated at 85y in this analysis to reduce likelihood of small sample sizes at older ages, given a baseline age range of 50-73y and up to 17 y of follow-up (between 2006–2010 and end of 2023). This restriction, which lead to a sample size of N = 202,946, was not made in the remaining parts of the analysis, that did not involve visualization with Kaplan–Meier curves. Details are provided on github: https://github.com/baydounm/UKB-paper21-supplementarydata

Fig. 2

Bubble plot displaying principal components analysis of 249 plasma metabolites, with varimax rotation (k = 15 principal components). Abbreviations: See list of abbreviations for protein abbreviations; AD = Alzheimer’s Disease; PRS = Polygenic Risk Score

Fig. 3

Volcano plot of plasma metabolomic biomarkers in relation to AD PRS: UK biobank 2006–2010. Abbreviations: See list of abbreviations for protein abbreviations; AD = Alzheimer’s Disease; PRS = Polygenic Risk Score. Note: Based on a series of multiple linear regression models, with main predictor being AD PRS and the outcome being each of 249 plasma metabolomic biomarkers (square-root transformed, z-scored). The y-axis is the predictor’s associated p-value on a -Log10 scale and the X-axis is the β coefficient (effect of AD PRS exposure on standardized z-scores of plasma metabolomic markers) from the multiple linear regression models. An estimate with a Bonferroni corrected p-value < 0.05 are marked by a different color and the plasma proteomic marker abbreviation is added for relatively stronger effect size of > 0.050 in absolute value (See UKB showcase URL: https://biobank.ndph.ox.ac.uk/showcase/). Details are provided on github: https://github.com/baydounm/UKB-paper21-supplementarydata

Fig. 4

Heatmap plots of percentage mediated from a series of 249 generalized structural equations models examining mediated effects of each metabolites for the total effect of the main AD PRS on all-cause dementia. Abbreviations: AD = Alzheimer’s Disease; PRS = Polygenic Risk Score. Notes: These are based on generalized structural equations models with each metabolite entered as an alternative mediator while adjusting for potential confounding exogenous variables. Red colored lines indicate a positive or consistent mediation whereby the indirect effect and the total effect have the same direction of association; Blue colored lines indicate a negative or inconsitent mediation whereby the indirect effect and the total effect have diverging directions of association (i.e. one is positive and the other is negative). Gray colored lines indicate non-significant mediating effect with p > 0.05. Details provided in supplementary datasheet 4 (nlcom_model 1: portion mediated; 2: total effect; 3: percentage mediated)