Topographical functional correlates of interindividual differences in executive functions in young healthy twins

Abstract

Executive functions (EF) are a set of higher-order cognitive abilities that enable goal-directed behavior by controlling lower-level operations. In the brain, those functions have been traditionally associated with activity in the Frontoparietal Network, but recent neuroimaging studies have challenged this view in favor of more widespread cortical involvement. In the present study, we aimed to explore whether the network that serves as critical hubs at rest, which we term network reliance, differentiate individuals as a function of their level of EF. Furthermore, we investigated whether such differences are driven by genetic as compared to environmental factors. For this purpose, resting-state functional magnetic resonance imaging data and the behavioral testing of 453 twins from the Colorado Longitudinal Twins Study were analyzed. Separate indices of EF performance were obtained according to a bifactor unity/diversity model, distinguishing between three independent components representing: Common EF, Shifting-specific and Updating-specific abilities. Through an approach of step-wise in silico network lesioning of the individual functional connectome, we show that interindividual differences in EF are associated with different dependencies on neural networks at rest. Furthermore, these patterns show evidence of mild heritability. Such findings add knowledge to the understanding of brain states at rest and their connection with human behavior, and how they might be shaped by genetic influences.

Keywords: Brain topology; Executive functions; Graph theory; Heritability; Twins study.

Fig. 1

Data acquisition and analysis workflow. a Structural and functional MRI data were acquired from 453 twins from the Colorado LTS dataset. b. BOLD signal fluctuations were then extracted from each of the 264 cortical nodes as defined by the Power Atlas (Power et al. 2011) and used to extract individual functional connectivity matrices from the Pearson’s r correlation value between each pair of ROIs. c Matrices were then thresholded to retain the 10% of the connection density. d Brain graph metrics were extracted, considering brain parcels as nodes and their functional connections as edges. e A matrix lesioning approach was employed to estimate the extent for which individual brain activity at rest relies on cortical nodes belonging to different cortical networks. To do so, all nodes in the brain were ranked based on their nodal degree, so that the most important hub in the brain was removed first during the lesioning process. After each lesion, brain topology measures were re-computed and the order of lesioning updated. f At the end of the lesioning process, we counted the number of nodes belonging to a given network that were lost at each stage of matrix lesioning. Networks which lost the greatest number of nodes during the first stages of lesioning (reduction of 12.5% of nodes) were considered to represent the networks on which individuals most rely during rest (i.e., network reliance). Finally, the influence of genetics and of environmental factors in determining network reliance was computed as the difference between MZ and DZ twins. *AUD auditory network, CING cingulo-opercular network, DAN dorsal attention network, DMN default mode network, FPN fronto-parietal network, LTS Colorado Longitudinal Twin Study, rs-fMRI resting-state functional magnetic resonance imaging, s-MRI structural magnetic resonance imaging, SMN sensorimotor network, SN salience network, SUB subcortical network, VAN ventral attention network, VIS visual network

Fig. 2

Network reliance at rest as a function of EF performance. Patterns of network reliance were computed based on the average number of nodes lost per network for each stage of matrix lesioning. Overall, degradation patterns appeared to differ as a function of EF performance. a Individuals scoring higher across EF tasks (cEF component) showed an initial greater loss of DMN nodes, opposite to the lower performers, who appeared to rely more on VIS nodes instead. Interestingly, the pattern switched along the lesioning process, with high and low performers losing visual and DMN/FPN nodes at the last stages of lesioning, respectively. b Performance scores for the shifting-specific (SHI) component proved that, even at rest, higher performers tend to rely more on CING and SUB network nodes, whereas those same networks appear of less relevance in low performers, who lose them only at the last stages of lesioning. c. Higher and lower performers at the updating-specific (UPD) component tended to equally rely more on bottom-up attentional networks (VAN), with a slight tendency for higher performers to also lose more DMN nodes. In the legend, brain size is indicative of the associated statistical significance of each finding. *AUD auditory network, cEF common executive function, CING cingulo-opercular network, DAN dorsal attention network, DMN default mode network, FPN fronto-parietal network, ns non-significant, SHI shifting-specific factor, SMN sensorimotor network, SN salience network, SUB subcortical network, UPD updating-specific factor, VAN ventral attention network, VIS visual network

Fig. 3

Phenotypic similarity across MZ and DZ twins. The average number of nodes lost for the FPN, DAN and DMN network along the 8 stages of matrix lesioning is shown. The greater the similarity in the profiles between MZ and DZ siblings, the narrower the shadowed area in-between profiles. *FPN fronto-parietal network, DAN dorsal attention network, DMN default mode network, MZ monozygotic, DZ dizygotic