Mapping-neurotransmitter-systems-to-the-structural-and-functional-organization-of-the-human-neocortex.md

Mapping neurotransmitter systems to the structural and functional organization of the human neocortex (2022)

by Justine Y. Hansen, Golia Shafiei, Ross D. Markello;Nature Neuroscience

https://github.com/netneurolab/hansen_receptors

https://www.nature.com/articles/s41593-022-01186-3

Contents

• Abstract
• Main
• Results
  • Receptor distributions reflect structural and functional organization
  • Receptor profiles shape oscillatory neural dynamics
  • Mapping receptors to cognitive function
  • Mapping receptors and transporters to disease vulnerability
  • Replication using autoradiography
  • Sensitivity and robustness analyses
• Discussion
• Comprehensive Brain Receptor Mapping in Health and Disease
• Methods
  • PET data acquisition
  • Autoradiography receptor data acquisition
  • Structural and functional data acquisition
  • Structural network reconstruction
  • Functional network reconstruction
  • Structure–function coupling
  • MEG power
  • ENIGMA cortical abnormality maps
  • Dominance analysis
  • Cognitive meta-analytic activation
  • Partial least squares analysis
  • Distance-dependent cross-validation
  • Null models
  • Reporting Summary
• Data availability
• Supplementary information

Abstract

Neurotransmitter Receptors:

• Support signal propagation in human brain
• Location and impact on emergent function not well understood
• No comprehensive atlas of receptors available

Collating Positron Emission Tomography (PET) Data:

• Constructed a whole-brain three-dimensional normative atlas of 19 receptors and transporters across nine neurotransmitter systems
• Data from over 1,200 healthy individuals used

Findings:

• Receptor profiles align with structural connectivity
• Mediate function, including neurophysiological oscillatory dynamics and resting-state hemodynamic functional connectivity

Topographic Gradient of Overlapping Receptor Distributions:

• Separates extrinsic and intrinsic psychological processes using Neurosynth cognitive atlas

Associations between Receptor Distributions and Cortical Abnormality Patterns:

• Found both expected and novel associations in an independently collected autoradiography dataset

Implications:

• Demonstrates how chemoarchitecture shapes brain structure and function
• Provides a new direction for studying multi-scale brain organization.

Main

Neurotransmitter Receptors and Brain Function

Concepts:

• Neurotransmitter receptors: Heterogeneously distributed across neocortex, modulate excitability and firing rate of cells, mediate transfer and propagation of electrical impulses, drive synaptic plasticity, shape network communication.
• Ionotropic vs. metabotropic receptors: Ionotropic affect membrane potential directly; metabotropic interact with intracellular second messengers.
• Neurotransmitter systems: Segregated and integrated information through specialized modules and hubs, respectively.

Challenges in Studying Neurotransmitter Receptors:

• Lack of comprehensive datasets for receptor distributions across multiple neurotransmitter systems.
• Autoradiography data available only in 44 cytoarchitectonically defined cortical areas; PET data limited by small cohorts and high cost.

Data Sharing Efforts:

• Comprehensive atlas of neurotransmitter receptor maps from 19 unique receptors, binding sites, and transporters across 9 neurotransmitter systems and over 1,200 healthy individuals: https://github.com/netneurolab/hansen_receptors.

Integrating Neurotransmitter Receptor Data with Other Modalities:

• Diffusion-weighted MRI and functional MRI: Neurotransmitter receptor densities follow brain's structural and functional connectomes.
• MEG: Neurotransmitter receptor densities shape oscillatory neural dynamics.
• Neurosynth functional activations: Spatially co-varying axis of neuromodulators and mood-related processes.
• ENIGMA cortical atrophy patterns: Specific receptor–disorder links.

Results

Neurotransmitter Receptor Profiles

PET Images:

• Collated PET images from 19 different neurotransmitter receptors, transporters, and receptor-binding sites across 9 neurotransmitter systems
• Included dopamine, norepinephrine, serotonin, acetylcholine, glutamate, GABA, histamine, cannabinoid, and opioid systems
• Acquired in healthy participants

Data Preprocessing:

• Parcellated PET tracer maps into the same 100 cortical regions
• Z-scored the data to mitigate acquisition and pre-processing variations

Analysis:

• Presented tracer maps for 19 unique neurotransmitter receptors and transporters from a combined total of 1,238 healthy participants
• Repeated analyses in an independently collected autoradiography dataset of 15 neurotransmitter receptors
• Examined across alternative brain parcellations

Results:

• Obtained mean receptor distribution maps for 19 different neurotransmitter receptors and transporters, totaling over 1,200 healthy participants
• Included non-displaceable binding potential (BPND), tracer distribution volume (VT), density (Bmax), standard uptake value ratio (SUVR) values

Note:

• Neurotransmitter receptor maps without citations refer to previously unpublished data
• Contact information for study principal investigators is provided in Supplementary Table 3
• Table 3 includes methodological details such as PET camera, number of males and females, modeling method, reference region, scan length, and modeling notes

Receptor distributions reflect structural and functional organization

Constructing a Cortical Neurotransmitter Receptor Atlas

Receptor Similarity:

• Quantified by correlating receptor density profiles of brain regions
• Decreases exponentially with Euclidean distance between brain region centroids
• Approximately normally distributed
• No single receptor or transporter exerts undue influence

Principal Component of Receptor Density:

• Represents a regional quantification of receptor similarity
• Separates insular and cingulate from somatomotor/posterior parietal regions
• Correlated with synapse density, supporting the notion that receptor expression depends on lamination

Stratifying Receptors:

• By biological mechanisms (excitatory/inhibitory, ionotropic/metabotropic, Gs-/Gi-/Gq-coupled)
• By neurotransmitter protein structure (monoamine/non-monoamine)

Relationship to Structural and Functional Connectivity:

• Receptor similarity is greater between anatomically connected brain regions
• Not due to spatial proximity or network topography, as shown by significance against surrogate structural connectivity matrices

Correlation of Receptor Similarity with Structural and Functional Connectivity:

• Significant correlation between receptor similarity and structural connectivity (P = 1.6 × 10−8, CI = [0.11, 0.23], two-sided) (Fig. 3a)
  • Receptor similarity greater between physically connected regions
  • Regression analysis shows positive correlation with structural connectivity
• Significant correlation between receptor similarity and functional connectivity (P = 7.1 × 10−61, CI = [0.20, 0.26], two-sided) (Fig. 3b)
  • Receptor similarity greater within same functional networks
  • Positive correlation with functional connectivity

Receptor Profiles and Structure–Function Coupling:

• Communicability of weighted structural connectome used to measure structure–function coupling
• Inclusion of receptor similarity improves prediction of regional functional connectivity in unimodal areas and paracentral lobule (Fig. 3c)
• Significance assessed against null distribution using a distance-dependent method.

Receptor profiles shape oscillatory neural dynamics

Neurotransmitter Receptors and Neural Oscillations

Relating Cortical Patterning of Neurotransmitter Receptors to Neural Oscillations:

• Analyzed MEG power spectra across six canonical frequency bands from HCP2930
• Fit multiple linear regression models that predict cortical power distribution of each frequency band from neurotransmitter receptor and transporter densities
• Cross-validated the model using a distance-dependent method
• Assessed significance against a spin-permuted null model (10,000 repetitions) and found all models except high-gamma are significant after FDR correction (_P_spin < 0.05, one-sided)

Receptor Densities and MEG-Derived Power:

• Close fit between receptor densities and MEG-derived power distributions
• Suggests overlapping spatial topographies of multiple neurotransmitter systems manifest as coherent oscillatory patterns

Dominance Analysis:

• Applied to identify independent variables contributing most to the fit
• Dominance analysis assigns proportion of final fit to each input variable for statistically significant models
• Found that MOR (opioid), H3 (histamine), and α4β2 make large contributions to lower-frequency (theta and alpha) as well as low-gamma power bands31
• Prominence of ionotropic receptors in autoradiography dataset replication32
• Inhibitory, non-monoamine and Gi-coupled receptors more dominant than excitatory, monoamine and Gs-/Gq-coupled receptors, respectively31 (Supplementary Fig. 5a)

Mapping receptors to cognitive function

Brain Mapping: Receptors and Cognitive Function

• Neurosynth meta-analytic task activation maps: derived from multiple cognitive tasks to identify brain regions activated during various functions
• Partial least squares (PLS) analysis: used to identify the relationship between neurotransmitter receptors/transporters and functional activation maps
• Significant latent variable (_P_spin = 0.010, one-tailed) representing 54% of covariance between receptor distributions and Neurosynth-derived cognitive functional activation
• Receptor scores and cognitive scores: reflect how well a brain area exhibits the dominant spatial pattern of receptor distributions and cognitive activations, respectively
• Receptor and cognitive score patterns reveal a sensory-fugal spatial gradient separating limbic, paralimbic, insular cortices from visual and somatosensory cortices
• Cross-validation using distance-dependent method (mean out-of-sample Pearson's r(98) = 0.54, _P_spin = 0.046, one-sided)) demonstrates a link between receptor distributions and cognitive specialization

Receptor loadings: correlation (Pearson's r) between each receptor's distribution across the cortex and PLS-derived scores; contribution of each receptor to the latent variable

• Almost all receptors/transporters have positive loading, with metabotropic dopaminergic and serotonergic receptors having the greatest loadings
• Cognitive processes with large positive loadings are enriched for emotional and affective processes such as 'emotion', 'fear', and 'valence'
• NET (norepinephrine transporter) has stable negative loading, co-varies with functions such as 'fixation', 'planning', and 'skill' in primarily unimodal regions.

Implications: These results demonstrate a direct link between cortex-wide molecular receptor distributions and functional specialization.

Mapping receptors and transporters to disease vulnerability

Neurotransmitter Receptors and Transporters in Diseases and Disorders

Identifying Neurotransmitter Receptors/Transporters:

• Important for developing new therapeutic drugs based on specific disorders
• Relating neurotransmitter receptors/transporters to cortical abnormality patterns across neurological, developmental, and psychiatric disorders

Methods:

• Used datasets from the ENIGMA consortium for 13 disorders:
  • 22q11.2 deletion syndrome
  • Attention deficit hyperactivity disorder (ADHD)
  • Autism spectrum disorder (ASD)
  • Idiopathic generalized epilepsy (IGE)
  • Right and left temporal lobe epilepsy
  • Depression
  • Obsessive-compulsive disorder (OCD)
  • Schizophrenia
  • Bipolar disorder (BD)
  • Obesity
  • Schizotypy
  • Parkinson's disease (PD)
• Fit a multiple regression model to predict each disorder's cortical abnormality pattern from receptor and transporter distributions
• Assessed the significance of each model against an FDR-corrected one-sided spatial autocorrelation-preserving null model
• Evaluated each model using distance-dependent cross-validation

Results:

• Receptor Distributions:
  • Some disorders are more heavily influenced by receptor distribution than others
  • IGE and schizotypy show low and non-significant correspondence with receptor distributions
  • ADHD, autism, and temporal lobe epilepsies show greater correspondence with receptor distributions
• Serotonin Transporter (5-HTT) Distributions:
  • Contribute more to OCD, schizophrenia, and BD profiles than any other receptors
• Mu-Opioid Receptor:
  • Strongest contributor to ADHD cortical abnormality patterns
  • Consistent with findings from animal models
• Unexpected Relationships:
  • In PD, dopamine receptors are not implicated
  • Serotonin receptors do not make large contributions to depression

Conclusion:

• These results present an initial step towards a comprehensive "look-up table" that relates neurotransmitter systems to multiple brain disorders.

Replication using autoradiography

Neurotransmitter Receptor Densities in the Brain

PET Imaging vs Autoradiography:

• PET imaging provides estimates for neurotransmitter receptor densities, but:
  • Densities are acquired from PET imaging alone
  • Quantification methods vary across radioligands, image acquisition protocols, and preprocessing
• Autoradiography is an alternative technique to measure receptor density:
  • Captures local densities at a defined number of postmortem brain sections
  • High cost and labor intensity limit the availability of a complete 3D autoradiography cross-cortex atlas

Authoradiography Dataset:

• Includes 15 neurotransmitter receptors (8 not included in PET dataset)
• Consists of ionotropic and metabotropic receptors, including excitatory glutamate, acetylcholine, and norepinephrine receptors

Similarity Between PET and Authoradiography:

• Receptor similarity is significantly correlated between the two datasets:
  • Pearson's r(1033) = 0.38, P = 6.7 × 10−38, CI = [0.33, 0.44]
• Receptor gradients are also correlated:
  • Pearson's r(44) = 0.51, _P_perm = 0.0001, CI = [0.26, 0.70], two-sided

Authoradiography Receptor Densities:

• Follow similar architectural patterns as PET-derived receptor densities:
  • Receptor similarity is non-significantly greater between structurally connected brain regions (P = 0.19)
  • Significantly correlated with structural connectivity (Pearson's r(329) = 0.39, P = 1.4 × 10−13, CI = [0.30, 0.48])
  • Greater in regions within the same intrinsic network (_P_spin = 0.03)
  • Significantly correlated with functional connectivity (Pearson's r(1033) = 0.21, P = 1.1 × 10−12, CI = [0.16, 0.28])
• Augment structure–function coupling in visual, paracentral, and somatomotor regions

Authoradiography Receptor Densities vs MEG Oscillations:

• AMPA, NMDA, GABAA, and α4β2 (all ionotropic receptors) are most dominant in fitting autoradiography neurotransmitter receptors to MEG power
• Confirms that fast oscillatory dynamics captured by MEG are closely related to fluctuations in neural activity modulated by ionotropic neurotransmitter receptors

Authoradiography Receptor Densities vs Cognitive Functional Activation and Disease Vulnerability:

• Authoradiography-derived receptor densities follow similar topographic gradients linking to Neurosynth-derived functional activations
• PET-derived and autoradiography-derived receptor and cognitive scores are correlated
• Loadings of receptors and cognitive processes are consistent
• Prominent associations between authoradiography-derived receptor densities and cortical abnormality patterns of multiple disorders, including a relationship between the ionotropic glutamate receptor kainate and depression

Sensitivity and robustness analyses

Findings:

• Methodological robustness: analyses repeated using different parcellation resolutions and receptor subsets confirm consistency of results
• Single neurotransmitter's influence: no single neurotransmitter receptor/transporter disproportionately influences receptor similarity, as evidenced by highly correlated original and iteratively removed receptor similarity matrices
• Age effects: age has negligible effect on reported findings, as shown by high correlation between age-regressed and original receptor density and similarity matrices. However, individual subject variability in neurotransmitter systems may not be captured by mean age analysis.

Methods:

• Parcellation resolution: results consistent using parcellations of 200 and 400 cortical regions
• Single receptor/transporter removal: highly correlated original and iteratively removed receptor similarity matrices confirm no disproportionate influence of a single neurotransmitter receptor/transporter on receptor similarity
• Age effects analysis: linear model fit between mean age of scanned participants and receptor density, resulting in age-regressed receptor density and similarity matrices with high correlation to original.

Discussion

Neurotransmitter Receptors and Brain Organization

Key Findings:

• Comprehensive 3D atlas of 19 neurotransmitter receptors and transporters
• Chemoarchitecture is a key layer in the multi-scale organization of the brain
• Neurotransmitter receptor profiles align with structural connectivity and mediate function, including neurophysiological oscillatory dynamics and resting-state hemodynamic functional connectivity
• Overlapping topographic distributions of receptors manifest as patterns of cognitive specialization and disease vulnerability

Background:

• Brain's structural architecture gives rise to its function
• Connectomics model represents brain's structural or functional architectures as regional nodes interconnected by links, with the assumption of homogenous nodes
• Emerging effort to annotate connectome with molecular, cellular, laminar attributes

Neurotransmitter Receptors and Transporters:

• Important molecular annotation for bridging brain structure to function
• Previous initiatives used autoradiography to map receptor densities in human and macaque brains
• Consistent results between autoradiography and PET datasets

Receptor Distribution and Brain Structure/Function:

• Prominent link between receptor distribution and brain structure and function
• Canonical electrophysiological frequency bands can be captured by overlapping topographies of multiple receptors
• Multivariate mapping between receptor profiles and cognitive activations

Implications:

• Serotonergic and dopaminergic receptors underlying patterns of cognitive activation related to affect
• Robust spatial concordance between multiple receptor maps and cortical abnormality profiles across brain disorders
• Key step toward developing therapies for specific syndromes is to reliably map them onto underlying neural systems

Comprehensive Brain Receptor Mapping in Health and Disease

Neurotransmitter Receptor Profiles and Disease Phenotypes

Background:

• Study findings on neurotransmitter receptors and their associations with disease phenotypes
• Some results have preliminary support in literature but not clinically adopted
  • Histamine H3 in PD: Rinne et al. (2002)
  • MOR in ADHD: Sagvolden et al. (2009)
  • D1 and NET in TLE: Costa et al. (2016), Giorgi et al. (2004)

Implications:

• Mapping disease phenotypes to receptor profiles can identify novel targets for pharmacotherapy
• Present report focuses on cortical thinning/thickening but should be expanded in future work
• Building on previous neurochemical and pharmacological causal studies
• Large-scale characterization of receptor systems should be validated and inspire future research

Potential Avenues for Future Research:

• Study changes in receptor architecture in healthy aging, across sexes, and mapping to subcortical structures
  • Dopamine D1 and D2: Seaman et al. (2019)
  • Serotonin transporter and receptor density: Karrer et al. (2019)
  • GABAergic differences: Cuypers et al. (2021)
• Greater understanding of multi-system receptor distributions across age, sex, and within subcortical structures

Methodological Considerations:

• Use of PET images with low spatial resolution and no laminar specificity
  • Replicated using autoradiography dataset, but comprehensive atlas needed for full understanding
• Normalized spatial distributions to focus on relative topographies
• Accounted for spatial dependencies using autocorrelation-preserving null models
• Restricted analyses to cortex, obscuring subcortical neuromodulatory systems
• Direct comparison between PET and autoradiography datasets not possible due to missing receptors in the PET datasets.

Methods

Data and Code Availability:

• Code and data for the analyses are accessible at github.com/netneurolab/hansen_receptors
• Volumetric PET images can be found in neuromaps at github.com/netneurolab/neuromaps, where they can be easily converted between template spaces (Markello et al., 2022, Nature Methods)

PET data acquisition

PET Images and Neurotransmitter Receptors/Transporters:

• Volumetric PET images collected for 19 different neurotransmitter receptors and transporters across multiple studies (n = 1,238; 718 males, 520 females)
• Protect patient confidentiality by averaging individual participant maps within studies before sharing
• Details: Table 1, Supplementary Table 3
• Only healthy participants were scanned using best practice imaging protocols recommended for each radioligand (ref.56)
• Images registered to MNI-ICBM 152 template and parcellated into 100, 200, or 400 regions according to Schaefer atlas (ref.12)
• Receptors/transporters with more than one mean image of the same tracer combined using weighted average (Supplementary Fig.13a)
• Each tracer map corresponding to each receptor/transporter z-scored and concatenated into a final region by receptor matrix of relative densities
• Comparisons between tracers shown for some neurotransmitter receptors/transporters: 5-HT1A, 5-HT1B, 5-HT2A, 5-HTT, CB1, D2, DAT, GABAA, MOR, and NET (Supplementary Fig.13b)
• Specific notes: 5-HTT and GABAA involve comparisons between the same tracers (DASB and flumazenil, respectively), but one map is converted to density using autoradiography data (ref.9)

Serotonin System Mapping: High-resolution in vivo atlas of human brain's serotonin system using multiple tracers: FLB457, raclopride, and fallypride.

• FLB457 is a reliable SERT tracer for mapping serotonin densities in the cortex.
• Raclopride has unreliable binding in the cortex, but its comparison to FLB457 and fallypride is presented for completeness.

Dopamine Receptor Mapping: Carfentanil (MOR) map collated from PET Turku Centre database due to overlap with an alternative map; no combination of tracers into a single mean map.

• Importance of choosing appropriate tracers for accurate mapping.

Synaptic Density Measurement: [11C]UCB-J, a PET tracer that binds to the synaptic vesicle glycoprotein 2A (SV2A), used to measure synapse density in 76 healthy adults.

• Data collected using HRRT PET camera for 90 minutes after injection and modeled using SRTM2 with centrum semiovale as reference.
• Group-averaged map presented in ref. [105].

Autoradiography receptor data acquisition

Receptor Autoradiography Data

• Originally acquired as described in Zilles & Palomero-Gallagher (2017)
• Fifteen neurotransmitter receptor densities across 44 cytoarchitectonically defined areas in three postmortem brains (age range: 72–77 years, two males)

Receptor Densities Included:

• See Supplementary Table 1 for a complete list
• See Supplementary Table 2 in Zilles & Palomero-Gallagher (2017) for originally reported densities
• Access machine-readable Python numpy files at https://github.com/AlGoulas/receptor_principles

Comparison to PET Data Analyses:

• Manually created region-to-region mapping between the 44 autoradiography cortical areas and the 50 left hemisphere Schaefer regions
• Four regions in Schaefer atlas did not have a suitable mapping to the autoradiography atlas, resulting in conversion to 46 Schaefer left hemisphere regions
• Concatenated and z-scored receptor densities to create a single map of receptor densities across the cortex.

Structural and functional data acquisition

Data Acquisition and Pre-Processing

Participants:

• 326 unrelated participants (age range: 22–35 years) from the HCP S900 release
• 145 males

Functional MRI Data:

• Two scans on day 1 and two scans on day 2
• Each scan approximately 15 minutes long
• TR = 720 ms

Diffusion-Weighted Imaging (DWI) Data:

• Available for all participants

Pre-Processing:

• All structural and functional MRI data pre-processed using HCP minimal pre-processing pipelines
• Detailed information on data acquisition and pre-processing available elsewhere

Structural network reconstruction

Preprocessing DWI Data using MRtrix3

• Fiber orientation distributions generated using multi-shell, multi-tissue constrained spherical deconvolution algorithm from MRtrix:
  • Estimated response function without co-registered T1 image (Dhollander et al., 2016; Jeurissen et al., 2014)
• White matter edges reconstructed using probabilistic streamline tractography based on generated fiber orientation distributions:
  • Improved accuracy with 2nd order integration (Tournier, Calamante & Connelly, 2010)
• Tract weights optimized by estimating appropriate cross-section multiplier for each streamline:
  • SIFT2 procedure (Smith et al., 2015)
• Connectivity matrix built for each participant using 100-region Schaefer parcellation:
  • Local-global parcellation of human cerebral cortex (Schaefer et al., 2018)

Group Consensus Binary Network Construction

• Preserve density and edge-length distributions of individual connectomes:
  • Distance-dependent consensus thresholds (Betzel et al., 2019)
• Assign weights to edges in group consensus network:
  • Average log-transformed streamline count across participants
• Scale edge weights between 0 and 1.

Functional network reconstruction

Preprocessing Steps for Functional MRI Data:

• Correction of gradient non-linearity, head motion, and geometric distortions using scan pairs with opposite phase encoding directions106
• Co-registration of corrected images to T1w structural MR images
• Brain extraction and normalization of whole brain intensity
• High-pass filtering (>2,000s full width at half maximum (FWHM)) to correct for scanner drifts
• Removing additional noise using the ICA-FIX process106,114
• Parcellation of pre-processed time-series into 100 cortical brain regions according to the Schaefer atlas12
• Construction of functional connectivity matrices as Pearson correlation coefficient between pairs of regional time series for each scan
• Group-average functional connectivity matrix construction by taking the mean of all individual and scan functional connectivity matrices.

Structure–function coupling

Brain Connectivity Metrics:

• Communicability: measure of diffusion between brain regions based on weighted average of all walks and paths between them.
• Calculated as: simple linear regression model fits regional communicability to functional connectivity.
• Represents diffusive communication in complex networks (Crofts & Higham, 2009; Estrada & Hatano, 2008).
• Bridges structure and function (Seguin et al., 2022).

Structure–function Coupling:

• Defined as the adjusted _R_2 of a simple linear regression model that fits:
  • Regional communicability to regional functional connectivity.
• Significance assessed against null distribution of adjusted _R_2 from a model with rotated receptor similarity vector (10,000 repetitions).
• Ensures robustness of increased _R_2 when receptor information is included in the model.

Receptor Similarity:

• Included as an additional independent variable in the structure–function coupling model.
• Represents similarity between a region of interest and every other region.
• Significance assessed against null distribution to ensure robustness of results.

MEG power

MEG Data Acquisition and Preprocessing

• Six-minute resting-state MEG time series: acquired from HCP (S1200 release) for 33 unrelated participants
• Participant demographics: age range 22–35 years, 17 males
• Complete MEG acquisition protocols can be found in HCP S1200 Release Manual

MEG Data Processing

• Power spectrum computed at vertex level across six frequency bands: delta (2–4 Hz), theta (5–7 Hz), alpha (8–12 Hz), beta (15–29 Hz), low gamma (30–59 Hz), high gamma (60–90 Hz) using Brainstorm
• Pre-processing:
  • Apply notch filters at 60, 120, 180, 240 and 300 Hz to remove artifacts
  • High-pass filter at 0.3 Hz to remove slow-wave and DC offset artifacts
• Source estimation:
  • Preprocessed sensor-level data used to obtain source activity on HCP's fsLR4k cortex surface for each participant
  • Head models computed using overlapping spheres, and data/noise covariance matrices estimated from MEG and noise recordings
  • Brainstorm's LCMV beamformers method applied to obtain source activity
• Power spectrum density (PSD) estimation:
  • Welch's method used with overlapping windows of length 4 seconds and 50% overlap
• Source-level power data parcellation: divided into 100 cortical regions for each frequency band

ENIGMA cortical abnormality maps

ENIGMA (Enhancing Neuroimaging Genetics through Meta-Analysis) Consortium:

• Data-sharing initiative for standardized image acquisition and processing pipelines
• Disorder maps from ENIGMA consortium and Enigma toolbox: https://github.com/MICA-MNI/ENIGMA (ref. 118)
• Patterns of cortical abnormality collected for various disorders: 22q11.2 deletion syndrome, ADHD, ASD, idiopathic generalized epilepsy (right temporal lobe, left temporal lobe), depression, OCD, schizophrenia, BD, obesity, schizotypy, and PD.

Disorders and Cortical Abnormalities:

• 22q11.2 deletion syndrome: Large-scale mapping of cortical alterations (ref. 119)
• ADHD: Brain imaging of the cortex in ADHD (ref. [120])
• ASD: Cortical and subcortical brain morphometry differences between patients with ASD and healthy individuals (ref. [121])
• Idiopathic generalized epilepsy: Structural brain abnormalities in the common epilepsies (ref. [122])
• Depression: Cortical abnormalities in adults and adolescents with major depression (ref. [123])
• OCD: Cortical abnormalities associated with pediatric and adult OCD (ref. [124])
• Schizophrenia: Cortical brain abnormalities in schizophrenia patients (ref. [125])
• BD: Cortical abnormalities in BD patients (ref. [126])
• Obesity: Brain structural abnormalities in obesity (ref. [127])
• Schizotypy: Cortical and subcortical neuroanatomical signatures of schizotypy (ref. [128])
• PD: International multicenter analysis of brain structure across clinical stages of PD (ref. [129])

Cortical Abnormalities:

• Decreases in cortical thickness: most disorders
• Increases in cortical thickness: 22q, ASD, and schizotypy

Data Collection:

• Over 21,000 scanned patients and almost 26,000 controls
• Adult patients (except for ASD) following identical processing protocols
• Values are z-scored effect sizes (Cohen’s d) of cortical thickness in patient populations versus healthy controls
• Native representation: Desikan–Killiany atlas (68 cortical regions) (ref. [130])

Imaging and Processing Protocols: [http://enigma.ini.usc.edu/protocols/]

Dominance analysis

Dominance Analysis

• Purpose: Determine relative contribution of independent variables to overall fit (adjusted _R_2) of multiple linear regression model
• Methodology:
  • Fit the same regression model on every combination of input variables (2_p_ - 1 submodels for a model with p input variables)
  • Calculate total dominance: average of the relative increase in _R_2 when adding a single input variable to a submodel, across all combinations
  • Sum of dominance equals total adjusted _R_2 of complete model
  • Normalize by total fit (_R_2) for comparability within and across models
• Advantages:
  • Accounts for predictor–predictor interactions
  • Interpretable
  • Partitions the total effect size across predictors

Cognitive meta-analytic activation

Neurosynth: Probabilistic Measures of Association between Voxels and Cognitive Processes

Background:

• Neurosynth is a meta-analytic tool synthesizing results from 15,000 published functional MRI studies
• Focuses on high-frequency keywords like 'pain' and 'attention' with voxel coordinates
• Provides probabilistic measures of association between voxels and cognitive processes

Data Collection:

• Searches for cognitive and behavioral terms in Neurosynth (Cognitive Atlas)
  • Umbrella terms: attention, emotion
  • Specific cognitive processes: visual attention, episodic memory
  • Behaviors: eating, sleeping
  • Emotional states: fear, anxiety
• Terms selected from Cognitive Atlas, a public ontology of cognitive science
• Coordinates reported by Neurosynth are parcellated according to the Schaefer-100 atlas and z-scored

Interpretation:

• Probabilistic measure reported by Neurosynth: quantitative representation of how regional fluctuations in activity relate to psychological processes
• Full list of cognitive processes is available in Supplementary Table 2.

Partial least squares analysis

Neurotransmitter Receptor Distributions and Functional Activation Analysis using Partial Least Squares (PLS)

Technique Overview:

• Unsupervised multivariate statistical technique: PLS
• Decomposes two datasets into orthogonal sets of latent variables with maximum covariance

Latent Variables:

• Receptor weights, cognitive weights and a singular value representing the covariance between receptor distributions and functional activations

Scoring Process:

1. Project original data onto respective weights
2. Assign brain regions receptor and cognitive scores
3. Compute receptor loadings: Pearson's correlation between receptor densities and receptor scores
4. Compute cognitive loadings: Pearson's correlation between functional activations and cognitive scores

Significance Analysis:

• Assess significance of latent variable based on singular value against spin-test
• Only the first significant latent variable was analyzed further
• Cross-validate empirical correlation between receptor and cognitive scores using distance-dependent validation

Limitations:

• Does not establish causal relationships between receptors and cognition
• Does not make specific univariate associations between receptors and cognitive function
• Does not preclude existence of additional relationships between receptors and cognitive function.

Distance-dependent cross-validation

Assessment of Multilinear Models

Robustness Assessment:

• Cross-validated using a distance-dependent method to assess the robustness of each multilinear model
• Applied to:
  • Every multilinear regression model (Figs. 3c, 4 and 6)
  • The PLS model (Fig. 5)
• Procedure:
  • Selected the 75% closest regions as the training set
  • Remaining 25% of brain regions as the test set
  • Stratification procedure to minimize dependence due to spatial autocorrelation

Multilinear Regression Models:

• Model fit on the training set
• Predicted test set output variable (regional functional connectivity, MEG power or disorder maps) was correlated to the empirical test set values
• Distribution of Pearson's correlations between predicted and empirical variables across all repetitions found in Supplementary Figs. 2, 3 and 7

PLS Analysis:

• Model fit on the training set
• Weights were projected onto the test set to calculate predicted receptor and cognitive scores
• Training and test sets defined as described above
• Correlation between receptor and cognitive score was calculated in both the training and test set
• Significance of mean out-of-sample correlation assessed against a permuted null model (Fig. 5d)

Null models

Spatial Autocorrelation-Preserving Permutation Tests

Assessing Statistical Significance of Associations Across Brain Regions:

• Termed "spin tests"
• Used to assess statistical significance of associations across brain regions
• Designed by Alexander-Bloch et al. (2018)24
• Implemented by Markello and Misic (2021)25

Creating a Surface-Based Representation:

• Defined spatial coordinates for each parcel on the FreeSurfer fsaverage surface
• Rotated parcel coordinates and reassigned parcels to nearest rotated parcel (10,000 repetitions)
• Handled medial wall cases separately

Performing Spin Test:

• Not applied to autradiography data
• Permutation test used instead

Testing Receptor Similarity in Connected vs. Unconnected Regions:

• Generates a null structural connectome preserving density, edge length, and degree distributions
• Swaps edges within distance bins to create rewired networks
• Computes the difference between mean receptor similarity of unconnected and connected edges
• Compares this difference to a null distribution computed on the rewired networks

Reporting Summary

Research Design Information

Find further details in the Nature Research Reporting Summary accompanying this article.

Data availability

Data Availability:

• Volumetric PET images, including receptor images and synaptic density: github.com/netneurolab/hansen_receptors
• Neuromaps conversion between template spaces: github.com/netneurolab/neuromaps
• Autoradiography data: Supplementary Table 2 of ref. 6
• HCP dataset: db.humanconnectome.org/
• Neurosynth data: neurosynth.org/
• ENIGMA datasets: ENIGMA consortium and the ENIGMA Toolbox (github.com/MICA-MNI/ENIGMA)
• Parcellation atlases: netneurotools (github.com/netneurolab/netneurotools)
  • Schaefer-100 and Desikan–Killiany atlas

Supplementary information

Supplementary Materials

• Figures 1–13 and Tables 1 and 2 can be found at this link
• Reporting Summary and Supplementary Table 3 are available at this link
• Excel file contains methodological details for each PET tracer.