Networks of tau distribution in Alzheimer's disease

Abstract

See Whitwell (doi:10.1093/brain/awy001) for a scientific commentary on this article.A stereotypical anatomical propagation of tau pathology has been described in Alzheimer's disease. According to recent concepts (network degeneration hypothesis), this propagation is thought to be indicative of misfolded tau proteins possibly spreading along functional networks. If true, tau pathology accumulation should correlate in functionally connected brain regions. Therefore, we examined whether independent components could be identified in the distribution pattern of in vivo tau pathology and whether these components correspond with specific functional connectivity networks. Twenty-two 18F-AV-1451 PET scans of patients with amnestic Alzheimer's disease (mean age = 66.00 ± 7.22 years, 14 males/eight females) were spatially normalized, intensity standardized to the cerebellum, and z-transformed using the mean and deviation image of a healthy control sample to assess Alzheimer's disease-related tau pathology. First, to detect distinct tau pathology networks, the deviation maps were subjected to an independent component analysis. Second, to investigate if regions of high tau burden are associated with functional connectivity networks, we extracted the region with the maximum z-value in each of the generated tau pathology networks and used them as seeds in a subsequent resting-state functional MRI analysis, conducted in a group of healthy adults (n = 26) who were part of the 1000 Functional Connectomes Project. Third, to examine if tau pathology co-localizes with functional connectivity networks, we quantified the spatial overlap between the seed-based networks and the corresponding tau pathology network by calculating the Dice similarity coefficient. Additionally, we assessed if the tau-dependent seed-based networks correspond with known functional resting-state networks. Finally, we examined the relevance of the identified components in regard to the neuropathological Braak stages. We identified 10 independently coherent tau pathology networks with the majority showing a symmetrical bi-hemispheric expansion and coinciding with highly functionally connected brain regions such as the precuneus and cingulate cortex. A fair-to-moderate overlap was observed between the tau pathology networks and corresponding seed-based networks (Dice range: 0.13-0.57), which in turn resembled known resting-state networks, particularly the default mode network (Dice range: 0.42-0.56). Moreover, greater tau burden in the tau pathology networks was associated with more advanced Braak stages. Using the data-driven approach of an independent component analysis, we observed a set of independently coherent tau pathology networks in Alzheimer's disease, which were associated with disease progression and coincided with functional networks previously reported to be impaired in Alzheimer's disease. Together, our results provide novel information regarding the impact of tau pathology networks on the mechanistic pathway of Alzheimer's disease.

Keywords: PET; independent component analysis; resting-state networks; tau pathology networks.