After separating the 47 items from the SDQ and Conners symptom ratings into the two dimensions of aggressive and cognitive impulsivity, we used the FBAT-PC software to evaluate their possible associations with 582 SNPs distributed through 14 relevant genes. The genes were selected because their expression affects 5-HT activity and closely-related neural functions and they have previously been reported to be associated with a diagnosis of ADHD. The focus on 5-HT-related features was based on descriptions of the relevance of this transmitter to the expression of impulsivity [4, 26, 53] and for the preferential transmission of polymorphisms influencing 5-HT function in ADHD (see introduction). Gene-wide significance was achieved only for a variant of the gene encoding the enzyme for the synthesis of adrenalin (PNMT) in cognitive impulsivity. However nominal significance was shown for genes affecting a range of elements of the 5-HT and DA systems in both forms of impulsivity, and that have previously been reported to be associated with ADHD.
The absence of prominent associations for neurotransmitter markers, apart from that with adrenalin synthesis through PNMT variants, was surprising. A gene-wide significant association was found for one SNP for PNMT with cognitive impulsivity, and a second was nominally associated with behavioral aggressive impulsivity from the 4 SNPs for PNMT that were examined. In both instances the trait was over-expressed (table 4). This was partly anticipated by a similar significant finding in a candidate gene analysis for association with ADHD from the IMAGE sample . PNMT is the rate-limiting enzyme in the synthesis of adrenalin that has transmitter and endocrine influences on brain function which can be stimulatory or inhibitory, respectively . The association of adrenalin with CNS arousal is significant in the context of the common attribution of hypoarousal to ADHD function (review: ). Further, one might anticipate that arousal levels would be pertinent to both forms of impulsivity under investigation here, but especially the cognitive attention-related form that could be a precursor to the behavioral, aggression-related form. Indeed, adrenalin levels in ADHD children are usually lower than normal [56–58] unless they show anxiety [59–61]. Low levels are associated with restlessness and aggressive outbursts . Response to treatment with methylphenidate tends to normalize levels of adrenalin [58, 63] and improve autonomic measures of low arousal (skin conductance levels: ). Thus there is broad support for a widespread role for impaired function of arousal and adrenalin systems in ADHD and the impulsivity that is characteristic of the disorder. From our current results, we would predict that this could be based in an altered expression or function of the PNMT gene. Considering the potential parallel involvement of 5-HT systems in impulsivity, and the demonstrated role of 5-HT in the control of autonomic function  it would be useful to look for tagging between variants influencing PNMT and 5-HT receptors.
In agreement with our expectations, analyses of the two forms of impulsivity are reflected at least nominally in similar but also in different associations with the genetic markers for monoamine activity. For the 5-HT system, association with synthesis (TPH2) and receptor-mediated mechanisms (HTR1E) may be relevant to both forms of impulsivity. Indeed the same SNP for HTR1E (rs1406946) was nominally associated with both types of impulsiveness in the dominant transmission model. It was over-expressed in the cognitive and under-expressed in the behavioral or aggressive form of impulsivity (table 4). The difference for 5-HT function is one of emphasis, with a number of markers pointing to small effects on synthesis (TPH2 3 SNPs), pre-and post-synaptic receptors (HTR1E, HTR2A, HTR3B: 5 SNPs) and reuptake (SERT: STin2) for cognitive (not behavioral) impulsivity. For cognitive impulsivity 4 SNPs for receptors (apart from HTR1E) were under-expressed in a recessive model, whereas the availability of 5-HT (e.g. TPH2 in a dominant model) was influenced by over-expression of relevant alleles. For the DA system, nominally significant markers were found for both types of impulsivity, but while the emphasis for the behavioral/aggressive form was on DRD4 (one SNP and in particular the 3-repeat allele was over-expressed), cognitive impulsivity was nominally associated with uptake (DAT1: 1 SNP) and the D1 receptor (DRD1: 2 SNPs), which were all under-expressed. Based on their anatomical distributions, this could reflect top-down mesocortical and bottom-up mesolimbic mechanisms of control, respectively .
The 5-HT component implies a potential influence of 5-HT activity in those children with ADHD who show aggressive behavioral and cognitive impulsivity both through its availability (TPH2-mediated synthesis and SERT-mediated uptake), and the efficacy of 5-HT action via the 5-HT2 and 5-HT1 receptor families . Generalizing in this context, activation of the 5-HT1B and 5-HT2A receptors tend to be 'stimulatory', while the 5-HT1A and 5-HT2C sites are more often 'inhibitory' . The 5-HT receptor mediated influence obtains especial significance with regard to the interaction of 5-HT with the DA system considering that ADHD is often considered a 'hypodopaminergic' disorder [69, 70] and the 5-HT and DA systems interact directly and widely through subcortical and cortical regions . Even if the genetic effects reported here are modest, there is intriguing support for our hypothesis of opposing types of interactions of DA and 5-HT in the two forms of impulsivity. Thus, in aggressive behavioral impulsivity a DRD4 SNP was overexpressed while those for HTR1E and TPH were underexpressed. In contrast in cognitive impulsivity SNPs for DRD1 and DAT1 were underexpressed while others for TPH2 and HTR1E were overexpressed.
Corrected significant associations for the 5-HT2A gene have also been reported recently for children and adults with ADHD  and correlations with hyperactive/impulsive symptoms were described for a normal population . However, other groups have seen no association [73, 74]. Unfortunately, little is known about the functional role of 5-HT1E sites that make up a quarter of those binding 5-HT in the forebrain in autoradiographic postmortem studies of cortical and hippocampal regions (CA3-4 >> CA1 ). In contrast, 5-HT3 receptors have been more widely studied, and are abundant in limbic and frontal regions . In contrast to the inhibitory functions attributed to HT1 receptors, stimulation of HT3 sites enhances DA release . Thus the nominal significance of the association of 4 SNPs for stimulatory receptors (HT2A and HT3B) with cognitive impulsivity is consistent with the opposing effects of 5-HT activity on the two forms of impulsiveness proposed in the introduction.
The presence of 3 SNPs for TPH2 with near gene-wide significant associations with cognitive impulsiveness along with the significant relationship for a SERT allele (as also indicated by ) would suggest that markers for the presynaptic availability of 5-HT merit continued attention. The importance of this is illustrated, for example, by the functional consequence of changes in SERT activity as expressed with the short allele. In this instance the binding potential of 5-HT receptors may alter, as demonstrated for the 5-HT1a binding site . In turn changes at this binding site (with inhibitory function) would strongly modulate transmitter release in DA systems. Considering the well-known interactions of these two monoaminergic systems , the potential for tagging between genetic loci should be sought in future studies.
Lastly, we studied SNPs representing 3 other genes (BDNF, FADS2, NURR1/NR4A2) on the grounds that their activity could influence neurotransmission including that based on 5-HT, and that some associations with ADHD have been reported . Brain derived neurotrophic factor (BDNF) facilitates the development of the release mechanisms for several neurotransmitters. It is known to have age-dependent effects on 5-HT transporter function . In BDNF-knockout mice this led to reduced 5-HT metabolism  and aggressive impulsive behavior . A nominal association with both forms of impulsiveness emerged from the present analysis, that may reflect a small effect on the development of monoamine systems in general for cases of ADHD that particularly reflect an etiology of delayed maturation . However, there is not much support for an association of BDNF markers with the diagnosis of ADHD . NURR1/NR4A2 is an orphan nuclear receptor and transcription factor crucial to the genetic control of development of DA neurons [85, 86]. The expression of NURR1/NR4A2 is essential for the induction of several markers of DA activity (e.g. tyrosine hydroxylase). However, there was no evidence here for the altered expression or function of NURR1/NR4A2. Finally, fatty-acid desaturase gene-2 (FADS2) is involved in regulating membrane synthesis and function and at least one polymorphism has been closely associated with ADHD . The ratio of omega-3 to omega-6 unsaturated fatty acids in neural membranes certainly influences their fluidity and the ability of neurons to conduct efficiently: this is a problem for ADHD with documented white matter anomalies  and potentially for the function of oligodendrocytes responsible for membrane synthesis . But we found no evidence that the FADS2 markers investigated were associated with the types of impulsivity studied here.
Limitations of the study
The design of the study involved recruiting children and adolescents who showed to a greater or lesser extent the features of impulsivity we sought. Alternative approaches could contrast children with and without these features, and compare pre-, post-pubertal and young adult subjects to investigate putative developmental changes in gene expression. A comparison of groups with high comorbidity for externalizing or internalizing symptoms would be instructive both for the integrity of our concepts of impulsivity in affect and cognition, as well as for identifying the grounds for separating subgroups of ADHD with 'comorbid features'. We would also point out the severe limitations of interpreting symptoms based on lay responses to questionnaires. It is important that the preliminary work described here be extended to neuropsychological testing that would contrast not only provoked aggression with impulsive decision making, but incorporate the role of reinforcement (e.g. delayed gratification).
There are also several limitations with regard to the sample and genotyping. Although this sample is reasonably sized for moderate genetic effects, the number of individuals could be larger to detect the modest genetic effects that one would expect to observe here. In addition, individuals enrolled in the IMAGE study are from several different places throughout Europe, including Spain and Israel. Although the FBAT statistic is robust to population stratification, the PBAT screening algorithm can be affected by this. Finally, current genome-wide scans are suggesting many genes that one could not have suspected of being of interest a short time ago. Therefore some candidate genes that may now seem important were not included in this analysis.