Altered network hub connectivity after acute LSD administration

Research interest in classic hallucinogens such as psilocybin, DMT and LSD has increased because their acute effects resemble some features of psychosis and because single or few administrations have shown promising, sometimes long-lasting, therapeutic effects in conditions such as anxiety and depression. Functional magnetic resonance imaging (fMRI) studies have begun to probe the neuronal correlates of these altered states by assessing functional connectivity (FC) within and between resting state networks (RSNs). Earlier reports found that LSD and psilocybin reduce coactivation within several RSNs while increasing connectivity between networks, but those studies were limited by small samples, differences in head motion between conditions, and a lack of independent replications. Müller and colleagues set out to replicate and extend prior findings by testing the acute effects of oral LSD (100 μg) versus placebo on resting-state FC in a double-blind, randomized, cross-over sample of healthy volunteers. The study used independent component analysis (ICA) to identify ten well-established RSNs and compared within-network and between-network connectivity across drug conditions. Exploratory seed-to-voxel analyses probed whole-brain changes in RSN connectivity, with particular attention to whether hub structures (for example thalamus and striatum) showed altered coupling. The investigators also explored associations between FC changes, subjective drug effects (5D-ASC), plasma LSD levels and physiological measures.

This was a randomized, placebo-controlled, double-blind, cross-over fMRI study. Each participant received 100 μg LSD in one session and placebo in the other, with at least seven days washout between sessions. Twenty healthy participants (10 male, 10 female; mean age 32 ± 11 years, range 25–60) were included in the analysed sample after exclusion of four datasets for excessive head motion. Exclusion criteria included major psychiatric disorder (personal or first-degree relative), significant medical conditions, recent or heavy illicit drug use, and pregnancy; only two participants had previously used a hallucinogen on a single occasion. Resting-state fMRI was acquired on a 3 T scanner (Magnetom Prisma). Participants kept their eyes closed; 300 volumes were collected (repetition time 1.8 s). Anatomical T1-weighted images were also obtained. Physiological measures (blood pressure, heart rate, body temperature) and venous blood for plasma LSD were taken two and three hours after administration; subjective effects were assessed three hours after dosing using the 5D-ASC questionnaire. Preprocessing combined FSL and SPM procedures: brain extraction, realignment, slice-timing correction, co-registration to T1, segmentation, normalization to MNI space and 5 mm smoothing. ICA-specific preprocessing included high-pass filtering (0.01 Hz) and dimension reduction to 20 components using PCA. Data quality steps for network and seed analyses included scrubbing (global signal z > 3 or composite motion > 0.5 mm), regression of six motion parameters, CompCor removal of five principal components from white matter and CSF, and band-pass filtering (0.008–0.09 Hz). Probabilistic ICA (MELODIC) was constrained to 20 components; after visual inspection eight components judged artifactual were removed, leaving 12 components that were matched to ten canonical RSN templates (visual networks 1–3, DMN, cerebellum, sensorimotor, auditory, executive control, frontoparietal 1–2). Within-network FC differences were tested using dual regression and paired permutation tests (Randomise) with threshold-free cluster enhancement (TFCE); a Bonferroni-adjusted family-wise error rate of p < 0.005 was used to account for testing ten networks. Between-network FC was assessed by extracting weighted ROI time series from the unthresholded ICA-derived RSNs and computing bivariate correlations between all network pairs, with paired t-tests and false discovery rate (FDR) correction across comparisons (significance at FDR p < 0.05). Seed-to-voxel analyses correlated each RSN time course with every brain voxel and used paired t-tests with cluster-level FDR correction (cluster-forming p < 0.001, cluster FDR p < 0.005). Exploratory Pearson correlations tested relationships between FC changes (Δ parameter estimates) and 5D-ASC total and subscale scores, the specific ego-dissolution item, plasma LSD at 2 h (immediately before scanning), and changes in physiological parameters. Multiple-comparison corrections were applied (FDR) where reported.

Sample and network identification: Of 20 ICA components retained after artifact removal, ten matched canonical RSN templates (visual 1–3, DMN, cerebellum, sensorimotor, auditory, executive control, frontoparietal 1–2). The final analysed sample comprised 20 subjects after excluding four for head motion; no significant motion differences between conditions were reported. Within-network connectivity: Compared with placebo, LSD acutely decreased coactivation within several RSNs. Significant reductions were observed in visual networks 1 and 3 (covering large extents of those networks), extensive portions of the sensorimotor network, the medial–posterior portion of the DMN, and a small cluster in the right inferior frontal gyrus within the auditory network. Small pockets of increased coactivation under LSD were noted outside some network masks (visual networks 2 and 3 in right superior parietal lobe and precuneus, and frontoparietal network 1 in left inferior frontal gyrus). Between-network connectivity: LSD produced widespread increases in FC between RSNs; no decreases were detected. All investigated networks showed some increase in between-network coupling, most prominently visual network 1, the cerebellum and the executive network. Seed-to-voxel and hub overlap: Seed-to-voxel analyses confirmed broad increases in RSN-to-voxel FC under LSD. Regions repeatedly implicated across seeds included cerebellar and occipital areas (lingual gyrus, intracalcarine cortex, precuneus), medial frontal areas (frontal medial cortex, frontal pole, anterior cingulate cortex) and subcortical structures (bilateral thalamus, putamen, pallidum, caudate). Overlap maps showed maximal convergence (>6 networks) in bilateral thalamus and caudate. When all ten RSNs were entered together to assess common effects, significant drug-related increases were observed in putative hub regions: thalamus, putamen, caudate, cerebellum, precuneus, anterior cingulate cortex and frontal pole/middle frontal gyrus. Associations with subjective effects, plasma and physiology: Two uncorrected correlations were reported between decreased FC within the auditory network and 5D-ASC dimensions (visionary restructuralization r = -0.53, p = 0.02; auditory alterations r = -0.49, p = 0.03), but these did not survive FDR correction. No other significant associations between within-network FC and subjective ratings were found; specifically, there was no relationship between DMN integrity and the ego-dissolution item (r = 0.06, p = 0.81). Plasma LSD levels at 2 h did not show significant correlations with FC measures. Some uncorrected correlations arose between changes in physiological measures (systolic and diastolic blood pressure, body temperature) and particular between-network connections, but none remained significant after multiple-comparison correction and there was no evidence for a systematic physiological confound.

Müller and colleagues interpret their findings as a replication and extension of prior work showing that acute LSD decreases coactivation within several RSNs while increasing connectivity between networks on a global scale. The within-network reductions observed here—particularly in visual, sensorimotor and medial–posterior DMN regions—align with some earlier LSD and psilocybin reports, although exact network-level overlap across studies was inconsistent. The investigators note that similar within-network decreases have also been reported after administration of the SSRI sertraline, which raises questions about the specificity of these within-network alterations to the subjective, hallucinogenic effects of LSD and suggests they could reflect nonspecific serotonergic modulation. A central novel emphasis of the present study is that seed-to-voxel and overlap analyses consistently implicated hub structures—notably bilateral thalamus and striatum (caudate, putamen, pallidum), plus precuneus and anterior cingulate—increased in their connectivity with multiple RSNs under LSD. The authors highlight that thalamocortical and corticostriato-thalamo-cortical circuitry alterations resemble patterns repeatedly reported in schizophrenia research, suggesting a possible mechanistic link for why hallucinogens can transiently induce psychosis-like phenomena in vulnerable individuals. At the same time, the authors refer to a recent model proposing that acute destabilisation of hub connectivity by hallucinogens could underlie their therapeutic effects, by allowing the emergence and later stabilisation of new connectivity patterns. The study's limitations as acknowledged by the authors include a relatively small sample size and a single moderately high oral dose of LSD, the indirect nature of fMRI as a measure of neuronal activity (with potential drug effects on neurovascular coupling), and physiological confounds that were measured at only one pre-scan time point. The fMRI environment itself might have influenced subjective experiences. Methodological choices—such as constraining ICA to 20 components to match prior templates—could also affect results. Finally, the authors stress that although some network perturbations mirror features of schizophrenia, hallucinogen-induced states and schizophrenia differ in symptom profile and phenomenology. Overall, the investigators conclude that while decreased within-RSN coactivation and broadly increased between-network connectivity are reproducible group-level effects of LSD, the lack of consistent associations with subjective measures and discrepancies across studies make it doubtful that any single FC metric is a reliable, characteristic neural signature of hallucinogen action. The consistent involvement of subcortical and cortical hub regions, however, provides a plausible neurobiological substrate linking acute psychedelic effects, psychosis models and hypotheses about therapeutic mechanisms.

The authors conclude that acute oral LSD reproduces previous observations of decreased functional connectivity within several RSNs and broadly increased connectivity between networks. However, they caution that these within- and between-network measures lack clear and specific associations with subjective drug effects and can resemble changes produced by non-hallucinogenic serotonergic drugs, calling into question their specificity as biomarkers of psychedelic states. Importantly, increased coupling between networks and cortical/subcortical hub structures (notably thalamus and striatum) was a robust finding and is consistent with patterns reported in psychosis and with theoretical models positing that transient alterations in hub connectivity may both produce psychotomimetic effects and enable therapeutic reorganisation of brain networks.