Heterogeneity of the Language Network in Autism Spectrum Disorders: A Data-driven Study of Neurophenotypes

Autism Spectrum Disorders (ASD) are heterogeneous developmental disorders associated with atypical functional connectivity (FC) and neuroanatomy. Language impairments affect individuals with ASD, but the neural underpinnings remain elusive, partly due to the heterogeneity across the ASD population. The current studies utilized multimodal neuroimaging to explore 1) differences in language network intrinsic FC (iFC) between children diagnosed with ASD and typically developing (TD) children; 2) whether there are distinguishable ASD subgroups based on language network iFC patterns, and 3) how these iFC subgroups relate to ASD subgroups derived from anatomical features of the language network. Study 1 (Gao et al., 2019): Seed-to-whole brain iFC analyses revealed that school-age children with ASD (n= 52) had increased iFC of language regions with posterior cingulate cortex and visual regions, in comparison to TD peers (n=50). Study 2: An expanded sample of 69 children with ASD showed atypically increased within-group heterogeneity of language network iFC. Latent profile analysis (LPA) of language iFC dimensions revealed three distinct ASD subtypes, each with lower language abilities than their TD peers (n= 60): one subgroup that was similar to the TD group in iFC, and two subgroups that exhibited broad under- and overconnectivity, respectively. Study 3: LPA of anatomical dimensions of combined cortical thickness (CT) and local gyrification index (lGI) in a cohort of 104 ASD children uncovered three distinct subgroups (sASD1-3): sASD1 characterized by increased lGI and lower language scores, sASD2 composed of older ASD participants with lower sociocommunicational symptoms, and sASD3 showed greater CT in left hemisphere language regions. Subgroup membership based on iFC- and morphology-based LPA was not clearly related. Neither subgroup type (iFC or structural) was related to diffusion indices of language-related white matter tracts. Overall, our findings expand on previous reports of increased heterogeneity and atypical language network iFC in ASD. Language network ASD subtypes and membership differed across imaging modalities. The existence of ASD subtypes with distinct iFC and anatomical patterns may explain conflicting results in the ASD imaging literature. Our findings underscore the need to focus on individual variability in ASD beyond conventional group-level analyses