Genetic influences on attention deficit hyperactivity disorder symptoms from age 2 to 3: A quantitative and molecular genetic investigation

Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder characterised by pervasive, age inappropriate behaviours of inattention, hyperactivity and impulsivity. The current definition of ADHD defines the age of onset of impairing symptoms as occurring before the age of 7 years, although formal diagnoses are not usually made before this age. However, early characteristics are good predictors of later appearing behavioural problems [1] and therefore, employing research strategies to identify developmental aetiological factors in young children remains important. It is well established that ADHD in children is highly heritable with estimates averaging at ~76% [2], with the same being true of ADHD symptoms in pre-school children [3]. However, genetic variation underlying these observed heritabilities is still not well understood.

Candidate gene studies in children have focused predominantly on genes of monoaminergic neurotransmitter systems, particularly dopamine. The main genes of interest in this research have been the dopamine transporter gene (DAT1) and dopamine receptor genes (DRDs). These choices have been informed by a dopamine hypothesis of ADHD, which stems from the action of stimulant medications such as methylphenidate and dexamphetamine which increase levels of available synaptic dopamine. These studies have proven relatively fruitful with robust associations between DRD4 and DRD5 with ADHD being identified in meta-analysis [4]. More recently, whole genome association analyses in both children and adults have provided some information on potential new candidates for follow up [5–7]. Of particular interest is the convergent finding of association with variants within CDH13, a gene that lies within the ADHD linkage region on chromosome 16p [5, 8]. This has provided new insights into the underlying genetics of ADHD and has allowed for new hypotheses to be formed for future research. However, there have been fewer molecular studies in preschool children, although there is some evidence to suggest that candidate genes from various neurotransmitter systems such as DAT1, synaptosome-associated Protein 25 (SNAP25) and the noradranaline transporter (NET1) may have some involvement [9].

It is apparent that these genes are not necessarily acting on the ADHD phenotype consistently throughout development, with a number of studies suggesting that although there is a general genetic stability across time from ages 2 through to 4 years [10]; 2, 3, 4 and 7 years [11]; 3 through 12 years [12] and 8 through to 14 years [13], there is also age-specific genetic variance. The implications of this are that association studies using heterogeneous samples are potentially losing information on age-specific effects of genotype on ADHD. Further, with the need for replication across studies it becomes very difficult to identify the causes of non-replication due to differences in sample demographics.

We have recently reported high heritability and genetic association between specific risk alleles and ADHD symptom scores in a population sample of 2-year old twins, with modest evidence of association being found for DAT1 and NET1 [14]. In the present analysis we have used the same sample to assess the degree to which genetic effects on ADHD symptoms are stable from ages 2 to 3 using quantitative genetic techniques. In addition to this analysis, we have studied previously reported ADHD risk alleles to identify any age-specific genetic associations. Candidate gene variants were chosen based on previous positive association with ADHD in either clinical or quantitative trait locus (QTL) analyses. Given the nature of the analyses we hypothesised that there would be substantial genetic overlap in ADHD symptom scores across ages, which would translate into a number of genetic variants at age 2 also being associated at age 3.

From the Cholesky decomposition model (Figure 1) we can estimate the degree to which A, C and E contribute to the co-variance of ADHD symptoms across time. C has been omitted from Figure 1 because of the lack of significant C on the variance at either age. All path estimates are provided from the most parsimonious AE model.

A large proportion of the additive genetic variance at age 2 was shared with that at age 3 (Figure 1), although there remained emerging age-specific effects (Figure 1). Indeed, dropping the age 3-specific A path from the Cholesky decomposition model resulted in a significant worsening in fit (χ² = 12.263, df = 1, p < 0.01), suggesting a contribution of genetics to both phenotypic stability and change. The effect of E on the covariation between ages was small, yet significant (Figure 1). Using unsquared path estimates from the Cholesky decomposition model, we can estimate the correlation between ADHD symptoms at age 2 and 3. In this case the phenotypic correlation between ages is calculated as (√0.79 × √0.48) + (√0.21 × √0.01) = 0.67. Additive genetic influences account for 93% of this correlation (bivariate heritability = ((√0.79 × (rG = 0.78) × √0.79)/0.67) × 100 = 93%).

In this study we investigated the genetic relationship between ADHD symptom scores at two time points in infancy. Consistent with previous reports we found ADHD scores to be highly heritable at age 2 and 3 years, providing evidence for the involvement of additive genetics on the variance of these measures, as well as identifying them as viable measures for molecular studies. Intraclass correlations for our ADHD measure were suggestive of predominantly additive genetic influences at both ages. However, the literature is mixed with regards the effects of dominance and contrast effects, a feature of ADHD that is often found in samples of older children [21]. Dominance and contrast effects are characterized by DZ correlations that are lower than half MZ correlations, and while there is evidence for dominance in symptoms of overactivity in young children [22], there is no evidence for these effects in other studies of activity and attention problems [23]. In light of the power needed to detect dominance and contrast effects [24] and given the lack of evidence for these effects in this study, we did not formally test for them, although future research in large samples using similar measures are needed to clarify this issue.

Phenotypic stability of ADHD symptoms across ages was moderate, producing inter-age correlations of 0.51 - 0.62 (twin 2 - twin 1), which is consistent with previous reports using samples of this age range [10]. The suggestion here is that while symptoms are consistent across ages for the most part, there remains developmental change, which is reflected in the newly emerging additive genetic variance at age 3, a variance component that is unaffected by error associated with fluctuations in evaluations. Prior research has shown a level of genetic stability on ADHD traits across numerous age ranges, including very young children [10, 11]. Our analyses concurred with these findings as we found that genetic effects at age 2 are largely shared with those acting at age 3. The suggestion here is that genetic variation that influences variance in ADHD scores at age 2 will be the same as those acting at age 3, on the most part. Having said that, unique effects of additive genetics at age 3 are significant, so while there is substantial genetic continuity across ages, emerging effects cannot be ignored. Unfortunately a limitation of this study was the limited power to assess sex × gene interaction effects in the quantitative analysis. This is an interesting area of research and one that should be considered in future research with more powerful samples, although at present there is little evidence for gene × sex interaction, at least in symptoms of overactivity [22].

Given the results from our quantitative analysis, it is interesting to consider the results of our molecular genetic analyses. At age 2, we found modest, nominally significant (p < 0.05) associations with four variants (DAT1 3'UTR VNTR, rs11568324, rs3785157 and rs998424). Although there were some associations in common at age 3 (DAT1 3'UTR VNTR and rs3785157), the association between ADHD scores and rs11568324 at age 2 did not replicate at age 3. Further, an age-3-specific association was observed with the DRD4 exon 3 VNTR and one SNP in 5-HTT (rs11080121), findings that are consistent with our quantitative genetic results. Although suggestive at this stage, these findings highlight problems of age-specific genotypic effects that may occur in demographically heterogeneous samples. We may speculate that these differences in genetic association are due to new effects emerging at age 3, implying developmental specificity in which phenotypic consequences of DNA polymorphisms are effectively masked until a particular developmental stage is reached. There are, however, alternative explanations. It might be that subtle differences in ratings between ages causes some manner of spurious association at either age independently, an issue that relates largely to the power of the sample and increases the chance of type I and II errors. In any case, from our analyses it is apparent that there are age-specific effects of genotype on ADHD symptom scores and is thus a factor that should be considered in genetic studies.

An interesting comparison to be drawn is one between this study and an analysis carried out by Mill et al. [9], who conducted a similar analysis in a population-based twin sample. Although they used a composite measure of ADHD symptom scores across 2, 3, 4 and 7 years for the main analysis, they also reported some individual time-point data. DAT1 was found to be associated with ADHD symptoms at ages 2 and 3, and our report therefore serves as a replication of these findings.

A further point for discussion is the observed difference between the AT and AW tests of association. At age 2, rs3785157 and rs998424 were significantly associated (nominal p < 0.05) only in the AW test. Given the increased power of the AT test to detect association in the absence of stratification, these results may be surprising, and may reflect between-family differences in child ratings. We are, however, unable to assign this observation to any stratification effects because of a non-significant finding in the AP test. This raises issues regarding the power of the sample to detect stratification and makes it difficult to conclude that there are in fact any significant differences in the between and within family components of association. However, of interest is that at age 3, larger discrepancies in effects of these two markers were observed between the AT and AW tests, an observation that is apparent in the AP test which displays significant evidence of stratification. This phenomenon is also seen for associations with the DAT1 3'UTR VNTR and DRD4 VNTR at age 3, where there is a decrease in p-value in the AW compared to the AT test, albeit with no significant difference in the AP test. Taken together, we conclude that there is evidence for stratification effects, an observation that is not unique to this study [9] and which may reflect between-family differences in rating styles. In particular, it is interesting to note that the pattern of DAT1 3'UTR VNTR associations in this study are the same as those observed by Mill et al. [9]. Both studies display greater significance for the AT than AW test at age 2, with the reverse effect at age 3. The suggestion is, therefore, that there may be new stratification effects emerging at age 3 that could contribute to the observed age-specific genotypic effects.

A major limitation of this study is the power of the sample to detect genetic association, especially if we consider convincing levels of significance to be in the order of p < 5 × 10^-7 [25]. Using the genetic power calculator http://pngu.mgh.harvard.edu/~purcell/gpc/ we estimated that the sample had 47% power to detect a QTL affecting 1% of the phenotypic variance and 71% power to detect a 5% QTL. Despite being underpowered, we detected nominal significance for a number of polymorphisms at ages 2 and 3, and although we cannot rule out the possibility of false positives, the study serves as a proof of principle, in that age-specific effects of genotype on behavioural measures is an issue to be addressed, especially in underpowered samples.

In this study we investigated the genetic relationship between ADHD symptom scores at age 2 and age 3. Although we found that the majority of genetic effects were shared across ages, there was room for some age-specificity. These inferences were borne out in the molecular genetic analyses, whereby associations seen at age 2 replicated at age 3. However, some observed associations were age-specific, which highlights this issue as an important one to consider in genetic association studies.

This report indicates that although the majority of genetic effects on ADHD symptom scores at age 2 are stable through to age 3, there remains significant emerging effects. As well as enabling us to better understand how genes contribute to the aetiology and origin of ADHD, the report also serves to highlight the importance of demographic homogeneity in molecular genetic studies.

The authors declare that they have no competing interests.