Combining network spread with protein aggregation correctly recapitulates empirical spatio-temporal progression of Alzheimer's tau pathology.

Abstract

Alzheimer's disease involves widespread and progressive deposition of misfolded protein tau, likely via "prion-like" trans-neuronal transmission along the brain's structural connectome. It is not well understood how tau aggregation and spread leads to stereotypical progression in the Alzheimer brain. Here we extend our previous Network Diffusion Model (NDM, [1]) to include these critical processes (see Figure 1): a) Tau monomer seeding and production in the entorhinal cortex; b) aggregation of monomers into larger oligomers and then tangles using well-known Smoluchowski equations that were previously successful for amyloid beta and prions [2]. c) These processes were integrated into NDM, whereby anatomic connections govern the transmission of tau between two distant but connected brain regions. This extended model, which we call Aggregation-Network-Diffusion (AND), was then thoroughly tested on empirical data (Table 1): MRI-derived atrophy, AV1451-PET and CSF biomarkers (Ab42, pTau) from ADNI and Korea cohort [3]. AND model exhibits key hallmarks of AD, accurately capturing the spatial distribution of empirical tau (Korea cohort) as well as regional atrophy of ADNI cohort (Figure 2). R values shown in figure are all highly significant (p<0.0001). Next we interrogated different regions serving as seeding site; the best cortical seeding site was entorhinal cortex, while the best subcortical seeding was in amygdala followed by hippocampus. We also fitted the AND model's total tau PET against patients' cerebrospinal fluid phosphorylated tau (p-tau181) profiles as a function of disease progression; we obtained R = 0.37, p < 1e-36. Sensitivity analysis suggests that diffusion rate is far more important than monomer production rate in governing long-term pathology accumulation (Figure 3). This unified quantitative and testable model has the potential to explain observed phenomena. It may also serve as a test-bed for future hypothesis generation and testing in silico, e.g. to mathematically test whether monomer production, aggregation or diffusion is the key bottleneck.