Decreased brain modularity after psilocybin therapy for depression

Depression remains highly prevalent and current antidepressant drugs have modest efficacy, problematic side effects, discontinuation issues and high relapse rates, motivating the search for new treatments. The authors describe depressive episodes using dynamical-systems language — as constrained, ‘‘attractor’’ states — and note that neuroimaging has repeatedly implicated abnormal functioning of higher-order networks such as the default mode network (DMN), executive network (EN) and salience network (SN). They further note that the 5-HT2A receptor, the principal binding site of classic serotonergic psychedelics such as psilocybin, is densely expressed across cortex overlapping these higher-order networks. Daws and colleagues set out to test whether psilocybin therapy produces sub-acute changes in spontaneous brain activity, operationalised as reduced brain network modularity in resting-state fMRI, and whether such changes relate to antidepressant outcomes. They focused on two clinical trials that included pre- and post-treatment fMRI: an open-label trial in treatment-resistant depression and a double-blind randomised controlled trial (DB-RCT) that compared psilocybin therapy with daily escitalopram. The investigators hypothesised that psilocybin would produce post-treatment network ‘‘desegregation’’ (lower modularity), that this effect would relate to clinical improvement, and that comparable changes would not be observed after escitalopram.

Two clinical trials with pre- and post-treatment resting-state fMRI were analysed. Trial 1 was an open-label study in treatment-resistant depression (TRD). Trial 2 was a two-arm double-blind randomised controlled trial (DB-RCT) in major depressive disorder (MDD) comparing psilocybin therapy with daily escitalopram. General exclusion criteria for both trials included personal or family history of psychosis, significant physical health risk, history of serious suicide attempts, pregnancy and MRI contraindications; the DB-RCT additionally excluded participants with SSRI contraindications or prior escitalopram use. Eligible patients completed telephone screening, provided written consent and had medical and psychiatric histories evaluated. Interventions differed by arm and trial. In the open-label TRD trial participants completed an eyes-closed resting-state fMRI at baseline and then received two oral psilocybin dosing days (DDs) one week apart: 10 mg on DD1 and 25 mg on DD2. Follow-up fMRI and clinical assessment took place one day after DD2 and clinical measures were repeated at 1 week, 3 months and 6 months. In the DB-RCT a neuroimaging subsample of 43 adults was analysed; 30 were randomised to the psilocybin-arm and 29 to the escitalopram-arm in the full trial. On DD1 participants received either 25 mg psilocybin (psilocybin-arm) or 1 mg psilocybin (escitalopram-arm, intended as negligible); DD2 repeated the same dosage three weeks later. Beginning one day after DD1, participants took daily capsules for six weeks: placebo capsules for the psilocybin-arm and escitalopram (10 mg for three weeks, then 20 mg) for the escitalopram-arm. Resting-state fMRI was recorded at baseline and three weeks after the second psilocybin dose (the final day of capsule ingestion). The principal clinical outcome was the Beck Depression Inventory 1A (BDI-1A), a patient-rated measure emphasising cognitive features of depression. In the open-label trial BDI was collected at baseline and 1 week, 3 months and 6 months post-DD2; in the DB-RCT BDI was collected at baseline and 2, 4 and 6 weeks post-DD1. Resting-state fMRI scans were acquired on a 3T Siemens Tim Trio scanner. Functional connectivity (FC) matrices were estimated from 100 cortical regions defined by a functional atlas. The main neuroimaging metric was network modularity, calculated via a common community detection algorithm that partitions regions into non-overlapping modules; higher modularity indicates greater segregation between networks. Study 1 additionally used functional cartography to estimate network recruitment (probability regions form communities with same-network regions) and integration (probability of cross-network community formation). Study 2 applied dynamic community detection using a sliding-window approach to derive network flexibility, defined as the average number of times regions changed community affiliation across time windows. Statistical analyses included within-subject comparisons of modularity between sessions, Pearson correlations between modularity and BDI scores (with false discovery rate correction for multiple comparisons where reported) and arm-by-timepoint ANOVA for clinical scores. The extracted text did not provide detailed preprocessing parameters here but referenced supplemental methods for MRI acquisition and preprocessing.

Samples and clinical outcomes: In the open-label TRD trial 19 patients were recruited and 3 were excluded for excessive head motion, leaving 16 patients for analysis (mean age 42.8, 4 female) with baseline BDI indicating severe depression (mean BDI=34.81). Marked and sustained reductions in BDI were reported: at 1 week post-treatment mean difference -21.0 points (95% CI -27.30 to -14.71, P<.001) and at 6 months mean difference -14.19 (95% CI -21.29 to -7.09, P<.001). In the DB-RCT neuroimaging sample 43 adults were analysed (mean age 42.7, 14 female). Across the neuroimaging sample, antidepressant effects favoured the psilocybin-arm with significant arm-by-timepoint interaction for BDI (ANOVA F=4.47; P=0.005). Pairwise comparisons relative to baseline favoured psilocybin at 2 weeks (mean difference -8.73; 95% CI -13.55 to -3.91, P=0.002), 4 weeks (mean difference -7.79; 95% CI -13.62 to -1.95, P=0.013) and 6 weeks (mean difference -8.78; 95% CI -15.58 to -1.97, P=0.013). Primary neuroimaging findings: Confirming the primary hypothesis, brain network modularity decreased after psilocybin therapy. In the open-label sample modularity was significantly reduced one day post-treatment (reported mean difference -0.29; P=.012). Note: the extracted CI reported alongside this mean difference appears inconsistent in direction (95% CI 0.07 to 0.50), so the CI as presented in the extraction is unclear. In the DB-RCT the psilocybin-arm also showed a significant reduction in modularity (mean difference -0.39; 95% CI -0.75 to -0.02, P=0.039). This modularity reduction was specific to the psilocybin-arm: the escitalopram-arm showed no change from baseline to week 6 (mean difference 0.01; 95% CI -0.35 to 0.33, P=0.945). Associations with clinical response: In the open-label trial lower post-treatment modularity correlated with better long-term clinical outcomes. After false discovery rate correction a strong correlation was observed at 6 months (Pearson r=0.64; P=.023), and the pre-to-post change in modularity correlated with change in BDI at 6 months (Pearson r=0.54; P=.033). In the DB-RCT individuals' decreases in modularity correlated with greater depression recovery at the 6-week primary endpoint within the psilocybin-arm (Pearson r=0.42; P=.025, one-tailed). No significant relationship between modularity change and BDI was observed in the escitalopram-arm (Pearson r=0.08; P=0.361, one-tailed). Network-specific and dynamic results: In the open-label sample psilocybin therapy produced decreased DMN recruitment (mean difference -0.54; 95% CI -0.92 to -0.15, P=.009) and increased DMN integration with frontoparietal networks, including DMN–EN (mean difference 0.53; 95% CI 0.15 to 0.90, P=.01) and DMN–SN (mean difference 0.55; 95% CI 0.14 to 0.95, P=.01). The DB-RCT dynamic analysis showed that post-treatment increases in network flexibility, particularly in the executive network, related to clinical improvement: increased EN dynamic flexibility correlated with greater recovery at 6 weeks (Pearson r=-0.76; P=0.001). Significant flexibility–outcome relationships also involved the salience and dorsal attention networks. No significant correlations between BDI change and dynamic flexibility were observed in the escitalopram-arm. The extracted text did not present a detailed adverse-events table or other safety data within the supplied sections.

Daws and colleagues interpret their convergent findings across two trials as evidence that psilocybin therapy is associated with decreased brain network modularity and increased inter-network integration post-treatment, and that these changes relate to antidepressant outcomes. They argue that these post-acute effects resemble, at an attenuated level, the increased global functional connectivity and broadened state-space reported during the acute psychedelic state. The authors propose a mechanistic model in which 5-HT2A receptor agonism by psilocybin dysregulates spontaneous cortical activity, temporarily ‘‘disintegrating’’ modular network structure and flattening the brain's functional energy landscape, thereby expanding available brain states; this expanded state-space is hypothesised to underlie improved mood, cognitive and psychological flexibility. Relative to escitalopram, the investigators emphasise that the modularity changes and their relationship to clinical improvement were specific to psilocybin; chronic escitalopram did not alter modularity in their sample. The authors suggest this specificity may reflect pharmacological differences, with escitalopram exerting broader serotonergic effects and a predominant action via inhibitory 5-HT1A receptors in limbic circuits rather than the cortical 5-HT2A-driven effects implicated for psychedelics. They also note that decreased modularity and increased flexibility of frontoparietal networks, particularly the executive and salience networks, are consistent with models that link rigid, internally focused cognition in depression to abnormally segregated brain community structure. The investigators acknowledge limitations and practical considerations. They recommend modularity metrics as potentially sensitive biomarkers but caution that fMRI data are noisy and that dynamic flexibility analyses require long, high-quality time series which can be challenging to obtain in clinical samples. They also state they did not formally assess cognitive flexibility in these trials, although they observed improvements in general cognitive functioning in the DB-RCT. Finally, the authors call for Phase III trials to pursue licensing and for pragmatic studies to address treatment optimisation and implementation, and they encourage continued use and development of network-level analyses in imaging studies of psychedelic therapies.

The authors conclude that decreased brain network modularity is a robust, reliable and potentially specific biomarker of response to psilocybin therapy for depression. They suggest this neural change may help explain psilocybin's antidepressant effects and support further clinical development of psilocybin therapy as a treatment option for depression.