Subcortical evidence for a contribution of arousal to fMRI studies of brain activity

Abstract

Cortical activity during periods of rest is punctuated by widespread, synchronous events in both electrophysiological and hemodynamic signals, but their behavioral relevance remains unclear. Here we report that these events correspond to momentary drops in cortical arousal and are associated with activity changes in the basal forebrain and thalamus. Combining fMRI and electrophysiology in macaques, we first establish that fMRI transients co-occur with spectral shifts in local field potentials (LFPs) toward low frequencies. Applying this knowledge to fMRI data from the human connectome project, we find that the fMRI transients are strongest in sensory cortices. Surprisingly, the positive cortical transients occur together with negative transients in focal subcortical areas known to be involved with arousal regulation, most notably the basal forebrain. This subcortical involvement, combined with the prototypical pattern of LFP spectral shifts, suggests that commonly observed widespread variations in fMRI cortical activity are associated with momentary drops in arousal.

Fig. 1

Global fMRI peaks and electrophysiological SST events in monkeys. a The large peaks in the global fMRI signal (the lower panel) are preceded by SST events in the mean spectrogram of concurrently acquired LFP signals (the upper panel). b The SST pattern was obtained on an independent ECoG data set from macaque monkeys by a previous study. c The SST pattern emerges with aligning and averaging the mean LFP spectrogram segments with respect to the top half of large global fMRI peak points (marked as red circles and dash lines in the example in a). In contrast, the control cases, which were obtained by reversing time locations of these peak points, do not show any obvious pattern. The results for monkey A, V, and S are averaged from 125, 401, and 197 data segments, respectively. d The 2D cross-correlation function between the ECoG SST pattern (b) and the global peak triggered LFP pattern shows a clear peak (top left, averaged across N = 16 sessions), and its cross-sectional profile at the zero-frequency shift reaches the maximum at a time lag of 5.2 s, as expected based on the fMRI hemodynamic delay (the blue curve in the right panel, the shadow represents regions within one S.E.M.). In contrast, the 2D cross-correlation function (bottom left) and the corresponding zero-frequency profile (the gray curve in the right panel) for the control case do not show clear peaks

Fig. 2

Global fMRI co-activation pattern in a representative human subject. a A global fMRI time course from a typical subject and four examples of instantaneous co-activation patterns at individual time points with large global signal. The time-averaged global co-activation pattern was generated by averaging all the high-signal (red) time points, which were defined to have a global signal larger than 0.2 and made up 16.6% of total number of time points. b The human fMRI co-activation pattern at time points with high global signal. The Z-score map suggests that the visual, auditory, and sensorimotor regions show much larger and more consistent signal increase than other brain areas at time points with large global signals, even though the 93.9% gray matter voxels (colored regions) show a statistically significant score (Z > 2.58, equivalent to p < 0.01, N = 18,725 time points). c The gamma power increases at the SST (within the dash box in Fig. 1b) seen in monkey electrophysiology show larger amplitude at the visual, auditory, and sensorimotor regions. The results here are adapted from ref.

Fig. 3

De-activation of nucleus basalis (NB) during the global fMRI co-activation. a The global signal co-activation map is overlaid on a high-resolution anatomical MRI scan and shown in three slices at sagittal (X = 11), coronal (Y = 2), and axial (Z = −10) planes, respectively. All coordinates are in the standard Montreal Neurological Institute (MNI) space, which offsets in Y and Z directions by 4 and 5 mm with respect to the “anatomical MNI space” previously used for the basal forebrain cytoarchitectonic mapping. b The de-activated region from coronal view is amplified and compared with the NB region shown in a human brain atlas. c 3D rendering of the de-activated regions shown in a. d The overlap (yellow) between the NB de-activation mask (light blue, corresponding to Z < −6) and a mask of Ch4 cell distribution (red) derived from the cytoarchitectonic map of the human basal forebrain is shown at an axial slice (Z = −10)

Fig. 4

De-activation of two other subcortical regions during global signal co-activation. The de-activation in the thalamus is situated at the dorsal midline region, as shown in three imaging slices at sagittal (X = 2), coronal (Y = −8), and axial (Z = 7) planes, respectively. b The de-activation in the midbrain is located right above the Pons, as shown in three imaging slices at sagittal (X = −8), coronal (Y = −14), and axial (Z = −18) planes, respectively. This midbrain structure is tentatively assigned to be the substantia nigra (SN) based on a comparison with the human brain atlas

Fig. 5

Temporal dynamics at global fMRI signal peaks. The averages of fMRI signals around the global signal peak points (N = 2134) at the nucleus basalis of the basal forebrain (blue, dash-dot line), the dorsal midline thalamus (green, dotted line), the midbrain structures (dark blue, dashed line), and the cortical region (orange, solid line). The control (gray, dash-dot line) is an average of fMRI signals around randomly selected time points at the nucleus basalis. The shadow regions represent the area within one standard error of the mean. The masks for de-activations at the nucleus basalis (blue), the dorsal midline thalamus (green), and the midbrain structure (dark blue) are shown at three axial slices at Z = −10, 7, and −18, respectively, along with the mask of the cortical co-activations (orange) on the right panel

Fig. 6

Reproducibility of the global signal co-activation pattern. Global signal co-activation maps derived from the DAY 1 (a) and DAY2 (b) data of the top 94 subjects with the largest global signal fluctuations (i.e., standard deviation) in DAY 1 data, from the DAY 1 data of 94 randomly selected subjects (c), from the DAY 1 data of top 20 subjects showing the largest global signal fluctuations (d), as well as from the DAY 1 data of top 94 subjects but without using ICA-FIX preprocessing (e). The result in a is the same as that shown in Fig. 3. Three slices are at Z = −10, 7, and −18, respectively. The white arrows and magenta circles in the top, middle, and bottom rows indicate de-activations in the nucleus basalis (NB) of the basal forebrain, the dorsal midline thalamus, and the midbrain structure, respectively