Psilocybin modulation of dynamic functional connectivity is associated with plasma psilocin and subjective effects

This preprint single-blind study (n=15) examined the association between resting-state dynamic functional connectivity (dFC) and plasma psilocin level (PPL) and subjective drug intensity (SDI) before and right after psilocybin intake (17mg/70kg) in healthy volunteers using fMRI. It was found that the acute effects of psilocybin may stem from drug-level associated decreases in the occurrence and duration of lateral and medial frontoparietal connectivity motifs in exchange for increases in a uniform connectivity structure.

Abstract

“Background: Psilocin, the neuroactive metabolite of psilocybin, is a serotonergic psychedelic that induces an acute altered state of consciousness, evokes lasting changes in mood and personality in healthy individuals, and has potential as an antidepressant treatment. Examining the acute effects of psilocin on resting-state dynamic functional connectivity implicates network-level connectivity motifs that may underlie acute and lasting behavioural and clinical effects.

Aim: Evaluate the association between resting-state dynamic functional connectivity (dFC) characteristics and plasma psilocin level (PPL) and subjective drug intensity (SDI) before and right after intake of a psychedelic dose of psilocybin in healthy humans.

Methods: Fifteen healthy individuals completed the study. Before and at multiple time points after psilocybin intake, we acquired 10-minute resting-state blood-oxygen-level-dependent functional magnetic resonance imaging scans. Leading Eigenvector Dynamics Analysis (LEiDA) and diametrical clustering were applied to estimate discrete, sequentially active brain states. We evaluated associations between the fractional occurrence of brain states during a scan session and PPL and SDI using linear mixed-effects models. We examined associations between brain state dwell time and PPL and SDI using frailty Cox proportional hazards survival analysis.

Results: Fractional occurrences for two brain states characterized by lateral frontoparietal and medial fronto-parietal-cingulate coherence were statistically significantly negatively associated with PPL and SDI. Dwell time for these brain states was negatively associated with SDI and, to a lesser extent, PPL. Conversely, fractional occurrence and dwell time of a fully connected brain state was positively associated with PPL and SDI.

Conclusion: Our findings suggest that the acute perceptual psychedelic effects induced by psilocybin may stem from drug-level associated decreases in the occurrence and duration of lateral and medial frontoparietal connectivity motifs in exchange for increases in a uniform connectivity structure. We apply and argue for a modified approach to modeling eigenvectors produced by LEiDA that more fully acknowledges their underlying structure. Together these findings contribute to a more comprehensive neurobiological framework underlying acute effects of serotonergic psychedelics.”

Authors: Anders S. Olsen, Anders Lykkebo-Valloee, Brice Ozenne, Martin K. Madsen, Dea S. Stenbaek, Sophia Armand, Morten Morup, Melanie Ganz, Gitte M. Knudsen & Patrick M. Fisher

Summary

A study was conducted at the Copenhagen University Hospital Rigshospitalet, Copenhagen, Denmark, to investigate the effects of smoking on the brain.

ORCID IDs

The neuroactive metabolite of psilocybin, psilocin, induces an acute altered state of consciousness, evokes lasting changes in mood and personality in healthy individuals, and has potential as an antidepressant treatment. We acquired 10-minute resting-state blood-oxygen-level-dependent functional magnetic resonance imaging scans of 15 healthy individuals and evaluated associations between resting-state dynamic functional connectivity characteristics and plasma psilocybin level and subjective drug intensity. Results suggest that the acute perceptual psychedelic effects induced by psilocybin may stem from drug-level associated decreases in the occurrence and duration of lateral and medial frontoparietal connectivity motifs in exchange for increases in a uniform connectivity structure.

Psilocybin, a psychedelic compound, induces an altered state of consciousness and positive effects on mood, well-being, and personality. These effects may inform future drug development programs and predict potential adverse drug effects. Previous studies have shown that psilocybin affects distributed functional brain connectivity patterns, but most have focused on “static” functional connectivity, which neglects relevant and observable neural dynamics arising from mind-wandering or ephemeral experiences. Dynamic functional connectivity (dFC) is a method for extracting informative, time-varying brain connectivity patterns. It has been used to evaluate psychedelic effects on rs-fMRI, including the effect of psilocybin on dFC.

We examined the effects of psilocybin on functional connectivity in alignment with an assessment of plasma psilocin levels and subjective effects throughout six hours of psilocybin administration. The results suggest that psilocybin effects may be informatively characterized by approaches that group sets of regions in a data-driven manner. LEiDA clusters eigenvectors using Euclidean k-means, but in practice eigenvectors are normalized to have unit length and arbitrary sign, which leads to sub-optimal clustering. We applied LEiDA with diametrical clustering to evaluate acute psilocybin effects on dFC with blood-oxygen-level-dependent (BOLD) rs-fMRI in 15 healthy participants. We determined the association between dFC characteristics and the psychopharmacological effects of psilocybin using plasma psilocin level and dwell time.

We found that subjective drug intensity was coupled to 5-HT2AR occupancy and baseline 5-HT2AR, and that PPL and SDI were associated with discrete brain state dwell time.

The LEiDA pipeline consists of extracting session and region-wise instantaneous BOLD phases, followed by eigenvalue decomposition of the associated phase coherence map, and diametrical clustering to derive discrete brain states.

Results

Seventy-two 10-minute rs-fMRI scan sessions were acquired across 15 healthy participants before and at multiple time points after oral psilocybin administration. The phase coherence map was computed for the 90 cortical and subcortical regions and clustered into k discrete brain states defined by centroids determined with diametrical clustering.

We estimated brain states using LEiDA and diametrical clustering with k=7 and observed a “global” brain state characterized by all centroid elements having the same sign. All other brain states were characterized by coherence loadings in both directions.

We used a linear mixed-effects model to determine the respective associations between brain states and permutation testing and Max-T correction. One brain state was statistically significantly negatively associated with both perpetuity and SDI.

PPL and SDI were negatively related to the average total time the brain occupies this frontoparietal state during a 10-min rs-fMRI scan. We observed a second brain state, frontoparietal state 2, which was negatively associated with both PPL and SDI, and a third brain state, fully connected state, which was positively associated with SDI.

We applied a Cox-proportional hazards frailty model to evaluate the association between PPL and SDI on dwell time for each individual brain state. The higher the PPL and SDI, the less average continuous time was spent in frontoparietal state 1. For kk 8, frontoparietal state 2 was negatively associated with both PPL and SDI, and for kk 4, dwell time of the fully connected state was positively associated with both PPL and SDI.

Table S3 lists centroids across all values of k that show a statistically significant association between PPL or SDI and either FO or dwell time.

The authors used linear mixed-effects models, Cox proportional hazards frailty models, and Bonferroni-Holm applied within-k to analyze the association between brain state fractional occurrence and dwell time with plasma psilocin level and subjective drug intensity.

Figure 4 shows that the frontoparietal state 1 brain state was associated with plasma psilocin level and subjective drug intensity in k=7 participants using linear mixed-effects models for fractional occurrence and frailty Cox proportional hazards models for dwell time.

We identified three brain states based on template centroids. The three states had very similar centroids across the range of k, and frontoparietal state 2 became more associated with PPL and SDI than frontoparietal state 1 at k = 8.

We ran diametrical clustering 1000 times and extracted the two frontoparietal states and the fully connected state by identifying the state most closely matching the relevant template states. Generally, we see high clustering stability regardless of initialization.

We compared the results of diametrical clustering and LEiDA Euclidean k-means for clustering frontoparietal states 1, 2, and the fully connected state. The results show that although there are clear similarities across the brain states paired between the two clustering methods, the magnitudes of these similarities are variable.

We measured the effects of acute psilocybin use on dynamic functional brain connectivity in healthy individuals. We found that the higher the subjective experience intensity and plasma psilocin levels, the lower the fractional occurrence of two discrete frontoparietal-like brain states and the average dwell time of these states. Frontoparietal states 1 and 2 were characterized by phase coherence between areas commonly assigned to a network described as the “frontoparietal” network, and by similarity in the regions with strong “negative” loadings. Despite methodological differences between the studies, the findings suggest that decreased frontoparietal connectivity is a critical neural characteristic of the psilocybin-induced drug experience. This finding implicates a systems-level neural correlate (frontoparietal state prevalence) to the relation perpetuity.

We observed that available psilocin was negatively associated with subjective intensity of the psychedelic experience, and that dwell time was negatively associated with PPL and SDI. However, this effect was statistically significant for only a subset of the evaluated number of brain states. Previous studies have used transition probability matrices to evaluate brain state switching mechanisms. However, the Cox proportional hazards model provides a more complete perspective on brain state dynamics. Previous studies applying LEiDA have reported alterations in a “fully connected” brain state, but here we observed a decrease in fractional occurrence and dwell time of frontoparietal connectivity dynamics, but an increase in fractional occurrence significantly associated with SDI, but not PPL.

Here we have presented the application of diametrical clustering, which acknowledges the antipodal symmetry along both directions of a given eigenvector, and is fundamentally more appropriate clustering method than k- means based on Euclidean distance. When clustering eigenvectors with similar numbers of positive and negative loadings, slight variations can result in sign flips that place otherwise similar eigenvectors in different areas of this region space. Diametrical clustering avoids this limitation by explicitly modeling vectors with unit length and arbitrary sign. We did not address the optimal number of brain states, but explored a range of k, 2 to 20, consistent with previous studies. Opportunities remain for developing methodology surrounding the clustering of dynamic BOLD time series, including using the Watson mixture model and the Bingham distribution to estimate cluster outlines and more objectively estimating how many brain states to include.

The author/funder of this preprint has granted medRxiv a license to display the preprint in medRxiv preprint doi: https://doi.org/10.1101/2021.12.17.21267992. The instantaneous leading eigenvector constructed as part of the LEiDA pipeline explains 58% of the variance. We have previously reported a negative association between static functional connectivity within a priori defined resting-state networks and PPL and SDI using rs-fMRI data. However, the current findings are different because the brain states resolved here are not easily translated to canonical resting-state networks. Together, our studies provide complementary perspectives on the associations between resting-state connectivity and PPL and SDI, including insight into the neurobiological mechanisms underlying lasting behavioral and clinical effects of psilocybin. A plasma psilocin level of 20 g/L, corresponding to 70% neocortex 5-HT2AR occupancy11, results in a more than 50% decrease in the fractional occurrence of frontoparietal state 1 (for k = 7), although two participants showed lower fractional occurrence values at baseline.

Psychedelics can alter functional connectivity dynamics and brain function, and these findings may provide complementary insights into the neural mechanisms underlying psychedelics. We did not have pulse and breathing rate data, and thus we could not directly regress physiological noise from our data. Additionally, we excluded two full scan sessions where motion artifacts were pervasive, but we cannot preclude motion-related effects on our results. We report that acute psilocybin-induced modulation of brain connectivity dynamics is significantly associated with PPL and SDI, and propose a new method for clustering eigenvectors.

Methods

A brief description of experimental procedures is provided here; a full description can be found elsewhere20. The author/funder grants medRxiv a license to display the preprint.

Fifteen healthy participants with no or limited prior experience with psychedelics were recruited for a brain imaging study.

Psilocybin was taken orally in multiples of 3 mg psilocybin capsules, dosed according to body weight, and participants received either psilocybin or a non-psychedelic drug (ketanserin). Functional neuroimaging data were acquired once before and at regular intervals after administration.

Neuroimaging data were acquired on a 3T Siemens Prisma scanner with a 64-channel head coil. BOLD fMRI data were acquired using a T2*-weighted gradient echo-planar imaging sequence.

The author/funder has granted medRxiv a license to display the preprint in medRxiv preprint doi: https://doi.org/10.1101/2021.12.17.21267992; this version posted December 17, 2021. Motion and signal variance artifacts were identified using Artifact detection Tool (ART), and 21,600 rs-fMRI volumes were included in subsequent analyses.

We estimated regional phase-series from BOLD signals by constructing an analytic signal from the BOLD signals and integrating the Hilbert transform. The instantaneous phase is a sawtooth curve representing the locally linear temporal phase angle variation, and encompasses oscillatory information in ss(tt) and (potentially spurious) amplitude information in aa(tt). The instantaneous phase coherence between brain region pairs was described using the symmetric phase coherence map AA for every time point tt.

This Clustering was made available under a CC-BY-NC-ND 4.0 International license by the author/funder. The Dimroth-Scheidegger-Watson distribution model can be used to disentangle clusters by estimating a Watson mixture model, and k-means can be used to update cluster centroid locations according to the squared Pearson correlation similarity measure. Diametrical clustering is a method for clustering data that uses the squared Pearson correlation between pairwise vectors. It is able to group correlated and anticorrelated unit norm vectors into the same cluster. We grouped the 21,600 leading eigenvectors into k clusters, which we denoted “brain states”. We matched specific centroids across different k by selecting a template, and then found the centroid that most closely matched this template in terms of squared Pearson correlation.

To identify associations between brain state dynamics and PPL and SDI, we calculated the fractional occurrence for each brain state, modeled the association using a random intercept linear mixed-effects model, and performed permutation testing with max-T adjustment and 100,000 permutations.

We used survival analysis to model state dwell time, and a Cox proportional hazards model with a frailty element to account for inter-subject variability was used. The estimated hazard ratio was proportional in the covariate level (PPL or SDI), and the associated confidence interval was defined as ee 1.96 ssss ( ) .

This preprint is made available under a CC-BY-NC-ND 4.0 International license and was not certified by peer review. The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Acknowledgements

We thank the MRI assistants, the biobank manager, the BAFA laboratory, the National Institute of Mental Health, the Glostrup Apotek company, and Sys Stybe Johansen and Kristian Linnet for their help.

Contributors

All authors contributed to data analysis, production of figures, and manuscript writing. ASO led the study, ALV analyzed data, BO supervised statistical methods, MKM designed the experiment, and DSS collected data.

Funding

The authors thank several funders for their support, which did not impact the study in any way.