Autistic Traits Are Both Distributed and Localised within Structural and Functional Brain Networks

Background: Autism, the autism spectrum and the broader autism phenotype are behaviourally defined constructs whose physiology remains murky. Impairments in theory-of-mind present a major aspect of autism's 'core' social communicative symptoms, although it's unclear to what extent such symptoms may be fractionable features which interact only later in development (Valla & Belmonte, 2013; doi: 10.1016/j.dr.2013.08.004). Behavioural measurements have tended to be coarse, failing to capture the often subtle variability of theory-of-mind across the autism spectrum, and this limitation holds doubly when it comes to dimensional variation beyond the autism spectrum: Most assays have used a binary outcome- success or failure at the theory-of-mind question- which cannot assess a dimensional continuum, and narrative presentations which can confound theory-of-mind with verbal working memory.

Objectives: We ask whether both psychometric measures of autistic traits and a continuous, reaction-time measure of theory-of-mind, in a visual presentation, might correlate with neural network integrity at whole-brain and regional scales.

Methods: In 30 right-handed normal volunteers (12 females) aged 27.29±2.88, we studied the relationship between anatomical (diffusion tensor imaging (DTI)) and physiological (500 s of 2 Hz eyes-closed rs-fMRI) measures of brain connectivity, and psychometric (Social Responsiveness Scale (SRS), Autism Spectrum Quotient (AQ), Systemising Quotient (SQ)) and behavioural (the Attention Network Test (ANT)), and key-press reaction time difference associated with a computer-game theory-of-mind condition) measures of autistic traits. The AAL-90 atlas was warped to functional scans and to grey/white segmented T1-weighted anatomical scans to label regions. A functional correlation matrix was constructed over mean fMRI time series within each pair of regions. Similarly, segmented T1-weighted anatomical scans and the entire diffusion-weighted series were rigidly registered to the corresponding first b=0 image, the AAL-90 atlas was warped to these, and probabilistic tractography was conducted from grey-matter voxels to construct an anatomical connection strength matrix over all region pairs. These functional and anatomical correlation matrices produced network measures of clustering coefficient, characteristic path length, and efficiency (Rubinov & Sporns, 2010; 10.1016/j.neuroimage.2009.10.003).

Results: For the functional connectivity data, high SRS scores were associated with strong clustering (r=+.45, p=.008), short path length (r=-.39, p=.018) and high efficiency (r=+.47, p=.006), globally and within regions spanning temporo-parietal and prefrontal cortices. Similar trends manifested for AQ scores. Likewise, delayed orienting of attention (lower orienting effect) in the ANT was associated with strong clustering (r=-.42, p=.0095), short path length (r=+.44, p=.007) and high efficiency (r=-.46., p=.0133). For the structural connectivity data, SQ was associated with strong clustering (r=+.37, p=.023) and efficiency (r=+.38, p=0.018) in a relation driven specifically by temporal pole; theory-of-mind reaction time trended towards relation with strong clustering (r=-.27, p=.0766) and efficiency (r=-.28, p=0.0655) overall, but this relationship became highly significant within right supramarginal gyrus.

Conclusions: Autistic traits as indexed by SRS, AQ and SQ, and slowed attention shifting in the ANT, are associated with more efficient distributed resting-state functional networks. Theory-of-mind, in contrast, relates largely to localised structural integrity within right temporoparietal junction. Measured localisation of function depends not only on the cognitive capacity assayed but also on the assay itself.