Tau protein spreads through functionally connected neurons in Alzheimer’s disease
Deborah N Schoonhoven,Emma M Coomans,Ana P Millan,Anne M van Nifterick,Denise Visser,Rik Ossenkoppele,Hayel Tuncel,Wiesje M. van der Flier,Sandeep SV Golla,Philip Scheltens,Arjan Hillebrand,Bart NM van Berckel,Cornelis J Stam,Alida A Gouw
DOI: https://doi.org/10.1002/alz.066494
2022-12-01
Abstract:Recent studies in Alzheimer’s disease (AD) suggest that tau proteins spread through the brain following neuronal connections. Several mechanisms could be involved in this process: spreading between brain regions that interact strongly (functionally connected); through the pattern of anatomical connections (structural connectivity); or simple diffusion to spatially adjacent regions (Euclidean distance (EC)). We investigated this by modelling the tau‐spreading process on these different networks, and compared the modelled tau‐depositions with tau‐depositions at several stages of the AD continuum as measured with in‐vivo 18F‐flortaucipir PET. We collected 10‐minutes eyes‐closed, whole‐head magnetoencephalography (MEG) recordings from 82 subjects (Table‐1) with and without Aβ‐pathology (CSF‐Aβ42 and/or amyloid‐PET) from the Amsterdam Dementia Cohort. Subjects were classed as controls (subjective cognitive decline (SCD) Aβ‐,n=25), preclinical AD (SCD Aβ+,n=16), mild cognitive impairment (MCI) due to AD (n=16) and probable AD dementia (n=25). Tau propagation was modelled as an epidemic process (susceptible‐infected (SI) model) on structural networks (Exponential Distance Rule (EDR) or EC), or functional networks (corrected‐amplitude envelope correlation (AEC‐c) in alpha (8‐13Hz) and beta (13‐30Hz) bands) derived from source‐reconstructed MEG data. The group‐level control network was used as backbone for the SI‐model to predict tau spread in the next stages of the AD continuum; parameters were tuned to produce an optimum correlation to the group‐specific tau propagation patterns as measured with dynamic 130‐minutes 18F‐flortaucipir PET, according to 50 regions‐of‐interest (ROIs) in an adjusted version of the probabilistic Hammers atlas. Tau propagation was modelled to start from one seed region (ROI14), which was selected based on its reliability to accurately measure tau deposition. The functional network predicted tau‐spread with highest accuracy in the preclinical AD stage (AEC‐c alpha r=0.59; AEC‐c beta r=0.58), followed by the structural network (r=0.45) and simple diffusion (r=0.44), Figure‐1, Figure‐2, Figure‐3. Prediction accuracy declined for the MCI and AD dementia stages (AEC‐c alpha r=0.50 and r=0.33; AEC‐c beta r=0.40 and r=0.37; structural r=0.29 and r=0.25; diffusion r=0.29 and r=0.25, respectively). Using a relatively simple SI‐model we predicted tau propagation in preclinical AD with an accuracy of approximately 60%. Our results suggest that tau spreads through functional connections, rather than structural connections or simple diffusion.
clinical neurology