Introduction
Attention deficit hyperactivity disorder (ADHD) is defined as age-inappropriate inattention and/or hyperactivity/impulsiveness (APA, 2013). It is one of the most common neurodevelopmental disorders with around 5% prevalence worldwide (Polanczyk et al. Reference Polanczyk, Willcutt, Salum, Kieling and Rohde2014). ADHD patients have deficits in executive functions (EF) such as inhibition, attention and working memory (Willcutt et al. Reference Willcutt, Sonuga-Barke, Nigg and Sergeant2008), underpinned by abnormalities in fronto-striatal, fronto-temporo-parietal and fronto-cerebellar networks (Hart et al. Reference Hart, Radua, Mataix-Cols and Rubia2012, Reference Hart, Radua, Nakao, Mataix-Cols and Rubia2013; Rubia et al. Reference Rubia, Alegria and Brinson2014a ). Moreover, they have deficits in timing functions (Noreika et al. Reference Noreika, Falter and Rubia2013) and in ‘hot’ EF, referring to EF involving motivation and affect such as reward-related decision-making (Kerr & Zelazo, Reference Kerr and Zelazo2004), as measured by temporal discounting (TD) and gambling tasks (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Noreika et al. Reference Noreika, Falter and Rubia2013). Nonetheless, there is heterogeneity in cognitive impairments, with some patients not showing impairments or only in some cognitive domains which are likely underpinned by different pathophysiological pathways (Sonuga-Barke, Reference Sonuga-Barke2003; Nigg et al. Reference Nigg, Willcutt, Doyle and Sonuga-Barke2005; Sonuga-Barke et al. Reference Sonuga-Barke, Bitsakou and Thompson2010).
TD tasks require choices between small immediate and larger delayed rewards and measure the extent to which a reward is subjectively discounted when delayed in time, i.e. the sensitivity to temporal delays measured in units of reward (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009). The ability to inhibit immediate rewards and wait for larger future rewards depends on well-developed frontal lobe-mediated motivation control and temporal foresight and is key for mature decision-making. TD matures with age (Christakou et al. Reference Christakou, Brammer and Rubia2011) and varies between individuals, with steeper TD, i.e. more rapidly decaying rates of reward discounting with increasing time (Richards et al. Reference Richards, Zhang, Mitchell and Wit1999), in more impulsive subjects (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Noreika et al. Reference Noreika, Falter and Rubia2013). In individually adjusted TD paradigms (Richards et al. Reference Richards, Mitchell, de Wit and Seiden1997; Christakou et al. Reference Christakou, Brammer and Rubia2011), the immediate reward is adjusted using an algorithm based on previous choices for different delays, converging towards the value of the participant's subjective equivalent of the fixed delayed reward (Richards et al. Reference Richards, Zhang, Mitchell and Wit1999). From this, a typically hyperbolic delay discounting function is calculated, the steepness of which indicates the individual TD rate, which is associated with impulsivity (Richards et al. Reference Richards, Zhang, Mitchell and Wit1999; Critchfield & Kollins, Reference Critchfield and Kollins2001).
ADHD patients are impaired in TD tasks (Noreika et al. Reference Noreika, Falter and Rubia2013), with, however, some negative findings, mostly in non-individually adjusted task versions (Sonuga-Barke et al. Reference Sonuga-Barke, Taylor, Sembi and Smith1992; Scheres et al. Reference Scheres, Dijkstra, Ainslie, Balkan, Reynolds, Sonuga-Barke and Castellanos2006, Reference Scheres, Tontsch, Thoeny and Kaczkurkin2010). Functional magnetic resonance imaging (fMRI) studies of TD in healthy adults implicate ventromedial-fronto-limbic networks of reward-based decision-making and dorsolateral and inferior-fronto-insula-striato-parietal networks of temporal foresight (Christakou et al. Reference Christakou, Brammer and Rubia2011; Wesley & Bickel, Reference Wesley and Bickel2014). Despite documented TD deficits in ADHD, few fMRI studies have investigated its neurofunctional correlates. ADHD adolescents showed underactivation relative to controls during delayed choices in an adjusted fMRI TD task in inferior frontal cortex (IFC), insula, striatal and cerebellar regions (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009) and significantly weaker correlations between better TD and activation during delayed choices in IFC, superior temporal lobes, insula, supplementary motor area and cerebellum (Chantiluke et al. Reference Chantiluke, Christakou, Murphy, Giampietro, Daly, Brammer, Murphy and Rubia2014d ). In adult ADHD, abnormal striato-limbic activation has been observed (Plichta et al. Reference Plichta, Vasic, Wolf, Lesch, Brummer, Jacob, Fallgatter and Grön2009).
Neurotransmitters such as serotonin (5-HT) are implicated in ADHD (Oades, Reference Oades2007, Reference Oades2008), potentially via modulation of these neural circuits. Converging evidence across methodologies shows that serotonergic systems may be dysfunctional in ADHD (Oades, Reference Oades2007), with evidence for chemical alterations of 5-HT systems, decreased 5-HT platelet levels (Spivak et al. Reference Spivak, Vered, Yoran-Hegesh, Averbuch, Mester, Graf and Weizman1999), and increased ADHD-related behaviour after 5-HT depletion in ADHD patients (Zepf et al. Reference Zepf, Gaber, Baurmann, Bubenzer, Konrad, Herpertz-Dahlmann, Stadler, Poustka and Wöckel2010). Second, there is evidence for an association between 5-HT-related polymorphisms and ADHD (Gizer et al. Reference Gizer, Ficks and Waldman2009; Rommelse et al. Reference Rommelse, Franke, Geurts, Hartman and Buitelaar2010) and methylphenidate treatment response (McGough et al. Reference McGough, McCracken, Loo, Manganiello, Leung, Tietjens, Trinh, Baweja, Suddath, Smalley, Hellemann and Sugar2009). Moreover, the selective serotonin reuptake inhibitor (SSRI) fluoxetine has been shown to be effective in reducing ADHD-related symptoms in children (Barrickman et al. Reference Barrickman, Noyes, Kuperman, Schumacher and Verda1991; Quintana et al. Reference Quintana, Butterbaugh, Purnell and Layman2007) and to improve the efficacy of stimulants in human and animal studies (Gammon & Brown, Reference Gammon and Brown1993; Findling, Reference Findling1996; Gainetdinov et al. Reference Gainetdinov, Wetsel, Jones, Levin, Jaber and Caron1999). Further, the concurrent administration of 5-HT and dopamine amino-acid precursors can improve ADHD symptoms (Hinz et al. Reference Hinz, Stein, Neff, Weinberg and Uncini2011). However, replication is needed as these studies are limited by co-morbid samples (Quintana et al. Reference Quintana, Butterbaugh, Purnell and Layman2007) and non-randomized trials in small samples (Barrickman et al. Reference Barrickman, Noyes, Kuperman, Schumacher and Verda1991). Last, in healthy adults, tryptophan depletion, which reduces brain 5-HT by up to 60%, down-regulates activation in key ADHD deficit areas of IFC and basal ganglia (Rubia et al. Reference Rubia, Lee, Cleare, Tunstall, Fu, Brammer and McGuire2005; Lamar et al. Reference Lamar, Cutter, Rubia, Brammer, Daly, Craig, Cleare and Murphy2009, Reference Lamar, Craig, Daly, Cutter, Tang, Brammer, Rubia and Murphy2014), which are up-regulated with 5-HT agonists (Del-Ben et al. Reference Del-Ben, Deakin, McKie, Delvai, Williams, Elliott, Dolan and Anderson2005). 5-HT has also been implicated in reward-based decision-making in healthy adults (Cools et al. Reference Cools, Nakamura and Daw2011; Rogers, Reference Rogers2011; Robinson et al. Reference Robinson, Cools and Sahakian2012), where striatal 5-HT levels have been shown to modulate choices of longer, delayed rewards (Schweighofer et al. Reference Schweighofer, Tanaka and Doya2007; Tanaka et al. Reference Tanaka, Schweighofer, Asahi, Shishida, Okamoto, Yamawaki and Doya2007; Doya, Reference Doya2008).
In conclusion, there is evidence that 5-HT is associated with ADHD, with impulsivity, in particular TD performance, and that it modulates IFC-striatal activation, a key ADHD deficit. Despite this, hardly any fMRI studies have tested the effects of 5-HT agonists on brain function in ADHD. We have previously shown that in ADHD children, the SSRI fluoxetine v. placebo up-regulated and normalized IFC-striatal and parietal underactivation during inhibition (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons, Murphy and Rubia2014a , Reference Chantiluke, Barrett, Giampietro, Santosh, Brammer, Simmons, Murphy and Rubia c ), and enhanced the deactivation in default mode network (DMN) regions during working memory (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons and Rubia2014b ).
This study therefore aimed to investigate the effect of a single dose of fluoxetine relative to placebo on brain activation in ADHD adolescents during a TD task. Further, to test for potential normalization effects of fluoxetine on abnormal brain activation in ADHD patients under placebo, we also compared brain activation during both drug conditions to that of healthy adolescents.
Based on previous findings, we hypothesized that ADHD adolescents under placebo would show steeper TD rates (Noreika et al. Reference Noreika, Falter and Rubia2013) and reduced IFC-insular-striatal activation during TD (Plichta et al. Reference Plichta, Vasic, Wolf, Lesch, Brummer, Jacob, Fallgatter and Grön2009; Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Chantiluke et al. Reference Chantiluke, Christakou, Murphy, Giampietro, Daly, Brammer, Murphy and Rubia2014d ). Further, based on our fMRI studies showing up-regulation and normalization with fluoxetine in task-relevant regions during related tasks of cognitive control in ADHD (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons, Murphy and Rubia2014a –Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons and Rubia c ), and evidence for 5-HT modulation of IFC-striatal regions in healthy adults (Rubia et al. Reference Rubia, Lee, Cleare, Tunstall, Fu, Brammer and McGuire2005; Lamar et al. Reference Lamar, Cutter, Rubia, Brammer, Daly, Craig, Cleare and Murphy2009), we hypothesized that fluoxetine would up-regulate IFC-insula-striato-parietal activation in patients and normalize regional underactivation relative to controls.
Method
Participants
Thirty-two right-handed (Oldfield, Reference Oldfield1971) boys with (N = 12) and without ADHD (N = 20) were recruited from local clinics and support groups, aged 11–17 years, with IQ⩾70 measured by the Wechsler Abbreviated Scale of Intelligence – Revised (WASI-R) short form (Wechsler, Reference Wechsler1999). ADHD boys had a clinical DSM-IV diagnosis of non-comorbid ADHD, inattentive/hyperactive-impulsive combined subtype, assessed using the standardized Maudsley diagnostic interview (Goldberg & Murray, Reference Goldberg and Murray2006). Patients scored above clinical threshold for ADHD symptoms on the Strengths and Difficulties Questionnaire (SDQ; Goodman & Scott, Reference Goodman and Scott1999) and Conner's Parent Rating Scale – Revised (CPRS-R; Conners et al. Reference Conners, Sitarenios, Parker and Epstein1998). They also scored below clinical threshold for autistic spectrum disorders on the Social Communication Questionnaire (SCQ; Rutter et al. Reference Rutter, Bailey and Lord2003). Nine ADHD boys were on psychostimulants but withheld medication for 48 h prior to scanning.
Patients were scanned twice in a randomized, double-blind, placebo-controlled design. Due to the half-life of fluoxetine (1–3 days) and its metabolite norfluoxetine (5–16 days) (Wong et al. Reference Wong, Bymaster and Engleman1995), scans were conducted 3–4 weeks apart. To ensure the fluoxetine dose had reached peak plasma levels (after 5–8 h; Wong et al. Reference Wong, Bymaster and Engleman1995), patients were scanned 5 h post-administration. Liquid fluoxetine was titrated to age and weight: boys 10–13 years and <30 kg received 8 mg, those >30 kg received 10 mg. Boys 14–17 years and <30 kg received 10 mg, and those >30 kg received 15 mg. Placebo was equivalent volumes of peppermint water, similar in taste and appearance to fluoxetine.
Twenty healthy age and handedness-matched control boys were recruited locally by advertisement and scanned once. Controls scored below clinical threshold on the SDQ, SCQ and CPRS-R and did not have any psychiatric condition.
Exclusion criteria for all participants were neurological disorders, drug/alcohol dependency and MRI contraindications.
Ethical statement
The study was conducted according to the latest version of the Declaration of Helsinki. Ethical approval was obtained from the local Research Ethics Committee. Study details were explained to both child and guardian, and written informed consent was obtained for all participants.
Temporal discounting fMRI paradigm
Prior to the first scan, subjects practised the 12-min task once in a ‘mock’ scanner. Subjects made a choice, by pressing a left or right button, between receiving a small amount of money immediately (£0–£100) or £100 in 1 week, month or year. Delay choices (20 trials of each delay length) were randomized, but the delayed option was consistently displayed on the right side, and the variable immediate choices on the left side of the screen to minimize potential sensorimotor mapping effects. Choices were displayed for 4 s, followed by a blank screen of at least 8 s (inter-trial interval: 12 s). The amount of immediate reward was adjusted through an algorithm based on previous choices which was calculated separately for each of the three delays. This narrows the range of values, converging into an indifference point where the immediate reward is considered by the subject to be equivalent to the delayed amount for the given delay (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Christakou et al. Reference Christakou, Brammer and Rubia2011). This algorithm ensures equal numbers of immediate and delayed choices to be contrasted in the fMRI analysis.
Analysis of performance data
To estimate the steepness of TD for each subject, the indifference value between the immediate amount and the delayed £100 for each delay was calculated, equivalent to the subject's subjective value of £100 after each delay, and defined as the midpoint between the lowest immediate reward chosen by the subject and the next lowest immediate reward available (i.e. the value of the immediate reward offered at which point the subject began to instead consistently choose the delayed reward) (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Christakou et al. Reference Christakou, Brammer and Rubia2011).
Reward is typically discounted as a hyperbolic decay function depending on amount, delay and a free impulsiveness indicator k, calculated by fitting a hyperbolic function to the indifference values for each delay (see Supplementary material).
However, the limitations of the fMRI task adaption, i.e. relatively few trials and only three delay points, limit the goodness-of-fit of the data to a nonlinear curve function. In addition, the distribution of k values was not normal, skewed by low-frequency and high-value outliers. Thus, TD was measured using the area under the curve (AUC) which is more appropriate for investigations with quantitative, inferential statistics (Myerson et al. Reference Myerson, Green and Warusawitharana2001). The normalized subjective values of the delayed £100 for each delay were plotted against the normalized delays and AUC of these plots were calculated for each participant, using this obtained value as the main dependent variable. AUC correlated inversely with k values (r = −0.898, p < 0.001), where smaller AUC values denote steeper discounting rates, indicating increased choice impulsivity.
A repeated-measures within-group analysis of variance (ANOVA) was conducted with patients with medication condition (placebo, fluoxetine) as a within-subjects variable to test for medication effects on TD. Two ANOVAs were conducted with group as independent variable and AUC as dependent measure to test for differences in TD performance between controls and ADHD patients on either placebo or fluoxetine. To test for potential main effects of drug administration order and of an interaction between order and drug condition, order was included as a between-subjects factor in the repeated-measures ANOVA.
fMRI image acquisition
Gradient-echo echo-planar MR imaging (EPI) data were acquired at King's College London, Institute of Psychiatry's Centre for Neuroimaging Sciences on a 3-T General Electric Signa HDx MRI scanner (GE Healthcare, UK). For details of scan acquisition, see Supplementary material.
fMRI image analysis
Event-related activation data were acquired in randomized trial presentation and analysed using the non-parametric XBAM software package developed at the Institute of Psychiatry, King's College London (www.brainmap.co.uk; Brammer et al. Reference Brammer, Bullmore, Simmons, Williams, Grasby, Howard, Woodruff and Rabe-Hesketh1997). The individual- and group-level analyses methods are described in detail elsewhere (Brammer et al. Reference Brammer, Bullmore, Simmons, Williams, Grasby, Howard, Woodruff and Rabe-Hesketh1997; Bullmore et al. Reference Bullmore, Brammer, Rabe-Hesketh, Curtis, Morris, Williams, Sharma and McGuire1999a , Reference Bullmore, Suckling, Overmeyer, Rabe-Hesketh, Taylor and Brammer b , Reference Bullmore, Long, Suckling, Fadili, Calvert, Zelaya, Carpenter and Brammer2001; Cubillo et al. Reference Cubillo, Smith, Barrett, Giampietro, Brammer, Simmons and Rubia2014b ) and in the Supplementary material.
ANCOVA of within-patient medication effects
To investigate medication effects on brain activation in the ADHD group, a within-group repeated-measures analysis of covariance (ANCOVA) with motion as covariate and medication condition as within-subjects factor was conducted using randomization-based testing for voxel- or cluster-wise differences, as described previously (Bullmore et al. Reference Bullmore, Suckling, Overmeyer, Rabe-Hesketh, Taylor and Brammer1999b , Reference Bullmore, Long, Suckling, Fadili, Calvert, Zelaya, Carpenter and Brammer2001) and in the Supplementary material. Voxel- and cluster-level statistical thresholds were set so as to obtain less than one false positive 3D cluster per map (p < 0.05 was used for voxel and p < 0.005 for cluster comparisons). The standardized blood oxygen level dependent (BOLD) response values for each participant were extracted for each of the significant clusters of the ANCOVA analyses and plotted to determine the direction of effects. Repeated-measures ANOVAs on the extracted BOLD response measures were conducted in patients to test for potential effects of scan-order and interactions between scan order and drug condition.
ANCOVA of between-group effects
One-way ANCOVAs with group as main factor and motion as covariate were conducted using randomization-based testing to test for case-control differences under placebo or fluoxetine (Bullmore et al. Reference Bullmore, Suckling, Overmeyer, Rabe-Hesketh, Taylor and Brammer1999b , Reference Bullmore, Long, Suckling, Fadili, Calvert, Zelaya, Carpenter and Brammer2001). For these comparisons, p < 0.05 (voxel-level) and p < 0.05 (cluster-level) were used. Standardized BOLD responses were then extracted from significant clusters for each participant and plotted to determine direction of effects.
Correlations with behaviour and IQ
To examine whether clusters which showed group effects in case-control comparisons were related to IQ or TD, the BOLD response in these clusters was extracted for each participant and Pearson correlations were performed with IQ and AUC in each group (ADHD placebo, ADHD fluoxetine, controls).
Results
Participant characteristics
Univariate ANOVA revealed no significant group differences in age, but IQ, which was lower in ADHD (Table 1). However, since low IQ is associated with ADHD (Bridgett & Walker, Reference Bridgett and Walker2006), IQ was not covaried as covarying for differences between groups that were not randomly selected violates ANCOVA assumptions (Miller & Chapman, Reference Miller and Chapman2001). Nonetheless, to assess potential effects of IQ on case-control comparisons, BOLD responses were correlated with IQ and analyses were repeated covarying for IQ. As expected, patients had significantly lower CPRS-R t scores, SDQ and SCQ scores than controls (Table 1).
Participants’ characteristics
s.d., Standard deviations; df, degrees of freedom; CPRS-R, Conner's Parent Rating Scale – Revised; SCQ, Social Communication Questionnaire; SDQ, Strengths and Difficulties Questionnaire; AUC, area under the curve.
a CPRS-R total T score could not be obtained for four control participants.
b SCQ scores could not be obtained from three control participants.
c SDQ scores could not be obtained from two control participants.
Performance data
For repeated-measures ANOVA in patients, no significant drug effects were found on AUC (F 1,11 = 0.08, p = n.s.), reaction time (RT) (F 1,11 = 0.01, p = n.s.) or omission errors (OM) (F 1,11 = 0.16, p = n.s.) (see Table 1). Case-control ANOVAs showed no differences in RT or OM, but controls had larger AUC than patients under placebo (F 1,30 = 4, p < 0.05) but no longer differed from patients under fluoxetine (F 1,30 = 2, p = n.s.), suggesting that fluoxetine normalized case-control performance differences (see Table 1).
Drug administration order had no main effect on the primary behavioural outcome of AUC in the ADHD group (F 1,10 = .07, p = n.s.) and there was no interaction between scan order and drug condition (F 1,10 = 1.31, p = n.s.).
fMRI data
Motion
No differences were found for largest head displacement in 3-dimensional space in the ADHD group under each drug condition (F 2,10 = 0.51, p = n.s.). Moreover, no group×displacement interaction was found between controls and ADHD under placebo (F 2,29 = 2.63, p = n.s.) or fluoxetine (F 2,29 = 2.54, p = n.s.). Nevertheless, to exclude potential effects of non-significant motion, motion parameters in 3D Euclidian space were included as covariates in the fMRI analyses.
Group brain activation maps for delayed minus immediate choices
For the contrast of delayed minus immediate choices, controls showed activation in dorsomedial PFC (dmPFC) and anterior cingulate cortex (ACC), insula, pre- and post-central gyri and parieto-occipital and cerebellar regions. ADHD patients on placebo showed activation in ACC, pre- and post-central gyrus, posterior cingulate (PCC), and occipito-cerebellar regions, while under fluoxetine they showed activation in right dorsolateral and inferior PFC (dlPFC/IFC)/insula extending into basal ganglia (BG), ACC, temporo-parietal and occipital cortices and cerebellum (see Supplementary material and Fig. S1).
Within-group differences between ADHD patients on placebo v. fluoxetine
Repeated-measures ANCOVA revealed a significant drug effect in a large cluster comprising right IFC, insula, precentral gyrus and superior temporal cortices extending into BG, which was enhanced under fluoxetine relative to placebo (Fig. 1, Table 2). Post-hoc calculations in SPSS version 21 (IBM Corp., USA) indicated an observed power of 89% (partial η 2 = 0.53). Activation in IFC was significantly negatively correlated in the placebo group with AUC (r = −0.676, p < 0.016). Under fluoxetine, however, the correlation was at a trend-level positive (r = 0.563, p = 0.057).
Within-patient comparisons. Axial sections show medication effects in the ADHD group. Red = fluoxetine > placebo; blue = placebo > fluoxetine. Also shown are the statistical measures of the blood oxygen level-dependent (BOLD) response for each of the brain regions that showed a significant effect of medication in patients. R, Right; L, left; IFC, inferior frontal cortex; STL, superior temporal lobe; PCC, posterior cingulate cortex; IPL, inferior parietal lobe; MFG, middle frontal gyrus. Talairach z coordinates are indicated for slice distance (in mm) from the intercommissural line. The right side of the image corresponds to the right side of the brain.
Within-patient comparisons of activation differences for delayed minus immediate choices
MNI, Montreal Neurological Institute; R, right; L, left; IFC, inferior frontal cortex; dlPFC, dorsolateral prefrontal cortex; STL, superior temporal lobe; PCC, posterior cingulate cortex; IPL, inferior parietal lobe. η 2 p refers to the effect size of the differences in activation between groups.
Under placebo relative to fluoxetine, patients had enhanced activation during delayed minus immediate choices in two clusters, one comprising bilateral cerebellar hemispheres and vermis, PCC, precuneus and occipital lobe, and the other in left pre- and post-central gyrus, extending into middle frontal gyrus and inferior parietal lobe (IPL) (Fig. 1, Table 2). Further, activation in the cerebellum/PCC/precuneus cluster was negatively correlated with the IFC cluster that was up-regulated under fluoxetine (r = −0.859, p < 0.001).
Drug administration order had no effect on within-group differences in BOLD response (right IFC: F 1,10 = 2.8, p = n.s.; cerebellum/occipital: F 1,10 = 0.40, p = n.s.; left pre-/post-central gyrus: F 1,10 = 0.88, p = n.s.), and there was no interaction between drug administration order and condition (IFC: F 1,10 = 0.07, p = n.s.; cerebellum/occipital: F 1,10 = 0.15, p = n.s.; left pre-/post-central gyrus: F 1,10 = 0.75, p = n.s.).
Between-group differences
Controls v. ADHD patients on placebo
Between-group ANCOVA showed significantly increased activation in controls relative to ADHD under placebo for delayed minus immediate choices in three right-hemispheric clusters comprising right pre- and post-central gyri, extending into IPL and insula. Patients on placebo showed increased activation relative to controls in left anterior cerebellum/occipital lobe (Table 3, Fig. 2 a). No significant correlations were observed between extracted BOLD response from these clusters and IQ. Further, ANCOVA with IQ as covariate showed that all significant clusters remained with the exception of right insula. No correlations were observed between AUC and extracted BOLD response in any clusters.
Case-control comparisons. Axial sections show the between-group ANCOVA findings between controls and patients under (a) placebo and (b) fluoxetine. Red = controls > ADHD; blue =ADHD > controls. Also shown are the statistical measures of the blood oxygen level-dependent (BOLD) response for each of the brain regions that showed a significant group effect. R, Right; L, left; IPL, inferior parietal lobe; CB, cerebellum. Talairach z-coordinates are indicated for slice distance (in mm) from the intercommissural line. The right side of the image corresponds to the right side of the brain.
Case-control comparisons of activation differences for delayed minus immediate choices
MNI, Montreal Neurological Institute; R, right; L, left; IPL, inferior parietal lobe. η 2 p refers to the effect size of differences in activation between groups.
Controls vs. ADHD patients on fluoxetine
Controls relative to ADHD patients on fluoxetine showed enhanced activation in left pre- and post-central gyri reaching into IPL. No clusters were increased in ADHD relative to controls (Table 3, Fig. 2 b). Thus, the right hemispheric clusters which were enhanced in activation in controls relative to ADHD under placebo were no longer observed, suggesting that fluoxetine normalized these activation differences. No significant correlations were observed between extracted BOLD response and IQ. ANCOVA with IQ as covariate showed that all significant clusters remained. No correlations were found between AUC and extracted BOLD response in significant clusters.
Discussion
Behaviourally, an acute dose of fluoxetine normalized TD abnormalities in ADHD relative to controls. At the brain level, in patients, fluoxetine relative to placebo significantly up-regulated activation in a large right-hemispheric IFC-premotor-insular-striatal cluster, which correlated trend-wise with better TD. Fluoxetine additionally down-regulated activation in presumably default mode network activations in PCC/precuneus and in pre- and post-central gyrus/IPL as well as cerebellum. Relative to controls, patients on placebo showed underactivation in right insula, pre-/post-central gyrus and IPL but enhanced activation in left anterior cerebellum/PCC. Fluoxetine normalized all case-control differences, due to up-regulation/down-regulation of these regions in patients, but lead to underactivation in left-hemispheric pre-/post-central gyrus/IPL in ADHD relative to controls, due to down-regulation of this activation in patients.
Fluoxetine relative to placebo up-regulated a large right-lateralized cluster in ADHD comprising IFC, premotor cortex, insula and BG, which was associated trend-wise with better TD. This right IFC-insular-striatal network comprises key regions for TD important for integrating external information with internal value representations to support goal-directed EF (Wittmann et al. Reference Wittmann, Leland and Paulus2007; Christakou et al. Reference Christakou, Brammer, Giampietro and Rubia2009, Reference Christakou, Gershman, Niv, Simmons, Brammer and Rubia2013a ; Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Wesley & Bickel, Reference Wesley and Bickel2014). Right IFC is a key cognitive control hub region, crucial for inhibiting immediate reward choices as well as for inter-temporal bridging and future reward representation (Wiener et al. Reference Wiener, Turkeltaub and Coslett2010; Radua et al. Reference Radua, Pozo, Gómez, Guillen-Grima and Ortuño2014). The BG are linked to a reward-valuation network that mediates reinforcement learning, reward-processing and inter-temporal bridging (Wittmann et al. Reference Wittmann, Leland and Paulus2007; Koch et al. Reference Koch, Oliveri and Caltagirone2009; Peters & Büchel, Reference Peters and Büchel2011) while the insula plays a role in future reward-value representation as well as timing functions including temporal foresight (Carter et al. Reference Carter, Meyer and Huettel2010; Wiener et al. Reference Wiener, Turkeltaub and Coslett2010; Radua et al. Reference Radua, Pozo, Gómez, Guillen-Grima and Ortuño2014; Wesley & Bickel, Reference Wesley and Bickel2014). In particular right hemispheric IFC, insula and BG have been shown to be consistently hypoactivated in ADHD in meta-analyses of EF tasks (Cortese et al. Reference Cortese, Kelly, Chabernaud, Proal, Di Martino, Milham and Castellanos2012; Hart et al. Reference Hart, Radua, Nakao, Mataix-Cols and Rubia2013), including TD (Rubia et al. Reference Rubia, Halari, Christakou and Taylor2009; Chantiluke et al. Reference Chantiluke, Christakou, Murphy, Giampietro, Daly, Brammer, Murphy and Rubia2014d ). Right IFC underactivation has further been shown to be disorder-specific relative to other childhood disorders such as conduct and obsessive-compulsive disorder (Rubia, Reference Rubia2011; Rubia et al. Reference Rubia, Alegria and Brinson2014a ; Norman et al. Reference Norman, Carlisi, Lukito, Hart, Mataix-Cols, Radua and Rubia2015). In this study, we only observed underactivation in ADHD relative to healthy adolescents in insula and premotor regions, rather than in IFC and basal ganglia, presumably due to low power.
Given that our recent meta-analysis findings of a consistent up-regulation with stimulants of right IFC, insula and BG activation in ADHD (Rubia et al. Reference Rubia, Alegria, Cubillo, Smith, Brammer and Radua2014b ), the findings suggest that 5-HT agonists may have comparable up-regulatory effects to stimulants. In fact, the up-regulated region in ventrolateral PFC reaching into anterior insula, putamen and superior temporal lobe is in a very similar location to the cluster observed in our meta-analysis of methylphenidate effects on ADHD brain function (Talairach coordinates: 42, 20, −12), with a sizable effect size of 1 relative to the meta-analytical effect size of 1.5 (Rubia et al. Reference Rubia, Alegria, Cubillo, Smith, Brammer and Radua2014b ). Further, it is markedly similar to the up-regulated IFC location in our fMRI studies of methylphenidate effects on inhibition and timing, with effect sizes of 0.7 and 0.2, respectively (Smith et al. Reference Smith, Cubillo, Barrett, Giampietro, Simmons, Brammer and Rubia2013; Cubillo et al. Reference Cubillo, Smith, Barrett, Giampietro, Brammer, Simmons and Rubia2014b ; Rubia et al. Reference Rubia, Alegria, Cubillo, Smith, Brammer and Radua2014b ). The finding of right IFC-striatal up-regulation together with normalization of behavioural TD deficits extends previous evidence for modulation of behavioural TD rates with 5-HT (Schweighofer et al. Reference Schweighofer, Tanaka and Doya2007, Reference Schweighofer, Bertin, Shishida, Okamoto, Tanaka, Yamawaki and Doya2008) and of IFC-striatal activation with 5-HT modulators such as tryptophan depletion and SSRIs in healthy adults (Del-Ben et al. Reference Del-Ben, Deakin, McKie, Delvai, Williams, Elliott, Dolan and Anderson2005; Rubia et al. Reference Rubia, Lee, Cleare, Tunstall, Fu, Brammer and McGuire2005; Lamar et al. Reference Lamar, Cutter, Rubia, Brammer, Daly, Craig, Cleare and Murphy2009) in the ADHD population. It also extends our previous findings in ADHD that fluoxetine enhances and normalizes frontal activation during other impulsiveness-related functions such as IFC-striatal regions during inhibition (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Santosh, Brammer, Simmons, Murphy and Rubia2014c ) and dlPFC during working memory (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons and Rubia2014b ). The findings of right IFC modulation suggest that indoleamine agonists have similar effects to catecholamine agonists on ADHD brain function, given that not only methylphenidate but also atomoxetine up-regulated right IFC activation during inhibition and timing (Smith et al. Reference Smith, Cubillo, Barrett, Giampietro, Simmons, Brammer and Rubia2013; Cubillo et al. Reference Cubillo, Smith, Barrett, Giampietro, Brammer, Simmons and Rubia2014a , Reference Cubillo, Smith, Barrett, Giampietro, Brammer, Simmons and Rubia b ).
The fact that fluoxetine normalized both underactivation in right pre-/post-central gyri, insula and IPL and behavioural TD deficits in ADHD is in line with the role of lateral fronto-insular-striato-parietal circuitry in intertemporal choice (McClure et al. Reference McClure, Laibson, Loewenstein and Cohen2004; Bickel et al. Reference Bickel, Pitcock, Yi and Angtuaco2009; Xu et al. Reference Xu, Liang, Wang, Li and Jiang2009) and for the modulation of these regions by 5-HT (Long et al. Reference Long, Kuhn and Platt2009; Cools et al. Reference Cools, Nakamura and Daw2011; Rogers, Reference Rogers2011; Robinson et al. Reference Robinson, Cools and Sahakian2012). Apart from fronto-insular-striatal regions, IPL are also consistently underactivated in ADHD during EF (Cortese et al. Reference Cortese, Kelly, Chabernaud, Proal, Di Martino, Milham and Castellanos2012; Hart et al. Reference Hart, Radua, Mataix-Cols and Rubia2012, Reference Hart, Radua, Nakao, Mataix-Cols and Rubia2013). We have found left IPL underactivation to be normalized in ADHD with fluoxetine during inhibition (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Santosh, Brammer, Simmons, Murphy and Rubia2014c ). The up-regulation and normalization with fluoxetine of insula, pre-/post-central and IPL deficits in ADHD thus provides promising novel evidence for modulatory effects of 5-HT agonists on typically dysfunctional fronto-insular-parietal systems in ADHD.
The down-regulation of PCC/precuneus under fluoxetine v. placebo likely reflects deactivation of the default mode network (DMN), comprised of ACC, precuneus and PCC, thought to represent mind-wandering, and which is typically anti-correlated with task-positive regions as it needs to be switched off during cognitive effort (Northoff et al. Reference Northoff, Qin and Nakao2010). This is reinforced by the negative correlation under placebo of this cluster with the IFC activation. There is accumulating evidence that the DMN is insufficiently deactivated and anticorrelated with task-positive activation in ADHD (Christakou et al. Reference Christakou, Murphy, Chantiluke, Cubillo, Smith, Giampietro, Daly, Ecker, Robertson and Murphy2013b ), leading to enhanced mind-wandering, poor attention, EF and timing functions. We have found a similar effect of fluoxetine enhancing the deactivation of PCC during working memory in ADHD patients (Chantiluke et al. Reference Chantiluke, Barrett, Giampietro, Brammer, Simmons and Rubia2014b ), which we also observed with methylphenidate and atomoxetine (Cubillo et al. Reference Cubillo, Smith, Barrett, Giampietro, Brammer, Simmons and Rubia2014a ). The finding suggests that fluoxetine, like catecholamine agonists (Rubia et al. Reference Rubia, Alegria, Cubillo, Smith, Brammer and Radua2014b ), can strengthen the weak deactivation of the DMN in ADHD, presumably improving mind-wandering. Given that the key functional deficits in ADHD are both reduced activation in key fronto-striato-parietal networks mediating EF as well as a reduced deactivation of the DMN (Rubia et al. Reference Rubia, Alegria and Brinson2014a ), the findings suggest that a 5-HT agonist positively modulates both ‘task-positive’ as well as ‘task-negative’ activation deficits of not switching off the DMN.
5-HT is relatively ubiquitous in the brain. However, 5-HT modulates specifically ADHD-relevant impulsivity-related functions mediated by ventrolateral-prefrontal regions which are dependent on 5-HT input such as inhibitory control and reward-related decision-making (Dalley & Roiser, Reference Dalley and Roiser2012). The up-regulation with a 5-HT agonist of key right-hemispheric IFC-striatal activation that is typically abnormal in ADHD suggests that abnormal 5-HT may be underlying abnormal activation in these networks and not just catecholamine systems, in line with accumulating evidence of a role of 5-HT in ADHD (Barrickman et al. Reference Barrickman, Noyes, Kuperman, Schumacher and Verda1991; Spivak et al. Reference Spivak, Vered, Yoran-Hegesh, Averbuch, Mester, Graf and Weizman1999; Quintana et al. Reference Quintana, Butterbaugh, Purnell and Layman2007; Oades, Reference Oades2008; Gizer et al. Reference Gizer, Ficks and Waldman2009; McGough et al. Reference McGough, McCracken, Loo, Manganiello, Leung, Tietjens, Trinh, Baweja, Suddath, Smalley, Hellemann and Sugar2009; Rommelse et al. Reference Rommelse, Franke, Geurts, Hartman and Buitelaar2010; Zepf et al. Reference Zepf, Gaber, Baurmann, Bubenzer, Konrad, Herpertz-Dahlmann, Stadler, Poustka and Wöckel2010; Hinz et al. Reference Hinz, Stein, Neff, Weinberg and Uncini2011). However, it cannot be ruled out that fluoxetine had no indirect effects on other neurotransmitter systems which are known to be influenced by 5-HT such as dopamine, acetylcholine and other monoamines (Mongeau et al. Reference Mongeau, Blier and De Montigny1997; Bymaster et al. Reference Bymaster, Zhang, Carter, Shaw, Chernet, Phebus, Wong and Perry2002; Oades, Reference Oades2008). All three main monoamine systems are likely to interact in a concerted manner to mediate impulsiveness-relevant functions (Dalley & Roiser, Reference Dalley and Roiser2012).
The strength of this study is a carefully selected non-co-morbid ADHD group. Limitations are a relatively small patient sample size and the fact that for ethical and financial considerations, the control group was scanned only once while patients were scanned twice. However, the randomization accounted for potential training effects, and order did not affect the results. The significantly lower IQ in the ADHD group, typical for the population, is a limitation, in particular because IQ impacts upon decision-making (Toplak et al. Reference Toplak, Sorge, Benoit, West and Stanovich2010).However, covariance and correlation findings did not suggest that IQ confounded group differences. Finally, long-term stimulant use affects brain function and structure, so deficit findings may have been mitigated by the majority of patients taking stimulant medication (Hart et al. Reference Hart, Radua, Mataix-Cols and Rubia2012; Rubia et al. Reference Rubia, Alegria and Brinson2014a ).
Conclusions
A single fluoxetine dose in ADHD up-regulated activation in key right IFC-premotor-insular-striatal circuitry that mediates TD and which correlated trend-wise with better TD, and enhanced the deactivation of posterior DMN regions. Moreover, fluoxetine, via up-regulation of these right hemispheric regions, normalized underactivation in ADHD under placebo relative to controls in right premotor-insular-parietal areas and behavioural TD abnormalities. The findings show for the first time that a 5-HT agonist can modulate right IFC-insular-striato-parietal neural mechanisms underlying poor temporal foresight in ADHD. While the study aim was to clarify the mechanism of action of an acute dose of fluoxetine, which has the advantage of revealing true drug effects not confounded by indirect symptom improvements after chronic administration, future studies need to assess longer-term effects, as clinical behavioural changes are typically observed after weeks of administration. Longer-term SSRI administration has been shown to lead to down-regulation of 5-HT1A receptors and 5-HT transporters (Lesch et al. Reference Lesch, Hoh, Schulte, Osterheider and Müller1991) and may well have different effects on brain function than acute doses, which are clinically more informative.
Supplementary material
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0033291715002731.
Acknowledgements
C.C. was supported by an NIHR BRC PhD studentship, and L.N. was supported by an MRC-IOP excellence PhD studentship.
This work was supported by the UK Department of Health via the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) for Mental Health at South London and the Maudsley NHS Foundation Trust and Institute of Psychiatry, Psychology and Neuroscience, King's College London.
Declaration of Interest
K.R. has received funding from Lilly for another project and consultancy fees from Lilly, Shire, and Novartis. M.B. is a consultant for P1 Vital Ltd, Oxford, UK. The other authors have no conflict of interests to declare.
