Development of a large-scale functional brain network during human non-rapid eye movement sleep

Abstract

Graph theoretical analysis of functional magnetic resonance imaging (fMRI) time series has revealed a small-world organization of slow-frequency blood oxygen level-dependent (BOLD) signal fluctuations during wakeful resting. In this study, we used graph theoretical measures to explore how physiological changes during sleep are reflected in functional connectivity and small-world network properties of a large-scale, low-frequency functional brain network. Twenty-five young and healthy participants fell asleep during a 26.7 min fMRI scan with simultaneous polysomnography. A maximum overlap discrete wavelet transformation was applied to fMRI time series extracted from 90 cortical and subcortical regions in normalized space after residualization of the raw signal against unspecific sources of signal fluctuations; functional connectivity analysis focused on the slow-frequency BOLD signal fluctuations between 0.03 and 0.06 Hz. We observed that in the transition from wakefulness to light sleep, thalamocortical connectivity was sharply reduced, whereas corticocortical connectivity increased; corticocortical connectivity subsequently broke down in slow-wave sleep. Local clustering values were closest to random values in light sleep, whereas slow-wave sleep was characterized by the highest clustering ratio (gamma). Our findings support the hypothesis that changes in consciousness in the descent to sleep are subserved by reduced thalamocortical connectivity at sleep onset and a breakdown of general connectivity in slow-wave sleep, with both processes limiting the capacity of the brain to integrate information across functional modules.

Figure 1.

Connectivity and small-world metrics throughout sleep at various thresholds, depicting the wavelet correlation threshold, clustering coefficient (C), gamma (C/C_random), path length (L), lambda (L/L_random), and small-worldness σ (gamma/lambda) at all possible thresholds for S0, S1, S2, and SW, for the frequency band 0.03–0.06 Hz. A, There was a significant effect of sleep on R_thresh, with post hoc tests indicating that values were higher in S1/S2 than in S0/SW (Table 1). B, A significant effect of sleep on C is shown, with post hoc tests revealing higher values in S1/S2 than in S0/SW. C, Shows the reverse pattern for gamma as C_random values were also highest in light sleep. D, E, The effect of sleep on L and lambda was not robustly significant. F, The significant effect of sleep on sigma (driven by gamma) is depicted, with post hoc tests revealing higher values in S0/SW than in S1/S2, or intermediate values for S0. **Significant at 0.001; *significant at 0.05.

Figure 2.

Differential connectivity graphs. Figure 2 illustrates differences in connections of 90 cortical and subcortical regions in the transition from S0 to S1 and S2, and finally to SW, with lines depicting a |ΔR|| > 0.20 between two stages. Top left, Illustrates the breakdown of thalamocortical connectivity from S0 to S1, whereas the top right displays the increased corticocortical connectivity in S1 compared with S0. There are few differences between S1 and S2, whereas a reduction in corticocortical connectivity in SW is depicted in the lower left panel (for the other contrasts, see supplemental Fig. S7, available at www.jneurosci.org as supplemental material; regions with at least 15 differential connections are described in supplemental Table S3 available at www.jneurosci.org as supplemental material). Note that the current threshold of |ΔR|| > 0.20 is an arbitrary threshold as z-values for correlation differences fell between [-1.5; 1.5] (see supplemental Figs. S8 and S9, available at www.jneurosci.org as supplemental material) and were not significant in univariate tests due to modest group sizes. Figure 2 is provided as an illustration of between-stage connectivity differences that are described in Table 3 or in the Results section.

Figure 3.

Algorithmic graphs of wakefulness and each NREM sleep stage. The networks represent binary connections, locally organized by a layout algorithm (Kamada and Kawai, 1989) implemented in the Pajek software package. The algorithm draws distances between nodes as a function of path length. Well-connected nodes (hubs) appear more central, whereas more isolated regions are located in the periphery. This algorithm further iteratively adjusts the positions of nodes and forces them to reduce the total energy of the system to a minimum (for labels, see supplemental Table S2, available at www.jneurosci.org as supplemental material). For this visualization, binary connection matrices were generated at a connection probability of 25% to create a network that was sparsely connected but not fragmented in S0.