Identifying heritable brain networks that may confer cognitive reserve to Alzheimer's disease

Abstract

BackgroundAlzheimer’s disease (AD) is a neurodegenerative disease that affects 50 million people worldwide (Breijyeh & Karaman, 2020). Among individuals with AD, cognitive dysfunction varies, giving rise to the hypothesis that neural function and anatomical substrates provide “cognitive reserve” against AD. This idea inspired the brain reserve hypothesis, which states that structural features of the brain may protect against AD‐related cognitive decline (Stern, 2012). One potential source of brain reserve is neural microanatomy, which varies across individuals (Wang et al., 2020) and may then explain the differential cognitive outcomes in AD associated with cognitive reserve.MethodWe utilized a population of genetically distinct AD mice produced by crossing the 5XFAD model of AD to five mouse strains (C57BL/6J, DBA/2J, FVB/NJ, CAST/EiJ, and WSB/EiJ). At 3 months of age, these mice underwent diffusion magnetic resonance imaging (dMRI), a collection of techniques that characterizes brain microstructure (Campbell & Pike, 2019), to probe variation of neural microanatomy in 430 regions of interest (ROIs). At 5 months of age, when cognitive decline is expected (Oakley et al., 2006; Devi & Ohno, 2010), these mice underwent contextual fear conditioning (CFC) and contextual fear memory (CFM) paradigms. Hierarchical clustering was performed on the correlations between dMRI measurements in each brain region influenced by strain and CFC and CFM performance.Result362 ROIs had dMRI metrics with significant effects of sex, strain, and/or 5XFAD genotype. Of these ROIs, 348 had a significant effect of strain, 68 had a significant effect of sex, and 30 or fewer had a significant effect of genotype or any variable interactions. We identified one network of ROIs in which nearly all dMRI parameters correlated positively and strongly to CFC performance and two networks whose dMRI measures correlated strongly to CFM performance, although the directionality of these correlations varied within these two networks.ConclusionOur work identifies three neural networks that are influenced by strain and correlate strongly to CFC and CFM performance. Further analyses should explore genetic factors influencing these networks and the structural and functional connectivity of brain regions in these networks.