Dynamical exploration of the repertoire of brain networks at rest is modulated by psilocybin

This experiment measured brain states (fMRI) under psilocybin infusion, and found more general coherence (communication) and lower frontoparietal network activity.

This paper used the same methods (but new participants) as ‘Neural correlates of the psychedelic state as determined by fMRI studies with psilocybin‘ (Carhart-Harris et al, 2012).

Abstract

“Growing evidence from the dynamical analysis of functional neuroimaging data suggests that brain function can be understood as the exploration of a repertoire of metastable connectivity patterns (‘functional brain networks’), which potentially underlie different mental processes. The present study characterizes how the brain’s dynamical exploration of resting-state networks is rapidly modulated by intravenous infusion of psilocybin, a tryptamine psychedelic found in “magic mushrooms”. We employed a data-driven approach to characterize recurrent functional connectivity patterns by focusing on the leading eigenvector of BOLD phase coherence at single-TR resolution. Recurrent BOLD phase-locking patterns (PL states) were assessed and statistically compared pre- and post-infusion of psilocybin in terms of their probability of occurrence and transition profiles. Results were validated using a placebo session. Recurrent BOLD PL states revealed high spatial overlap with canonical resting-state networks. Notably, a PL state forming a frontoparietal subsystem was strongly destabilized after psilocybin injection, with a concomitant increase in the probability of occurrence of another PL state characterized by global BOLD phase coherence. These findings provide evidence of network-specific neuromodulation by psilocybin and represent one of the first attempts at bridging molecular pharmacodynamics and whole-brain network dynamics.“

Authors: Louis-David Lord, Paul Expert, Selen Atasoy, Leor Roseman, Kristina Rapuano, Renaud Lambiotte, David J. Nutt, Gustavo Deco, Robin L. Carhart-Harris, Morten L. Kringelbach & Joana Cabral

Summary

1. Introduction

Brain dynamics can be understood as the exploration of activity configurations over both space and time, and may be related to cognitive function. Although the mechanisms driving the spontaneous formation and dissolution of functional networks remain under debate, recent evidence suggests that transitions between brain states are organized in a hierarchical manner.

Psilocybin, a prodrug of psilocin, changes the exploration of the brain’s dynamical repertoire, leading to broad unconstrained perception and cognition, hyper-associative cognition, and a breakdown in the perception of time, space and selfhood. Psilocybin has potent psychoactive effects due to its agonist activity at the serotonin 2A (5-HT2A) receptor, and it may help understand the functional mechanisms underlying the recently demonstrated therapeutic potential of psilocybin for disorders including depression, anxiety and addiction.

Psilocybin increases the variance of intra-network synchrony over time, and increases the entropy of the motif sequence in a specific network of four brain regions. However, an increase in local complexity does not necessarily imply higher randomness at the larger scale. The serotonergic psychedelics psilocybin and LSD induce a different type of functional integration, characterized by greater global integration, but this has not yet been explored.

We used a data-driven approach to identify recurrent patterns of BOLD phase coherence across subjects and quantified differences in their probability of occurrence and transition profiles before and after the infusion of psilocybin. The results were compared to well-established resting-state networks described in the literature.

3. Participants

Nine participants were scanned, and all had used psilocybin at least once before, but not within 6 weeks of the study.

5. Experimental protocol

All subjects underwent two 12 min eyes-closed resting state fMRI scans over separate sessions, at least 7 days apart, with each session consisting of a psilocybin or placebo injection. The last 5 min of each scan were used for the relevant analyses. Subjects rated the subjective effects of psilocybin as much stronger than placebo, and reported never falling asleep during either scanning session.

6.2. fMRI acquisition

fMRI data were processed using MELODIC, part of FSL, with the default parameters: motion correction, non-brain removal, spatial smoothing, grand-mean intensity normalization, and linear de-trending.

We used the Anatomical Automatic Labeling (AAL) atlas to parcellate the MNI brain into N 1 4 90 cortical and sub-cortical non-cerebellar brain areas, and then band-pass filtered the BOLD signals in each of these 90 brain areas.

7. Dynamic BOLD phase-locking analysis

Fig. 1B shows that at a single time point N 14 90 BOLD phases are placed at the center of gravity of each brain area.

7.1. Leading eigenvector of the phase-locking matrix

We employed a method termed Leading Eigenvector Dynamics Analysis (LEiDA) to characterize the evolution of the dPL matrix over time with reduced dimensionality. The magnitude of each element in V1(t) indicates the strength with which brain areas belong to the communities in which they are placed. We use a convention that ensures that most elements have negative values, and this substantially reduces the dimensionality of the data while still explaining most of the variance of BOLD phase coherence.

8.1. Detection of recurrent BOLD phase-locking patterns

We applied a k-means clustering algorithm to divide the set of 1800 eigenvectors into a predefined number of clusters, with higher k revealing more rare and more fine-grained patterns.

For each partition model considered, k cluster centroids are obtained in the shape of Nx1 vectors, which represent the average vector of each cluster. These centroids can be rendered onto a cortical surface to facilitate visualization and interpretation of recurrent BOLD phase-locking states.

8.2. Probability of occurrence and switching profiles of PL states

We clustered the PL states and calculated the probability of occurrence of each state for each TR in each scanning session. We then compared the probability of occurrence of each PL state before and after psilocybin injection using a permutation-based paired t-test.

8.3. Selection of the optimal number of PL states

In the current study, we used a data mining approach to detect meaningful activity patterns (PL states) following the administration of psilocybin, and left the placebo session as a validation dataset.

We varied the number of clusters between 5 and 10, and examined how the probability of occurrence of each PL state changed after the injection of psilocybin.

We used the same PL states obtained previously for the psilocybin session for the selected optimal k, and ran a single iteration of the k-means algorithm with k 1 4 7, defining as ‘start vectors’ the 7 cluster centroids VC obtained before.

8.4. Comparison with resting-state networks

We transformed the 7 RSNs in 2 mm3 MNI space into 7 vectors with 90 elements each, and computed the bivariate correlation with the centroid vectors VC.

8.5. Global order and metastability

Using a perspective from dynamical systems theory, we investigated how psilocybin affected the order and stability of BOLD signals.

10.1. Repertoire of PL states reveals canonical resting-state networks

We divided the set of leading eigenvectors into seven clusters, and plotted the resulting PL states as a bar plot, a network, and a matrix. The resulting PL states differed most significantly in terms of probability of occurrence after psilocybin injection.

The most probable BOLD phase-locking state corresponds to a state where all BOLD signals are following one main direction, but non-global patterns also shape the phase alignment.

10.2. Consistency of between-condition differences across partition models

We examined a range of PL states obtained with k-means clustering and found that more fine-grained, less frequent and often less symmetric networks were revealed.

For each partition into k PL states, the p-values obtained from between-condition comparisons are shown in terms of probability of occurrence of each of the corresponding k PL states. The Bonferroni correction is applied to prevent false positives.

10.5. Dynamical exploration in a low-dimensional manifold

The occurrence of different functional networks can be interpreted mechanistically in terms of excursions into distinct regions of a low-dimensional manifold of network configurations. Psilocybin injection reduces the number of excursions into the region associated to the frontoparietal PL state.

10.6. Trajectories between BOLD phase-locking states

We found that the probability of transitioning from any given PL state to the frontoparietal PL state 3 was consistently reduced under psilocybin effects, whereas all PL states except one were more likely to transition toward the PL state of global coherence.

11. Discussion

The present study investigated how the dynamical exploration of the brain’s repertoire of discrete functional networks at rest is significantly altered by a psychoactive molecule, here the serotonergic psychedelic psilocybin. The results suggest that the frontoparietal control system is related to the psychoactive effects of the drug.

Psilocybin significantly modified the BOLD spectral content in a distributed Frontoparietal network, and decreased power at low frequencies and the power spectrum scaling exponent. This is consistent with prior human neuroimaging studies of psilocybin’s effects. The Frontoparietal system plays a key role in cognitive control by focusing attention on goal-relevant information through top-down mechanisms. The Frontoparietal system is disengaged under moderate-high doses of psilocybin, which would be a reasonable way of describing the present dosage.

Psilocybin’s psychoactive effects are primarily due to its agonist activity at the serotonin 2A (5-HT2A) receptor. The distribution and density of 5-HT2A receptors in the human brain was recently estimated, and a strong qualitative overlap was found between 5-HT2A receptor expression on the lateral surface of the brain and the regions implicated in the Frontoparietal PL state detected in the present study.

Studies have shown that 5-HT2A receptors are highly expressed in layer 5 pyramidal neurons, and that their stimulation increases the excitability of these neurons. However, the dysregulated excitation that follows their stimulation may destabilize the lateral frontoparietal network, weakening its stability within the brain’s broader repertoire.

The serotonergic psychedelics psilocybin and LSD induce an alternative type of functional integration, characterized by greater global integration and increased metastability of brain states. The results presented here align with prior findings and shed further light on the functional significance of the psychedelic state. The brain spends more time in a globally coherent, highly integrated state, which serves as the dominant attractor, absorbing the majority of inter-state transitions.

Psilocybin increased the metastability of brain dynamics at the whole-brain scale by calculating the standard deviation of the order parameter of the system over time. This result is consistent with prior fMRI investigations of the same psilocybin dataset.

We used a within-subject design to analyze fMRI data, and found that there were significant between-condition differences despite the fact that each scanning condition consisted of only 100 fMRI volumes. Moreover, the changes in the fractional occupancy of specific PL states after the psilocybin infusion was unrelated to participant motion during the scan.

The phase-locked patterns detected automatically with LEiDA are similar to the canonical resting-state networks obtained from ICA analysis on >1000 subjects. Considering larger sample sizes and higher spatial resolution is likely to provide additional insights on the mechanisms underlying the formation of these BOLD phase-locking patterns. We note that the LEiDA approach allows for the analysis of additional temporal properties of BOLD phase-locking patterns, but the temporal resolution used in the current dataset was too slow to capture more rapid dynamics.

Psilocybin, a serotonergic psychedelic, modulates the brain’s dynamical repertoire by selectively destabilizing a brain functional network (i.e., frontoparietal control system) and promoting transitions towards a globally coherent PL state. This could have exciting implications for the design of novel therapeutics for neuropsychiatric disorders.

The present study used a data-driven dynamical functional connectivity analysis to investigate how psilocybin changes the brain’s exploration of functional network states under task-free conditions. It found that the psychedelic state is characterized by a different type of brain integration.