Intracranial Electrophysiology of the Human Default Network

Abstract

The human default network (DN) plays a critical role in internally directed cognition, behavior, and neuropsychiatric disease. Despite much progress with functional neuroimaging, persistent questions still linger concerning the electrophysiological underpinnings, fast temporal dynamics, and causal importance of the DN. Here, we review how direct intracranial recording and stimulation of the DN provides a unique combination of high spatiotemporal resolution and causal information that speaks directly to many of these outstanding questions. Our synthesis highlights the electrophysiological basis of activation, suppression, and connectivity of the DN, each key areas of debate in the literature. Integrating these unique electrophysiological data with extant neuroimaging findings will help lay the foundation for a mechanistic account of DN function in human behavior and cognition.

Keywords: ECoG; default network; electrocorticography; iEEG; intracranial electroencephalography.

Figure I. Intracranial EEG refers to the invasive measure of electrical brain potentials via two common methods

A) Electrocorticography (ECoG) utilizes strips or grids of electrode arrays placed subdurally on the cortical surface. B) Stereo-EEG (sEEG) utilizes penetrating depth electrode arrays based on stereotactic coordinates for targeting deeper brain structures.

Figure 1. Intracranial electrophysiology of the human default network

(A) The human DN as assessed using intrinsic connectivity fMRI during the resting state in 1000 participants [9]. The network is distributed throughout the brain: in the frontal lobe, medial prefrontal cortex (PFC), dorsomedial PFC, and ventrolateral PFC; in the temporal lobe, temporopolar cortex, lateral temporal cortex, and aspects of the medial temporal lobe (MTL); and in the parietal lobe, the posterior cingulate cortex and retrosplenial cortex medially, and angular gyrus laterally. Although there is general consensus on this collection of regions, specific network borders vary depending on methods and modality used, and the notion of what constitutes the DN continues to evolve [55]. Because much of the MTL falls outside DN boundaries using standard methods, as a rule we do not discuss iEEG investigations of the MTL throughout this review unless they bear directly on questions relating to the function of other DN regions (for a reviewing covering many iEEG studies involving the MTL, see ref. [103]). (B) Combining iEEG data from cohorts of patients can provide comprehensive coverage of the brain, including all cortical DN regions (here we show 1,955 electrode sites from 16 patients). There are two basic types of iEEG recording [19]. (i) Electrocorticography (ECoG), using subdural grids or strips of circular (plate-shaped) electrodes placed on the brain’s surface, can record from lateral or medial cortical DN areas. Grids often contain dozens of contacts, allowing coverage of, and simultaneous recording from, a substantial proportion of cortex. (ii) Stereotactic EEG (sEEG) uses cylindrical depth electrodes which penetrate through the skull and brain tissue and can reach deep medial cortical and subcortical structures, including midline DN regions. Each depth electrode typically contains 6–10 contacts. For more details, see [19] and Box 1.

Figure 2. Unique contributions to understanding default network deactivations and activations from intracranial EEG

(A) Default network deactivations in response to a visuomotor task consists largely of rapid suppression of power in higher frequency ranges (>30 Hz). Hot colors indicate increased power, and cool colors decreased power, at given frequencies. These and similar findings from other studies constitute a major advance in understanding the electrophysiological basis of DN deactivations observed with neuroimaging. (B) Timing of deactivation matters: whereas neuroimaging findings might suggest that the entire DN deactivates simultaneously, iEEG shows that there are subtle (yet probably important) timing differences in the onset of higher frequency ‘gamma band deactivations’ (GBD) in response to the presentation of complex visual task stimuli. (C) The duration of deactivations is also of functional significance: a more difficult visual search task elicited significantly longer suppressions of high frequency power in default regions. (D) Certain neuronal populations in posteromedial cortex (PMC) are significantly activated (electrodes with red fill color) by a variety of internally-directed cognitive tasks, (E) eliciting power increases specifically at higher frequencies (hot colors indicate increased power, cool colors decreased power). (F) The mean magnitude of PMC activations in the high-gamma band (70–180 Hz) can differentiate between conditions, with the largest activations for the most episodic/memory-like judgments. In contrast, significant suppression of high-gamma power is evident during an externally-directed math task (i.e., judging the accuracy of simple arithmetic equations). (G) Intraregional functional heterogeneity within the PMC. Certain populations show preferential activation in response to memory retrieval, others are selectively active during rest periods, and yet others show mixed patterns. Default regions often treated as functionally homogeneous in neuroimaging investigations in fact contain populations with very diverse response profiles. Panel A reproduced, with permission, from [34]. Panels B and C reproduced, with permission, from [27]. Panels D–F reproduced, with permission, from [28]. Panel I modified, with permission, from [35].

Figure 3. Intrinsic connectivity, internetwork interactions, and electrical stimulation of the default network

(A) Considerable correspondence between maps of intrinsic functional connectivity as measured with fMRI BOLD (two leftmost panels) and iEEG power restricted to high gamma power (50–150 Hz) (rightmost panel) across lateral DN regions as well as other brain networks. Color indicates strength of correlated spontaneous activity at each point with a seed region in the medial PFC DN hub (white-rimmed circle in leftmost panel). Black circles highlight positive correlations with other default regions for both methods, and conversely, white boxes underscore anticorrelations with various non-default regions. (B) Using electrodes in the posterior cingulate cortex as seed regions, significant intrinsic connectivity is observed with the angular gyrus while performing an episodic memory task, at rest, and during sleep. (C) Spatial correlation of intrinsic connectivity maps across iEEG and fMRI BOLD in the default mode network (DMN) is driven most strongly by power in the theta (4–8 Hz) and higher frequency (50–100 Hz) ranges. Other networks (somatomotor network, SMN; dorsal attention network, DAN; and frontoparietal control network, FPC) show similarities but also differences in their correspondence spectra. (D) Precise timing of HFB activations in a DN hub (here, PMC) during an episodic memory task compared to its deactivation during math and activation of superior parietal lobule (SPL) regions during the same conditions. The delayed activation and deactivation of the PMC compared to SPL suggests that the DN lies higher in a hierarchy of information flow through the brain. (E) HF-EBS along the medial occipital and dorsomedial parietal cortical surface often elicits visual and motor effects, respectively, but has no effect within posterior parietal DN regions (PCC/RSC). Panel A reproduced, with permission, from [70]. Panel B reproduced, with permission, from [30]. Panel C reproduced, with permission, from [71]. Panel D reproduced, with permission, from [104]. Panel E reproduced, with permission, from [95].