Comparison of Functional Connectivity Abnormalities in Autism and Williams Syndrome

Background:  Williams Syndrome and autism both exhibit characteristic social impairments, but differ in the specific social phenotype, with autism exhibiting social avoidance and Williams Syndrome exhibiting enthusiasm for social engagement but impaired skills in social interaction and also associated cognitive impairment. Differences in relationships between social brain regions in each condition may offer clues to how these divergent social impairment phenotypes may arise.

Objectives:  We compared brain functional connectivity within the social brain for a large subset of autism patients from the ABIDE dataset and a smaller group of individuals with Williams Syndrome as well as individuals with typical development obtained locally.

Methods:  Functional connectivity MRI measurements were performed during viewing of Looney Tunes cartoon videos in 13 subjects with Williams Syndrome and 10 typically developing subjects, grouop matched for age and sex (age range 14-35). Each subject viewed up to 7 cartoons of 7 minutes each, for a total of 67 7-minute scans in controls and 71 7-minute scans in Williams Syndrome participants. Imaging was acquired using multiband BOLD protocol on a Siemens Trio 3T scanner (TR = 730 ms, 2 x 2 x 2 mm resolution, 32 channel head coil). Postprocessing included motion correction, coregistration to MP2RAGE anatomic image, normalization to MNI space, and voxelwise regression of degraded ROIs within white matter, CSF, facial soft tissues, and time series from heart rate and respiratory waveforms obtained during scanning. Time series were extracted for 333 cortical regions parcellating cortical gray matter, 14 subcortical regions obtained from FreeSurfer-derived subject-specific segmentation of deep gray nuceli, and 14 cerebellar ROIs. Functional connectivity was measured in each pair of ROIs. Independently, functional connectivity was measured from the ABIDE dataset consisting of 972 autism and control subjects acquired from multiple sites with conventional resting state acquisition. Postprocessing was analagous with exception of heart rate and respiratory waveform regression. Regions showing connectivity abnormalities in each dataset were compared to a map of the social brain obtained from a reverse inference from the NeuroSynth.org website for the term "social," identifying regions significantly associated with the term social in the functional neuroimaging literature. Significant differences in connectivity were established using false discovery rate correction for multiple comparisons across all region pairs.

Results: Individuals with Williams Syndrome, compared to typically developing controls, exhibited widespread, significant decreases in cortico-cortical and cortico-subcortical connectivity. Regions most frequently involved in atypically decreased connectivity in Williams Syndrome included hubs of the social brain: medial prefrontal, inferior parietal, superior temporal sulcus, and amygdalar regions. Individuals with autism showed similar regions of greatest cortical underconnectivity, but also exhibited additional atypical connections not seen in Williams Syndrome, including decreased left-right homotopic connectivity and increased connectivity between default mode and salience networks.

Conclusions: We observe similar underconnectivity of social brain regions in Williams Syndrome and autism, but with differences in the functional connectivity pattern: homotopic underconnectivity, corticostriatal overconnectivity, and increased connectivity between default mode and attentional networks in autism, but more widespread corticocortical underconnectivity in Williams Syndrome. Such differences may contribute to differences in social impairment phenotype between the two conditions.