Psilocybin modulates functional connectivity of the amygdala during emotional face discrimination

Psilocybin is a serotonergic modulator that acts strongly in limbic structures and prefrontal hubs. The emotional content of faces engages an evolutionarily conserved network—including fusiform gyrus, superior temporal gyrus, amygdala and orbitofrontal cortex—that functions in salience detection and is implicated in affective and anxiety disorders. Previous work has shown that psilocybin alters electrophysiological measures and can attenuate amygdala reactivity in block-design fMRI, but effects on event-related amygdala connectivity during rapid emotional face discrimination had not been investigated. Grimm and colleagues designed the present study to test whether acute psilocybin alters the amygdala's seed-to-voxel connectivity during an event-related emotional face discrimination task. They hypothesised that psilocybin would selectively change the amygdala's connectivity with other limbic and prefrontal regions while participants discriminated angry, happy or fearful faces from neutral faces. The authors emphasised the use of an event-related design because rapid amygdala habituation may obscure subtle, time-locked connectivity effects in block paradigms.

This was a randomized, double-blind, placebo-controlled, within-subject cross-over study in which 25 healthy right-handed volunteers were recruited; after quality exclusions the final analysis included 18 participants. Each participant attended two scanning sessions at least 14 days apart and received either oral psilocybin (0.16 mg/kg) or placebo in counterbalanced order. Verbal intelligence, mood (PANAS), state anxiety (STAI-state) and altered states of consciousness (5D-ASC) were assessed before and 210 minutes after drug administration. The study received ethical approval and participants gave informed consent. Participants completed three separate event-related emotional face discrimination fMRI sessions in randomized order: angry versus neutral, happy versus neutral and fearful (sad in some labels) versus neutral. Each session used 10 emotional and 10 neutral faces, each repeated pseudorandomly to yield 40 emotional and 40 neutral trials per session. Faces were displayed for 2000 ms and participants indicated via button press whether each face was emotional or neutral; interstimulus intervals followed a Poisson distribution with mean ~5 s. Reaction times and hit rates were recorded. Imaging was performed on a 3.0-T scanner. Functional data were preprocessed in SPM12 (motion correction, slice timing, spatial normalisation to MNI space, resampling to 2 mm isotropic and 8-mm smoothing). First-level general linear models (GLMs) modelled each trial onset convolved with the canonical haemodynamic response function; six motion parameters were included and a high-pass filter (1/128 Hz) and serial autocorrelation correction were applied. To estimate event-related functional connectivity the authors used a beta-series approach: each trial produced a beta estimate, beta-series for a condition were concatenated, and condition-specific correlations between the amygdala seed time series and whole-brain voxels were computed using the CONN toolbox. Mean timeseries were extracted from left and right amygdala (AAL atlas), nuisance regressors (motion, derivatives, CSF and white matter principal components via aCompCor) were removed, and beta-series correlation maps were generated separately for angry, happy and fearful conditions. Statistical analysis used repeated-measures models. Behavioural data (hit rate, reaction times) were analysed with a linear mixed model accounting for within-subject effects and drug (psilocybin versus placebo) as a factor. For fMRI seed-to-voxel analyses the authors computed within-subject contrasts (emotional versus session-specific neutral) and tested psilocybin versus placebo in second-level repeated-measures models using CONN. Multiple comparisons were addressed with cluster-wise whole-brain topological FDR correction and a cluster-defining voxel threshold of p=0.001. Correlations between extracted connectivity values and self-report measures (PANAS, STAI, 5D-ASC) and behaviour were also evaluated, with correction for multiple tests where specified.

Behaviourally, psilocybin significantly slowed reaction times across all face types: the drug main effect on reaction time was F(1,17)=24.03, p<0.001. Correct classification (hit rate) was not significantly altered by psilocybin (F(1,17)<2.25, p>0.15), and there was no significant drug × face-type interaction (F(1,17)=0.70, p=0.50). Self-report effects on anxiety, affect and altered states were reported elsewhere and were not the primary focus of this extracted text. The task-related amygdala seed-to-voxel connectivity pattern (main effect across emotional faces) revealed strong coupling with contralateral amygdala, striatum (caudate and putamen), lateral occipital cortex and fusiform gyrus, insula, thalamus, frontal pole and supplementary motor cortex (cluster-wise pFDR<0.05, voxel threshold p=0.001). These regions matched the expected face-processing and salience-detection network. Comparing psilocybin to placebo for emotion-versus-neutral contrasts, two focal connectivity decreases reached whole-brain significance. In the angry versus neutral contrast, psilocybin reduced coupling between the right amygdala and left striatum (caudate/putamen). Extraction of the significant z-transformed connectivity values from this cluster did not yield significant correlations with self-report questionnaires. In the happy versus neutral contrast, psilocybin decreased connectivity between the right amygdala and the medial frontal pole (BA 10); lower connectivity in this cluster during the psilocybin session correlated with lower state anxiety after drug intake (r(16) = -0.63, p=0.005) and with lower self-rated depressed mood (r(16) = -0.6, p=0.008). No whole-brain significant effects were observed for fearful versus neutral faces, and the left amygdala seed did not show significant happy or angry contrasts. There was no significant overall emotion type by drug interaction. Correlations between reaction times and the extracted connectivity values were non-significant for both happy and angry conditions (p>0.34 and p>0.44, respectively).

Grimm and colleagues interpret their findings as evidence that acute psilocybin modulates amygdala-centred networks engaged in emotional face discrimination. Behaviourally, the drug produced a non-specific slowing of responses without impairing accuracy. Neurally, the authors observed decreased amygdala–striatum connectivity during angry face discrimination and decreased amygdala–medial frontal pole connectivity during happy face discrimination, with no significant effects for fearful faces. The authors relate the amygdala–striatum decrease to modulation of salience evaluation: both regions are established hubs for assessing cue saliency and may form a relevance-detection system. They note that the striatum is often discussed in dopaminergic terms and cite preclinical evidence that psilocybin's active metabolite elevates striatal dopamine, but they stress that neurotransmitter-specific mechanisms were not tested here. Reduced amygdala–frontal pole connectivity is discussed as a potential disruption of prefrontal cognitive control over the amygdala; the observed negative correlations between this connectivity and state anxiety/depressed mood during psilocybin are reported, but the authors caution that their analysis cannot determine directionality of influence (top-down versus bottom-up). Several limitations are acknowledged. The sample comprised healthy volunteers and cannot speak to therapeutic effects in clinical populations. Trial-level conscious recognition was imperfect (relatively low hit rate), which complicates interpretation though the authors note amygdala responses can occur without conscious awareness. Attentional changes may explain reduced connectivity, since psilocybin has been linked to attenuated attentional performance; the current design could not disambiguate general versus selective attention effects. The temporal resolution of BOLD fMRI (TR=2.3 s) precludes resolving rapid temporal differences suggested by EEG studies, and cluster-wise correction methods have known limitations. Finally, the authors highlight that while 5-HT2A receptor agonism is a plausible mechanism given previous PET/fMRI work, they did not test receptor specificity (for example, by co-administration of a 5-HT2A antagonist). In conclusion, the study reports spatially specific reductions in amygdala connectivity with key salience-related hubs during psilocybin intake, which the authors suggest may reflect altered attention and salience attribution. They propose that future work should combine task-based connectivity, electrophysiology or graph-analytic approaches, and should assess longitudinal and clinical-sample effects to determine whether such connectivity changes relate to therapeutic outcomes.