Dynamic coupling of whole-brain neuronal and neurotransmitter systems

This computational paper (2020) describes a model that predicts whole-brain activity in light of the functional coupling between neuroanatomical and neuromodulatory systems and applies this model to demonstrate how the effects of psilocybin on the brain arise out of the mutual interaction between serotonergic (5-HT2A) receptor modulation and the anatomy of the raphe nucleus. The results provide evidence for how the integration of dMRI (anatomy), fMRI (functional neuronal activity), and PET (neurotransmitter system) at the whole-brain level is necessary for properly predicting brain dynamics as a result of the mutual coupling between a dual system.

Abstract

Introduction: Remarkable progress has come from whole-brain models linking anatomy and function. Paradoxically, it is not clear how a neuronal dynamical system running in the fixed human anatomical connectome can give rise to the rich changes in the functional repertoire associated with human brain function, which is impossible to explain through long-term plasticity. Neuromodulation evolved to allow for such flexibility by dynamically updating the effectivity of the fixed anatomical connectivity.

Methods: Here, we introduce a theoretical framework modelling the dynamical mutual coupling between the neuronal and neurotransmitter systems.

Results: We demonstrate that this framework is crucial to advance our understanding of whole-brain dynamics by bidirectional coupling of the two systems through combining multimodal neuroimaging data (diffusion magnetic resonance imaging [dMRI], functional magnetic resonance imaging [fMRI], and positron-electron tomography [PET]) to explain the functional effects of specific serotoninergic receptor (5-HT2AR) stimulation with psilocybin in healthy humans.

Discussion: This advance provides an understanding of why psilocybin is showing considerable promise as a therapeutic intervention for neuropsychiatric disorders including depression, anxiety, and addiction. Overall, these insights demonstrate that the whole-brain mutual coupling between the neuronal and the neurotransmission systems is essential for understanding the remarkable flexibility of human brain function despite having to rely on fixed anatomical connectivity.”

Authors: Morten L. Kringelbach,  Josephine Cruzat,  Joana Cabral, Gitte Moos Knudsen, Robin Carhart-Harris, Peter C. Whybrow, Nikos K. Logothetis & Gustavo Deco

Summary

Whole-brain models linking anatomy and function have shown remarkable progress. However, it is not clear how a neuronal dynamical system running in the fixed human anatomical connectome can give rise to the rich changes in the functional repertoire associated with human brain function.

serotonin | PET | psilocybin | neurotransmitter | whole-brain modeling

Human connectomics has revealed how function arises from structure, and how anatomy can give rise to a complex dynamical neuronal system. However, the anatomical connectome alone cannot explain the full palette of brain function, so evolution has found a solution by dynamically modulating the connectome over time.

We used diffusion magnetic resonance imaging, functional magnetic resonance imaging, and positron electron tomography to create a whole-brain model that describes the coupling between neuronal and neurotransmitter systems. This model explained the effects of psychedelics on brain activity.

Significance

In a technical tour de force, we have combined multimodal neuroimaging data to causally explain the functional effects of specific serotoninergic receptor (5-HT2AR) stimulation with psilocybin in healthy humans.

The authors designed the research, performed the research, contributed new reagents/analytic tools, analyzed the data, and wrote the paper. They declare no competing interest.

Psilocybin, a prodrug of psilocin, acts through the serotonin 2A receptor (5-HT2AR) and has been shown to affect resting-state activity on healthy human participants.

We used a standard whole-brain model to simulate the neuronal system and then coupled the neuronal and neurotransmitter systems by simulating the release-and-reuptake dynamics of the serotonin neurotransmitter and the mutual coupling with the neuronal system.

Recent studies have begun to demonstrate the functional neuroanatomy underlying the experience of unconstrained cognition and enhanced mind-wandering reported following psilocybin, which may be a potential treatment for neuropsychiatric disorders.

Results

The mutually coupled neuronal – neurotransmitter whole-brain model fits the empirical neuroimaging BOLD data by describing the time evolution of the ensemble activity of the different neural populations building up the spiking network, coupled to the dynamics of the neurotransmitter concentration level.

The neurotransmitter system is shown in Fig. 1F, and the coupling from the neuronal to the neurotransmitter system is given by inserting the neuronal population firing rate.

We used psilocybin to study the modulation of the serotonin system by analyzing the neuroimaging data from a group of healthy participants being administered psilocybin i.v. The raphe nucleus is the main source of serotonin neurotransmission.

We fitted a whole-brain model using a framework for describing brain states as ensembles of probabilistic “clouds” in a given state space. These clouds contain time-varying pseudo-states that can distinguish between brain states.

We used the leading eigenvector dynamics analysis (LEiDA) method to extract a probabilistic metastable substate space for the empirical psilocybin data. The obtained PMS space contains three different substates, and the probability and lifetime of each substate can be significantly different between the two conditions.

We fitted the whole-brain model to the placebo condition PMS space using only the neuronal system and found that the optimal values were KLD = 0.002 and ME = 0.05.

We used a mutually coupled whole-brain model to predict the changes in brain dynamics under the effects of psilocybin. The optimal fitting (global minimum) of the active condition of the psilocybin data was found at WES = 0.3 and WIS = 0.1.

A mutually coupled whole-brain model allows us to obtain further insights into the underlying dynamics of neurotransmission involved in psilocybin (in this case for serotonin). This model shows that coupling the neuronal and neurotransmitter systems is important.

We used empirical 5-HT2A receptor densities to compare the optimal fit of the receptor binding maps with the results of randomly shuffling the receptor densities, and found that the 5-HT2A receptor plays the main role in the effects of psilocybin.

Discussion

Here, we showed how a dynamically coupled whole-brain model can explain the paradoxical flexibility of human brain function despite the reliance on a fixed anatomical connectome. We used a mutually coupled dynamical system that couples the neuronal and neuromodulator systems at the whole-brain level.

The results show that the interaction between the neuronal and neuromodulator systems at the whole-brain level is fundamental for explaining the empirical data, and that the integration of dMRI, fMRI, and PET at the whole-brain level is necessary for predicting properly brain dynamics.

We changed the distribution of regional receptor densities to investigate the importance of the receptor distribution in explaining the empirical data, and found that the precise anatomical location and density of 5-HT2A was crucial.

Psilocybin has been demonstrated to play a role in rebalancing the human brain in treatment-resistant depression. This is thought to depend on activation of serotonin 2A receptors, thereby initiating a multilevel plasticity.

The mutually coupled whole-brain model sheds important light on our understanding of human brain function. It was initially based on neuronal modeling, but recently it was shown that even better results can be obtained when the dynamics of the neuronal system is influenced by static neuromodulation.

We fit a whole-brain model to the PMS space of the placebo condition and a psilocybin condition using only the neuronal system and then the coupled neuronal – neurotransmission whole-brain model. The most significantly optimal fit is found at WES = 0.3 and WIS = 0.1.

Here, we have demonstrated the importance of having a mutually coupled whole-brain model that involves anatomical connectivity and two dynamical neuromodulation systems.

A framework for modeling human brain function in health and disease is proposed, which combines multimodal neuroimaging from dMRI, fMRI, and PET at the whole-brain level. This framework can be used to make predictions that can be tested causally with electromagnetic stimulation or pharmacological interventions.

Materials and Methods

We use a pipeline to integrate structural and functional connectivity data with neurotransmission data to model the placebo and psilocybin responses.

Structural connectivity, functional dynamics, neurotransmitter receptor density, whole-brain neuronal model, mutually coupled neuronal and neurotransmission whole-brain model, empirical fitting of mutually coupled whole-brain model to neuroimaging data were estimated.

Based on previous whole-brain studies, we used the AAL atlas but considered only 90 cortical and subcortical noncerebellar brain regions. Structural connectivity was integrated using this atlas.

16 healthy right-handed participants underwent a 3-T Siemens Skyra scanner MRI with 62 optimal nonlinear diffusion gradient directions at b = 1,500 s/mm2. Two dMRI datasets were collected: one with anterior-to-posterior phase encoding direction and the other acquired in the opposite direction.

We used the structural connectivity between 90 AAL regions in the dMRI dataset described above to generate structural brain networks for each participant. The structural brain networks were generated using the dMRI data acquired with different phase encoding to optimize signal in difficult regions.

The FSL diffusion toolbox (Fdt) was used to process the dMRI dataset. The probtrackx tool was used to improve the tracking sensitivity of nondominant fiber populations in the human brain.

We sampled 5,000 streamlines per voxel to compute the connectivity probability from a given seed voxel to another voxel. This allowed us to compute the connectivity probability from a region to another region.

We computed the connectivity probability of 89 brain regions to each other using tractography, and then computed the unidirectional connectivity between two areas by averaging these two connectivity probabilities. This represents the structural connectivity network organization of the brain.

We used the Harvard Ascending Arousal Network Atlas to estimate the connectivity between the raphe nucleus region and the rest of the brain, and then used LeadDBS Mapper within Matlab to estimate the projections from the raphe nucleus to each AAL region.

The AAL parcellation is the least homogeneous parcellation scheme compared to a number of other parcellation schemes, but it yields excellent significant results in the whole-brain literature in general and is highly suitable for our very extensive computational demands.

Nine healthy subjects were included in a fMRI study on psilocybin. They had all used psilocybin at least once before, but not within 6 wk of the study.

Subjects underwent two 12-min eyes-closed resting-state fMRI scans, one with psilocybin and one without. Psilocybin was injected i.v. 6 min after the start of the scans and lasted 60 s.

Anatomical scans were performed on a 3-T GE HDx system with 1-mm isotropic voxel resolution.

Two BOLD-weighted fMRI data sets were acquired using a gradient echo planer imaging sequence. The data sets were acquired in an interleaved fashion using 35 oblique axial slices.

The fMRI data were preprocessed using MELODIC 3.14 (60), part of FSL (FMRIB’s Software Library), and averaged across all voxels within each region defined in the AAL atlas.

We conduct the LEiDA analysis on the neuroimaging data to extract the PMSs, which are the amounts of interregional BOLD signal synchrony at each time point.

The phase coherence between two regions with temporarily aligned BOLD signals at a given time point is estimated using the Hilbert transform for each BOLD regional time course. The resulting dFC(t ) matrix is symmetric across the diagonal and represents the phase coherence of the brain at each time point. The leading eigenvector, V1(t ), is used to characterize the evolution of the dFC(t ) matrix over time.

We identified PMSs in the fMRI data by clustering the leading eigenvectors and used HCP Workbench to render the PMSs onto a cortical surface. The optimal number of clusters was 3 according to Silhouette analysis and the minimal P value for significant differences between probabilities between conditions.

We computed the probability of occurrence and lifetime of each metastable substate in each condition, and the switching matrix to capture the trajectories of FC dynamics in a directional manner. We then applied a permutation-based paired t test to assess differences between conditions.

210 healthy male and female controls were included in the study, and 232 PET scans were taken. Of these, 189 participants had only one scan, 20 participants had two scans, and 1 participant had three scans.

We extracted receptor density maps for each individual AAL region 5-HT1A, 5-HT1B, 5-HT2A, and 5-HT4 as well as 5-HTT using standard FSL tools on the freely available receptor density maps in MNI space. We then modeled the whole brain as a network of neurons using a dynamic mean field model.

The input currents received by the inhibitory and excitatory pools of neurons are converted into firing rates by the neuronal response functions, H(E, I), using the input – output function of Abbott and Chance (65). The GABA receptors control the average firing rate.

NMDA receptors control the synaptic gating variable of excitatory pools. The weight of recurrent excitation is 1.4 and the overall effective external input is 0.382 nA.

We used the DMF model based on Wong and Wang (64) to emulate the resting-state condition, and adjusted the inhibition weight for each node such that the firing rate remained clamped at 3 Hz.

We used parcellated structural and functional MRI data from 90 cortical and subcortical brain regions to simulate resting-state activity of nine participants. The optimal working point of the system was found by running 200 simulations with a global coupling factor of G between 0 and 2.5.

To transform the simulated mean field activity from our DMF model into a BOLD signal, the generalized hemodynamic model of Stephan et al. (70) was used. This model involves computing the BOLD signal in each brain area.

The whole-brain neuronal and neurotransmitter systems are mutually coupled through simulating the release-and-reuptake dynamics of the serotonin neurotransmitter and the firing rate activity of the raphe brain region.

The interaction between the neuronal and neurotransmitter system is defined by the first term on the right-hand side in Eq. 13, and the second term on the right-hand side is defined by the maximum reuptake rate and the half-maximum substrate concentration.

The reverse coupling is described in Eqs. 14-16, where the excitatory pyramidal and GABAergic inhibitory neurons generate an extra current.

The density of a serotonin receptor Rn weighs the serotonin modulated current vector Mn, which is given by the sigmoid-like function.

We fitted a mutually coupled whole-brain model to neuroimaging data using a 10 J 0.1 s time interval.

We computed the entropy rate S of a Markov chain, with N states and transition matrix P, and found that the stationary probability of state i is given by p(i).

For each transition matrix, we obtained the stationary distribution and computed the entropy rate. We then compared the two transition probability matrices.

The human connectome is a structural description of the human brain. Human brain networks function in connectome-specific harmonic waves, and whole-brain effective connectivity can be used to extract orthogonal subject- and condition-specific signatures from fMRI data. A multimodal neuroimaging model using serotonin receptor maps explains non-linear functional effects of LSD, and a high-resolution in vivo atlas of the human brain’s serotonin system is modulated by psilocybin.

Psilocybin, ketamine, LSD and a combination of these drugs increases spontaneous MEG signal diversity, which may be a factor in the treatment of anxiety in patients with advanced-stage cancer. Psilocybin is used in the treatment of tobacco addiction, treatment-resistant depression, and to understand neuropsychiatric disorders. The brain’s functional connectivity evolves on multiple time-scales over a static structural connectome.

Resting-state functional connectivity emerges from structurally and dynamically shaped slow linear fluctuations, and local excitation-inhibition ratio impacts the whole brain dynamics. C. E. John, E. A. Budygin, Y. Mateo, S. R. Jones, and G. Deco studied dopamine and serotonin in the ventral tegmental area and substantia nigra of the mouse, respectively. The stability of BOLD fMRI correlations during rest is related to cognitive performance in healthy older adults, and fast transient networks in spontaneous human brain activity are a source of information for neuromodulation of attention. Cholinergic control of information coding, how psychedelics work, and relating structure and function in the human brain are discussed. Functional connectivity dynamics: Modeling the switching behavior of the resting state, T. Pfeffer et al., Catecholamines alter the intrinsic variability of cortical population activity and perception, G. Deco et al., Whole-brain computational modelling of empirical MEG data, M. Jenkinson et al., Improved optimization for the robust and accurate linear registration and motion correction.

T. E. Behrens et al. described uncertainty in diffusion-weighted MR imaging and B. L. Edlow described the neuroanatomic connectivity of the human ascending arousal system critical to consciousness and its disorders. A. Horn, A. A. Kühn, Lead-DBS, K. Setsompop, E. M. Gordon, Probabilistic conversion of neurosurgical DBS electrode coordinates into MNI space, C. F. Beckmann, S. M. Smith, The dynamic functional connectome, Cimbi, and the center for integrated molecular brain imaging are some of the articles mentioned. K. F. Wong, X. J. Wang, L. F. Abbott, F. S. Chance, Drivers and modulators from push-pull and balanced synaptic input, B. D. Burns, A. Webb, Unit activity in monkey parietal cortex related to haptic perception and temporary memory, W. R. Softky, C. Koch, Neuroimage 38, 387 – 401 (2007). Functional connectivity in the motor cortex of resting human brain using echo-planar MRI was studied. Cortical hubs were revealed by intrinsic functional connectivity and were related to Alzheimer’s disease.

Study details

Topics studied
Neuroscience

Study characteristics
Theory Building Bio/Neuro

Authors

Authors associated with this publication with profiles on Blossom

Robin Carhart-Harris
Dr. Robin Carhart-Harris is the Founding Director of the Neuroscape Psychedelics Division at UCSF. Previously he led the Psychedelic group at Imperial College London.

PDF of Dynamic coupling of whole-brain neuronal and neurotransmitter systems