Amplification and Suppression of Distinct Brainwide Activity Patterns by Catecholamines

Ruud L. van den Brink, Sander Nieuwenhuis, Tobias H. DonnerJournal of Neuroscience 2018 · DOI

TL;DR

This study challenges the long-held assumption that catecholamines (like dopamine and norepinephrine) affect brain activity uniformly. By using fMRI and a novel decomposition technique, researchers found that these neuromodulators can both amplify and suppress specific patterns of brain activity, with these effects being spatially organized and linked to the distribution of different catecholamine receptors in the brain.

What Problem Does This Paper Address?

For decades, the prevailing model of catecholamine function in the brain has been one of widespread, homogeneous influence. This view stems from the fact that these neurotransmitters are released from brainstem nuclei with extensive projections throughout the forebrain. However, this model doesn't fully account for the fact that the brain possesses a diverse array of catecholamine receptors, each with distinct functional roles and varying densities across different brain regions. This paper addresses the gap between the presumed homogeneous effects of catecholamines and the known heterogeneous nature of their receptor systems, questioning whether their impact on brain-wide network dynamics is truly uniform or, instead, spatially specific.

Methods

The core of this study's methodological innovation lies in its use of generalized eigenvalue decomposition applied to resting-state fMRI data. This technique allows for the identification of spatial patterns of correlated brain activity (termed "spatial modes") that are differentially expressed between two conditions – in this case, placebo versus atomoxetine (a drug that increases catecholamine levels).

Here's a breakdown of the key steps:

  • Pharmacological Intervention: Participants received either atomoxetine or a placebo in a double-blind, crossover design.
  • fMRI Data Acquisition: Resting-state fMRI scans were collected under both conditions.
  • Generalized Eigenvalue Decomposition: This is the central analytical tool. It was applied to the covariance matrices of fMRI signal fluctuations from the placebo and atomoxetine conditions. This decomposition identifies spatial modes that are either amplified or suppressed by atomoxetine.
  • Spatial Mode Imaging: The identified spatial modes were then projected onto voxel-level fMRI data to visualize their precise neuroanatomical distribution.
  • Correlation with Gene Expression: The spatial maps of these modes were correlated with publicly available gene expression data for various catecholamine receptors (e.g., alpha-1, alpha-2, beta norepinephrine receptors, D1-like, D2-like dopamine receptors) from postmortem brain samples.
  • Control Analyses: A series of rigorous control analyses were performed to ensure the robustness of the findings. These included:
  • Cross-validation: Assessing the reliability of spatial modes across different data segments.
  • Parcellation Scheme Independence: Repeating analyses with different brain atlases (anatomical vs. functional).
  • Global Signal Regression: Ruling out artifacts related to global brain signal fluctuations.
  • Peripheral Physiology Control: Ensuring that changes in heart and breath rate didn't confound the results.
  • Session-Related Effects Control: Using pre-pill fMRI data to distinguish drug effects from session differences.
  • Comparison with ICA: Validating the spatial modes against established resting-state networks derived from Independent Component Analysis (ICA).

Datasets

  • Human fMRI Data: Resting-state fMRI data from 24 healthy participants (19 females, 5 males) collected during placebo and atomoxetine conditions. This dataset was reanalyzed from a previous study by the same research group (van den Brink et al., 2016).
  • Allen Brain Atlas: Postmortem human brain gene expression data, specifically focusing on the expression patterns of catecholamine receptor genes. This dataset was used to link the identified spatial modes to the underlying neurochemical architecture.

Key Insights and Ideas

  • Beyond Homogeneity: The most significant insight is the direct challenge to the idea of homogeneous catecholaminergic effects. The brain doesn't just get a general "boost" from these neuromodulators; specific circuits are differentially affected.
  • Opposing Effects: Catecholamines can both amplify and suppress brain activity patterns. This duality is crucial for understanding their complex roles.
  • Receptor-Driven Specificity: The spatial patterns of these amplified and suppressed activity are not random. They strongly correlate with the known distribution of specific catecholamine receptor genes in the brain. This provides a clear mechanistic link.
  • D2 Dopamine Receptors & Amplification: The spatial mode showing increased brain activity under atomoxetine was linked to the expression of D2-like dopamine receptors. This suggests a role for these receptors in disinhibition or enhanced signal transmission.
  • Alpha-1 Norepinephrine Receptors & Suppression: Conversely, the spatial mode showing decreased brain activity was associated with alpha-1 norepinephrine receptors, hinting at their role in dampening neural interactions.
  • Beta Receptors & Bipolar Effects: Beta norepinephrine receptors showed associations with both amplified and suppressed patterns, but with opposite signs, underscoring their complex modulatory roles.
  • Spatial Modes as a Novel Tool: The generalized eigenvalue decomposition approach proved effective in uncovering these spatially specific effects, going beyond traditional methods that might average out such fine-grained differences.

Key Results

  • Two Distinct Spatial Modes: The analysis identified two primary spatial patterns of fMRI signal fluctuations that were significantly modulated by atomoxetine.
  • Mode 1 (Atomoxetine > Placebo): This pattern showed amplified cofluctuations and was associated with the spatial distribution of D2-like dopamine receptors.
  • Mode 2 (Placebo > Atomoxetine): This pattern showed suppressed cofluctuations and was associated with the spatial distribution of α₁ norepinephrine receptors.
  • Beta Receptor Association: Both modes were associated with β norepinephrine receptors, but with opposite correlation signs, suggesting a complex role for these receptors.
  • Reliable Discrimination: The identified spatial modes reliably discriminated between the atomoxetine and placebo conditions, even when analyzed on short segments of fMRI data (ROC index of 0.62 for both modes, p < 0.002).
  • Specificity to Catecholamines: The spatial correlations between the identified modes and gene expression were specific to catecholamine receptors, showing no significant association with acetylcholine or NMDA receptors.
  • Robustness Across Atlases: The findings were consistent when using different brain parcellation schemes (anatomical AAL atlas and functional Craddock atlas).

Limitations and Future Work

  • Parcellation Dependence: While robust across two schemes, the spatial modes might differ with radically different parcellation densities.
  • Two-Condition Limit: The generalized eigenvalue decomposition approach is primarily suited for comparing two conditions, limiting its application to longitudinal designs without modification.
  • Focus on First Mode: The study focused on the first spatial modes due to their dominant explanatory power and orthogonality constraints. However, other "signal" modes also showed statistically significant differences, suggesting a richer landscape of catecholaminergic effects that could be further explored.
  • Future Directions: The authors suggest that future work could test the hypothesis of a shift towards goal-oriented stimulus processing, as hinted at by the lateralization shifts observed. Further research could also explore the precise mechanistic origins of the observed correlation changes and the role of different receptor subtypes in more detail. The findings also provide crucial constraints for developing more realistic computational models of neuromodulatory effects.

Summary updated

Abstract

The widely projecting catecholaminergic (norepinephrine and dopamine) neurotransmitter systems profoundly shape the state of neuronal networks in the forebrain. Current models posit that the effects of catecholaminergic modulation on network dynamics are homogeneous across the brain. However, the brain is equipped with a variety of catecholamine receptors with distinct functional effects and heterogeneous density across brain regions. Consequently, catecholaminergic effects on brainwide network dynamics might be more spatially specific than assumed. We tested this idea through the analysis of fMRI measurements performed in humans (19 females, 5 males) at "rest" under pharmacological (atomoxetine-induced) elevation of catecholamine levels. We used a linear decomposition technique to identify spatial patterns of correlated fMRI signal fluctuations that were either increased or decreased by atomoxetine. This yielded two distinct spatial patterns, each expressing reliable and specific drug effects. The spatial structure of both fluctuation patterns resembled the spatial distribution of the expression of catecholamine receptor genes: α<sub>1</sub> norepinephrine receptors (for the fluctuation pattern: placebo > atomoxetine), D<sub>2</sub>-like dopamine receptors (pattern: atomoxetine > placebo), and β norepinephrine receptors (for both patterns, with correlations of opposite sign). We conclude that catecholaminergic effects on the forebrain are spatially more structured than traditionally assumed and at least in part explained by the heterogeneous distribution of various catecholamine receptors. Our findings link catecholaminergic effects on large-scale brain networks to low-level characteristics of the underlying neurotransmitter systems. They also provide key constraints for the development of realistic models of neuromodulatory effects on large-scale brain network dynamics.<b>SIGNIFICANCE STATEMENT</b> The catecholamines norepinephrine and dopamine are an important class of modulatory neurotransmitters. Because of the widespread and diffuse release of these neuromodulators, it has commonly been assumed that their effects on neural interactions are homogeneous across the brain. Here, we present results from the human brain that challenge this view. We pharmacologically increased catecholamine levels and imaged the effects on the spontaneous covariations between brainwide fMRI signals at "rest." We identified two distinct spatial patterns of covariations: one that was amplified and another that was suppressed by catecholamines. Each pattern was associated with the heterogeneous spatial distribution of the expression of distinct catecholamine receptor genes. Our results provide novel insights into the catecholaminergic modulation of large-scale human brain dynamics.
BibTeX
@article{brink2018amplification,
  author = {Ruud L. van den Brink and Sander Nieuwenhuis and Tobias H. Donner},
  title = {{Amplification and Suppression of Distinct Brainwide Activity Patterns by Catecholamines}},
  journal = {Journal of Neuroscience},
  year = {2018},
  doi = {10.1523/jneurosci.0514-18.2018},
}
Figure