Atypical Circuit Level Brain Activity in 3 Month Old Infants at Risk for Autism Spectrum Disorder

Background: Heterogeneous genetic and environmental etiologies of autism spectrum disorder (ASD) converge upon circuit level brain disruption well before behavioral symptoms emerge (DeLaTorre, 2017). Prompt detection of early changes in brain development could facilitate earlier identification or enhanced stratification of risk, which in turn can improve the timing of early intervention. Traditional algorithms to examine neural signal are limited in infants, due to inherent data collection challenges.

Objectives: Here we take a data driven approach, developing a robust pipeline to maximize EEG data quality to study cortex-wide electrophysiological differences in infants at risk for ASD, as measured by alpha oscillations (6-12 Hz), specifically phase coherence (APC; Dickinson et al., 2017). ‘High risk’ is defined based on an older ASD sibling (Charman et al., 2017). The objectives of the present study were, 1) Develop a robust signal processing pipeline appropriate for infant EEG data, 2) Quantify differences in APC in infants who develop early signs of ASD symptoms (at 18 months), and 3) Examine whether APC at 3 months predict later behavioral outcomes (at 18 months).

Methods: Spontaneous EEG data were collected during the first year of life from high and low risk infants, with focus here on the earliest timepoint (3 months). Infants were grouped into ‘high risk’ (ADOS-t CSS ≥ 4, N=14) and ‘typically developing’ (CSS<4, N=49) groups based on the presence of social communication impairments at 18 months, and the Mullen scales of early learning were used to assess overall development. We designed processing methods which allowed us to maximize data quality, while retaining enough data to calculate stable estimates of signal characteristics. This involved 1) extensively cleaning data using artifact subspace reconstruction (ASR) and independent component analysis (ICA), 2) transforming EEG data into current source density (CSD) using a Laplacian transform, and 3) calculating APC between every possible electrode pair, obviating the need for a priori assumptions regarding spatial patterns of atypical APC.

Results: 1) The pipeline presented here was successful in minimizing sources of noise in the data, and also mitigated the effects of volume conduction (particularly relevant to connectivity analyses), allowing us to retain 100% of the EEG recordings collected at 3 months of age. 2) FDR-corrected permutation testing reveal that APC is increased at 3 months in the group of infants who demonstrate ASD behaviors at 18 months. These group differences were most statistically evident in one long-range interhemispheric electrode pair (ASD: M=0.33; typical: M=0.23; P<.000003). 3) Across all participants, long range APC at 3 months was inversely related to verbal cognition at 18 months (R=-.56, P=.02).

Conclusions: The present data suggest that the dynamics of circuit level brain activity are altered early in infants who later show symptoms of ASD. The hyperconnectivity demonstrated here is consistent with structural findings of increased white matter during infancy in ASD (Wolff et al., 2012). We will discuss the potential utility of circuit level biomarkers to 1) objectively identify neurodevelopmental disruptions early in life, 2) support individualized prognoses, and 3) inform neurobiological targets of early intervention.