Stepwise connectivity patterns along the gradients of brain organization in Alzheimer's disease

187 

Abstract Submission 

Jazlynn Tan¹, Min Su Kang², Yi-Hsuan Yeh², Gleb Bezgin³, Nesrine Rahmouni⁴, Firoza Lussier⁴, Seok Jun Hong⁵, Jean-Paul Soucy⁶, Serge Gauthier⁴, Boris Bernhardt⁷, Sandra Black^2,8, Pedro Rosa-Neto^4,6, Maged Goubran^2,1,9, Julie Ottoy²

¹Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada, ²LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute, Toronto, Ontario, Canada, ³Neuroinformatics for Personalized Medicine lab, Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada, ⁴Translational Neuroimaging laboratory, McGill Centre for Studies in Aging, Montreal, Quebec, Canada, ⁵Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Gyeonggi-do, Republic of Korea, ⁶McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Quebec, Canada, ⁷Montreal Neurological Institute and Hospital, Montreal, Quebec, Canada, ⁸Department of Medicine (Division of Neurology), University of Toronto, Toronto, Ontario, Canada, ⁹Physical Sciences Platform, Sunnybrook Research Institute, University of Toronto, Toronto, Ontario, Canada

Jazlynn Tan  
Department of Medical Biophysics, University of Toronto
Toronto, Ontario, Canada

Min Su Kang  
LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute
Toronto, Ontario, Canada

Yi-Hsuan Yeh  
LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute
Toronto, Ontario, Canada

Gleb Bezgin  
Neuroinformatics for Personalized Medicine lab, Montreal Neurological Institute, McGill University
Montreal, Quebec, Canada

Nesrine Rahmouni  
Translational Neuroimaging laboratory, McGill Centre for Studies in Aging
Montreal, Quebec, Canada

Firoza Lussier  
Translational Neuroimaging laboratory, McGill Centre for Studies in Aging
Montreal, Quebec, Canada

Seok Jun Hong  
Center for Neuroscience Imaging Research, Institute for Basic Science
Suwon, Gyeonggi-do, Republic of Korea

Jean-Paul Soucy  
McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University
Montreal, Quebec, Canada

Serge Gauthier  
Translational Neuroimaging laboratory, McGill Centre for Studies in Aging
Montreal, Quebec, Canada

Boris Bernhardt  
Montreal Neurological Institute and Hospital
Montreal, Quebec, Canada

Sandra Black  
LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute|Department of Medicine (Division of Neurology), University of Toronto
Toronto, Ontario, Canada|Toronto, Ontario, Canada

Pedro Rosa-Neto  
Translational Neuroimaging laboratory, McGill Centre for Studies in Aging|McConnell Brain Imaging Centre, Montreal Neurological Institute and Hospital, McGill University
Montreal, Quebec, Canada|Montreal, Quebec, Canada

Maged Goubran  
LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute|Department of Medical Biophysics, University of Toronto|Physical Sciences Platform, Sunnybrook Research Institute, University of Toronto
Toronto, Ontario, Canada|Toronto, Ontario, Canada|Toronto, Ontario, Canada

Julie Ottoy, PhD  
LC Campbell Cognitive Neurology Unit, Hurvitz Brain Sciences Program, Sunnybrook Research Institute
Toronto, Ontario, Canada

In Alzheimer's Disease (AD), the entorhinal cortex (EC) is recognized as one of the earliest sites of tau tangle deposition. Existing studies have predominantly focused on tau propagation along direct (seed-to-target) neural connections between brain regions (Sepulcre et al. 2018). Here, we hypothesize that exploring indirect, multi-step connections adds new insights on the spread of AD in the brain. We first employ graph theory-based stepwise connectivity (Sepulcre et al. 2012) to elucidate multi-step functional and structural connections between the EC and the rest of the brain. We then implement a novel integration of stepwise connectivity with low-dimensional gradient space (Margulies et al. 2016) to elucidate connectivity trajectories along the major axes of functional and structural brain organization.

We acquired resting-state functional MRI (rs-fMRI) and diffusion-weighted MRI (dMRI) in 213 participants from the Translational Biomarkers in Aging and Dementia (TRIAD) cohort, including 103 cognitively normal Aβ-negative controls, 35 cognitively normal Aβ-positive (CN A+) and 75 cognitively impaired Aβ-positive (CI) participants. Subject-specific functional and structural connectomes were estimated using regional time series correlations (Esteban et al. 2019) and probabilistic fiber tractography (Tournier et al. 2019), respectively, with parcellations from a high-resolution atlas adapted from Glasser et al. (2016). We then employed functional or structural stepwise connectivity (SFC or SSC) analyses (Sepulcre et al. 2012) to unveil higher-order indirect connectivity patterns between the EC and the rest of the brain. The SFC or SSC value assigned to a region denotes the number of walks of a particular edge length (1 to 7 edges) to reach the EC from that region (Fig 1A). Groupwise (within-subject normalized) SFC/SSC values were compared via linear regression adjusted for age, sex, and APOE-ε4. Finally, we investigated these stepwise connectivity patterns within a coordinate system spanned by the principal components ('gradients') explaining the most variance in connectivity after non-linear dimensionality reduction (Margulies et al. 2016).

SFC was highest closest to the EC seed (step 1) and propagated to regions of the default-mode network at step 2 before shifting to sensorimotor regions at steps 3-7. SSC from the EC propagated from posterior (step 1-2) to anterior (step 3-7) regions (Fig 1B). Group comparisons revealed hypoconnectivity from the EC to temporal and posterior regions in CI compared to controls. Conversely, hyperconnectivity from the EC to frontoparietal and sensorimotor regions were observed in CI compared to controls (Fig 1C). In functional gradient space, CN A+ showed accelerated SFC propagation from the EC to the rest of the brain (Fig 2A, yellow pixels), which may be compensatory connectivity in preclinical stages. Later-stage CI subjects showed diminished SFC to the default-mode at the transmodal pole of gradient 1 (Fig 2A: green pixels; 2B: blue t-stats), with accelerated propagation to the sensorimotor regions at the unimodal pole of gradient 1 which were not revealed in the standard SFC analysis (Fig 2A: yellow pixels, 2B: red t-stats). Finally, in structural gradient space, propagation was restrained in the temporal-posterior pole of the structural gradient in CI compared to controls (Fig 2C: yellow pixels, 2D: blue t-stats).

Using a novel integrated stepwise connectivity and gradient approach, we demonstrated widespread network reorganization in AD affecting both short and long connections. Combining the stepwise connectivity and gradient space allows new insight previously inaccessible through conventional analyses in anatomical space. It unveils how AD affects connectivity strength along the major axes of brain organization.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Connectivity (eg. functional, effective, structural) ²

Degenerative Disease

Modeling

MRI

Other - gradient; Alzheimer’s disease; stepwise connectivity; neuroimaging; multi-modal