Main

Your mission, should you choose to accept it, is to use the candidate gene approach to find susceptibility genes for child psychiatric disorders. The hopes of patients and their families are at stake. All you have are a few clinical measures, some DNA, a PCR machine, and a handful of well-worn candidate genes. Good luck.

Mission impossible? Association and linkage studies of candidate gene polymorphisms in common psychiatric disorders now have such a dreadful reputation that many have almost given up hope of success without a major breakthrough from linkage analysis. To date, the only truly robust findings are the associations between the ApoE gene and Alzheimer's disease (and even this was initially flagged by the results of linkage analysis,1 and between aldehyde dehydrogenase 2 and protection against alcoholism.2 There have been many intriguing findings in disorders such as schizophrenia, addictions, affective and eating disorders but so far these have all fallen some way short of the burden of proof required to prove an association, ie, widespread replication combined with functional evidence of causality. There is one ray of hope however, in attention deficit hyperactivity disorder (ADHD). Indeed, ADHD appears to be a shining example of how things might be if the world were an easier place. Not coincidentally, there are six articles in this issue of Molecular Psychiatry on this topic.3, 4, 5, 6, 7, 8

Out of the relatively small number of studies reported so far, evidence for association has been reported and subsequently replicated with the dopamine D4 receptor gene (DRD4), the dopamine transporter gene (DAT1), and the dopamine D5 receptor gene (DRD5).

The most robust of these three findings is the association between the 48-bp VNTR in exon 3 of the dopamine D4 receptor gene and ADHD. That association was first published in the second issue of Molecular Psychiatry in 1996.9 To date, eight published studies have shown positive association between the seven repeat allele and ADHD,9, 10, 11, 12, 13, 14, 15, 36 whereas four are negative,16, 17, 37, 38 giving a respectable but modest genotypic relative risk of around 1.8. These replications were found on a background of a wide range of sampling procedures, including a highly selected population of Ritalin responders to a sample of twin children selected for the top and bottom deciles on a hyperactivity rating scale and a sample of adult ADHD probands using both case-control and within-family tests of association. Two further studies in press add more weight to this body of evidence: Sunohara et al18 examined ADHD family trios from Canada and the USA, and again replicated excess transmission of allele 7, and a second study found evidence of distorted transmission of DRD4 haplotypes.19

At face value, this represents a major achievement in psychiatric genetics: an association finding which has been observed in the overwhelming majority of attempts at replication, most of them published in this journal. In the present issue of Molecular Psychiatry, Holmes et al3 present data from case-control and family trio samples with children diagnosed by ICD-10 or DSM criteria. In line with most previous studies, they found an increase in the frequency of the 7-repeat allele in ADHD probands compared to normal controls. Surprisingly, however, this was not seen when the same probands and their parents were analysed by TDT. Why? There seems to be no obvious technical explanation from the TDT other than reduced statistical power. One explanation is that there is no true association in the sample, and the excess of allele seven seen in the unrelated cases is a result of population stratification. This should be straightforward to test using the procedure for detecting population stratification devised by Pritchard and Rosenberg.20 However, this seems unlikely as an explanation since the frequency of the 7-repeat allele in their case and control samples are very close to those reported in other samples derived from Caucasians of European origin. Furthermore five other groups have reported excess transmission to ADHD probands from heterozygote parents.10, 12, 15, 18, 19

Overall, association between the 7-repeat allele and ADHD looks convincing. But there are still a number of puzzling questions. ADHD appears to be just as common in Asia, where the seven repeat allele is very rare.21 One explanation for this stems from the fact that the 7-repeat allele, when considered on its own, only confers a modest risk for ADHD. The rarity of the allele in Asia may not therefore make a lot of difference to the overall prevalence of ADHD on a background of genetic heterogeneity. The main alternative explanation yet to be excluded is that the 7-repeat itself does not confer susceptibility to ADHD, but is acting as a marker for a nearby functional polymorphism and the pattern of LD across the gene differs in Asian populations. In this scenario allele-7 may not be a major component of risk haplotypes among Asians.

The exon III VNTR is not the only functional polymorphism in the gene. For example, 14 variants have been identified within the upstream regulatory regions, including a single nucleotide polymorphism which has been shown to reduce gene expression by 40% and a 120-bp repeat VNTR. McCracken et al,4 also in the current issue of Molecular Psychiatry, analysed the 120-bp promoter VNTR and found significant evidence for transmission distortion of the longer 240-bp allele, particularly with the inattentive subtype of the disorder. Within the same sample the 7-repeat allele of the exon III VNTR showed a trend towards an excess of transmission but was not significant. Since the 120/240-bp promoter VNTR is in linkage disequilibrium with the exon III VNTR, it is possible that it is the real susceptibility allele (thus explaining the previously observed association with the 7-repeat allele through association). However this hypothesis gained little support from the haplotype analysis in this sample. Another possibility is that one or more of the upstream polymorphisms interact additively with the 7-repeat allele to increase risk for ADHD. This 120/240-bp VNTR may also be unstable, since a 381-bp insertion which may be part of the same allelic system has also been reported in cloned DNA,22 raising the possibility of somatic heterogeneity for allele size in the brain. Detection of this effect will require post-mortem studies.

A role for DRD5 and DAT1 in ADHD susceptibility is less certain at present, since two studies in the current issue do not provide strong support of their roles in ADHD. Both of these genes have shown replicated association with ADHD: four published studies have found association with DAT123, 24, 25, 26 and two with the ‘148’-bp allele of DRD5.9, 22 DAT1 has some real biology as well: not only are DAT1 knockout mice hyperactive with elevated dopaminergic tone, they respond paradoxically to psychostimulants, making them a passable model for ADHD.27 Two studies published in this issue of Molecular Psychiatry, used family trios and the TDT: Holmes et al3 failed to find association between DAT1 and ADHD, and Barr et al5 failed to find association with the microsatellite markers in DRD5, although they did find a small but non-significant excess of the 148-bp allele. Even so, association with both these genes has been found more often than not, and the jury is still out regarding their involvement.

More recently the serotonin system has been considered in ADHD. Serotonin also plays a role in activity: serotonin-releasing drugs such as substituted amphetamines have a strong effect on motor activity. In addition, the paradoxical effect of psychostimulants on the DAT1 knockout mouse model of ADHD is dependent on serotonin, raising the possibility of a role for serotonin in both aetiology and treatment response. Quist et al,26 also in this issue of Molecular Psychiatry, present data showing a marginal association of the 5-HT2A gene His452Tyr allele with ADHD. With several groups now having large TDT samples, there should be no difficulty in confirming or refuting this hypothesis.

What is really needed is a powerful set of concordant and discordant sibling pairs.28 These can be used for linkage to disclose the location of previously unsuspected loci and would also provide a powerful resource for association mapping using both between and within-family tests of linkage and association.29 Independent confirmation of association findings by linkage would be an outstanding result. A comprehensive approach would adopt the dual approach of linkage for initial localisation of genes of moderate effect, followed by association for final gene identification and confirmation and the detection of genes of smaller effect.

Dissecting the phenotype?

One important step forward will be to characterise the effect of these specific polymorphic variants on a range of behavioural and neuro-psychological measures to characterise the range of phenotypic expression and gain insights into the neuro-cognitive processes underlying ADHD. Also in this issue of Molecular Psychiatry, Jorm et al,7 in a carefully performed and powerful study, examined the role of the serotonin transporter promoter VNTR (5-HTTLPR) in 660 children in a longitudinal study of temperament from 4 months to 16 years old. Unfortunately the association they found was in the opposite direction of that expected: between longer alleles of the VNTR and anxiety. A similar approach taken by Rowe et al11 with ADHD and the 7-repeat of DRD4 showed no effect on levels of sub-traits of inattention, or hyperactivity/impulsitivity, although analysis of DAT1 by the Waldman et al study25 found evidence for association and linkage to dimensional measures of hyperactivity and impulsivity. This may be a risky approach in the absence of formal evidence for a genetic contribution to sub-traits of a syndrome; by subdividing an illness one might be moving further away from the underlying genetic effects into the realm of stochastic variation in symptomatology and a minefield of multiple testing. Perhaps it is worth remembering the razor of William of Occam: ‘entities must not be multiplied beyond what is necessary’.

Where to now?

The problem with many of these findings, even when replication looks promising, is demonstrating causality by understanding the biological pathways through which genetic variation causes human disease traits. Proving the precise involvement of alterations in dopamine and serotonin neurotransmission in ADHD will be difficult. There are a number of possibilities, including characterisation of the effects of associated polymorphisms on protein function and gene expression, post-mortem studies which examine gene expression and proteomics, and in vivo imaging methods such as Single Photon Emission Tomography (SPET).

Looking at the brain

SPET is an elegant way of examining neurochemistry, which relies on the emission of photons from nulcides such as 123I attached to a drug or ligand. This provides real time receptor occupancy and ‘off-rate’ analysis in humans. In the cover article of this issue of Molecular Psychiatry, Dahlstrom et al8 have used this approach to examine monoamine transporters in the brains of children and adolescents with and without depression, using the monoamine ligand βCIT. Their finding, that depressive child and adolescent patients had significantly higher serotonin transporter availability in the hypothalamic/midbrain area is intriguing, and has three possible explanations, that of decreased serotonin levels but normal SERT, an increase in synapses or increased SERT density. The first of these, where the ligand is able to gain easier access to the transporter is the most attractive, as it is consistent with work in primates and supports the importance of low serotonin in depression. However the contrast with SPECT studies in depressed adults (where availability is reduced) needs careful explanation.

This approach demonstrates a potentially important way forward for ADHD. Two studies have already used SPECT imaging to examine DAT1 density in ADHD using [123I] altropane,30, 31 and both found substantial increases in transporter density. A positron emission tomography study32 also found dopaminergic abnormalities, high midbrain [18F]DOPA accumulation indicative of altered dopamine decarboxylase activity. A new generation of highly specific SPECT ligands which can pick out individual receptors and transporters promises an exciting way forward for the analysis of neurotransmitter systems implicated in ADHD. Perhaps the most elegant studies will involve genotyping and imaging in tandem so that cases can be selected genotypically as well as phenotypically. However these studies are difficult and expensive to do: administration of radioactivity to children needs careful ethical consideration and analysis of normal control children may not pass this test.

Bring on mighty mouse

Candidate gene studies in obesity have benefited from the explosion in our understanding of the ways in which regulatory systems integrate to regulate appetite and body composition. The catalyst for this revolution was the mouse: the cloning of the genes underlying spontaneous mutants such as obese, tubby and mahogany unlocked the door to a plethora of candidate genes now implicated as causes of human obesity. Likewise ADHD research would benefit tremendously from a deeper understanding of hyperkinetic behaviour, and unlike many psychiatric phenotypes, hyperkinesis is amenable to animal modelling. Analysis of transgenic mice, such as the DAT1 knockout model of ADHD, is a useful approach, but again is candidate gene driven: you must first have the gene to knock it out. The Coloboma mouse is another intriguing ADHD model: this animal, which has a spontaneous contiguous gene deletion including the SNAP 25 gene, a protein integral to synaptic vesicle fusion and neurotransmitter release, also has amphetamine-responsive hyperkinesis which can be rescued by a SNAP 25 transgene.33 Unfortunately alterations in activity levels, particularly in response to novel environments or stimuli are a common consequence of knocking out or mutating genes, so it is not at all clear which, if any, of these mice make good models of the human condition. A far better ADHD behavioural phenotype has been described in a strain of spontaneously hypertensive rats (SHR),34 which unlike most other overactive animals shows normalisation of activity levels when presented with novel environments. Furthermore SHR show an altered response rate to delayed reinforcement schedules seen typically in ADHD children. Both these behavioural measures in SHR respond to methylphenidate.

A powerful alternative approach is to make use of high-throughput mouse mutagenesis, where tens of thousands of mutant mice can be generated and subjected to simple screens for physical, biochemical, behavioural35 and neurological phenotypes. A glance at the ENU mouse mutagenesis program at Harwell, UK (http://www.mgu.har.mrc.ac.uk/shirpadata/) reveals two dozen or so mice with high locomotor activity. With further testing, especially with respect to the effects of psychostimulants, interaction with novel stimuli and altered response to reinforcers, this collection might prove to be a goldmine for functional biology and the identification of candidate genes for hyperkinesis research.