This past year we have continued to evaluate the neural characteristics, and associated behavioral consequences, of Autism Spectrum Disorders (ASD). Prevailing theories suggest that ASD results from impaired brain communication due to aberrant timing or coordination of neuronal firing patterns (impaired synchrony). However, it remains debated whether synchrony abnormalities in ASD are among local and/or long-range circuits, are circuit-specific or are generalized, reflect increased (hyper-) synchrony and/or decreased (hypo-) synchrony, and whether they are frequency band specific or are distributed across the frequency spectrum. In previous studies, we used data driven techniques in conjunction with both fMRI and structural data to show that the abnormalities in ASD brain networks are most prominent within brain regions that support social functions. To provide an additional test of this hypothesis, as well as determine how oscillatory neural dynamics are affected in ASD, we recorded spontaneous MEG data in 17 high functioning adolescents and adults with ASD and 18 controls matched on age, IQ, and sex, and equated for motion during the scanning session. We used a method we recently developed to look at all-to-all synchronization across the brain in conjunction with data driven analyses to compare local and long-distance synchrony in a frequency-specific manner. We found that individuals with ASD showed local increased or hypersynchrony in the theta band (4-7 Hz) in lateral occipitotemporal cortex. We also observed long-range decreased or hyposynchronous activity in the alpha band (10-13 Hz) that was most prominent in neural circuitry underpinning social processing. The magnitude of this alpha band hyposynchrony was correlated with social symptom severity. These results suggest that while ASD is associated with both decreased long-range synchrony and increased posterior local synchrony - with each effect limited to a specific frequency band - impairments in social functioning may be most related to decreased alpha synchronization between critical nodes of the social processing network (Ghuman et al., 2016). Our lab, as well as many others, have also documented deficient neural connectivity in the ASD brain via analyses of slowly fluctuating neural activity recorded while subjects rest quietly in the MRI scanner. However, the viability of this measure of neural connectivity as a diagnostic biomarker of ASD remains to be investigated. To address this important question we used a large number of machine learning classifiers to distinguish ASD from typically developing control subjects using resting-state fMRI data obtained from our own subjects (59 high functioning males with ASD and 59 age- and IQ-matched typically developing males) and an additional age and IQ-matched cohort of 178 individuals (89 ASD; 89 TD) from the Autism Brain Imaging Data Exchange (ABIDE) open-access dataset for replication. High classification accuracy was achieved through several analysis methods (peak accuracy 76.67%). However, classification via behavioral measures consistently surpassed all findings based on measures of neural connectivity (peak accuracy 95.19%). Thus, while individuals can be classified as having ASD with statistically significant accuracy based on their resting-state scans alone, this method falls short of biomarker standards. Nevertheless, the connectivity measures provided further evidence that ASD is characterized by dysfunction of large-scale functional brain networks, particularly those involved in social information processing (Plitt et al., 2015a). Though typically identified in childhood, the defining social and communication symptoms of autism spectrum disorders (ASD) persist throughout the lifespan. Although higher IQ, less severe symptoms, and stronger adaptive behaviors moderately predict positive outcomes in ASD, there is substantial outcome variability even among high functioning individuals with ASD (Farley et al., 2009). Thus, there is a critical need to identify predictors of adulthood outcomes in individuals with ASD. The work described above identified three functional networks that best discriminated ASD from TD individuals (the so-called salience network, default-mode network, and the frontoparietal task control network (2). In this study we used those networks to constrain our search for brain networks that were predictive of longitudinal change in symptoms, a realm in which resting state fMRI may be able to serve as a prognostic biomarker. Previous studies have found various behavioral assessments (such as intelligence quotient (IQ), early language ability, and baseline autistic traits and adaptive behavior scores) to be predictive of outcome, but most of the variance in functioning remained unexplained by such factors. We investigated to what extent functional brain connectivity measures could predict the variance left unexplained by age and behavioral measures of outcome; adaptive behaviors (as measured by the Adaptive Behavior Assessment System, ABAS) and autistic traits (as measured by the Social Responsiveness Scale) at least 1 year post scan (mean follow-up latency = 3 years). We found that connectivity involving all three networks was highly predictive of future autistic traits and the change in autistic traits and adaptive behavior over the same time period. Furthermore, functional connectivity involving the salience network, which is predominantly composed of the anterior insula and the dorsal anterior cingulate, was extremely accurate at predicting improvement in adaptive behaviors (Plitt et al., 2015b).
Showing the most recent 10 out of 45 publications
