Brain responses to speech predict early language outcomes in children with autism

An essential component of current research initiatives on brain disorders is to enhance our understanding of normal and atypical brain development and how certain trajectories impact or underlie disease-related phenotypes. Studying brain disorders is inherently difficult, mostly due to their heterogeneity, genetic and biological complexity, and high comorbidity with other disorders.

One strategy to enhance our understanding of these disorders is to identify core features that lend themselves to systematic and longitudinal investigations. Autism spectrum disorder (ASD) is a poorly understood developmental disorder with tremendous heterogeneity in both clinical phenotypes and later clinical outcomes. A study, recently published in Neuron by Eric Courchesne’s laboratory, examined neural correlates of early language development in infants and toddlers with ASD. Early language development is strongly associated with clinical outcomes in some children with ASD, such that delayed or typical language skills are usually associated with better clinical outcomes, while children who are minimally- or non-verbal have poorer outcomes. The study aimed to determine whether early language development and associated neural subtypes within ASD could have predictive power for later clinical outcomes. 

The investigators used a large sample of children that included those aged 12-48 months with ASD, non-ASD language/developmentally delayed subjects, and typically developing infants and toddlers from the general population. Language development was assessed using multiple standardized scales. Neural response to speech was assessed using functional neuroimaging of language relevant bilateral lateral frontal and superior temporal cortices. Within the ASD group, cohorts were divided into ‘good’ or ‘poor’ classifications in early childhood based on their receptive and expressive scores on the Mullen Scales of Early Learning (i.e., integrated analysis of language, motor, and perceptual abilities). ‘Good’ or ‘poor’ were considered to be one standard deviation above or below the mean, respectively.

Each ‘subtype’ displayed differential developmental trajectories of language ability from 12-48 months of age. In the ASD-Good subgroup, language steadily progressed along trajectories of those with typical development, while language trajectories steadily declined in the ASD-Poor subgroup. Moreover, these trajectories corresponded to different neural subtypes within children with ASD. Children with ASD and typically developing language, along with the non-ASD language delayed children, displayed activation of the superior temporal cortical areas and activation patterns across these areas similar to typically developing children. The ASD-Poor subgroup displayed hypoactivation of these areas, lack of functional differentiation between types of speech in patterns of activation, and typical responses to general auditory processing, compared to the other groups. The authors posit that neural systems are generally intact for auditory processing in the ASD-Poor subgroup individuals and deficits are specific to language and speech.

Courchesne and colleagues then determined whether patterns of activation across brain networks of language correlated with language ability within all children. In children with better language, language areas, subcortical regions of reward, emotion and memory were normatively recruited, whereas the somatosensory area, thalamus, motor, premotor, and primary visual areas of the brain were recruited less. These relationships are almost completely reversed in children with ASD during early development from 12-48 months of age. In the ASD-Good subgroup whom had improving language scores, these relationships approached those displayed by typically developing children and children with language delay.

Together, these findings suggest there are neural responses to speech that correlate with early language development in those with or without ASD, and these correlates can be use to parse subtypes of children with ASD. The authors extended these findings by leveraging the longitudinal design of the study to investigate whether early neural response and/or behavioral measures (i.e., language test scores) could reliably predict language ability later in development. They found when combined, that early neural and behavioral information could be used to classify ‘good’ versus ‘poor’ language outcomes in toddlers with ASD, better than either criterion alone. The authors go onto to stress the potential utility of early identification of biomarkers in personalizing intervention and ultimately improving clinical outcomes in children with autism.

Publication

1. Courchesne E et al. (2015) Different functional neural substrates for good and poor language outcome in autism. Neuron 86:567-577. doi: 10.1016/j.neuron.2015.03.023