Effective connectivity of functionally anticorrelated networks under LSD

Psilocybin is a serotonergic psychedelic whose agonism at 5-HT2A receptors produces profound alterations of perception, emotion and self-experience, including graded experiences of ego dissolution. Previous work has implicated the amygdala and large-scale resting-state networks (RSNs) — notably the default mode network (DMN), central executive network (CEN) and salience network (SN) — in emotional appraisal, self-related cognition and psychopathology. However, the directional (effective) connectivity that links cortical RSNs with the amygdala during the acute psychedelic state, and how such directed interactions relate to subjective reports, remain poorly characterised. Stoliker and colleagues set out to test whether psilocybin alters top-down effective connectivity from RSNs to the amygdala and whether these directed changes associate with aspects of the acute subjective experience. Using spectral dynamic causal modelling (spDCM) to infer directional influences among predefined cortical ROIs and the amygdala, the study specifically examined changes in DMN–amygdala, CEN–amygdala and SN–amygdala interactions and related those changes to two dimensions of the 5D-ASC (changed meaning of percepts and blissful state). The primary hypothesis was reduced top-down effective connectivity from RSNs to the amygdala under psilocybin, with such changes linked to altered meaning and positive affective experiences.

The study used a double-blind, randomised, placebo-controlled, cross-over design. Healthy volunteers were recruited and screened; the extracted text reports 24 participants (12 males, 11 females; mean age 26.3 years; range 20–40), with one subject not completing the study. Standard psychiatric and medical screens excluded current or past major psychiatric disorders and other medical contraindications. Participants abstained from prescription and illicit drugs for two weeks prior to testing and from alcohol for 24 hours before sessions; urine screens verified compliance. Participants received oral psilocybin (reported in the Design section as 0.2 mg/kg) or placebo in two sessions separated by two weeks. Resting-state fMRI scans of 10 minutes were acquired at 20, 40 and 70 minutes after administration, but only the 70-minute scans were analysed for this paper. Subjective measures from the five dimensions of altered states of consciousness scale (5D-ASC), specifically the subscales “changed meaning of percepts” and “blissful state”, were collected 360 minutes after administration and used in brain–behaviour analyses. Regions of interest (ROIs) for the DMN, CEN, SN and the bilateral amygdala were identified using Neurosynth peak coordinates and refined by expert visual inspection; time series were extracted as the first principal component within 6 mm spheres around those coordinates. Preprocessing regressed six head motion parameters, white matter and cerebrospinal fluid signals, and included global signal regression. One subject was excluded because no activation was found in one or more ROIs. Three fully connected DCM models were specified, one per network plus the amygdala, using 11 ROIs in total and no exogenous inputs. Spectral DCM was inverted for each subject and condition to estimate effective connectivity that best explained the observed cross-spectral density; model fit averaged 90.9% for placebo and 89.8% for psilocybin. Group-level inferences were made using parametric empirical Bayes (PEB), a Bayesian hierarchical GLM that accounts for parameter uncertainty. For inference the authors reported effects exceeding a posterior probability threshold of > 0.99, described as very strong evidence.

The principal result was a pattern of decreased top-down effective connectivity from cortical RSNs to the amygdala under psilocybin, with several within-network changes reported at a posterior probability > 0.99. Connections not meeting this threshold were not reported. DMN–amygdala findings: Compared with placebo, the posterior cingulate cortex (PCC) showed decreased self-inhibition under psilocybin (which the authors interpret as increased synaptic gain or sensitivity to inputs). In addition, effective connectivity from the PCC to the left amygdala was decreased relative to placebo. CEN–amygdala findings: Multiple directed changes were observed among dorsolateral prefrontal cortices (dlPFC), lateral posterior parietal cortex (lPPC) and the amygdala. The right and left dlPFC (rdlPFC, ldlPFC) showed increased self-inhibition, whereas the left lPPC showed decreased self-inhibition. Directed effects included decreased effective connectivity from the lPPC to the left amygdala, increased connectivity from the lPPC to the rdlPFC and decreased connectivity from the lPPC to the right amygdala. The ldlPFC decreased its influence on the right amygdala and increased connectivity to the lPPC, while the rdlPFC increased connectivity to the ldlPFC. The right amygdala also showed increased directed excitation to the lPPC. Effect sizes were reported as posterior expectations in Hz for connections (self-connections were log-scaled); all reported results met the stated posterior probability threshold. Behavioural associations: Several directed connectivity changes correlated with the two 5D-ASC dimensions examined. For the DMN, decreased PCC self-inhibition was associated with higher ratings of blissful state, and bidirectional inhibitory coupling between the PCC and left amygdala was associated with changed meaning of percepts. Within the CEN network, excitation from the right amygdala to the lPPC correlated with blissful state. Both changed meaning of percepts and blissful state associated with decreased self-inhibition of the left lPPC. Excitatory effective connectivity from rdlPFC to ldlPFC and from ldlPFC to lPPC was associated with changed meaning of percepts. Additional notes from the analyses: The authors did not observe strong effective connectivity changes involving the medial prefrontal cortex (mPFC) in these analyses. DCM model fit statistics and the strict posterior probability threshold were reported; however, the extracted text does not clearly report the exact final sample size used for the DCM group analysis after exclusions.

Stoliker and colleagues interpret their findings as evidence that acute psilocybin reduces the recruitment of the amygdala by cortical RSNs involved in self-referential processing, executive evaluation and salience detection. They highlight increased inhibitory directed influences from cortical nodes (including PCC, dlPFC, lPPC, dACC and left anterior insula) to the amygdala under psilocybin, suggesting cortical cognitive resources are less likely to engage amygdala-mediated emotional responses during the acute psychedelic state. The observed decrease in PCC self-inhibition (interpreted as increased synaptic gain) was linked to reports of blissful state, while inhibitory PCC–left amygdala coupling related to changed meaning of percepts, leading the authors to propose that altered cortex–amygdala interactions may underpin changes to self-related emotion and the reappraisal of perceptual meaning. Within the CEN, the pattern of directed effects was more complex. The authors note increased self-inhibition in bilateral dlPFC alongside decreased self-inhibition of left lPPC and increased directed excitation among dlPFC regions. Two interpretations are offered: one likens the pattern to “hyper-frontality” seen in worry or acute psychosis, a resemblance that may be influenced by the implanting stress of scanning; the other frames the pattern as enhanced evaluative processing that could support cognitive reappraisal and resilience. The authors favour a therapeutic reading in light of observed associations between CEN-directed connectivity and changed meaning of percepts, but they acknowledge ambiguity. Regarding the SN, inhibition from dACC and left AI to the right amygdala was taken to indicate reduced salience detection under psilocybin, a change the authors suggest could be therapeutically beneficial in internalising disorders yet potentially related to psychotic-like symptoms at higher doses. They also report increased self-inhibition in bilateral amygdala and right AI, but these changes were not linked to the measured subjective effects. The authors acknowledge several limitations. The study analysed only the 70-minute post-administration resting scan and lacked pre/post scans that would strengthen causal interpretation. Region selection and network definitions can vary by method and may affect results; additional ROIs (ventral striatum, precuneus, parahippocampus and amygdala subnuclei) were not examined here. Preprocessing choices — including global signal regression — and limited temporal resolution of fMRI were noted as potential constraints. The sample was small and comprised healthy volunteers receiving a dose below those now common in clinical trials, limiting generalisability to patient populations and to different dosing regimens. Finally, the authors recommend future work using modalities with higher temporal resolution (EEG/MEG), task-based designs and more detailed subcortical parcellation to refine understanding. In conclusion, the study proposes that attenuation of RSN-to-amygdala effective connectivity and altered within-RSN directed interactions under psilocybin are candidate neural markers of altered self-related emotion and meaning, and that these directed connectivity changes merit further investigation as potential neural substrates of psychedelic therapeutic effects.