The Quantitative Autism Score (QAS): A Tool to Unravel Genetic Associations

Background:  The clinical diagnosis of ASD has changed considerably since its introduction into the formal nomenclature, reflecting conceptualization of ASD as a multi-dimensional phenotype. We hypothesize that genetic variants modify the expression of ASD symptomatology within subjects diagnosed with ASD, and specifically that variation in ASD candidate genes will lead to an increased number of core autism features. To evaluate this hypothesis, we developed a quantitative autism score (QAS) using variables on the ADI-R which reflect ASD features that have remained consistent throughout the changing diagnostic criteria (core ASD features).

Objectives: To identify genes that explain variation in the number of core ASD features.

Methods: The QAS was developed using the ADI-R, the semi-structured informant interview used to classify individuals for research studies in ASD. The QAS was defined using ADI-R algorithm items which consistently distinguish ASD from non-ASD or are present in early development and persist. The 25 QAS items are scored as present/absent and summed to yield a total score (0-25). Individuals with higher scores are those with more core features of ASD. The score was then calculated in 1118 ASD subjects from the Hussman Institute for Human Genomics (HIHG) and the Simons Simplex Collections. These individuals also had DNA sequence data available from a 17Mb custom capture covering 681 genes within regions identified by GWAS of ASD (Hussman et al 2013). SKAT-O (Li et al 2012) was used to conduct gene-based and single-variant tests for association with QAS as a quantitative trait. We examined combinations of synonymous, non-synonymous, missense, stop, loss-of-function and splice variants in different hypothesis tests. A Bonferroni correction for the number of genes tested was used as a significance threshold for each hypothesis with an experiment-wise significance level of 0.05. 

Results: Values for the QAS in the 1118 subjects ranged from 4 to 25 with a mean value of 18.06 (sd=3.942). We found significant association for the gene CDH4 (p= 9.20E-06) when all exonic variants were included in the gene-based test. This gene is a neuronal cell adhesion molecule known to play a role in brain segmentation and neuronal outgrowth and is a member of the cadherin family of genes, many of which have previously identified as ASD candidate genes and have been implicated by genetic association and sequencing for rare variants. We also found a significant association for the LGR5 gene when all exonic variants were included (p=3.59E-05) and when only missense variants were included (p=7.67-05).  LGR5 is known to be implicated in neuronal specification in the nervous system. Single-variant tests within these two genes identified two synonymous variants in CDH4 and two variants (1 synonymous and 1 missense) in LGR5 associated (p<0.05) with QAS.  

These results are in the process of being tested in three different replication samples.

Conclusions: Our study identified two genes, CDH4 and LGR5, associated with the number of core ASD features.  Our findings add support to the importance of genes involved in neurogenesis as well as to the role of phenotypic variations within ASDs.