Abstract
Background
Bipolar disorder (BD) and unipolar depression (UD) can be difficult to distinguish clinically, particularly during episodes of depression. In this study we test for differences between BD, UD, and healthy control (HC) adults regarding within- and between-session changes in BOLD response during implicit emotional processing.
Methods
During fMRI, HC adults (N=19) and depressed adults with UD (N=19) and BD (N=16) performed an implicit emotion-processing task. Each participant was scanned twice, separated by 6-months, resulting in 108 scans. BOLD response and linear change in BOLD response were examined within and between sessions.
Results
We observed within-session linear decreases in BOLD signal (irrespective of group, condition, or session) in the left amygdala, a right-sided temporo-parietal region, and a right-sided fronto-insular region. Furthermore, we observed group differences in within-session BOLD signal change (p<0.05, FWE corrected) in a left-sided striatal-insular-thalamic region. Individuals with BD demonstrated a linear decrease in BOLD signal compared to HC (p<0.008, FWE corrected) across this region and compared to UD in the posterior insula portion of the region (p<0.008, FWE corrected). Finally, we observed main effects of emotional valence in bilateral visuo-spatial processing regions as well as in the left and right amygdala.
Conclusions
Individuals with BD demonstrated linear attenuation of BOLD response to emotional stimuli within left-sided striatal-insular-thalamic regions. Individuals with BD may either have experienced abnormal habituation in this region or disengaged quickly from processing the emotional stimuli, despite comparable task performance. This pattern may represent an underlying pathophysiological process associated with BD that differs from UD.
Keywords: Emotion processing, Bipolar Disorder, Major Depressive Disorder, fMRI, BOLD attenuation, Whole brain analysis
Introduction
Bipolar disorder can be difficult to diagnose correctly, particularly during depressive episodes. The delay between initial treatment-seeking and correct diagnosis can span more than a decade (1,2). Even when a bipolar disorder diagnosis is made, as many as 17%–27% of individuals will go on to be misdiagnosed by other treatment providers later in the course of their illnesses (3,4). The potential value of identifying biologically based markers that could help differentiate bipolar and unipolar illness is clear (e.g., 5,6), and there has been a recent increase in the number of functional neuroimaging studies aimed at helping to identify such differences (6,7). The majority of this work, however, has examined neural activity as a function of average Blood Oxygen Level Dependent (BOLD) response during a single scanning session. In the current work, we test for differences between bipolar depressed (BD), unipolar depressed (UD), and healthy control (HC) individuals regarding the temporal dynamics of the BOLD response during emotion processing within and between neuroimaging sessions. We argue that examining the dynamics of neural regions involved in emotion processing may provide important information about the pathophysiology of the illness.
Prior neuroimaging work has revealed abnormalities in several key circuits among individuals with BD compared to healthy individuals. In a recent review, Phillips and Swartz (6) identified that individuals with BD demonstrate dysfunction in cortico-limbic regions during emotion regulation coupled with hypersensitivity to positive and rewarding stimuli in striatal and frontal reward systems. Abnormalities that can help to differentiate individuals with BD from those with UD have been more challenging to identify (7).
Across a small number of studies, differences between UD and BD to positive and negatively valenced emotional faces have been observed in limbic (8–11), subcortical (12), visuospatial (11), and prefrontal regions (10,12). Different BOLD activity patterns have emerged across studies, however. For example, in one study from our research group, we (9) observed increased activity in the amygdala to faces displaying a mild sad emotion in BD relative to UD, HC, and individuals with BD who were in remission. In another study from our group, using a subsample of the Session 1 data analyzed below (11), we likewise observed increased activity to dynamically emerging facial displays of sadness in BD relative to UD, but this time in several temporo-parietal regions involved in visuo-spatial processing. By contrast, in a set of studies using pattern classification techniques, Grotegard and colleagues (10,13) observed the opposite pattern. That is, they found that increased amygdala activation to negative (10) and sad (13) faces was key to accurately classifying individuals with UD, whereas increased amygdala activity to happy faces was more important in classifying individuals with BD.
One way to begin to account for the varied emotional valence effects in neural response between unipolar and bipolar disorder is suggested by Grotegard and colleagues (13) who note that emotion processing is a dynamic mechanism that unfolds over time. They suggest that different findings may obtain depending on when in that time course the system is probed (also see, 14). We agree that examining the temporal dynamics of emotion processing is an important next step in the search to identify biomarkers that may differ between UD and BD.
In the current work, we examine differences among BD, UD, and HC individuals regarding mean BOLD activity as well as changes in the BOLD signal that occur between scanning sessions, separated by a six-month interval, and changes that occur over time within scanning sessions in response to dynamically emerging positive and negative facial displays. The six-month interval was chosen to be representative of the durations of typical major depressive episodes (see, 15) and, therefore, to allow sufficient time for symptom reductions. Following our approach in prior work (9,11,12), and given the findings noted above, we focused the analyses on between group differences in the amygdala and across the whole-brain. Based on our prior work (9,11), we hypothesized increased mean BOLD response to negative emotions among adults with bipolar relative to unipolar depression across sessions. Whereas we expected to observe attenuation of the BOLD signal across groups and conditions in regions involved in emotional processing, including the amygdala (potentially representing habituation effects over time), based on related findings from previous work (13), we expected to observe greater attenuation of BOLD signal to happy faces among the adults with unipolar versus bipolar depression, both within- and between-sessions. Finally, an exploratory aim of the project was to examine the degree to which changes in brain activity across sessions tracked with changes in symptoms. We expected that normalization of brain function in patient groups would be associated with symptom reduction.
Materials and Methods
Participants
The 57 participants for the current study were a subset of individuals who participated in a larger examination of neural differences between bipolar and unipolar depression (R01-MH076971, PI:MLP). The current sample represents those who were scanned at an initial visit (Session 1) and who were scanned again six months later (Session 2): 21 currently depressed adults diagnosed with UD, 17 currently depressed adults diagnosed with BD, and 19 HC participants. All participants were right handed, all were native English speakers, and all were screened with the Structured Clinical Interview for Psychiatric Disorders (16). In all diagnostic groups, participants were excluded if they had a history of head injury, systemic medical illness known to interfere with blood flow or brain function, cognitive impairment (score < 24 Mini-Mental State Examination, 17), color-blindness, premorbid IQ estimate < 85 (18), psychosis (including any psychotic mood symptoms), borderline personality disorder (19), and standard MRI exclusion criteria (e.g. metallic objects in the body). Participants in the UD and BD groups were also excluded if they met criteria for an alcohol/substance use disorder within 2 months prior to the scan. For the HC group, additional exclusion criteria included any current or prior alcohol or substance use disorder and any personal or immediate-family history of an Axis I disorder. All participants in all groups were required to pass urine drug screens and breath alcohol tests prior to both scanning sessions. In the original project from which these data originated, participant groups were age-and gender-matched. Because the current sample consisted of the subset of participants who returned for a second scan, rigid gender matching was not possible, but rather will be addressed statistically (see below). Three participants, one from the BD group and two from the UD group, were excluded due to behavioral performance of <75% correct during the scan. Thus, the final sample included 19 adults with UD, 16 adults with BD, and 19 HC adults (Table 1), each of whom were scanned twice, six months apart, resulting in a total of 108 scans. All participants were followed naturalistically during the six-month assessment window. No specific intervention was administered, and participants were free to receive, augment, or terminate treatment as needed. The study protocol was approved by the University of Pittsburgh Institutional Review Board. After hearing a complete description of the study, participants provided written informed consent.
Table 1.
Demographics and Behavior.
| Variable | UD(M=19) | BD(N=16) | HC(N=19) | Effect of Group | Post-hoc |
|---|---|---|---|---|---|
| Demographics: | |||||
| %Female | 78.95% | 93.75% | 57.89% | χ2(2)=6.26* | BD > HC* |
| Age, M(SD) | 32.53 (7.24) | 35.22 (8.91) | 32.17 (6.28) | F(2,51)=0.84 | |
| IQ (NART Estimate), M(SD) | 111.34 (10.29) | 114.10 (8.84) | 113.19 (7.68) | F(2,51)=0.43 | |
| Behavior (Session 1): | |||||
| Accuracy, M(SD) | 95.32% (4.94%) | 92.84% (7.01%) | 96.86% (3.00%) | χ2(2)=4.14a | |
| Reaction Time, M(SD) | 959.72 (145.96) | 997.57 (154.16) | 948.70 (185.44) | χ2(2)=1.63a | |
| Behavior (Session 2): | |||||
| Accuracy, M(SD) | 94.48% (5.78%) | 93.97% (7.64%) | 92.03% (7.76%) | χ2(2)=1.07a | |
| Reaction Time, M(SD) | 961.04 (161.06) | 985.65 (184.55) | 941.66 (106.96) | χ2(2)=0.15a | |
Note. UD = Unipolar Depressed, BD = Bipolar Depression, HC = Healthy Control.
Due to non-normality, the non-parametric Kruskal-Wallis test was used for comparisons. HRSD=Hamilton Rating Scale for Depression, YMRS=Young Mania Rating Scale,
p<0.05
Clinical Measures
Demographic, family history, medication, and clinical information were collected through self-report questionnaires and clinical interviews. Depressive symptoms were assessed using the 25-item Hamilton Rating Scale for Depression (HRSD, 20), and manic symptoms were assessed using the Young Mania Rating Scale (YMRS, 21). Current and trait anxiety was assessed using the Spielberger State-Trait Anxiety Inventory (22). IQ was estimated with the National Adult Reading Test (NART, 18).
Paradigm
Participants completed a 12.5-minute emotional dynamic faces task during fMRI. Stimuli comprised faces from the NimStim set (23) that were morphed in 5% increments, from neutral (0% emotion) to 100% emotion for 4 emotions: happy, sad, angry and fear (24–27). Morphed faces were collated into 1s movies progressing from 0% to 100% emotional display. In control trials, movies comprised a simple shape (dark oval) superimposed on a light-grey oval, with similar structural characteristics to the face stimuli, which subsequently morphed into a larger shape, approximating the movement of the morphed faces. Participants were asked to use one of three fingers to press a button indicating the color of a semi-transparent foreground color flash (orange, blue, or yellow) that appeared during the mid 200ms–650ms of the 1s presentation. The emotional faces were task-irrelevant and, thus, were processed implicitly. Each condition consisted of 36 total trials (see Supplement for additional details).
Data Analyses
Neuroimaging data acquisition
Structural and functional fMRI data were collected using a 3-Tesla Siemens Trio MRI scanner. See Supplement for detailed acquisition parameters.
Functional neuroimaging data preprocessing
Data were preprocessed with statistical parametric mapping software version 12b (SPM12b; http://www.fil.ion.ucl.ac.uk/spm). During preprocessing, data were motion corrected, co-registered, segmented and normalized into Montreal Neurological Institute [MNI] space, resampled to 3×3×3mm3, and smoothed with an 8-mm full-width at half maximum (FWHM) Gaussian kernel. AFNI 3dDespike was used to reconstruct scans with abnormal spikes in signal intensity.
Functional neuroimaging data analyses
Each of the four emotion conditions (anger, fear, sad, happy) was entered as separate conditions in the design matrix, as was the shape condition, which served as the baseline. The experiment was treated as a mixed block/event-related design at the first level, and events were convolved to the canonical hemodynamic response function. Standard first-level, within-subject general linear models were fit. The six movement parameters from the preprocessing procedure, along with their first-order derivatives, were entered as covariates of no interest. We also included an additional nuisance regressor to capture potential physiological confounds in the global signal (calculated as the average signal from voxels located within a white matter and CSF mask and described in detail in an earlier methodological examination of first-level model specification using some of the data reported below, 28). Finally, changes in BOLD signal within each scanning session were modeled by the addition of orthogonal first- and second-order polynomial terms corresponding to each event type. In order to examine our hypotheses regarding BOLD response to negative emotions, the anger-, fear-, and sad- minus shape contrasts were pooled to form one negative-emotion-versus-shape contrast. The decision to pool these contrasts was made a-priori, given statistical power constraints, the absence of any specific hypotheses regarding differential group effects of BOLD attenuation to different negative facial expressions, and previous studies using this approach in other patient samples (e.g., 29). First-level models were calculated separately for each testing occasion, resulting in separate contrast estimates for each.
Whole brain analyses
Across the whole brain, we examined one between-subjects factor (group) and two within-subjects factors (session and emotion-valence), resulting in a 3(group)×2(session)×2(emotion-valence) repeated measures ANOVA. We utilized the partitioned-errors approach (30), implemented in the GLM Flex suite of tools (http://mrtools.mgh.harvard.edu/index.php/Main_Page) to ensure that the proper error terms were used for each contrast. Outliers, calculated separately for each voxel on the basis of Cook’s D, were removed. Separate second level models were estimated to examine mean BOLD signal and linear changes in BOLD signal within sessions.
Inference and multiple comparison correction
Inferences were made on the basis of cluster level statistics (cluster forming threshold p<0.005, uncorrected, extent threshold p<0.05 corrected). AFNI 3dClusterSim software was used to implement Monte Carlo simulations to determine the familywise error corrected cluster extent thresholds, using the smoothness values for each error term estimated separately from the second level models above. The required cluster extents for mean BOLD activity were: a) k > 100 for group, b) k > 101 for session, c) k > 89 for emotion-valence, and d) k > 92 for session-by-emotion-valence interaction effects. The corresponding values for linear within-session changes in BOLD signal were k > 102, k > 97, k > 84, and k > 90, respectively. We computed all post-hoc tests of main- and interaction-effects using the same whole-brain cluster correction techniques. The omnibus FWE alpha levels of these tests were Bonferroni corrected for the number of tests performed. Only those portions of the significant (whole-brain cluster-corrected) post-hoc clusters that overlapped with the initial main or interaction effects are discussed below. Furthermore, only statistically significant results are discussed (see Supplement for discussion non-significant trends, p<0.10, in the data).
Clinical and demographic data analyses
Because voxelwise repeated measures ANCOVAs are not possible using standard neuroimaging software, in order to examine the effects of demographic and clinical variables, we extracted beta values representing the contrasts of interest from those regions identified in the analyses described above. We next examined whether demographic or clinical variables could account for any observed group differences. To examine differences that emerged among all three groups (UD, BD, and HC), we examined the effects of sex, age, and IQ. To examine differences between the two clinical groups, we examined: illness duration, current antidepressant medication (y/n), current antipsychotic medication (y/n), current mood stabilizer medication (y/n), current benzodiazepine (y/n), number of prior mood disorder episodes, total 25-item HRSD score, total YMRS score, total state-anxiety score, and total trait-anxiety score. Repeated measures maximum likelihood models, estimated with SAS version 9.4 statistical software (SAS Institute Inc., Cary, NC), were constructed in order to incorporate changes in these variables between testing occasions. Because these analyses were intended to identify potential confounds, we first employed a liberal p-value threshold of p < 0.10 to identify those variables that exhibited at least a trend-level significant bivariate relationship with the extracted data. Next, we added all of the variables that crossed this threshold to a model that included group membership.
Region of interest analysis (ROI)
To examine activity in the right and left amygdala we extracted data from anatomical amygdala masks, as defined in the Wake Forest Toolbox PickAtlas Talairach Daemon template(31). We used SAS 9.4 to estimate repeated measures maximum likelihood models to examine the effects of group, emotional-valence, session and relevant covariates (described above).
Results
Demographic and behavioral data are presented in Table 1. The ratio of males to females differed as a function of group membership. Post-hoc contrasts revealed that the ratio was more balanced in HC compared to BD groups. Overall, task accuracy was high, and there were no significant effects of group regarding behavioral performance. Clinical characteristics are displayed in Table 2. The two patient groups differed on three baseline clinical variables. Participants with BD experienced a greater number of previous episodes of illness, and a higher proportion were taking antipsychotic and mood stabilizing medications. On average, symptoms of depression (t(34)=−4.31, p<.001) and sub-threshold mania (t(33)= −3.55, p=0.001) were significantly reduced between sessions. No differences in symptom change were observed between BD and UD (Table 2).
Table 2.
Clinical Characteristics.
| Variable | UD(M=19) | BD(N=16) | Effect of Group |
|---|---|---|---|
| Session 1 | |||
|
| |||
| # Episodes, M(SD) | 2.58 (0.69) | 5.06 (1.95) | F(1,33)=26.96*** |
| HRSD, M(SD) | 26.53 (4.86) | 24.50 (8.25) | F(1,33)=0.81 |
| YMRS, M(SD) | 3.95 (2.37) | 3.94 (2.57) | F(1,33)<0.01 |
| State Anxiety, M(SD) | 57.42 (8.14) | 54.81 (11.46) | F(1,33)=0.62 |
| Trait Anxiety, M(SD) | 58.47 (10.32) | 58.33 (8.79)a | F(1,32)<0.01 |
| Duration(years), M(SD) | 14.11 (8.31) | 17.78 (8.50) | F(1,33)=0.20 |
| Diagnostic Comorbidities (%): | |||
| Current Anxiety | 63.16% | 37.50% | χ2(1)=2.29 |
| Current Somatoform | 0.00% | 6.25% | Fisher’s Exact: NS |
| Current Eating | 0.00% | 0.00% | – |
| Lifetime Drug/Alcohol | 31.58% | 50.00% | χ2(1)=1.23 |
| Medication (% Prescribed): | |||
| Antidepressant | 63.16% | 43.75% | χ2(1)=1.32 |
| Antipsychotic | 0.00% | 50.00% | χ2(1)=12.31*** |
| Mood Stabilizer | 10.53% | 62.50% | χ2(1)=10.41*** |
| Benzodiazepine | 15.79% | 18.75% | χ2(1)=0.05 |
|
| |||
| Change (Session 2 − Session 1) | |||
|
| |||
| Additional Episodes | 10.53% | 0.00% | Fisher’s Exact: NS |
| HRSD Change, M(SD) | −7.16 (6.85) | −7.56 (13.17) | F(1,33)=0.01 |
| YMRS Change, M(SD) | −2.5 (3.29)a | −1.63 (3.63) | F(1,32)=0.54 |
| State Anxiety Change, M(SD) | −8.17 (11.27)a | −1.60 (17.66)a | F(1,31)=0.21 |
| Primary Diagnosis Change (%)b | Fisher’s Exact: NS | ||
| Partial Remission | 0.00% | 12.50% | |
| Remission | 15.79% | 31.25% | |
| Medication (% Prescribed): | |||
| Antidepressant | Fisher’s Exact: NS | ||
| % Increase | 10.53% | 6.25% | |
| % Decrease | 0.00% | 6.25% | |
| Antipsychotic | Fisher’s Exact: NS | ||
| % Increase | 5.26% | 6.25% | |
| % Decrease | 0.00% | 6.25% | |
| Mood Stabilizer | Fisher’s Exact: NS | ||
| % Increase | 5.26% | 6.25% | |
| % Decrease | 5.26% | 6.25% | |
| Benzodiazepine | Fisher’s Exact: NS | ||
| % Increase | 5.26% | 0.00% | |
| % Decrease | 0.00% | 0.00% | |
Note. UD = Unipolar Depressed, BD = Bipolar Depression, HC = Healthy Control, NS = Non-Significant.
Data from one participant was missing.
Percentages concern changes to the primary diagnoses corresponding to group membership. No UD individuals experienced episodes of mania or hypomania or converted to bipolar disorder spectrum diagnosis. HRSD=Hamilton Rating Scale for Depression, YMRS=Young Mania Rating Scale,
p<0.001
Mean BOLD Response
Whole brain analyses
In the examination of average BOLD activity, we observed a significant main effect of emotional-valence in six clusters (Figure 1, Table 3). Two clusters demonstrated increased BOLD response to positive emotional stimuli across groups and sessions. These left- and ride-sided clusters included portions of bilateral thalamus and the hippocampus. Four clusters demonstrated increased activity to negative stimuli. These included bilateral visual regions, portions of Lobules I–V of the cerebellum, and a cluster in the frontal pole.
Figure 1.
Main Effect of Emotional Valence (Mean BOLD Signal). Clusters represent significant effects of emotional valence (negative vs positive facial displays of emotion) for average BOLD activity, across groups and across sessions. Plotted values represent t statistics. Green-blue clusters represent regions wherein the negative-emotion-minus-shape-contrast was significantly larger than the positive-emotion-minus-shape contrast. Red-orange clusters represent the reverse. All effects are familywise error cluster-corrected at p<0.05, using an initial cluster-forming threshold of p<0.005 and Monte-Carlo methods to determine minimum cluster extents.
Table 3.
Coordinates of Clusters Showing Significant Activations.
| Contrast Region | K | x | y | z |
|---|---|---|---|---|
| Mean BOLD Response | ||||
| Main effect of condition (negative > positive) | ||||
| R. Inf./Mid. Occipital Gyrus, Fusiform Gyrus | 527 | 45 | −76 | −5 |
| L. Inf. Occipital Gyrus, Fusiform Gyrus | 274 | −36 | −82 | −11 |
| Cerebellar Lobules I–IV, V | 125 | 9 | −31 | −11 |
| Frontal pole | 120 | −12 | 56 | 28 |
| Main effect of condition (positive > negative) | ||||
| R. Hippocampus, Thalamus | 228 | 12 | −34 | 16 |
| L. Hippocampus, Thalamus | 155 | −9 | −34 | 16 |
|
| ||||
| Linear BOLD Attenuation | ||||
| General attenuation | ||||
| R. Sup. Temporal Gyrus, Mid. Temporal Gyrus, | 116 | 57 | −55 | 10 |
| Inf. Parietal Lobule, Supramarginal Gyrus | ||||
| R. Anterior Insula, Inf. Frontal Gyrus, BA 47 | 114 | 45 | 26 | −8 |
| Main effect of condition (negative > positive) | ||||
| Cuneus, L./R. Calcarine, L./R. Lingual | 485 | −12 | −88 | 1 |
| L./R. Precuneus | 108 | 6 | −61 | 61 |
| Main effect of group | ||||
| L. Thalamus, Post. Insula, Caudate, Putamen | 201 | −33 | −10 | −8 |
Note. R. = Right, L. = Left, Inf. = Inferior, Mid. = Middle, Sup. = Superior, Post. = Posterior
Amygdala ROI
In the analysis of data extracted from the left and right amygdala, we observed significant main effects of emotional-valence regarding average BOLD response (F(1,51)=7.85,p=0.007 on the left; F(1,51)=9.91,p=0.003 on the right). For both sides, average BOLD response was higher for negative compared to positive faces.
Linear BOLD Attenuation within Sessions
Whole brain analyses
We observed general within-session attenuation (linear decreases) of the BOLD signal to the emotional stimuli over time during the task in two clusters, one in a right-sided fronto-insular region and one in a right-sided temporo-parietal region (Figure 2a, Table 3). Further, we observed a significant effect of emotional valence in a large medial occipital cluster and a cluster in bilateral posterior precuneus (Figure 2b, Table 3). Both clusters evidenced linear increases in within-session BOLD response over time to negative emotional stimuli relative to positive stimuli.
Figure 2.
Within Session BOLD Attenuation. A) The clusters represent significant BOLD signal attenuation across groups, emotional-valence conditions, and sessions during an implicit emotion-processing task (whole-brain, Monte-Carlo cluster-corrected FWE p<0.05). Plotted values represent t statistics. B) The clusters represent significant effects of emotional-valence (positive vs negative emotional faces) regarding within-session linear change in BOLD response. Plotted values represent F statistics. The bar graph is provided for descriptive purposes only. It depicts extracted beta estimates from the regions depicted. Bars represent the slope of the linear change in BOLD signal. Negative values represent decreases in BOLD activity over time; positive values represent increases in BOLD activity. Error bars represent +/−1 standard error.
We observed a significant main effect of group on within-session BOLD attenuation in a large cluster in left insular-striato-thalamic regions (Figure 3a). Independent, whole-brain cluster-corrected (FWE p<0.0083) post-hoc analyses revealed that the effect was driven primarily by a greater within-session linear attenuation in BOLD signal for the BD group compared to the UD group in the posterior insula portion of the region, and compared to the HC group across the region (Figure 3b. See Supplement for further information). No HC versus UD comparisons were significant (all ks < 21).
Figure 3.
Main Effect of Group (Within Session BOLD Attenuation). A) The cluster represents a significant effect of group (bipolar depressed [BD] vs healthy control [HC] vs unipolar depressed [UD]) regarding within-session linear change in BOLD response in a left striatal-insular-thalamic region during the implicit emotion-processing task (whole-brain, Monte-Carlo cluster-corrected FWE p<0.05). Plotted values represent F statistics. B) The clusters represent whole-brain, cluster-corrected (FWE p<0.0083) post-hoc comparisons of BP versus HC (yellow) and BP versus UD (Green). Plotted clusters represent the overlap between the significant clusters identified in the whole-brain post-hoc analysis and the main-effect of group (panel A). Full whole-brain post-hoc results are presented in the Supplement. Plotted values represent t-statistics. C) The bar graph is provided for descriptive purposes only, it depicts extracted beta estimates from the region depicted in Panel A. Bars represent the slope of the linear change in BOLD signal. Negative values represent decreases in BOLD activity over time; positive values represent increases in BOLD activity. Error bars represent +/−1 standard error.
Potential confounds
We examined whether clinical, demographic, or behavioral effects could account for the group difference in attenuation of BOLD signal in the left insular-striato-thalamic region. Neither sex, age, or IQ was associated with attenuation of activity in this region across the full sample (all Fs(1,52)<2.24, p>0.14). Regarding possible clinical confounds among those in the two depressed groups (UD and BD), two variables crossed the initial p<0.10 threshold: antipsychotic and mood stabilizing medications, however neither remained significant when added to the model containing group and session (both Fs<0.23, ps>0.66). Moreover, within the BD group, neither medication was associated with estimates of BOLD signal attenuation in this region (both Fs< 0.05, ps>0.87). See the Supplement for secondary analyses regarding effects of the gender of the facial images.
Although behavior on the task (i.e., color matching), was not directly related to the emotional content of the images, behavioral performance might serve as a proxy for task engagement. In order to determine whether the group differences in attenuation could be explained by differences in behavioral performance, we computed linear mixed-effects models of the trajectories of reaction times (using SAS PROC MIXED) and accuracy (using SAS PROC GLIMMIX). We observed no evidence that individuals with bipolar disorder differed from the other groups with respect to the trajectories of reaction times or accuracies over the task (Fs(1, 54)<1.55, ps>0.21).
Amygdala ROI
We observed significant general within-session BOLD attenuation in the left amygdala, irrespective of group, condition, or session (t(53)= −3.14,p=0.003). On the right, we also observed a main effect of emotional valence whereby greater within-session attenuation occurred for the negative emotional faces compared to the positive faces (F(1,51)=5.14, p=0.03).
Discussion
Contrary to our hypotheses, we did not observe greater average BOLD activity in the BD group compared to the UD group during the negative emotional faces condition, nor did we observe attenuation of BOLD signal to positive faces either within- or between-sessions among members of the UD group. Rather, the primary finding of this study was that in a large area that spanned portions of the left thalamus, striatum, and posterior insula, individuals with BD displayed abnormal BOLD signal attenuation, relative to UD in the posterior insula and to HC individuals in all three regions, during an implicit emotion processing task. These effects were observed despite the fact that no corresponding decrements were observed in behavioral performance. That is, the individuals with BD did not seem to be disengaging from the task perse, but rather, they showed a linear decline in BOLD signal in response to the emotional information contained in the task. Furthermore, we did not observe any evidence that the group differences regarding attenuation varied across the two sessions.
In addition, we observed general within-session attenuation of BOLD signal in the left amygdala and in right fronto-insular and temporo-parietal regions. This attenuation was observed irrespective of group, condition or session. Activity in these regions has been associated with social-affective processing (32–34), and these findings suggest that over time, these regions became generally less activated to the emotional facial stimuli. Given that no differences were observed between the patient and HC samples in these regions, such attenuation is likely normative.
By contrast, our primary findings regarding abnormal attenuation of an insular-striato-thalamic system among BD compared to the other two groups suggests that part of the pathophysiology of BD could involve the additional, and abnormal, habituation of this system, potentially subserving attention, sustained arousal, and behavioral responses to emotional stimuli as function of the perceived salience of those stimuli (35–40). Our previous findings indicated elevated left-sided striatal and ventrolateral prefrontal cortical activity during uncertain reward anticipation in individuals with BD relative to HC and UD (41–43). The uncertainty of the outcomes in these studies may have increased the perceived salience of task stimuli to individuals with BD and led to increases in BOLD response in these regions. We have interpreted this pattern of abnormally elevated activity as a neural mechanism underlying greater encoding of future reward potential, which may be evident in individuals with higher levels of impulsive sensation seeking and risky decision-making, characteristics of BD. By contrast, the pattern of abnormally increased attenuation in the left-sided thalamo-striatal-insula cluster in individuals with BD in the present study may reflect the relative similarity and predictability of the stimuli within each emotional block, and the absence of any uncertainty regarding task outcomes. Future research will be needed to investigate further the interplay between salience and emotion processing in bipolar disorder, as well as the real-world consequences of altered functioning in these systems.
In addition to the group differences in BOLD attenuation, we also observed main effects of emotional valence in average BOLD response in several brain regions that have been implicated more generally in emotional face processing, including the left and right amygdala (e.g., 44,45,46). Furthermore we observed a pattern of within-session changes in BOLD response whereby response in the right amygdala attenuated to a greater degree for negative than to positive faces – a pattern of findings similar to those of Phillips and colleagues (47) who observed a more rapidly decreasing BOLD response in the right amygdala to fearful faces. We also observed a pattern of within-session strengthening of BOLD response to negative faces in occipital and posterior parietal regions, potentially representing increased activity in portions of the dorsal visual stream over time to the negative emotional images. These results are in line with those of Goldberg and colleagues (48) who reported preferential activity in areas of the dorsal visual stream to dynamic, naturalistic movie clips depicting emotional scenes. Finally, it is not immediately clear why we did not observe emotional-valence-by-group interaction effects or significant changes in BOLD signal across sessions. One possibility is that although the sample size of the current study is the largest to date to compare the longitudinal functioning of the emotion processing system among bipolar and unipolar depression, it was too small to examine more fine-grained distinctions in the functioning of the system for different discreet emotions.
Limitations
The results of the current study must be understood in the context of several limitations. First, given the naturalistic follow-up design, the groups were unbalanced with respect to gender. Gender was not a significant predictor of BOLD activity, but future work will be needed to verify that the observed BOLD attenuation effects replicate in both men and women. Second, we did not have access to data from a sufficiently large group of remitted UD or BD participants to examine their data separately. We did examine whether symptom changes were correlated with the observed effects and found no evidence of any such association. Finally, many of the UD and BD adults were taking psychotropic medications, and the two groups differed in the proportion of participants taking antipsychotic and mood stabilizing medications. None of the effects of medications remained significant when examined in statistical models that also contained a term representing group membership. Due to the sample sizes and the naturalistic design, it is not possible strictly to parcel out the unique effects of medications and group membership. Although we cannot rule out the possibility that medication affected BOLD response, the most likely explanation of the pattern of results that we observed is that the medications were associated with BOLD activity in bivariate models because of their correlation with group membership, rather than the reverse (see 49,50).
Conclusion
The findings from this study highlight the importance of examining changes in the BOLD signal over time for understanding the pathophysiology of bipolar disorder. Specifically, they suggest that individuals with bipolar disorder differ both from healthy control participants and from individuals with unipolar depression in the degree to which left-sided striato-insular-thalamic activity is sustained during emotion processing. This may represent an important pathophysiological process in bipolar disorder, which if replicated in independent samples, could represent a core deficit associated with the illness.
Supplementary Material
Acknowledgments
All work was carried out within the Department of Psychiatry, University of Pittsburgh. Neuroimaging data were collected at the Magnetic Resonance Research Center (MRRC), University of Pittsburgh. We thank the MRRC staff for their help acquiring neuroimaging data.
This work was supported by the National Institute of Mental Health (MP, MH076971) and the Pittsburgh Foundation (MP). Additionally, JF receives financial support from MH097889.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Financial Disclosures
MLP serves as a consultant to Roche Pharmaceuticals. All other authors report no biomedical financial interests or potential conflicts of interest.
References
- 1.Lish JD, Dime-Meenan S, Whybrow PC. The National Depressive and Manic-depressive Association (DMDA) survey of bipolar members. J Affect Disord. 1994;31:281–94. doi: 10.1016/0165-0327(94)90104-x. [DOI] [PubMed] [Google Scholar]
- 2.Hirschfeld RMA, Lewis L, Vornik LA. Perceptions and impact of bipolar disorder: How far have we really come? Results of the National Depressive and Manic-Depressive Association 2000 Survey of Individuals with Bipolar Disorder. J Clin Psychiatry. 2003;64:161–74. [PubMed] [Google Scholar]
- 3.Stensland MD, Schultz JF, Frytak JR. Diagnosis of unipolar depression following initial identification of bipolar disorder: A common and costly misdiagnosis. J Clin Psychiatry. 2008;69:749–58. doi: 10.4088/jcp.v69n0508. [DOI] [PubMed] [Google Scholar]
- 4.Stensland MD, Schultz JF, Frytak JR. Depression diagnoses following the identification of bipolar disorder: Costly incongruent diagnoses. BMC Psychiatry. 2010;10:39. doi: 10.1186/1471-244X-10-39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Phillips ML, Kupfer DJ. Bipolar disorder diagnosis: Challenges and future directions. The Lancet. 2013;381:1663–71. doi: 10.1016/S0140-6736(13)60989-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Phillips ML, Swartz HA. A critical appraisal of neuroimaging studies of bipolar disorder: Toward a new conceptualization of underlying neural circuitry and a road map for future research. A J Psychiatry. 2014;171:829. doi: 10.1176/appi.ajp.2014.13081008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Almeida JRC, Phillips ML. Distinguishing between unipolar depression and bipolar depression: Current and future clinical and neuroimaging perspectives. Biol Psychiatry. 2012;73:111–8. doi: 10.1016/j.biopsych.2012.06.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Almeida JRC, Mechelli A, Hassel S, Versace A, Kupfer DJ, Phillips ML. Abnormally increased effective connectivity between parahippocampal gyrus and ventromedial prefrontal regions during emotion labeling in bipolar disorder. Psychiatry Research: Neuroimaging. 2009;174:195–201. doi: 10.1016/j.pscychresns.2009.04.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Almeida JRC, Versace A, Hassel S, Kupfer DJ, Phillips ML. Elevated amygdala activity to sad facial expressions: A state marker of bipolar but not unipolar depression. Biol Psychiatry. 2010;67:414–21. doi: 10.1016/j.biopsych.2009.09.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Grotegerd D, Suslow T, Bauer J, Ohrmann P, Arolt V, Stuhrmann A, et al. Discriminating unipolar and bipolar depression by means of fmri and pattern classification: A pilot study. Eur Arch Psychiatry Clin Neurosci. 2012;263:119–31. doi: 10.1007/s00406-012-0329-4. [DOI] [PubMed] [Google Scholar]
- 11.Fournier JC, Keener MT, Almeida J, Kronhaus DM, Phillips ML. Amygdala and whole-brain activity to emotional faces distinguishes major depressive disorder and bipolar disorder. Bipolar Disorders. 2013;15:741–52. doi: 10.1111/bdi.12106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lawrence NS, Williams AM, Surguladze S, Giampietro V, Brammer MJ, Andrew C, et al. Subcortical and ventral prefrontal cortical neural responses to facial expressions distinguish patients with bipolar disorder and major depression. Biol Psychiatry. 2004;55:578–87. doi: 10.1016/j.biopsych.2003.11.017. [DOI] [PubMed] [Google Scholar]
- 13.Grotegerd D, Stuhrmann A, Kugel H, Schmidt S, Redlich R, Zwanzger P, et al. Amygdala excitability to subliminally presented emotional faces distinguishes unipolar and bipolar depression: An fMRI and pattern classification study. Hum Brain Mapp. 2013;35:2995–3007. doi: 10.1002/hbm.22380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Suslow T, Kugel H, Ohrmann P, Stuhrmann A, Grotegerd D, Redlich R, et al. Neural correlates of affective priming effects based on masked facial emotion: An fMRI study. Psychiatry Research: Neuroimaging. 2013;211:239–45. doi: 10.1016/j.pscychresns.2012.09.008. [DOI] [PubMed] [Google Scholar]
- 15.American Psychiatric Association. Practice Guideline for the Treatment of Patients with Major Depressive Disorder. 3rd. Washington, DC: American Psychiatric Association; 2010. [PubMed] [Google Scholar]
- 16.First M, Spitzer R, Gibbon M, Willians J, Benjamin L. Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) New York: Biometric Research Department, New York State Psychiatric Institute; 1995. [Google Scholar]
- 17.Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–98. doi: 10.1016/0022-3956(75)90026-6. [DOI] [PubMed] [Google Scholar]
- 18.Blair J, Spreen O. Predicting premorbid IQ: A revision of the National Adult Reading Test. The Clinical Neuropsychologist. 1989;3:129–36. [Google Scholar]
- 19.First M, Gibbon M, Spitzer R, Williams J, Benjamin L. Structured Clinical Interview for DSM-IV Axis II Personality Disorders, (SCID-II) Washington, D.C: American Psychiatric Press, Inc; 1997. [Google Scholar]
- 20.Hamilton MA. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56–62. doi: 10.1136/jnnp.23.1.56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Young RC, Biggs JT, Ziegler VE, Meyer DA. A rating scale for mania: Reliability, validity and sensitivity. Br J Psychiatry. 1978;133:429–35. doi: 10.1192/bjp.133.5.429. [DOI] [PubMed] [Google Scholar]
- 22.Spielberger C, Gorsuch R, Lushene R, Vagg P, Jacobs G. Manual for the State-Trait Anxiety Inventory (Form Y) Palo Alto, CA: Consulting Psychologists Press, Inc; 1983. [Google Scholar]
- 23.Tottenham N, Tanaka J, Leon A, McCarry T, Nurse M, Hare T, et al. The NimStim set of facial expressions: judgments from untrained research participants. Psychiatry Res. 2009;168:242–9. doi: 10.1016/j.psychres.2008.05.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Perlman SB, Almeida JRC, Kronhaus DM, Versace A, Labarbara EJ, Klein CR, et al. Amygdala activity and prefrontal cortex-amygdala effective connectivity to emerging emotional faces distinguish remitted and depressed mood states in bipolar disorder. Bipolar Disorders. 2012;14:162–74. doi: 10.1111/j.1399-5618.2012.00999.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Fournier J, Keener M, Mullin B, Hafeman D, LaBarbara E, Stiffler R, et al. Heterogeneity of amygdala response to happy faces in major depressive disorder: The impact of lifetime sub-threshold mania. Psychol Med. 2013;43:293–302. doi: 10.1017/S0033291712000918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Keener MT, Fournier JC, Mullin BC, Kronhaus D, Perlman SB, LaBarbara E, et al. Dissociable patterns of medial prefrontal and amygdala activity to face identity versus emotion in bipolar disorder. Psychol Med. 2012;42:1913–24. doi: 10.1017/S0033291711002935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Almeida JRC, Kronhaus DM, Sibille EL, Langenecker SA, Versace A, Labarbara EJ, et al. Abnormal left-sided orbitomedial prefrontal cortical-amygdala connectivity during happy and fear face processing: A potential neural mechanism of female MDD. Frontiers in Psychiatry. 2011;2:69. doi: 10.3389/fpsyt.2011.00069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Fournier JC, Chase HW, Almeida J, Phillips ML. Model specification and the reliability of fMRI results: Implications for longitudinal neuroimaging studies in psychiatry. PLoS ONE. 2014;9:e105169. doi: 10.1371/journal.pone.0105169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Manelis A, Ladouceur CD, Graur S, Monk K, Bonar LK, Hickey MB, et al. Altered amygdala-prefrontal response to facial emotion in offspring of parents with bipolar disorder. Brain. 2015;138:2777–90. doi: 10.1093/brain/awv176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Penny W, Henson R. Anysis of Variance. In: Friston K, Ashburner J, Kiebel S, Nichols T, Penny W, editors. Statistical Parametric Mapping: The Analysis of Functional Brain Images. New York: Academic Press; 2007. pp. 166–77. [Google Scholar]
- 31.Maldjian J, Laurienti P, Kraft R, Burdette J. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI datasets. Neuroimage. 2003;19:1233–9. doi: 10.1016/s1053-8119(03)00169-1. [DOI] [PubMed] [Google Scholar]
- 32.Frith U, Frith CD. Development and neurophysiology of mentalizing. Philosophical Transactions of the Royal Society B: Biological Sciences. 2003;358:459–73. doi: 10.1098/rstb.2002.1218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Gu X, Hof PR, Friston KJ, Fan J. Anterior insular cortex and emotional awareness. J Comp Neurol. 2013;521:3371–88. doi: 10.1002/cne.23368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Saxe R, Kanwisher N. People thinking about thinking people: The role of the temporo-parietal junction in “theory of mind”. Neuroimage. 2003;19:1835–42. doi: 10.1016/s1053-8119(03)00230-1. [DOI] [PubMed] [Google Scholar]
- 35.Schiff ND, Shah SA, Hudson AE, Nauvel T, Kalik SF, Purpura KP. Gating of attentional effort through the central thalamus. J Neurophysiol. 2013;109:1152–63. doi: 10.1152/jn.00317.2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Langner R, Eickhoff SB. Sustaining attention to simple tasks: a meta-analytic review of the neural mechanisms of vigilant attention. Psychol Bull. 2013;139:870–900. doi: 10.1037/a0030694. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Byne W, Hazlett EA, Buchsbaum MS, Kemether E. The thalamus and schizophrenia: Current status of research. Acta Neuropathol. 2009;117:347–68. doi: 10.1007/s00401-008-0404-0. [DOI] [PubMed] [Google Scholar]
- 38.Craig AD. How do you feel? Interoception: The sense of the physiological condition of the body. Nat Rev Neurosci. 2002;3:655–66. doi: 10.1038/nrn894. [DOI] [PubMed] [Google Scholar]
- 39.Harrison NA, Gray MA, Gianaros PJ, Critchley HD. The Embodiment of Emotional Feelings in the Brain. The Journal of Neuroscience. 2010;30:12878–84. doi: 10.1523/JNEUROSCI.1725-10.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Duerden EG, Arsalidou M, Lee M, Taylor MJ. Lateralization of affective processing in the insula. Neuroimage. 2013;78:159–75. doi: 10.1016/j.neuroimage.2013.04.014. [DOI] [PubMed] [Google Scholar]
- 41.Chase HW, Nusslock R, Almeida JR, Forbes EE, Labarbara EJ, Phillips ML. Dissociable patterns of abnormal frontal cortical activation during anticipation of an uncertain reward or loss in bipolar versus major depression. Bipolar disorders. 2013;15:839–54. doi: 10.1111/bdi.12132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Nusslock R, Almeida JR, Forbes EE, Versace A, Frank E, Labarbara EJ, et al. Waiting to win: Elevated striatal and orbitofrontal cortical activity during reward anticipation in euthymic bipolar disorder adults. Bipolar disorders. 2012;14:249–60. doi: 10.1111/j.1399-5618.2012.01012.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Caseras X, Lawrence NS, Murphy K, Wise RG, Phillips ML. Ventral striatum activity in response to reward: differences between bipolar I and II disorders. A J Psychiatry. 2013;170:533–41. doi: 10.1176/appi.ajp.2012.12020169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Kober H, Barrett LF, Joseph J, Bliss-Moreau E, Lindquist K, Wager TD. Functional grouping and cortical–subcortical interactions in emotion: A meta-analysis of neuroimaging studies. Neuroimage. 2008;42:998–1031. doi: 10.1016/j.neuroimage.2008.03.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Waddell S. Functional atlas of emotional faces processing: A voxel-based meta-analysis of 105 functional magnetic resonance imaging studies. J Psychiatry Neurosci. 2009;34:418–32. [PMC free article] [PubMed] [Google Scholar]
- 46.Wager TD, Barrett LF, Bliss-Moreau E, lindquist k, Duncan S, Kober H, et al. The Neuroimaging of Emotion. In: Lewis M, Haviland-Jones JM, Barrett LF, editors. Handbook of Emotion. 3rd. New York, NY: Guilford Press; 2008. pp. 249–71. [Google Scholar]
- 47.Phillips ML, Medford N, Young AW, Williams L, Williams SC, Bullmore ET, et al. Time courses of left and right amygdalar responses to fearful facial expressions. Hum Brain Mapp. 2001;12:193–202. doi: 10.1002/1097-0193(200104)12:4<193::AID-HBM1015>3.0.CO;2-A. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Goldberg H, Preminger S, Malach R. The emotion–action link? Naturalistic emotional stimuli preferentially activate the human dorsal visual stream. Neuroimage. 2014;84:254–64. doi: 10.1016/j.neuroimage.2013.08.032. [DOI] [PubMed] [Google Scholar]
- 49.Phillips ML, Travis MJ, Fagiolini A, Kupfer DJ. Medication effects in neuroimaging studies of bipolar disorder. A J Psychiatry. 2008;165:313–20. doi: 10.1176/appi.ajp.2007.07071066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Hafeman DM, Chang KD, Garrett AS, Sanders EM, Phillips ML. Effects of medication on neuroimaging findings in bipolar disorder: An updated review. Bipolar Disorders. 2012;14:375–410. doi: 10.1111/j.1399-5618.2012.01023.x. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.