Structural Language Abilities Are Related to Cortical Structure and Covariance in Autism Spectrum Disorders

Background: Autism spectrum disorders (ASD) are characterized by significant difficulties in language and communication. Recent changes in diagnostic criteria define language difficulties as distinct from core communication symptoms. Neuroimaging evidence suggests that language impairments in ASD may be related to atypical connectivity between fronto-temporal brain areas (Courchesne et al, 2007). However, the relationship of language and communication abilities to brain structure remains unknown.  

Objectives: The objectives of the current study were to measure cortical thickness and anatomical covariance associated with language and communication in ASD compared with typically developing (TD) controls.

Methods:  Participants were 46 ASD and 50 TD males, matched on age (mean=12.8 years, SD=3.06), with IQ>75. ASD group was diagnosed using ADI-R and ADOS. Structural language and communication abilities in ASD were assessed using CELF-4 (Clinical Evaluation of Language fundamentals, Semel et al, 1996) and CCC-2 (Children’s Communication Checklist-2, Bishop, 1998), respectively. High-resolution T1-weighted images obtained for all participants were analyzed using the CIVET pipeline (Ad’dabagh et al, 2006) to calculate cortical thickness (20mm smoothing). Seed-based analysis of anatomical covariance was performed to measure the Pearson correlation coefficient, across subjects, between cortical thickness at a seed vertex and all other vertices to generate a group map of covariance (Evans, 2013). Seed loci known to recruit the language structural covariance network (SCN) were selected in left inferior frontal gyrus (IFG), left superior temporal pole (STP) and their right hemisphere homologues. Statistical analyses were performed using vertex-wise general linear interaction models with age, site, full-scale IQ and brain volume as nuisance variables and vertex-wise seed thickness as variable of interest. Modulation of SCN strength by language and communication ability in ASD was measured by adding terms for CELF-4 and CCC-2 scores in the models respectively. All results were corrected for multiple comparisons at p<0.05 using random field theory.  

Results: Results showed increased cortical thickness in ASD versus TD in left fronto-temporal regions. SCNs for both groups were mapped from 4 seed loci. While SCNs for controls reflected intrinsic connectivity networks described in earlier studies, SCNs for ASD showed widespread disruption, especially for the left STP seed. A direct comparison of the SCNs between TD and ASD revealed reduced covariance of the left STP seed with a region in the right frontal loci in ASD, suggesting decreased bilateral interactions of the left hemisphere language network. Furthermore, alterations in both cortical thickness and covariance were modulated by structural language ability as measured by CELF-4 of the ASD group but not communicative function (as measured by CCC-2).

Conclusions: Our findings reflect distinct differences in cortical structure and covariance of fronto-temporal regions in ASD, which are best explained by their structural language abilities but not communicative abilities. These differences further indicate the importance of structural language abilities in the study of altered fronto-temporal cortical structure and covariance in ASD. They also suggest that diagnostic specifiers, such as language, can be useful tools for understanding heterogeneity while maintaining the generalizability of findings in brain structural differences, much more than either symptom severity or cognitive ability.