Fig. 2

Changes of the mtDNAcn are primarily associated with tau in the DLPFC and TDP-43 in the PCC. (A) The bars indicate the mtDNAcn’s variance explained (partial R²) by different pathologies, cognitive measures, and demographics in the ROSMAP cohort. Each Variable was analyzed separately in a regression model with mtDNAcn as outcome adjusted for sex and age. The cognitive variables were additionally adjusted for education. The results for sex and age were obtained from a model adjusted for global AD pathology. Colors indicate brain regions. Asterisks indicate significance levels obtained by F-tests (*** for p ≤ 0.001, ** for p ≤ 0.01, * for p ≤ 0.05, and · for p ≤ 0.1). The sample size available for each pathology and brain region is denoted on the y axis. (B) Boxplot shows the mtDNAcn in the PCC for different stages of TDP-43 pathology. Wilcoxon rank-sum test was used to calculate p values. (C) The bars indicate the mtDNAcn’s variance explained (partial R²) by cell type proportions estimated from RNA-seq data in a subset of n = 327 DLPFC samples. Each cell type proportion was analyzed separately in a regression model with mtDNAcn as outcome adjusted for sex and age. Significance levels were obtained and indicated by asterisks as in (A). (D) Forest plot shows the result from a multivariable regression model with mtDNAcn in the DLPFC as outcome and the pathologic and demographic variables denoted on the y axis as explanatory variables. Estimated coefficients are shown as dots and the line segments represent the respective 95% confidence intervals. Continuous variables were z-standardized. Categorical variables were dichotomized and the factor level and case numbers corresponding to the plotted coefficient are denoted in brackets under the variable name. t-test was applied to calculate p values. A total of n = 288 cases with complete observations of all variables were used to fit the model