Substructure of the brain's Cingulo-Opercular network

Abstract

The Cingulo-Opercular network (CON) is an executive network of the human brain that regulates actions. CON is composed of many widely distributed cortical regions that are involved in top-down control over both lower-level (i.e., motor) and higher-level (i.e., cognitive) functions, as well as in processing of painful stimuli. Given the topographical and functional heterogeneity of the CON, we investigated whether subnetworks within the CON support separable aspects of action control. Using precision functional mapping (PFM) in 15 participants with > 5 hours of resting state functional connectivity (RSFC) and task data, we identified three anatomically and functionally distinct CON subnetworks within each individual. These three distinct subnetworks were linked to Decisions, Actions, and Feedback (including pain processing), respectively, in convergence with a meta-analytic task database. These Decision, Action and Feedback subnetworks represent pathways by which the brain establishes top-down goals, transforms those goals into actions, implemented as movements, and processes critical action feedback such as pain.

Keywords: Cingulo-Opercular network; action control; cognitive control; functional connectivity; precision functional mapping.

Figure 1:. Three Cingulo-Opercular Subnetworks.

The undifferentiated CON (left) and CON subnetworks (right) in an exemplar participant (P01) with 356 minutes of resting-state fMRI data. Three distinct subnetworks were consistently identified in every individual: an Anterior subnetwork in dorsal anterior cingulate, dorsal anterior insula, and anterior PFC (green); a Central subnetwork in more posterior dorsal anterior cingulate extending up to dmPFC, middle/posterior insula, and regions just anterior and posterior of the central sulcus (yellow); and a Lateral subnetwork in middle insula, supramarginal gyrus, and the pars marginalis of the cingulate (blue). See Figure S1 for all individual participants.

Figure 2:. Cortical and subcortical distribution of CON subnetworks across participants.

A-C) Density maps illustrate the number of participants with overlapping subnetwork representations at each point in cortex (left), thalamus and striatum (top right), and cerebellum (bottom right), for the A) Anterior, B) Central, and C) Lateral subnetworks. Maps were thresholded to retain points at which at least three participants exhibited overlap. Differential scaling of density maps in cortex and subcortex was employed because of increased cross-participant variability (due to lower SNR) in subcortex. D) A winner-take-all map illustrates the relative topographies of all three subnetworks.

Figure 3:. Meta-Analytic Network Annotation of Cingulo-Opercular subnetworks.

Subnetworks from the cross-participants winner-take-all analysis (Fig 2D) were matched to spatial activation distributions that had task descriptor terms in the Neurosynth database . Word clouds illustrate terms more associated with activation patterns of each Cingulo-Opercular subnetwork compared to the other subnetworks (tested for each term via one-way ANOVA). Larger font size indicates higher frequency of the term. Terms shown in black are significant at p < 0.05 (unc.); colored terms are significant at p < 0.05, FDR corrected for the 742 terms tested.

Figure 4:. Functional connectivity of CON subnetworks.

A) A spring embedded plot in an exemplar participant (P01) illustrates the preferential connectivity of Cingulo-Opercular subnetworks to large-scale functional networks. For clarity of visualization, only networks most closely associated with the subnetworks are shown. See Figure S2 for all individual participants. B) Across participants, individual-specific Cingulo-Opercular subnetworks demonstrate preferential connectivity to other individual-specific large-scale networks. The radial axis indicates the strength of functional connectivity Z(r) between each CON subnetwork and each large-scale functional network. Negative connectivity values are not represented. Colors and spatial locations of CON subnetworks (left) and other large-scale networks (right) are shown in the exemplar participant at the bottom.

Figure 5:. Temporal ordering of CON signals in the action output hierarchy.

Temporal ordering of subnetwork fMRI signals for each individual-specific CON subnetwork, as well as for the SCAN and the Somatomotor-hand networks. Values are averaged across vertices within each subnetwork and across participants. Standard error across participants is indicated by error bars. A one-way ANOVA indicated a significant main effect of subnetwork/network identity (p = 0.0007). * indicates p < 0.05 for post-hoc paired t-tests. Inset shows subnetwork and network topography for an example participant (P01). Prior electrophysiology work suggests that later infra-slow activity (here, the Feedback subnetwork) corresponds to earlier delta-band (0.5–4Hz) activity .

Figure 6:. Differentiation of Cingulo-Opercular subnetworks based on task activations.

Ten participants performed motor tasks (including flexure of right and left toes, open/closing of right and left hand, and left-right movement of the tongue), as well as a Spatial and a Verbal Discrimination task. The radial axis indicates the z-score of each condition relative to baseline fixation, averaged across all participants and across all vertices in each individual-specific CON subnetwork. Inset shows subnetwork topography for an example participant (P01).

Figure 7:. Model of CON subnetworks within a hierarchical organization of goal and action control.

Our proposed model represented on the cortex (left) and in schematic form (right). Reward-driven valuations from the Salience network feed into CON subnetworks, which establish goals (Decision subnetwork) and goal-related action plans (Action subnetwork), and project those action plans to the SCAN and to effector-specific M1 to execute an action. Fast somatosensory information resulting from the action is processed by the Action subnetwork, which modifies action plans. Slower goal-related visual and painful outcomes of the action are processed by the Feedback subnetwork, which modifies goals represented by the Decision subnetwork. Red arrows indicate feedforward projections; blue arrows indicate feedback.