Introduction
The disruptive behavioral disorders (DBDs) of conduct disorder (CD) and oppositional defiant disorder (ODD) are characterized by aggressive behavior, poor emotional regulation, and relationship difficulties (APA, 2013). There have been recent claims that the impairments shown by these patients might relate to forms of dysfunction in several different neurocognitive mechanisms that manifest as specific forms of behavioral disturbance (Blair, Reference Blair2013). Thus, considerable data indicate that a group of youth with DBDs show reduced amygdala responses to distress cues, the degree of which is positively associated with callous-unemotional (CU) traits (i.e. reduced guilt and empathy) and instrumental aggression (White et al. Reference White, Marsh, Fowler, Schechter, Adalio, Pope, Sinclair, Pine and Blair2012; Lozier et al. Reference Lozier, Cardinale, VanMeter and Marsh2014). These youths also showed atypical responses to reward and punishment within ventromedial prefrontal cortex (vmPFC) and caudate compared to typically developed youth (Finger et al. Reference Finger, Marsh, Mitchell, Reid, Sims, Budhani, Kosson, Chen, Towbin, Leibenluft, Pine and Blair2008, Reference Finger, Marsh, Blair, Reid, Sims, Ng, Pine and Blair2011) and youth with attention deficit hyperactivity disorder (ADHD; Finger et al. Reference Finger, Marsh, Mitchell, Reid, Sims, Budhani, Kosson, Chen, Towbin, Leibenluft, Pine and Blair2008). A recent study suggested that these functional difference reflect compromised representation of reinforcement expectancies within the vmPFC and aberrant prediction error signaling within the caudate (White et al. Reference White, Pope, Sinclair, Fowler, Brislin, Williams, Pine and Blair2013). There are also some data indicating a second group of youths show increased amygdala responses to threat and low CU traits (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Blair, Reference Blair2013; Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014). Indeed, the importance of CU traits is recognized in DSM-5 with the inclusion of the limited pro-social emotions specifier for CD (APA, 2013).
Two additional neurocognitive mechanisms have been hypothesized, when dysfunctional, to increase the risk of antisocial behavior/relate to the behavior problems of youth with CD (Patrick et al. Reference Patrick, Fowles and Krueger2009; Young et al. Reference Young, Friedman, Miyake, Willcutt, Corley, Haberstick and Hewitt2009; Miyake & Friedman, Reference Miyake and Friedman2012; Blair et al. Reference Blair, Vythilingam, Crowe, McCaffrey, Ng, Wu, Scaramozza, Mondillo, Pine, Charney and Blair2013). The first of these is top-down attention [related to prefrontal (dorsomedial and lateral) regions; Buhle et al. Reference Buhle, Silvers, Wager, Lopez, Onyemekwu, Kober, Weber and Ochsner2014]. Top-down attention is implicated in emotional regulation in explicit cognitive reappraisal paradigms where subjects alter stimulus representations by priming non-emotional features (Buhle et al. Reference Buhle, Silvers, Wager, Lopez, Onyemekwu, Kober, Weber and Ochsner2014) and implicit attention distraction paradigms (e.g. the Affective Stroop task; aST) where subjects prime task features at the expense of the representation of emotional distracters (Pessoa et al. Reference Pessoa, Padmala and Morland2005; Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007). As such dysfunction in top-down attention might lead to emotional dysregulation and an increased risk for reactive aggression. Indeed, studies have reported increased amygdala responses to negative stimuli in youth with conduct problems and low CU traits (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014). It has been suggested that this might reflect deficient top-down attention-based emotion regulation (Blair, Reference Blair2013; Blair et al. Reference Blair, Vythilingam, Crowe, McCaffrey, Ng, Wu, Scaramozza, Mondillo, Pine, Charney and Blair2013). However, no previous functional magnetic resonance imaging (fMRI) work has investigated this possibility.
The second neurocognitive mechanism hypothesized, when dysfunctional, to increase the risk of antisocial behavior/relate to the behavior problems of youth with CD is response control/response inhibition (Patrick et al. Reference Patrick, Fowles and Krueger2009; Young et al. Reference Young, Friedman, Miyake, Willcutt, Corley, Haberstick and Hewitt2009; Miyake & Friedman, Reference Miyake and Friedman2012). Response control/response inhibition is thought to be mediated by dorsomedial and inferior frontal/anterior insula cortices (Criaud & Boulinguez, Reference Criaud and Boulinguez2013). Response control/inhibition is important for avoiding sub-optimal choices and can be indexed by the Stop, Go/No-Go and Stroop tasks (Criaud & Boulinguez, Reference Criaud and Boulinguez2013). Impairment in response control/inhibition should result in an individual who will ‘impulsively’ express behaviors (including antisocial behaviors) that are non-optimal for the situation. Such impairment has also been associated with an increased risk for antisocial behavior (Patrick et al. Reference Patrick, Fowles and Krueger2009; Young et al. Reference Young, Friedman, Miyake, Willcutt, Corley, Haberstick and Hewitt2009; Miyake & Friedman, Reference Miyake and Friedman2012).
The current study uses the aST to investigate emotional responding, automatic emotion regulation and response control/inhibition in youth with DBD and high and low callous-unemotional traits (HCU/LCU). Considerable data, including from the aST and related tasks, demonstrate that the performance of a cognitive task reduces the response within the amygdala to an emotional stimulus (Critchley et al. Reference Critchley, Daly, Phillips, Brammer, Bullmore, Williams, Van Amelsvoort, Robertson, David and Murphy2000; Pessoa et al. Reference Pessoa, McKenna, Gutierrez and Ungerleider2002; Erthal et al. Reference Erthal, de Oliveira, Mocaiber, Pereira, Machado-Pinheiro, Volchan and Pessoa2005; Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007; Mitchell et al. Reference Mitchell, Nakic, Fridberg, Kamel, Pine and Blair2007); i.e. participants undertaking paradigms such as the aST demonstrate automatic emotion regulation. Our goal in using the aST is to elucidate the neurocircuitry dysfunction related to symptom manifestation across disorders (in this case CD and ODD), thus departing from diagnosis-based approach to a mechanism-based approach towards the understanding of pathophysiology in DBD (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010; Cuthbert & Insel, Reference Cuthbert and Insel2013).
We predicted: (i) consistent with previous work (White et al. Reference White, Marsh, Fowler, Schechter, Adalio, Pope, Sinclair, Pine and Blair2012; Lozier et al. Reference Lozier, Cardinale, VanMeter and Marsh2014; Baker et al. Reference Baker, Clanton, Rogers and De Brito2015), DBD-HCU youth would show reduced amygdala responsiveness to threatening stimuli relative to healthy youth; (ii) consistent with previous work (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014, Baker et al. Reference Baker, Clanton, Rogers and De Brito2015), DBD-LCU youth would show increased amygdala responsiveness to threatening stimuli relative to healthy youth; (iii) and that amygdala responsiveness would be inversely associated with CU traits in youth with DBD; (iv) on the basis of previous hypotheses (Blair et al. Reference Blair, Vythilingam, Crowe, McCaffrey, Ng, Wu, Scaramozza, Mondillo, Pine, Charney and Blair2013), we predicted DBD-LCU youth would show reduced recruitment of attention-based emotion regulation regions (dorsomedial and lateral frontal cortex; Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007) relative to healthy youth and DBD-HCU youth; (v) and that DBD youth would show reduced recruitment of regions implicated in response control (anterior insula/inferior frontal and dorsomedial frontal cortex) relative to healthy youth with responsiveness being inversely associated with ADHD symptoms in DBD youth; and (vi) consistent with previous functional connectivity studies (Marsh et al. Reference Marsh, Finger, Mitchell, Reid, Sims, Kosson, Towbin, Leibenluft, Pine and Blair2008; Herpers et al. Reference Herpers, Scheepers, Bons, Buitelaar and Rommelse2014), we predicted DBD-HCU youth would show reduced connectivity between the amygdala and cortical regions to threatening stimuli relative to healthy youth and DBD-LCU youth and that level of connectivity would be inversely associated with CU traits in youth with DBD.
Method
Participants
Sixty-seven youths participated: 29 healthy and 38 with DBD (CD/ODD). Participants were recruited from the community through newspaper adverts, fliers, and referrals from area mental health practitioners. Four participants (healthy n = 1, DBD n = 3) were excluded (due to, for example, excessive movement). Thus, data from 28 healthy (average age = 13.88 years, 13 females) and 35 DBD (average age = 14.81 years, 13 females) participants were analyzed (see Table 1). Statements of informed assent/consent were obtained from participating children/parents. This study was approved by the NIMH IRB.
Characteristics of healthy youth, and youth with DBD
DBD, Disruptive behavioral disorder; LCU, low callous-unemotional; HCU, high callous-unemotional; CD, conduct disorder; ODD, oppositional defiant disorder; ADHD, attention deficit hyperactivity disorder; SA, substance abuse.
a Between DBD-LCU and DBD-HCU.
b All cannabinoid abuse except one in HCU group (alcohol abuse).
c DBD-LCU < DBD + HCU: t 33 = 7.817, p = 0.000.
d Healthy youth < DBD-LCU: t 43 = 9.379, p = 0.000; DBD-LCU = DBD + HCU: t 26 = 0.540, p = 0.594.
e Atomoxetine (n = 1); lamotrigine (n = 1); lamotrigine + aripiprazole (n = 1); amphetamine, intuitiv, risperidone (n = 1).
f Amphetamine (n = 4); atomoxetine + intuitiv (n = 1).
Participants’ parents completed the Inventory of Callous-Unemotional Traits – Parent Version (ICU-P). The ICU-P is a 24-item scale assessing CU traits in youth with good construct validity (Frick, Reference Frick2004; Kimonis et al. Reference Kimonis, Frick, Skeem, Marsee, Cruise, Munoz, Aucoin and Morris2008) and reliability (Cronbach's α = 0.81). Following previous studies (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Lozier et al. Reference Lozier, Cardinale, VanMeter and Marsh2014), we divided the patients with DBD into two groups on the basis of a median split of the ICU-P scores (median score = 42; LCU/HCU: n = 17/18) (see Table 1). Previous community sample studies reveal average ICU scores between, for example, 22 and 31 (standard deviation: 7.88–10.98; Roose et al. Reference Roose, Bijttebier, Decoene, Claes and Frick2010; Byrd et al. Reference Byrd, Kahn and Pardini2013). In contrast, clinical samples of patients with DBD and forensic samples reveal average ICU scores of 41 (White et al. Reference White, Cruise and Frick2009). As such, all HCU groups showed a level of CU that was above average for patients with DBD and notably greater than that shown by healthy populations.
All youths and their parents completed the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; Kaufman et al. Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci, Williamson and Ryan1997). Assessments were conducted by a doctoral-level clinician and supervised by expert child/adolescent psychiatrists. The K-SADS has demonstrated good validity and inter-rater reliability (Kaufman et al. Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci, Williamson and Ryan1997). The parents of 26/28 healthy youths and 28/35 youths with DBD completed the Conners Parent Rating Scale for ADHD – version 2 (Conners et al. Reference Conners, Sitarenios, Parker and Epstein1998). IQ was assessed with the Wechsler Abbreviated Scale of Intelligence (2-subtest form; Wechsler, Reference Wechsler1999). Exclusion criteria are listed in the Supplementary material, section 1. The groups did not differ significantly in terms of age, sex, handedness or IQ (see Table 1).
Experimental task
We used an adapted version of the aST described previously (Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007; Hwang et al. Reference Hwang, White, Nolan, Sinclair and Blair2014) (see Fig. 1). In each trial, participants saw a central fixation point (400 ms), a positive, neutral, or negative image (400 ms), either a numerical array on task trials, or a blank screen on view trials (400 ms), the same image previously displayed (400 ms), and a second blank screen (1300 ms). For task trials, participants pressed a button corresponding to how many numbers were displayed (numerosity: 3–6). On congruent trials, numerosity matched the actual number values displayed (e.g. three for 3 s). On incongruent trials, numerosity did not match the number values displayed (e.g. four for 3 s or six for 3 s). The numerical gap between numerosity and the number values ranged between 1 (e.g. four for 3 s) and 3 (e.g. six for 3 s). Participants were free to respond at any time between the initial numerical presentation and the end of the blank screen display (response window: 1700 ms). Participants made no response for view trials.
Example trial sequences. (a) Negative view trial; (b) negative congruent trial; (c) negative incongruent trial.
The images consisted of 48 positive, 48 negative, and 48 neutral pictures selected from the International Affective Picture System (Lang & Cuthbert, Reference Lang and Cuthbert2005) (see Supplementary material, section 2 for mean valence and arousal values by stimulus class). Participants completed two runs. Each involved 288 trials [32 in each nine categories (3 image type × 3 task type)] and 96 fixation trials (each of 2500 ms length to generate a baseline). Trial order was randomized across participants.
Image acquisition and analysis
Whole-brain blood oxygen level-dependent (BOLD) fMRI data were acquired using a 3-T GE MRI scanner. Following sagittal localization, functional T2*-weighted images were acquired using an echo-planar single-shot gradient echo pulse sequence with a matrix of 64 × 64 mm2, repetition time (TR) of 3000 ms, echo time (TE) of 30 ms, field of view (FOV) of 240 mm, and voxels of 3.75 × 3.75 × 4 mm3. Images were acquired in 30 continuous 4 mm axial slices per brain volume across two runs. The duration of each run was 8 min 13 s. In the same session, a high-resolution T1-weighed anatomical image was acquired to aid with spatial normalization (three-dimensional Spoiled GRASS; TR = 8.1 ms; TE = 3.2 ms, flip angle 20°; FOV = 240 mm, 128 axial slices, thickness = 1.0 mm; 256 × 256 acquisition matrix).
fMRI analysis
Data were analyzed within the framework of a random effects general linear model using Analysis of Functional Neuroimages (AFNI). Both individual- and group-level analyses were conducted. The first five volumes in each scan series, collected before equilibrium magnetization was reached, were discarded. Motion correction was performed by registering all volumes in the EPI dataset to a volume that was collected shortly before acquisition of the high-resolution anatomical dataset.
The EPI datasets for each participant were spatially smoothed (using an isotropic 6 mm Gaussian kernel) to reduce the influence of anatomical variability among the individual maps in generating group maps. Next, the time-series data were normalized by dividing the signal intensity of a voxel at each time point by the mean signal intensity of that voxel for each run and multiplying the result by 100. Resultant regression coefficients represented a percent signal change from the mean. The model involved six motion regressors and the following nine task regressors: negative congruent, negative incongruent, negative view, neutral congruent, neutral incongruent, neutral view, positive congruent, positive incongruent and positive view. A regressor modeling incorrect responses was also included. All regressors were convolved with a canonical hemodynamic response function (HRF) to account for the slow hemodynamic response (with time point commencing at time of first image onset). There was no significant regressor collinearity.
The participants’ anatomical scans were individually registered to the Talairach and Tournoux atlas (Talairach & Tournoux, Reference Talairach and Tournoux1988). The individuals’ functional EPI data were then registered to their Talairach anatomical scan within AFNI. Linear regression modeling was performed using the 10 regressors (nine task plus incorrect responses) described earlier, plus regressors to model a first-order baseline drift function. This produced β coefficients and associated t statistics for each voxel and regressor.
The BOLD data were analyzed via a 3 (group: healthy youth, youth with DBD-LCU, youth with DBD-HCU) × 3 (emotion: negative, positive, neutral) × 3 (task: congruent, incongruent, view) ANOVA. Statistical maps were created for each main effect and interaction by thresholding at a single-voxel p value of p < 0.005. ClustSim was then applied to these results yielding a minimum cluster size (22 voxels) with a map-wise false-positive probability of p < 0.05, corrected for multiple comparisons.
Given our a priori hypotheses, regions of interest (ROIs), taken from the AFNI software's anatomical maps (TT_Daemon atlas) were obtained from the amygdala (Talairach & Tournoux, Reference Talairach and Tournoux1988). A small volume-corrected ROI analysis via ClustSim was used on these regions (initial threshold: p < 0.02, k = 13, corrected p < 0.05).
Follow-up analyses were performed to facilitate interpretations. For these analyses, average percent signal change was measured across all voxels within each ROI generated from the functional masks, and data were analyzed using appropriate follow-up independent t tests within SPSS v. 22 (SPSS Inc. USA).
Context-dependent psychophysiological interaction (gPPI) analysis
A gPPI analysis was conducted to examine group differences in functional connectivity following the method described by McLaren and colleagues (Reference McLaren, Ries, Xu and Johnson2012). Our main goal was to examine group differences in functional connectivity between the amygdala and cortical regions. We took as a seed the region of right amygdala (coordinates: 25.5, −1.5, −12.5) showing a main effect of emotion from the main ANOVA conducted on the BOLD data (see Supplementary Table S2). This seed can be considered relatively unbiased by group membership as it was identified by main effect of emotion (i.e. significant activity to emotion was seen within all groups). The average activation from this seed region was extracted across the time series. Interaction regressors were created by multiplying each of these average time series with nine task time-course vectors (one for each task and emotion condition) which were coded 1 or 0 for task and emotion condition present or absent. The average activation for the seeds was entered into a linear regression model along with the nine interaction regressors and six motion regressors. A 3 (group) ×3 (task) ×3 (emotion) whole-brain repeated-measures ANOVA was then applied to the data, and the regions showing significant group×emotion interaction were reported.
Results
Behavioral data
Two 3 (group: DBD-HCU, DBD-LCU, healthy) ×3 (emotion: positive, neutral, negative) ×2 (task: congruent, incongruent) ANOVAs were applied to the reaction time (RT) and accuracy data (see Supplementary Table S1). With respect to RT, there was a significant main effect of task (incongruent > congruent: F 1,60 = 169.349, p < 0.001) and a trend for emotion (F 2,59 = 2.898, p = 0.059, negative and positive > neutral, t 62 = 1.746 and 1.924, p = 0.086 and 0.059, respectively). With respect to accuracy, there was a significant main effect of task (incongruent < congruent: F 1,60 = 22.565, p < 0.001) and group (F 2,60 = 4.578, p = 0.014; DBD-LCU < DBD-HCU and healthy youth, t 43,33 = 2.352 and 2.486, p = 0.023 and 0.018, respectively). The performance of youth with DBD-HCU and healthy youth did not significantly differ (t 44 = 1.056, p = 0.297). No other main effects or interactions for either ANOVA were significant.
Movement data
There were no significant group differences in movement parameters (F 1,60 = 1.484–2.981, p > 0.1).
MRI data: main analysis
A whole-brain 3 (group) ×3 (emotion) ×3 (task) ANOVA was applied to the BOLD data. This revealed regions showing significant group × emotion, group × task and group × task × emotion interactions. Regions showing main effects of task and emotion and task × emotion interactions are presented in the Supplementary material, section 3.
Group × emotion interaction
There was a group×emotion interaction within left vmPFC and right (but not left) amygdala ROI (see Table 2, Fig. 2a, d ). Within both regions, youth with DBD-HCU showed significantly decreased activation to negative relative to neutral stimuli, compared to healthy youth and youth with DBD-LCU who did not significantly differ (vmPFC: t = 3.573 and 3.891, p < 0.001; right amygdala: t = 2.491 and 2.312, p = 0.017 and 0.027) (see Fig. 2b, e ). There were no group differences in either region's response to positive relative to neutral stimuli (F = 0.827 and 1.771, p > 0.05).
Regions showing a significant group×emotion interaction: (a) blood oxygen level-dependent (BOLD) response data: left ventromedial prefrontal cortex (vmPFC) (coordinates: −4.5, 43.5, −12.5, at = 0.005); (b) parameter estimates for left vmPFC; (c) negative correlation between symptom severity of callous-unemotional trait measured by the Inventory of Callous-Unemotional Traits (ICU) (x-axis) and BOLD response parameter estimates of negative relative to neutral trials (y-axis) in left vmPFC; (d) right amygdala ROI (at p = 0.05); (e) parameter estimates for this region: context-dependent psychophysiological interaction data with right amygdala seed; (f) left inferior frontal gyrus (coordinates: −40.5, 40.5, −0.5 at p = 0.005) and; (g) parameter estimates for this region. Abbreviations: Neg, negative; Neu, neutral; Pos, positive; Healthy, healthy youth; LCU, youth with DBD-LCU; HCU, youth with DBD-HCU. * Significant contrasts for interaction variables (p < 0.05). The results are shown on the Talairach space.
(a) Brain regions showing a significant interaction in comparison between healthy youth, youth with DBD-LCU and youth with DBD-HCU. (b) Brain regions showing a significant interaction of connectivity with right amygdala seed in comparison between healthy youth, youth with DBD-LCU, and youth with DBD-HCU
DBD, Disruptive behavioral disorder; LCU, low callous-unemotional; HCU, high callous-unemotional; BA, Brodmann area.
a According to the Talairach Daemon Atlas (http://www.nitrc.org/projects/tal-daemon).
b Below the ClusterSim cluster size (22 voxels).
Group × task interaction
There was a group×task interaction within left insula (see Table 2). Within this region, youth with DBD-LCU and youth with DBD-HCU did not differ (t = 1.542, p = 0.133). However, both showed a significantly decreased differential response to incongruent task trials relative to view trials (t = 3.471 and 2.579, p = 0.001 and 0.013) and to incongruent relative to congruent task trials (t = 3.517 and 3.406, p = 0.001) compared to healthy youth (see Fig. 3d, e ). There were no group differences in differential response to congruent relative to view trials (t = 0.621 and 0.118, p = 0.538 and 0.907). No other regions showed a significant group × task interaction. While left insula did not survive multiple comparison correction (k = 22), this likely reflects a Type II error; the reduced insula activity (as well as left inferior parietal lobule) was seen in both groups of youths with DBD and an exploratory ANOVA contrasting healthy youth with a combined DBD group revealed a highly significant group×task interaction within this region (k = 57) (see Fig. 3a, b ). The only other region showing a significant group×task interaction for this second analysis was the left inferior parietal lobule (k = 28; for analysis details, see Supplementary material, section 4).
Regions showing a significant group×task interaction: (a) bilateral insula (coordinates: 37.5, −13.5, −6.5; −37.5, 7.5, −0.5 at p = 0.005) via Healthy v. DBD ANOVA; (b) parameter estimates for this region; (c) negative correlation between ADHD symptom severity measured by Conners Parent Rating Scale (x-axis) and blood oxygen level-dependent response parameter estimates of congruent relative to view trials (y-axis); (d) left insula (coordinates: −37.5, 7.5, −0.5 at p = 0.005) via Healthy v. DBD-LCU v. DBD-HCU ANOVA; (e) parameter estimates for this region. Abbreviations: Incong, incongruent trial; Cong, congruent trial; View, view trial; Healthy, healthy youth; DBD, youth with DBD; LCU, youth with DBD-LCU; HCU, youth with DBD-HCU. * Significant contracts for interaction variables (p < 0.05). The results are shown on the Talairach space.
Group × task × emotion interaction
There was a significant group×emotion×task interaction within right superior frontal gyrus and bilateral caudate (see Table 2). Within both regions, youth with DBD-HCU showed greater activity on negative incongruent trials relative to comparison groups (t = 2.013–3.319, p = 0.002–0.005), and youth with DBD-LCU showed greater activity on positive incongruent trials than both comparison groups (t = 2.438–2.668, p = 0.011–0.020). All other contrasts were not significant except within bilateral caudate where healthy youth showed less activity than DBD-LCU for negative congruent trials (t = 2.146, p = 0.038).
MRI results: gPPI results
A 3 (group) ×3 (emotion) ×3 (task) ANOVA was conducted on the gPPI data using the right amygdala seed. Regions displaying a significant group×emotion interaction included left inferior frontal gyrus, left posterior cingulate gyrus, left caudate, and left insula (see Table 2). Youth with DBD-LCU compared to healthy youth and youth with DBD-HCU showed significantly reduced connectivity between the right amygdala seed and these regions in response to emotional (negative and positive) relative to neutral stimuli (t = 2.452–5.355, p = 0.000–0.018); although for left posterior cingulate gyrus in response to negative relative to neutral stimuli (t = 1.916, p = 0.062; see Fig. 2f, g for left inferior frontal gyrus). Healthy youth and youth with DBD-HCU showed no significant differences in gPPI connectivity (t = 0.067–1.947, p = 0.058–0.947), except that youth with DBD-HCU showed significantly increased connectivity between right amygdala and caudate in response to positive relative to neutral stimuli relative to healthy youth (t = 2.152, p = 0.037).
Correlations with symptom severity
Ten correlations were conducted examining the relationship between BOLD response parameters and symptom severity in the patients with DBD. CU symptom severity was negatively correlated with differential (negative-neutral) BOLD response for the vmPFC (r = −0.370, p = 0.026) but not the amygdala (r = −0.238, p > 0.05) (see Fig. 2c ). However, it was positively correlated with amygdala-caudate connectivity in response to the negative relative to neutral stimuli (r = 0.420, p = 0.012). ADHD symptom severity (as indexed by the Conners Parent Rating Scale) was negatively correlated with differential (congruent-view, but not incongruent-view or incongruent-congruent) BOLD response within left insula (r = −0.477, p = 0.010) (see Fig. 3c ). Following a reviewer's suggestion and for completion, we also examined the relationship between CU symptom severity and differential (incongruent-view, congruent-view, and incongruent-congruent) BOLD response within left insula. However, these were non-significant (r = 0.188, −0.084, 0.105, p = 0.280, 0.630, 0.459, respectively). In addition, we examined the relationship between ADHD symptom severity and differential (negative-neutral) BOLD response for the vmPFC and amygdala. However, these were also non-significant (vmPFC: r = −0.028, p = 0.889; amygdala: r = −0.174, p = 0.377).
Potential confounds
We conducted analyses excluding youth on psychotropic medications and substance abusers. These analyses revealed similar results to the main analysis reported above (see Supplementary material, sections 5 and 6).
Discussion
We investigated emotional responding, automatic emotion regulation and response control/inhibition in youth with DBD and HCU/LCU. There were two main results: first, youth with DBD-HCU showed significantly decreased left vmPFC and right amygdala activation to negative relative to neutral stimuli, compared to healthy youth and youth with DBD-LCU. Moreover, the vmPFC response to negative v. neutral stimuli was inversely related to level of CU traits in the patients with DBD. Second, youth with DBD (LCU and HCU) showed decreased activation of bilateral insula on task trials relative to healthy youth. Insula responsiveness was inversely related to ADHD symptomatology in youth with DBD.
In line with our first prediction and previous work (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; White et al. Reference White, Marsh, Fowler, Schechter, Adalio, Pope, Sinclair, Pine and Blair2012; Lozier et al. Reference Lozier, Cardinale, VanMeter and Marsh2014), youth with DBD-HCU showed reduced amygdala responses to threat stimuli relative to comparison youth. In addition, the current study extends the literature in two ways. First, it indicates that reduced amygdala recruitment is specific for negative relative to positive emotional stimuli (though it remains possible that amygdala responding to happy expressions may be disrupted in youth with DBD-HCU; cf. Fusar-Poli et al. Reference Fusar-Poli, Placentino, Carletti, Landi, Allen, Surguladze, Benedetti, Abbamonte, Gasparotti, Barale, Perez, McGuire and Politi2009). Second, it indicates dysfunction in emotional responding in both the amygdala and vmPFC. vmPFC and amygdala responsiveness to negative relative to neutral stimuli were correlated in all three groups (see Supplementary material, section 7) and level of vmPFC response was inversely related to CU traits level in patients with DBD. The relationship between the amygdala and vmPFC is complex. vmPFC may regulate amygdala activity (Milad & Quirk, Reference Milad and Quirk2012). However, vmPFC lesions suppress amygdala activity and ‘protect’ the individual from the development of PTSD/depression (Koenigs & Grafman, Reference Koenigs and Grafman2009). These latter results are consistent with a more interactive role where valence information is provided by the amygdala to vmPFC for representation (Schoenbaum et al. Reference Schoenbaum, Roesch and Stalnaker2006). We assume that the current data of decreased activation in vmPFC and amygdala for youth with HCU reflects a failure in this interaction (cf. Marsh et al. Reference Marsh, Finger, Mitchell, Reid, Sims, Kosson, Towbin, Leibenluft, Pine and Blair2008; Motzkin et al. Reference Motzkin, Newman, Kiehl and Koenigs2011).
In contrast to our second prediction, youth with DBD-LCU did not show significantly increased amygdala responses to threat stimuli relative to healthy youths (only a non-significant trend) though their amygdala responses to threat stimuli were significantly greater than those of youth with HCU. It should be noted that while some previous studies have reported increased amygdala responses to negative stimuli in youth with conduct problems and LCU (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014), not all studies have (Lozier et al. Reference Lozier, Cardinale, VanMeter and Marsh2014). However, previous work has consistently shown, as was seen here, that youth with DBD-LCU show increased amygdala responses relative to youth with DBD-HCU [Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; White et al. Reference White, Marsh, Fowler, Schechter, Adalio, Pope, Sinclair, Pine and Blair2012; cf. prediction (iii), Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014]. Moreover, it is worth noting that while the group with DBD-HCU was selected for showing elevated CU traits, the group with DBD-LCU was selected for not showing CU traits. Selecting a second group of youth with DBD for impairment potentially associated with heightened threat sensitivity (possibly irritability; Thomas et al. Reference Thomas, Hall, Skup, Jenkins, Pine and Leibenluft2011) might prove beneficial in future research.
Our fourth prediction, that DBD-LCU patients would show reduced recruitment of attention-based emotion regulation regions (dorsomedial and lateral frontal cortex; Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007) relative to healthy youth and DBD-HCU patients was not supported. The suggestion had been that increased emotional responsiveness in youth with DBD-LCU might reflect a failure on top-down attention-driven emotion regulation (cf. Blair et al. Reference Blair, Smith, Mitchell, Morton, Vythilingam, Pessoa, Fridberg, Zametkin, Sturman, Nelson, Drevets, Pine, Martin and Blair2007). However, no regions showed significant group×task or group×task×emotion interactions that were consistent with reduced recruitment of systems implicated in top-down attention in patients with DBD-LCU. It should be noted though that youth with DBD-LCU showed decreased connectivity between the amygdala and both (right/left) insula and (right/left) inferior frontal cortex. These regions have been implicated in some accounts of emotion regulation in previous work (Davidson et al. Reference Davidson, Putnam and Larson2000 Gold et al. Reference Gold, Morey and McCarthy2015). The youth with DBD-LCU thus show some pathology consistent with impaired emotional regulation. However, it is important to note that though they showed significantly increased amygdala and vmPFC responsiveness to negative stimuli relative only to youth with DBD-HCU (there was increased responsiveness relative to healthy youth but this was not statistically significant). As such data from this study did not indicate heightened responsiveness to aversive stimuli (i.e. emotional dysregulation) in youth with DBD-LCU relative to healthy youth though this has been reported in other studies (Viding et al. Reference Viding, Sebastian, Dadds, Lockwood, Cecil, De Brito and McCrory2012; Sebastian et al. Reference Sebastian, McCrory, Dadds, Cecil, Lockwood, Hyde, De Brito and Viding2014).
In line with predictions, youth with DBD (HCU and LCU) showed reduced task-related bilateral anterior insula cortex activity, a region implicated in response control (Chambers et al. Reference Chambers, Garavan and Bellgrove2009), during incongruent trials relative to healthy youth. This is consistent with previous reports of insula dysfunction in DBD (Crowley et al. Reference Crowley, Dalwani, Mikulich-Gilbertson, Du, Lejuez, Raymond and Banich2010; Fairchild et al. Reference Fairchild, Hagan, Passamonti, Walsh, Goodyer and Calder2014; White et al. Reference White, Fowler, Sinclair, Schechter, Majestic, Pine and Blair2014), and the relationship between dysfunctional response inhibition and externalizing behaviors (Young et al. Reference Young, Friedman, Miyake, Willcutt, Corley, Haberstick and Hewitt2009; Patrick et al. Reference Patrick, Venables, Yancey, Hicks, Nelson and Kramer2013) particularly impulsivity (Loeber et al. Reference Loeber, Burke and Pardini2009). BOLD responses in anterior insula cortex correlated inversely with ADHD symptom severity in youth with DBD. This indicates a second form of pathophysiology in DBD related not to CU but rather impulsiveness and doubtless exacerbating antisocial and risky behavior (such as substance abuse) in these youth (Crowley et al. Reference Crowley, Dalwani, Mikulich-Gilbertson, Du, Lejuez, Raymond and Banich2010; Blair et al. Reference Blair, Vythilingam, Crowe, McCaffrey, Ng, Wu, Scaramozza, Mondillo, Pine, Charney and Blair2013).
In contrast to our final prediction, youth with DBD-HCU did not show reduced connectivity between the amygdala and cortical regions to threatening stimuli relative to comparison youth and DBD-LCU youth. Instead, youth with DBD-HCU showed increased connectivity between right amygdala and caudate in response to positive relative to neutral stimuli compared to healthy youth. This contrasts with previous functional connectivity findings indicating reduced connectivity between the amygdala and particularly vmPFC in patients with high psychopathic traits (Marsh et al. Reference Marsh, Finger, Mitchell, Reid, Sims, Kosson, Towbin, Leibenluft, Pine and Blair2008, Reference Marsh, Finger, Fowler, Jurkowitz, Schechter, Yu, Pine and Blair2011; Motzkin et al. Reference Motzkin, Newman, Kiehl and Koenigs2011). However, it should be noted that these previous studies reflect connectivity during either resting state (Motzkin et al. Reference Motzkin, Newman, Kiehl and Koenigs2011) or across all task conditions (Marsh et al. Reference Marsh, Finger, Mitchell, Reid, Sims, Kosson, Towbin, Leibenluft, Pine and Blair2008, Reference Marsh, Finger, Fowler, Jurkowitz, Schechter, Yu, Pine and Blair2011). The current study investigated group differences in differential connectivity across specific conditions. It is thus possible that while amygdala-vmPFC global connectivity is reduced in youth with elevated CU traits, any increase in connectivity for emotional relative to neutral stimuli is comparable for youth with DD-HCU and healthy youth.
Implications for treatment
Our results support suggestions that CU traits/emotional responsiveness should be considered when assessing patients with DBD. Youth with DBD-HCU showed decreased activation in the areas of emotional responsiveness including amygdala and vmPFC, which may lead to exercising proactive aggression, whereas youth with DBD-LCU showed decreased connectivity between amygdala and inferior frontal cortex which may lead to difficulty in emotion regulation and in turn exercising more of reactive aggression (Blair et al. Reference Blair, Vythilingam, Crowe, McCaffrey, Ng, Wu, Scaramozza, Mondillo, Pine, Charney and Blair2013, Reference Blair, Leibenluft and Pine2014). Optimal treatment for patients with hypo-emotionality may differ from patients with hyper-emotionality. Indeed, youth with high CU traits benefit less from current interventions (Frick et al. Reference Frick, Ray, Thornton and Kahn2014). The current data are not particularly supportive of interventions designed to augment emotional regulation in youth with DBD-LCU but this may reflect patient assessment and/or the form of emotion regulation assessed (Gyurak et al. Reference Gyurak, Gross and Etkin2011). The current data support suggestions that response control dependent on anterior insula cortex might be deficient in patients with DBD (Young et al. Reference Young, Friedman, Miyake, Willcutt, Corley, Haberstick and Hewitt2009; Patrick et al. Reference Patrick, Venables, Yancey, Hicks, Nelson and Kramer2013) and it may well be an integral part of assessment for youth with DBD.
Limitations and conclusion
Two caveats should be considered: first, we included youth with substance abuse and those receiving medication treatment. However, subsequent analyses excluding these subjects yielded similar results to the main analysis (see Supplementary material, sections 5 and 6). Second, there were relatively few group differences in behavioral task performance. The patients with DBD-LCU were less accurate in their responding than both the healthy youth and youth with DBD-HCU who did not differ in performance. However, this was seen for both congruent and incongruent trials and there were no group differences in impact of emotional distracters. This likely reflects the relatively minor differential effect of emotional distracters in this task (there were only trends for trials involving positive or negative emotional distracters to be slower than trials involving neutral distracters). Given the relatively weak impact of these distracters here, it is less surprising that we did not observe, for example, an anticipated reduction in interference from emotional distracters in the youth with DBD-HCU (Mitchell et al. Reference Mitchell, Richell, Leonard and Blair2006). More arousing/negatively valenced distracters might have been more successful in producing group differences in behavior. However, such stimuli are unlikely to be considered ethical for research with adolescents.
In summary, we demonstrated two forms of pathophysiology in youth with DBD that related to different forms of behavioral impairment. One is associated with reduced amygdala and vmPFC responses to negative stimuli and related to increased CU traits. Another with reduced insula responses during response control and related to ADHD symptoms. Appropriate assessment/intervention will need to be individualized according to specific pathophysiology of youth with DBD.
Supplementary material
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0033291716000118.
Acknowledgements
Drs Hwang, White, Sinclair, and Blair are with the Intramural Research Program at the National Institute of Mental Health, National Institutes of Health. This work was supported by the Intramural Research Program at the National Institute of Mental Health, National Institutes of Health under grant number 1-ZIA-MH002860-08 to Dr Blair. Ethics approval for this study was granted by the NIH Combined Neuroscience Institutional Review Board under protocol number 05-M-0105.
