Adolescence and reward processing

Adolescence is characterized by increased social interactions, risk taking, impulsivity, and reward sensitivity (Casey, Jones, & Hare, Reference Casey, Jones and Hare2008). This crucial transitional developmental period is also marked by drastic neurobiological and hormonal changes. Preclinical studies have shown that various components of the dopamine system go through extensive structural changes during adolescence, i.e. there is a peak in dopamine transporter and D1-like receptor density in the rat striatum (Moll et al., Reference Moll, Mehnert, Wicker, Bock, Rothenberger, Rüther and Huether2000; Tarazi, Tomasini, & Baldessarini, Reference Tarazi, Tomasini and Baldessarini1999), and ventral tegmental area (VTA) dopamine neurons in adolescent rats fire faster than in adult rats (McCutcheon et al., Reference McCutcheon, Conrad, Carr, Ford, McGehee and Marinelli2012). Neuroimaging work strongly supports the notion that the greater reward-seeking behavior during adolescence is reflected by a hyper-responsive reward system in the brain (Galvan, Reference Galvan2010). In addition, developmental cognitive neuroscience studies have shown adolescent-specific influences on reward system function, with elevated recruitment of subcortical limbic systems relative to top-down (cortical) control systems during adolescence compared to childhood and adulthood (Galvan et al., Reference Galvan, Hare, Parra, Penn, Voss, Glover and Casey2006; Steinberg et al., Reference Steinberg, Albert, Cauffman, Banich, Graham and Woolard2008). Casey, Jones, and Somerville (Reference Casey, Jones and Somerville2011) have proposed a neurobiological model of adolescent development describing the imbalance between an accelerated responsiveness to motivational cues and immaturity in cognitive control (Casey et al., Reference Casey, Jones and Somerville2011).

These findings suggest that structural and functional changes in corticosubcortical circuitry during this developmental period may predispose adolescents to a higher sensitivity to reward-related cues, such as cooperative peer-interaction.

Susceptibility to peer influence, brain function, and risk taking behavior

Neuroimaging studies have shown activation of the mesolimbic dopamine reward system in response to peer-interactions, particularly cooperative and socially shared context as well as emotional (positive and negative) experiences recruit the VS (Schilbach et al., Reference Schilbach, Wilms, Eickhoff, Romanzetti, Tepest, Bente and Vogeley2010; Tabibnia & Lieberman, Reference Tabibnia and Lieberman2007; Wagner et al., Reference Wagner, Galli, Schott, Wold, van der Schalk, Manstead and Walter2015). Likewise, other human brain imaging studies revealed individual differences in dopaminergic neurotransmission in reward-related circuits, which may lead to different responses to environmental cues, such as peer influences. Chein and others (Reference Chein, Albert, O'Brien, Uckert and Steinberg2011) demonstrated that, in the awareness of their peers, adolescents showed an amplified activity in reward-related brain regions, including the VS, which had a subsequent risk-promoting effect (Chein, Albert, O'Brien, Uckert, & Steinberg, Reference Chein, Albert, O'Brien, Uckert and Steinberg2011). The adolescent brain's incentive processing system would thus differentially respond to the potential reward of risky choices, dependent on the presence or absence of peers.

Neuroimaging studies have also reported significant group differences between antisocial adolescents and healthy controls concerning reward-sensitivity (Cohn et al., Reference Cohn, Veltman, Pape, van Lith, Vermeiren, van den Brink and Popma2015; Gatzke-Kopp et al., Reference Gatzke-Kopp, Beauchaine, Shannon, Chipman, Fleming, Crowell and Aylward2009; Matthys, Vanderschuren, Schutter, & Lochman, Reference Matthys, Vanderschuren, Schutter and Lochman2012). Buckholtz and others demonstrated that individuals displaying impulsive-antisocial behavior show heightened dopamine release in the nucleus accumbens during monetary reinforcement, compared to healthy controls (Buckholtz et al., Reference Buckholtz, Treadway, Cowan, Woodward, Benning and Li2010). In addition, a multi-modality functional imaging study found that maladaptive decision-making in incarcerated offenders was associated with a disrupted regulation in the cortico-striatal circuit (Hosking et al., Reference Hosking, Kastman, Dorfman, Samanez-Larkin, Baskin-Sommers, Kiehl and Buckholtz2017). Neuropsychological and physiological studies have shed light on the amygdala's multifaceted role in reward processing, expanding beyond its well-known association with fear conditioning and the processing of negative emotions (Morrison & Salzman, Reference Morrison and Salzman2010). The amygdala has been shown to play a key role in learning about the positive value of stimuli, demonstrating its involvement in reward representation and stimulus-reward learning (Baxter & Murray, Reference Baxter and Murray2002). Lastly, extensive preclinical and human research emphasizes the critical role of the medial prefrontal cortex (mPFC) in incentive processing (Tzschentke, Reference Tzschentke2000; Xue et al., Reference Xue, Lu, Levin, Weller, Li and Bechara2009). Notably, previous fMRI studies have implicated the mPFC as a pivotal region mediating the effects of stress on reward sensitivity (Ironside, Kumar, Kang, & Pizzagalli, Reference Ironside, Kumar, Kang and Pizzagalli2018; Treadway, Buckholtz, & Zald, Reference Treadway, Buckholtz and Zald2013). Furthermore, the mPFC is instrumental in integrating sensory information with learned emotional values and exerting cortical control over reward-based behavioral output (Pastor & Medina, Reference Pastor and Medina2021).

In light of these findings, the present study explored whether the strength of the association between DPA and DBD could be linked to the responsiveness of reward-circuitries. Neuroimaging research has identified the ventral striatum, amygdala, and mPFC as crucial brain areas involved in reward processing (Baxter & Murray, Reference Baxter and Murray2002; Haber, Reference Haber and Gottfried2011; Murray, Reference Murray2007; O'Doherty, Reference O'Doherty2004). To assess sensitivity to reward, neural activity in the VS, amygdala, and mPFC during the fMRI-based monetary incentive delay (MID) task, which involves reward anticipation and outcome response, was employed as an intermediate phenotype. While the MID paradigm serves as a well-established and controlled measure of reward sensitivity, it is crucial to acknowledge that social rewards may entail more intricate and nuanced processes that extend beyond simple monetary incentives. Prior research has demonstrated that tangible and quantitative social rewards hold greater incentive power among children and adolescents compared to monetary rewards (Wang, Liu, & Shi, Reference Wang, Liu and Shi2017). Interestingly, although studies have shown that the neural basis for anticipating social approval is similar to that of anticipating monetary rewards, men exhibit a wider network of mesolimbic brain regions activated during the prospect of monetary rewards, whereas limited activation is observed for social rewards (Spreckelmeyer et al., Reference Spreckelmeyer, Krach, Kohls, Rademacher, Irmak, Konrad and Gründer2009). We hypothesized that increased reward-sensitivity is related to DBD symptoms, particularly in the context of DPA.

Delinquent peer affiliation

DPA was assessed by a self-report questionnaire developed by the RADAR study (Research on Adolescent Development and Relationships; (37)). This questionnaire consists of eleven items, scored on a three-point scale, of which the first half measured the peer delinquency, whereas the second half measured the level of peer affiliation. The first six items were summed to index ‘crime level peers’ and were scored using a three-choice response format (0 = none of my friends are like that, 1 = just a few of my friends are like that, 2 = most of my friends are like that). The last five items were summed in the same way to index peer affiliation. The DPA score (ranging from 0 to 120) was displayed by multiplying crime level peers score with the peer affiliation score. At baseline interview, the internal consistency of the two subscales was adequate (peer delinquency, Cronbachs α = 0.83; peer affiliation, α = 0.74).

MID task

We used the MID paradigm as described by Cohn and others that consisted of 72 six second trials (Knutson, Fong, Adams, Varner, & Hommer, Reference Knutson, Fong, Adams, Varner and Hommer2001b). During the MID-task subjects anticipated on a potential monetary reward, punishment or no consequences depending on the type of cue. Cues signaling the potential of winning €0.50 were denoted by circles, the potential of losing €0.50 were denoted by squares and no monetary outcome was denoted by triangles. Within each trial, participants saw one of three cue shapes (cue, 2000 msec), fixated on a crosshair as they waited during a variable interval (delay, 2000–2500 msec), and then responded to a white target square that appeared for a variable length of time (target 160–260 msec) with a button press. After target presentation, the outcome (outcome, 1650 msec), which followed the disappearance of the target, notified participants whether they had won or lost money during that trial and indicated their cumulative total at that point. The MID paradigm has been tested extensively and successfully recruits the reward regions of the brain, which is also reflected in the expected activation in the limbic and frontostriatal neurocircuitry in the present study (Cohn et al., Reference Cohn, Veltman, Pape, van Lith, Vermeiren, van den Brink and Popma2015; Knutson et al., Reference Knutson, Adams, Fong and Hommer2001a; Knutson et al., Reference Knutson, Fong, Adams, Varner and Hommer2001b).

Our study focused on reward sensitivity by observing the reward trials. The outcome of these reward trials could be either (1) ‘reward outcome’ if the participant pressed the button fast enough during target presentation or (2) ‘no reward outcome’ if the participant was too slow. Trial types were pseudo randomly ordered within each session. Task difficulty, based on reaction times collected during the practice session before scanning, was set such that each participant should succeed on 66% of his or her target responses (Knutson et al., Reference Knutson, Adams, Fong and Hommer2001a).

fMRI acquisition

Functional MRI data were obtained on a second occasion, during which participants were scanned in a Philips 3 T Intera MRI-scanner in the Academic Medical Center Amsterdam. Neuroimaging generally took place within 1.5 months after the behavioral session, with a few exceptions where the scan was performed after 2 months. During this scan protocol, T1 weighted anatomical scans were acquired using a 8-channel SENSE head-coil and consisted of 180 sagittal 1 mm thickness slices, with an in-plane resolution of 1 × 1 mm (FOV 256X256X180 mm, TR 9.0 ms, TE 3.5 ms). Furthermore, 204 T2* weighted echo planar images (EPI) were acquired, each volume consisting of 38 ascending slices of 3 mm thickness and 2.29 × 2.29 in-plane resolution, parallel to the anterior commissure – posterior commissure line (FOV 220 × 220 mm, TR 2300, TE 30 ms).

fMRI analysis

Functional MRI data were processed using SPM8 (Statistical Parametric Mapping, Wellcome Department of Imaging Neuroscience, London, UK) in MATLAB (Mathworks, Inc., Sherborn, MA). Analyses focused on changes in blood oxygen level-dependent (BOLD) contrast that occurred during anticipation and outcome in the monetary reward condition of the MID task. Preprocessing included spatial realignment, unwarping, slice-time correction, spatial normalization into standard Montreal Neurological Institute space based on the segmented anatomical scan and spatially smoothening with an 8 mm full width half-maximum Gaussian kernel.

First and second level analyses

Changes in the BOLD signal for each subject (i.e. ‘activation’) were tested on the basis of linear combinations of estimated GLM parameters (beta values). These changes, displayed by individual contrast images, correspond to the percent signal change (effect size). First level analyses were performed by modeling two separate reward regressors for each trial type: anticipation period and outcome period (40). Realignment parameters were included in the statistical model as additional regressors to correct for motion effects. Then, contrast images were calculated to assess reward anticipation (reward trial anticipation > neutral trial anticipation) and reward outcome (reward trial hit > reward trial miss). Next, to compute group-level statistics, these individual contrast images were entered into second-level analyses. The second-level data were then extracted from MATLAB and used to perform multiple regression analyses in SPSS.

Statistical analyses

Power calculations were conducted using G-Power software (Faul, Erdfelder, Lang, & Buchner, Reference Faul, Erdfelder, Lang and Buchner2007) (Version 3.1.9.7). For the multiple linear regression analysis, we considered small (F ² = 0.02), medium (F ² = 0.15), and large (F ² = 0.35) effect sizes, accounting for 12 predictors and a corrected alpha level of 0.008. The estimated power was 0.03 for small effect sizes, 0.54 for medium effect sizes, and 0.98 for large effect sizes, respectively. Zero-order correlations were calculated to inspect the relationship between the variables. These indicated that the neural activity during the tasks in the VS, amygdala and mPFC was highly correlated across hemispheres (see correlation matrix, online Supplementary Table S1; for main effects of the MID task, see online Supplementary Figure S1). In our analyses, we therefore utilized the average activation of regions of interest (i.e. total VS) to examine the functional brain correlates. Then, to reduce structural multicollinearity between the main effects of the predictors and their interaction terms and to facilitate interpretation of our parameter estimates, we mean-centered all predictors. Second, for all predictors and for the DBD outcome measure, outliers were defined as the scores deviating more than three s.d. from the mean. These outliers (N = 6) were constrained at their lower and upper boundaries (mean + /- three s.d.), so that no extreme score carried excessive weight. We applied a square root transformation of the DBD variable, to obtain a non-significant departure from normality (see online Supplementary Figure SI). We observed only minor differences between the outcomes before and after transformation, therefore, for the purpose of interpretation, we report the standardized regression coefficients and p-values for the untransformed DBD score throughout this study. We then ran a multiple linear regression model to test for the interaction effect of DPA and reward sensitivity on DBD outcome, correcting for the possible confounders sex, SES, age, and the covariate interaction terms. Since reward sensitivity was indexed by the neural response in three brain regions (striatum, amygdala, and mPFC), during two conditions (reward anticipation and outcome), six separate multiple linear regression analyses – each including twelve predictors – were run. In order to account for multiple testing, the threshold for significance of the interaction terms was corrected for 6 independent tests and hence chosen as p < 0.00833. Lastly, in case of nominally significant (p < 0.05) interaction of a covariate and DPA on DBD, we performed separate follow-up analyses in strata defined by the covariate to further explore the relationship between DPA and DBD.