Imaging Biomarkers for Early Detection of Autism in Tuberous Sclerosis Complex

Background: Tuberous Sclerosis Complex (TSC) is a monogenetic disorder with high penetrance of autism spectrum disorder (ASD). Our previous work has characterized imaging features associated with ASD in adult and pediatric TSC patients. We determined that there are underlying microstructural integrity alterations in TSC patients with ASD that are not present in TSC patients without ASD, and that they exist throughout the brain, with varying degrees of severity, and are prevalent in the corpus callosum and arcuate fasciculus. There is an urgent unmet need for imaging biomarkers that can be used in the first year of life to identify and direct early intervention towards infants most likely to develop ASD.

Objectives: In light of our previous findings of abnormality in the corpus callosum and arcuate fasciculus, we set out to identify a biomarker of ASD in TSC through prospective and longitudinal study of microstructural development of the arcuate fasciculus, a fronto-temporal white matter (WM) fiber bundle implicated in language development, and in the corpus callosum, a large interhemispheric fiber bundle and collector of stochastically-distributed disease burden in TSC.

Methods: 40 participants with TSC underwent longitudinal 3T diffusion tensor imaging (DTI) within the first year of life and at 12, 24, and 36 months of age. Participants were assessed for early ASD symptoms at 12 months of age with the Autism Observation Scale for Infants (AOSI). Bilateral regions of interest were automatically defined along the arcuate fasciculi, located in the WM underlying Geschwind’s Territory, Wernicke’s Area, posterior to Broca’s area and in the body of the corpus callosum at the midline [Figure 1]. Diffusion properties in each region were measured. The ability of diffusion properties in each region to predict high or low AOSI total score at 12 months of age was evaluated with ROC analysis of the longitudinal sample, and in a subsample of data, restricted to scans acquired before 12 months of age.

Results: From the longitudinal sample, prediction of high or low AOSI total score was best achieved with fractional anisotropy of WM underlying left Geschwind’s territory (AUC=87.20, sensitivity=86.67, specificity=80.00) and left Wernicke’s area (AUC=85.07, sensitivity=86.67, specificity=76.00), while radial diffusivity of WM underlying left Wernicke’s area conferred a negative predictive value of 100% [Table 1a]. Prior to 12 months of age, WM microstructure underlying right Wernicke’s area was most predictive of high or low AOSI total score (radial diffusivity, AUC=76.89 ; mean diffusivity, AUC=76.14) and left Wernicke’s area (fractional anisotropy, AUC=75.76) [Table 1b]. AUC of corpus callosum diffusion properties did not exceed 75.00 in the longitudinal or 12 month sample, and therefore were not the best predictors of ASD symptoms at 12 months of age.

Conclusions: Temporoparietal white matter that supports language development within the first year of life is associated with later diagnosis of ASD in TSC. Although associated with ASD in adults, microstructure of the corpus callosum was not predictive of later diagnosis of ASD, suggesting its role as an accumulator of disease burden in TSC.