Situating tau pathology and neuroinflammation along the principal gradients of brain organisation in Alzheimer’s disease

Abstract

AbstractBackgroundThe spread of abnormal tau along connectivity‐based networks is a hallmark of Alzheimer’s disease (AD). In addition, activated microglia was shown to affect tau spread across networks. However, traditional network approaches using atlas‐defined regions suffer from assuming within‐region homogenous connectivity, as well as abrupt and linear connectivity changes between the regions; while, in fact, brain structure and function exist as complex overlapping axes of connectivity variation. Here, we studied smooth/continuous spatial transitions or ‘gradients’ of connectivity and their link to gradients of pathology in AD. This work provides novel insights into brain architecture driving pathology spread.MethodWe enrolled 220 individuals including 106 cognitively normal Aβ‐negative (A‐) and 37 A+, and 77 MCI/AD A+ (TRIAD). From diffusion‐weighted MRI and fMRI data, we estimated region‐to‐region fiber count and correlated timeseries, respectively, and averaged subject‐wise connectivity matrices to create structural and functional ‘template connectomes’. From 18F‐MK6240 and 11C‐PBR28 PET data, we extracted regional tau and neuroinflammation SUVR values, respectively, and created covariance matrices for each disease stage. Next, the template connectomes/covariance matrices were subjected to non‐linear dimensionality reduction to extract embedded components (‘gradients’). We correlated the primary gradients of functional or structural connectivity (G1FC/G1SC) with the primary and secondary gradients of tau or neuroinflammation (G1TAU/G2TAU; G1INFLAM/G2INFLAM). We employed three brain atlases for validation: high‐resolution customized Glasser‐atlas (1332 equally‐sized subregions), DKT, and Schaefer‐100.ResultG1FC replicated the previously reported ‘uni‐to‐transmodal’ functional gradient (Margulies 2016) and correlated strongly with G1TAU in cognitively impaired A+ (Fig.1 A); illustrating a common topographic variation in tau spread and functional connectivity in symptomatic stages. Conversely, G1SC ran between medial‐temporal‐lobe (MTL)/posterior and frontal/anterior regions, and correlated strongly with both G2TAU (Fig.1 B) and G1INFLAM (Fig.1 C) in cognitively (un)impaired A+; illustrating a common topographic variation in tau/microglia and structural connectivity already present at early disease stages. Different brain atlases led to similar results.ConclusionWe identified two gradients of continuous topographic variation in tau and neuroinflammation, potentially reflecting heterogeneity of AD‐subgroups. While microglial activation and early‐stage tau propagation may be largely dominated by structural organisation, later‐stage tau propagation may be more heterogeneously associated with variation in structural and particularly functional connectivity.