Multimodal Stratification of ASD with the Help of DTI, MEG, and MRS

Background: Efficient white matter and synapses are both necessary for accurate and rapid encoding of auditory sensory information. In autism spectrum disorder (ASD), slow auditory response latencies may contribute to abnormal auditory processing and, ultimately, language impairment. A structure-function relationship between M50 latency and diffusion MR indices of white matter maturation has been shown in typically developing children. However, the M50 latencies observed in ASD are heterogeneous and not well explained by white matter microstructure. GABA is an inhibitory neurotransmitter and abnormally low auditory system GABA has been observed in ASD. This study hypothesizes that M50 is dominantly modulated by white matter microstructure (indexed by dMRI) in one subgroup of ASD and is modulated by synaptic transmission function (indexed by GABA MRS) in a separate biologically-distinct subgroup.

Objectives: This multimodal study uses dMRI, GABA MRS and magnetoencephalography (MEG) to identify the multiple physiological mechanisms which underlie auditory processing efficiency in ASD and to potentially stratify children according to differing biological etiologies.

Methods: Participants included 46 children with ASD (mean age = 11.5±2.3) and 27 typically developing (TD) children (mean age = 10.9±2.1) with evaluable DTI, MRS and MEG data. DTI was acquired at 3T with 30 gradient directions at b=1000s/mm² and voxel size 2x2x2mm. The MNI template was used to measure DTI parameters including FA from Heschl’s gyrus. MRS was performed using a single voxel macromolecule-suppressed, edited MEGAPRESS sequence (TR=1500ms, TE=80ms, 128 transient pairs, acq. time ~6min) with voxels (4x3x2cm) placed in left and right superior temporal gyrus and expressed relative to the reference creatine. The M50 response component to auditory stimulus tones was measured with MEG and identified as the first source localized post stimulus peak in the interval 50-150ms.

Results: A mixed model of M50 with hemisphere, diagnosis, GABA level, and cross terms revealed a main effect of FA (p<0.02) and an interaction between GABA and diagnosis (p<0.01). Analyzed separately in controls alone, increasing FA (p<0.05) and decreasing GABA (p<0.01) were correlated with shorter (faster) M50. In the ASD group, the relationship between FA, GABA and M50 was uncoupled. A subgroup of ASD subjects was identified with long (slow) M50 latencies relative to the typically developing M50 vs GABA curve (>95^th percentile). Outlier group FA (0.287±0.004) was lower than FA among non-outliers (0.296±0.002, p<0.05), suggesting latency was prolonged in this subgroup compared to the “typical for their GABA level” value, perhaps through immature white matter, with commensurately slow conduction velocity.

Conclusions: This study observed higher GABA levels were associated with slower M50 responses in typically developing children. This relationship is possibly related to the inhibitory role of GABA in cortico-cortical circuitry mediating the M50 response, after the initial thalamocortical volley. Uncoupling of GABA to M50 as well as diffusion MR to M50 was observed overall in ASD, suggesting heterogeneous subpopulations. This study suggests stratification of ASD subjects may be performed based on which physiological factors dominantly modulate or limit the speed of sensory input. Stratification based on such mechanistic etiology can potentially pave the way for tailored/personalized therapy.