Cerebello-basal Ganglia Networks and Cortical Network Global Efficiency

Abstract

The cerebellum (CB) and basal ganglia (BG) each have topographically distinct functional subregions that are functionally and anatomically interconnected with cortical regions through discrete thalamic loops and with each other via disynaptic connections, with previous work detailing high levels of functional connectivity between these phylogenetically ancient regions. It was posited that this CB-BG network provides support for cortical systems processing, spanning cognitive, emotional, and motor domains, implying subcortical network measures are strongly related to cortical network measures (Bostan & Strick, 2018); however, it is currently unknown how network measures within distinct CB-BG networks relate to cortical network measures. Here, 122 regions of interest comprising cognitive and motor CB-BG networks and 7 canonical cortical resting-state were used to investigate whether the integration (quantified using global efficiency, GE) of cognitive CB-BG network (CCBN) nodes and their segregation from motor CB-BG network (MCBN) nodes is related to cortical network GE and segregation in 233 non-related, righthanded participants (Human Connectome Project-1200). CCBN GE positively correlated with GE in the default mode, motor, and auditory networks and MCBN GE positively correlated with GE in all networks except the default mode and emotional. MCBN segregation was related to MN segregation. These findings highlight the CB-BG network's role in executive function, task switching, and verbal working memory. This work has implications for understanding cortical network organization and cortical-subcortical interactions in healthy adults and may help in deciphering subcortical differences seen in disease states.

Cerebellar and basal ganglia motor network predicts trait depression and hyperactivity

“…We also replicated and extended our prior work detailing correlations between CB-BG and cortical networks (Jackson and Bernard, 2022) by expanding the sample size 3-fold and introducing additional controls to probe the network relationships. We found CCBN GE strongly predicted DMN and AN GE, moderately predicted MN GE, and did not predict EN GE, largely consistent with our previous findings and suggestive of CCBN involvement in self-referential thought and auditory/verbal working memory (Fox et al, 2015).…”

“…The cerebellum (CB) and basal ganglia (BG) are each cytoarchitectonically unique and thought to be foundational to cortical processing, each supporting cognition with specific computations associated with predictive learning (Flanagan and Wing, 1997;Morton and Bastian, 2006;Wolpert et al, 2011) and reinforcement learning (Schultz et al, 1998(Schultz et al, , 2000Cohen and Frank, 2009;Ito and Doya, 2011), respectively. Previous work by our lab found a highly-connected resting-state CB-BG network in healthy young adults (Hausman et al, 2019) and a follow-up study using graph theory metrics confirmed distinct motor and cognitive functional networks that were differentially related to cortical functional networks (Jackson and Bernard, 2022). Notably, the CB and BG have each been separately implicated in multiple psychiatric and neurological disorders (e.g., depression, bipolar disorder, psychosis; reviewed in Ring, 2002;Hoppenbrouwers et al, 2008;Macpherson and Hikida, 2019).…”

“…Additionally, due to associations between working memory tasks and motor systems (Balsters et al, 2013;Von der Gablentz et al, 2015;Bernard et al, 2020;Clark et al, 2020), we hypothesized MCBN GE would predict scores on working memory tasks. Additionally, we previously found CB-BG GE is associated with the GE of canonical resting-state networks in a sample with no history of mental illness, illicit drug use, or alcohol and marijuana abuse (Jackson and Bernard, 2022). Here, we looked to replicate and extend these findings by broadening the selection criteria to include all subjects in the HCP 1200 release and by using a regression model to allow for the control of other potential explanatory variables such as alcohol and drug abuse.…”

“…Additionally, to account for the use of both left and right ROIs in some cortical networks and for the unilateral nature of subcortical networks (Pelzer et al, 2017), all lateralized (defined as more than 3 mm from the midline in the x direction) ROIs without a reported contralateral match were mirrored in the x direction. Supplementary Table 1 provides details for the 122 resulting ROIs (for a visual depiction of their locations, please see Jackson and Bernard, 2022).…”

“…To compute graph theory components, edges were defined as thresholded correlation coefficients and nodes were defined as the ROIs in each network as described above (Stein et al, 2007;Fair et al, 2009;Geerligs et al, 2015;Hausman et al, 2019). We previously investigated differences in network measures when different thresholds were used and found no substantial differences; as such, we used β > 0.1 to define edges, a correlation coefficient threshold commonly used in the literature (i.e., Fair et al, 2009) and consistent with prior work (Jackson and Bernard, 2022). To define interconnectedness, networklevel GE, a graph theory metric of how efficiently information in a network travels, was computed separately for each of the nine networks (Latora and Marchiori, 2001;Achard and Bullmore, 2007;Whitfield-Gabrieli and Nieto-Castanon, 2012;Sheffield et al, 2016).…”