Effect Sizes of Deletions and Duplications on Autism Risk Across the Genome

Abstract

Objective: Deleterious copy number variants (CNVs) are identified in up to 20% of individuals with autism. However, levels of autism risk conferred by most rare CNVs remain unknown. The authors recently developed statistical models to estimate the effect size on IQ of all CNVs, including undocumented ones. In this study, the authors extended this model to autism susceptibility.

Methods: The authors identified CNVs in two autism populations (Simons Simplex Collection and MSSNG) and two unselected populations (IMAGEN and Saguenay Youth Study). Statistical models were used to test nine quantitative variables associated with genes encompassed in CNVs to explain their effects on IQ, autism susceptibility, and behavioral domains.

Results: The "probability of being loss-of-function intolerant" (pLI) best explains the effect of CNVs on IQ and autism risk. Deleting 1 point of pLI decreases IQ by 2.6 points in autism and unselected populations. The effect of duplications on IQ is threefold smaller. Autism susceptibility increases when deleting or duplicating any point of pLI. This is true for individuals with high or low IQ and after removing de novo and known recurrent neuropsychiatric CNVs. When CNV effects on IQ are accounted for, autism susceptibility remains mostly unchanged for duplications but decreases for deletions. Model estimates for autism risk overlap with previously published observations. Deletions and duplications differentially affect social communication, behavior, and phonological memory, whereas both equally affect motor skills.

Conclusions: Autism risk conferred by duplications is less influenced by IQ compared with deletions. The model applied in this study, trained on CNVs encompassing >4,500 genes, suggests highly polygenic properties of gene dosage with respect to autism risk and IQ loss. These models will help to interpret CNVs identified in the clinic.

Keywords: Autism Spectrum Disorder; Copy Number Variants; IQ.

Figure 1.. Methodological pipeline

Probands from the SSC and MSSNG are defined as individuals recruited on the basis of a diagnosis of autism. Siblings and parents from the SSC did not meet diagnostic criteria for autism. 10 individuals from IMAGEN and none in SYS met criteria for ASD (as estimated by the Development and Well-Being Assessment, DAWBA). 1,490 unaffected siblings from the SSC, 3,660 unaffected parents from the SSC and 1,465 individuals from the general population carry at least one CNV ≥ 50kb (Table S1 in the online supplement). ^(a) Microarray quality control and CNV selection and annotation were performed as previously published (15) (Methods in the online supplement); ^(b) The model used and available data for each phenotype are detailed in Table 1 and Tables S7, S8, S9, and S10 in the online supplement. SSC: Simon Simplex Collection; SYS: Saguenay Youth Study; CNV: copy number variants; SD: standard deviation; N.A.: Not applicable; NVIQ: non-verbal intelligence quotient; SRS: Social Responsiveness Scale; ADOS: Autism Diagnostic Observation Schedule; ADI-R: Autism Diagnostic Interview-Revised; CBCL: Child behaviour Checklist.

Figure 2.. Effect of gene dosage on NVIQ and autism susceptibility

(A) Effect-size of a deleted (DEL, red) or duplicated (DUP, blue) point of pLI on NVIQ in autistic probands and unselected populations. Y axis values are z-scores for NVIQ (e.g. 0.2 z-score=3 points of NVIQ). (B) Autism risk conferred by a deleted (red) or duplicated (blue) point of pLI. Y axis values are odds ratios (OR) computed using a logistic regression to explain an autism diagnosis. Controls include unaffected siblings or unselected populations. Replication was performed using autistic probands from the MSSNG dataset and the unselected populations. (C) Autism risk potentially mediated by NVIQ on the pooled dataset. Y axis values are ORs computed using a counterfactual-based mediation analysis on a logistic regression. Direct effects of CNVs are those not mediated by NVIQ. Indirect are those potentially mediated by NVIQ. Percentage of effect mediated by NVIQ = (indirect / total effect) × 100. Total effects are those computed without adjusting for NVIQ. (D) We compared the risk of autism estimated by our model and the risk observed in previous published studies on 16 recurrent CNVs (4, 9, 22) (Table S3 in the online supplement). OR estimates from the model overlap with the 95% confidence interval (CI) of ORs from previous publications for 14 recurrent CNVs. For three CNVs, the horizontal dotted arrows represent the extreme variability for ORs reported in previous publications. Values are detailed in Table S3 in the online supplement. (1) DEL 15q11.2, (2) DUP 15q11.2, (3) DUP 16p11.2 distal, (4) DUP 15q13.3, (5) DUP 16p13.11, (6) DEL 1q21.1, (7) DEL 16p11.2 distal, (8) DUP 22q11.2, (9) DEL 17p12, (10) DEL 16p13.11, (11) DUP 1q21.1, (12) DEL 16p11.2, (13) DEL 15q13.3, (14) DUP 16p11.2, (15) DEL 17q12, (16) DEL 3q29, (17) DUP 7q11.23 and (18) DUP 17p11.2. (E) Genes covered by deletions (red), duplications (blue), or both (black) in autistic probands. The grey line represents the excess of genes encompassed in CNVs in autistic probands relative to the unselected populations. The Y-axis represents the number of distinct genes encompassed in CNVs (each gene is only counted once). The X-axis represents the number of individuals. The mean and 95%CI confidence interval were obtained using 1,000 iterations (bootstrap procedure). (F) Distribution of the estimated effects of deletions (red) and duplications (blue). Full line: Estimated effects computed for all CNVs ≥ 50Kb identified in both autism cohorts. Dotted line: Autism risk computed for any CNV of 1MB across the genome including at least one gene with a pLI annotation.

Figure 3.. Effect of gene dosage on phenotypic measurements

(A) Effect-size of a deleted (red) or duplicated (blue) point of pLI on the z-scored Social Responsiveness Scale (SRS) in autistic probands from the SSC pooled with their unaffected relatives (siblings and parents) and the unselected populations from the IMAGEN dataset. Effects were measured with and without adjustment for the diagnosis of autism, and NVIQ. The Y-axis represents the estimated effect of pLI on the SRS z-score computed using the mean and standard deviation of unaffected individuals (0.10 z-score = 1.28 SRS raw score point). The analysis adjusting for NVIQ only contains probands from the SSC and individuals from Imagen. (B) Effect-size of a deleted or duplicated point of pLI on the z-score of the Child behaviour Checklist (CBCL) in autistic probands from the SSC pooled with their unaffected siblings. Effects were measured with and without adjustment for the diagnosis of autism, and NVIQ. The estimates were originally computed as OR using a negative binomial regression. The Y-axis represents the estimated effect of pLI on z-scored CBCL computed using the mean and standard deviation of unaffected individuals (0.10 z-score = 1.52 CBCL raw score point). The analysis adjusting for NVIQ is only performed on probands from the SSC. (C, D) Effect-size of a deleted (red) or duplicated (blue) point of pLI on continuous (C) and categorical (D) phenotypes in autistic probands from the SSC unadjusted for NVIQ. Results adjusted for NVIQ are detailed in Figure S4. Y-axis values of panel (C) are measures z-scored using normative data (Table S7) except for DCDQ which was z-scored using the SSC autistic proband group. Y-axis values of panel (D) are odds ratios computed by logistic regression. The significance threshold was computed (51) to account for multiple testing: 0.0027. (✱) Effects significant after adjusting for NVIQ. CTOPP: Comprehensive Test of Phonological Processing; VABS-II: Vineland Adaptive behaviour Rating Scales - Second Edition; DCDQ: Developmental Coordination Disorder Questionnaire.

Figure 4.. Interpretation of the effect of gene dosage on autism risk

Summary interpretation of the differential effects of deletions and duplications on autism risk. Effect-sizes of deletion on most autism specifiers are larger than those of duplications. Duplications and to a lesser extent deletions increase the probability of an autism diagnosis after adjusting for their effect on NVIQ (transparent arrows).