Classical and non-classical psychedelic drugs induce common network changes in human cortex

Dai and colleagues situate their work in the context of incomplete understanding of the neurobiology of the psychedelic experience. They note that classical psychedelics such as LSD primarily act at the serotonergic 5-HT2 receptor, while non-classical drugs such as ketamine and nitrous oxide act at different molecular targets but can produce overlapping phenomenology, increased neurophysiologic complexity, and changes in oscillatory activity and brain state repertoire. Prior to this study there had been no resting-state fMRI characterisation of nitrous oxide at psychedelic concentrations in humans, and it remained unclear whether diverse psychedelics share drug-invariant network-level signatures. The study therefore set out to identify common brain network changes induced by both classical and non-classical psychedelics. Using a primary, prospective fMRI dataset of nitrous oxide administration and secondary analyses of existing ketamine and LSD fMRI datasets, the investigators compared within- and between-network functional connectivity changes associated with each drug. Propofol sedation was included as a non-psychedelic control to help distinguish general state-dependent connectivity changes from psychedelic-specific effects.

The study combined one prospectively collected dataset of nitrous oxide with three secondary datasets (ketamine, LSD, and propofol) drawn from prior studies or open repositories. Dataset 1 (nitrous oxide) comprised 16 healthy participants (8 female; mean age 24.6 ± 3.7 years) scanned at the University of Michigan. Participants underwent two resting-state scans: baseline (100% oxygen, 20 min) and during inhaled subanesthetic nitrous oxide (35% in oxygen) for 40 min; two subjects were excluded for excessive motion. Altered states of consciousness were assessed with an 11-dimension questionnaire before and after nitrous oxide exposure. Standard clinical monitoring and measures to mitigate adverse effects were used. Dataset 2 (ketamine) included 12 right-handed participants (5 female; mean age 41.4 ± 8.6 years) scanned at Fudan University; the investigators analysed the subanesthetic, pre-loss-of-responsiveness period during intravenous ketamine infusion (stepwise dosing totalling up to 1.0 mg/kg in the analysed window). Dataset 3 (LSD) used open-access preprocessed data from 15 participants (5 female; mean age 38.4 ± 8.6 years) who received 75 μg intravenous LSD with scanning approximately 70 min after dosing. Dataset 4 (propofol) comprised previously published data from 17 participants during light, subanesthetic sedation (target-controlled infusion at an effect-site concentration of 1.3 μg/ml). Scanner platforms and acquisition parameters varied across datasets; each dataset also included high-resolution anatomical scans for co‑registration. Functional MRI preprocessing was performed largely in the CONN toolbox with procedures including slice timing correction, rigid-body motion correction and frame-wise displacement (FD) calculation, co-registration to anatomy, spatial normalisation to MNI space, band-pass filtering, regression of nuisance signals (motion derivatives, white matter, CSF), and spatial smoothing (6 mm FWHM). For the LSD dataset only, preprocessed data from the repository were used; that pipeline included de-spiking, scrubbing with an FD threshold of 0.4 and replacement of scrubbed volumes with local means. The investigators primarily applied an FD threshold of 0.9 mm to preserve at least five minutes of continuous data in the nitrous oxide cohort and then applied the same threshold to ketamine and propofol for consistency; sensitivity analyses using FD = 0.4 produced sparser but broadly similar results. Functional connectivity analyses comprised two complementary approaches. ROI-to-ROI connectivity used a CONN HCP-ICA parcellation derived from the Human Connectome Project (covering seven cortical networks plus a cerebellar network and 32 ROIs) to produce Fisher-transformed pairwise correlation matrices and network-level multivariate tests of intra- and inter-network connectivity. Seed-based voxelwise analyses used the right temporoparietal junction (TPJ) as a seed to map whole-brain connectivity changes. Statistical inference for ROI and voxel analyses employed parametric cluster-based methods with false discovery rate (FDR) control at pFDR < 0.05 for multiple comparisons. Within-group paired comparisons (drug versus each dataset's own baseline) were used because of heterogeneity in scanner parameters; Spearman correlations tested relationships between nitrous-oxide-induced connectivity changes and altered-states questionnaire score changes.

Nitrous oxide produced robust subjective effects on the altered-states questionnaire: all 11 dimensions showed significant increases versus the pre-nitrous-oxide baseline. Reported t-statistics and FDR-corrected p-values included, for example, disembodiment t(12) = 5.302, pFDR = 0.002; blissful state t(12) = 3.692, pFDR = 0.011; and experience of unity t(12) = 3.315, pFDR = 0.013. Among dimensions, disembodiment showed the largest change, consistent with nitrous oxide's dissociative character. At the network level, nitrous oxide increased between-network connectivity and reduced within-network connectivity. Significant between-network increases included visual–salience (pFDR = 0.007), dorsal attention–frontoparietal (pFDR = 0.025), sensorimotor–language (pFDR = 0.025), dorsal attention–language (pFDR = 0.025), salience–default mode (pFDR = 0.025), and dorsal attention–default mode (pFDR = 0.025). Decreases in within-network connectivity were observed in the salience and language networks (both pFDR = 0.025). Ketamine and LSD showed broadly similar network-level patterns. The ketamine analysis enhanced between-network connectivity including frontoparietal–default mode (the extracted text does not clearly report the associated p-value). LSD increased multiple between-network connections with reported pFDR values such as visual–language (pFDR = 0.003), dorsal attention–language (pFDR = 0.020), language–default mode (pFDR = 0.020), visual–default mode (pFDR = 0.025), dorsal attention–default mode (pFDR = 0.025), salience–default mode (pFDR = 0.035), sensorimotor–default mode (pFDR = 0.042), and frontoparietal–default mode (pFDR = 0.042). LSD decreased within-network connectivity in the sensorimotor and dorsal attention networks (both pFDR = 0.030). Direct comparisons among the three psychedelic drugs showed some uncorrected differences (p < 0.05) but none remained significant after FDR correction. ROI-level analyses identified four functional connectivity cluster pairs consistently affected by all three psychedelics: right lateral parietal/TPJ—left intraparietal sulcus (IPS), right lateral parietal—right IPS, precuneus—left IPS, and right lateral parietal—left anterior insula. These changes primarily bridged the default mode and dorsal attention networks, with the right lateral parietal—left anterior insula pair spanning default mode and salience networks. To test whether these patterns were specific to psychedelic states, the investigators analysed propofol sedation: propofol did not show the general decrease in within-network connectivity characteristic of the psychedelics and only one cluster pair overlapped with the psychedelic effects (right lateral parietal—left anterior insula). After excluding connectivity changes shared with propofol, three common psychedelic-altered pairs remained: right TPJ/lateral parietal to bilateral IPS and precuneus to left IPS. Seed-based TPJ maps corroborated the ROI findings: across the three psychedelic drugs, TPJ connectivity increases overlapped in the bilateral IPS. In contrast, the TPJ seed map for propofol was located in occipital cortex and did not overlap the psychedelic pattern. Finally, in the nitrous oxide dataset, changes in TPJ-to-right IPS functional connectivity correlated with subjective intensity on five altered-states subscales: disembodiment (pFDR = 0.018), impaired control and cognition (pFDR = 0.018), anxiety (pFDR = 0.018), changed meaning of percepts (pFDR = 0.019), and experience of unity (pFDR = 0.046). Sensitivity analyses using an FD threshold of 0.4 produced sparser results but were broadly similar to those obtained with FD = 0.9 mm.

The investigators interpret their findings as evidence that both non-classical (nitrous oxide, ketamine) and classical (LSD) psychedelics produce a common large-scale network signature: reduced within-network functional connectivity together with increased between-network connectivity. They highlight the repeated involvement of posterior cortical nodes—particularly the right temporoparietal junction (TPJ), intraparietal sulcus (IPS) and precuneus—which reside in the authors' described ‘‘posterior cortical hot zone’’ and have been proposed to support the contents of conscious experience. TPJ, in particular, was consistently implicated by both ROI-to-ROI and seed-based analyses and is noted for its roles in multisensory integration and body ownership; modulation of TPJ connectivity is proposed as a plausible contributor to psychedelic phenomenology. A key comparative element of the study was the propofol control: propofol produced opposite network changes (enhanced within-network and reduced between-network connectivity) and overlapped with psychedelics only in an anterior insula connection, which the investigators therefore regarded as a nonspecific effect of pharmacological state change. The authors suggest that the shared network reconfiguration across chemically diverse psychedelics implies that the psychedelic experience may be better characterised at the level of network dynamics than by any single molecular target such as the 5-HT2 receptor. They further note potential relevance for therapeutic mechanisms, citing nitrous oxide, ketamine, and classical psychedelics' antidepressant effects and proposing that common network correlates might inform mechanistic understanding of therapeutic benefits. The authors acknowledge several limitations. Heterogeneity in acquisition and protocols across datasets and institutions—plus the use of publicly preprocessed LSD data with a different pipeline—limits comparability. The spatial resolution of 3T fMRI constrains inference about subcortical structures. Only the nitrous oxide dataset included prospective altered-states assessments, so subjective–neural comparisons across drugs are limited. Molecular mechanisms differ across drugs and participants varied in age and dosing protocols, restricting claims about a single common mechanism. The authors also note that while they observed shared TPJ–IPS effects, they did not statistically demonstrate precise identity of network reconfigurations across drugs and recommend further studies, including other psychedelics such as psilocybin and DMT. Despite these caveats, they conclude that this is the first fMRI characterisation of psychedelic-dose nitrous oxide and that the work identifies posterior cortical network alterations that are common to both classical and non-classical psychedelics.