Ventral Striatum Activity in Response to Reward: Differences Between Bipolar I and II Disorders

Euthymic bipolar I and bipolar II patients were recruited from a preexisting database of well-characterized patients participating in ongoing molecular genetic studies at Cardiff University and through local community mental health teams. Comparison subjects were recruited from the community using advertisements.

A qualified clinical psychologist (X.C.) administered the Mini-International Neuropsychiatric Interview (20) to confirm diagnosis and exclude participants with a recent history (<1 year) of substance abuse or dependence or borderline personality disorder. To quantify the potential presence of residual symptoms of depression and hypomania or mania, the Hamilton Depression Rating Scale (HAM-D) (21) and the Young Mania Rating Scale (22) were also administered. Euthymia was defined as the absence of episodes of depression, mania, or hypomania for 2 months before scanning, based on clinical interview, plus scores <10 on the HAM-D and the Young Mania Rating Scale.

Healthy comparison subjects were also interviewed using the Mini-International Neuropsychiatric Interview, the HAM-D, and the Young Mania Rating Scale. Healthy participants were excluded for the same criteria as patients as well as for any personal or family history of affective or psychotic disorders or current psychiatric disorder.

All participants completed a questionnaire (the BIS/BAS scales) assessing the strength of the behavioral inhibition and behavioral activation systems (23), which regulate aversive motives and inhibition of action in response to negative stimulation (inhibition system), and appetitive motives and preparation to move toward desirable stimuli (activation system). The behavioral activation system is divided into three scales—drive, fun seeking, and reward responsiveness—that account for different aspects of incentive sensitivity. Patients also completed the Hypomanic Checklist–32 (24), a screening tool for bipolar disorder that allows for a detailed description of hypomanic and manic episodes. Verbal IQ was assessed in all participants using the National Adult Reading Test (25).

Two bipolar I patients were excluded because of concerns that they were no longer euthymic on the day of the scan. The final sample included 20 healthy comparison subjects and 32 bipolar patients (17 with bipolar I disorder and 15 with bipolar II disorder). One healthy comparison subject and one bipolar I patient were excluded from the reward paradigm analyses because they had more than three missing trials, suggesting poor engagement or understanding of the task.

All participants gave written informed consent and were paid £20 for participating. The study was approved by the local National Health System research ethics committee.

We employed a monetary reward-processing task that has been shown to activate the ventral striatum (16, 26). The task was back-projected so that participants could see it through a mirror mounted on the head coil. Participants guessed the numeric value of cards in order to win or avoid losing money. Each trial included an anticipation phase, during which participants waited for the outcome of their guess, followed by the actual outcome (see the supplementary material in the data supplement that accompanies the online edition of this article). Participants practiced the task before scanning.

Participants performed the task during an optimized functional MRI (fMRI) data acquisition protocol. An anatomical brain scan was also obtained using a 3-T General Electric HDx MRI scanner. fMRI data processing was performed using FMRIB’s Software Library (www.fmrib.ox.ac.uk/fsl), and preprocessing the functional data followed standard methods (see the online data supplement). The task was modeled within the general linear modeling framework, with crosshair periods as baseline stimuli. The anticipation periods were divided into independent events representing the initial 2 seconds and the remaining anticipation time. Since we assumed the effect of reward anticipation to be more prominent at the beginning of the anticipation, because of potential habituation effects, all the analyses reported here refer to the initial 2 seconds of the anticipation phase.

An a posteriori analysis with time as the within-subject factor (the initial 2 seconds compared with the rest of the anticipation period) and group as the between-subjects factor revealed a significant time-by-group interaction on activity in the ventral striatum (F=4.54, df=2, 47, p<0.05), reflecting a reduction in the group differences as the anticipation period progressed. This result supported our a priori strategy of focusing our analyses on the initial 2 seconds of the anticipation period. We were also interested in all positive outcomes; therefore, our outcome variable included trials in which participants won after anticipating reward and trials in which participants avoided losing after anticipating punishment. Thus, our contrasts of interest were anticipation of reward relative to intertrial crosshair baseline and positive outcome relative to intertrial crosshair baseline. The full model also included events for anticipation of punishment and negative outcomes. The resulting functional images were converted to the standard space of the Montreal Neurological Institute (MNI) using FMRIB’s linear registration tool (FLIRT). Higher-level analysis was carried out using FMRIB’s local analysis of mixed effects (FLAME) (27).

Since our hypotheses were strongly focused on the ventral striatum, region-of-interest analyses were conducted within a mask of two 8-mm radius spheres (Figure 1) based on MNI coordinates (right: x=9, y=9, z=−8; left: x=−9, y=9, z=−8) from previous meta-analyses (28, 29). 3dClustSim within AFNI was used on the normalized images (voxel size, 2 mm isotropic) to determine that a corrected p<0.05 within this region of interest required a voxel-wise threshold of p<0.05 coupled with a cluster size criterion of 41 voxels. Subsequently, we conducted exploratory whole-brain analyses using a voxel-wise Z statistic >2 and a cluster-corrected threshold of p<0.05. In both region-of-interest and whole-brain analyses, we conducted a voxel-wise analysis of variance (ANOVA) to examine the main group factor, and we performed post hoc pairwise between-group comparisons as appropriate. Although not directly related to the objectives of this study, results of the three (group) by two (condition: reward anticipation relative to punishment anticipation) ANOVA and the analysis of the effect of group on punishment anticipation are presented in the online data supplement.

To test our second hypothesis, we examined relationships between ventral striatal activity (highest peak percentage blood-oxygen-level-dependent [BOLD] signal change within our predefined region of interest) during reward anticipation and outcome and behavioral activation system scores using correlation analyses. The relationships between ventral striatal activity and residual manic or depressive symptoms (HAM-D, Young Mania Rating Scale, and Hypomanic Checklist–32 scores) were also investigated. To avoid floor effects produced by the high frequency of zero scores on clinical measures in healthy comparison subjects, correlation analyses with these scales were performed only for bipolar patients. Given the nonnormal distribution of ventral striatal activity, nonparametric Spearman’s rho analyses were conducted.

Structural MRI data were processed with the fully automated segmentation procedure in Freesurfer (http://surfer.nmr.mgh.harvard.edu/; see the online data supplement). We focused on examining ventral striatal volume, including the putamen and nucleus accumbens, which have been shown to be contiguous ventral striatal regions in humans (30). Following subcortical segmentation, we compared the volumes of these regions between groups using ANOVA and pairwise comparisons, and we performed correlation analyses to examine relationships with peak ventral striatal activity (as for the above correlation).

Demographic and clinical characteristics were compared across groups using ANOVA followed by pairwise comparisons or chi-square tests as appropriate.

Post hoc analysis of the reaction time to the guessing cues that preceded the reward anticipation periods revealed no differences between groups, suggesting similar levels of engagement.

Table 1 summarizes participants’ demographic and clinical characteristics. Groups were matched for gender, age, and estimated verbal IQ (p>0.1 in all cases). Healthy comparison subjects had lower scores than both bipolar patient groups on residual mood symptoms measured by the HAM-D (F=6.18, df=2, 49, p<0.005) and the Young Mania Rating Scale (F=29.37, df=2, 49, p<0.005); the two bipolar groups had similar scores, although bipolar I patients tended to have higher Young Mania Rating Scale scores (p=0.07). The bipolar patient groups did not differ on the number of manic symptoms experienced during “highs” (Hypomanic Checklist–32), age at first mood episode, age at first bipolar diagnosis, or time from first mood episode until diagnosis. Slightly under half of the bipolar patients had identified a family history of bipolarity in first-degree relatives, the distribution of which did not differ between bipolar I and bipolar II disorder. However, among those not reporting a family history of bipolarity, 64% reported the presence of depression in at least one first-degree relative. No group differences were observed on any of the behavioral activation system scales, but both bipolar patient groups had greater behavioral inhibition system scores than healthy comparison subjects (F=11.61, df=2, 47, p<0.001) (data from two participants were missing for the BIS/BAS questionnaire).

Most bipolar patients were taking psychotropic medication; only one bipolar I patient and five bipolar II patients had been medication free for more than 2 months at the time of scanning. Bipolar I patients tended to be taking more antipsychotic medications than bipolar II patients, but the difference fell short of significance (χ²=4.29, p=0.07). The proportion of bipolar I and bipolar II patients with DSM-IV comorbid anxiety diagnoses did not differ.

Region-of-interest analysis for reward anticipation revealed significant ventral striatal activity for the entire sample. A main effect of group was observed (F=3.61, df=2, 47, p<0.05), with greater bilateral ventral striatal activity in bipolar II patients than in healthy comparison subjects (p<0.05, 169 voxels) and bipolar I patients (p<0.05, 131 voxels). No differences were observed between bipolar I patients and healthy comparison subjects (Figure 1).

Region-of-interest analysis for positive outcome revealed a cluster of activity that approached but did not meet significance in the entire sample (p<0.05, 25 voxels). However, a main effect of group was observed (F=3.35, df=2, 47, p<0.05), with greater right ventral striatal activity in bipolar I patients than in bipolar II patients (p<0.05, 51 voxels). Bipolar I patients also exhibited significantly greater left ventral striatal activity than healthy comparison subjects, but this difference did not survive cluster-wise correction (p<0.05, 15 voxels).

For the entire sample, significant activity during reward anticipation was observed in the left superior temporal cortex, insula, putamen and claustrum, ventrolateral prefrontal cortex, and precentral gyrus (Figure 1). No significant overall group effect was observed in these regions. Planned pairwise comparisons did, however, reveal significantly greater activity in bipolar II patients relative to healthy comparison subjects and bipolar I patients in the left ventrolateral prefrontal cortex, insula, precentral gyrus, and middle and superior temporal cortex. Bipolar II patients also exhibited greater activity than healthy comparison subjects in the caudate nuclei bilaterally and the left dorsolateral prefrontal cortex (Figure 1 and Table 2). No significant differences were observed between bipolar I patients and healthy comparison subjects.

Whole-brain analysis of activity to positive outcome did not yield any significant cluster of activity or differences between groups.

Supporting our second a priori hypothesis, left and right ventral striatal activity during reward anticipation was positively correlated with scores on the behavioral activation system fun-seeking scale across all individuals (r_s=0.312, p<0.05; r_s=0.385, p<0.01, respectively) (Figure 2). No significant correlations were observed between ventral striatal activity and scores on the HAM-D, the Young Mania Rating Scale, and the Hypomanic Checklist–32 scores.

No significant relationships were observed between ventral striatal activity to positive outcome and any behavioral or clinical scores.

Examination of striatal brain volumes indicated significant differences in the left putamen volume between groups (F=3.24, df=2, 48, p<0.05), with bipolar II patients showing enlargement relative to bipolar I patients (mean=5,834 mm³ [SD=648] and mean=5,278 mm³ [SD=701], respectively; p<0.05), as well as relative to healthy comparison subjects, although the latter difference fell short of significance (mean=5,420 mm³ [SD=512]; p=0.064). These analyses survived after controlling for intracranial volume. No differences were observed in the right putamen or accumbens volumes. Across the whole sample, left putamen volume was positively correlated with left ventral striatal activity during reward anticipation (r_s=0.33, p<0.05) (Figure 2).

As there was a trend toward more bipolar I than bipolar II patients taking antipsychotics, we repeated the above comparisons after excluding those patients taking this class of medication (bipolar I patients, N=8; bipolar II patients, N=12). Again, bipolar II patients had greater ventral striatal activity during reward anticipation than bipolar I patients (p<0.05, 53 voxels) and healthy comparison subjects (p<0.05, 185 voxels). The bipolar I > bipolar II difference in ventral striatal activity during positive outcome no longer reached significance. Similarly, the differences in whole-brain activity between bipolar groups during reward anticipation were not significant here, although bipolar II patients still showed greater activity than healthy comparison subjects in the left ventrolateral prefrontal cortex, insula, caudate, putamen, superior temporal cortex, and precentral gyrus.

Following the approach used in other studies (17, 31, 32), we converted antipsychotic medication into chlorpromazine equivalents (coded 0 for those not taking antipsychotics) and used this index as a covariate. Region-of-interest analyses revealed significantly greater ventral striatal activity in bipolar II patients during reward anticipation than in healthy comparison subjects (p<0.05, 172 voxels) and bipolar I patients (p<0.05, 60 voxels). During positive outcome, the bipolar I > bipolar II difference in ventral striatal activity did not reach significance (p<0.05, 33 voxels). As before, whole-brain analysis failed to replicate significant differences between bipolar I and bipolar II patients, but still revealed significantly greater activity in the left superior temporal cortex, left insula, left caudate, and left ventrolateral prefrontal cortex in bipolar II patients relative to healthy comparison subjects.

Covarying for antipsychotic medication, a nonsignificant main effect of group on left putamen volume was still observed (F=2.57, df=2, 48, p=0.08).