A Genetics-First Approach to Dissecting the Heterogeneity of Autism: Phenotypic Comparison of Autism Risk Copy Number Variants

Abstract

Objective: Certain copy number variants (CNVs) greatly increase the risk of autism. The authors conducted a genetics-first study to investigate whether heterogeneity in the clinical presentation of autism is underpinned by specific genotype-phenotype relationships.

Methods: This international study included 547 individuals (mean age, 12.3 years [SD=4.2], 54% male) who were ascertained on the basis of having a genetic diagnosis of a rare CNV associated with high risk of autism (82 16p11.2 deletion carriers, 50 16p11.2 duplication carriers, 370 22q11.2 deletion carriers, and 45 22q11.2 duplication carriers), as well as 2,027 individuals (mean age, 9.1 years [SD=4.9], 86% male) with autism of heterogeneous etiology. Assessments included the Autism Diagnostic Interview-Revised and IQ testing.

Results: The four genetic variant groups differed in autism symptom severity, autism subdomain profile, and IQ profile. However, substantial variability was observed in phenotypic outcome in individual genetic variant groups (74%-97% of the variance, depending on the trait), whereas variability between groups was low (1%-21%, depending on the trait). CNV carriers who met autism criteria were compared with individuals with heterogeneous autism, and a range of profile differences were identified. When clinical cutoff scores were applied, 54% of individuals with one of the four CNVs who did not meet full autism diagnostic criteria had elevated levels of autistic traits.

Conclusions: Many CNV carriers do not meet full diagnostic criteria for autism but nevertheless meet clinical cutoffs for autistic traits. Although profile differences between variants were observed, there is considerable variability in clinical symptoms in the same variant.

Keywords: Autism; Copy Number Variants; Genetics.

Figure 1:. Score variability between genetic variant groups

This plot visualises the between genetic variant group variation data presented in Table 2. Between group eta squared values are plotted on a scale of 0% variance to 100% of the variance. These values represent the proportion of variation in phenotypic outcome predicted by genetic variant group. A value close to 0% would indicate a non-specific model whereby different genotypes lead to similar phenotypic outcomes. As value close to 100% would indicate a highly specific model whereby different genotypes lead to different and discrete phenotypic outcomes. The bars indicate 95% confidence intervals.

Figure 2:

Domain profiles of the genetic variant + autism groups FSIQ, Full Scale Intelligence Quotient; VIQ, Verbal Intelligence Quotient; PIQ, Performance Intelligence Quotient ; ADI-R, Autism Diagnostic Interview. 2A To visualise how the “genetic variant + autism” groups differed, a heatmap plot was generated by transforming IQ and ADI-R scores of the “genetic variant + autism” groups to z-scores, dendrograms showing the clustering of CNVs and phenotypes were generated using methods described for 3A. Scores for each “genetic variant + autism” were standardized into z scores relative to each other and were adjusted for sex, age and site. The z scores were constructed so that a negative score always denoted a poorer performance. Black indicates a relative deficit in that neuropsychiatric domain compared to other CNV carriers, yellow represents a relative strength compared to other CNV carriers. Hierarchical clustering, for the purposes of presentation (indicated by the dendrogram), was performed using Ward’s method and Euclidian distance. 2B To visualise the profiles of the “genetic variant + autism” groups relative to the heterogeneous autism group phenotypic scores were standardised to z-scores, using the mean and SD of the heterogeneous autism group as reference—i.e., the difference in the individual’s score and the mean score for the entire autism heterogeneous group was divided by the SD for the heterogeneous autism group. The z-scores were adjusted for sex, age and site. We constructed these Z scores so that a negative score for a -CNV carrier indicated a worse outcome.

Figure 2:

Domain profiles of the genetic variant + autism groups FSIQ, Full Scale Intelligence Quotient; VIQ, Verbal Intelligence Quotient; PIQ, Performance Intelligence Quotient ; ADI-R, Autism Diagnostic Interview. 2A To visualise how the “genetic variant + autism” groups differed, a heatmap plot was generated by transforming IQ and ADI-R scores of the “genetic variant + autism” groups to z-scores, dendrograms showing the clustering of CNVs and phenotypes were generated using methods described for 3A. Scores for each “genetic variant + autism” were standardized into z scores relative to each other and were adjusted for sex, age and site. The z scores were constructed so that a negative score always denoted a poorer performance. Black indicates a relative deficit in that neuropsychiatric domain compared to other CNV carriers, yellow represents a relative strength compared to other CNV carriers. Hierarchical clustering, for the purposes of presentation (indicated by the dendrogram), was performed using Ward’s method and Euclidian distance. 2B To visualise the profiles of the “genetic variant + autism” groups relative to the heterogeneous autism group phenotypic scores were standardised to z-scores, using the mean and SD of the heterogeneous autism group as reference—i.e., the difference in the individual’s score and the mean score for the entire autism heterogeneous group was divided by the SD for the heterogeneous autism group. The z-scores were adjusted for sex, age and site. We constructed these Z scores so that a negative score for a -CNV carrier indicated a worse outcome.