|Home | About | Journals | Submit | Contact Us | Français|
Pediatric bipolar disorder (BD), a highly debilitating illness, is characterized by amygdala abnormalities, i.e., volume reduction and hyperactivation during face processing. Evidence of perturbed amygdala functional connectivity with other brain regions would implicate a distributed neural circuit in the pathophysiology of BD, and would further elucidate the neural mechanisms associated with BD face emotion misinterpretation.
Thirty-three BD and 24 healthy age, gender, and IQ-matched subjects completed a functional magnetic resonance imaging (fMRI) task of face emotion identification in which attention was directed to emotional (hostility, fearfulness) and nonemotional (nose width) aspects of faces. Voxel-wise analyses examined whole brain functional connectivity with the left amygdala.
Compared to healthy subjects, BD subjects had significantly reduced connectivity between the left amygdala and two regions: right posterior cingulate/precuneus and right fusiform gyrus/parahippocampal gyrus. Deficits were evident regardless of mood state and comorbid diagnoses.
BD youth exhibit deficient connectivity between the amygdala and temporal association cortical regions previously implicated in processing facial expressions and social stimuli. In conjunction with previously documented volumetric and functional perturbations in these brain regions, dysfunction in this distributed neural circuit may begin to clarify the pathophysiology of the face emotion misperceptions and social deficits seen in BD youth.
Pediatric bipolar disorder (BD) is the focus of increasing research interest. Approximately 1% of children and adolescents are diagnosed with BD (Johnson et al., 2000), a highly debilitating psychiatric illness. One study found that nearly 65% of BD youth had attempted suicide and 75% had been psychiatrically hospitalized (Dickstein et al., 2005b). Typically, BD youth have deficient social skills and few friendships (Geller et al., 2000; Lewinsohn et al., 2003).
Recent studies document impaired face-emotion labeling in BD youth, a deficit which has been previously associated with social impairment (De Sonneville et al., 2002). Specifically, BD youth misclassify emotional facial expressions (Guyer et al., 2007; McClure et al., 2005), and misinterpret neutral faces as being significantly more threatening than do controls (Rich et al., 2006). In the latter study, negative misinterpretations by BD youth were associated with hyperactivation of the left amygdala: increasingly negative misperceptions of neutral expressions were correlated with increased amygdala activation. The amygdala mediates emotional processing (LeDoux, 2000), and studies document amygdala volumetric deficits in BD youth (Blumberg et al., 2005; Chang et al., 2005; Chen et al., 2004; Delbello et al., 2004; Dickstein et al., 2005a). Thus, affective neuroscience research using structural and functional magnetic resonance imaging (MRI) implicates the amygdala in pediatric BD.
However, brain regions are not isolated entities but instead form highly interconnected neural circuits. Therefore, it is important to not only identify brain regions that are dysfunctional in pediatric BD, but also to ascertain whether the connections among nodes of neural circuits differ between BD patients and controls. We examined this with functional connectivity, which provides a measure of the temporal correlation between neurophysiological measurements from different brain regions (Friston et al., 1993; Friston et al., 1997). While this technique cannot describe signal direction nor account for intermediate or multiple connections, it can elucidate the extent to which the activity in one brain area is associated with the activity in another brain area.
The current study compares BD youth to healthy controls on neural functional connectivity during a face processing task. The left amygdala served as our reference area based on our prior study which identified amygdala hyperactivation when BD youth misinterpreted face expressions (Rich et al., 2006). Because this was the first study of functional connectivity in BD youth, we viewed this study as exploratory. However, while we mapped amygdala connectivity across the entire brain, we were particularly interested in the temporal association cortex, given its role in face-processing (Critchley et al., 2000; Haxby et al., 2000).
BD (N = 33) and control (N = 24) subjects were recruited as described previously (Rich et al., 2006). The National Institute of Mental Health (NIMH) IRB approved this study. Parents and children gave written informed consent/assent. BD inclusion criteria required subjects ages 7–18 to meet DSM-IV (American Psychiatric Association, 1994) criteria for BD, with the strict requirements of a history of at least one full duration hypomanic or manic episode – i.e., lasting ≥4 days for hypomania or ≥7 days for mania – with abnormally elevated or expansive mood and/or grandiosity, and at least three DSM-IV criterion ‘B’ mania symptoms (Leibenluft et al., 2003). Diagnoses were made by master’s-level clinicians using the Kiddie-Schedule for Affective Disorders (K-SADS-PL; Kaufman et al., 1997). Control subjects and a parent also completed the K-SADS-PL to ensure that the subject had no psychiatric history. The control sample included three pairs of siblings. As a result, in secondary analyses we randomly removed a sibling from each pair to determine if this altered our results (see Results below). To evaluate current mood, clinicians administered the Children’s Depression Rating Scale (CDRS; Poznanski et al., 1984) and Young Mania Rating Scale (YMRS; Young et al., 1978) to patients and their parents within 24 hours of scanning. General functioning was measured using the Children’s Global Assessment Scale (CGAS; Shaffer et al., 1983).
Exclusion criteria included IQ <70, pervasive developmental disorder, psychosis that interfered with study compliance, unstable medical illness, and substance abuse within three months.
As described elsewhere (Rich et al., 2006), the faces fMRI task consisted of participants viewing four face types (happy, angry, fearful, and neutral) selected from standardized sets (Ekman & Friesen, 1976; Gur et al., 2001). Subjects used a button box (MRI Devices, Waukesha, WI) to conduct one of three ratings of the face on a 5-point scale: 1) the threat-level of the face (‘How hostile is the face?’); 2) their fearful response to the face (‘How afraid are you?’); or 3) a non-emotional facial feature (‘How wide is the nose?’). A fourth passive face viewing condition was also included.
The 160-trial run was divided into blocks comprised of 10 trials: eight faces (two of each face type) and 2 fixations. At the beginning of each block, subjects were presented with a screen instructing them to conduct one of the three rating types or perform passive viewing throughout that block. Face stimuli consisted of 32 different actors, with each actor presenting just one face emotion. Order of face type/fixation were randomized within each block, and order of block type (e.g., ratings type/passive viewing) and actor-emotion pairings were randomized across participants. For each block, instructions were presented for 3000 ms and each face or fixation was displayed for 4000 ms, followed by a 750–1250 ms inter-trial interval. We used the rapid event related paradigm of Friston et al. (1998) and Zarahn and Slifstein (2001), which allowed us to analyze trials by face and ratings type. Behavioral data consisted of face ratings and the reaction time (RT) required to make these ratings.
Whole-brain fMRI data were acquired on a General Electric Signa 3T scanner. Following localization and shimming, T2* weighted images were acquired using echo-planar single-shot gradient echo imaging (64 × 64 matrix, TR = 2000 ms, TE = 40 ms, FOV = 240 mm, 3.75 × 3.75 × 5 mm voxels, 23 contiguous 5 mm axial slices parallel to the AC-PC line). High-resolution T1 weighted anatomical images were acquired to aid with spatial normalization (180 1 mm sagittal slices, FOV = 256, NEX = 1, TR = 11.4 ms, TE = 4.4 ms, matrix = 256 × 256, TI = 300 ms, bandwidth = 130 Hz/pixel, 22 kHz/256 pixels).
Analyses were conducted with Statistical Parametric Mapping (SPM) software (SPM99, Wellcome Department of Neurology, London, UK). Data were corrected for slice timing and motion, co-registered to the anatomical data, spatially normalized, and re-sliced into isotropic 2 mm voxels. The data were spatially smoothed with a Gaussian kernel (FWHM = 8). See Rich et al. (2006) for a more detailed description of the initial data analysis procedures.
For each subject, functional connectivity was measured by examining the temporal correlation of the signal between the left amygdala and the signal from each of the other voxels in the brain (Friston et al., 1997; Greicius et al., 2003). This signal was collected for the entire duration of the experiment, i.e., across all face stimuli and ratings. At each voxel in the brain, the time series was normalized by the grand mean (the mean over all voxels in the brain and all time points), detrended and then bandpass filtered (frequency cutoffs at .0083 and .15 Hz). A general linear model was applied to this data using SPM. The user-specified single regressor of interest was the amygdala time-series; specifically, we extracted the time-series from the left amygdala by taking the mean over all amygdalar voxels at each time point in the temporally filtered time series. The global signal and six motion parameters were treated as regressors of no interest. The regression coefficient corresponding to the regressor of interest was entered into a second, group-level analysis. The threshold for significance was p < .0001 uncorrected. We transformed Montreal Neurological Institute (MNI) coordinates in SPM to Talairach coordinates and then labeled the regions using Analyses of Functional Neuroimages (AFNI).
For functional connectivity analyses of fMRI data, we used t-tests generated from the group-level random effects model. These tests examined amygdala connectivity within each sample and group differences between BD subjects and controls. To minimize Type I errors, the Greenhouse–Geisser procedure was applied when appropriate. Data were analyzed using SPSS version 14.0 (SPSS, 2005).
BD patients (N = 33) and controls (N = 24) did not differ on age (BD = 14.4 ± 3.0 yrs; control = 14.4 ± 2.2 yrs), sex (male: BD = 39.4%; control = 37.5%), or IQ (BD = 108.6 ± 10.9; control = 110.6 ± 12.7) as measured with the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999). There was a significantly greater proportion of Caucasians in the BD sample (X2 = 11.01, p = .01). Demographic similarity was maintained when we randomly removed one sibling from each of the three sibling pairs. Criteria for Bipolar I were met by 90.9% (N = 30) of BD patients; 75.8% (N = 25) had at least one comorbid diagnosis (mean number of diagnoses = 2.7 ± 1.4); 81.8% (N = 27) of patients were medicated when scanned (mean = 2.5 ± 1.7 medications per subject).
CDRS and YMRS scores showed that 54.5% (N = 18) of BD patients were euthymic at the time of scanning, 33.3% (N = 11) were hypomanic (YMRS > 12 but <26, CDRS < 40), 9.1% (N = 3) were depressed (CDRS > 40, YMRS <12), and 3.0% (N = 1) were mixed (CDRS > 40 and YMRS > 12). Patients’ CGAS score (54.0 ± 14.0) indicated moderate impairment.
Relative to healthy subjects, BD youth exhibited significantly lower functional connectivity between the left amygdala and two regions: 1) a region bordering the right posterior cingulate and precuneus (24, −66, 20) [t(56) = 4.18, p < .0001 uncorrected] (Figure 1 and Table 2), and 2) a region bordering the right fusiform gyrus (FG) and parahippocampal gyrus (PHG) (24, −38, −14) [t(56) = 4.01, p < .0001 uncorrected] (Figure 2 and Table 2).
When we randomly removed one sibling from each of the three sibling pairs and used peak voxel data to compare the groups, controls continued to have significantly greater connectivity than BD youth in the posterior cingulate/precuneus [t(53) = 6.86, p < .0001] and the FG/PHG [t(53) = 7.03, p < .01].
Connectivity was not correlated with BD patients’ mood during testing, as measured by CDRS or YMRS scores. In addition, an ANOVA examined functional connectivity between euthymic BD subjects (N = 18), noneuthymic BD subjects (N = 15), and controls. There were group differences at both the posterior cingulate/precuneus [F(2,54) = 13.75, p < .001] and FG/PHG [F(2,54) = 3.95, p = .03]. Specifically, both euthymic and noneuthymic BD youth had significantly lower connectivity than controls between the amygdala and both neural regions [posterior cingulate/precuneus: euthymic BD subjects vs. controls (p < .001), noneuthymic BD subjects vs. controls (p = .04); FG/PHG: euthymic BD subjects vs. controls (p = .01), noneuthymic BD subjects vs. controls (p = .05)]. Thus, weaker neural connectivity in BD youth was not related to mood state.
Regarding comorbid diagnoses, separate ANOVAs comparing controls to BD subjects with vs. without comorbid attention deficit hyperactivity disorder (ADHD) (with: N = 16; without: N = 17), oppositional defiant disorder (ODD) (with: N = 8; without: N = 25), and anxiety disorders (with: N = 13; without: N = 20) found that BD youth without these comorbid disorders had significantly lower connectivity than controls at both the posterior cingulate/precuneus [ADHD: F(2,54) = 9.23, p < .001; BD without ADHD vs. controls p = .003; ODD: F(2,54) = 11.04, p < .001, BD without ODD vs. controls p = .001; anxiety: F(2,54) = 10.20, p < .001; BD without anxiety vs. controls p < .001)] and FG/PHG regions [ADHD: F(2,54) = 5.62, p = .006; BD without ADHD vs. controls p = .001; ODD: F(2,54) = 4.20, p = .02, BD without ODD vs. controls p = .005; anxiety: F(2,54) = 4.60, p = .02; BD without anxiety vs. controls p = .01)] Thus, connectivity deficits were evident in BD youth without ADHD, ODD, or anxiety.
Examining the impact of medication, there were too few unmedicated BD subjects (N = 6) to compare this sample to controls. Bivariate correlations did not find a significant relationship between number of medications and connectivity in BD youth [posterior cingulate/precuneus (r = .24, p = .17); FG/PHG (r = .15, p = .42)]. We then conducted ANOVA comparisons of controls to those BD subjects on vs. off specific classes of medications (i.e., antipsychotics, mood stabilizers, antidepressants, anticonvulsants, stimulants, sedatives, lithium; none mutually exclusive). When each medication class was examined separately (e.g., BD youth taking lithium and possibly other medications vs. BD youth not taking lithium but possibly taking other medications), there was significantly diminished connectivity between the amygdala and posterior cingulate/precuneus and FG/PHG in BD youth not receiving that medication class vs. controls (all p’s < .02). Thus, lower connectivity in BD youth does not appear to be driven by the presence of a particular medication.
Prior behavioral and fMRI studies demonstrate face processing deficits in pediatric bipolar disorder (BD; McClure et al., 2005; Rich et al., 2006). Specifically, compared to healthy subjects, BD youth: 1) view neutral faces as more threatening; and 2) have increased left amygdala activity when rating their fear of, or the hostility on, neutral faces (Rich et al., 2006). To expand on these findings, we compared functional connectivity (Friston et al., 1997) between the left amygdala and other brain regions in BD youth versus healthy controls when processing facial expressions. Functional connectivity provides a measure of the temporal correlation of activity between different brain regions. We found that BD youth have significantly less functional connectivity between the left amygdala and a network involving a region bordering the right posterior cingulate/precuneus, and another region bordering the right fusiform gyrus (FG)/parahippocampal gyrus (PHG).
Our results indicate that BD youth display impaired signal communication in neural networks critical to processing faces and emotional stimuli. Both the FG and posterior cingulate are associated with the visual processing of facial expressions. Specifically, the FG is specialized in processing face identity and, to a lesser extent, face expression (Haxby et al., 2000; Kanwisher et al., 1997), while the posterior cingulate responds to threatening stimuli that are of immediate social importance (Critchley et al., 2000). Further, the PHG has been implicated in emotional and cognitive learning, specifically associating stimuli with affective meaning (Fanselow & LeDoux, 1999; Yaniv et al., 2000). Face emotion classification is thought to reflect, in part, the amygdala-FG-posterior cingulate network (Haxby et al., 2002), while the amygdala receives input from the PHG (Shi & Cassell, 1999). Thus, disruption of these circuits may explain partly the face emotion misperception seen in BD youth.
The previously documented amygdala hyperactivity in BD youth (Rich et al., 2006) may reflect the end result of deficient signal transmission from a network involving the FG, posterior cingulate, or PHG to the amygdala. The amygdala regulates the attribution of emotional significance to threatening faces (Calder et al., 2001). Thus, if the amygdala fails to receive from the FG, posterior cingulate, or PHG information regarding face expression or stimulus affective meaning, BD youth may aberrantly perceive neutral faces as threatening. Conversely, information is transmitted from the amygdala to regions of the temporal association cortex (LeDoux, 2000). Impaired connectivity in BD youth may mean that such information is transmitted incompletely to regions such as the FG or posterior cingulate, possibly resulting in face emotion misperceptions.
In addition to elucidating possible neural correlates of face misperception in pediatric BD, our results may help to explain the social deficits seen in these youth (Geller et al., 2000; Lewinsohn et al., 2003). The coordinated function of the amygdala-FG-posterior cingulate circuit mediates not only face emotion processing, as already noted, but also a diverse set of social functions, including interpretation of social signals and regulation of emotional and behavioral responses to these signals (Amaral et al., 1992; Critchley et al., 2000; Haxby et al., 2002). Further, the PHG is thought to be the ‘gateway’ for cortical sensory input to the amygdala before this information is sent to neural regions responsible for regulating behavior (Kolb, 1990). Thus, disruption of these circuits, as tentatively documented here, is likely to impair social signal processing and emotional and behavioral regulation, which in turn is likely to result in the social deficits that characterize BD youth.
The involvement of the posterior cingulate, precuneus, and PHG also indicate that memory may be a factor in the face misperception and social deficits seen in BD youth. Memory impairments are thought to be a trait marker of adult and pediatric BD (Martinez-Aran et al., 2000; Pavuluri et al., 2006b), and prior research finds diminished recall of face expressions by BD youth (Dickstein et al., in press). The posterior cingulate (Maddock et al., 2003) and precuneus (Cavanna & Trimble, 2006) play primary roles in retrieving prior experiences and regulating the impact of emotional salience on memory, and the amygdale–PHG circuit has been implicated in the formation of aversive emotional memories (Fanselow & LeDoux, 1999; Yaniv et al., 2000). Impaired connectivity between the amygdala and the posterior cingulate, FG, and PHG may lead to inadequate recall of prior social encounters which would otherwise serve to shape BD youth’s face processing and social function.
This study broadens our understanding of the pathophysiology of pediatric BD. Our results expand upon prior research which has documented in BD youth volumetric deficits and functional aberrations in the FG (Frazier et al., 2005a; Pavuluri et al., 2006a), posterior cingulate (Chang et al., 2004; Kaur et al., 2005), and amygdala (Blumberg et al., 2003; Blumberg et al., 2005; Chang et al., 2005; Chen et al., 2004; Delbello et al., 2004; Dickstein et al., 2005a; Rich et al., 2006). Thus, neuroimaging research is consistent in implicating volumetric, functional, and now connectivity impairments of the amygdala, FG, and posterior cingulate in the pathophysiology of pediatric BD.
Although memory deficits are thought to be trait markers of BD, as noted above, there is heterogeneity in studies investigating potential neural correlates. We are not aware of functional nor structural studies reporting perturbations in the precuneus or PHG in pediatric BD. In BD adults, memory tasks have evoked both heightened (Deckersbach et al., 2006) and diminished (Malhi et al., 2007) PHG activation, as well as hypoactivation of the precuneus (Malhi et al., 2007). If discussion of the neural correlates of memory is expanded to include the hippocampal formation, results of structural studies are inconsistent in both BD youth (Blumberg et al., 2003; Chang et al., 2005; Chen et al., 2004; Dickstein et al., 2005a; Frazier et al., 2005b) and adults (Altshuler et al., 2000; Hauser et al., 2000; Strakowski et al., 1999; Strasser et al., 2005; Yucel et al., 2007). Hippocampal hyperactivation has been reported in BD adults during a memory task involving word lists (Deckersbach et al., 2006), but it is unclear how this relates to a visual task such as face perception. Further, there are no reports of hippocampal functional perturbations in pediatric BD. Clearly, additional research is required to better understand the role of the precuneus and PHG, or even more generally the hippocampal formation, in BD.
A primary limitation of our study is the clinical diversity of our BD sample, though this is typical of that seen in other samples of BD youth (Birmaher et al., 2006). While there were high rates of comorbidity in our BD patients, the fact that BD youth without comorbid ADHD, ODD, or anxiety had significantly lower connectivity than controls indicates that the deficit may reflect BD itself rather than the presence of a co-occurring disorder.
Another possible confounding variable is mood state, given that in BD adults face emotional identification and its neural correlates may differ between euthymic (Bozikas et al., 2006; Getz et al., 2003; Harmer et al., 2002; Yurgelun-Todd et al., 2000) and manic (Lembke & Ketter, 2002; Lennox et al., 2004) subjects. However, we found that both euthymic and noneuthymic BD youth had significantly lower functional connectivity than controls, indicating that these neural deficits may be trait-based and not mood-state dependent.
Medication status is an additional potential confound. Since most BD patients were medicated, we were unable to fully determine the extent to which our results were associated with BD or, instead, with medication effects. However, there was no relationship between number of medications and connectivity, and lower connectivity in BD youth did not seem to be due to the presence of a particular class of medication. The results of all post hoc analyses should be interpreted with caution because they are subject to Type I error due to multiple comparisons and Type II error due to small sample sizes.
Finally, it is important to note that an alternate explanation for our results is differences in measurement error between the groups, which would impact the estimated regression coefficient magnitude via attenuation bias (van Belle & Fisher, 1996). Further, it is possible that a hypothesis- driven study may fail to confirm the preliminary results presented here.
In sum, our results in BD youth suggest deficient functional connectivity among neural regions mediating the evaluation of facial emotional expressions. When we examined connectivity between the left amygdala and the entire brain, exploratory data implicate regions primarily associated with processing face expressions (fusiform gyrus), social information (the posterior cingulate/precuneus), and emotional learning and memory (parahippocampal gyrus, precuneus). Given that this is the first study of functional connectivity in BD youth, our results should be viewed as preliminary. The use of samples with different clinical characteristics, or alternate methodology parameters and cognitive-emotional tasks, may identify other networks with impaired connectivity in BD youth. Since our technique measures connectivity strength but not direction, future studies using effective connectivity methods will enable exploration of the directionality of neural circuitry and provide a better understanding of dysfunction in the limbic-temporal association cortex circuit in BD youth.
This research was supported by the Intramural Research Program of the NIH, NIMH. We gratefully acknowledge the children and families of patients and controls. We also acknowledge the staff of the Bipolar Spectrum Disorders Unit at the NIMH.
Conflict of interest statement: No conflicts declared.