Tau accumulation and degeneration differ across functional networks in atypical Alzheimer's disease phenotypes

Background Recent models of Alzheimer's disease (AD) posit that pathogenic tau propagates through interconnected structural networks, suggesting amnestic and non‐amnestic AD (aAD & naAD) phenotypes could reflect differential involvement of functional networks. We tested this hypothesis by contrasting longitudinal grey matter (GM) atrophy and tau accumulation across AD clinical phenotypes in 7 canonical functional networks. Methods We analyzed longitudinal 3‐Tesla T1‐weighted MRI from 122 patients [aAD=49; logopenic primary progressive aphasia (lvPPA)=31; posterior cortical atrophy (PCA)=23; behavioral/dysexecutive AD (bvAD)=6; corticobasal syndrome (CBS)=6; mixed naAD=4] with autopsy or CSF evidence of AD pathology and 91 cognitively normal controls (Table 1). Thirty‐one patients (aAD=9; lvPPA=7; PCA=9; bvAD=2; CBS=3; naAD=1) had longitudinal flortaucipir PET (FTP‐PET) data. We computed baseline values and annualized change for GM volume and FTP‐PET standardized uptake value ratios (SUVRs) in the 100‐label, 7‐network parcellation of Schaefer et al. (2018). Linear mixed effects models (α=0.05, corrected) adjusting for MMSE assessed group differences in atrophy and tau accumulation. Results All groups exhibited significant baseline and longitudinal atrophy vs. controls in multiple networks (Figs. 1‐2). Differences in longitudinal atrophy rate (Fig. 2) were observed in the dorsal attention network [PCA>lvPPA, t(371)=3.1, paAD, t(371)=3.1, paAD, t(371)=3.3, pbvAD, t(371)=3.7, p