ADHD: Progressing From Genetic Discoveries to Biological Insights

Attention deficit hyperactivity disorder (ADHD) is one of the most heritable of all neuropsychiatric disorders (1). Despite being so highly disruptive and impairing, relatively few effective treatments are available, and outcomes tend to be poor. Gaining insight into the pathophysiology that underlies ADHD represents a crucial first step toward developing novel and effective modes of diagnosis and treatment. However, the complex multifactorial nature of psychiatric disorders, coupled with inaccessibility to live brain tissue, poses a serious challenge to investigations of the biology of these disorders. Genetic studies provide a useful strategy for gaining insight into biology, and this understanding has provided a strong impetus for conducting genome-wide discovery studies of many different psychiatric disorders, including ADHD (2).

Recent genome-wide association studies (GWASs) have shown that ADHD is associated with multiple rare and common genetic variants (1, 3, 4). Deleterious mutations of large effect size tend to be rare because they are more rapidly removed from the population by natural selection, while common variants by themselves have small effect sizes. Rare mutations thus are especially interesting from the perspective of offering a window into biology. However, discoveries from genome-wide discovery studies represent only the first step. That is because an association between ADHD and a gene variant does not necessarily tell us which specific gene or genes are involved or the biological mechanisms that might be involved (2). Typically, GWASs only identify genomic regions in which potential risk genes lie. Thus, further work is needed.

The article by Harich et al. in this issue (5) sets out to prioritize genes implicated by GWASs of rare copy number variants (CNVs) in ADHD. CNVs are one class of rare gene variant that involve alterations to chromosomal structure. Deletions and duplications of DNA have been observed to be associated with ADHD, other neurodevelopmental disorders such as autism spectrum disorder (ASD), intellectual disability, and Tourette’s syndrome, as well as psychiatric disorders such as schizophrenia and depression (6). Like all classes of genetic variation, CNVs show marked pleiotropy across different psychiatric disorders, and there is also prominent variation in the severity and nature of clinical features manifested by carriers (7, 8). For example, CNVs associated with ADHD overlap with those implicated in ASD and schizophrenia (9–11).

Although CNVs at different locations on the genome have been reported to be associated with ADHD, no study to date has been large enough to implicate individual CNVs at a genome-wide significant level, with the exception of one pooled analysis of 15q13.3 duplications (12). Another challenge is that CNVs typically encompass multiple genes. Thus, the first tranche of case-control CNV studies of ADHD have not led yet to the identification of risk genes. The article by Harich et al. addresses this gap and moves the field of ADHD genetics forward. The authors identify “high-priority” genes by taking the CNVs reported in 11 published studies of ADHD and using a comprehensive set of bioinformatic approaches to analyze them. They capitalize on publicly available biological data resources and also examine convergence of results across different methods. Triangulation of evidence across different designs and methods is increasingly recognized as crucial for epidemiological and etiological research (13), and here we see its application in genetics.

The authors initially identified 2,241 potential genes within CNVs reported to be associated with ADHD from the 11 published studies. This is a prohibitively large number of genes if the aim is to investigate gene function in the laboratory. The authors’ aim was to select the most plausible genes from this very long list in different ways. They first included only genes that were completely deleted or partially truncated or were entirely duplicated. Some CNVs overlap and thus cover even larger stretches of DNA, and the degree of overlap is variable in different individuals. The authors narrowed down the region of interest by selecting the minimal region common to all these overlaps. They then selected messenger RNA coding genes that were present in at least two cases and excluded those observed in control subjects. This yielded a list of 432 high-ranking genes.

Next, the authors used an atlas known as BrainSpan to test the biological plausibility of these high-ranking genes. Genes contain the instructions for manufacturing proteins, but this process first requires DNA to be transcribed to (expressed as) messenger RNA. The BrainSpan atlas maps the RNA profile of human brain across development, including during the prenatal period (14). The authors found significant evidence of enrichment for coexpressed genes among the high-ranking genes; that is, the high-ranking genes appeared to be expressed together in the brain at the same time. Further analysis of these genes using a protein-protein interaction database identified several protein networks that involved 62 proteins.

One interesting aspect of this study is that the authors further conducted a set of alternative cross-species analyses to test the plausibility of the 432 high-ranking genes. Using cross-species data, they observed that 18 of the 432 priority genes were associated with cross-species phenotypes that resemble features of ADHD (e.g., hyperactivity). Examination of a cross-species biological database of interactions enabled the authors to identify an additional 48 genes that interact with these 18 genes, providing a total of 66 priority genes implicated by cross-species data.

Finally, the authors tested for convergence between the human and cross-species approaches and identified a final list of 26 highest-confidence ADHD genes that were observed across all these approaches (human and cross-species). These highest-confidence genes also showed significant evidence of being coexpressed in brain tissue, in the prenatal period and later in development, and over half of the genes showed broad expression patterns across different tissue types (i.e., not just brain). They also highlight that the higher-confidence genes link to several interconnected protein networks that cross human and other species.

Rare CNVs are not the only class of genetic variation associated with ADHD risk. Rare sequence variants and common gene variants have also been shown to be associated with ADHD (1). The most recent and largest published GWAS identified 12 genome-wide significant loci (3). Given the polygenic nature of ADHD, a key question was whether any of the highest-confidence CNV genes identified by Harich et al. overlap with genes implicated by common gene variants. The authors examine this and identify two of their priority genes, POLR3C and RBFOX1, that map within genomic regions showing common variant associations.

This study is an important one, as it moves the field forward, one step closer to biological insights from ADHD genetic discovery. The authors conducted an elegant and comprehensive set of bioinformatic and statistical analyses, using published CNV data, to infer genes that are most plausibly involved in ADHD pathogenesis. The next steps will be to conduct laboratory-based animal and cellular studies of the prioritized genes to fully understand the molecular and developmental mechanisms involved. Laboratory work examining the function of genes is expensive and time-consuming, so prioritizing genes in this way is invaluable.

As the authors point out, there are limitations. First, studies of ADHD rare mutations trail behind other neurodevelopmental disorders (e.g., ASD, intellectual disability) and many psychiatric disorders (e.g., schizophrenia), despite being so heritable. Discovery sample sizes remain small when very large samples are needed to identify genome-wide significant loci (because of allowing for multiple testing). The CNVs examined in this study are not genome-wide significant. Also, the authors had to rely on case-control CNV data, which means the CNVs will include ones that are inherited from the affected offspring’s parents and those that arise de novo. Rare de novo mutations in offspring, when parent-offspring trios are studied, are especially powerful because these mutations are more likely pathogenic and thus can provide important biological insights. Yet, unlike for ASD, intellectual disability, and schizophrenia, there have been very few such studies for ADHD. Another challenge faced by the authors is that the original 11 published CNV studies all used different approaches in defining their CNVs. Furthermore, CNVs do not work in isolation. ADHD is multifactorial in nature; multiple genetic variants as well as environmental risk factors contribute. Finally, while the approach used by the authors is important and the highlighted genes help prioritize ones of high interest, the study cannot confirm that these genes are causally involved in ADHD. This means that much more work, including further laboratory work, will be needed before findings result in clinical translation.

How soon will this type of genetics research help clinicians and patients? The answer is that while there have been rapid and enormous advances with genetic discovery, initial genetic association findings do not tell us about what specific genes and mechanisms are involved, so the approaches used by Harich et al. represent an important step in understanding the biology of the disorder. Continued gene variant discovery in larger patient samples, further bioinformatic analyses, and laboratory-based and clinical studies will all be needed not only to uncover pathophysiology and test interventions but also to move toward evaluating potential clinical applications.