Adolescent neural response to reward is related to participant sex and task motivation

Abstract

Risky decision making is prominent during adolescence, perhaps contributed to by heightened sensation seeking and ongoing maturation of reward and dopamine systems in the brain, which are, in part, modulated by sex hormones. In this study, we examined sex differences in the neural substrates of reward sensitivity during a risky decision-making task and hypothesized that compared with girls, boys would show heightened brain activation in reward-relevant regions, particularly the nucleus accumbens, during reward receipt. Further, we hypothesized that testosterone and estradiol levels would mediate this sex difference. Moreover, we predicted boys would make more risky choices on the task. While boys showed increased nucleus accumbens blood oxygen level-dependent (BOLD) response relative to girls, sex hormones did not mediate this effect. As predicted, boys made a higher percentage of risky decisions during the task. Interestingly, boys also self-reported more motivation to perform well and earn money on the task, while girls self-reported higher state anxiety prior to the scan session. Motivation to earn money partially mediated the effect of sex on nucleus accumbens activity during reward. Previous research shows that increased motivation and salience of reinforcers is linked with more robust striatal BOLD response, therefore psychosocial factors, in addition to sex, may play an important role in reward sensitivity. Elucidating neurobiological mechanisms that support adolescent sex differences in risky decision making has important implications for understanding individual differences that lead to advantageous and adverse behaviors that affect health outcomes.

1. Introduction

Following perinatal neural organization, adolescence marks a second wave of plasticity, during which numerous behavioral, social, and physiological changes occur that act to re-organize and activate the brain (Spear, 2013). This extended brain plasticity can be viewed as a double-edged sword, serving to augment vulnerability to biological and psychological insult, as well as support healthy neurodevelopment (Telzer, 2016). Processing of rewarding stimuli is particularly relevant during the adolescent period, given the rise in sensation seeking, which may contribute to increased reward sensitivity and risk taking in some youth (Romer & Hennessy, 2007). Dysregulated reward processing has been linked with affective and substance use disorders, the incidence of which increase substantially during adolescence (Davey, Yucel, & Allen, 2008; Ernst, Pine, & Hardin, 2006; Fairchild, 2011; MacPherson, Magidson, Reynolds, Kahler, & Lejuez, 2010). As such, elucidating the neural mechanisms underlying adolescent reward sensitivity may help in promoting beneficial, rather than adverse, neuroplastic change.

Psychobiological models of adolescent risk taking posit an imbalance between reward processing and self-control, mirrored by enhanced functional activation of reward-sensitive regions (i.e. striatum, including nucleus accumbens) and diminished activation of self-regulatory brain regions (i.e. medial prefrontal cortex), which drives risk taking via inefficient regulation of reward-sensitive brain regions by self-regulatory regions (Casey, 2015; Ernst, 2014; Smith, Chein, & Steinberg, 2013; Somerville, Jones, & Casey, 2010). However, there is a paucity of data showing a direct relationship between reward sensitivity and risk taking (Braams, Peper, van der Heide, Peters, & Crone, 2016; Braams, van Duijvenvoorde, Peper, & Crone, 2015; Galvan et al., 2006; van Duijvenvoorde et al., 2014, 2015; Vorobyev, Kwon, Moe, Parkkola, & Hamalainen, 2015), likely because there is substantial individual variability in reward sensitivity (Bjork & Pardini, 2015; Braams et al., 2015; Chick, 2015; Cservenka, Herting, Seghete, Hudson, & Nagel, 2012). Some of this variability may be due to individual differences in personality traits, such as sensation seeking (Cservenka et al., 2012; van Duijvenvoorde et al., 2014) and impulsivity (Forbes et al., 2009; Piray, den Ouden, van der Schaaf, Toni, & Cools, 2015). Moreover, the link between reward sensitivity and risk taking may be partly explained by pubertal influences (Forbes et al., 2010; Urosevic, Collins, Muetzel, Lim, & Luciana, 2014), given that puberty has been shown to correlate with sensation seeking (Forbes & Dahl, 2010; Martin et al., 2002, 2006; Steinberg, 2004; Steinberg et al., 2008), reward sensitivity (Urosevic et al., 2014) and nucleus accumbens activity in response to rewards (Braams et al., 2015). Indeed, there is evidence that pubertal increases in sensation seeking predict real-world risky behavior, such as substance use (Kirillova, Vanyukov, Gavaler, Pajer, & Tarter, 2001; Martin et al., 2002).

Gonadal hormones, which are re-activated at the onset of puberty, have also been linked to reward processing. Previous work in adolescents showed a positive association between striatal activity in response to reward and endogenous levels of testosterone (Braams et al., 2015; Op de Macks et al., 2011) and estradiol (Op de Macks et al., 2011) in both males and females. Moreover, sex hormone levels have been positively associated with risk-taking behavior in adolescence (de Water, Braams, Crone, & Peper, 2013; Martin, Mainous, Curry, & Martin, 1999; Peper, Mandl, et al., 2013; Peters, Jolles, Van Duijvenvoorde, Crone, & Peper, 2015; Vermeersch, T’Sjoen, Kaufman, & Vincke, 2008a, 2008b). In studies that compared boys and girls directly, there is more evidence of a positive relationship between sex hormones and risky behavior in boys relative to girls (de Water et al., 2013; Peper, de Reus, van den Heuvel, & Schutter, 2015; Peters et al., 2015), or compared to evidence indicating no sex difference (Peper, Koolschijn, & Crone, 2013). In young adults, sex hormone levels have been shown to predict risky behavior in both sexes to the same degree (Braams et al., 2016; Mehta, Welker, Zilioli, & Carre, 2015; Nguyen et al., 2016; Stanton, Liening, & Schultheiss, 2011). The majority of this research supports a link between testosterone and risk taking (Braams et al., 2016; de Water et al., 2013; Martin et al., 1999; Mehta et al., 2015; Nguyen et al., 2016; Peper et al., 2015; Peper, Koolschijn, et al., 2013; Peters et al., 2015; Stanton et al., 2011; Vermeersch et al., 2008b), while a subset of studies also support a positive association between estradiol and risk taking (de Water et al., 2013; Martin et al., 1999; Peper et al., 2015; Vermeersch et al., 2008a). Only two studies have examined the relationship between reward sensitivity, as indexed by nucleus accumbens activity, sex hormones and risk taking (Braams et al., 2015, 2016). One of these studies reported that puberty, testosterone and risk taking explained nucleus accumbens activation during a gambling game in both males and females (Braams et al., 2015). The second study indicated that testosterone levels, but not nucleus accumbens activation during the same gambling task, predicted risky behavior, as indexed by self-reported alcohol use, two years later in males and females (Braams et al., 2016). The mechanism linking sex hormones, reward sensitivity and risk taking remains to be fully elucidated; however, the extant literature suggests that both testosterone and estradiol may be important in explaining risk-taking behavior during adolescence, particularly in boys.

Intriguingly, sex differences in striatal reactivity during reward processing have not been reported or examined in previous studies of adolescents (Braams et al., 2015, 2016; Forbes et al., 2010; Op de Macks et al., 2011). This is somewhat surprising, given the presence of sex differences in pubertal maturation, sex hormone levels (Tanner & Whitehouse, 1976), prefrontal cortical maturation (on average, girls mature approximately two years earlier than boys) (Lenroot et al., 2007) and sensation seeking (on average, boys report more sensation seeking than girls) (Romer & Hennessy, 2007; Steinberg et al., 2008; Zuckerman & Kuhlman, 2000). Thus, sex may be an important variable to consider for understanding individual differences in reward sensitivity and risk taking during adolescence. Indeed, one of the primary neurotransmitters involved in reward processing - dopamine (Berridge & Kringelbach, 2008) - develops in a sexually dimorphic manner during adolescence. Studies in rodents demonstrate enhanced dopamine release in females compared to males due to elevations in estradiol levels during puberty (Di Paolo, Rouillard, & Bedard, 1985; Sarvari et al., 2014; Xiao & Becker, 1994). In contrast, testosterone metabolites have been shown to mediate reward response following direct administration into the nucleus accumbens, which may be mediated by binding at γ-Aminobutyric acid (GABA) (Frye, Park, Tanaka, Rosellini, & Svare, 2001) and dopamine (Mhillaj et al., 2015) receptors. Additionally, both sex hormones have been shown to influence sensation seeking in adolescence (Kerschbaum, Ruemer, Weisshuhn, & Klimesch, 2006; Vermeersch, T’Sjoen, Kaufman, & Vincke, 2009), indicating a role for sex hormones in dopamine activity and sensation seeking. Thus, examining the influence of sensation seeking and sex hormones on potential sex differences in reward sensitivity may inform psychobiological models of risk taking in adolescence.

The current study adds to this literature by examining sex differences in reward processing in a large sample of healthy adolescents, as well as the potentially mediating influence of sex hormones on observed sex differences. We hypothesized boys would show increased blood oxygen level-dependent (BOLD) response in the striatum, including nucleus accumbens, during reward receipt feedback, as well as heightened risky behavior during a risky decision-making task, compared to girls. These hypotheses were based on research showing higher sensation seeking in adolescent boys (Romer & Hennessy, 2007; Steinberg et al., 2008) and delayed prefrontal gray matter maturation in boys, compared to age-matched girls (Lenroot et al., 2007). We also predicted testosterone and estradiol would mediate sex differences in nucleus accumbens BOLD response, given their important role in pubertal development, sensation seeking and in modulating reward-relevant brain regions (Braams et al., 2015; Di Paolo et al., 1985; Frye et al., 2001; Op de Macks et al., 2011; Sarvari et al., 2014; Xiao & Becker, 1994).

2. Material and methods

2.1. Participant screening and exclusionary criteria

Participants underwent comprehensive structured interviews by trained research assistants to determine eligibility. Youth and parents completed separate structured telephone interviews that included the Diagnostic Interview Schedule for Children Predictive Scales (Lucas et al., 2001), the Family History Assessment Module (Rice et al., 1995), and the Brief Lifetime version of the Customary Drinking and Drug Use Record (Brown et al., 1998). Exclusionary criteria included current diagnosis of DSM-IV disorders (lifetime history of DSM-IV disorders was not assessed), significant substance use (>10 lifetime alcoholic drinks or >2 drinks/occasion, >5 uses of marijuana, any other drug use, or >4 cigarettes per day), neurological illness/head trauma, serious medical problems, prenatal exposure to drugs or alcohol, reported history of psychotic disorders in biological parents, current medication that may affect neural (e.g. psychoactive medication) or endocrine (e.g. birth control) function, the inability of a parent to provide family history information, left-handedness (Edinburgh Handedness Inventory, Oldfield, 1971), pregnancy, and MRI contraindications. This study was reviewed and approved by the Oregon Health & Science University’s (OHSU) Institutional Review Board. Written assent and consent was obtained from all children and their parents, respectively.

Two-hundred one participants from an ongoing longitudinal study of adolescent neurodevelopment completed a reward processing task (see Section 2.4). From this sample, 21 were excluded due to missing sex hormone data or values that exceeded normal ranges based on sex, age, and pubertal status (see Section 2.3). An additional 5 participants were excluded for excessive head motion during scanning (see Section 2.6) and 8 participants were excluded for taking birth control or other endocrine-disrupting medication. The data for the remaining 167 participants were used in subsequent data analyses.

2.2. Participant characteristics and questionnaires

Eligible youth were administered the following: a 2-subtest (Vocabulary and Matrix Reasoning) version of the Wechsler Abbreviated Scale of Intelligence (Wechsler, 1999), the Pubertal Development Scale (PDS) (Petersen, Crockett, Richards, & Boxer, 1988), the Children’s Sleep Habits Questionnaire (CSHQ) (Owens, Spirito, & McGuinn, 2000) and the Impulsive Sensation Seeking scale from the Zuckerman-Kuhlman Personality Questionnaire (Zuckerman, Kuhlman, Joireman, Teta, & Kraft, 1993). In addition, parents completed the Hollingshead Index of Social Position to determine family socioeconomic status (SES) (Hollingshead, 1975). Prior to the scan session, participants filled out the state anxiety sub-scale from the State-Trait Anxiety Inventory (STAI) (Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983). Measures on sleep habits, SES and state anxiety were collected because they have been shown to impact reward processing (Forbes et al., 2012; Gianaros et al., 2011; Hasler et al., 2012; Hasler, Sitnick, Shaw, & Forbes, 2013; Holm et al., 2009; Kumar et al., 2014; Mullin et al., 2013; Telzer, Fuligni, Lieberman, & Galvan, 2013). At the conclusion of the scan, participants completed an Exit Questionnaire assessing their motivation to perform well on the task (i.e. “How important was it for you to do well?” rated on a scale from 1 to 5, or “Not important at all” to “Very important”) and earn money (i.e. “How much did earning money motivate you?” rated on a scale from 1 to 4, or “Not at all” to “Very much”), as well as general feelings about winning (e.g. “On average, how did you feel when you won $7 on this wheel?” rated on a scale from 1 to 10 or “Very Sad” to “Very Happy”) and not winning (e.g. “On average, how did you feel when you did not win $1 on this wheel?” rated on a scale from 1 to 10 or “Very Sad” to “Very Happy”) during low- and high-risk trials (see Section 2.4).

2.3. Sex hormone assays

Serum sex hormone levels were measured within seven days of the scan procedure. Four mL of blood was collected via venipuncture from all subjects at the Oregon Clinical and Translational Research Institute. To reduce diurnal heterogeneity of hormone levels, blood was collected between 7:00 and 10:00 a.m. In addition, samples from post-menarche girls were drawn during the follicular phase of the menstrual cycle (days 1–10) to further minimize variability, as well as interactions with progesterone (Gillies & McArthur, 2010; Wallach, 2000). Menstrual cycle phase was determined by self-report. Testosterone levels were determined by Coat-A-Count radioimmunoassay (Diagnostic Product Corp., Los Angeles, CA). The intra-assay and inter-assay CVs were 7.0% and 7.4%, respectively, with a lower level of detection of 10 ng/dL. Normal levels of testosterone range from <7 to 75 and <7 to 1200 ng/dL, in pubertal girls and boys, respectively (Wallach, 2000). Estradiol levels were determined using the DSL-4800 Ultra-sensitive Estradiol Radioimmunoassay Kit (Beckman Coulter, Fullerton, CA). The intraassay and interassay CVs were 7.4% and 12.6%, respectively, with a lower level of detection of >2.2 pg/mL. Normal levels of estradiol range from <2 to 350 pg/mL and <2 to 40 pg/mL in pubertal girls and boys, respectively (Wallach, 2000). Hormone levels were examined to ensure none exceeded expected levels, as determined by norms for age, sex, and pubertal status.

2.4. Reward processing task

A modified version of the Wheel of Fortune (WOF) Task (Cservenka & Nagel, 2012), adapted from the original WOF paradigm (Ernst et al., 2004), was used to assess neural response during reward processing. Details of the task have been described in depth previously (Cservenka & Nagel, 2012). Briefly, the WOF is a computerized decision-making task in which participants chose between two options associated with distinct probabilities of winning various monetary amounts, represented as portions of a wheel adding up to 100% (10/90, 30/70, or 50/50%). Selection of the low probability/high magnitude option of the wheel (10% chance of winning $7 or 30% chance of winning $2) was considered a risky choice, while selection of the high probability/low magnitude option (90% or 70% chance of winning $1) represented a safe choice. Lastly, selection of a wheel with equal probabilities and magnitudes (50% of winning $2) was considered a chance or neutral choice. Seventy-two trials were presented over two 10-min runs, with each run including 12 10/90, 14 30/70, and 10 50/50 probability wheels. In order to “Win” a trial, a participant’s selection had to match the computer’s choice, based on predefined probabilities, while a choice that did not match, resulted in a “No Win” trial. Participants were instructed to select the portion of the wheel they thought would win them money and to try to win as much money as possible because they would receive “a portion” of their total earnings at the end of the scan session. Each trial was 10.5 s and included a selection (3 s), anticipation (3.5 s) and feedback (4 s) phase, with intertrial fixation intervals jittered between 1 and 11 s. Trial numbers included in the decision making and anticipation phases of the task were determined by participant selections (safe, risky or neutral), while the number of trials for the reward receipt (feedback) phase included wins from all trial types. Only the feedback phase (combining wins across all types of trials) was analyzed in the present study, because it offers the most power for statistical analysis (i.e. number of trials is not limited by participant selections) (Jones, Cservenka, & Nagel, 2016; Steele et al., 2016). During this phase, the screen indicated whether the participant won or did not win money during that trial, as well as the cumulative dollar amount won up to that point. To confirm participant attention during this phase, youth were asked to indicate whether or not they won money during each trial. The task was displayed with E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA). Average accuracy and reaction times (RT) for all phases across different trial probabilities/magnitudes were recorded for both runs of the task.

2.5. MRI data acquisition

Youth were scanned on a 3 Tesla Siemens Magnetom Tim Trio system (Siemens Medical Solutions, Erlangen, Germany) using a twelve-channel head coil at the Advanced Imaging Research Center at OHSU. Functional images were collected in the axial plane oblique to the anterior – posterior commissure, using a high-angular resolution T2*-weighted echo-planar BOLD sequence (TR = 2000 ms, TE = 30 ms, matrix = 240 × 176, FOV = 256 mm, flip angle = 90°, 33 slices, no gap, slice thickness = 3.8 mm, 300 repetitions/run). A whole-brain, high-resolution structural image series was collected in the sagittal plane using a T1-weighted MPRAGE scanning sequence (TI = 900 ms, flip angle = 10°, TE = 3.58 ms, TR = 2300 ms, matrix = 256 × 240, FOV = 240 mm, slice thickness = 1 mm, 33 slices) for co-registration to functional data.

2.6. Image processing

Imaging data were processed and analyzed using Analysis of Functional NeuroImages (AFNI) (Cox, 1996) using the following steps: slice time correction, correction for head movement, spatial smoothing with a 6.0 mm full-width half-maximum Gaussian kernel, within-run intensity normalization to a whole-brain signal, and co-registration of functional images to the anatomical image. TRs exceeding movement of 2.5 mm or 2.5° in any of the three displacement or three rotational parameters, respectively, were censored. To assess within-run motion, an average root mean square (RMS) value was calculated using these six motion parameters and compared across groups. Any subject with a two-run average RMS value exceeding 1.5 mm was excluded from analyses.

Regressors representing selection, anticipation and feedback trials of the task were modeled. Stimulus times corresponding to the onset time of each phase and duration of the event coded as the length of each phase were convolved with a gamma-variate hemodynamic response function (Cohen, 1997). The estimated implicit baseline model corresponded to mean BOLD signal from the entire time course of the task, linear drift, periods of fixation, and nuisance regressors (Cox, 1996). Functional data were then transformed into Talairach space (Talairach & Tournoux, 1988) and resampled into 3 mm³ voxels. Contrast images Win - No Win, No Win - baseline, and Win - baseline were used for fMRI statistical analysis.

2.7. Statistical analyses

2.7.1. Demographics and behavior

Demographic and task performance data were examined for normality and occurrence of outliers using SPSS Statistics 20 (Armonk, NY: IBM Corp.). Non-normal variables with absolute skew/kurtosis values exceeding 2.0 were log transformed. Sex differences for self-reported race were assessed with chi-square analysis, while differences in PDS and responses on the post-scan Exit Questionnaire were determined with a Mann-Whitney U analysis. All other variables were examined with independent samples t-tests.

2.7.2. Task activation-masked fMRI analysis

To best represent task-related activity for both boys and girls, task activation maps were created for each sex. One-sample t-tests of task activation were thresholded at a voxel level of p < 0.001, uncorrected, summed and binarized to create a task-relevant mask (33,903 total voxels). By using this approach, we limit the detection of findings to those pertinent to the task conditions, as previously published (Cservenka et al., 2012; Cservenka, Jones, & Nagel, 2015; Jones et al., 2016). Sex differences were then examined within this reward-related activity mask. Sex differences in Win - No Win activation were examined with an analysis of covariance (ANCOVA) that included age as a covariate because previous studies have demonstrated changes in reward processing across this period of development (Braams et al., 2015; Forbes et al., 2010; Op de Macks et al., 2011). Although pubertal status was significantly different between boys and girls (Table 1), it was not included as a covariate due to its strong collinearity with age. Contributing effects of puberty were examined post hoc in SPSS. AFNI’s AlphaSim (Cox, 1996) was used to correct for between-group multiple comparisons at a voxel- and cluster-level (threshold of p < 0.001 and α < 0.01, respectively; minimum cluster size = 19 voxels).

Table 1.

Demographic and behavioral measures.

	Boys (n = 96)	Girls (n = 71)	Statistic
Age (years)	14.2 (1.3)	14.4 (1.3)	t165 = 0.63
Range	12.3–17.0	12.0–16.9
Pubertya	3.1 (0.9)	4.0 (0.6)	Z = 6.45**
Socioeconomic statusb	29.9 (13.5)	30.5 (14.6)	t165 = 0.28
IQc	110.3 (12.4)	110.8 (10.0)	t165 = 0.31
Log Testosterone (ng/dL)	2.3 (0.5)	1.2 (0.3)	t153.37 = 20.68**
Range	7.1–809.5	2.4–53.3
Log Estradiol (pg/dL)	1.2 (0.2)	1.5 (0.2)	t162.76 = 11.12**
Range	6.1–38.0	11.9–85.0
Sensation Seekingd	42.5 (21.0)	42.8 (22.0)	t165 = 0.09
CHSQ Sleepinesse	5.8 (1.7)	6.2 (1.9)	t165 = 1.24
RMS head movementf	0.31	0.27	t165 = 1.19

^aCrockett Pubertal Development Scale; values range from 1 to 5, with larger values referring to more advanced pubertal development.

^bHollingshead Index of Social Position; larger values indicate lower socioeconomic status (middle class corresponds to 32–47 range); parent-rated.

^cWechsler Abbreviated Scale of Intelligence (2-subtest version).

^dImpulsive Sensation Seeking scale from Zuckerman-Kuhlman Personality Questionnaire.

^eChildren’s Sleep Habits Questionnaire.

^fRoot mean square; index of averaged within-run motion.

^**Statistical significance at p < 0.001.

2.7.3. Region of interest and mediation analyses

To test if sex hormones mediated the relationship between sex and striatal BOLD response, a region of interest (ROI) analysis of bilateral nucleus accumbens was conducted. A mask of the nucleus accumbens was created with the following Talairach coordinates: 12, −8, −8 (right) and −12, −8, −8 (left) corresponding to this region. Surrounding the peak coordinates, two spheres with 4 mm radii where created (20 total voxels; Fig. 3A). Next, an ANCOVA comparing sex differences in reward processing (Win - No Win), while controlling for age, was conducted. AFNI’s AlphaSim was used again to account for multiple comparisons at a voxel- and cluster-level threshold of p < 0.05 and α < 0.05, respectively, which yielded a minimum cluster size of 5 voxels. A more liberal voxel- and cluster-wise threshold was implemented relative to the task activation-masked ANCOVA to account for the small ROI. Values of percent BOLD signal change from significant clusters were included in a nonparametric bootstrapping procedure (5000 re-samples) to assess mediation by sex hormones (Hayes, 2013). The model included sex (independent variable), sex hormone (mediator variable), percent BOLD signal change from significant clusters (dependent variables), and age as a covariate.

Fig. 3.

Boys showed greater BOLD response in nucleus accumbens during reward processing. (A) Bilateral nucleus accumbens mask (peak coordinates: 12, −8, −8 and −12, −8, −8) overlaid on standard Talairach atlas. (B) Statistical map of sex differences in nucleus accumbens region of interest analysis overlaid on a standard Talairach template. Percent BOLD signal change in the Win – No Win contrast is displayed for one significant cluster in right nucleus accumbens (p < 0.05 voxel and α < 0.05 cluster correction) in which boys had significantly more activation than girls. (C) Mean ± SEM percent BOLD signal change in right nucleus accumbens is plotted by trial type.

2.7.4. Sex hormone linear regressions

To examine the relationships between sex hormones and reward-relevant brain activation, multiple regressions with either log testosterone or log estradiol (controlling for age) were conducted separately for boys and girls. Like the previous analysis, multiple regressions were restricted by the task-activation masks created for each sex. Results of the linear regressions were corrected for multiple comparisons with a voxel-wise correction of p < 0.001 and cluster-wise correction of α < 0.01. Because task activation maps differed slightly by sex, the minimum cluster sizes were 15 voxels for girls (of 11,039 total voxels) and 16 voxels for boys (of 32,885 total voxels).

3. Results

3.1. Participant characteristics and task behavior

Details on participant characteristics can be found in Table 1. Boys and girls did not differ in age, SES, IQ, sensation seeking, or general quality of sleep. Two-run averaged RMS head movement values exceeded 1.5 mm for five participants that were subsequently excluded from further analyses; RMS was not statistically different between boys and girls of the remaining sample. Sex hormone levels were not normally distributed and underwent log transformation. Boys had statistically greater serum levels of log testosterone, while girls had statistically higher log estradiol levels. Girls also reported more advanced pubertal maturation. Sensation seeking was not statistically different by sex, nor was it correlated with pubertal status or age across the whole sample. However, examined separately by sex, sensation seeking was correlated to puberty in boys, but not girls, while sensation seeking was not correlated to age in either boys or girls. Self-reported state anxiety prior to the scan session was significantly higher in girls (T-score = 42.7 ± 5.7), compared to boys (T-score = 39.6 ± 7.5; t₁₆₄ = 2.95, p = 0.004). State anxiety was positively correlated with log estradiol across the whole sample, but not within sex. Percent of risky selections was statistically higher in boys (64.5%), compared to girls (56.2%; t₁₆₄ = 2.00, p < 0.05), but was not correlated with sensation seeking in the whole sample or by sex. Log testosterone and estradiol were not significantly related to percent of risky selections across the whole sample or by sex. Correlation tables can be found in Supplementary Material.

Responses on the post-scan Exit Questionnaire revealed that compared to girls, boys believed it was more important to perform well on the task (Z = 2.08, p = 0.04) and were more motivated by money (Z = 2.01, p < 0.05). All participants understood the task, and boys and girls did not feel differently after winning or losing (all Z ≤ 1.62, p ≥ 0.11). Additionally, log estradiol was negatively correlated with ‘Importance of Performance’ and ‘Motivation to Earn Money’ across the whole sample, but not within sex. Log testosterone was not related to either measure (Supplementary Material).

3.2. Sex differences in BOLD response during reward processing

3.2.1. Task activation-masked fMRI analysis

Sex differences in reward processing were assessed with an ANCOVA (controlling for age) masked with a task activation-related mask. Sex differences in percent BOLD signal change during Win - No Win were such that boys showed an increase in response, as compared to girls, in the following brain regions: left lentiform nucleus (extending to thalamus, putamen and caudate nucleus), right lentiform nucleus (extending to putamen, caudate nucleus and nucleus accumbens), right thalamus, left paracentral lobule (extending to cingulate cortex and precuneus), right superior parietal lobule (extending to precuneus and inferior parietal lobule) and right fusiform gyrus (extending to lingual gyrus) (Table 2; Figs. 1 and 2). Effect sizes for these results were in the medium range (Partial η² ~ 0.10) (Table 2).

Table 2.

Sex differences in BOLD response during reward processing.

	Voxels	x	y	z	Partial η2a
Peak Coordinate Brain Region
Boys > Girls
L Lentiform/Caudate nucleus/Thalamus/Putamen	163	−19	−13	−7	0.10
L Paracentral gyrus/Cingulate cortex/Precuneus	124	−1	−28	44	0.10
R Superior/Inferior parietal lobule/Precuneus	87	32	−52	56	0.12
R Lentiform nucleus/Putamen	43	26	−10	5	0.09
R Thalamus	40	5	−7	11	0.09
R Lentiform/Caudate nucleus/Nucleus accumbens	34	14	5	−10	0.11
R Fusiform/Lingual gyrus	33	23	−67	−10	0.10
R Lentiform nucleus/Thalamus	19	23	−16	−7	0.10
Region of Interest Analysis
Boys > Girls
R Nucleus accumbens	9	14	8	−4	0.05

^aPartial η² effect sizes range from small (0.01) to medium (0.09) to large (0.25). R = right; L = left.

Fig. 1.

Boys showed greater striatal BOLD response during reward processing. Statistical maps of sex differences in percent BOLD activation in Win – No Win contrast (controlling for age) overlaid on a standard Talairach template are depicted here. Boys showed more activation (orange) than girls (blue) in bilateral lentiform nucleus (extending to caudate nucleus, putamen, nucleus accumbens and thalamus). Percent BOLD signal change in these regions is also depicted by trial type (No Win and Win) and their contrast (Win – No Win). In all cases, boys showed higher percent BOLD signal change during Win trials, compared to girls. In contrast, there were no sex differences in No Win BOLD activation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.

Boys showed greater cortical BOLD response during reward processing. Statistical maps of sex differences in percent BOLD activation in Win – No Win contrast (controlling for age) overlaid on a standard Talairach template are depicted here. Boys showed more activation (orange) than girls (blue) in paracentral gyrus (extending to cingulate cortex and precuneus), superior/inferior parietal lobule (extending to precuneus) and fusiform gyrus (extending to lingual gyrus). Percent BOLD signal change in these regions is also depicted by trial type (No Win and Win) and their contrast (Win – No Win). In all cases, boys showed higher percent BOLD signal change during Win trials, compared to girls. In contrast, there were no sex differences in No Win BOLD activation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Percent BOLD signal change values from significant clusters of group differences were plotted for Win - baseline and No Win -baseline contrasts to examine whether group differences were driven by changes in BOLD response during Win trials, No Win trials, or both. Boys and girls showed comparable BOLD response during No Win trials; however, they differed significantly in their BOLD response during Win trials. In all clusters, boys showed greater BOLD signal during Win trials, relative to girls (Figs. 1 and 2).

To confirm that sex differences in reward processing were not attributed to differences in task performance, percent of risky selections, which was significantly higher in boys compared to girls (Table 1), was examined post hoc. Addition of a percent risky selections covariate did not change the significant effects of sex on percent BOLD signal change at any cluster (all F(1,162) ≥ 12.99, p < 0.001). However, percent of risky selections also explained some variance of BOLD activation during Win – No Win in the following clusters: left lentiform nucleus/caudate nucleus/thalamus/putamen (F(1,162) = 7.61, p = 0.006, partial η² = 0.05), right lentiform nucleus/putamen (F(1,162) = 4.76, p = 0.03, partial η² = 0.03), right lentiform nucleus/caudate nucleus/nucleus accumbens (F(1,162) = 6.31, p = 0.01, partial η² = 0.04), right fusiform/lingual gyrus (F(1,162) = 7.04, p = 0.009, partial η² = 0.04) and right lentiform nucleus/thalamus (F(1,162) = 11.09, p = 0.001, partial η² = 0.06). Percent of risky selections did not explain variance in Win – No Win BOLD response in left paracentral gyrus/cingulate cortex/precuneus (F(1,162) = 2.77, p = 0.10), right superior/inferior parietal lobule/precuneus (F(1,162) = 2.70, p = 0.10) or right thalamus (F(1,162) = 3.71, p = 0.06).

Further, given the relevance of pubertal status in the neuromaturation of reward processing (Braams et al., 2015; Forbes et al., 2010), its effect on reward response was also analyzed post hoc. With the addition of a PDS covariate, the results of the ANCOVA replicated; pubertal status did not change the significant effects of sex on percent BOLD signal change at any cluster (all F (1,163) ≥ 11.43, p ≤ 0.001). However, PDS explained additional variance of Win – No Win BOLD response in right fusiform/lingual gyrus (F(1,163) = 3.92, p < 0.05, partial η² = 0.02), such that PDS was associated with an increase in fusiform gyrus activation independent of age and sex. PDS did not explain a significant amount of variance in the remaining clusters (all F(1,163) ≤ 2.88, p ≥ 0.09).

3.3. ROI and mediation analyses

Sex differences in reward processing were assessed with an ANCOVA (controlling for age) masked with a nucleus accumbens mask (Fig. 3A). Boys showed more BOLD activation in Win - No Win contrasts compared to girls in right, but not left nucleus accumbens (Table 2, Fig. 3B and C). Log testosterone was positively correlated with right nucleus accumbens percent BOLD signal change, controlling for age, across the whole sample (partial ² = 0.22, p = 0.005; when girls and boys where examined separately, this correlation did not stand (all partial r² ≤ −0.10, p ≥ 0.41)); therefore, a nonparametric mediation analysis was pursued (Hayes, 2013). Log testosterone did not statistically mediate the effect of sex on nucleus accumbens percent BOLD signal change (Bootstrapped CI₉₅: −0.0834, 0.0859). A mediation analysis with estradiol was not pursued because estradiol did not relate to nucleus accumbens BOLD response, controlling for age (partial r² = −0.10, p = 0.22).

Since we observed sex differences in ‘Importance of Performance’, ‘Motivation to Earn Money’ and state anxiety - variables known to impact striatal response to rewards (Kumar et al., 2014; Lighthall et al., 2012; Zink, Pagnoni, Martin-Skurski, Chappelow, & Berns, 2004) - we examined correlations between right nucleus accumbens BOLD response and these variables. Neither ‘Importance of Performance’ (ρ = 0.10, p = 0.21) nor state anxiety (r² = −0.03, p = 0.72) were related to nucleus accumbens BOLD response. ‘Motivation to Earn Money’ was significantly related to nucleus accumbens BOLD activity (ρ = 0.17, p = 0.04), thus, a mediation analysis was pursued. ‘Motivation to Earn Money’ partially mediated the effect of sex on nucleus accumbens activity (Bootstrapped CI₉₅: 0.0005, 0.0303); the direct effect of sex on nucleus accumbens activity remained statistically significant (CI₉₅ = 0.0301, 0.1556).

3.4. Task-activation masked sex hormone linear regressions

To determine any associations between sex hormones and BOLD response during reward processing, multiple regressions (controlling for age) with testosterone or estradiol as regressors were conducted in girls and boys separately. These analyses did not yield significant effects of testosterone or estradiol in either boys or girls.

4. Discussion

In this study, sex differences in brain response to reward were observed in a large and carefully matched sample of adolescent boys and girls. Specifically, BOLD response following notification of receipt of monetary rewards was higher in males, relative to females, in several reward-relevant brain regions (Liu, Hairston, Schrier, & Fan, 2011; Mohr, Biele, & Heekeren, 2010), including the nucleus accumbens. Although testosterone was positively related with nucleus accumbens BOLD response, it did not mediate the relationship between sex and nucleus accumbens activation during reward processing. Estradiol was not related to nucleus accumbens response. Notably, although self-reported sensation seeking was not different by sex, boys made a higher percentage of risky selections on the task. In addition, boys reported higher motivation to perform well and earn money on the task, while girls reported more state anxiety, both of which have been shown to impact reward processing (Kumar et al., 2014; Lighthall et al., 2012; Zink et al., 2004). Notably, motivation to earn money on the task partially mediated the effect of sex on nucleus accumbens BOLD signal during reward processing, indicating that both sex and task motivation play an important role in determining striatal reactivity to rewards.

4.1. Sex differences mechanisms: sensation seeking

While our primary hypothesis was supported (i.e. increased nucleus accumbens BOLD response to reward in males), the mechanisms underlying this difference were not as predicted. First, it was hypothesized that males would show increased activation of nucleus accumbens during receipt of rewards, due in part to previous evidence showing heightened sensation seeking in age-matched males versus females (Romer & Hennessy, 2007; Steinberg et al., 2008). Interestingly, no sex differences in sensation seeking were observed in this study; however, sensation seeking and percent of risky selections were positively correlated only in males, who made a statistically greater number of risky selections compared to girls. The higher rates of risky selections in males may still be partly due to sensation seeking, but not impulsive sensation seeking per se. The Impulsive Sensation Seeking scale (Zuckerman et al., 1993) employed in the present study measures a specific type of sensation seeking that hinges on impulsivity; however, not all sensation seeking is done impulsively; a different assessment tool may have detected sex differences in sensation seeking that may better explain group differences in BOLD response during reward.

4.2. Sex differences mechanisms: testosterone

Second, a link between testosterone and ventral striatal brain response during reward processing has been shown in early adolescent samples (Braams et al., 2015; Op de Macks et al., 2011) and animal models (Frye et al., 2001); therefore, testosterone was a relevant mechanistic target explaining group differences in reward processing. Although testosterone was related to nucleus accumbens BOLD response in the entire sample, replicating previous work (Braams et al., 2015; Op de Macks et al., 2011), it did not explain sex differences in reward processing BOLD activation. Moreover, testosterone did not relate to reward processing in either sex when regressed directly onto reward feedback BOLD signal, indicating that testosterone alone may not predict a neural response to reward, but might be correlated to this process insofar as it relates to sex and the biological processes that differentiate males and females during adolescence. However, it is possible that the reduction in sample size when conducting these analyses separately in males and females may have reduced our power to detect an effect. Future work should address these questions with much larger samples in order to account for the broad range of individual variability in sex hormones levels during adolescence.

4.3. Sex differences mechanisms: estradiol

Estradiol is another neurophysiological mechanism that may contribute to sex differences in reward-related BOLD response, as it has been shown to modulate dopaminergic systems (Di Paolo et al., 1985; Sarvari et al., 2014; Xiao & Becker, 1994) and relate to impulsivity (Smith, Sierra, Oppler, & Boettiger, 2014) and risk taking (de Water et al., 2013). Indeed, estradiol levels in adolescent girls have been associated with dorsal striatal BOLD response during reward processing (Op de Macks et al., 2011), and brain response during reward anticipation and receipt has been shown to fluctuate across the menstrual cycle, at least in adults (Bayer, Bandurski, & Sommer, 2013; Dreher et al., 2007; Ossewaarde et al., 2011). An effect of estradiol was explored in the present study by relating estradiol levels to BOLD response in reward-relevant brain regions that differentiated boys and girls, including the nucleus accumbens. However, estradiol was not significantly related to reward-related brain response. Moreover, regression of estradiol with reward-related BOLD response did not yield any significant effects in either boys or girls. It is possible that our hormone data collection approach obfuscated the relationship between estradiol and BOLD activation to reward. In the case of girls, the sampling of sex hormones levels was truncated to the first 10 days of the follicular phase; thus, establishing a physiological context through which individual differences in estradiol and testosterone could be interpreted. Given that sex hormones, particularly estradiol, fluctuate across the menstrual cycle, sampling these hormones during the same phase provides a reference point and added meaning to the values. However, one limitation of sampling from only one phase of the menstrual cycle, particularly the early follicular phase when estradiol levels are at their lowest, is that our hormone values have restricted variance. Reductions in estradiol, like those observed in premenstrual and early follicular phases of the menstrual cycle, acutely down-regulate endogenous dopamine activity (Di Paolo et al., 1985; Thompson & Moss, 1994), which in turn, diminishes reward signaling from the ventral striatum (Tzschentke & Schmidt, 2000). Therefore, the relatively low levels of estradiol, reflective of the follicular phase, observed in our female sample may have precipitated reductions in striatal BOLD response via blunted dopamine signaling during reward outcome. This interpretation may explain the weaker nucleus accumbens BOLD response observed in girls, relative to boys, during reward processing. Research comparing hormone states across the menstrual cycle is needed to confirm an independent role for estradiol in reward-related brain response during adolescence.

4.4. Sex differences mechanisms: motivation

Psychosocial factors can also interact with physiological and reward processes. Previous research has shown that degree of striatal BOLD response depends, in part, on the saliency of a reinforcer (Zink et al., 2004). Based on post-scanning questionnaire responses, in the current study, boys self-reported higher motivation to obtain monetary rewards and perform well on the task, which may have made the task more salient for males than females. This is supported by the finding that males made more risky selections during the WOF task. Further, percent of risky selections explained a significant amount of variance (in a model controlling for sex and age) in the activation of several reward-relevant brain regions that differentiated boys and girls, particularly in ventral and dorsal striatum, which may indicate that sex differences in activation of these regions underlies sex differences in percent of risky selections made during the task. Interestingly, ‘Motivation to Earn Money’ partially mediated the effect of sex on nucleus accumbens activation, suggesting that both sex and motivation are critical components of salience and sensitivity to rewards. Importantly, girls reported more state anxiety prior to their scan sessions, which can blunt striatal BOLD response (Kumar et al., 2014; Lighthall et al., 2012); thus, we cannot rule out the possibility that anxiety during the scan, which was not measured, may have impacted BOLD activity. The combination of heightened motivation in males and enhanced state anxiety in females may contribute to the sex differences in nucleus accumbens BOLD response observed in this adolescent sample; however, this remains to be tested. In fact, individual differences in such variables may explain why previous studies have not found sex differences in reward sensitivity of the striatum (Braams et al., 2015; Forbes et al., 2010; Op de Macks et al., 2011); motivation, and not sex per se, may be a better predictor of nucleus accumbens reward activation. Future studies must parametrically modulate motivation and examine reward-relevant brain regions outside the ventral striatum. Regardless, we can conclude that task motivation is an important factor when assessing adolescent sex differences in reward sensitivity, as a function of striatal BOLD response.

4.5. Limitations

The primary limitation of this study was the method for analyzing sex hormone levels. Blood samples were collected the week of MRI scanning, thus providing a measure of trait, rather than state sex hormone levels. In the case of the females, this trait was specific to the early follicular phase of the menstrual cycle. However, a single measurement does not reliably describe a trait that can vary as much as sex hormone levels. In an attempt to account for diurnal heterogeneity, blood samples were collected before 10 a.m. for all participants; however, we cannot ascertain the reliability of this measure across days. Measurement of sex hormone levels immediately prior to scanning or administration of exogenous sex hormones would provide some confidence that acute mechanisms of action could be at play. Consideration of other variables that influence sex hormone levels would also be beneficial. Indeed, both genetic and environmental factors influence testosterone levels in adolescence (Harden, Kretsch, Tackett, & Tucker-Drob, 2014). For instance, maternal social instability during pregnancy or lactation leads to a delay in increase of testosterone in male adolescents (Siegeler et al., 2013), while anticipation of a challenge or competition, as well as winning lead to more acute increases in testosterone concentration in young men (Booth, Shelley, Mazur, Tharp, & Kittok, 1989; van der Meij, Buunk, Almela, & Salvador, 2010; Zilioli & Watson, 2014). Understanding variability in the factors linked to reward sensitivity and risk taking during adolescence will offer the best opportunity to uncover the mechanism(s) underlying problem risky behavior.

5. Conclusions

In sum, sex differences in BOLD activity during reward processing were observed in a large sample of healthy adolescents. Regions related to reward processing, including nucleus accumbens (Liu et al., 2011; Mohr et al., 2010), were recruited more robustly in males during reward trials. Sex hormones did not mediate the effect of sex on nucleus accumbens activation even though testosterone was positively correlated with activation of this region, indicating that sex may be a stronger predictor of reward sensitivity than sex hormones. Task motivation partially mediated the effect of sex on nucleus accumbens BOLD response, suggesting that motivation may serve to explain sex differences in reward sensitivity. Importantly, our findings of sex differences in reward processing BOLD activity had medium effect sizes, which emphasize the notion that neurobiological sex differences are nuanced. However, studying subtle differences in male and female brains, particularly during adolescence, can help elucidate healthy developmental trajectories and individual differences in psychosocial and neurophysiological factors that affect relevant processes, such as risk taking and reward sensitivity, which can promote both beneficial and adverse neuroplastic events with long-term consequences.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.bandc.2016.10.003.