This neuroimaging model (2018) of the whole brain (under LSD influence) offers causal (and non-linear) mechanisms linking neuromodulation and neuronal activity.
Abstract
“Understanding the underlying mechanisms of the human brain in health and disease will require models with necessary and sufficient details to explain how function emerges from the underlying anatomy and is shaped by neuromodulation. Here, we provide such a detailed causal explanation using a whole-brain model integrating multimodal imaging in healthy human participants undergoing manipulation of the serotonin system. Specifically, we combined anatomical data from diffusion magnetic resonance imaging (dMRI) and functional magnetic resonance imaging (fMRI) with neurotransmitter data obtained with positron emission tomography (PET) of the detailed serotonin 2A receptor (5-HT2AR) density map. This allowed us to model the resting state (with and without concurrent music listening) and mechanistically explain the functional effects of 5-HT2AR stimulation with lysergic acid diethylamide (LSD) on healthy participants. The whole-brain model used a dynamical mean-field quantitative description of populations of excitatory and inhibitory neurons as well as the associated synaptic dynamics, where the neuronal gain function of the model is modulated by the 5-HT2AR density. The model identified the causative mechanisms for the non-linear interactions between the neuronal and neurotransmitter system, which are uniquely linked to (1) the underlying anatomical connectivity, (2) the modulation by the specific brainwide distribution of neurotransmitter receptor density, and (3) the non-linear interactions between the two. Taking neuromodulatory activity into account when modeling global brain dynamics will lead to novel insights into human brain function in health and disease and opens exciting possibilities for drug discovery and design in neuropsychiatric disorders.”
Notes
Neuromodulation: “The alteration of nerve activity through targeted delivery of a stimulus, such as electrical stimulation or chemical agents, to specific neurological sites in the body.”
Summary
INTRODUCTION
Brain activity results from the self-organization of large neural networks, emerging from complex recursive non-linear interactions between interconnected neural populations. Whole-brain models relying on the mean activity and variance of excitatory and inhibitory neuronal populations have been increasingly studied.
Neuronal population models rely on structural and functional connectivity of a number of anatomically defined brain regions, but the activity of these populations is strongly modulated by the synergistic interactions of the diffuse ascending systems, including acetylcholine, various monoamines, and tryptamines.
To address the problem of neurotransmitter modulation of whole-brain activity dynamics, we combined standard anatomical and functional maps of the human brain with a detailed map of 5-HT2A receptor density of the neuromodulator serotonin. We added the receptor maps to the standard whole-brain model by investigating how gain values can be adapted by the local regional values of 5-HT2AR density.
RESULTS
We used a whole-brain model to explain how neuromodulation is coupled with the neuronal system, and how function emerges from the underlying anatomy. The model only uses two parameters: a neuronal parameter and a neuromodulator parameter.
We investigated the influence of neuromodulation on the neuronal system using LSD to enhance the emotional response and produce greater feelings of wonder and transcendence compared with listening to music after placebo.
A model was fitted to the placebo condition and the LSD condition, and the role of the empirical 5-HT2AR was assessed by comparing the LSD maps with neuromodulatory maps of randomly shuffled 5-HT2AR densities.
The whole-brain model was composed of 90 anatomically delineated brain regions linked by the structural connectivity (SC) matrix of fiber densities obtained by tractography. The dynamic neuronal mean-field model represented the activity of each region and the inter-regional coupling was scaled by a single global parameter, G.
Explaining the Influence of Neuromodulation
The whole-brain mean-field model was optimized on the basis of diverse structural, functional, and regulatory data. The optimal global coupling parameter value was subsequently used to explain the LSD condition by selectively changing the neuronal gain of each region according to the empirical measured 5-HT2AR density.
The gain values of the model were adapted by taking into account the local regional values of 5-HT2AR density. This allowed us to quantitatively ascribe the LSD-induced whole-brain activity dynamics to one type (i.e., 5-HT2A) of serotonergic modulation of brainwide neural responses.
Results of Fitting Whole-Brain Neuromodulation Model to Empirical Data
We first estimated the optimal coupling parameter G such that the whole-brain model optimally fits the placebo condition. The results show that G = 2.1 at minimum of green line in Figure 3A indicates an excellent fit.
The neuromodulatory effects of LSD were modeled by estimating the neuronal gain function, and this approach significantly changed the fit, revealing an optimal sE value of approximately 0.2. In contrast, trying to improve the fit with the placebo condition did not show any improvement.
To demonstrate that the LSD function is dependent on the precise distribution of the 5-HT2AR, we randomly shuffled the empirical 5-HT2AR densities across the regions at the optimal gain value sE.
We used a number of different strategies to test the specificity of receptor-binding maps, and found that the 5-HT2AR receptor-binding map performed the best, followed by the 5-HT1A, 5-HT1B, 5-HT4, and 5-HTT receptor-binding maps.
We tested the generalization capability of our model on a random 50% subset of participants and found that the training performance was very similar to the original results, while the testing performance was equally good.
DISCUSSION
We created a whole-brain model integrating neuronal and neuromodulation multimodal data from dMRI and fMRI with neurotransmitter data obtained with PET, which revealed a detailed whole-brain map of 5-HT2AR densities. This allowed us to explain the functional effects of 5-HT2AR stimulation with well-known 5-HT2AR agonist compound LSD in healthy participants.
The serotonin (5-hydroxytryptamine, 5-HT) system, which is an evolutionary conserved neuromodulator/neurotransmitter, regulates many psychophysiological functions and modulates cognitive capacities. It is synthesized within the brainstem’s raphe nuclei (dorsal and median), and has 14 subtypes, including the 5-HT2A receptor, which mediates adaptive plasticity.
The present results extend the recent findings of Shine and colleagues to the human whole-brain level and emphasize the role of PET-based binding receptors for global brain dynamics. Furthermore, future research should seek to describe the full entanglement of the two very different neural and neurotransmitter dynamical systems with very different timescales.
Numerous investigations in human and animals have demonstrated that functional whole-brain activity depends both on effective connectivity and so-called brain states, and that such evolving activity patterns are affected by diseases and may eventually be used to predict approaching criticalities.
A mechanistic understanding of the imbalances found in neuropsychiatric disorders may help open up rational ways for effective brain interventions and identify biomarkers stratifying a broad illness phenotype into a finite number of treatment-relevant subgroups.
Based on the current findings, a pipeline could be used to combine structural, functional, and neurotransmitter neuroimaging data for modeling a disease state. Once the model is established, the regional drug receptor modulation can be optimized.
We have demonstrated how to integrate anatomical brainwide distribution of neuromodulatory activity into a whole-brain computational model to provide new causative insights into human brain function.
STAR+METHODS
Anatomical connectivity, functional connectivity, neurotransmitter density were measured using diffusion magnetic resonance imaging, functional magnetic resonance imaging, and positron emission tomography.
Functional dataset using functional magnetic resonance imaging
All 15 participants provided written informed consent to participate in the study, which was approved by the National Research Ethics Service committee London-West London and conducted in accordance with the revised declaration of Helsinki (2000), the International Committee on Harmonization of Good Clinical Practice guidelines, and National Health Service Research Governance Framework.
Neurotransmitter dataset using positron emission tomography
The participants were healthy male and female controls from the Cimbi database, and a total of 232 positron emission tomography scans and corresponding structural MRI scans were acquired for 210 individual participants.
Study setting and overview
Screening took place at Imperial’s clinical research facility at the Hammersmith hospital campus. Participants received placebo and LSD on separate study days.
On scanning days, volunteers arrived at the study center at 8:00am and gave a urine test, breathalyzer test, and were given 75 mg of LSD intravenously followed by an infusion of saline. They were then placed inside a mock MRI scanner for 60 min.
Participants reported subjective drug effects between 5 to 15 min post-dosing, which reached a peak between 60 to 90 min post-dosing. The study psychiatrist assessed the participant’s suitability for discharge.
The BOLD scanning consisted of three eyes-closed resting state scans, each lasting seven minutes. The second scan included listening to two excerpts of music from two songs by ambient artist Robert Rich, and the volume-maximization and broadband compression was performed using Ableton Live 9 software.
- Neurotransmitter density
PET data were acquired on a Siemens HRRT scanner in list mode with a 2 mm in-plane resolution using [11C]Cimbi-36 as a selective serotonin 2A (5-HT2A) receptor agonist radioligand. Structural MRI data were acquired on four different Siemens scanners with standard parameters.
Further processing was performed with FreeSurfer 5.3 using a surface and a volume stream, corrected for gradient nonlinearities, and PET – MR coregistration was estimated using boundary-based registration.
QUANTIFICATION AND STATISTICAL ANALYSIS
We provide a general overview of the analysis pipeline used to integrate structural and functional connectivity with neurotransmission in healthy participants.
Structural connectivity, functional connectivity and neuromodulation data were integrated into a whole-brain computational model to fit the placebo condition.
Tractography
We used the structural connectivity between 90 AAL regions obtained in a previous study averaged across 16 healthy young adults (5 females, mean SD age: 24.75 2.54) to generate the structural connectivity maps for each participant using the dMRI data acquired.
We estimated the whole-brain network using probabilistic tractography, and averaged data across participants. The AAL parcellation scheme appeared to be a good choice for this purpose, because it yields excellent significant results in the whole-brain literature in general. We used the FSL diffusion toolbox (Fdt) to process the diffusion MRI data and estimated the local probability distribution of fiber direction at each voxel.
A connectivity probability matrix was calculated from 5000 streamlines per voxel to calculate the connectivity probability from a seed voxel to another voxel. This matrix was extended to the region level.
We applied the AAL90 template to the b0 image in diffusion MRI space, then to the T1 template of ICBM152 in MNI space, and then to the AAL templates.
Preprocessing
We first preprocessed the fMRI data using MELODIC 3.14, part of FSL, and corrected head motion using mcflirt. The BOLD signals were extracted and averaged for each participant and for each brain state using the AAL atlas. We computed the static functional connectivity (FC) and the functional connectivity dynamics (FCD) matrices, and compared the distributions of the elements of those matrices between empirical and whole-brain model level of fitting using the Kolmogorov-Smirnov (KS) distance and correlation between the FC matrices.
We computed the FC over 30 TR with increments of 2 TR and then computed a time-versus-time matrix of Functional Connectivity Dynamics (FCD), which revealed a characteristic checkered pattern indicative of spontaneous switching between different FC patterns. The distribution of FCD values contained valuable information regarding the time-dependencies of resting-state activity.
- Parcellation
We used the AAL atlas to integrate structural, functional and neuromodulation data and extracted the average receptor density for each individual AAL region.
- Whole-Brain Dynamic Mean Field Model
We used a network model to simulate spontaneous brain activity at the whole-brain level. The activity in each brain area is represented by the Dynamic Mean Field (DMF) model proposed by Deco et al. .
The input-output function of Abbott and Chance [80] is used to model the firing rate of neurons in the brain. The synaptic gating variable of excitatory pools is controlled by NMDA receptors with a decay time constant tNMDA = 0.1 s and g = 0.641, whereas the synaptic gating variable of inhibitory pools depends on GABA receptors.
We considered N = 90 brain areas in our whole-brain network model, and adjusted the coupling factor G to emulate the empirical resting-state scans from 15 participants.
We used a generalized hemodynamic model to transform the simulated mean field activity into a BOLD signal, which is then used to control blood flow and deoxyhemoglobin content.
The BOLD signal in each area n is a static nonlinear function of volume and deoxyhemoglobin content.
- Serotonergic Modulation
Serotonin injection modulates the firing rate of neurons with 5-HT receptors in cats and rats, indicating that 5-HT2A receptors have both excitatory and inhibitory effects.