Thread by @singletonion

New preprint!
biorxiv.org/content/10.110…
Excited to share my first paper with @amykooz and others.
@loopyluppi @RCarhartHarris @josecruzat @LeorRoseman Gustavo Deco, Morten Kringelbach, and @estamatakis
@CompPhd @WeillCornell @ccig_cambridge @Imperial_PRG

Here’s what we did: 🧵

Some background
The RElaxed Beliefs Under pSychedelics (REBUS) model by @RCarhartHarris & Friston proposes that increased neural entropy observed under psychedelics (psychs) relaxes the brain’s prior beliefs & their influence over incoming info. These priors shape our perception.

This may explain how psychs treat various mental illnesses; their common cause being pathologically overweighted priors. “I don’t deserve love”, for ex. This may also explain various forms of visual/auditory distortions under psychs. Belief that walls don’t move has less weight.

Dysregulation of these priors is thought to open up / flatten the brain’s variational free-energy landscape. Using recent methods by @EliCornblath @DaniSBassett et al, we tested this hypothesis by mapping the brain’s transition energy landscape using Network Control Theory (NCT).

We clustered fMRI time-series (eyes-closed, resting-state) of 15 subjects who were scanned on two separate days under placebo/LSD and identified recurrent states of co-activation that we would study the temporal dynamics of.

We characterized these “brain-states” based on cosine similarity with Yeo resting-state networks (RSNs). Two states are dominated by bottom-up activation through the somatomotor network (SOM+/-), and two are dominated by top-down activation in the frontoparietal network (FPN+/-).

We found that LSD subjects spent more of their time in the SOM+/- states compared with PL. This was due to LSD subjects dwelling in these states for longer lengths of time, rather than differences in the rate of appearance.

NCT offers a framework to quantify the ease of state transitions in a dynamical system. We calculated the transition energy, i.e. the min amount of E that would needed to induce transitions between states.

Mapping each individual’s LSD and PL landscape, we find that LSD significantly reduces the energy required to transition between every possible combination of states confirming our initial hypothesis.

Our mechanistic hypothesis was that serotonin 2a receptor agonism (LSD’s primary target) is responsible for the flattening. NCT uses a set of ‘control points’, where the energy is injected into the system. We used PET-generated maps of the human serotonin system (Beliveau 2017).

Using 2a weighted inputs to recalculate the placebo landscape resulted in a flattening effect similar to that of LSD. To test the spatial specificity of this observation we randomly shuffled our inputs 10,000x and found the true distribution resulted in lower transition energies.

We also calculated transition energies using distributions of other serotonin receptors and the serotonin transporter as weights and found the 2a receptor was particularly well situated for transition energy reduction.

In addition, we found that the empirically observed reduction of TE by LSD correlated with more dynamic activity at the individual level. Those who saw greater reductions by LSD saw their average state dwell time decrease, and appearance rates increase!

Lasly, one could imagine a state sequence with rapid switching, but in a predictable way [e.g. 1 2 3 4 1 2 3 4 ..]. We calculated the Lempel-Ziv complexity of individual’s brain-state time series and found E reduction by LSD also correlated with increased entropy of the sequence.

By combining fMRI, dMRI, PET and NCT, we provide support for a fundamental theory of the mech. of action of psychedelics by showing that LSD flattens the brain’s energy landscape, allowing for more facile and frequent state transitions and more temporally diverse brain activity.

This work highlights the potential of receptor-informed NCT to provide mechanistic insights into pharmacological modulation of brain dynamics. If you made it this far, thank you!
@amykooz will present the work at #CCNS tomorrow 2:55 EST here -> crowdcast.io/ccns

This was made possible by an awesome team & variety of funding sources. #NSFfunded

For more on the work that inspired our questions:
pharmrev.aspetjournals.org/content/71/3/3…

For more on the work that inspired our methods:
nature.com/articles/s4200…