Graph Theory Analysis of Structural Connectivity in Limbic‐predominant Age‐related TDP‐43 Encephalopathy Neuropathological Change (LATE‐NC)

Abstract

AbstractBackgroundThe anatomical distribution of limbic‐predominant age‐related TDP‐43 encephalopathy neuropathological change (LATE‐NC) typically involves the amygdala, hippocampus, and other subcortical, temporal and frontal lobe regions. But its impact on the structural integrity of this network of regions remains unknown. In this study, we tested the hypothesis that LATE‐NC is associated with abnormalities in graph theoretic measures of structural connectivity in the network of regions known to be involved in LATE‐NC.MethodThis study included 148 deceased older adults participating in Rush Memory and Aging Project and Religious Orders Study (Fig.1A). Ex‐vivo diffusion tensor imaging (DTI) scans were acquired on 3T scanners, followed by detailed neuropathologic examination by a board‐certified neuropathologist (Fig.1B). Fractional anisotropy (FA) maps derived from DTI scans were used to define region‐to‐region structural connectivity in a network of 19 regions known to be involved in LATE‐NC using graph theory‐based network measures (Fig.2). General linear models were used to investigate the association of global and local graph measures with LATE‐NC stages, controlling for other neuropathologies (Alzheimer’s disease, Lewy bodies, arteriolosclerosis, atherosclerosis, cerebral amyloid angiopathy, gross and microscopic infarcts), demographics (age at death, sex, education), total white matter hyperintensity volume, postmortem intervals and scanner variables. Statistical significance was set at p<0.05.ResultLinear regression revealed that greater LATE‐NC stage was associated with increased characteristic path length and with lower values of the remaining global network measures (p<0.05) (Fig.3A). LATE‐NC was also associated with lower local efficiency, increased path length, lower nodal strength, and lower clustering coefficient in 74%, 79%, 32%, and 63% of the nodes, respectively (p<0.05) (Fig.3B). Finally, LATE‐NC was associated with both greater (in 21% of nodes) and lower (in 16% of nodes) eigenvector centrality, and with both greater and lower betweenness centrality in 21% of nodes, respectively (p<0.05) (Fig.3B).ConclusionUsing graph theory and a hypothesis‐driven approach, we found significant degradation of structural integrity in white matter of the network of regions affected by LATE‐NC. These findings may shed light on the effects of LATE‐NC on brain connectivity, and in combination with other imaging and clinical information, may aid towards the development of tools for in‐vivo detection of this devastating neuropathology.