Increased amygdala responses to emotional faces after psilocybin for treatment-resistant depression

Roseman and colleagues situate their work in the recent renaissance of clinical research on classic psychedelics, noting growing evidence that psilocybin combined with psychological support can produce rapid reductions in depressive and anxiety symptoms. They outline a neurobiological rationale focused on the amygdala, a subcortical structure sensitive to emotionally salient stimuli that has been implicated in depression. Previous functional MRI studies in untreated depressed patients have often reported amygdala hyper-reactivity to negative stimuli, and conventional antidepressants such as SSRIs tend to attenuate amygdala responses early in treatment. The study aimed to test whether amygdala responses to emotional faces are altered after psilocybin treatment in people with treatment-resistant depression (TRD). Using a validated emotional faces fMRI paradigm, the investigators predicted that post-treatment amygdala reactivity would change relative to baseline and that such changes would relate to clinical measures of depressive symptoms. They paid particular attention to the fearful versus neutral contrast because prior SSRI work has emphasised attenuation to negative stimuli, and they also anticipated that aspects of the acute psychedelic experience might relate to subsequent changes in amygdala function.

This was an open-label interventional study of psilocybin with psychological support in patients with moderate to severe treatment-resistant major depression. Ethical approval was obtained from the relevant UK bodies and all participants provided written informed consent. Twenty patients were enrolled; 19 completed both pre- and post-treatment fMRI scans. Demographics for the scanned sample were six females, mean age 44.7 years (SD 10.9, range 27–64). Inclusion required a Hamilton Depression Rating Scale (HAM-D) score ≥16 and failure to improve after at least two adequate antidepressant courses. Participants were asked to be free of antidepressant medication for at least two weeks prior to study procedures. Exclusion criteria included current or past psychotic disorder, first-degree relative with psychosis, significant medical illness contraindicating participation, history of serious suicidal attempts requiring hospitalisation, history of mania, blood/needle phobia, pregnancy, and current drug or alcohol dependence. The intervention comprised preparatory psychological sessions (about 4 hours) followed by two psilocybin-assisted therapy sessions one week apart. The first session was a low ‘‘test’’ dose of 10 mg p.o. and the second a higher therapeutic dose of 25 mg p.o. During dosing, participants lay with eyes closed listening to music while two therapists provided nondirective supportive care. Baseline fMRI scanning occurred prior to any psychological or pharmacological intervention; the post-treatment scan was performed the morning after the high-dose session (both scans at approximately 10:00am). The sample size for imaging analyses was 19 (one of 20 did not complete scans). Imaging was performed on a 3T Siemens Tim Trio with a 12-channel head coil. Anatomical MPRAGE and T2*-weighted echo-planar images were acquired; functional scans used 3 mm isotropic voxels, TR 2000 ms, TE 31 ms. The emotional faces task was an 8-minute block design presenting fearful, happy and neutral expressions from the Karolinska Directed Emotional Faces set. Five faces (3 s each) of the same expression were shown per 15 s block, interleaved with 15 s rest blocks; two task versions (reverse block order) were counterbalanced across visits. Participants passively viewed faces and pressed a button with each new face to confirm attention. Preprocessing combined tools from FSL, AFNI, Freesurfer and ANTS and included motion correction, brain extraction, rigid and non-linear registration to MNI space, scrubbing using a framewise displacement threshold of 0.90 mm, spatial smoothing (6 mm FWHM), high-pass filtering (0.01 Hz) and regression of six motion parameters. Mean framewise displacement did not differ before versus after treatment and the proportion of scrubbed volumes was low (mean 2.3% before and after). Region-of-interest (ROI) and voxelwise analyses focused on the amygdala. Two complementary approaches were used: mean signal within left and right amygdala ROIs and voxelwise testing inside a bilateral amygdala mask (Harvard–Oxford atlas, probability > 50%). First-level GLMs (FEAT, FSL) modelled fearful, happy and neutral onsets convolved with a haemodynamic response, with contrasts for each emotion versus baseline and contrasts of fearful > neutral and happy > neutral (five contrasts total). Group-level mixed-effects analyses (FLAME 1+2) compared before versus after treatment. Voxelwise results used a cluster-forming threshold Z > 2.3 with cluster correction at p < .05. To examine associations with clinical outcome, voxelwise analyses within a right amygdala mask used clinical measures as regressors; the primary clinical outcomes pre-specified for this imaging analysis were Beck Depression Inventory (BDI) at 1 week and in-scanner state depression ratings (0–20) comparing baseline with one day post-treatment. The Quick Inventory of Depressive Symptomatology (QIDS) was added post-hoc to allow assessment across multiple time points (1 day, 1, 2, 3 and 5 weeks). State and trait anxiety measures (in-scanner rating and STAI at 1 week) were also explored.

Nineteen patients completed both fMRI sessions and were included in the imaging analyses. ROI analyses showed increased BOLD responses in the right amygdala after treatment for fearful faces (p = .001) and for happy faces (p = .022), with a trend for neutral faces (p = .066). After correcting for multiple tests (five contrasts × two ROIs), only the fearful-face increase in the right amygdala remained statistically significant. No significant pre–post changes were observed in the left amygdala for fearful, happy or neutral faces (p = .11, p = .95 and p = .2, respectively). Voxelwise analysis within the amygdala mask revealed significantly greater right amygdala responses post-treatment for fearful, happy and neutral faces, and for the fearful > neutral contrast. These increases localised to a consistent region of the right amygdala. Whole-brain exploratory analysis indicated additional post-treatment increases in visual cortical areas for the fearful, happy and neutral contrasts; the fearful > neutral whole-brain effect was more restricted, appearing primarily in the right amygdala and right middle temporal gyrus. Clinical outcomes showed substantial early improvement: one day after psilocybin, 15 of 19 patients (79%) had a clinically meaningful state-response (≥50% reduction in in-scanner depressed-mood rating). Changes in right amygdala activation for fearful > neutral faces were positively associated with better clinical outcomes. Specifically, greater post-treatment right amygdala reactivity to fearful > neutral predicted larger reductions in in-scanner state depression and in trait depression measured by BDI at 1 week. At 1 week the mean BDI change reported in the extracted text was 10.2 ± 5.3, with response and remission rates of 63.2% and 57.9% respectively; these clinical improvements were related to increased right amygdala responses to fearful > neutral. Using QIDS, response rates at 1 day (68.4%), 1 week (63.2%) and 3 weeks (63.2%) were associated with greater post-treatment right amygdala reactivity for fearful > neutral. QIDS response at 2 weeks (57.9%) and 5 weeks (47.3%) did not show a relationship with amygdala changes. The authors also report that responders and remitters tended to show increased amygdala reactivity while non-responders and non-remitters showed decreased reactivity for fearful > neutral. Regarding relapse, 13 subjects were responders at 1 day and 9 maintained response at 5 weeks; comparison between the 4 who relapsed and 9 who did not showed no difference in amygdala activation. The extracted text does not provide detailed adverse-event data or other safety outcomes.

Roseman and colleagues interpret their main finding—an increase in right amygdala responses to emotional faces one day after psilocybin treatment—as fundamentally different from the commonly reported SSRI effect of reduced amygdala reactivity to negative stimuli. They argue that whereas SSRIs have been proposed to exert therapeutic effects in part by attenuating amygdala responses (potentially blunting emotional reactivity), psilocybin with psychological support may instead revive or amplify emotional responsiveness, enabling patients to confront and work through difficult emotions. The observed association between increased right amygdala responses to fearful > neutral faces and subsequent clinical improvement at one week is presented as consistent with this interpretation. The authors position their results relative to prior studies, noting a discrepancy with an earlier report of reduced amygdala responses during the acute psilocybin experience in healthy volunteers. They caution that acute effects during intoxication may differ from post‑acute changes and that apparent attenuations in imaging studies can sometimes reflect reduced task engagement rather than true functional downregulation. Qualitative data from the same clinical trial are cited to support the notion that patients commonly describe greater emotional acceptance and the ability to undergo cathartic emotional processing after psilocybin. Limitations acknowledged by the investigators include the open-label, uncontrolled design, which prevents disentangling the specific contribution of psilocybin from psychological support or non-specific factors. They note the restricted imaging time‑point (one day post-treatment) and the absence of acute-phase imaging in this sample. Additional limitations discussed are the lack of a healthy control group and the absence of active comparator arms such as psilocybin without psychological support or a pharmacological control (e.g. SSRI). The authors recommend future studies with longer follow-up imaging, acute and post-acute imaging, multimodal approaches combining PET and fMRI to relate receptor changes to function, and controlled designs to test replicability and specificity of the effects. In concluding remarks within the Discussion, the authors characterise psilocybin as a potentially rapid-acting, low-abuse-potential intervention for major depression that may operate through a distinct neurobiological mechanism from SSRIs by enhancing emotional connectedness. They present selected patient quotations describing emotional release and increased acceptance as illustrative of the phenomenological changes that may accompany the observed neurobiological effects. The investigators call for replication and further work to determine whether enhanced amygdala responsiveness is causally linked to enduring antidepressant benefits.