This pre-print tested the hypothesis that LSD and psilocybin reduce whole-brain hierarchical (top-down) organization. The results of a model that looks at fMRI data confirms this hypothesis.
Abstract
“LSD and psilocybin are serotonergic psychedelic compounds with potential in the treatment of mental health disorders. Past neuroimaging investigations have revealed that both compounds can elicit significant changes to whole-brain functional organization and dynamics. A recent proposal linked past findings into a unified model and hypothesized reduced whole-brain hierarchical organization as a key mechanism underlying the psychedelic state, but this has yet to be directly tested. We applied a non-linear dimensionality reduction technique previously used to map hierarchical connectivity gradients to pharmacological resting-state fMRI data to assess cortical organization in the LSD and psilocybin state.
Results supported our primary hypothesis: The principal gradient of cortical connectivity, describing a hierarchy from unimodal to transmodal cortex, was significantly flattened under both drugs relative to their respective placebo conditions. Between-condition contrasts revealed that this was driven by a reduction of functional differentiation at both hierarchical extremes – default and frontoparietal networks at the upper end, and somatomotor at the lower. Gradient-based connectivity mapping confirmed that this was underpinned by increased unimodal-transmodal crosstalk. In addition, LSDdependent principal gradient changes tracked changes in self-reported ego-dissolution. Results involving the second and third gradient, which respectively represent axes of sensory and executive differentiation, also showed significant alterations across both drugs. These findings provide support for a recent mechanistic model of the psychedelic state relevant to therapeutic applications of psychedelics. More fundamentally, we provide the first evidence that macroscale connectivity gradients are sensitive to a pharmacological manipulation, specifically highlighting an important relationship between cortical organization and serotonergic modulation.“
Authors: Manesh Girn, Leor Roseman, Boris Bernhardt, Jonathan Smallwood, Robin L. Carhart-Harris & R. Nathan Spreng
Summary
LSD and psilocybin can alter whole-brain functional organization and dynamics, and a non-linear dimensionality reduction technique was applied to pharmacological resting-state fMRI data to assess cortical organization in the psychedelic state. The results supported the primary hypothesis that reduced whole-brain hierarchical organization underlies the psychedelic state.
The past decade has seen a resurgence of scientific interest in serotonergic psychedelic compounds such as lysergic acid diethylamide (LSD), psilocybin, and dimethyltryptamine (DMT)/ayahuasca, primarily motivated by suggestive findings from preliminary clinical trials for depression, end-of-life-anxiety, alcoholism, and tobacco addiction. Psychedelic administration increases global functional integration, global functional connectivity, and complexity of brain dynamics, as reflected by increases in regional and population-level entropy/complexity, as well as increases in the repertoire or dynamic range of functional connectivity states.
The recently proposed RElaxed Beliefs Under Psychedelics (REBUS) model unifies past psychological and neural findings with psychedelics into a common theoretical framework based on hierarchical predictive coding and the Free Energy Principle.
Gradient-mapping techniques have emerged as valuable tools to characterize brain functional organization. These techniques have been used to identify a principal component describing a gradient of functional connectivity spanning from unimodal sensorimotor regions to transmodal association regions centered on the default network.
We investigated how serotonergic drugs LSD and psilocybin alter cortical hierarchy by applying gradient-mapping analyses to fMRI datasets collected with two serotonergic psychedelic compounds. We found that the principal gradient contracted, indicating reduced hierarchical differentiation, and that crosstalk between unimodal and transmodal cortices increased.
Resting-state BOLD fMRI data were acquired in 20 subjects for each of LSD and placebo conditions, with conditions spaced 2 weeks apart. Ego-dissolution scores were collected via intra-scanner visual analogue scale ratings, whereas complex imagery scores were derived from the ASC questionnaire.
Resting-state BOLD fMRI data were acquired in 15 subjects for each of psilocybin and placebo conditions, with conditions spaced 2 weeks apart. Ego-dissolution and complex imagery scores were collected via intra-scanner visual analogue scale ratings and the 11-factor altered states of consciousness questionnaire.
Data collection was approved by the National Research Ethics Service committee London-West London and was conducted under a Home Office license.
Both datasets underwent an identical preprocessing protocol, and 4 subjects were excluded from the LSD dataset and 6 subjects were excluded from the psilocybin dataset due to excessive head motion.
Pre-processing and denoising steps were performed on the BOLD resting-state fMRI data for each dataset, including removal of the first three volumes, de-spiking, slice time correction, motion correction, brain extraction, rigid body registration, non-linear registration, scrubbing, spatial smoothing, and band-pass filtering.
Gradient-mapping was performed on cortical surfaces using the BrainSpace toolbox, and a similarity matrix was generated by computing cosine similarity on the transformed z-matrix. This similarity matrix was then used to embed diffusion maps in the surface.
Diffusion map embedding is a non-linear manifold learning technique that estimates low-dimensional sets of embedding components from a high-dimensional similarity matrix. It is controlled by a single parameter, which allows both global and local relationships between data points to influence the estimation. To enable comparisons across subjects, Procrustes rotation was performed on gradient axes to match them across subjects. Surface-based linear models were used to conduct analyses.
Gradient-based connectivity mapping was used to examine whether unimodal regions preferentially exhibited increased FC with transmodal regions as opposed to non-specific global increases. T-tests were applied to evaluate drug vs. placebo differences at each bin for each of unimodal-specific and transmodal-specific FC.
We computed seedmaps for each gradient percentile bin and compared across drug and placebo conditions to examine whole-brain FC.
We used gradient-mapping techniques to characterize differences in macroscale cortical gradients in each of LSD and psilocybin states relative to their respective placebo states. The algorithm revealed 98 mutually orthogonal gradient components per subject.
Principal gradient histograms indicated a contraction on both sides of the gradient in the drug conditions relative to the respective placebo conditions for both datasets.
We found that both LSD and psilocybin reduced hierarchical differentiation along the unimodal-transmodal axis, and that both unimodal-proximal regions and transmodal-proximal regions became less differentiated in each drug condition relative to placebo.
We quantitatively assessed between-condition differences in gradient score values at the vertex- and network-wise levels. We found that both drugs affected the principal gradient similarly. LSD and psilocybin affect the connectivity of several brain regions, including the precuneus, premotor cortex, superior and inferior frontal gyrus, superior parietal lobule, and Wernicke’s area. These results indicate a reduction in differentiation along the hierarchical axis of cortical connectivity.
We applied a gradient-based connectivity mapping approach to evaluate whether the observed changes in gradient values were specifically consistent with increased cross-talk between unimodal and transmodal cortices, as hypothesized by the REBUS model.
We found that lower percentile bins of unimodal-specific FC were significantly reduced with LSD compared to placebo, and that higher percentile bins were significantly increased with LSD compared to placebo.
We computed whole-brain seedmaps based on each gradient bin ROI and compared across conditions at both the vertex-wise and network-wise level. Results indicated that lower percentile bins exhibit significantly reduced FC with somatosensory and visual networks and increased connectivity with the frontoparietal control network.
We found that the LSD state contracted the macroscale functional hierarchy, which was consistent with greater unimodal-transmodal crosstalk. The second and third gradients of macroscale functional organization explained additional variance in the LSD, psilocybin, and placebo conditions.
LSD-induced increases in lateral and ventral visual cortex and the middle and superior frontal gyrus bilaterally, while decreases were found in somatomotor and auditory cortex. Psilocybin-induced changes were similar, but qualitatively different, suggesting differential effects on sensory differentiation across these drugs.
LSD and psilocybin increase the somatomotor vortex, lateral and medial visual cortex, and the retrosplenial/posterior cingulate cortex, while decreasing the posterior middle temporal gyrus, inferior frontal gyrus, left premotor cortex, and right supramarginal gyrus.
Gradient manifolds were created to visualize the relationship between the three gradients examined in this study and how they differ across drug and placebo conditions. Psilocybin showed a greater contraction of the third gradient than LSD.
To ascertain the subjective relevance of changes in macroscale gradients during the psychedelic state, we assessed relationships between drug-placebo gradient differences scores and two self-report measures of ego dissolution and complex imagery. We found that greater hierarchical differentiation in the left medial prefrontal cortex was associated with increased ego dissolution.
To investigate psychedelic-induced changes in cortical functional organization, researchers characterized macroscale cortical gradients after LSD and psilocybin administration. They found that this gradient exhibited a significant contraction in both LSD and psilocybin states, reflective of less differentiated hierarchical organization. The topographical changes across the LSD and psilocybin datasets were similar, with the exception of reduced differentiation of visual and somatomotor cortex along the second gradient. Additionally, there was a significant association between LSD-related changes in the principal gradient and self-reported ego-dissolution.
Past neuroimaging investigations with LSD and psilocybin have revealed that they elicit a complex set of changes to both static and dynamic FC, as well as to the entropy/complexity of regional timeseries. This work has often attempted to manage the complexity of the data by focusing on specific large-scale network interactions or on general trends. However, this approach obscures important nuances in the structure of the data and makes it difficult to compare findings across datasets. This paper uses a cortical gradient approach to study the effects of serotonergic psychedelics by identifying specific alterations in cortical information processing.
LSD-dependent changes in the unimodal-transmodal hierarchical gradient were symmetrical, with decreases occurring in regions at the highest end of the hierarchy. This was underpinned by hierarchically specific changes in FC.
In the psychedelic state, the brain becomes more integrative, with unimodal regions becoming less modular and less differentiated from the globally distributed network of transmodal hubs. This is consistent with a reduction in the number of intervening processing steps between low-level sensorimotor cortex and high-level association cortex.
Psychedelics can blur the distinction between internal and external/subject-object processing, and increase the influence of internal mentation on perceptual processing. The phenomenon of ‘ego-dissolution’ is a central component of the psychedelic experience and is associated with so-called ‘mystical-type experiences’ which are core predictors of the therapeutic efficacy of psychedelic treatment approaches. We observed a significant positive association between LSD principal gradient scores and ego-dissolution within left medial prefrontal cortex/anterior cingulate cortex.
We examined changes in the second and third gradient of cortical connectivity, and found that the second gradient was reduced with LSD, whereas the third gradient was increased with psilocybin. This could be related to certain cognitive impairments present while under the influence of these drugs.
To our knowledge, this is the first investigation which has shown that serotonergic manipulations can alter cortical gradients and that these changes can be paralleled by specific changes in conscious experience.
Although there is consistent evidence suggesting the safety of psychedelics in controlled research settings, large datasets are currently difficult to collect due to hurdles pertaining to funding and ethics board approval.
The present study reveals that the whole-brain effects of LSD and psilocybin can be represented as a contraction of the brain’s macroscale functional hierarchy, and provides evidence of reductions in sensory and executive region differentiation.
This study received funding from the National Sciences and Engineering Research Council of Canada, the Canadian Institutes of Health Research, and the National Sciences and Engineering Research Council of Canada.