Involvement of the dorsal striatum in the processing of happy faces in Borderline Personality Disorder

Background Emotion dysfunction and anhedonia are main problems in borderline personality disorder (BPD). In the present functional MRI (fMRI) study, we investigated neural activation in the processing of happy faces and its correlates with habitual emotion acceptance in patients with BPD. Methods 22 women with BPD and 26 female healthy controls watched movie clips of happy and neutral faces during fMRI without any instruction of emotion regulation. To associate neural activation with habitual emotion acceptance, we included individual scores of the Emotion Acceptance Questionnaire (EAQ) as a covariate in brain data analysis. Results All participants showed amygdala, temporal and occipital activation in the processing of happy compared to neutral faces. Additionally, healthy controls showed activation of the hippocampus and patients with BPD showed dorso-striatal (caudate, putamen) activation in the happy compared to the neutral condition. Compared with healthy controls, patients with BPD showed significantly more caudate activation. We could not find significant correlations with the EAQ. Our results indicate caudate hyperactivation in patients with BPD during the processing of happy faces. Although patients reported significantly less emotion acceptance of positive emotions, an association with neural activation was not detectable.

Patients undergo intense negative emotions and report less positive affective states (3,4). The lack of happiness in everyday life is problematic because positive emotions enhance resilience (5), psychosocial well-being (6) and physical health (7). According to Fredrickson and Joiner (8), frequent positive emotions broaden coping mechanisms (e.g. positive cognition) which again promote positive affect in a self-sustaining circulatory mechanism. Moreover, patients with BPD suffer from general emotional instability (1). This is problematic, because a stable experience of positive emotions in everyday life is important for overall psychological health (9).
For several authors, impaired emotion regulation is the primary emotional dysfunction in BPD patients and the main cause of further emotional symptoms in BPD (10)(11)(12). Emotion regulation is defined as a set of actions aimed at influencing which emotions we have and how we experience and express those emotions (13). Patients with BPD try to suppress affective arousal, not only in negative, but also in positive affective states (14). In a conceptual context, emotion suppression can be understood as a variation of emotion non-acceptance, opposing emotion acceptance (15,16). It is known that attempts to suppress affective reactions lead to increased negative and decreased positive emotions in the long term (17). Emotion non-acceptance in form of suppression essentially contributes to the persistence of BPD symptoms (18) and diminishes positive experiences (19). While it is obvious that the suppression of negative emotions might be desirable, it is a matter of debate why patients hardly accept positive emotions. It has been suggested, that the attempt to avoid physiological reactions generalizes to positive emotions and that, as a result, positive emotions are perceived as being threatening (20).
FMRI research indicates that several brain areas are crucial for the processing of positive emotions: ventral and dorsal striatum (e.g. putamen, caudate), limbic areas (e.g. amygdala, hippocampus), thalamus, anterior cingulate cortex as well as prefrontal areas and fusiform gyrus (21)(22)(23)(24)(25)(26). While in healthy people the processing of positive feelings was extensively investigated, studies focusing on neural activity of positive emotions in patients with BPD are rare. Koenigsberg et al. (27) found that patients with BPD showed increased activation in the temporal and inferior frontal gyrus as well as in the cerebellum, the premotor cortex and the posterior cingulate cortex compared to healthy controls when viewing positive pictures compared to a rest condition. A study by Hazlett et al. (28) found greater activation of the amygdala in patients with BPD compared to healthy controls in new and familiar emotional pictures with pleasant content compared to neutral pictures.
To our knowledge, the effects of emotion acceptance on neural activity during the processing of positive emotions in patients with BPD have not yet been investigated. A study on negative emotions showed that patients with BPD demonstrated a reduction of amygdala hyperactivation while watching emotional compared to non-emotional pictures 12 months after DBT treatment that included improvement of emotion acceptance among other emotion regulation skills (29). Moreover, our study group found that the dispositional (habitual) non-acceptance of negative emotions was correlated with more activity in caudate and putamen in patients with BPD while watching fearful vs. neutral movie scenes of human faces (30). In the latter studies, no active or instructed emotion regulation was required.
Considering the importance of positive emotions and emotion acceptance for mental health, one aim of our study was to further investigate the neural correlates of positive affect in patients with BPD. We hypothesized altered activation of striato-limbic (e.g. amygdala, caudate, hippocampus) areas as well as prefrontal and temporal brain areas and the anterior cingulate cortex in patients with BPD compared to healthy controls after experiencing emotionally positive stimuli. Accounting for the lack of studies considering habitual emotion acceptance, another aim of the study was the exploration of neural activation being associated with subjective dispositional emotion acceptance.

Methods And Materials Participants
In the current study, we included 22 female subjects with BPD and 26 female healthy control subjects (table 1). All participants were right-handed with normal or corrected-to-normal vision. Patients were recruited at the Department of Psychiatry and Psychotherapy Evangelisches Klinikum Bethel, Germany. Healthy participants were recruited via local advertisements. Exclusion criteria for all participants were current or previous severe medical conditions (e.g. cancer) or neurological disorders (e.g. epilepsy) and incompatibilities with MRI-investigations (e.g. pregnancy). Control subjects with any lifetime psychiatric treatment were excluded. Moreover, exclusion criteria comprised changes of psychotropic medication within the last 14 days, the intake of benzodiazepines, severe actual comorbid diagnoses (e.g. affective disorders, substance dependence or anorexia nervosa) that were not remitted for at least 6 months and DBT treatment within the last 6 months. Patients with schizophrenia were entirely excluded. In sum, 49 Patients with BPD were informed about the study but 27 had to be excluded from participation due to aforementioned conditions. Fourteen patients of our study were treated with psychotropic medication (antidepressant: n = 13, neuroleptics: n = 6, antiepileptics: n = 1). Eleven patients with BPD hat no comorbid diagnosis. Eight patients had one and three patients had two additional diagnoses. Comorbid diagnoses were posttraumatic stress disorder (n = 8), social phobia (n = 3), specific phobia (n = 1), agoraphobia (n = 1) and obsessive-compulsive disorder (n = 1).
Diagnoses were made by experienced and trained clinical psychologists on the basis of the German version of the Structural Clinical Interview for DSM-IV (SCID-I) and the BPS-section of the SCID-II (SCID-II; 31). In average, patients with BPD met 6.7 out of 9 BPD criteria (SD = 1.12). Accordingly, the Borderline Symptom List (short version; BSL-23; 32) showed more symptoms in the BPD group than in the HC group.
Ethical approval was obtained by the Ethics Committee of the University of Muenster, Germany on January 14th, 2014. All participants gave their written informed consent prior to study participation.
An allowance of 50 € was given to all participants.

Experimental procedure
We administered a modified version of the fearful face paradigm during fMRI (33) in order to investigate the neural correlates of the emotional response to happy faces. The original paradigm included movie sequences of fearful faces and landscapes. In our paradigm, we replaced fearful faces with happy faces and landscapes with neutral faces in order to enhance contrast specificity. This modification allowed us to link effects to the emotional valence of the presented faces while controlling for effects associated with movement perception or face recognition. Ninety-six different short movie sequences of male and female happy (happy condition) or neutral (neutral condition) faces were shown. Participants were instructed to passively watch the movie sequences; no active response was required. The experimental design is illustrated in Fig. 1.
Each of the experimental trials is composed of three phases: introduction, stimulus presentation (movie sequences) and pause. During the introduction phase, the words 'Los geht's' (english: 'Here we go') were displayed in white letters on a black screen and indicated the beginning of each trial.
The introduction phase had a duration of 3200 ms in average with a variable jitter between 1700 ms and 4700 ms (34). After the introduction, movie sequences were presented with a duration of 16000 ms. In these movie sequences, either happy (condition 'happy') or neutral faces (condition 'neutral') were shown. After the movie sequence, a pause followed, indicated by the word 'Pause' that was presented in white letters on black screen for 1700 ms. During this pause, a white fixation cross Two main scales are provided by the EAQ: acceptance of positive and acceptance of negative emotions. As we investigated responses to happy faces, we only included the acceptance of positive emotions scale comprising 16 items. In our sample, Cronbach's α was α = .93 for the healthy controls and α = .91 for the BPD group. The PANAS (38) was administered to assess positive and negative affectivity and includes 10 items for each condition that have to be rated on a 5-point-Likert-scale Imaging data acquisition All brain data were acquired in the MARA epilepsy centre (Bielefeld, Germany) with a 3-Tesla Siemens Magnetom Verio whole-body tomograph (Siemens, Erlangen, Germany) using a quantum gradient and a standard 12-channel head coil. Functional images were obtained by using a single T2-weighted gradient echo planar imaging (EPI) sequence with the following parameters: slice thickness = 4 mm (1 mm gap), TR = 2100 ms, TE = 30 ms, Flip angle = 90°, FOV = 192 × 192 mm, matrix size = 64 × 64, voxel size = 3 × 3 × 4 mm. The number of volumes was 377, each containing 30 axial slices of the whole brain and measured in descending order parallel to the hippocampus. During scanning, participants lay in a supine position in the MR scanner. An adjustable head holder was used to restrict potential head movements. Visual stimuli were shown on a back-projection screen (1024 pixels x 768 pixels) near the tube end using the software "Presentation" (Neurobehavioral Systems, Albany, CA, USA). Participants watched the screen via a dual-mirror that was mounted to the head coil. Before the EPI sequence, field map sequences were applied to control for magnetic field inhomogeneities. After the EPI sequence, high-resolution anatomical images were acquired using a T1-weighted, three-

Data analysis
Behavioral data. Behavioral data were analyzed with IBM SPSS Statistics 20.0 (SPSS Inc.) using repeated measures analyses of variance (for PANAS) and two-sample t-tests (for EAQ). All levels of significance were α ≤ 0.05 and tests were two-tailed.
Imaging data. Imaging data were analyzed using SPM12 (Statistical Parametric Mapping 12, Wellcome Department of Cognitive Neurology, London, UK; http://www.fil.ucl.ac.uk/spm). We discarded the first three images of every EPI-recording session to account for the time needed for the magnetic field to realize a steady state. The EPI data were preprocessed including field mapping, movement and slice time correction, co-registration to the individuals' structural images, 12 parameter non-linear normalization (3 mm x 3 mm x 3 mm) into the Montreal Neurological Institute (MNI) reference space, and smoothing (FWHM = 9 mm). Participants whose head movements exceeded the size of one voxel were excluded from analysis (n = 0). Functional imaging data were analyzed using a general linear model with five regressors, including one regressor for the movie sequences showing happy faces (condition 'happy), one regressor for the movie sequences showing neutral faces (condition 'neutral'), two regressors for the respective pauses, and one combined regressor for the introductions.
Additionally, the six alignment parameters were included into the design matrix as movement regressors of no interest. Regressors were convolved using the hemodynamic response function as provided in SPM 12 and design matrices were highpass filtered (128 s).
For the identification of neural activation associated with the emotional response to happiness, activation during the movie sequences showing happy faces (condition 'happy') was contrasted to activation during the movie sequences showing neutral faces (condition 'neutral'). This contrast ('happy minus neutral') was first computed for each subject on the first level. To identify happinessrelated neural networks in BPD-patients and HC-subjects as well as network differences between these groups, contrast images were then applied to second level random-effects within-and betweengroup analyses. To examine the impact of the level of habitual emotion acceptance on brain activation in the single experimental groups, we additionally conducted within-group analyses with the individual EAQ values (total score for acceptance of positive emotions) as covariate.
At first, all analyses in SPM were performed at whole-brain level. Whole-brain results were tested on the cluster level using a threshold of Z > 3.1 with a minimum cluster size of k ≥ 20 voxels and a cluster significance threshold of p ≤ 0.05 (family-wise error (FWE) corrected for multiple comparisons).
Moreover, we used a region-of-interest (ROI) to examine the neural response to happiness in small brain regions (amygdala, hippocampus, thalamus, insula, striatum) that are associated with positive emotion processing (39)(40)(41). The corresponding ROI masks were taken from the automated anatomical labeling atlas (AAL; 42), which is implemented in the Wake Forest University (WFU) PickAtlas, an automated software toolbox for generating ROI masks based on the Talairach Daemon database (43)(44)(45)(46). Contrast values from the contrast 'happiness minus neutral' were extracted using the SPM Marsbar toolbox (47) in order to further analyze these contrasts using SPSS. Contrast values represented BOLD signal changes from the 'neutral' to the 'happy' condition.

Results
Behavioral data: habitual emotion acceptance fMRI data For healthy controls (table 2a, Fig. 2), cluster-based whole-brain analysis identified two large clusters with activation peaks in left middle occipital gyrus and right middle temporal gyrus in the 'happy compared to the 'neutral' condition (contrast 'happy minus neutral'). Additional ROI-analyses revealed increased activation in the bilateral amygdala and the left hippocampus (table 2b).
For the patient group (table 2c, Fig. 2), cluster-based whole-brain analysis also revealed increased occipital and temporal but also increased precentral and striatal activation in the 'happy' compared to  (table 2d).
Between-group comparisons revealed no significant whole-brain differences between patients with BPD and healthy controls for the contrast 'happy minus neutral'. However, ROI-analyses revealed increased activation in bilateral caudate in patients with BPD compared to healthy controls for the contrast 'happy minus neutral' (table 2e, Fig. 3). Healthy controls did not show increased activation compared to patients with BPD in any of the ROIs. In both groups, neural activation in positive compared to neutral emotions did not correlate with emotion acceptance of positive emotions.

Discussion
In the current work, we explored the neural correlates of positive emotions in patients with BPD and healthy controls. We found temporal, limbic (including the amygdala) and occipital activation in both study groups. Our main findings show that patients with BPD exhibit a hyperactivation of striatal brain areas and particularly show stronger activation in bilateral caudate than healthy controls when they experience positive emotions. Other than expected, the disposition to accept positive emotions did not correlate with neural activation during positive emotion processing.
We found hyperactivation of the bilateral caudate during the processing of positive emotions in patients with BPD compared to healthy controls. The caudate is crucial for planning goal-directed behavior and for the processing of reward-related information (48)(49)(50). Together with the putamen and the ventral striatum, the caudate is part of an emotional-motivational system (51). Caudate hyperactivation in our study may reflect stronger emotional arousal in patients with BPD compared to healthy controls. Patients with BPD probably experience emotional states independent of the specific emotional valence (e.g. happiness or fear) more intensively (52). Moreover, Kim and Hamann (53) found activation in the dorsal striatum (caudate and putamen) and the pallidum in healthy controls when trying to increase positive emotions compared to a simple watching condition. In contrast to Kim and Hamann (53), participants of our study were not instructed to regulate their emotional reaction. Therefore, another explanation of our findings might be a habitual attempt of patients with BPD to regulate their emotional arousal. As caudal activation was only observable in patients with BPD, healthy controls possibly experienced less needs to regulate their emotional reaction than patients with BPD. However, our correlational analysis of habitual emotion acceptance and neural activity did not show significant results. Nevertheless, patients with BPD reported less emotion acceptance of positive emotions than healthy controls which is in line with previous findings (14). As emotion acceptance was assessed using a self-report questionnaire, a negative bias in subjective recall of emotional experience in patients with BPD (54) might have exaggerated the rating of emotion non-acceptance in the patient group. It is possible that the actual emotional experience during our experiment did not differ that much between patients with BPD and healthy controls on an acceptance-related neural level but patients might have remembered EAQ-relevant everyday situations differently than healthy controls.
Although the group-difference in hippocampal activity did not reach statistical significance, we found activation of the hippocampus in healthy controls but not in patients with BPD in the processing of positive emotions. The hippocampus is important for emotion processing and memory (55) with impaired functioning in mental disorders (56,57).,. In a recent study, Lambert et al. (58) found, that early-life stress (e.g. violence exposure in childhood) might lead to impaired hippocampal development. Thus, missing hippocampal activity in patients with BPD when positive emotions were induced may be attributable to stressful childhood events that are often reported in patients with BPD (59).
Our findings of amygdala activation in positive emotion processing in both groups is in line with previous findings (24,40). In a PET-study, Hamann et al. (60) found increased activation of the left amygdala when healthy subjects processed pictures conveying positive emotions and "interesting" images without apparent emotional content (e.g. rhinoceros). The authors suggested, that activation of the amygdala might reflect a vigilant reaction rather than the experience of a particular emotion. In line with that, Phelps (61) outlined that the amygdala appears to navigate attention to relevant stimuli and that this attention is facilitated by emotional context. Becauseamygdala activation in our study occurred in both study groups without significant group differences, we could not confirm altered amygdala activation in patients with BPD during the processing of positive emotions.
According van Zutphen et al. (62) findings of altered amygdala involvement in emotion processing in patients with BPD are rather inconsistent, resulting from methodical (e.g. fMRI acquisition settings like slice thickness) or individual (e.g. dissociation) differences. In contrast to our methodological approach, van Zutphen et al. (62) rather suggests a more liberal exploration of the amygdala with an uncorrected significance threshold of p ≤ 0.05. Moreover, we found activation of middle and superior occipital and temporal regions in the processing of happy faces that was also comparable in both groups. Occipitotemporal activation is associated with the perception of faces and their emotional expressions (63,64) as well as with positive emotion processing (53). This indicates stronger visual perception of happy compared to neutral faces in all participants of our study. The study groups did not seem to differ in the perception of happy (facial) expressions.
In total, our study has several strengths. To make sure that confounding aspects (e.g. processing of facial attributes) were eliminated, we used stimuli (facial expressions) that only differed with respect to the emotional content. Moreover, we did not use any kind of cognitive task that could have influenced emotion processing. On the contrary, some limitations have to be discussed, too. First, some patients also suffered from other diagnoses (e.g PTSD). However, comorbid disorders are typical for patients with BPD and their exclusion would have resulted in a non-representative study group.
Second, psychotropic medication may have influenced emotion processing of our patients. Third, we did not implement immediate emotional valence ratings after each stimuli presentation. However, although this might have enabled a more detailed estimation of emotion processing in our study, it might also have influenced neural activity (65). Forth, the impact of our positive emotion stimuli might not have been strong enough to elicit neural reactions that could be associated with habitual

Declarations
Acknowledgements: The authors thank Violetta Swiatek for her expertise and kind help during the fMRI investigations. We also thank all participants for their willingness to take part in this study. We thank the German Research Foundation ("Deutsche Forschungsgemeinschaft", DFG) for the financial support.
Competing interests: None declared.  Whole-brain activation in patients with BPD (blue) and healthy controls (green). Displayed are the whole-brain results of within-group cluster-level analyses with cluster significance thresholds of p ≤ 0.05 (family-wise error (FWE) corrected for multiple comparisons) and minimum cluster sizes of 20 voxels. Tables.docx