Intrinsic Structural Covariation Links 28 Cerebellum Sub-Regions to the Cerebral Cortex

1957 

Abstract Submission 

Zilong Wang¹, Jörn Diedrichsen², Karin Saltoun¹, Christopher Steele^3,4, Sheeba Arnold-Anteraper^5,6,7, Jeremy Schmahmann⁸, Danilo Bzdok¹

¹McConnell Brain Imaging Centre (BIC), Montreal Neurological Institute (MNI), McGill University, Montréal, Quebec, ²The Brain and Mind Institute, University of Western Ontario, London, Ontario, ³Department of Psychology, Concordia University, Montréal, Quebec, ⁴Department of Neurology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ⁵Department of Psychology, Northeastern University, Boston, MA, ⁶Carle Foundation Hospital, Urbana, IL, ⁷Alan and Lorraine Bressler Clinical and Research Program for Autism Spectrum Disorder, Massachusetts General Hospital, Boston, MA, ⁸Ataxia Center, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA

Zilong Wang  
McConnell Brain Imaging Centre (BIC), Montreal Neurological Institute (MNI), McGill University
Montréal, Quebec

Jörn Diedrichsen  
The Brain and Mind Institute, University of Western Ontario
London, Ontario

Karin Saltoun  
McConnell Brain Imaging Centre (BIC), Montreal Neurological Institute (MNI), McGill University
Montréal, Quebec

Christopher Steele  
Department of Psychology, Concordia University|Department of Neurology, Max Planck Institute for Human Cognitive and Brain Sciences
Montréal, Quebec|Leipzig, Germany

Sheeba Anteraper, PhD  
Department of Psychology, Northeastern University|Carle Foundation Hospital|Alan and Lorraine Bressler Clinical and Research Program for Autism Spectrum Disorder, Massachusetts General Hospital
Boston, MA|Urbana, IL|Boston, MA

Jeremy Schmahmann  
Ataxia Center, Department of Neurology, Massachusetts General Hospital and Harvard Medical School
Boston, MA

Danilo Bzdok  
McConnell Brain Imaging Centre (BIC), Montreal Neurological Institute (MNI), McGill University
Montréal, Quebec

The cerebellum, long associated solely with motor control and skill acquisition, is now understood to play a role in non-motor and higher cognitive functions¹. Tract-tracing studies in monkeys have shown distinct cortico-cerebellar circuits for areas like the prefrontal and somatomotor cortex², but similar invasive studies are infeasible in humans and impossible to scale to the whole brain³. Evolutionarily, the associative cortical and cerebellar regions enlarge in humans, compared to other species⁴. Human development aligns with evolutionary evidence such that regions supporting higher cognition in the cortex and the cerebellum mature later in life⁵. However, this "big" and "small" cortex of the human brain were routinely studied in isolation. Efforts to parcel the cerebellum mainly take a winner-takes-all approach⁶, allowing only one possibility about the functional organization of the cerebellum, which may lead to an incomplete view. We hypothesize that there are multiple structural variation patterns that occur simultaneously and are spread throughout the cortex-cerebellum complex. We also hypothesize that each cerebellar subregion can be linked to several brain phenomena at the same time, and vice versa.

We leveraged structural whole-brain scans and 977 in-depth phenotype measurements for 38527 individuals from UKBB cohort. We parcellated the entire cerebellar cortex into 28 regions with a new functional atlas⁷, and segmented the cerebral cortex into 100 subregions ⁸. Using partial least squares regression (PLSR), we estimated their structural covariation patterns (modes) at the population level. Each derived mode characterized a latent cortical score and cerebellar score that represented linear combinations of the original cortical and cerebellar measurements with maximal covariance. Robustness of the modes was determined by permutation test. We further profiled the key phenotypes associated with each mode's latent variables to understand their real-world implications by means of phenome-wide association assays.

Our analysis uncovered three significant population-level cortex-cerebellum covariation modes. The first and most explanatory mode revealed the interplay between broad higher-associative regions, excluding dorsal attention network (DAN), and visual, sensorimotor regions in both the cortex and cerebellum (Fig.1A). The cortical latent variable in the first mode was linked to complex reasoning, cardiovascular dieases, while the cerebellar latent variable was associated with psychomotor speed, physical activity and angiogenesis (Fig.2A). The second mode contrasted visual regions and key nodes in DAN with frontal, anterior temporal associative regions implicated in default, limbic, executive control and salience network (Fig.1B). Both its latent variables were strongly associated with watching TV and complex reasoning (Fig.2B). The third mode showed an ipsilateral pattern such that each side of cerebellum varied in the same direction as the ipsilateral side of the cortex, with few exceptions (Fig.1C).

Our first two modes are consistent with the higher-lower divergence of neural systems, as well as the classical double motor representation and the recently proposed triple nonmotor representation in the cerebellum ⁹. Our results support the anticorrelation between visual-attention and other higher order cognitive systems in the cerebellum, like in the cortex. Our third mode indicates a greater proportion of ipsilateral mechanisms between the cerebellum and cortex that may be overshadowed by the contralateral pathways. The distinct phenotype profiles for each mode and their latent variables revealed unique brain phenomena - behavior links that weighed differently in the cerebellum and the cortex. These findings greatly contribute to our understanding of the intricate interplay among the cortex, cerebellum and behaviors.

fMRI Connectivity and Network Modeling ²

Multivariate Approaches ¹

Anatomy and Functional Systems

Cortical Anatomy and Brain Mapping

Brain Atlases

Cerebellum

Computational Neuroscience

Cortex

Data analysis

Machine Learning

MRI

Phenotype-Genotype

Statistical Methods

Structures

Systems