Therapeutic mechanisms of psilocybin: Changes in amygdala and prefrontal functional connectivity during emotional processing after psilocybin for treatment-resistant depression

Serotonergic psychedelics such as psilocybin have shown therapeutic promise for conditions including depression, yet the neural mechanisms that underlie their clinical efficacy remain poorly understood. Previous work in this cohort of patients with treatment-resistant depression (TRD) identified increased right amygdala responsiveness to emotional faces one day after a high-dose psilocybin session, and this amygdala change correlated with clinical improvement. At the same time, other post-acute studies of psychedelics have reported alterations in resting-state connectivity and regional perfusion, suggesting that neural changes beyond the acute drug state may contribute to sustained symptom benefit. Mertens and colleagues set out to probe task-related functional connectivity (FC) changes that might accompany the earlier amygdala activation findings. Specifically, they tested whether psilocybin treatment leads to altered amygdala FC during an emotional face perception task, with a primary hypothesis of decreased amygdala–ventromedial prefrontal cortex (vmPFC) connectivity after treatment. The investigators also examined whether any FC changes related to clinical outcomes at one week, focusing on depressive symptoms and rumination as primary and secondary endpoints respectively.

This was an open-label feasibility study of psilocybin-assisted therapy in patients with TRD. Nineteen patients (six female; mean age 44.7 years) completed pre- and post-treatment fMRI scans and form the analysed sample. Participants received two psilocybin-assisted therapy sessions one week apart: an initial 10 mg oral test dose followed by a 25 mg oral therapeutic dose. Psychological support comprised a 4-hour preparation session, in-session and peri‑acute support during dosing, and a single integration session the morning after the high dose. Patients were requested to discontinue antidepressant medication before study entry; most complied, but two participants did not fully adhere (one continued venlafaxine throughout; another discontinued tramadol only one week before the high dose). Key inclusion criteria were moderate-to-severe major depression (HAM-D score ≥16) and treatment resistance defined as failure to respond to at least two pharmacologically distinct antidepressant trials. Functional MRI data were acquired on a 3T Siemens Tim Trio. The task was a blocked emotional face paradigm presenting fearful, neutral and happy faces (15 s blocks, faces shown for 3 s each) across an eight-minute run. Participants pressed a button on each face to ensure attention. Baseline scanning occurred before any intervention; the post-treatment scan was performed the morning after the high-dose psilocybin session. Preprocessing combined tools from FSL, AFNI, Freesurfer and ANTS and included motion correction, brain extraction, registration to standard space, scrubbing for framewise displacement >0.90 mm, spatial smoothing (6 mm FWHM), high-pass filtering (0.01 Hz) and regression of motion and CSF/white matter nuisance signals. The principal analytic approach was seed-based psychophysiological interaction (PPI) analysis, which assesses task-related changes in FC between a seed region's time-series and other brain areas. Separate PPI models used bilateral amygdala and a predefined vmPFC seed. Two PPI variants were run: a simple model contrasting all faces versus rest (primary, more powered) and a complex model modelling each expression (happy, fearful, neutral) separately (secondary). Contrast images comparing post- versus pre-treatment were entered into mixed-effects higher-level analyses (FLAME-1) with whole-brain cluster correction (z>2.3, cluster p<0.05). For hypothesis-driven tests, complementary ROIs (vmPFC when amygdala was seed and vice versa) were used as pre-threshold masks. Clinical measures included the Beck Depression Inventory (BDI) as the primary outcome at one week, the Ruminative Response Scale (RRS) for rumination, and the State-Trait Anxiety Inventory (STAI) and in-scanner state ratings for anxiety. Relations between FC changes (beta estimates) and clinical outcomes were assessed with Pearson correlations and independent t-tests; multiple comparisons were handled using Bonferroni corrections applied per cluster and construct.

Task-related FC changes after psilocybin differed by seed and expression. Using the amygdala as the seed, the simple PPI model (faces>rest) revealed increased connectivity post-treatment between the amygdala and visual cortical regions, including intracalcarine and supracalcarine cortex, cuneus, precuneus and right lateral occipital cortex. In the complex model, stronger amygdala–visual cortex connectivity was observed for happy (zmax=3.96, p<0.001) and neutral (zmax=3.31, p<0.001) faces but not for fearful faces; contrasts of fearful versus neutral did not show significant differences. When the vmPFC served as the seed, the primary PPI likewise showed greater post-treatment FC with left occipital and parietal visual areas (supracalcarine/intracalcarine cortex, precuneus, lingual gyrus). Importantly for the a priori hypothesis, pre-thresholded analyses (masking by the bilateral amygdala) indicated that vmPFC–right amygdala FC was higher before than after psilocybin during face processing—that is, vmPFC→right amygdala connectivity decreased post-treatment. The complex model clarified that vmPFC connectivity increases after treatment were condition-dependent: vmPFC showed enhanced FC with the right angular gyrus and lateral occipital cortex during fearful faces (zmax=3.74, p=0.044), with occipito-parietal visual cortex during happy faces (zmax=3.32, p=0.040), and with left cerebellum during neutral faces (zmax=3.54, p<0.001). Masked analyses indicated reduced vmPFC–right amygdala FC post-treatment for fearful (zmax=3.01, p=0.032) and neutral (zmax=3.47, p=0.041) faces but not for happy faces. Correlations between PPI-derived FC and amygdala activation from the task analysis were limited. The vmPFC×fearful PPI contrast did not correlate with amygdala task responses after treatment, nor did change scores; however, a significant negative correlation was present pre-treatment (r=-0.48, p=0.040), indicating that higher vmPFC–right amygdala FC before treatment associated with lower amygdala reactivity to fearful faces at baseline. Clinically, depressive symptoms (BDI) fell substantially by one week (mean reduction 22.26, SD=11.37, p<0.001), with 63.2% of participants meeting response criteria and 57.9% achieving remission at that time-point. Rumination (RRS) decreased significantly from baseline to one week and remained lower than baseline at three months, showing a quadratic pattern (drop at one week, partial rise by three months). After Bonferroni correction, amygdala FC measures were not significantly correlated with clinical outcomes. Absolute vmPFC–right amygdala FC post-treatment correlated with rumination at one week (r=0.54, p=0.018), such that lower connectivity related to lower rumination; this association survived the corrected α of 0.025. FC between the vmPFC and an occipital-parietal cluster post-treatment correlated with BDI at one week (r=-0.46, p=0.048) and with its change score (r=0.50, p=0.029). Responders and remitters differed in vmPFC–occipital-parietal FC change, with the remitter contrast remaining significant after Bonferroni correction. A larger increase in vmPFC–occipital-parietal FC change also correlated with greater decreases in STAI (r=0.54, p=0.017), though this latter finding narrowly missed the stricter corrected α of 0.013. No significant relations were found with in-scanner anxiety ratings.

Mertens and colleagues interpret the findings as partial support for their prediction that psilocybin alters amygdala–prefrontal circuitry during emotional processing. The principal observation was reduced vmPFC–right amygdala functional connectivity during face viewing after treatment; this effect was driven mainly by fearful and neutral faces. One plausible interpretation offered is that reduced top-down inhibitory input from vmPFC might disinhibit the amygdala, consistent with the previously reported post-treatment increase in amygdala reactivity. The investigators link this neural pattern to clinical improvements in rumination: lower vmPFC–amygdala connectivity after treatment associated with lower rumination scores at one week, and rumination itself showed a marked decline following psilocybin. The authors also highlight increased connectivity of both the amygdala and vmPFC with occipital–parietal visual regions post-treatment. They suggest these changes could reflect greater task engagement, enhanced emotional stimulus processing, or strengthened amygdala–visual cortex circuitry involved in detecting emotionally salient stimuli. Correlations between vmPFC–occipital-parietal connectivity and depressive and anxiety measures at one week further suggest behavioural relevance, although several of these associations did not survive conservative correction for multiple testing. The paper acknowledges several important limitations. The study was open-label without placebo or active control, the sample size was small and comprised a highly selected TRD population, and two participants did not complete recommended washout of medications. Psychophysiological interaction analyses have limited power and do not provide directional or causal inference about information flow, so the proposed reduction in top-down control remains speculative. Multiple comparisons across two PPI models and two ROIs further weaken inferential strength. The authors recommend that future work employ larger, placebo-controlled designs where feasible, incorporate methods capable of assessing causal connectivity (for example dynamic causal modelling or magnetoencephalography), consider eye-tracking or reaction-time regressors to control for attention, and attempt to disentangle drug effects from psychological-support effects.

The investigators conclude that psilocybin-assisted therapy for TRD produced post‑acute changes in task-related functional connectivity: decreased vmPFC–right amygdala connectivity during face processing that related to lower rumination, together with increased vmPFC and amygdala connectivity to occipital–parietal visual cortices that related to symptom change. They propose that increased amygdala responsiveness together with reduced prefrontal control may facilitate a renewed emotional sensitivity and acceptance that contributes to therapeutic benefit. Given the open-label design, small sample and analytic limitations, these conclusions are presented as preliminary and in need of replication with larger, better-controlled studies using methods that can test directional interactions.