In this study, we present observations on de novo CNVs from the largest published trio-based study of ADHD to date. The overall mutation rate for de novo CNV carriers was 4.6%, which is a similar rate to that observed for two other DSM-5 childhood neurodevelopmental disorders: ASD and Tourette disorder13,14, and somewhat higher than observed for schizophrenia7. Notably, the rate we observed is substantially higher than that which has been reported for controls24 and the previous and only trio-based study of ADHD that observed an overall rate of 1.7%16. Aside from the lower power of this previous, smaller study, one potential explanation for the difference is that, unlike Lionel and colleagues16, we did not exclude those with an IQ < 70. However, in the current study, only three of the 13 individuals who were carrying de novo CNVs had an estimated IQ < 70, with one further individual who had missing IQ data. The de novo CNV rate in children with ADHD without ID (IQ > 70) was 3.5% in our study, compared to the 1.7% rate previously observed16.
The identified CNVs affected a total of 262 genes. None of the genes spanned by the de novo CNVs have been highlighted by the only large exome sequencing study of ADHD to date3 or implicated in ADHD based on a recent, large case–control GWAS2. However, several of the CNVs have been implicated by previously published ADHD case–control CNV association studies (which will include inherited and de novo CNVs); these include the robustly implicated 16p13.11 duplication and the 22q11.2 duplication4,5,6.
The overlap of de novo CNV loci in the regions 15q13.1–13.2, 16p13.11, 16p12.2 and 22q11.21 with loci previously implicated in ASD, DD/ID, Tourette disorder and schizophrenia is consistent with the notion that neuropsychiatric CNVs often have associations with a broad range of neurodevelopmental and psychiatric phenotypes27. Similar findings have emerged from a sequencing study of ADHD, which highlighted strong overlap in the genes implicated in ADHD and ASD3, whereas GWAS demonstrate a moderate (rg = 0.36) genetic correlation between ADHD and ASD28, and a previous CNV study suggested that the same biological pathways are impacted by CNVs in ADHD and ASD29. Our study adds to the growing body of literature supporting the relatively recent reconceptualization of ADHD as a neurodevelopmental disorder, as evidenced by its DSM-5 definition.
However although there is overlap in implicated CNV regions (see Table 2), the type of CNV is usually but not always the same. In particular, we observed a de novo 22q11 duplication and although such duplications have been observed to be associated with risk of other neurodevelopmental disorders, they are protective for schizophrenia22, whereas deletions at this locus (not observed here) are associated with schizophrenia as well as other neurodevelopmental impairments21.
One potential limitation of the study is that the proband DNA was more likely to come from saliva than blood (68.6% samples), whereas the parental DNA samples were more likely to come from blood (28.7% samples from saliva in mothers and 44.2% in fathers). Although the QC metrics were somewhat poorer for the saliva samples, given the careful QC protocol and examination of all the traces (which were definitive; see Supplement), the rate of de novo CNVs in this study is unlikely to be overstated. Collecting blood samples from children with severe neurodevelopmental disorders poses a challenge for genetic research studies. Furthermore, although the comparison of the results of this study to previously published studies is limited by such differences in source of DNA, different genotyping arrays and variable CNV-calling protocols, our focus on well-defined large, rare CNV loci is a robust approach that is unlikely to be biased towards detecting false positives, but may have resulted in missing valid de novo CNVs in the sample.
Although ADHD is highly heritable, complex and polygenic, rare mutation discovery has been slower than for many other similarly heritable and polygenic disorders that include DD/ID, schizophrenia, ASD and Tourette disorder. For example, de novo CNV data are currently only available for around a total of 500 trios (Lionel and colleagues16 and the current study) and there have been no large, trio-based sequencing studies of ADHD. To identify associated variants, as well as damaging and likely causal rare mutations, will require very large sample sizes of ADHD that are not yet globally available.
In summary, we present findings which suggest that de novo CNVs likely contribute to ADHD risk and highlight the need for larger discovery studies for future biological insights.