Storm in a coffee cup: caffeine modifies brain activation to social signals of threat

Abstract

Caffeine, an adenosine A₁ and A(2A) receptor antagonist, is the most popular psychostimulant drug in the world, but it is also anxiogenic. The neural correlates of caffeine-induced anxiety are currently unknown. This study investigated the effects of caffeine on brain regions implicated in social threat processing and anxiety. Participants were 14 healthy male non/infrequent caffeine consumers. In a double-blind placebo-controlled crossover design, they underwent blood oxygenation level-dependent functional magnetic resonance imaging (fMRI) while performing an emotional face processing task 1 h after receiving caffeine (250 mg) or placebo in two fMRI sessions (counterbalanced, 1-week washout). They rated anxiety and mental alertness, and their blood pressure was measured, before and 2 h after treatment. Results showed that caffeine induced threat-related (angry/fearful faces > happy faces) midbrain-periaqueductal gray activation and abolished threat-related medial prefrontal cortex wall activation. Effects of caffeine on extent of threat-related amygdala activation correlated negatively with level of dietary caffeine intake. In concurrence with these changes in threat-related brain activation, caffeine increased self-rated anxiety and diastolic blood pressure. Caffeine did not affect primary visual cortex activation. These results are the first to demonstrate potential neural correlates of the anxiogenic effect of caffeine, and they implicate the amygdala as a key site for caffeine tolerance.

Fig. 1

Schematic diagram of the Emotional Face Processing Task (Hariri et al., 2002). The diagram shows the first 9 blocks of one of the task runs. Each 24-s block consisted of 6 (4-s) matching trials, and for each trial participants were required to indicate which of two probes presented at the bottom of the image (left or right) was identical to the target (centre top) image. Duration of the ‘match instructions’ was jittered between 2 and 4 s (2, 2.5, 3, 3.5 or 4 s). The contrast of interest was angry/fearful face processing minus happy face processing. Examples of the stimuli presented in the blocks of matching neutral faces, simple shapes (C/E, circles/ellipses; S/R, squares/rectangles) and doors are shown. The fixation cross is also shown.

Fig. 2

Colored regions represent the probabilistic maps of (A) the midbrain-PAG ROI displayed on an axial and sagittal slice and (B) the mPFC wall ROI displayed on a sagittal slice. All slices are T1-weighted MNI templates and all coordinates are in MNI space. Colorbar represents probability levels coded by different colors; yellow color marks those voxels that are included in the ROI with 100% probability (C) Effects of caffeine on mean percent (%) BOLD signal change for the threat (angry/fearful faces) minus non-threat (happy faces) contrast in a priori ROIs. Error bars show 95% within-subject confidence intervals (Loftus and Masson, 1994). Effect size is indicated by r (Rosnow and Rosenthal, 2003).

Fig. 3

(A) Colored region represents the probabilistic map of the BLA ROI displayed on a coronal slice of a single T1-weighted template in MNI space (MNI y = −6). Colorbar represents probability levels coded by different colors; yellow color marks those voxels that are included in the ROI with 100% probability (B) Scatterplot of effects of caffeine on mean percent (%) BOLD signal change for the threat (angry/fearful faces) minus non-threat (happy faces) contrast in the a priori BLA ROI as a function of level of dietary caffeine intake. Data are signal change meaned across the left and right BLA ROIs. The 95% confidence interval for r based on Fisher’s r–z transformation is displayed. The bootstrap 95% confidence interval (5000 iterations) for r was −0.90 ≤ r ≤ −0.17 (Efron, 1988).