The dopamine metabolism in the brain is a function of the presynaptic synthesis, release and reuptake of dopamine. Dopamine is synthesized from the amino acid tyrosine, which is first converted to L-dihydroxyphenylalanine (L-DOPA), and then to dopamine by the enzyme DOPA decarboxylase. The synthesis has been studied in vivo in subjects with ADHD with [¹⁸F]fluorodopa in two PET studies [19,20]. In the first of these studies, the authors found that adult subjects with ADHD had reduced levels of [¹⁸F]fluorodopa in the prefrontal cortex in comparison with control subjects. In a second study, they instead reported an increased level of [¹⁸F]fluorodopa in the right midbrain in adolescents with ADHD. Three independent groups, using SPECT, have reported increased density of the dopamine transporter (DAT) in the striatum in adults and children with ADHD [21-23]. However, these findings have not been replicated by other groups reporting unaltered DAT binding in the striatum [24] and lower DAT binding in the midbrain [25]. Hence, previous PET and SPECT studies indicate that there is an abnormal dopamine metabolism in subjects with ADHD, although the nature of the disturbances remains to be clarified.

In the present study we hypothesized that the behavioural problems of subjects with ADHD are associated with altered dopamine synthesis. We used the radio L-[¹¹C]DOPA, (chemically identical to native L-DOPA), which is transported into the presynaptic neurons. There it is converted by the enzyme DOPA decarboxylase to L-[¹¹C]dopamine and stored in storage vesicles. Therefore, PET data obtained by L-[¹¹C]DOPA reflect DOPA decarboxylase activity and dopamine storage processes. The properties of the radiotracer L-[¹¹C]DOPA differ in several important aspects from its more commonly used fluorine labelled analogue [26]. The turnover rate of L-[¹¹C]DOPA is state-dependently modulated by drugs that affect dopaminergic systems and is thus also a measure of the functional state of the dopaminergic neurons [27-29].

The PLS regression coefficient indicates which variables are most responsible for the separation of the two groups (Figure ​(Figure2).2). Such coefficients are interpreted in the same way as in ordinary regression analysis and, consequently, positive coefficients indicate positive correlations and the size of the coefficient reflects the strength of the correlation. In this case, the dependent variable is a discriminant variable denoting the ADHD diagnosis, so that positive coefficients indicate increases in comparison with the control group, and negative coefficients indicate decreases. Thus, the dominance of negative columns in Figure ​Figure22 shows that ADHD subjects generally have lower Ki. Furthermore, the regression coefficients suggest that predominantly subcortical Ki values (dark blue) are low, while cortical Ki values (light blue) are less affected.

To further explore the relationship between the Ki indices and ADHD symptoms, a PLS regression model was created. This describes the dependence of the attention score (obtained using DSM-IV) on the Ki indices in the ADHD patients. The attention scores were taken as the Y variables, and Ki values as the X variables. One patient with a remarkably low attention (an outlier) who exhibited a good response to methylphenidate was excluded from this regression model (see patient 7; Table ​Table1).1). The model was statistically significant, indicating a strong relation between Ki and attention (R²X_cum = 0.46, R²Y_cum = 0.982, Q²_cum = 0.726 for the two first components). The R2Ycum statistic is analogous to the coefficient of determination (R2) used in correlation analysis. In this model, R2Y is large, 98%, reflecting a strong correlation between the Ki and the attention score data. The relations between Ki values and attention are further visualized in Figure ​Figure3,3, where the PLS regression coefficients are shown. Most cortical variables tend to be positively related to attention, whereas the subcortical Ki indices, with few exceptions, are negatively correlated to attention. Thus, low subcortical Ki values are associated both with the ADHD diagnosis per se, as demonstrated by the first PLS model (Figure ​(Figure2),2), and with more severe attention deficits within the ADHD group (Figure ​(Figure3).3). Cortical variables appear to be less important with respect to group discrimination, but have a positive correlation to attention within the ADHD group. When patient 7 with the deviating score was included, the model showed the similar trend, but did not reach significance due to the increased variation in the small sample (n = 8).

There are several limitations of this study. Most of our subjects with ADHD were not drug naïve, and although all were medication-free for at least 1 week, the washout period may not have been sufficient to reset the dopamine system to a basal level. In addition, psychostimulants are known to produce long term effects, e.g., dendrite growth in prefrontal cortex of rodents [41] and down-regulation of the dopamine D2 receptor in primates [42]. Therefore, at this point we cannot exclude any effects of the medication and therefore new studies need to be performed on drug naïve subjects. Moreover, two of the subjects with ADHD had additional neurological disorders. In particular, the subject with Tourette's syndrome could be a confounder since this condition is associated with abnormal dopamine metabolism [43]. However, it should be noted that the method used to establish statistical significance of the multivariate models, cross-validation, is based on the calculation of how stable the results are as different subjects are excluded. Models in which the relationships observed rely on "outliers" will therefore not be considered statistically significant by the cross-validation criterion. Exclusion of the two comorbid cases did not affect the results. Hence, although the sample size is small and includes individuals with comorbid conditions, the applications of multivariate statistical methods are able to demonstrate significant results.

The low L [¹¹C]DOPA Ki values suggest that children with ADHD differ from controls in their metabolism of DOPA over a range of cortical and subcortical regions. It is most likely that the low values reflect aberrations in the presynaptic synthesis and release of dopamine. These results support earlier theories that a hypofunction of the dopamine system underlies the behavioural symptoms in ADHD [44]. This theory was based on the positive effects of stimulant drugs in children with ADHD and that such drugs facilitate the endogenous monoamine transmission by blocking the dopamine transporter. Later, Volkow and co-workers have in a series of elegant PET studies shown that methylphenidate increases the endogenous level of dopamine and that it interacts with mental tasks triggering dopamine release [45,46]. However, there has been little support for lower dopamine synthesis in previous SPECT and PET studies in subjects with ADHD. PET studies on the uptake of [¹⁸F]fluorodopa have demonstrated that there is a decreased uptake in the frontal cortex and an increased uptake in the midbrain [19,20]. The latter result is opposite to ours, obtained by studying the uptake of L-[¹¹C]DOPA. The differences might be attributable to the different specificity of the radioligands. As a radiotracer, L-[¹¹C]DOPA seems to detect changes in the functional tone in the dopaminergic systems more readily than its fluorine labelled analogue [26].

Other SPECT and PET studies of dopamine transporter binding in ADHD have reported increased binding capacity in the striatum of adults and children with ADHD [21-23,47]. However, this result was not replicated by other groups, who instead reported either an unaltered or a reduced DAT- binding [24,48]. In a recent PET study on adolescent boys with ADHD, we used a newly developed cocaine analogue radioligand, [¹¹C]PE2I, that binds selectively to the dopamine transporter [25,49]. We could not confirm the first SPECT studies indicating an increased DAT binding in striatum, but rather obtained results more closely in agreement with the studies indicating an unaltered or decreased binding capacity. In addition, the high performance of 3D PET and the good characteristics of the new radioligand allowed us to detect a clear reduction of the DAT binding in the midbrain. These two recent studies from our group, indicating a lower dopamine synthesis and a reduced density of DAT in the midbrain, correspond well to the earlier theories of reduced dopamine signalling [44]. From our results it is not possible to determine the primary cause of the imbalance. A reduction of the presynaptic dopamine synthesis and release could be compensated for by decreased DAT activity. Genetic studies in subjects with ADHD have identified polymorphism in several specific genes that encode components of catecholamine signalling system, among them polymorphism in the DAT gene [13]. In some subjects, a lower uptake of dopamine from the synaptic cleft may thus be the primary cause. This could subsequently induce secondary adaptations of the dopamine release and synthesis via presynaptic autoreceptors [50]. Recently, in an animal model of ADHD (using the spontaneously hypertensive rat), the expression of specific genes involved in dopamine neuron differentiation and functioning during the postnatal development of the midbrain was studied [51]. The data showed transient reductions of the expression of tyrosine hydroxylase and dopamine transporter genes, suggesting a down-regulation of the dopamine transmission at this stage of development, which is in line with the present study and which supports the hypodopaminergic hypothesis of ADHD.

In the present study, ADHD subjects tended to display lower Ki values over most regions. However, with closer examination, a region-specific pattern emerges. The most profound decreases are seen in subcortical areas (Figure ​(Figure22 and Table ​Table1),1), such as the nucleus accumbens, putamen and the midbrain, while cortical areas show smaller decreases, and even a tendency towards increases in some prefrontal areas. This indicates that there are regional differences and that the ADHD specific alteration of the dopamine synthesis is region specific. The possibility of different dopamine abnormalities in different brain regions in ADHD has been discussed previously by, e.g., Castellanos [52], who suggested that there was a difference between areas innervated by the two dopamine pathways originating from the substantia nigra and the ventral tegmental area, respectively. The motor hyperactivity and impulsivity could be caused by overactivity in the basal ganglia, innervated by the substantia nigra, while the inattention could be due to an underactivity in the cortical areas innervated by the ventral tegmentum. Our data corresponds with this hypothesis to a certain extent, i.e., our data indicate that the DOPA metabolism in the basal ganglia seems to have a different pattern to that in the cortical areas. However, our data on the DOPA metabolism is hard to translate into terms of increased or reduced activity in the neural networks controlling the functional activity because the dopamine may have both facilitatory and inhibitory effects, depending on, e.g., the profile of the local dopamine receptor population on the postsynaptic neurons [53].

There was a strong relationship between attention and the regional Ki values within the ADHD Group. First, the results indicate that the Ki pattern not only discriminates between the ADHD group and the controls, but that it is also related to the severity of the attention difficulties of the individuals with ADHD. Most subcortical Ki values are inversely related to attention, i.e., the lower the Ki values the more problems the subject has with inattention. Conversely, in several cortical regions the Ki are positively related to the severity of symptoms. Again, it appears that the subcortical and cortical DA systems are differentially affected in patients with ADHD and that the resulting behavioural effects differ.