Connectome-harmonic decomposition of human brain activity reveals dynamical repertoire re-organisation under LSD

This randomized, single-blind, placebo-controlled, within-subjects, crossover study (n=12) examined the combined effects of LSD (75 μg) and listening to music on dynamical changes of brain states using a novel connectome-specific harmonic decomposition method. Beyond expanding the repertoire of dynamic brain states, LSD induced a uniquely stable pattern of neural activity with more complex brain dynamics compared to the placebo condition.

Abstract

“Introduction: Recent studies have started to elucidate the effects of lysergic acid diethylamide (LSD) on the human brain but the underlying dynamics are not yet fully understood.

Methods: Here we used ’connectome-harmonic decomposition’, a novel method to investigate the dynamical changes in brain states. We found that LSD alters the energy and the power of individual harmonic brain states in a frequency-selective manner.

Results: Remarkably, this leads to an expansion of the repertoire of active brain states, suggestive of a general re-organization of brain dynamics given the non-random increase in co-activation across frequencies.

Discussion: Interestingly, the frequency distribution of the active repertoire of brain states under LSD closely follows power-laws indicating a re-organization of the dynamics at the edge of criticality. Beyond the present findings, these methods open up for a better understanding of the complex brain dynamics in health and disease.”

Authors: Selen Atasoy, Leor Roseman, Mendel Kaelen, Morten L. Kringelbach, Gustavo Deco & Robin L. Carhart-Harris

Notes

Selen Atasoy also provides a great explanation of these results in this video:

• Singing the mind: Connectome-harmonic signatures of psychedelics, sleep and meditation

Summary

LSD was discovered in 1943 and was used to explore and understand consciousness, psychopathology and mental illness for 15 years. Now, with significant advances in functional neuroimaging, psychology is beginning to embrace and exploit this powerful scientific tool.

Previous fMRI and MEG work has reported increased visual cortex blood flow, increased whole-brain functional integration, and decreased oscillatory power across a broad frequency range under LSD. This work describes a novel method to decompose resting-state fMRI data under LSD and placebo into a set of independent, frequency-specific brain states.

Our technique capitalizes on the identification of the brain states as harmonic modes of the brain’s structural connectivity and the decomposition of fMRI-measured cortical activity patterns into these harmonic brain states. This approach provides a set of independent and fully synchronous brain states with increasing spatial frequency. The activation of harmonic brain states composes the complex spatio-temporal dynamics of cortical activity, and the evaluation of fundamental properties of these harmonic brain states enables novel tools for neuroscience.

Figure 1 shows the workflow of using T1 and DTI data to reconstruct the cortical surface between gray and white matter, and to estimate the power of activation of each of these brain states for each time instance.

Here we evaluate the power and energy of different frequency connectome harmonics to characterize brain activity in the LSD-induced psychedelic state, and further investigate whether the changes in the activation of individual harmonics lead to any variations in the dynamical repertoire of these harmonics brain states.

Results

Here we recorded fMRI data from 12 subjects in 6 different conditions: LSD, placebo, LSD and PCB while listening to music, LSD and PCB after the music session. The results show that music can evoke emotions, which are emphasized by the effect of psychedelics.

We decomposed fMRI recordings of 12 subjects in 6 different conditions into the activity of frequency-specific brain states (cortical patterns). We investigated two fundamental properties of these harmonic brain states: power of activation and energy of each of these harmonic brain states.

We measured the total power and total energy of all brain states in 6 conditions: LSD, PCB, LSD with-music, PCB with-music, LSD after-music, PCB after-music. The results revealed that LSD increases these measures.

We estimated the probability distribution of energy values for 6 conditions and found that the probability distributions were significantly different between LSD and PCB conditions and that there was no significant difference between LSD and placebo conditions. Music increased the probability of reaching the characteristic energy state in both LSD and PCB conditions, and in the placebo condition. This effect was also found in the LSD after-music condition. LSD increased the sensitivity of cortical dynamics to the effect of music, and also increased the speed at which cortical dynamics changed after the offset of music under the influence of LSD.

The energy increase of brain activity under LSD may be due to more brain states contributing or the same brain states contributing with more power.

LSD extends the repertoire of active brain states, and psilocybin-induced psychedelic state shows greater diversity in functional connectivity motives accompanied by increased variance in temporal oscillations. The probability distribution of conditions under LSD and placebo conditions shows a clear decrease for small magnitude activations.

Figure 2 shows the changes in energy of brain states under LSD, and the probability distribution of total energy values for all 6 conditions. Connectome harmonics quantized into 15 levels of wavenumbers k for conditions LSD vs. PCB, LSD with-music vs. PCB with-music, LSD after-music vs. PCB after-music. Stars indicate significant differences (p 0.01, Monte-Carlo simulations after Bonferroni correction).

We investigated which brain states demonstrated increased activity under the effect of LSD by discretising the connectome-harmonic spectrum into 15 levels of wavenumbers k in the log-space and analysing the energy changes within each of these parts of the harmonic spectrum for each condition and each subject separately.

For all 3 conditions, LSD increased the energy of all quantized levels of wavenumbers (p 0.01, Monte-Carlo simulations after Bonferroni correction, Fig. 2e), and the mean and width of the normal distribution increased in all LSD conditions, although slightly less for LSD with-music condition. Significant decreases in energy were found for all low frequency brain states, and for both increased energy of high frequencies and decreased energy of low frequencies. Figure 2i shows that LSD increases the energy of brain states with larger wavenumbers and expands the repertoire of active brain states by significantly increasing the activity of high frequency brain states.

We investigated whether LSD-induced expansion of the repertoire of active brain states occurred in a structured or random fashion by examining the degree of co-activation of different frequency brain states. Under the influence of LSD, we observed a significant decrease in cross-frequency correlations within the low-frequency brain states, and a significant increase in cross-frequency correlations within the higher frequency brain states, indicating the influence of music on the co-activation of brain states within this frequency range.

LSD significantly increased cross-frequency correlations across the complete spectrum of brain states, but the effect of LSD faded over time, so the increase in cross-frequency correlations was not found in the sequential comparison of PCB scans. Under the influence of LSD, music decreased cross-frequency correlations in the low to mid range frequencies, but increased cross-frequency correlations throughout the connectome-harmonic spectrum. This result indicates that the fading effect of LSD and the influence of music both contribute to observed changes in cross-frequency correlations.

Figure 3 shows the cross-frequency correlations between different conditions, and the significant differences between condition LSD, PCB, LSD with-music, PCB with-music, LSD after-music, PCB after-music are marked with stars.

LSD causes a re-organization of brain dynamics rather than a total randomization. This type of non-random expansion of the state repertoire naturally occurs in dynamical systems when they approach criticality.

Power laws and whole-brain criticality are two concepts that explain how the brain works. Criticality is a delicate balance between two extreme tendencies, order and disorder, where complexity increases and certain functional advantages may emerge.

The power-law distributions are a key characteristic of critical dynamics in large scale brain networks, and are found consistently across wakefulness, deep-sleep, REM sleep and anaesthetics induced loss of consciousness, but are slightly deteriorated in wakefulness, tend to diminish in cognitive load and recover during sleep.

LSD’s psychoactive effects are mediated by serotonin 2A receptor (5-HT2AR) agonism, which induces cortical excitation. Increased cortical excitation via increased 5-HT2AR signalling is a plausible mechanism by which LSD may tune cortical dynamics towards criticality.

Based on LSD’s pharmacology, we investigated LSD’s effects on the brain in the context of criticality. We observed that the power-law distributions of connectome harmonics followed a slight cut-off, indicative of the slight deviation to the subcritical regime.

We measured the goodness of fit of power-laws for three different conditions: before music, with music and after music. The results suggest that brain dynamics reside close to criticality in both conditions, with LSD tuning them further towards criticality. The power-law exponent of maximum and mean-power distribution decreased significantly in all conditions with LSD compared to placebo, and was only significantly decreased in the first scan; LSD vs. placebo condition. This indicates increased power of high frequency connectome harmonics.

Figure 4 shows that the power laws in connectome harmonic decomposition are linear with respect to wavenumber for LSD vs. PCB, LSD with-music vs. PCB with-music and LSD after-music vs. PCB after-music.

LSD-induced brain activity changes correlate with subjective experience, and five key facets of the LSD experience were investigated: complex imagery, simple hallucinations, emotional arousal, positive mood, and ego-dissolution.

We explored the correlations between the activation of different brain states and subjective experiences by measuring the energy differences between LSD and PCB conditions and then evaluating the subjective ratings.

In the low frequency range k 1 – 200 corresponding to [0 – 0.01%] of connectome-harmonic spectrum, we observed a decrease in the mean energy as well as in the energy fluctuations under LSD at an individual subject level. This decrease was significantly correlated with the subjective ratings of ego-dissolution and emotional arousal.

Correlations between energy changes of connectome harmonics and subjective experiences were demonstrated in Figure 5, including significant correlations between difference in mean energy of connectome harmonics and ego dissolution and emotional arousal, as well as significant correlations between difference in energy fluctuations of connectome harmonics and positive mood.

The energy change within the low frequency range did not significantly correlate with ratings of positive mood under LSD, but a broader range of spatial frequencies did show significant correlation with the intensity of positive mood induced by LSD.

Although we observed high correlations between energy changes of individual harmonics and subjective ratings, we did not find any significant correlations when considering the whole connectome-harmonic spectrum.

Correlations between subjective ratings and connectivity of resting state networks were examined using multiple correlation analysis. The results showed that the connectivity of different groups of networks mutually correlated to the intensity of subjective experience.

We identified the following networks and defined the following subgroups: fronto-parietal network, visual, visual-AUD, PAR-pOP, pOP-rFP, DMN-SAL, DMN-lFP, DMN-rFP, DMN-pOP, SAL-lFP, SAL-rFP, lFP-visual, rFP-visual.

Connectome harmonics are patterns of synchronous activity emerging on the cortex for different frequency oscillations. These patterns can be decomposed into frequency-specific functional connectivity patterns.

The direct application of multiple correlations between subjective ratings and connectivity changes of RSNs showed differences of correlations, but the indirect application revealed significant correlations between the intensity of certain subjective experiences and groups of RSNs.

We observed significant correlations between the connectivity changes of the visual and sensory (visual-auditory) networks and ratings of simple hallucinations and complex imagery. Moreover, the coupled connectivity of the visual networks with the left and right fronto-parietal networks correlated with ratings of both complex imagery and simple hallucinations.

We found significant correlations between the DMN connectivity and the intensity of emotional arousal, positive mood, and ego dissolution. The abnormal DMN-SAL functional connectivity has previously been reported under LSD10 and psilocybin54. We found that the coupled connectivity changes of the SAL with lFP alone or FPN (lFP and rFP together) significantly correlated with the ratings of emotional arousal, positive mood and ego-dissolution, whereas the correlation of the SAL with rPF alone was less significant. The increased functional connectivity between the DMN and other networks, particularly the DMN-rFP and DMN-pOP, suggested a link between the increased between-RSN functional coupling and emotional arousal under the effect of LSD.

Discussion

Here, we used a novel connectome-specific harmonic decomposition method to study LSD-induced changes in brain activity.

The estimation of harmonic brain states relies solely on the structural connectivity (human connectome) and thus is independent from the fMRI data itself. The connectome harmonics provide fully synchronous (spatial) patterns of activation, which can be used to describe the brain dynamics in different conditions.

LSD alters the energy and power of individual harmonic brain states in a frequency-selective manner and enriches the connectome-harmonic repertoire. Moreover, the co-activation patterns of these brain states are highly correlated over time.

The increased diversity of brain states under LSD is thought to underlie an expanded capacity for information encoding and an enhanced efficiency of processing. This is accompanied by a tuning towards criticality. Our results suggest that brain dynamics in both LSD and placebo conditions reside close to criticality, with slight deviations to the subcritical regime under placebo, while the induction of LSD tunes brain dynamics further towards criticality.

The presented method goes beyond conventional fMRI analyses to measure changes in brain activity under psychedelics, by revealing how fundamental properties of harmonic brain states relate to important principles of dynamical systems, such as whole-brain criticality.

The LSD-induced shift towards criticality may be a mechanism underlying increased sensitivity to the context under LSD and psychedelics more generally, and may therefore represent the beginnings of a mechanistic explanation of these principles, which would have profound implications for psychology.

Studies have shown that deviations from criticality may be symptomatic or even causative of certain psychiatric disorders. LSD may be able to restore the critical balance between ordered and disordered states in the brain.

Brain dynamics at the edge of criticality have been hypothesized to constitute the neural basis of creativity, where an optimal balance between stability (cognitive control) and flexibility (spontaneous thought) may enable the generation of novel and potentially useful ideas.

Studies exploring the neural basis of jazz improvisation reveal that the number of musical notes played during improvisation is significantly higher compared with memorized play of the same piece, hence leading to an increase of novel information. Likewise, brain dynamics at the edge of criticality enable maximally novel dynamics.

This interpretation is supported by previous studies that have associated psycho-pathology with deviations of brain dynamics from criticality, as well as by the shared genetic roots of schizophrenia, bipolar disorder, psychosis and creativity.

We found significant correlations between the energy changes of different connectome harmonics and the connectivity changes of the RSNs and the intensity of different subjective experiences, supporting the idea that abnormal DMN connectivity may involve or even be mediated by altered emotional processing.

Increased between-RSN functional connectivity under LSD has been found to correlate with emotional arousal, positive mood and ego dissolution, suggesting a potential link between these effects and the psychedelic state.

In patients with schizophrenia, increased connectivity between the lFP and temporal and parietal regions was also found, and this connectivity was correlated with ratings of emotional arousal, positive mood and ego-dissolution, as well as with ratings of simple hallucinations.

We have applied a new methodology to study the LSD state, which reveals a shift in brain dynamics towards whole-brain criticality. This method opens-up an opportunity to explore the neural signatures of other psychological traits and states.

This study was approved by the UK National Health Service research ethics committee, West-London, and all participants gave informed consent.

Participants attended two sessions of scanning days (LSD and placebo) 14 days apart, and were given 2 non-music fMRI BOLD runs and 1 music fMRI BOLD run within each session. The scans were completed 135 min postinfusion.

Methods

20 healthy subjects were scanned with fMRI in 6 different conditions: LSD, placebo, LSD and PCB while listening to music, and LSD and PCB after music session. 12 subjects were used for this analysis.

Participants were asked to rate their LSD experience on a scale of 1 to 5 (complex imagery, simple hallucinations, emotional arousal, positive mood, ego-dissolution).

We developed a new technique to decompose spatio-temporal recordings of brain activity into temporal evolution of brain states. These brain states are estimated as the harmonics of macro-scale structural connectivity of the human brain.

Recently, it has been shown that the harmonic modes of structural connectivity of the human brain predict patterns of correlated neural activity. These patterns, called connectome harmonics, have a wavenumber that is proportional to the spatial frequency of the particular structural connectivity of the human brain.

We used data from the Human Connectome Project, WU-Minn Consortium, to estimate connectome harmonics. All data were preprocessed according to the HCP protocol and no additional preprocessing was performed.

To estimate the connectome-harmonic basis, the cortical surfaces separating the white and grey matter were reconstructed from T1-weighted MRI data and represented as a graph with 10,242 nodes for each hemisphere. The white matter cortico-cortical and thalamo-cortical fibres were extracted using a deterministic tractography algorithm.

We formed a graph representation of the human connectome using nodes and edges, and used an undirected, unweighted graph to represent the adjacency (connectivity) matrix of each subject.

We average the adjacency matrices of 10 subjects to get an average structural connectivity of all subjects. We then compute the connectome Laplacian.

We calculate the connectome harmonics of a subject by solving the eigenvalue problem for the adjacency matrix and degree matrix.

To represent fMRI data in cortical surface coordinates of connectome harmonics, fMRI scans were projected onto the cortical coordinates and the spatial cortical activity pattern was decomposed into the activity of connectome harmonics.

The temporal activity of each connectome harmonic was estimated by projecting the fMRI data onto that particular harmonic.

The power of each connectome harmonic is computed as the strength of activation of a particular connectome harmonic, and the energy of a brain state is estimated by combining the strength of activation of a particular connectome harmonic with its own intrinsic energy.

The total energy and power of a brain state are computed by summing over all time points and dividing by the orthonormal connectome harmonics. The upper bound of the power is determined by the cortical activity pattern, whereas the lower bound is 0.

Power-law analysis was performed on the connectome harmonics and the relations between maximum power, average power as well as power fluctuations were evaluated. The line fitting and the estimation of the critical exponent were performed in MATLAB.

Estimation of RSN connectivity was performed using an independent sample of participants as part of the Human Connectome Project.

Multiple correlations between RSN connectivity and subjective ratings were estimated using multiple correlation coefficient53. These correlations were evaluated using the connectome-harmonic correlates of the 24 dimensional vectors of FC changes and subjective ratings.

Acknowledgements

The authors thank Andrea Insabato for valuable discussions on significance analysis, and the Beckley foundation, the ERC Advanced Grant, and the Spanish Research Project PSI2016-75688-P (AEI/FEDER), and the European Union’s Horizon 2020 research and innovation programme.

Additional Information

This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, provided you give appropriate credit to the original author(s) and the source.