Topologically Reorganized Connectivity Architecture of Default-Mode, Executive-Control, and Salience Networks across Working Memory Task Loads

Abstract

The human brain is topologically organized into a set of spatially distributed, functionally specific networks. Of these networks, the default-mode network (DMN), executive-control network (ECN), and salience network (SN) have received the most attention recently for their vital roles in cognitive functions. However, very little is known about whether and how the interactions within and between these 3 networks would be modulated by cognitive demands. Here, we employed graph-based modularity analysis to identify the DMN, ECN, and SN during an N-back working memory (WM) task and further investigated the modulation of intra- and inter-network interactions at different cognitive loads. As the task load elevated, functional connectivity decreased within the DMN while increased within the ECN, and the SN connected more with both the DMN and ECN. Within-network connectivity of the ventral and dorsal posterior cingulate cortex was differentially modulated by cognitive load. Further, the superior parietal regions in the ECN showed increased internetwork connections at higher WM loads, and these increases correlated positively with WM task performance. Together, these findings advance our understanding of dynamic integrations of specialized brain systems in response to cognitive demands and may serve as a baseline for assessing potential disruptions of these interactions in pathological conditions.

Keywords: connectomics; default-mode network; executive-control network; module; salience network.

Figure 1.

Illustration of module identification results. (A) Task mask generated from a combination of activation/deactivation maps across 1-back, 2-back, and 3-back against 0-back. Modularity analyses were performed within this mask. Brain areas showing task-related activation are depicted in red; areas showing task-related deactivation are depicted in blue. (B) The mean modularity across subjects obtained for 0-back, 1-back, 2-back, and 3-back at network sparsities ranging from 1% to 5%. (C) Module partitions of the 0-back brain graph thresholded at the highest density of 1%. (D) Maps of the DMN (blue), ECN (red), and SN (yellow) at 0-back, 1-back, 2-back, and 3-back. (E) A spring-embedded layout of nodes and edges within the DMN, ECN, and SN networks at 0-back. The network was visualized with the Pajek program (Batagelj and Mrvar 1998).

Figure 2.

Load-dependent differences in intranetwork connections within the DMN (A) and the ECN (B), and internetwork connections between the SN and the DMN (C) and between the SN and the ECN (D). Error bars refer to SE. **P_corrected < 0.05, *P_uncorrected < 0.05.

Figure 3.

Load-dependent effects in WD. (A) The WD map at 0-back, 1-back, 2-back, and 3-back. (B) Repeated-measures ANOVA revealed significantly modulated WD values by task load in the ACC (1) and ventral PCC (2), dorsal PCC (3), and left IPL (4). (C) Significant differences in WD across the 4 task loads in the 4 regions. Error bars refer to SE. *P_corrected < 0.05.

Figure 4.

Load-dependent effects in PC. (A) The PC map at 0-back, 1-back, 2-back, and 3-back. (B) Repeated-measures ANOVA revealed significantly modulated PC values by task load in the rACC (1) and PCC (2), left SPL (3), and right MFG (4). (C) Significant differences in PC across the 4 task loads in the 4 regions. Error bars refer to SE. *P_corrected < 0.05.

Figure 5.

Relationship between WM task performance and PC in the left SPL at 3-back. (A) PC in the region of the left SPL showed significant modulation effect of task loads (3-back–0-back). (B) Plot of the average changes in PC (3-back/0-back) in the left SPL against dprime (3-back/0-back).

Figure 6.

Schematic illustration of the load-dependent reorganization of intra- and inter-module connectivity among the DMN, SN, and ECN. As task load increases, connectivity within DMN decreased whereas connectivity within ECN increases. SN showed more connections to both DMN and ECN. The inset figure circled by dashed lines shows the heterogeneous changes in intermodule connectivity for ventral (purple) and dorsal (dark blue) PCC subregions. The lines match the color of the brain modules or regions. Module-level intermodule connections are depicted in green color. PCC, posterior cingulate cortex. The 3D brain regions were drawn using the BrainNet viewer (www.nitrc.org/projects/bnv (Xia et al. 2013)).