Consistent with previous reports (Hariri et al., 2000
; Lieberman et al., 2005
), we found negative functional connectivity between the amygdala and VLPFC in healthy subjects during the cognitive evaluation (labeling) of affective stimuli. That is, when normal control subjects evaluate a facial expression by identifying an emotion displayed by a face and assign a verbal descriptor to that emotion, a neural network becomes activated in which the VLPFC appears to exert an inhibitory effect on amygdala. This negative functional connectivity pattern was significantly reduced in manic patients, who instead demonstrated an increased BOLD response in the amygdala and a decreased BOLD response in VLPFC during the cognitive evaluation task compared to control subjects. Only 3 known studies of bipolar mania have previously targeted prefrontal function using behavioral inhibition tasks in conjunction with fMRI and of these, all have reported decreased activation of the VLPFC in comparison with healthy subjects (Blumberg et al., 2003
; Elliott et al., 2004
; Altshuler et al., 2005b
). Our replication of decreased function in VLPFC using a different task paradigm underscores the role of this frontal region in bipolar mania and extends previous findings by linking this functional deficit with an increased responsivity of the left amygdala.
Disruptions of frontal-limbic networks are beginning to emerge in other studies of emotional disorders, including schizophrenia (Meyer-Lindenberg et al., 2001
; Das et al., 2006
), borderline personality disorder (New et al., 2007) and unipolar depression (Pezawas et al., 2005
). To our knowledge, this is the first study to examine the functional connectivity of mood regulatory networks in bipolar disorder. A disruption of the fronto-limbic circuitry in bipolar mania may have special clinical relevance; an inverse correlation was found between the current episode's severity of mania and the left VLPFC-left amygdala functional connectivity. An inverse correlation was also found between the number of prior manias and the strength of right VLPFC-left amygdala connectivity such that a greater number of prior manias was associated with a more weakened connectivity between fronto-limbic centers. Finally, a significant negative correlation was also found between illness duration and bilateral VLPFC-left amygdala functional connectivity. Future studies that examine whether each manic episode cumulatively contributes to a permanent alteration in right sided fronto-limbic emotion modulatory circuits would be of interest, as would studies that examine whether medication significantly alters fronto-limbic connectivity.
In contrast to findings from previous studies (Getz et al., 2003
; Lembke and Ketter, 2002
) bipolar subjects in our sample did not perform significantly worse on the behavioral measures of emotion perception or emotion labeling compared to healthy subjects. We believe this to be because, unlike previous studies, our facial emotion task was not difficult. The present study was conducted not to ascertain whether there are subtle perceptual differences between groups, but to show how strong emotional faces might differentially activate the amygdala. Thus, deficits in the recognition of facial emotion if present, were not discernible in our bipolar sample using the affective labeling task employed here. This is an area worthy of further exploration however, as pathological hyperactivity of the amygdala may be critical in how affective information is perceived. Both studies of healthy individuals and unipolar depressed subjects have shown that the degree of amygdala hyperactivity is directly correlated with negative biases in the perception of facial emotion (Cooney et al., 2007; Dannlowski et al., 2007
). Future paradigms that therefore employ more difficult facial recognition paradigms may allow for evaluation of correlation between deficits in facial recognition in bipolar disorder and amygdala activation.
Our finding of increased amygdala response in manic patients is not unique to the state of mania. Hyperactivity of this structure has been reported in studies of euthymic patients both in cases where the valence of perceived emotional facial expressions were positive (Lawrence et al., 2004
) and negative (Yurgelun-Todd et al., 2000
; Lawrence et al., 2004
). One recent study showed that a heightened response of the amygdala occurs even to neutral faces in a sample bipolar-euthymic youths (Rich et al., 2006
). Taken in context with other studies showing that a reduction in ventral prefrontal function (Malhi et al., 2005
; Strakowski et al., 2005a
) as well as a heightened emotional reactivity (Lovejoy et al., 1995) persists beyond manic and depressive states, it seems likely that a weakening of frontal regulation of limbic centers may indeed be a trait related abnormality that could contribute to vulnerabilities in mood.
Increased response of the amygdala in bipolar mania provides one possible explanation for the enlarged structure of the amygdala reported by our group (Altshuler et al., 1998
), and others (Brambilla et al., 2003, Strakowski et al., 1999), although not all groups have observed this (see Blumberg et al., 2003b). A reduction in modulatory/inhibitory VLPFC inputs could lead to increases in amygdala firing, and trigger dendritic arborization in this subcortical region (Vyas et al., 2002). Frontal dysfunction in turn may be caused by reductions in frontal lobe gray matter (Frangou, 2005, Lopez-Larson et al., 2002, Lyoo et al., 2004). Future studies are required however to further assess the relationship between brain structure and function, and the impact of mood state.
While manic patients in our sample demonstrated a robust increase in activation of the amygdala relative to healthy subjects in both the “perceive emotion” and “label emotion” conditions, previous findings regarding amygdala function in bipolar mania have been inconsistent (Lennox et al., 2004
; Chen et al., 2006
). It may be possible that differences in ours and other's findings are due to variations in the clinical profiles of the manic subjects included in each study, however we suspect that differences in activation are more likely the result of variations in activation paradigms (Critchley et al., 2000
). Chen et al (2006)
who scanned bipolar-manic, bipolar-depressed and healthy subjects found that subjects with mania elicited a greater activation of the amygdala compared to healthy subjects when they rated the intensity of a color shaded over an emotional face, but these same patients elicited a reduced activation of the amygdala compared to healthy subjects when they rated the emotional intensity of that face. Amygdala activations during mania are therefore likely strongly influenced both by the type of activation task employed and/or the by the level of attentional engagement during the processing of emotional stimuli (Chen et al., 2006
). Thus, future studies that further delineate the impact of task design on amygdala functioning and their interaction with mood state are needed.
Although greater VLPFC activation and negative functional connectivity with the amygdala in healthy subjects was predicted, its apparent bilaterality was surprising, given previous reports of right sided frontal connectivity (Hariri et al., 2000
; Ochsner et al., 2004
; Lieberman et al., 2005
). Lack of power may be to blame, as the current study included only nine subjects in each group, whereas studies showing right sided prefrontal-amygdala connectivity have included between 16 and 24 subjects. Functional neuroimaging studies of manic subjects however are commonly based on small samples; previous reports have included as few as 5 (Blumberg et al., 1999
), 6 (Rubinsztein et al., 2001
), 8 (Elliott et al., 2004
) and 11 manic subjects (Blumberg et al., 2003
). The logistical problem of having a manic patient remain at rest for a period of time no doubt accounts for the small numbers. In the current study, more severely manic patients were excluded from further analysis due to excessive motion.
Group differences in VLPFC-amygdala coupling were the primary focus of this study, but several other regions were differentially activated and found to functionally connected with the amygdala between groups. However, because these regions were not part of our a priori hypotheses and as such, did not survive a whole brain corrected threshold of P<0.05, an interpretation of their role in emotion processing is not included here. Future studies that observe these regions to be differentially activated between groups, perhaps in a study employing a larger number of subjects, would help to ensure the validity of these findings.
To our knowledge, this is the first study to evaluate the functional connectivity between frontal and limbic regions in bipolar disorder. Our findings should be considered in light of several limitations. First, PPI is only able to account for the functional connectivity between a seed region and regions to which activity in the seed region is directly coupled. Therefore, future studies that utilize more complex connectivity analyses, such as dynamic causal modeling (Stephan et al., 2005) or structural equation modeling (Horwitz et al., 1999) are needed to be able to account for the effect of additional brain regions in emotion regulatory circuits on amygdala activity.
Second, many of our subjects were on antimanic medications at the time of scanning, and the impact of these medications on blood flow is unknown. Medication effects may contribute to the left amygdala response in patients, but other studies of subjects with mood disorders observed both pre- and post-medication intervention have shown that pharmacological treatment either did not detectably change blood flow (Oliver et al., 1998; Theodore et al., 2000) or reduced it (Leiderman et al., 1991
; Gaillard et al., 1996
; Blumberg et al., 2005
Third, reduced activation in VLPFC in manic subjects may be the result of their attending less to the task; however, accuracy and response times suggest that all subjects were attending to the task and make this explanation of our findings unlikely. Moreover, activity in the fusiform gyrus, a region which consistently responds to viewing of faces and which can be directly modulated by attention (Pessoa et al., 2002
), was not significantly different between groups.
Fourth, more severely manic patients had to be removed from the current analysis due to motion thus limiting the range of YMRS scores in our sample. This, combined with the fact that most of our subjects were on medication, constrains the ability to draw clear conclusions from this study. However, even within the narrow YMRS range, correlations were seen between mania severity and negative connectivity.
Finally, PPI does not resolve causality, therefore it may be argued that the amygdala suppresses VLPFC output. We consider this possibility to be unlikely however, since more unilateral projections travel from VLPFC to the amygdala than vice versa (Cavada et al., 2000
) and at least a proportion of these efferents are directly inhibitory (Amaral et al., 1992
; Cavada et al., 2000
; Ghashghaei et al., 2002). Additionally, lesion or inactivation of the ventral PFC has been found to additionally facilitate affective behaviors, whereas stimulation of the PFC suppresses these behaviors (al Maskati et al., 1989; Dias et al., 1996
In summary, consistent with previous reports we found negative (i.e. modulatory) functional connectivity between the left amygdala and bilateral VLPFC in healthy subjects during the cognitive evaluation of affective stimuli. This connectivity pattern was significantly decreased in bipolar-manic patients, who additionally demonstrated increases in the response of the amygdala and decreases in the response of the VLPFC. Our finding of decreased functioning in the VLPFC using a different task paradigm from previous studies underscores a role for dysfunction in ventrolateral prefrontal brain regions in bipolar mania and extends these findings by linking this deficit with increased responsivity of the amygdala.