A polygenic risk score derived from episodic memory weighted genetic variants is associated with cognitive decline in preclinical Alzheimer's disease

Studies of Alzheimer's disease risk-weighted polygenic risk scores (PRSs) for cognitive performance have reported inconsistent associations. This inconsistency is particularly evident when PRSs are assessed independent of APOE genotype. As such, the development and assessment of phenotype-specific weightings to derive PRSs for cognitive decline in preclinical AD is warranted. To this end a episodic memory-weighted PRS (emPRS) was derived and assessed against decline in cognitive performance in 226 healthy cognitively normal older adults with high brain A beta-amyloid burden participants from the Australian Imaging, Biomarkers and Lifestyle (AIBL) study. The effect size for decline in a verbal episodic memory was determined individually for 27 genetic variants in a reference sample (n = 151). These were then summed to generate a emPRS either including APOE (emPRS((c) over bar APOE)) or excluding APOE (emPRS((s) over bar APOE)). Resultant emPRS were then evaluated, in a test sample (n = 75), against decline in global cognition, verbal episodic memory and a pre-Alzheimer's cognitive composite (AIBL-PACC) over 7.5 years. The mean (SD) age of the 226 participants was 72.2 (6.6) years and 116 (51.3%) were female. Reference and test samples did not differ significantly demographically. Whilst no association of emPRSs were observed with baseline cognition, the emPRS((c) over bar APOE) was associated with longitudinal global cognition (-0.237, P = 0.0002), verbal episodic memory (-0.259, P = 0.00003) and the AIBL-PACC (-0.381, P = 0.02). The emPRS((s) over bar APOE) was also associated with global cognition (-0.169, P = 0.021) and verbal episodic memory (-0.208, P = 0.004). Stratification by APOE epsilon 4 revealed that the association between the emPRS and verbal episodic memory was limited to carriage of no epsilon 4 or one epsilon 4 allele. This was also observed for global cognition. The emPRS and rates of decline in AIBL-PACC were associated in those carrying one epsilon 4 allele. Overall, the described novel emPRS has utility for the prediction of decline in cognition in preclinical AD. This study provides evidence to support the further use and evaluation of phenotype weightings in PRS development