Exome Sequencing and Clinical Correlates in 250 ASD Trios

Background: The diagnoses of ASD is mainly based on i) defficiencies in social interaction and communication and ii) restrictive and repetitive behavioral patters, interests or activities, including sensorial alterations. The behavioral assessment sufficient for a clinical diagnosis has limited use in research field and it is also likely that it will also have limited use for personalized medicine approaches, where trying to identify different biological subtypes seems the appropriate move forward .

In addition to important polygenic component, a great variety of single-gene disorders and chromosomal abnormalities have been described across latest genetic studies. Proper biological pathways and/or specific physiopathological mechanisms involved such as chromatin remodeling, redox system regulation or transcriptional regulation processes have been identified as candidate gene sets enriched in deleterious rare, inherited and de novo variation [Lai et al. 2013; de Rubeis et al 2014; Kosmicki 2017].

Objectives: In this study, we aimed to differentiate ASD patients based on rare disrupting and common genetic predisposing variation found to be enriched in main gene networks previously reported in ASD, as synapse, chromatin remodelling or transcription regulation. We extensively phenotipically characterize each group based on clinical manifestations, including restricted and repetitive behaviors, unusual interests, developmental regression, epilepsy or gastrointestinal disease.

Methods: 250 Spanish ASD trios were recruited at Hospital Universitario Gregorio Marañón. Demographic, clinical and neuropsychological variables were collected from trios samples to perform cluster analysis with a combination of hierarchical and k-mean methodology. Exome from blood DNA was sequenced as part of the Autism Sequencing Consortium (ASC) dataset. Bioinformatic pipeline was constructed based on Broad Institute recommendations from alignment, base Quality Control or variant calling, and ANNOVAR was used for variant annotation from vcf files. For rare variant analysisi, we filtered rare (MAF < 0.1) and deleterous variation, making use of a myriad of existent and recently published algorithms to described funtional, brain-expressed and intolerant variation. To analyze common polygenic variation, we imputed whole genome variants using Michigan Imputation Server, utilized PGC available data as discovery sample and used PRSice to poligenic risk score calculations in our target sample. General linear model was used to statistically analyze associations between genetic and phenotypic variables collected.

Results: ASD sample was biologically characterized based on both rare and common predisposing genetic variation. We found disctintive patterns of common varaition and de novo rare disrupting variation affecting intolerant genes (pLI > 0.9) across identified groups. Biological groups were exhaustively characterized, according to several clinical variables as intelectual disability (ID), developmental regression, psychiatric comorbility, gastrointestinal pathology or pro-inflamatory status. We also report suggestive results in specific traits with several types of variation, as epilepsy and disrupting mutations within astrocyte related genes (p = 4.5 x 10^-4) or post-zygotic mutations dependence with diagnostic categories (P = 5 x 10^-5).

Conclusions: We claim the usefulness of differentiating biological clusters with diagnosable atributes not only to allow the establishment of stratification criteria for clinical purposes, but also to improve decision making about clinical trials designs to improve pharmacological treatment choice.