Dopamine Signaling Modulates the Stability and Integration of Intrinsic Brain Networks

Abstract

Dopaminergic projections are hypothesized to stabilize neural signaling and neural representations, but how they shape regional information processing and large-scale network interactions remains unclear. Here we investigated effects of lowered dopamine levels on within-region temporal signal variability (measured by sample entropy) and between-region functional connectivity (measured by pairwise temporal correlations) in the healthy brain at rest. The acute phenylalanine and tyrosine depletion (APTD) method was used to decrease dopamine synthesis in 51 healthy participants who underwent resting-state functional MRI (fMRI) scanning. Functional connectivity and regional signal variability were estimated for each participant. Multivariate partial least squares (PLS) analysis was used to statistically assess changes in signal variability following APTD as compared with the balanced control treatment. The analysis captured a pattern of increased regional signal variability following dopamine depletion. Changes in hemodynamic signal variability were concomitant with changes in functional connectivity, such that nodes with greatest increase in signal variability following dopamine depletion also experienced greatest decrease in functional connectivity. Our results suggest that dopamine may act to stabilize neural signaling, particularly in networks related to motor function and orienting attention towards behaviorally-relevant stimuli. Moreover, dopamine-dependent signal variability is critically associated with functional embedding of individual areas in large-scale networks.

Figure 1.

Sample entropy of a time series | (a) An example of a BOLD signal is shown, where the x-axis is time and the y-axis is the amplitude. Signal variability is calculated using sample entropy analysis. Sample entropy (SE) measures the conditional probability that any two sequences of data points with length $\mathit{m}+1$ will be similar to one another under the condition that they were similar for the first $\mathit{m}$ points. The similarity criterion $\mathit{r}$ represents the tolerance of algorithm to accept matches in the time series. (b) An example of a BOLD signal in its original form (left). The same signal, with the time points reordered by amplitude (right). (c) Standard deviation of the signal is the same for both the original and reordered signal; however, sample entropy of the reordered signal drastically decreases compared with sample entropy of the original signal.

Figure 2.

Dopamine depletion increases signal variability | (a) PLS analysis identified a significant contrast between patterns of signal variability in depletion (APTD) versus non-depletion (BAL) conditions (permuted P =0.014). (b) The change in signal variability of each node is given by a bootstrap ratio for that node: such that a positive bootstrap ratio shows increase in signal variability of the node following dopamine depletion, while a negative bootstrap ratio shows the opposite. Bootstrap ratios are depicted at the finest resolution (1015 nodes), showing that dopamine depletion increases signal variability at most nodes. (c) Bootstrap ratios are shown in 3D space sagittally and axially. Corresponding results are shown for all resolutions in Figure S1.

Figure 3.

Node- and network-level effects of dopamine depletion | (a) The top 10% of the nodes (i.e., top 100 nodes) that had the largest increase in signal variability (largest bootstrap ratios) following dopamine depletion. Each bar shows the magnitude of bootstrap ratio of a node and is colored based on the community assignment of that node (Mišić et al. 2015a). Somatomotor (yellow) and salience (green) networks appear over-represented compared with other networks. (b) The mean change in signal variability is calculated for each network and assessed by permutation tests (10 000 repetitions). Signal variability increases most in the salience and somatomotor networks following dopamine depletion, and these are the only two networks where this effect is statistically significant. (c) Changes in mean signal variability are depicted for somatomor and salience networks (significance obtained by permutation tests; FDR corrected (Benjamini and and Hochberg 1995)). SM = somatomotor, SAL = salience, FPN = fronto-parietal, VA = ventral attention, SUB = subcortical areas, DMN = default mode, VIS = visual, DA = dorsal attention, TEM = temporal.

Figure 4.

Relating signal variability and functional connectivity | (a) Mean changes in functional connectivity following dopamine depletion were estimated across all nodes and correlated with changes in within-region signal variability. Changes in functional connectivity are related to changes in signal variability, such that the larger the increase in signal variability, the larger the decrease in functional connectivity. (b) Mean changes in functional connectivity for intrinsic networks are correlated with mean changes in local signal variability in those networks. There is a clear anti-correlation between the two, consistent with the result in part (a). (c) The mean changes in functional connectivity was calculated for each network and assessed by permutation tests (10 000 repetitions). Mean connectivity significantly decreases in temporal, salience and somatomotor networks. Somatomotor and salience networks also experience significant increase in local variability (Fig. 3). (d) Mean functional connectivity in depletion (APTD) versus non-depletion (BAL) conditions, shown for nodes belonging to the temporal (TEM), somatomotor (SM) and salience (SAL) networks. Functional connectivity decreases in all instances (permutation test; FDR corrected).

Figure 5.

The effects of dopamine depletion on cohesion and integration of specific intrinsic networks | (a) Mean participation coefficient, indexing the diversity of inter-network connectivity, significantly decreases in somatomotor (SM) and salience (SAL) networks after dopamine depletion (using 10 000 permutation tests; FDR corrected). (b) Within-module degree z-score, indexing within-network connectivity, remains unaffected.