Mediation Analysis Identifies Amyloid Imaging Genetic Patterns for Cognitive Outcomes in Alzheimer’s Disease

Abstract

AbstractBackgroundPrevious genetic studies of Alzheimer’s Disease (AD) utilized features derived from florbetapir (AV45) PET imaging (Lee et al., BMC Genomics 85(2022)). Genetic analysis of florbetaben (FBB) PET FreeSurfer‐defined ROI‐specific SUVrs provided by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) can extend this work by nominating ROIs likely to play a specific role in mediating genetic effects on AD‐related cognitive decline. Nominated ROIs may be suitable targets for biomedical interventions that slow the progression of AD.MethodGenetic and neuroimaging data were obtained from the ADNI 3 cohort. Amyloid burden measurements (SUVrs) were extracted and intensity‐normalized using the provided whole‐cerebellum reference region SUVr (Landau et al., 2021 ADNI). Imputed genotyping data included 5,402,690 variants from 308 individuals. Genome‐wide association studies predicted FBB changes in each of 68 FreeSurfer cortical ROIs utilizing age, gender, years of education, and the first 10 principal components as covariates. Within each individual imaging feature, FDR‐corrected P value thresholds determined significance (ɑ = 0.05). Mediation analysis combined identified SNP‐imaging feature associations with clinical cognitive score data to determine if changes in amyloid burden mediate genetic effects on clinical outcomes.Result149 unique SNPs and 28 unique imaging features were implicated in 881 FBB SNP‐feature associations (P values as low as 8.97e‐9). Although most associations are linked to known AD genes, a number of SNPs on chromosomes 18, 15, 14, 12, 6, 4, and 2 were also identified (Fig 1). Significant GWAS associations were carried forward to identify 83 mediation relationships (Bonferroni P < 1.56e‐6). An especially large number of significant causal mediation relationships implicated the entorhinal cortex and parahippocampal gyri, both of which play a significant role in learning and memory (Fig 2).ConclusionGenetic association traced some of the variation in FBB measures of amyloid burden to genetic drivers throughout the genome. Mediation analysis confirms that the FBB measures in the entorhinal cortex and parahippocampal gyri mediate the effects of some genetic findings on learning and global cognition. Future analyses of FBB measures as imaging endophenotypes may inspire pathway analyses and integrate additional clinical or biological information.