Microstructural changes in the reward system are associated with post-stroke depression

Abstract

Background: Studies of lesion location have been unsuccessful in identifying mappings between single brain regions and post-stroke depression (PSD). Based on studies implicating the reward system in major depressive disorder without stroke, we investigated structural correlates within this system and their associations with PSD.

Methods: The study enrolled 16 healthy controls, 12 stroke patients with PSD and 34 stroke patients free of PSD. Participants underwent 3T structural and diffusion MRI. Graph theoretical measures were used to examine global topology and whole-brain connectome analyses were employed to assess differences in the interregional connectivity matrix between groups. Structural correlates specific to the reward system were examined from grey matter volumes and by reconstructing its main white matter pathways, namely the medial forebrain bundle and cingulum connections, using deterministic tractography. Fractional anisotropy (FA) was derived as a measure of microstructural organization, and extracellular free-water (FW) as a possible proxy of neuroinflammation.

Results: Subnetworks of decreased FA-weighted and increased FW-weighted connectivity were observed in patients with PSD relative to healthy controls. These networks subsumed the majority of regions constituting the reward system. Within the reward system, FA and FW of major connection pathways and grey matter volume were collectively predictive of PSD, explaining 37.8% of the variance in depression severity.

Conclusions: PSD is associated with grey matter volume loss, reduced FA and increased extracellular FW in the reward system, similar to features observed in major depression without stroke. Structural characterization of the reward system is a promising biomarker of vulnerability to depression after stroke.

Keywords: Connectome; Grey matter; Magnetic resonance imaging (MRI); Neuroinflammation; Post-stroke depression (PSD); White matter.

Fig. 1

A) Cingulum bundle subdivisions and B) Medial forebrain bundle (MFB) reconstructed with anatomically constrained tractography from one representative participant.

Fig. 2

Lesion locations mapped for all participants in standard (MNI) space does not reveal a pattern of association between GDS scores and lesion location. A) Lesions mapped for participants without depression, i.e. GDS scores from 0 (green) to 9 (blue). B) Lesions mapped for participants with depression, i.e. GDS scores from 10 (yellow) to 27 (red). C) VLSM maps calculated for GDS scores of 46 stroke patients. The (uncorrected) cluster includes areas of the medial forebrain bundle (MFB) and the putamen in the left hemisphere. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

A) Network of significantly reduced fractional anisotropy (FA)-weighted connectivity in the group of stroke patients with depression compared to the healthy control group. B) Network of significantly increased free-water (FW)-weighted connectivity in the group of stroke patients with depression compared to the healthy control group. T-statistics are set to a supra-threshold of 3, which corresponds to p = 0.001. Subnetworks are significant at p_FWE < 0.05. Node colour reflects number of connections in subnetwork (degree of a node in the subnetwork), ranging from blue (few connections) to pink (many connections). Edge weights reflect t-statistic magnitude, whereby thicker edges correspond to higher t-statistics. C) Grey matter (warm colors) and white matter (cold colors) structures constituting the reward system. The medial forebrain bundle (cyan) and the cingulum bundle (green) interconnect the grey matter structures of the reward system. Connectograms of the significant D) FA and E) FW networks. Node color correspond to different lobes and subcortical regions: pink = frontal lobe, blue = temporal lobe, purple = parietal lobe, magenta = occipital lobe, yellow = subcortical structures, light orange = insular cortex, orange = cerebellum. Connection color represents F-statistics magnitude, ranging from cyan (low F-statistic value) to green (high F-statistic value). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

Correlations between Observed GDS scores plotted on the x-axis and predicted GDS scores from the regression models plotted on the y-axis. A) Correlations excluding the statistical outlier on GDS scores in the training set (upper scatterplot) and the test set (lower scatterplot). B) Correlations including the statistical outlier on GDS scores in the training set (upper scatterplot) and the test set (lower scatterplot).