Comparison of Phenotypic Severity Associated with Autism Risk CNVs in Clinically Identified Samples of High Functioning Males and Females

Background: Clinically identified samples of ASD typically exhibit a male:female sex ratio > 4:1. Many genetic studies report females possess a greater genetic ‘burden’ and substantially lower IQ than males with an equivalent phenotype (SFARI Foundation Autism Research Initiative). Current estimates point to 10-20% of ASD being caused by Copy Number Variants (SFARI database). It is not known whether apparently large sex differences in ASD prevalence (especially in normal/high IQ samples) reflect a failure of ascertainment of ‘high functioning’ females, rather than any substantial differences in genetic risk, as measured by the burden of CNVs.

Objectives: To investigate in a 'high-functioning' sample of clinically identified autistic children, with IQs in the normal/high range, the impact of ASD-risk CNVs (as compiled by SFARI) on ASD symptoms and cognitive abilities. To compare and contrast male and female phenotypes in those with and without ‘pathogenic CNVs’. To test the hypothesis that females with pathogenic CNVs have more severe phenotypes than males with equivalent genetic risk.

Methods: Subjects were consecutive referrals to the National Centre for High Functioning Autism in London, UK. The autistic traits of 253 participants with autism were measured using a standardized interview (ADI algorithm) and the Autism Diagnostic Observation Schedule (ADOS). IQ was tested with the Wechsler Intelligence Scale for Children-IV (UK). Salivary or buccal cells were collected from the participants for DNA extraction. The microarray analysis carried out in the North East Thames Regional Genetics Laboratory was applied on Affymetrix 750K oligo SNP array. Ratio plots generated by infoQuant Fusion v6 software to show the copy number losses or gains were analysed with an electronic reference file, supplied by Affymetrix, which was based on averaging the results from 1000 sex-matched individuals from 4 major ethnicities.

Results: We compared male:female phenotypes in terms of ADI algorithm scores, ADOS algorithm scores, and IQ subscales. The mean verbal IQ of females with no pathogenic CNV was 91.7 (SD 20.6); that of females with a pathogenic CNV was 104.2 (SD 19.9). Comparing males and females without CNVs (233), no significant gender differences were found in terms of severity of phenotypes, in any domain of the ADI, ADOS, or IQ subscale. In those with pathogenic CNV (20) the effect sizes of gender differences in ADI algorithm scores were in the range 0.03-0.6, with females tending to show more severe phenotypes in social interaction and repetitive/stereotyped behaviours. On all ADOS subscales, the rated phenotype was more severe in males (effect sizes in range 0.4 – 1.4). In the sample with a pathogenic CNV, verbal and performance IQ were higher in females than males (effect sizes in range 0.5-0.6). CNV burden was no greater in females than in males.

Conclusions: There has been a probable ascertainment bias in recruitment studies, leading to erroneous conclusions about the impact of pathogenic ASD-related CNV in females. Females with ASD who are at genetic risk, but who have normal/high range IQ, are being missed by clinical research methodologies. They have milder phenotypes than males when assessed by standardized observation.