Basis of executive functions in fine-grained architecture of cortical and subcortical human brain networks

Abstract

Theoretical models suggest that executive functions rely on both domain-general and domain-specific processes. Supporting this view, prior brain imaging studies have revealed that executive activations converge and diverge within broadly characterized brain networks. However, the lack of precise anatomical mappings has impeded our understanding of the interplay between domain-general and domain-specific processes. To address this challenge, we used the high-resolution multimodal magnetic resonance imaging approach of the Human Connectome Project to scan participants performing 3 canonical executive tasks: n-back, rule switching, and stop signal. The results reveal that, at the individual level, different executive activations converge within 9 domain-general territories distributed in frontal, parietal, and temporal cortices. Each task exhibits a unique topography characterized by finely detailed activation gradients within domain-general territory shifted toward adjacent resting-state networks; n-back activations shift toward the default mode, rule switching toward dorsal attention, and stop signal toward cingulo-opercular networks. Importantly, the strongest activations arise at multimodal neurobiological definitions of network borders. Matching results are seen in circumscribed regions of the caudate nucleus, thalamus, and cerebellum. The shifting peaks of local gradients at the intersection of task-specific networks provide a novel mechanistic insight into how partially-specialized networks interact with neighboring domain-general territories to generate distinct executive functions.

Keywords: cognitive control; executive functions; fMRI; multiple demand.

Fig. 1

(a) The 9 MD patches displayed on cortical surface (left) and a flattened left surface (right) as revealed by average activations of 449 subjects based on 3 cognitively demanding contrasts from Assem et al. (2020): 2 > 0 n-back, hard>easy reasoning, math>story. (b) Extended MD system from Assem et al. (2020). Core MD regions are colored in bright green surrounded by black borders and individually labeled. Penumbra MD regions are colored in dark green. Data available at: http://balsa.wustl.edu/r27NL. (c) Flat cortical maps overlaid with group average activations for each executive contrast in the current study. Green borders surround core MD areas, with the 9 coarser-scale patches labeled on the left hemisphere. Right column shows example activation from a single subject on the left hemisphere. See Supplementary Fig. 1 for more single subject activations. All single subject data are available at: http://balsa.wustl.edu/x8M0q.

Fig. 2

Illustration of the 3 tasks performed in the current study. Note stimuli were either faces or houses.

Fig. 3

The unity of executive functions. (a) Cortical parcels showing conjunction of significant activation in each of the 3 executive contrasts (P < 0.05 Bonferroni corrected). All unique areas identified in either hemisphere are projected on the left hemisphere, and colored according to RSN membership from Ji et al. (2019). Black borders surround core MD, white borders surround 2020-penumbra areas. Note PEF’s RSN membership is CON on the right hemisphere. Data available at: http://balsa.wustl.edu/P2MXl. (b) Areal responses to each of the 3 contrasts. A colored X means the area did not survive Bonferroni correction for 360 areas (P < 0.05); red n-back, blue stop, green switch). Colors of areal names show RSN membership (color scheme as in (a), with the addition of red = DMN). (c) Subject overlap map of cortical vertices that were significantly activated in all 3 executive contrasts for individual subjects (P < 0.05 FDR corrected). Black borders surround core MD. Colored borders show RSN membership (CON = blue, DAN = green). Data available at: http://balsa.wustl.edu/7x6l7.

Fig. 4

The diversity of executive functions. (a) Task functional preferences. Each vertex is colored with the task that significantly activated it more than each of the other 2 tasks (P < 0.05 FDR corrected across vertices and Bonferroni corrected for 3 tasks; red: 3 > 1 n-back, green: switch>no switch, blue: stop>no stop). Core MD areas are surrounded by a black border. (b) Canonical RSNs from the HCP based 12 network parcellation by Ji et al. (2019) (red: DMN, green: DAN, blue: CON, yellow with black borders: core MD in FPN, orange with gray borders: noncore MD FPN). Note similarity in the topographical organization with each task preference in (a). Data available at: http://balsa.wustl.edu/647Nr. (c) Task activations for each of the 5 networks in (b). Error bars are SEMs. Darker colored bars for left hemisphere, lighter colored bars for right hemisphere. Horizontal black lines compare significance between tasks collapsed across hemispheres (P < 0.05 Bonferroni corrected for 3 tasks and 5 networks).

Fig. 5

Sub-areal task preferences. (a) Cortical projection of the RGB color weighted normalized task profiles. Reddish colors mean stronger n-back activity, bluish colors mean stronger stop contrast activity and greenish colors mean stronger switch contrast activity. Core MD areas are surrounded by black borders. (b) Vertex-level statistical comparison of activations within core MD regions. N-back preferring vertices are in red, switch vertices in green, and stop vertices are in blue. White vertices denote nonsignificant statistical differences between tasks (P < 0.05 FDR corrected). Surrounding core MD regions (black borders) are canonical RSNs from Ji et al. (2019) (red: DMN, green: DAN, blue: CON). Data available at: http://balsa.wustl.edu/1p0wB.

Fig. 6

Core MD connectivity gradients. (a) Top row: connectivity map of 3 seeds within right p9-46v. Bottom row: group average activations for each executive contrast. Core MD regions shown with green outlines. Data available at: http://balsa.wustl.edu/5BPVB. (b) Core MD vertices colored using a winner-take all approach: blue, red, green for vertices where more subjects overlapped for stop, n-back, switch, respectively. Surrounding core MD regions (black borders) are canonical RSNs from Ji et al. (2019) (red: DMN, green: DAN, blue: CON). Data available at: http://balsa.wustl.edu/n82L9.

Fig. 7

Peak activations at region borders. (a) Subject overlap map of top 5% activated voxels for each contrast. Core MD borders are colored in black and the remaining MMP1.0 areas borders are in light gray. RSNs are colored as follows: DMN is red, DAN is green, and CON is blue. Data available at: http://balsa.wustl.edu/gmNwX. (b) First column represents a close up of cortical areas of interest (top: lateral prefrontal, middle: lateral parietal, bottom: medial frontal). Remaining columns display example activation profiles near areal borders, highlighting the hemisphere –right or left- with the strongest pattern. N-back activations are in red, stop in blue, and switch in green. Shaded areas represent SEMs. The location of the border is marked by a vertical gray line at the zero point of the x axis.

Fig. 8

Conjunctions and task preferences in subcortex and cerebellum. (a) Subcortical axial slices and a cerebellar flat map showing surviving voxels in caudate, thalamus, and cerebellum for the conjunction of significantly activated voxels for each of the 3 executive contrasts (P < 0.05 FDR corrected for each structure separately). (b) Voxels belonging to each network (yellow = MD, green = DAN, blue = CON, red = DMN). MD caudate, thalamic, and cerebellar voxels are from Assem et al. (2020). The other 3 RSN definitions are from Ji et al. (2019) [see Materials and Methods for more details]. Data available at: http://balsa.wustl.edu/Bgp6w. (c)–(e) Bar plots of activations across subjects for each hemisphere, network, and structure. Darker colored bars are for left hemisphere and lighter colored ones for the right hemisphere. Error bars are SEMs.