Linkage Analysis of Whole Exome Sequence Data in Multiplex Autism Families Including Cholesterol Covariates
The analysis of multiplex autism families may provide valuable insights into the risk of developing autism spectrum disorder (ASD). It is not known whether “sporadic” ASD is truly sporadic or a consequence of reduced penetrance, but our previous analyses of de novo variation suggest that multiplex families have a different underlying etiology. Linkage analysis is a method that is once again increasing in popularity for analyzing next-generation sequence data as it can leverage family structure to increase power compared to population-based methods and multiplex families may be enriched for high-penetrance, functional rare variants. It has also been observed that individuals with Smith-Lemli-Opitz syndrome show many autistic features and abnormal cholesterol measures have been demonstrated in individuals with ASD who do not have SLOS.
We aimed to analyze whole exome sequence (WES) data in multiplex ASD families from the Autism Genetic Resource Exchange collection using linkage analysis methods designed to incorporate covariate measures to see if some families show stronger evidence of linkage to particular loci in the presence of these covariate measures.
WES was performed in 69 families with 2-3 affected children from the AGRE collection. Quality filtering was performed in GoldenHelix SVS and Mendelian inconsistencies removed using Sibpair. Seven covariate measures were tested; total cholesterol, HDL cholesterol, ApoA1, ApoB, triglycerides and two categorical variables, hypercholesterolaemia and hypocholesterolaemia. Incorporating covariates into categorical trait linkage analysis is more challenging than quantitative traits, however some nonparametric methods exist. Here we use IBDReg, an approach that uses quasi-likelihood measures to test for which covariates might be influencing evidence for linkage at particular loci.
Significant evidence of linkage to ASD when including covariates was found on chromosome 1 (p=2.2x10-7), chromosome 2 (p=7.2x10-6) and chromosome 21 (p=6.1x10-6) for hypercholesterolaemia as the covariate and chromosomes 1 (p=1.1x10-7), 2 (p=1.6x10-6), and 21 (p=5.1x10-6) when hypocholesterolaemia were used as covariates. Several other regions with suggestive evidence of linkage of ASD were also observed for all covariates except ApoB, which had no signals even in the suggestive range.
Including covariates which may be relevant to ASD may help with interpretating WES results, particularly in families with multiple affected children. Different covariates produced different evidence of linkage to ASD, depending on which covariate was included, which may assist in classifying families with different underlying ASD etiologies. Interpreting the results of these kinds of linkage analyses is challenging and simulations are currently underway to calculate empirical p values. However, all of our signals have previously been reported in AGRE families or other ASD cohorts. The signals on chromosome 1 and 2 are close to reported linkage peaks in the literature, in particular the signal on chromosome 2 is in the 2q21.2 region which is part of the 2q21-q33 region which has been repeatedly reported in linkage analyses of ASD families. The locus on chromosome 21 is close to a region previously reported in a subset of ASD families with language regression. These regions are being examined in detail to attempt to identify the linked causal variants.