henhaobio
Sign in

Nature NeuroscienceNeurosciencePeer-reviewed journal article

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

Justine Y. HanseniD, Golia ShafieiiD, Ross D. MarkelloiD, Kelly A. SmartiD, Sylvia M. L. CoxiD, Martin NørgaardiD, Vincent BeliveauiD, Yanjun WuiD, Jean‐Dominique GallezotiD, Étienne AumontiD, Stijn ServaesiD, Stephanie G. ScalaiD, Jonathan M. DuBoisiD, Gabriel WainsteiniD, Gleb BezginiD, Thomas FunckiD, Taylor W. SchmitziD, R. Nathan SprengiD, Marian GaloviciD, Matthias J. KoeppiD, John Sidney DuncaniD, Jonathan ColesiD, Tim D. FryeriD, Franklin I. AigbirhioiD, Colm J. McGinnityiD, , Bratislav MišićiD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Neurotransmitter receptors support the propagation of signals in the human brain. How receptor systems are situated within macroscale neuroanatomy and how they shape emergent function remains poorly understood, and there exists no comprehensive atlas of receptors. Here we collate positron emission tomography data from >1 200 healthy individuals to construct a whole-brain 3-D normative atlas of 19 receptors and transporters across 9 different neurotransmitter systems. We find that receptor profiles align with structural connectivity and mediate function, including neurophysiological oscillatory dynamics and resting state hemodynamic functional connectivity. Using the Neurosynth cognitive atlas, we uncover a topographic gradient of overlapping receptor distributions that separates extrinsic and intrinsic psychological processes. Finally, we find both expected and novel associations between receptor distributions and cortical thinning patterns across 13 disorders. We replicate all findings in an independently collected autoradiography dataset. This work demonstrates how chemoarchitecture shapes brain structure and function, providing a new direction for studying multi-scale brain organization.

In-depth summary

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 2 (11 Jan 2022) · CC BY 4.0

Neurotransmitter receptor/transporter PET maps (19 targets, 1239 healthy participants) align with connectomes, explain MEG power, and track cognitive gradients and cortical thinning in 13 disorders.

Connectome models typically assume homogeneous nodes, and comprehensive open receptor data were lacking: the main autoradiography set covers only 44 cortical areas from three post-mortem brains, while radiation risk discourages multi-tracer scans per person and cost keeps PET control cohorts small. Pooling PET maps from many groups into an openly shared, harmonized cortical atlas lets chemoarchitecture be tested against connectivity, oscillations, cognition, and disease.

Key results

  1. ShownReceptor similarity (correlation of regional receptor profiles) is higher between structurally connected pairs and within intrinsic networks, tracks functional connectivity (FC) beyond spatial proximity, and improves regional structure-function coupling models.

    Similarity higher for connected pairs than density-, degree- and edge length-preserving surrogates (p = 0.0001, 10 000 repetitions) and within Yeo-Krienen 7-network systems (pspin = 0.016); correlation with FC net of Euclidean distance r = 0.24 (p = 7 × 10^−33); receptor term raised coupling chiefly in inferior temporal and dorsolateral prefrontal cortex (F > Fcritical). · Receptor distributions reflect structural and…; Structure-function coupling; Null models; Figure legends

  2. ShownReceptor/transporter densities closely fit the cortical distribution of resting magnetoencephalography (MEG) power across six frequency bands; GABAA and mu-opioid receptor (MOR) maps are among the most dominant predictors, and ionotropic dominance recurs with autoradiography.

    MEG from 33 Human Connectome Project participants; six-band fits adjusted R^2 0.85 to 0.94. Against spin nulls, MOR showed high dominance for theta and alpha, and GABAA was significant for alpha, low gamma and high gamma (pspin < 0.05). With 15 autoradiography receptors, AMPA, NMDA and GABAA were significantly dominant; α4β2 ranked high but was not significant. · Receptor profiles shape oscillatory neural…; Replication using autoradiography; MEG power; Figure legends

  3. ShownOne significant partial least squares (PLS) latent variable couples receptor densities with Neurosynth meta-analytic activation for 123 cognitive terms, forming a sensory-fugal gradient that separates affective and interoceptive processes from sensory, attentional, and action-related ones.

    pspin = 0.036 (10 000 repetitions), 44% of covariance; receptor-cognitive score r = 0.79 in-sample, mean training 0.86, mean test 0.29 (figure legend: 0.32). Most negative loadings: fear, emotion, valence with MOR, D1, 5-HT1A; most positive: visual attention, spatial attention, action. Scores separate limbic, paralimbic and insular from visual and somatosensory cortex. · Mapping receptors to cognitive function; Cognitive meta-analytic activation; Partial least squares analysis; Figure legends

  4. SuggestedReceptor/transporter maps fit cortical thinning patterns of 13 ENIGMA disorders to varying degrees and nominate receptor-disorder pairs, such as the serotonin transporter (5-HTT) with OCD, schizophrenia and bipolar disorder, and MOR with ADHD; some expected links did not emerge.

    Fits (adjusted R^2) between 0.17 and 0.76 using maps from over 21 000 patients vs almost 26 000 controls; stronger for ADHD, autism, temporal lobe epilepsies, weaker for idiopathic generalized epilepsy, OCD, schizotypy. 5-HTT was not significantly dominant for depression, and dopamine receptors were not implicated in Parkinson's disease (cortex-only analysis). · Mapping receptors and transporters to disease…; ENIGMA cortical thinning maps; Figure legends

  5. ShownMost analyses replicate with an independent post-mortem autoradiography set of 15 receptors, and PET receptor similarity is robust to dropping any single map, finer parcellations, and regressing out tracer-map mean age.

    Autoradiography vs PET similarity r = 0.32 (p = 1.9 × 10^−14); autoradiography similarity higher for connected pairs (p = 0.04), correlated with FC (r = 0.34, p = 9.4 × 10^−16), not significantly higher within networks (pspin = 0.11). Leave-one-out similarity r > 0.98; age-regressed density r = 0.81, similarity r = 0.98; 114- and 219-region parcellations agree. · Replication using autoradiography; Sensitivity and robustness analyses; Figure legends

Methods, data and reproducibility

Group-average PET maps of 19 receptors/transporters were parcellated (68 Desikan-Killiany regions) and z-scored; receptor similarity and densities were related to connectomes, MEG power, Neurosynth activation (partial least squares), and ENIGMA thinning via regression with dominance analysis, spin/rewiring nulls, and distance-dependent cross-validation.

  • Group-average PET maps of 19 receptors/transporters spanning 9 neurotransmitter systems, pooled across studies and parcellated to 68 cortical regions · human · n = 1239 healthy participants
  • Post-mortem receptor autoradiography densities for 15 receptors in 44 cytoarchitectonic areas, remapped to 33 left-hemisphere Desikan-Killiany regions · human · n = three brains (age 72 to 77, 2 males)
  • Lausanne diffusion spectrum imaging (DSI) and resting-state fMRI used to build group-consensus structural (24.6% density) and functional connectomes · human · n = 70 healthy young adults (27 females); 69 with fMRI · Zenodo record 2872624
  • Human Connectome Project S1200 resting eyes-open MEG (6 min), source-level power in six bands from delta to high gamma · human · n = 33 unrelated participants (age 22 to 35, 17 males)
  • Neurosynth meta-analytic association maps (from more than 15 000 fMRI studies) for terms shared with the Cognitive Atlas · human · n = 123 cognitive terms
  • ENIGMA cortical thickness effect-size (Cohen's d) maps for 13 neurological, psychiatric, and neurodevelopmental disorders (adult patients) · human · n = over 21 000 patients vs almost 26 000 controls
Methods & tools
PET tracers incl. FLB457, DASB, flumazenil, carfentanil, FEOBV, ABP688, P943, Receptor autoradiography (layer-averaged densities), Desikan-Killiany parcellation (68, 114, 219 regions), Connectome Mapping Toolkit (deterministic streamline tractography), bctpy (shortest path length), Brainstorm (linearly constrained minimum variance beamformer; Welch power spectra), Neurosynth and Cognitive Atlas, ENIGMA toolbox, Dominance analysis (dominance-analysis package), Partial least squares (PLS) with bootstrap loadings, Spin test and degree/edge-length-preserving rewiring null models, neuromaps
Software & environment
MNI-ICBM 152 nonlinear 2009c asymmetric template; Desikan-Killiany 68/114/219-region parcellations; Brainstorm on HCP fsLR4k surface; bctpy; Connectome Mapping Toolkit; FreeSurfer fsaverage sphere for spin nulls.
Experimental conditions
Lausanne MRI: 3-T Siemens Trio, 32-channel coil; diffusion 128 volumes, max b-value 8 000 s/mm^2, 2.2 × 2.2 × 3.0 mm; BOLD EPI TR 1 920 ms, 280 volumes. MEG: 6-minute eyes-open rest (HCP). PET scanners listed only in Table S3; compute hardware not stated.
Reproducibility gaps
Only study-level group-average PET maps are shared (individual scans withheld for confidentiality); three tracer maps were unpublished when shared, with PI contacts in Table S3; only mean age per tracer map available; no software versions stated.
Limitations (6)
  • PET captures tracer uptake only at coarse spatial resolution and without laminar specificity; despite autoradiography and finer-parcellation replications, a laminar-resolved receptor atlas is needed to fully understand regional effects. · stated by the authors
  • Tracer maps were acquired in different participants, scanners, and sites with radioligand-specific protocols, so analyses rely on z-scored relative topographies rather than absolute receptor densities. · stated by the authors
  • Analyses cover cortex only, owing to data availability and known cortical-subcortical differences in tracer uptake; this likely explains why dopamine receptors were not implicated in Parkinson's disease. · stated by the authors
  • Disease mapping uses cortical thinning alone, which may miss primary pathophysiology (5-HTT was not dominant for depression); age and pathology effects on receptor density remain open, and only per-map mean ages were available. · stated by the authors
  • Regressions fit 19 intercorrelated predictors to 68 regions; quoted adjusted R^2 values are in-sample, cross-validated accuracy for MEG and thinning models is only plotted, and reporting conflicts (training split 75% in Methods vs 25% in legends; PLS test r 0.29 vs 0.32). · noted in this summary
  • Autoradiography replication is partial despite the abstract's claim that all findings replicate: it spans 33 left-hemisphere regions from three brains aged 72 to 77, and within-network similarity was not significant there (pspin = 0.11). · noted in this summary

Read this if you need an openly shared, cortex-wide PET atlas of neurotransmitter receptors to annotate connectomes, MEG dynamics, cognitive maps, or disorder-specific cortical thinning.

Citations
799
FWCI
72.81
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISHänisch '23Ceballos '24Lotter '26Hansen '26Goulas '21Hansen '22
77% semantic matchNeuroscience
Benjamin Hänisch, Justine Y. Hansen, Boris C. Bernhardt+4 more·eLife·Jul 13, 2023· 25 citations