Mapping Pharmacologically-induced Functional Reorganisation onto the Brain’s Neurotransmitter Landscape

This theory-building preprint (2022) uses data from fMRI and PET scans (of the brain) to show how different pharmacological interventions (including psychedelics) interact with neurotransmitters. The results show opposite routes, but similar mechanisms, regarding how psychedelics and anaesthetics (e.g. ketamine) interact with the brain (molecular chemoarchitecture).

Abstract

“To understand how pharmacological interventions can exert their powerful effects on brain function, we need to understand how they engage the brain’s rich neurotransmitter landscape. Here, we bridge microscale molecular chemoarchitecture and pharmacologically-induced macroscale functional reorganisation, by relating the regional distribution of 18 neurotransmitter receptors and transporters obtained from Positron Emission Tomography, and the regional changes in functional MRI connectivity induced by 7 different mind-altering drugs including anaesthetics, psychedelics, and cognitive enhancers. Our results reveal that psychoactive drugs exert their effects on brain function by engaging multiple neurotransmitter systems. Intriguingly, the effects of both anaesthetics and psychedelics on brain function, though opposite, are organised along hierarchical gradients of brain structure and function. Finally, we show that regional co-susceptibility to pharmacological interventions recapitulates co-susceptibility to disorder-induced structural alterations. Collectively, these results highlight rich statistical patterns relating molecular chemoarchitecture and drug-induced reorganisation of the brain’s functional architecture.”

Authors: Andrea I. Luppi, Justine Y. Hansen, Ram Adapa, Robin L. Carhart-Harris, Leor Roseman, Daniel Golkowski, Andreas Ranft, Rüdiger Ilg, Denis Jordan, Alexander R. D. Peattie, Anne E. Manktelow, Draulio B. de Araujo, Stefano L. Sensi, Adrian M. Owen, Lorina Naci, David K. Menon, Bratislav Misic, Emmanuel A. Stamatakis

Summary

We compared the regional distribution of 18 neurotransmitter receptors and transporters obtained from Positron Emission Tomography with the regional changes in functional MRI connectivity induced by 7 different mind-altering drugs. Our results revealed that psychoactive drugs exert their effects on brain function by engaging multiple neurotransmitter systems.

Psychoactive compounds provide neuroscientists with a means to manipulate the brain without requiring surgical intervention, and are therefore a prominent tool for causal investigation of brain and cognitive function in healthy humans 1.

Mind-altering drugs such as anaesthesia, cognitive enhancers, and psychedelics play a fundamental role in modern clinical practice, and are increasingly being investigated for their potential to treat psychiatric conditions.

Pharmacological agents alter the brain’s neurotransmitter landscape, modulating the functional configurations of neuronal circuits and ultimately shaping network-wide communication. Several psychoactive drugs appear to exert their effects on the mind and brain primarily through one or few specific neurotransmitters.

The primary molecular target of a drug is not sufficient to understand a drug’s effects on brain function, because the brain’s intricate, nested feedback loops and recurrent pathways of connectivity may mean that even a relatively selective drug can influence unrelated systems beyond what may be apparent from in vitro studies. Different pharmacological agents can exert intriguingly similar effects on both the mind and brain, despite acting on different pathways.

The human brain exhibits rich patterns of anatomical, functional, cytoarchitectonic, and molecular variations, which extend to the regional distribution of different neurotransmitter receptors and transporters. This is important for understanding how drugs influence neurotransmission.

Here, we used PET scanning to map the neurotransmitter landscape of drug-induced alterations in the brain’s functional connectivity, and resting-state functional MRI to characterise a drug’s effects on the brain’s spontaneous activity, without the interference of any specific task.

This study aims to map the relationships between cortical neurotransmitter distributions and a variety of psychoactive pharmacological agents, and their effects on functional connectivity.

We combined data from multiple studies to establish a relationship between neurotransmitter systems and pharmacologically-induced reorganisation of the brain’s functional architecture. We considered propofol, sevoflurane, modafinil, methylphenidate, ketamine, ayahuasca and LSD.

We considered the cortical distribution of 14 neurotransmitter receptors and 4 transporters, obtained from in vivo Positron Emission Tomography, and parcellated the results into 100 functionally defined regions according to the Schaefer atlas.

We reasoned that regions that express similar patterns of receptors and transporters should exhibit similar patterns of susceptibility to drug-induced functional reorganisation.

To address this question, we computed matrices of pharmacological co-susceptibility and neurotransmitter co-expression between pairs of regions, and regressed out the exponential trend with Euclidean distance.

We found that pharmacological co-susceptibility is significantly correlated with neurotransmitter profile similarity, which means that regions that exhibit shared chemoarchitecture respond similarly across pharmacological perturbations.

To synthesise the two datasets, we employed a multivariate association technique, Partial Least Squares correlation, which identified patterns of maximum covariance between drug-induced effects on functional connectivity, and cortical distributions of neurotransmitter expression.

This analysis indicated that there were two statistically significant latent variables relating pharmacologically-induced functional reorganisation to neurotransmitter profiles, which were cross-validated using a distance-dependent method.

For each latent variable, a neurotransmitter and a drug score are associated with a brain region. Neurotransmitter loadings showed a distinction between pharmacological effects into two groups, with anaesthetics on one end and psychedelics and cognitive enhancers on the other.

The second latent variable (PLS2) separated dorsal and ventral aspects of the brain largely based on neurotransmitter loadings and drug loadings.

Neurotransmitters and drugs whose activity correlates positively with the score pattern covary with one another in the positively scored regions.

We observed that the regional distribution of PLS1 scores corresponds to the main axis of covariance between neurotransmitter expression and pharmacologically-induced functional reorganisation. We also observed that PLS1 scores correspond to the principal gradient of regional prevalence of different kinds of information, from redundancy to synergy.

We observed significant correlations between cortical hierarchies and neurotransmitter and drug scores, and between anaesthetics and psychedelics, except for methylphenidate, which exhibited no significant correlations.

The first two PLS latent variables revealed a stark division between transporters and receptors, which discriminates between anaesthetics and other psychoactive substances. The second latent variable also revealed a strong relationship between drug interventions and receptor systems.

We combined 11 spatial maps of cortical thickness abnormalities made available by the Enhancing Neuro Imaging Genetics Through Meta Analysis (ENIGMA) consortium to explore the relationship between functional co-susceptibility of different regions to transient pharmacological perturbations and structural perturbations resulting from different neurological, neurodevelopmental, and psychiatric disorders.

We found a correlation between regional patterns of cortical abnormality across 11 disorders and a previously derived matrix of regional co-susceptibility to pharmacological perturbations. This suggests that there may be common patterns of regional co-susceptibility to different kinds of perturbations.

We used a non-linear dimensionality reduction algorithm to obtain joint gradients of variation from pharmacological and disease-associated co-susceptibility. We found that the first two gradients coincide with the two principal gradients of functional connectivity of Margulies et al 57.

Here, we mapped the functional chemoarchitecture of the human brain, by relating the regional reorganisation of fMRI functional connectivity induced by 8 different mind-altering drugs to the cortical distribution of 18 neurotransmitter receptors and transporters obtained from PET 31.

Using our computational framework, we found that psychoactive drugs affect the brain via multiple neurotransmitter systems, and that the effects of these drugs are topographically organised along multiple hierarchical gradients of brain function, anatomy, and neurobiology.

The present results add another dimension to recent work that found widespread involvement of multiple receptors in psychedelic experiences. Psychedelics and anaesthetics also exert opposite effects on brain activity and connectivity, as well as on structure-function coupling.

The main division we observed in terms of neurotransmitters is between receptors and transporters, which displayed opposite associations with drug-induced effects. Psychedelics and GABA-ergic anaesthetics have been shown to have potent effects on higher-order association cortices.

Multiple aspects of neuroanatomy may contribute to the hierarchical organisation of mind-altering drug effects. These include the fact that receptors are organised along the brain’s sensory-to-association hierarchical axis and that transmodal cortices are especially susceptible to multiple pharmacological influences.

We observed that pharmacologically-induced changes in functional connectivity correlate with the map of regional cerebral blood flow, and that greater cerebral blood flow in transmodal cortices facilitates especially high availability of the drug in these regions.

We conjecture that the hierarchical organisation of pharmacologically induced changes in functional connectivity may be explained by the fact that transmodal association cortices are especially diverse in their receptor profiles, rich in some key receptors, and more susceptible to pharmacological intervention.

We found that regions with similar functional connectivity are also more similar in their susceptibility to cortical alterations associated with a variety of neuropsychiatric disorders. This suggests that the principal gradient of neurotransmitter expression is particularly relevant for predicting a wide spectrum of disease-specific cortical morphology.

The results reported here open new possibilities for mapping the effects of potent pharmacological interventions on the brain’s functional architecture. It is intriguing that regions of the brain that are structurally most vulnerable to disorder may also be the ones most susceptible to pharmacological intervention.

Although the main strength of our study is our extensive coverage of both neurotransmitters and pharmacological data, it is important to acknowledge that neither is complete.

Although we have endeavoured to mitigate scanner and acquisition differences by using a within-subject design and re-preprocessing all data with the same pipeline, we cannot exclude some residual influence of such differences on our results. Similar considerations apply to the PET and ENIGMA datasets, as well as to the ENIGMA disorder data: the datasets do not directly reflect changes in tissue volume, but rather the effect size of patient-control statistical comparisons.

Although we report a macroscale spatial association between neurotransmitter expression and pharmacologically-induced functional reorganisation that is statistically unexpected based on autocorrelation alone, caution is warranted when drawing inferences from statistical results to the underlying biology.

We conducted a PET study on the cortex, but did not include the thalamus, brainstem, and other subcortical structures. We expect that future work combining these approaches will provide richer insights.

We mapped receptors and drugs in separate cohorts of individuals, and identified spatially correlated patterns. These results generate empirically testable hypotheses about which neurotransmitters may be involved with the macroscale effects of different drugs on brain function.

Models of the brain’s functional connectivity can be enriched with further information, such as regional myelination, specific receptors and ion channels, to reflect neurotransmitter influences.

We mapped the functional chemoarchitecture of the human brain by relating regional changes in fMRI functional connectivity to the regional distribution of 18 neurotransmitter receptors and transporters obtained from PET. This work provides a computational framework to characterise how mind-altering drugs engage the brain’s rich neurotransmitter landscape.

Propofol is the most common agent used for intravenous induction and maintenance of general anaesthesia, and is used both in the operating room and for scientific studies because of its rapid action and minimal effects on regional cerebral blood flow and the coupling between blood flow and metabolism.

The Western University propofol data were collected between May and November 2014 at the Robarts Research Institute, Western University, London, Ontario (Canada), and have been published before 18,144,184,185. Healthy volunteers were recruited and provided written informed consent.

Resting-state fMRI data were acquired at different propofol levels: awake, deep anaesthesia (corresponding to Ramsay score of 5) and also during post-anaesthetic recovery. Participants’ level of responsiveness was evaluated using a verbal memory recall and a computer-based auditory target-detection task.

Propofol was administered intravenously using an AS50 auto syringe infusion pump. A pharmacokinetic model delivered via the infusion pump was used to achieve target-controlled infusion of propofol, and Ramsay level 5 was achieved when participants stopped responding to verbal commands.

Once Ramsay sedation level 5 was reached, participants stopped responding to both tasks and propofol concentrations were measured. Oxygen was titrated to maintain SpO2 above 96%.

At Ramsay 5 level, participants remained capable of spontaneous cardiovascular function and ventilation. However, intubation during scanning could not be used to ensure airway security, so scanner time was minimised to ensure return to normal breathing following deep sedation.

Once in the scanner, participants relaxed with closed eyes and listened to a story through headphones for 8 minutes. The present analysis focuses on the resting-state data only.

Anatomical scanning was performed with a 3D MPRAGE 777 sequence, using the following parameters: TA = 5min, TE = 4.25ms, 240×256 matrix size, 9 degrees flip angle 18.

The Cambridge University propofol dataset contains data from 25 healthy volunteer subjects. They gave informed consent to participate in the study and were provided with standard information about intravenous cannulation, blood sampling and MRI scanning.

Propofol was administered intravenously as a “target controlled infusion” (plasma concentration mode) to three different subjects. MRI was acquired at each stage, and also at Recovery, and measurements of heart rate, non-invasive blood pressure, and oxygen saturation were recorded at regular intervals.

815 MRI data were acquired on a Siemens Trio 3T scanner (WBIC, Cambridge) for 25 healthy subjects. 10 were excluded, either because of missing scans (n=2) or due to excessive motion in the scanner (n=8, 5mm maximum motion threshold).

Sevoflurane is an inhalational anaesthetic that acts via GABA-A receptors, NMDA, AMPA and nicotinic ACh receptors. It may also interact with Na+, K+ and HCN channels.

Twenty healthy adult men were recruited through campus notices and personal contact, and compensated for their participation in the study.

Before inclusion, patients underwent a focused physical examination, a resting electrocardiogram, and were assessed for any previous neurologic or psychiatric disorder, chronic intake of medication, hardness of hearing or deafness, absence of fluency in German, and presence of metal implants.

Sevoflurane concentrations were chosen so that subjects tolerated artificial ventilation and burst-suppression was reached in all 864 participants. An intermediate concentration of 3.0 vol% was also used to make group comparisons feasible. Sevoflurane mixed with oxygen was administered via a tight-fitting facemask using an fMRI-compatible anaesthesia machine. The fraction of inspired oxygen was set at 0.8 and mechanical ventilation was adjusted to maintain end-tidal carbon dioxide at steady concentrations of 33 1.7 mmHg during BS, 34 1.12 mmHg during 3 vol%, and 33 1.49 mmHg during 2 vol%. Norepinephrine was given to maintain the mean arterial blood pressure close to baseline values, and sevoflurane concentration was gradually increased until the EEG showed burst-suppression with suppression periods of at least 1,000 ms and about 50% suppression of electrical activity.

After an equilibration time of 15 min, sevoflurane concentrations were reduced to two times the concentration at LOR, but most subjects moved or did not tolerate the laryngeal mask any more.

Sevoflurane administration was terminated, the scanner table was slid out of the MRI scanner to monitor post-anaesthetic recovery, and the volunteer was manually ventilated until spontaneous ventilation returned. A Brice interview was administered to assess for awareness during sevoflurane exposure, and a resting-state combined fMRI-EEG scan was acquired.

In the original study, functional MRI and EEG data were collected on a 3-Tesla magnetic resonance imaging scanner. In the present work, we only considered fMRI data and used data from the Awake, 3% vol, 927 and Recovery scans.

Ketamine is a multi-faceted drug that can act as a dissociative anaesthetic or atypical psychedelic, and has also found recent use as a fast-acting antidepressant. Its mechanisms of action are yet to be fully elucidated, but it appears to be primarily an antagonist of NMDA and HCN1 receptors.

21 participants were recruited via advertisements placed throughout central Cambridge, UK, and underwent a screening interview. All were right-handed, free of current or previous psychiatric or neurological disorder or substance abuse problems, and had no history of cardiovascular illness or family history of psychiatric disorder/substance abuse.

Participants received a continuous computer-controlled intravenous infusion of a racemic ketamine solution until a targeted plasma concentration of 100 ng/ml was reached. A saline infusion was administered on the other occasion, and blood samples were drawn before and after the resting fMRI scan.

All MRI and assessment procedures were identical across assessment occasions, and participants were instructed to close their eyes and let the minds wander without going to sleep. T2*-weighted echo-989 planar images were acquired under eyes-closed resting-state conditions, and high-resolution anatomical T1 images were acquired using a three-dimensional magnetic-prepared rapid gradient echo sequence.

LSD is perhaps the best-known among classic psychedelics, inducing a powerful state of altered consciousness with subjective experiences including hallucinations and “ego dissolution” 46,48,49.

LSD’s main neural and subjective effects originate from its agonism of the 5HT2A receptor, and the spatial distribution of the 5HT2A receptor in the brain has been shown to correlate with the effects of LSD on brain activity and connectivity.

The data employed in this study have been extensively published, and were collected in accordance with the revised declaration of Helsinki (2000). Imperial College London sponsored the research, which was conducted under a Home Office license for research with schedule 1 drugs. Twenty healthy volunteers with previous experience using psychedelic drugs were scanned.

1046 volunteers underwent two scans, 14 days apart, with one scan being a placebo and the other an active dose of LSD. They reported marked alterations of consciousness under LSD.

The infusion of drug/placebo was administered over 2 min and occurred 115 min before the resting-state scans were initiated. ASL and BOLD scanning consisted of three seven-minute eyes closed resting state scans.

In 15 subjects, BOLD-weighted fMRI data were acquired using a gradient echo planer imaging sequence on a 3T GE HDx system. The subjects were eyes-closed, resting state without stimulation, and the second BOLD scan involved listening to music.

Ayahuasca is a tea made from two plants, Psychotria viridis and Banisteriopsis caapi, that contains tryptamine, DMT, and beta-carboline alkaloids. These alkaloids prevent the degradation of monoamine neurotransmitters and thus increase their levels.

Data were obtained from 9 healthy adult volunteers who had ingested 120 – 200 mL of Ayahuasca known to contain 0.8 mg/mL of DMT and 0.21 mg/mL of harmine. All volunteers had at least 8 years of formal education and were not under medication for at least 3 months prior to the scanning session.

1109 volunteers underwent fMRI scanning before and 40 minutes after Ayahuasca intake. They were instructed to close their eyes and remain awake and at rest.

fMRI images were obtained in a 1.5 T scanner (Siemens, Magneton Vision) using an EPI-BOLD like sequence comprising 150 volumes, and a multiplanar reconstructed gradient-echo sequence with the following parameters. The final dataset included 8 subjects.

Modafinil is a wakefulness promoting drug used to treat sleep disorders, attention and memory problems, ADHD, and mood disorders. It influences several neurotransmitters, including the dopamine and noradrenaline transporters, the locus coeruleus noradrenergic firing, and the histamine system.

The modafinil dataset was recruited throughout February 2011, drug/placebo administration and fMRI acquisitions started on March 2011, went on until January 2012, and the study was completed with the last fMRI session in January 2012.

Twenty six young male right-handed adults with comparable levels of education were enrolled. They were instructed to maintain their usual amount of nicotine and caffeine intake and avoid alcohol consumption in the 12h before the initiation of the study.

Study subjects received a single dose of modafinil or a placebo, and were scanned before and 3 hours after ingesting the drug/placebo. All but one reported no modafinil-induced side effects or alterations in the sleep-wake cycle.

Using a Philips Achieva 3T Scanner, 140 functional volumes were acquired, consisting of 30 transaxial slices per run, with a volume TR of 1,671 ms. High resolution structural images were acquired at the end of the three rs-fMRI runs.

Methylphenidate is used to treat ADHD and narcolepsy. It inhibits the reuptake of dopamine and noradrenaline and has a minor affinity for the 5-HT1A receptor.

The methylphenidate dataset used here was obtained from patients suffering from traumatic brain injury. The patients were referred from the Addenbrooke’s Neurosciences Critical Care Unit Follow-Up Clinic, Addenbrooke’s Traumatic Brain Injury Clinic and The Royal London Hospital Intensive Care Unit.

Thirty-eight volunteers were recruited for the study, 17 into the TBI arm and 21 into the healthy control arm. Exclusion criteria included left-handedness, history of drug/alcohol abuse, history of psychiatric or neurological disorders, and physical handicaps.

The study consisted of two visits (separated by 2 – 4 weeks) for both groups of participants. They were randomly allocated in a Latin square design to receive one of the two interventions on the first visit (a placebo tablet or 30 mg tablet of methylphenidate), and the alternate intervention on the second visit. After 75 min, the volunteers completed a MRI scan that included fMRI and structural image acquisition, but without any pharmacological intervention.

Methylphenidate dataset: MRI data were acquired on a Siemens Trio 3-Tesla MR system. Diffusion tensor imaging (DTI) data were acquired without diffusion weighting (b=0) on one volume.

Functional imaging data were acquired using an echo-planar imaging (EPI) sequence with parameters TR = 2,000 ms, TE = 30 mm, Flip Angle = 78, FOV 192192 mm2, in-plane resolution 3.03.0 mm, 32 slices 3.0 mm thick with a gap of 0.75 mm between slices.

We applied a standard preprocessing pipeline to each dataset and condition, consisting of the following steps: removal of the first five volumes, to achieve steady state magnetization; motion correction; slice-timing correction; identification of outlier volumes for subsequent scrubbing.

We applied the anatomical CompCor method of denoising to the functional data, which involves regressing out the first five principal components attributable to white matter and cerebrospinal fluid signal, six subject-specific realignment parameters, and scrubbing outliers identified by A RT.

The step of global signal regression (GSR) has received substantial attention in the literature as a denoising method. However, recent work has demonstrated that the global signal contains information about various states of consciousness.

We summarised pharmacological effects on brain function as a vector of regional functional connectivity deltas for each subject at each condition, with each intervention being quantified with respect to the mean of controls’ node strength values.

We used Positron Emission Tomography to estimate receptor densities for 18 receptors and transporters across 9 neurotransmitter systems, and then combined the densities for receptors with more than one mean image of the same tracer.

PLS analysis was used to relate regional neurotransmitter density to pharmacologically-induced functional connectivity changes. The first PLS component (PLS1) is the linear combination of the weighted neurotransmitter density scores that have a brain expression map that covaries the most with the map of regional FC changes.

In this study, the effect size of a drug can be estimated by comparing the covariance between neurotransmitter density and drug-induced FC changes. Positively weighed neurotransmitters covary with positively weighed drug-induced FC changes, and negatively weighed neurotransmitters covary with negatively weighed drug-induced FC changes.

The score pattern for each latent variable was computed by projecting the original data onto the singular vector weights and then computing the Pearson’s correlation between each individual variable and the PLS analysis derived neurotransmitter score pattern.

We tested the statistical significance of each PLS component by permuting the response variables 1,000 times, while considering the spatial dependency of the data, and then randomly rotating the parcels. PLS analysis is used to determine if neurotransmitter density and drug-induced FC changes are spatially correlated. The spin test is applied to the singular values of the latent variables, producing a null distribution of singular values.

We quantified the spatial similarity of pharmacologically-induced patterns of change in functional connectivity strength with several canonical maps of hierarchical brain organisation, including the anatomical gradient of intracortical myelination, evolutionary cortical expansion, the principal component of variation in gene expression, and the principal gradient of regional prevalence of different kinds of information.

The ENIGMA consortium collected cortical thickness data from over 21,000 patients with 11 neurological, neurodevelopmental, and psychiatric disorders, and compared them to almost 26,000 healthy controls. The results are presented as z-scored effect sizes.

We constructed 11-element vectors of disorder abnormality for every brain region and correlated the abnormality vectors between brain regions to quantify disorder co-susceptibility.