Advances in the Genetics of ADHD

When the Human Genome Project was initiated, among its main promises were the identification of disease-relevant genes, a reorganization of diagnostic processes, and the translation of that information into new and more effective treatment strategies (1). There are several clear examples of successes of the inclusion of genomics in medicine (2), which serve as templates for future genomic investigations (3). To date, however, genomics has had less of an impact on psychiatric disorders, and it has not yet led to new FDA-approved psychiatric treatments. Two articles in this issue make inroads for psychiatry by using attention deficit hyperactivity disorder (ADHD) as a model to demonstrate the potential impact of human genomics.

Most psychiatric illnesses are relatively common, with lifetime prevalence rates in the range of 1%–40% of the population. By the mid-1990s, the common disease–common variant hypothesis was developed, suggesting that the genetic component of common diseases would be explained by common variants of a relatively limited number of genes, with each variant having a moderate or larger effect (4). For several diseases—including diabetes, macular degeneration, and Crohn's disease—this approach has been successful, with common variants of three to 100 genes explaining a significant portion of the heritability of the disorder (2). However, genomic contributions to psychiatric diseases have proven more difficult to elucidate. Psychiatric illness risk may be associated with a much larger number of genes, with each gene variant making only a very small contribution to genetic vulnerability. For example, although genetic contributions may account for up to 80% of the risk for schizophrenia, a recent analysis suggests that much of that risk may be explained by common variants of perhaps thousands of genes, each variant of extremely small effect, with even the largest effects accounting at most for 1%–2% of the genetic risk (5). If this is true, it may explain why large genome-wide association studies have failed to replicably identify specific genetic contributions.

The multiple rare variant model is an alternative genetic model in which genetic contributions to disease are related to rare variants, with the average variant having a larger effect (6). However, if the disease is common and the variant is rare, only a small percentage of cases will be attributable to any specific variant, and association between disease and any specific rare variant will be difficult to demonstrate. One strategy is to focus on large deletions or duplications in the genome. This approach has two advantages. First, these deletions or duplications are large enough to cover an entire gene or multiple genes, meaning a decrease or increase in the number of copies of whole genes within an individual; this variation in the number of whole copies is termed copy number variants (CNVs). The assumption is that this alteration in the number of whole copies of a gene will either decrease or increase protein production to a level where the effect is moderate to large. Second, while the specific beginning and end points of the deletion/duplication may vary across individuals, the deletions are large enough that they overlap, and individuals with overlapping deletions/duplications can be grouped together.

Williams and colleagues, reporting in this issue (7) used the CNV approach to examine large (>100 kb), rare (frequency <1%) variant contributions to ADHD in over 700 children and adolescents compared with a representative U.S. sample of over 2,400 individuals. All individuals were of European ancestry. Across the entire genome, the ADHD group had 1.15 times as many large CNVs as the comparison group. This significant elevation in CNVs appears to be due primarily to an increase in large duplications in genomic areas that include known genes and, in particular, near genes that have been identified as potentially involved in schizophrenia or autism. There were no significant differences in duplications in genomic areas without known genes or in deletions across the genome.

While the total number of large duplications across the genome was higher in the ADHD group, there were no specific loci with a significantly greater rate of duplications. There was a nonsignificant suggestion of increased duplication rates at chromosome 15q13.3, the location of CHRNA7, the gene for the α7 nicotinic receptor. To further investigate this suggestive funding, 15q13.3 duplication rates were assessed and found also to be increased in a replication sample drawn from four mostly independent studies of ADHD. In sum, these results suggest that large, rare duplication variants are associated with ADHD. CHRNA7 duplications are one likely source of increased ADHD risk, occurring in 1.25% of cases, compared with 0.61% of comparison subjects.

A second article in this issue, by Stergiakouli and colleagues (8), went one step further in their genetic analysis, trying to integrate the common disease–common variant and multiple rare variant hypotheses. Using standard single-nucleotide polymorphism (SNP) genome-wide association analysis (which focuses on common variants) with a subsample of the cases utilized by Williams et al. combined with a large Icelandic sample, Stergiakouli et al. obtained the expected result: there was a suggestion of several loci where common variants might be associated with altered ADHD risk, but when adjusted for multiple testing, no variant met statistical standards for clear association. Combining genes into a smaller number of gene sets (based on metabolic pathways or similar evolutionary origin) led to similar results—several pathways or gene sets that may be involved, but no associations reaching statistical significance. Suggestive gene sets identified from the genome-wide association study (the common variant approach) appeared to congregate in the same pathways that were also identified as potentially important by increased rates of CNVs (the rare variant approach), consistent with the hypothesis that both rare larger-effect variants and more common smaller-effect variants contribute to genetic vulnerability (6). The only significant finding for convergence between SNP and CNV analysis for a single gene occurred at CHRNA7, the same gene identified by the Williams et al. group. Stergiakouli and colleagues conclude that given larger sample sizes and increasing understanding of neurobiological pathways, genetic approaches will further clarify genetic variants associated with risk for ADHD. Although this convergence of information is helpful, both articles note that for families that carry either the CHRNA7 duplication or the CHRNA7 SNP associated with the illness, it is still not clear how often ADHD (or other illnesses) occurs in children who inherit them.

Identifying etiologic genetic contributors to neuropsychiatric illnesses such as ADHD has proven more difficult than anticipated, providing less potentially clinically relevant information than has been the case for many nonpsychiatric medical illnesses. The hope is that genetic information will provide new targets for medication development, clarify the mechanisms of the illness, and perhaps lead to biomarkers that identify relevant subtypes of the illness. CHRNA7 has also been implicated in other psychiatric conditions, including schizophrenia, bipolar disorder, autism, and nicotine addiction (9–12), suggesting that overlap in etiology may require the field to rethink its diagnostic classification and that the development of drugs targeting nicotinic receptors for treatment of these other illnesses may accordingly have some transferability to ADHD. Williams et al. note that reactions to stress and the release of dopamine have both been associated with CHRNA7 and the α7 nicotinic acetylcholine receptor, the product of the gene. They also note a modest association of the CHRNA7 CNV with conduct disorder, an ADHD comorbidity that often foreshadows more severe adult psychopathology, including sociopathy and substance abuse.