Protective effects of mitochondrial genome abundance on Alzheimer’s Disease: a Mendelian randomization study

Abstract Background Mitochondrial DNA copy number (mtDNAcn) is a surrogate measure of mitochondrial function that is associated with higher cognitive performance and reduced risk of AD neuropathology. Here, we use two‐sample Mendelian randomization (MR) to estimate the causal association of blood‐based mtDNAcn on AD/dementia while controlling for latent confounders, sample overlap, and testing bidirectional effects. Method GWAS Summary statistics of mtDNAcn were procured from three separate studies using various bioinformatic methods to estimate mtDNAcn; microarray, whole exome sequencing (WES), and whole genome sequencing (WGS) datasets using the UK Biobank (UKBB) and Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium. Longchamps et al., 2021. uses quantitative PCR (CHARGE) and WES, WGS (UKBB). Hagg et al., 2021 use microarray approximation, as well as WGS and qPCR. Chong et al., 2022 employ the novel AutoMitoC in addition to WES data (UKBB). Kunkle et al., 2019 employed late‐onset clinical AD cases/controls (n = 94,437), and Bellenguez et al., 2022 (n = 487,511) used AD/dementia from clinical and proxy cases/controls. MR estimates the causal association between mtDNAcn and AD and AD/dementia using IVW analysis, and MR Egger, Weighted Mode, and Weighted Median as sensitivity analysis. Latent Heritable Confounder‐MR (LHC‐MR) was used to investigate bidirectional effects in the presence of sample overlap between the mtDNAcn datasets and AD/Dementia. Result There were no significant causal associations of mtDNAcn on AD or AD/dementia using univariate MR analysis. However, using LHC‐MR, higher genetically predicted mtDNAcn was causally associated with AD/dementia using all three mtDNAcn datasets (OR [95%CI]: Chong = 0.84 [0.78, 0.90], p < 0.001; Longchamps = 0.89 [0.82, 0.96], p = 0.003; Hagg = 0.84 [0.78, 0.91], p < 0.001). In the reverse direction, genetic liability for AD/dementia was causally associated with higher mtDNAcn levels (β [SE]: 0.10 [0.05], p = 0.017). Conclusion Higher blood‐based mtDNAcn was causally associated with reduced risk of AD, with limited evidence to suggest a bidirectional effect. Discrepancies between univariate MR results and LHC‐MR may be explained by the increased power and the inclusion of a latent confounder.