Linking triadic autism symptoms to distinct features of functional brain connectivity

Background:   An understanding of how the brain produces cognition ultimately depends on knowledge of its large-scale organization and wiring. The human brain undergoes protracted structural and functional changes during which it constructs dedicated large-scale brain networks comprised of discrete, interconnected, brain regions. Aberrations in brain connectivity have been widely reported in autism spectrum disorders (ASD), but their relation to its three core symptoms – social communication, language and repetitive-restricted behaviors are unknown. 

Objectives:   To examine connectivity of large-scale brain networks and to determine whether specific networks can distinguish children with ASD from typically developing (TD) children and predict symptom severity in children with ASD. 

Methods:   Intrinsic functional connectivity of large-scale brain networks, machine learning based classifiers built to discriminate children with ASD from TD children based on specific brain networks, and relations between brain connectivity and core symptoms of ASD. 

Results:   First, across three different cohorts of children, between the ages of 7 and 12, and over 100 subjects obtained from the open-source Autism Brain Imaging Data Exchange (ABIDE) database, we found evidence for hyper-connectivity at the whole-brain level. There were significantly more links that showed hyper-connectivity than hypo-connectivity in ASD, and the degree of hyper-connectivity predicted severity of social deficits. Second, we also observed stronger functional connectivity within several large-scale brain networks in children with ASD compared with TD children. Hyper-connectivity in ASD was most prominently observed in the salience, default mode, fronto-temporal, motor, and visual networks. This hyper-connectivity result was replicated in an independent cohort obtained from ABIDE. Using maps of each individual’s salience network, children with ASD could be discriminated from TD children with a classification accuracy of 78%, with 75% sensitivity and 80% specificity. The salience network showed the highest classification accuracy among all networks examined, and signals in this network predicted restricted and repetitive behavior scores. The classifier discriminated ASD from TD in the independent sample with 83% accuracy, 67% sensitivity, and 100% specificity. Third, examination of the voice-selective cortex in children with ASD revealed a striking pattern of under-connectivity between left-hemisphere posterior temporal sulcus and distributed nodes of the dopaminergic reward pathway, including bilateral ventral tegmental areas and nucleus accumbens, left-hemisphere insula, orbitofrontal cortex, and ventromedial prefrontal cortex. Furthermore, the degree of under-connectivity between voice-selective cortex and reward pathways predicted symptom severity for language communication deficits in children with ASD.

Conclusions:   The three core deficits in children with ASD can be linked to distinct features of atypical functional brain connectivity in the disorder. Although under-connectivity has been posited to be a hallmark of atypical brain organization in autism, emerging findings in children with ASD are painting a decidedly more complex picture, one that has thrown into sharp relief the challenges facing our understanding of brain connectivity in autism. At the same time, they open new possibilities for a deeper understanding of the neurobiological origins of disorder.