Abnormal brain synchrony in Down Syndrome

Abstract

Down Syndrome is the most common genetic cause for intellectual disability, yet the pathophysiology of cognitive impairment in Down Syndrome is unknown. We compared fMRI scans of 15 individuals with Down Syndrome to 14 typically developing control subjects while they viewed 50 min of cartoon video clips. There was widespread increased synchrony between brain regions, with only a small subset of strong, distant connections showing underconnectivity in Down Syndrome. Brain regions showing negative correlations were less anticorrelated and were among the most strongly affected connections in the brain. Increased correlation was observed between all of the distributed brain networks studied, with the strongest internetwork correlation in subjects with the lowest performance IQ. A functional parcellation of the brain showed simplified network structure in Down Syndrome organized by local connectivity. Despite increased interregional synchrony, intersubject correlation to the cartoon stimuli was lower in Down Syndrome, indicating that increased synchrony had a temporal pattern that was not in response to environmental stimuli, but idiosyncratic to each Down Syndrome subject. Short-range, increased synchrony was not observed in a comparison sample of 447 autism vs. 517 control subjects from the Autism Brain Imaging Exchange (ABIDE) collection of resting state fMRI data, and increased internetwork synchrony was only observed between the default mode and attentional networks in autism. These findings suggest immature development of connectivity in Down Syndrome with impaired ability to integrate information from distant brain regions into coherent distributed networks.

Fig. 1

Age distribution of subjects used in the analysis. Down Syndrome and control participants are shown above, with ABIDE sample (both autism and control) shown below. Histograms show age in 5-year bins.

Fig. 2

Between-network connectivity differences. Mean time series were obtained from voxels comprising 7 non-overlapping distributed networks in the brain for each subject and Fisher-transformed correlation was measured in each subject between the time series for each pair of networks. Top left; Mean correlation values for control and Down Syndrome groups are shown for each pair of networks. Vertical and horizontal lines show standard error of the mean across subjects for each network pair. The diagonal line shows y = x. Green circles show results for mean inter-network correlation using a 0.1 mm threshold for motion scrubbing in the Down Syndrome group, and red circles show analogous results without any motion scrubbing. Top right; Analogous scatter plot of between-network synchrony is shown for the ABIDE dataset. Vertical and horizontal lines show standard error of the mean across subjects for each network pair. Bottom left; Significant increased (colored) between-network synchrony was found for 14 of 21 network pairs using FDR q < 0.05 across network pairs. T-scores for a two tailed t-test for each network pair are shown by the color scale. Bottom right; Significant increased between-network synchrony was observed only for 2 network pairs in the autism sample.

Fig. 3

Inverse relationship of internetwork correlation and performance IQ in Down Syndrome. Top; Scatter plot showing mean Fisher-transformed internetwork correlation (averaged across all network pairs) compared to performance IQ for Down Syndrome subjects. Bottom; Significant correlations between individual network pairs and Down Syndrome performance IQ, with acceptable False Discovery Rate q < 0.05 among all network pairs.

Fig. 4

Intergroup differences in functional connectivity with respect to Euclidean distance and correlation strength of connections. 26.3 million “connections” were grouped into bins based on distance between ROIs in mm and mean correlation for the connection in a sample of 1019 subjects from the 1000 Functional Connectome and ADHD 200 datasets. Within each bin, the mean T-score (two-tailed t-test) between Down Syndrome and control groups (left) or autism and control groups (right) is shown. Above; Mean T-score between groups for all “connections” in the bin. Below; Mean T-score between groups for all “connections” in the bin with p-value < 0.001.

Fig. 5

Spatial distribution of “connections” that differed between control and patient groups (at p-value < 0.001). Color scale shows the number of significant connections that had a given ROI as an endpoint. The top 50% of ROIs are shown in warm colors for autism and cool colors for Down Syndrome. ROIs common to both autism and Down Syndrome are colored in magenta. Images are in radiological format, with subject left on image right.

Fig. 6

Parcellation of the brain into functional communities for Down Syndrome, healthy control, and large control samples. The Down Syndrome and healthy control data were obtained during cartoon viewing, while the large control sample was obtained from shorter epochs of resting state data. Each of the images was thresholded at 7 networks. Therefore, the red/brown regions of each image are in fact compilations of multiple networks.

Fig. 7

Intersubject synchronization to cartoon stimuli. Top; Intersubject synchronization between each pair of subjects. The entire 50 minute time series was correlated in each of 7266 ROIs, and Fisher-transformed intersubject correlation was averaged across all ROIs for each subject pair, with the result shown in color in the plot. Below; Significant intersubject correlation among control/control and DS/DS subject pairs as well as for significant differences between the two sets of subject pairs, determined by a Wilcoxon rank sum test. An acceptable False Discovery Rate (FDR, q < 0.05) was allowed for the set of all p-values for intersubject synchronization measurements from all regions and all subject pairs. The minimum correlation associated with p-values less than this threshold was chosen as a significant correlation threshold, and regions for which mean correlation across subject pairs was greater than this correlation value were deemed significant and shown in color on the images. Color scale shows mean Fisher-transformed correlation across all subject pairs for a given region.