Altered brain structural networks in the APP/PS1 mice: evidence from multi-shell diffusion imaging

• ¹ University of Antwerp, Belgium

Alzheimer’s disease is one of the most common neurodegenerative diseases characterized by cognitive deficits, memory loss and brain cerebral atrophy. Numerous studies suggested that Amyloid β (Aβ) and tau protein are key elements during the development of the disease. The APP/PS1 is one of the most widely used transgenic mouse models in AD-related research. In this model, the expression of human APP transgene is almost 3-fold higher than endogenous murine APP and mice develop many AD-like deficits including white matter alterations that underlie the connectivity patterns of the brain. Diffusion weighted magnetic resonance imaging (dwMRI) is a technique allowing the investigation of the brain white matter microstructures. Application of graph theoretical approaches to dwMRI provides a new approach to identify the altered network properties reflected in graph parameters including small-world attributes, global and nodal properties and rich club organization. In humans, many studies already demonstrated the altered network properties in AD patients. However, few studies used graph theory to detect network differences in mouse models to study therapeutic approaches. In this study, we used 19 male transgenic APP/PS1 mice and 20 male WT C57BL6/J mice at age of 8 months as subjects and acquired seven b-values (400, 800, 1200, 1600, 2000, 2400, and 2800 s/mm2), a total of 60 different diffusion gradient directions for each b-value, using a 7T Pharmascan MRI scanner (Bruker, Germany). Diffusion imaging processing and analysis were conducted using MRtrix3 and ANTs and whole brain tractography was terminated if angle > 45° or fractional anisotropy (FA) < 0.2. Then, we used a brain template with 236 ROIs to define network nodes and used the logarithm of the estimated numbers of streamlines (NOS) between nodes as the weights of network connections. In this way, we constructed the average network for each mouse group. Then, we subdivided the areas into hub and non-hub regions based on their degree and into five brain modules (Isocortex, Pallium, Sub-pallium, Diencephalon, Midbrain and Hindbrain) based on previous literature and compared the connectivity patterns between each type of connection (rich club, feeder, local). We found that the brain structural networks in both groups (APP/PS1 and WT) demonstrated small-world properties. Comparing across groups, the clustering coefficient and normalized shortest path length of the APP/PS1 mice were significantly decreased in comparison to WT. In addition, nodal properties of some areas such as the somatosensory and cingulate cortex demonstrated increased nodal strength in the AD-model mice.