Functional connectivity measures after psilocybin inform a novel hypothesis of early psychosis

Psilocybin is a classic tryptamine psychedelic and the prodrug of psilocin; it produces a range of subjective effects from mild perceptual changes to profound existential experiences. Functional MRI studies of spontaneous, task-free activity have established large-scale resting-state networks (RSNs) such as the default-mode network (DMN) and task-positive networks (TPNs). Previous work shows that DMN activity is associated with internally oriented cognition and that the DMN and TPNs are typically anti-correlated or orthogonal, a relationship thought to support the separation of introspective and externally focused mental states. Alterations in DMN-TPN coupling have been reported in clinical and pharmacological contexts, including increased DMN-TPN coupling in psychosis and with sedative anaesthesia, while thalamocortical coupling has been linked to arousal. Carhart-Harris and colleagues set out to test how psilocybin affects resting-state functional connectivity, focusing on two hypotheses: that psilocybin would increase coupling between the DMN and TPNs (reducing their normal orthogonality), and that thalamocortical connectivity would be preserved under psilocybin, distinguishing the psychedelic state from sedative states. By comparing psilocybin to placebo in healthy volunteers using complementary independent component analysis (ICA) and seed-based approaches, the study aimed to characterise between-network and thalamocortical connectivity changes and relate them to subjective effects.

Fifteen healthy volunteers (13 males, 2 females; mean age 32, SD 8.9) who had used psilocybin previously but not within 6 weeks were enrolled. Inclusion criteria required participants to be at least 21 years old with no personal or first-degree family history of major psychiatric disorder, no substance dependence, cardiovascular disease, or prior significant adverse reaction to a hallucinogen. Each subject underwent two eyes-closed resting-state BOLD fMRI scans on separate occasions at least 7 days apart: one following an intravenous saline placebo (10 ml over 60 s) and one following psilocybin (2 mg in 10 ml saline, infused over 60 s). The 12-minute scans began injections 6 minutes into acquisition; the subjective effects were rated postscan using visual analogue scales. Imaging used a 3T GE HDx system. Anatomical T1 scans preceded functional runs. BOLD echo-planar imaging parameters were TR/TE = 3000/35 ms with 53 oblique axial slices at 3 × 3 × 3 mm voxel size. For ICA, FSL MELODIC derived 20 group components from concatenated pre-injection data (first 6 minutes, 100 volumes per scan). After excluding components judged to be noise (white matter, ventricles, extracerebral), 11 functionally meaningful RSNs were retained, including an anterior DMN (aDMN) and posterior DMN (pDMN), dorsal attention network (DAN), salience network (SAL), auditory (AUD), and frontoparietal networks. Between-network FC was assessed both with ICA-derived components and with a seed-based approach. For ICA-based analyses the aDMN time series (post-injection, last 100 volumes) served as the dependent variable in linear regression models run in SPSS; each other RSN was entered as an independent variable together with nine noise components to remove non-neuronal variance. Unstandardised regression coefficients were compared across conditions using paired two-tailed t tests and Pearson correlations with selected subjective items; Bonferroni correction was applied for multiple comparisons. In the seed-based analysis a ventromedial PFC (vmPFC) seed defined DMN (vmPFC-positive) and TPN (vmPFC-negative) spatial masks; global grey matter signal regression was not performed. Time series for DMN and TPN (pre- and post-injection, excluding 40 volumes around injection) were entered into linear regressions with white matter, CSF and motion regressors, and regression coefficients were compared across conditions. Thalamic connectivity was tested using a bilateral thalamic anatomical mask transformed to each subject's functional space. Thalamic time series served as the dependent variable in regressions with DMN and TPN time series as independent variables; white matter, CSF and motion were included as nuisance regressors. The study also examined movement differences between conditions by calculating mean movement per volume and tested whether movement differences correlated with observed FC changes. Correlational analyses tested relationships between altered DMN–RSN coupling and five questionnaire items chosen for relevance to ego-boundary disturbance and drug intensity.

ICA identified 11 meaningful RSNs including an anterior and posterior DMN. Comparing post-injection connectivity under psilocybin versus placebo, significant increases in FC between the aDMN and four networks were observed: salience network (P = .0002), right frontoparietal network (P = .0003), auditory network (P = .0006), and dorsal attention network (P = .0009). A decrease in coupling between anterior and posterior DMN (P = .02) was suggested but did not survive Bonferroni correction (corrected α = 0.005). Seed-based vmPFC analyses produced DMN and TPN masks and confirmed increased DMN–TPN functional connectivity after psilocybin: psilocybin versus placebo post-scan comparison yielded P = .001, and post- versus pre-psilocybin comparison yielded P = .009. These complementary analyses therefore converged on increased coupling between DMN and task-positive networks under psilocybin. Correlational analyses between altered DMN–RSN FC and subjective ratings produced some suggestive relationships but none that survived correction for multiple comparisons. For example, the item "my thinking was muddled" showed a relationship with increased DMN–right frontoparietal FC at P = .02, but the corrected significance threshold was P = .002; ratings of overall drug intensity showed a non-significant suggestion of association with increased DMN–TPN FC (P = .16). Thalamic connectivity analyses did not show the reductions reported in sedation studies. Thalamic–DMN FC showed a non-significant increase after psilocybin, while thalamic–TPN FC showed a significant increase; no correlations were found between increased thalamocortical FC and subjective ratings of drug intensity. Movement analyses revealed greater mean movement per volume under psilocybin (mean 0.1 mm, SD 0.05) than placebo (mean 0.06 mm, SD 0.015), P < .01. However, correlational tests found no relationship between between-condition movement differences and the observed increases in between-network coupling (aDMN–DAN, aDMN–rFPN, aDMN–SAL, aDMN–AUD), arguing against movement as the primary driver of the FC findings.

Independent component and seed-based analyses converged on a principal finding: psilocybin increased coupling between the DMN and several task-positive networks, indicating reduced orthogonality between internally and externally oriented systems. The affected networks included canonical TPNs (dorsal attention, salience, right frontoparietal) and an auditory network. Because DMN and TPN activity is normally segregated, the authors interpret increased DMN–TPN FC as a loss of distinctiveness between introspective and externally focused modes of cognition in the psychedelic state. Comparing these results with prior pharmacological work, the investigators note that increased DMN–TPN coupling has also been observed with propofol sedation, but the psychedelic state differs phenomenologically: participants did not report reduced consciousness under psilocybin. To reconcile this, they highlight thalamocortical findings: whereas propofol is associated with marked thalamocortical decoupling and reduced arousal, psilocybin produced preserved or even increased thalamocortical connectivity. From this pattern the authors propose that increased DMN–TPN coupling in the presence of preserved thalamocortical connectivity reflects a change in the mode of consciousness rather than a decline in arousal or awareness. The discussion links the neuroimaging results to clinical and experiential phenomena. Increased DMN–TPN coupling has been reported in schizophrenia and in people at high risk for psychosis; the authors suggest that reduced network orthogonality may help explain disturbances of ego boundaries and the blurring of internal and external distinction seen in early psychosis. They also note parallels with nondual meditation and spiritual-type experiences, where altered DMN function and reports of "ego dissolution" are common. The authors frame the psychedelic and early psychotic states as conditions of relative neural disorganisation accompanied by heightened plasticity, and they advance the idea that psychedelics can serve as models of the prodrome to psychosis and as tools for studying constructs such as the sense of self. The investigators address potential confounds: although movement was greater under psilocybin, movement regressors were included and no correlation was found between movement differences and FC changes. They acknowledge that correlations between FC changes and subjective measures were suggestive but did not survive correction. Finally, they emphasise that this is the first assessment of between-network FC after a psychedelic and that the pattern of increased DMN–TPN coupling together with preserved thalamocortical connectivity may distinguish the psychedelic state from sedation and bear relevance to early psychosis and mystical-type experiences.