Over 20 years ago Kim and Choi  demonstrated that the breakdown of tryptophan and kynurenine can produce toxic metabolites (3HK and quinolinate) in neocortical tissue, but an alternative pathway produces kynurenate that can be protective. Kynurenine metabolism in the brain occurs largely in the glia that envelop glutamatergic synapses . Among the numerous responses of the glia to excitotoxicity are alterations in the proinflammatory cytokines, such as TNF-α and IFN-γ , where the balance with antiinflammatory cytokines influences tryptophan metabolism . Recently there have been numerous reports of associations of kynurenine metabolites and cytokines with brain damage and psychiatric disorders [8, 36–40]. While we only found modest changes of selected markers of these cytokines and metabolites in our pilot study of ADHD , we hypothesized here that key symptoms and poor attentional function would reflect alterations of the activity of these markers. Features of the kynurenine metabolic pathway and neuroinflammatory activity of cytokines have not been explored before in the context of ADHD.
We report a preliminary exploration of whether products of tryptophan metabolism and the cytokines that modulate this metabolism predict some of these symptoms, and their laboratory measures in children with a strict diagnosis of ADHD (combined type) contrasted with those developing normally. The study is predicated on increasing evidence that alterations in these two biochemical domains are reflected in both the features of ADHD [21, 41] and animal models of these functions [42, 43].
Those with the diagnosis of ADHD had clearly more symptoms of inattention, hyperactivity-impulsivity and opposition than those without the diagnosis. On a test of sustained attention the children with ADHD showed more motor activity, errors of omission, a slower RT and more RT variability. Most cytokine levels were modestly increased in children with ADHD (not S100B and IL-1β) and all tended to normalize on psychostimulant medication. While tryptophan availability was slightly increased in the ADHD group, 3HK levels tended to be lower than in the controls .
The cytokines and associations with symptoms
The DSM ratings of total symptom severity in the ADHD group were predicted by increasing levels of IL-16, and the anti-inflammatory IL-13, but by decreasing levels of S100B. The anti-inflammatory relationship was modestly evident for inattentive symptoms in the ADHD group, and arguably effective in countering dysfunction in that higher levels of IL-13 in the controls were related to fewer symptoms of inattention. The role of IL-16 can be pro-or anti-inflammatory in different situations [44, 45]. Here, in the ADHD group, IL-16 levels additionally predicted hyperactive-impulsive symptoms that were also modestly associated with decreased S100B. The IL-16 relationship with hyperactive symptoms might not be expected in so far as 2 genetic polymorphisms have been associated with the inattentive type of ADHD . However, the difference may reflect other features of the two populations influencing the neuroinflammatory role of IL-16, such as allergy (see below). The association of symptom severity with decreasing S100B levels is consistent with our prediction of inefficient glial function in ADHD  rather than an overproduction resulting from apoptosis and brain damage reported for other major psychiatric disorders [10, 47, 48].
Only proinflammatory cytokine activity related to oppositional symptom ratings. Decreasing levels of IL-2, and to a minor extent TNF-a, were associated with increasing signs of opposition. We suggest the negative direction of association mirrors consistently the frequent reports of increased IL-2 activity in disorders such as schizophrenia and Parkinsonism with a very different clinical profile [40, 49, 50]. On the other hand we note a relationship between increasing IL-6 and oppositional behaviour in the control group. We tentatively suggest that, as IL-6 levels increased with BMI and the controls were marginally older than the patients, such an increase reflects the normal parallel development of boisterous behaviour often associated with adolescence onset. This interpretation is also consistent with the well-studied associations of IL-6 levels with physical and psychological stress .
Sensitivity to allergy reflected opposing trends for pro-and antiinflammatory cytokines. Children sensitive to allergy showed 15% more IL-16, 32% more IL-13 and similar levels of IL-10 . But for the ADHD children who experienced allergic responses, these responses tended to be predicted by decreasing levels of the proinflammatory TNF-α and increases of the antiinflammatory IL-13. For controls the trend went in the opposite direction for IL-13, but this was balanced by modest increases of another antiinflammatory cytokine, IL-10. The difference between groups in the proinflammatory response is quantitative, but that for the antiinflammatory response is qualitative. This physiological response may underlie a subgroup of ADHD children with high allergic responsivity and for whom ADHD symptoms are alleviated by removal of the allergens from the diet [52–54]. Further, IL-10 secretion is facilitated by physical or psychological stress  and could be responsible for the decreases of tryptophan metabolism  discussed below.
The cytokines and associations with CPT performance
During the CPT both the duration of major movements and the quantity of small fidgety activity (microevents) were directly measured. Increasing IL-16 levels not only predicted hyperactive symptoms (above) but clearly predicted quantitative measures of motor activity, especially (but not only) in ADHD children. The associations extended to error omission and commission in the medicated children that may reflect motor as well as cognitive control. The associations of IL-16 with motor activity and allergy in this study clearly add to the concept of this being a pleiotropic cytokine, for which not all the roles have yet been elaborated.
Accuracy in terms of the errors made is more usually thought of as a cognitive than a motoric feature. Our data suggest that decreasing levels of the anti-inflammatory IL-13 may not only relate to inattentive symptoms (above), but specifically to increases of errors of omission. But, both this association and the tendency for errors of commission in medicated children to be reduced as pro-inflammatory TNF-α and IL-6 levels increased were modest trends. This should be examined closely in future studies. Nonetheless the increase of IL-2 with decreased RT in the ADHDmed group supports the proinflammatory effect on attention-related function.
Associations of the coefficient of variance of the RT is of especial interest, first as variability is the core feature of the hypothesis of glial impairment in ADHD , and secondly as it has been proposed as an endophenotype for ADHD (introduction). Here, across all subjects, increasing RT variability was related to decreasing levels of TNF-α and with increasing levels of IFN-γ.
Numerous studies confirm the involvement of proinflammatory cytokines in cognitive performance both in animal models and in the treatment of cancer patients. For example IL-2 treatment can interfere with long term potentiation in the hippocampal slice  and the formation of working memory in patients . Treating rodents with IL-6 can impair measures of memory [59, 60], yet in knockout mice recall can be facilitated [61, 62]: the relationship of IL-6 to cognitive function may be linear or U-shaped . However, IL-6 expression depends on the nature of the allele carried (CC/GC genotype) and this variable has too rarely been considered [64, 65]: the C allele may dominate in ADHD while the protective G allele may predominate in typically developing children . The characteristics of TNF-α and IFN-γ have much in common with other proinflammatory interleukins. Too much  or too little TNF-α  can impair cognitive function, but moderate levels are essential . Variation of these levels can influence the turnover of each of the monoamines . Thus it is no surprise that TNF-α levels relate to accuracy and the variability of response in ADHD. However, the relationship of the effects with those of IFN-γ require closer study, as the physiological response to each varies with the measure and over time .
Tryptophan metabolism and associations with symptoms and CPT performance
Even after controlling for age, BMI and SES there were tendencies across subjects for a build-up of tryptophan (and its increased availability) to be associated with symptoms of inattention, RT and its variability as well as increases of errors of omission. In contrast decreased tryptophan levels (more breakdown) related to errors of commission. However, tryptophan metabolism was not related to motor activity. This was unexpected as the motor regions of the brain receive a dense 5-HT innervation .
For controls, slower RTs but for the ADHDmed group faster RTs were clearly associated with lower kynurenine levels. There was an indication that low levels of kynurenine for controls was reflected by the trend for reduced variability and an associated increase of metabolism in the 5-HT pathway. For medicated children the implication is of further glial metabolism in the kynurenine pathway that is facilitated by psychostimulant treatment . Other trend relationships (e.g. 3HK with inattention, 5-HIAA with anxiety) were equivocal. The lack of anticipated changes of 5-HT metabolism related to antisocial/oppositional behaviour [73, 74] may simply reflect the extent of 5-HT metabolism from peripheral sources.
The absence of clear associations for tryptophan metabolism with symptoms or attention related performance in ADHD is itself of interest. An overexpression of indoleamine 2, 3 dioxygenase facilitates degradation to kynurenine (and its potentially toxic metabolites) and reduced 5-HT metabolism . Such activity is promoted by proinflammatory cytokines . This scenario fits evidence for the development of psychotic symptoms  and major depression . But, our results tend to support the opposite effect in ADHD with increases of tryptophan availability relating to inattention and variable responses. Indeed, here there was even modest support for previous evidence for an involvement of 5-HT metabolism .
We had anticipated that the cytokine balance would favour kynurenine breakdown to either kynurenate or 3HK. On the basis of their glutamatergic antagonist and agonist activity these metabolites (and the 3HK metabolite quinolinate) have been ascribed a neuroprotective and toxic role, respectively . But in fact our typically developing children rather than the ADHD group showed descriptively higher levels of kynurenate and 3HK. Yet, these measures did not predict symptoms or CPT performance. We speculate that the ADHD children in fact exhibit a relative lack of some of the other characteristics of these metabolites. For example, kynurenate also has α7 cholinergic antagonist properties  and higher levels antagonize rather than facilitate AMPA receptor transmission . 3HK has antioxidant capability () as well as the toxic potential to assist in developmental pruning . Indeed, it is unclear as yet whether energy metabolism in developing children would be impaired by the NMDA agonism of quinolinate  or facilitated through further metabolism to NAD(+) .
Our conclusions are severely curtailed by not having measured other metabolites (picolinic and, quinolinic acid, NAD(+)), enzyme activity or the expression of genetic variants for these. Also in a pilot study the robustness of the data is naturally compromised by the small number of children that could be recruited and persuaded to offer a blood sample. However, this exploration with 56 children indicates a number of lines of inquiry that are not worth pursuing and others that might merit further study. Indeed, although the number of variables examined raises questions about the prevalence of false positive results, the statistical procedure included confirmatory tests and false discovery assessment to limit this. Cytokine measures, renowned for their variability, were relatively consistent with remarkably few outliers. Future work should seek replication with a longitudinal design to demonstrate consistency of the data in a larger sample. Such a design should account for developmental age, match for gender, and control for medication yet recruit cases with a range of comorbid attributes to clarify their contribution to the diagnostic condition.