Use of Specific Language Constructs for a Family Genetics Study of Autism and Language Impairment

Background: Over the past decade there have been multiple publications comparing the language behaviors of a subset of individuals with autism who have spoken language to individuals who have been diagnosed with Specific Language Impairment (SLI). One of the motivations for these comparisons is to determine if there is a genetic link specifically in the language domain. Several genetic studies have already linked both disorders to genes on chromosome 7q using a language phenotype.  While SLI is described as the failure to develop adequate language skills in the absence of any neurological or environmental influences, language issues are only part of the defining characteristics of autism. Both ASDs and SLI are complex disorders with potentially multiple gene involvement, so there is no reason not to believe that a subset of individuals with autism and a subset with SLI may share genetic influences. It has been difficult to replicate behavioral and linkage studies due to differences in recruited samples and variables used to create phenotypes. However, most agree that the more refined a phenotype is, the better the chance of it becoming a strong behavioral biomarker for autism. 

Objectives: As part of a larger family study of the genetic basis of autism (New Jersey Language and Autism Genetics Study- NJLAGS), we developed language -based phenotypes for linkage and association studies of autism and SLI. We found that family members with language-based learning impairments (LLI) performed similarly to verbal family members who met criteria for ASD when broad language criteria were used such as oral language impairment (LI) and written language impairment (RI) (Flax et al., under review). In this next step of phenotypic characterization, our goal is to create more discrete language phenotypes for ASD and SLI, compare them to the current broader language phenotypes, then determine if they may be more advantageous for the creation of behavioral biomarkers.

Methods: Seven language constructs (subtypes) were created using task item analysis and test construct validity from the 22 subtest variables of the NJLAGS language testing battery. Subtypes include Language Comprehension, Expression, Phonological Processing, Verbal Memory, Higher Order Language, Processing Speed, Language Structure, and Written Language. Using bivariate correlational analyses, we looked separately at those individuals who were previously categorized as affected for ASD or LLI from the NJLAGS study and examined the strength of these newly created constructs to see if they may in fact serve as more discrete language phenotypes for future genetic analyses.

Results: Overall, there were more significant associations among the subtests included in each language construct for the ASD group than in the LLI group. Additionally, while there was some agreement between the two groups on which subtests within a construct were associated, there was greater variability in the number of subtests that were correlated and the strength of the correlations for the LLI group. 

Conclusions: Using specific language constructs to characterize language phenotypes is a reasonable method for decreasing genetic heterogeneity in family linkage and association studies.