Task-evoked Negative BOLD Response and Functional Connectivity in the Default Mode Network are Representative of Two Overlapping but Separate Neurophysiological Processes

Abstract

The topography of the default mode network (DMN) can be obtained with one of two different functional magnetic resonance imaging (fMRI) methods: either from the spontaneous but organized synchrony of the low-frequency fluctuations in resting-state fMRI (rs-fMRI), known as "functional connectivity", or from the consistent and robust deactivations in task-based fMRI (tb-fMRI), here referred to as the "negative BOLD response" (NBR). These two methods are fundamentally different, but their results are often used interchangeably to describe the brain's resting-state, baseline, or intrinsic activity. While the DMN was initially defined by consistent task-based decreases in blood flow in a set of specific brain regions using PET imaging, recently nearly all studies on the DMN employ functional connectivity in rs-fMRI. In this study, we first show the high level of spatial overlap between NBR and functional connectivity of the DMN extracted from the same tb-fMRI scan; then, we demonstrate that the NBR in putative DMN regions can be significantly altered without causing any change in their overlapping functional connectivity. Furthermore, we present evidence that in the DMN, the NBR is more closely related to task performance than the functional connectivity. We conclude that the NBR and functional connectivity of the DMN reflect two separate but overlapping neurophysiological processes, and thus should be differentiated in studies investigating brain-behavior relationships in both healthy and diseased populations. Our findings further raise the possibility that the macro-scale networks of the human brain might internally exhibit a hierarchical functional architecture.

Figure 1

The design of the attention switching sensory-motor tasks using an event-related fMRI paradigm. The blue line denotes the random timing for duration and onset of the visual stimuli, and the red line shows the same for audio stimuli in a typical participant. The green bars represent the time of the response to either the visual or audio stimulus. There were at least 55 visual and 55 audio stimuli in each run, with mean duration of 1.2 sec and a range of 0.5 to 3.5 seconds. In one run, participants were instructed to attend to visual stimuli (top row) and press a button as soon as each visual stimulus was terminated, while ignoring audio stimuli. In another run, they were instructed to attend to audio stimuli (bottom row) and press a button as soon as each audio stimulus was terminated, while ignoring visual stimuli.

Figure 2

Spatial overlap between regions with significant negative BOLD response and functional connectivity in DMN. Voxel-wise significant z statistics are mapped onto a semi-inflated cortical surface for better visualization (slice-based mapping can be found in Fig. S1). Dark blue denotes the spatial extent of the regions with significant (z < −4 after multiple comparisons correction) NBR to visual- and audio-attended stimuli during two separate tb-fMRI scans. Light blue indicates their overlap. Red indicates the spatial extent of the regions with significant functional connectivity (z > 4 after multiple comparisons correction) extracted from two tb-fMRI scans (visual-attended and audio-attended tasks) as well as one rs-fMRI scan. Light red and orange highlight the overlap of the two and three functional connectivity networks, respectively. Dark-yellow and yellow indicate the overlap of four and all five of the aforementioned networks, respectively. See Fig. S2 for individually depicted and detailed spatial pattern of each functional connectivity network and task-related BOLD response illustrated in this figure.

Figure 3

Attention-specificity of the negative BOLD response in the DMN regions. Spatial extent of the voxels with significant PBR (activated) and NBR (deactivated) during sensory-motor tasks using z-statistics thresholded at |z| > 4 for (a) attended visual, (b) unattended audio, (c) unattended visual, and (d) attended audio stimuli. The z-statistics for PBR are color-coded with warm colors (red-yellow), and those for NBR with cold colors (blue-light blue). The solid, light blue color represents the mask of the regions that have significantly higher magnitude of the NBR for attended stimuli versus unattended stimuli (z > 2.3, after cluster-wise multiple comparisons correction). The voxel-wise significant z statistics are mapped onto a semi-inflated cortical surface for a better visualization (slice-based mapping can be found in Fig. S3). Note that almost no significant NBR (deactivation) is present for unattended stimuli. The solid light blue color masks out most of the color-coded spatial maps of the deactivated area in the visual-attended BOLD response and partially in the audio-attended BOLD response. See the bottom row of the Fig. S2 for color-coded spatial maps of deactivated regions for both visual- and audio-attended tasks.

Figure 4

Consistency in the strength of the DMN functional connectivity during tasks and rest. The distribution of the subject-wise strength of the functional connectivity in the DMN regions extracted from attended visual (in blue), and attended audio (in red) tb-fMRI scans as well as from the rs-fMRI scan (in green) are illustrated with different violin plots. Pair-wise student t-test reveals no significant difference between any pairs of the three distributions (visual versus rest: t = 0.2, p > 0.8; audio versus rest: t = 0.3, p > 0.7; visual versus audio: t = 0.7, p > 0.4), highlighting that functional connectivity strength remains intact during tasks and rest.

Figure 5

The DMN functional connectivity and NBR are differentially related to task performance. Subject-wise median response time correlates with the subject-wise magnitude of the NBR in the DMN during (a) attended visual (β = 0.24, p < 0.02) and (b) attended audio (β = 0.17, p < 0.02) tb-fMRI scans. However, it does not correlate with the subject-wise expression of functional connectivity in DMN regions during (c) attended visual (β = −0.018, p > 0.2) and (d) attended audio (β = −0.018, p > 0.01) tb-fMRI scans, providing evidence for differential level of involvement of the two fMRI measurements in task execution. Each dot represents a single subject and the line presents the linear fit to the data.

Figure 6

Spatial overlap between the DMN’s functional connectivity and NBR during 2-back task. Voxel-wise significant z statistics are mapped onto a semi-inflated cortical surface for a better visualization (slice-based mapping can be found in Fig. S7). Blue denotes the spatial extent of the regions with significant (z < −4 after multiple comparisons correction) NBR during 2-back working memory task. Red depicts the spatial extent of the regions with significant functional connectivity (z > 4 after multiple comparisons correction) extracted from the same 2-back tb-fMRI scan. Light red indicates the spatial extent of the regions with significant functional connectivity (z > 4 after multiple comparisons correction) extracted from the rs-fMRI scan. Orange highlights the overlap of the two functional connectivity networks. Yellow highlights the overlap of the two functional connectivity networks and the NBR during the 2-back working memory task.

Figure 7

Consistency in the strength of the DMN functional connectivity during 2-back working memory task and rest. The distribution of the subject-wise strength of the functional connectivity in the DMN regions extracted from the 2-back working memory task (in blue), and from the rs-fMRI scan (in red) are illustrated with different violin plots. Pair-wise student t-test reveals no significant difference between the two distributions (t = 0.5, p > 0.6), highlighting that in comparison to rest, functional connectivity expression remains intact during working memory task.

Figure 8

Unlike the DMN’s functional connectivity, its negative BOLD response correlates with performance during a 2-back working memory task. Subject-wise median response time correlates with the subject-wise magnitude of the NBR in the DMN regions during a 2-back working memory task (β = 83.16, p < 0.0002). However, it does not correlate with the subject-wise expression of functional connectivity in DMN regions during the same tb-fMRI scan (β = 5.73, p > 0.14), providing evidence for differential level of involvement of the two fMRI measurements in task execution. Each dot represents a single subject and the line presents the linear fit to the data.