Psilocybin desynchronizes brain networks

This was a randomised cross-over precision functional mapping study in healthy adults (n = 7, age 18–45) conducted at Washington University in St Louis. Eligibility required at least one prior lifetime psychedelic exposure and excluded current psychiatric disorders; one participant (P2) was excluded from most analyses because they could not tolerate fMRI while on psilocybin. Participants completed repeated MRI sessions (roughly every other day), including at least three non-drug baseline-type visits and two drug dosing visits (one psilocybin 25 mg, one methylphenidate 40 mg), with dosing-day scans acquired 60–180 minutes after ingestion. A replication protocol invited participants back 6–12 months later for baseline scans, an unblinded psilocybin session and 1–2 after sessions. Neuroimaging used a Siemens Prisma scanner with a multiband, multi-echo echo-planar imaging sequence (2 mm isotropic voxels, TR 1,761 ms) and two 15-minute resting-state fMRI runs per session. A validated auditory–visual matching perceptual task was administered in some sessions (48 trials, event-related) to probe task modulation; participants performed this task with >80% accuracy during drug sessions. Processing included thermal-noise removal (NORDIC), slice- and distortion-correction, optimal echo combination, bandpass filtering and scrubbing at a 0.3 mm framewise displacement threshold; task regressors were applied when indicated. Head motion and drowsiness were monitored in real time. Analyses emphasised individualised parcellation and network mapping: vertex-wise and parcel-wise FC matrices were generated, and individualspecific network assignments were derived with Infomap across multiple thresholds; area-level parcels were identified using boundary detection on dense connectomes. The primary outcome was precision functional mapping of FC change (distance between a scan's FC and that participant's baseline FC averaged across baseline visits). Whole-brain FC change used root-mean-squared Euclidean distance of linearised FC matrices and was normalised relative to day-to-day variability. A linear mixed-effects (LME) model treated drug condition, task, framewise displacement and drug × task as fixed effects and participant and MRI session as random effects; this model was applied vertex-wise and to other dependent variables. Statistical inference used wild bootstrapping with Rademacher resampling and threshold-free cluster enhancement (TFCE); many reported P values are uncorrected and the authors highlighted differences when P < 0.001. The investigators quantified spatial desynchronization with normalized global spatial complexity (NGSC), the normalized entropy of eigenvalues from temporal principal component analysis of the dense BOLD timeseries; NGSC ranges from 0 (maximal synchrony) to 1 (maximal desynchrony). Multidimensional scaling (MDS) was applied to parcellated FC matrices to extract dominant data-driven dimensions (notably a 'psychedelic' dimension, dimension 1). Subjective experience was assessed with the 30-item Mystical Experience Questionnaire (MEQ30). Replication and generalisation were tested using previously published i.v. psilocybin and LSD datasets and stimulant effects were compared with ABCD Study stimulant use data.

Acute effects: Psilocybin induced large, widespread FC changes across cortex and subcortex, with strongest effects in association cortex and the default mode network (DMN). Association cortex showed greater mean FC change (mean (s.d.) 0.44 (0.03)) than primary cortex (0.36 (0.05)). Subcortical loci with pronounced disruption included DMN-connected thalamic regions, basal ganglia, cerebellar DMN areas and focal anterior hippocampus coordinates (MNI ±24, -20, -16). By contrast, MTP-associated FC changes were smaller and localised mainly to sensorimotor systems and resembled day-to-day variability. Magnitude and normalised comparisons: When normalised to day-to-day variability (set to 1), average whole-brain FC change values were: day-to-day 1; task 1.22; MTP 1.10; high head motion 1.29; psilocybin 3.52; between-person 3.53. Thus, psilocybin produced more than threefold greater FC change than MTP (post hoc t-test P = 3.6 × 10^-6, uncorrected) and effects were comparable in magnitude to differences between individuals. Relationship to subjective experience: Whole-brain FC change tracked strongly with MEQ30 total scores across drug sessions and participants (LME predicting MEQ30: effect of FC change t(13) = 7.68, P = 3.5 × 10^-6, uncorrected; whole-brain regression r^2 = 0.81). The MEQ30 subscale most tightly associated with FC change was the transcendence of time and space (for example, 'loss of your usual sense of time or space', r^2 = 0.86), though MEQ30 dimensions were highly intercorrelated. Head motion did not account for MEQ30 scores. A likelihood ratio test indicated inter-individual variability in psilocybin brain effects exceeded measurement error (participant-specific response P = 0.00245, uncorrected). Data-driven dimension and cross-dataset replication: Classical MDS isolated a dominant dimension (dimension 1) that separated psilocybin scans from others and corresponded to reduced segregation between the DMN and typically anticorrelated networks (fronto-parietal, dorsal attention, salience, action-mode). Applying the same dimension weights to external psilocybin and LSD datasets showed increased dimension 1 scores in nearly every participant, suggesting a common psychedelic FC dimension across drugs and datasets. Desynchronization (NGSC): Psilocybin acutely increased NGSC (whole-brain LME estimate 0.0510, 95% CI 0.0343–0.0676; t(265) = 6.8; P = 2.0 × 10^-6, uncorrected), indicating increased spatial entropy or desynchronization of BOLD signals; values returned to baseline by the next session. Area-level NGSC increases were also significant (estimate 0.0149, 95% CI 0.0071–0.0228; t(265) = 3.74; P = 2.30 × 10^-4). NGSC correlated with subjective intensity (MEQ30: r = 0.80, P = 3.52 × 10^-4 after one outlier removal). The spatial distribution of NGSC changes correlated with PET-derived 5-HT2A receptor density (NGSC psilocybin vs Cimbi-36 binding r = 0.39, P = 1.9 × 10^-13; NGSC LSD vs Cimbi-36 r = 0.32, P = 4.5 × 10^-9). Task modulation: Performing a simple perceptual matching task during scans significantly attenuated psilocybin-associated FC disruption and NGSC increases (LME interaction task × psilocybin: FC change P = 5.49 × 10^-5; NGSC P = 4.82 × 10^-8, uncorrected). Task-evoked responses in primary visual cortex were reduced by psilocybin, replicating classic findings that psychedelics reduce sensory-evoked responses. Persistent effects: Whole-brain FC largely returned to baseline 1–21 days after psilocybin (normalized FC change ~1.05 (0.94, 1.27)), but a targeted screen identified a significant persistent decrease in FC between the anterior hippocampus and the DMN during the 3-week post-psilocybin window (LME mean change 0.095; P pre–post psilocybin = 0.0033, uncorrected). AntHip–DMN FC mean (95% CI) was 0.180 (0.169, 0.192) pre-psilocybin and 0.163 (0.150, 0.176) post-psilocybin. No persistent anterior hippocampus FC change was observed after MTP (equivalence test and LME non-significant). AntHip–DMN FC appeared to return to baseline by the 6–12 month replication visit, but that replication sample (n = 4 with one pre-psilocybin visit each) was underpowered to detect small changes.

The investigators interpret their findings as evidence that psilocybin produces large-scale desynchronization of brain activity that dissolves the usual segregation of functional networks, especially the DMN and its interactions with association networks. This desynchronization, quantified by NGSC, links micro-scale electrophysiological observations of reduced neuronal co-activation with macro-scale FC changes observed with fMRI and is tightly related to the intensity of the subjective psychedelic experience measured by MEQ30. Siegel and colleagues argue that the desynchronization mechanism may underlie both the acute phenomenology of psychedelics (altered sense of self, time and space) and trigger homeostatic plasticity processes that lead to persistent neurotrophic changes. They highlight the persistent reduction in anterior hippocampus–DMN connectivity as a candidate neuroanatomical correlate of pro-plasticity and therapeutic effects: the anterior hippocampus is a site of synaptogenesis in animal models and is implicated in self-related and present-moment processing. The authors also note that task engagement substantially attenuates psilocybin-driven desynchronization, providing a neural basis for the clinical practice of 'grounding' during psychedelic-assisted therapy and demonstrating context dependence of drug effects. Key limitations acknowledged in the text include the small sample size, the use of non-depressed healthy volunteers (so inference to clinical antidepressant mechanisms requires patient studies), and that many P values reported are uncorrected; the replication visit was underpowered to detect small persistent effects. The authors emphasise that their precision, individual-level mapping approach enabled detection of these effects and that future work should measure human neurotrophic markers and conduct precision studies in patient populations to verify the proposed antidepressant mechanism and to further link cellular, circuit and psychological levels of effect.