Altered Organization of the Connectome in Pre-School Aged Children with Autism Spectrum Disorder

Background:    It is now largely recognized that Autism Spectrum Disorder (ASD) is a syndrome that involves alterations in brain connectivity. Previous diffusion-weighted imaging (DWI) research in late childhood has emphasized microstructural alterations of white matter pathways in Autism by using voxel-wise statistics. However, thorough analyses of whole-brain network properties of the structural connectome in early childhood are, thus far, limited. 

Objectives:   In the current study, we expand on the literature by using probabilistic tractography along with a graph theoretical measurement, termed communicability, to determine what aspects of brain communication might be altered in children with Autism.

Methods:    T1-weighted structural images and DWI were acquired in a sample of 55 children with ASD (age range: 3-5 years) and 13 typically developing (TD) children matched on age and gender. Scanning was performed while children were sleeping and images were visually inspected to exclude subjects with motion or other noticeable artifact. Voxelwise diffusion was modeled using constrained spherical deconvolution (Tournier et al., 2008) and probabilistic tractography was performed throughout all voxels of white matter. The Freesurfer cortical gray matter parcellation of 83 regions was then used to generate a whole-brain structural connectivity matrix (the connectome) for each subject. In order to understand how alterations in the structural connectome might relate to altered functional brain communication, we applied a measure termed communicability. Communicability measures the ease with which information can travel between two nodes (Crofts and Higham, 2009) by calculating a weighted sum of all paths between them. Comparisons of communicability across all node pairs was conducted between ASD and TD subjects using the Network-Based Statistic (Zalesky et al., 2010). In addition, multi-dimensional scaling was applied to child matrices to examine the variability of brain-wide communicability profiles within the TD and ASD populations.

Results:    Structural networks of children with ASD showed remarkable agreement with their TD counterparts across much of the region set when comparing node strength (i.e. the sum of a node’s connection strengths) and node communicability (the sum of a node’s communicabilities). However, results indicated reduced communicability in ASD between a subset of nodes including regions of the frontal cortex and the caudate (T-statistics>3, FWE-corrected P<.04). Results implicate the lateral orbitofronal cortex most heavily, although the frontal operculum and superior frontal cortex also had reduced communicabilities. Our multi-dimensional scaling analysis identified a small number of subjects with ASD (n=6) whose pattern of brain-wide communicability differed markedly from the central tendency of the TD group and the remainder of the ASD sample.

Conclusions:    Consistent with previous work, our findings suggest that from an early age ASD may be linked with alterations in brain communicability. In particular, connections throughout the integrative systems of the frontal cortex appear most atypical. Importantly, while our findings were robust across the group, not all children shared the same brain phenotype - highlighting sample heterogeneity. Future work will focus on characterizing this heterogeneity, and identifying the behavioral, genetic, and immunological correlates of these connectome phenomena.