The functional neuroanatomy of bipolar disorder: a review of neuroimaging findings
The authors review existing structural and functional neuroimaging studies of patients with bipolar disorder and discuss how these investigations enhance our understanding of the neurophysiology of this illness. Findings from structural magnetic resonance imaging (MRI) studies suggest that some abnormalities, such as those in prefrontal cortical areas (SGPFC), striatum and amygdala exist early in the course of illness and, therefore, potentially, predate illness onset. In contrast, other abnormalities, such as those found in the cerebellar vermis, lateral ventricles and other prefrontal regions (eg, left inferior), appear to develop with repeated affective episodes, and may represent the effects of illness progression and associated factors.
Magnetic resonance spectroscopy investigations have revealed abnormalities of membrane and second messenger metabolism, as well as bioenergetics, in striatum and prefrontal cortex. Functional imaging studies report activation differences between bipolar and healthy controls in these same anterior limibic regions. Together, these studies support a model of bipolar disorder that involves dysfunction within subcortical (striatal–thalamic)–prefrontal networks and the associated limbic modulating regions (amygdala, midline cerebellum). These studies suggest that, in bipolar disorder, there may be diminished prefrontal modulation of subcortical and medial temporal structures within the anterior limbic network (eg, amygdala, anterior striatum and thalamus) that results in dysregulation of mood. Future prospective and longitudinal studies focusing on these specific relationships are necessary to clarify the functional neuroanatomy of bipolar disorder.
MRI, MRS, fMRI, neuroimaging, neurophysiology
Bipolar disorder is one of the most common and disabling conditions affecting humankind. Prevalence estimates suggest that 1.5–3.0% of the population will develop bipolar disorder,1,2 which is the sixth leading cause of disability worldwide.3 Despite being a common and important psychiatric illness, the specific neurophysiologic basis of bipolar disorder is unknown.
However, in the past 15 years or so, refinement of neuroimaging techniques, particularly magnetic resonance imaging (MRI), positron emission tomography (PET), and more recently magnetic resonance spectroscopy (MRS) and functional MRI (fMRI), have produced a proliferation of studies that have attempted to clarify the neural substrates of bipolar disorder. The major symptoms of bipolar disorder, namely affective instability, neurovegetative abnormalities, impulsivity and psychosis, suggest that anterior limbic brain networks controlling these behaviors are dysfunctional. These networks consist of traditional limbic structures such as the amygdala modulating well-recognized iterative prefrontal–striatal–thalamic circuits that control complex socioemotional behaviors. One would therefore predict that abnormalities would occur within these brain networks in patients with bipolar disorder. In this review, we will update our previous reviews of this topic4,5,6 and synthesize information in order to inform hypotheses of the functional neuroanatomy of bipolar disorder.
Top of page
In the past decade, MRI has built upon older computed X-ray tomography (CT) studies to provide detailed in vivo investigations of the neuroanatomy of bipolar disorder. Morphometric neuroimaging provides a means to identify specific neuroanatomic abnormalities that may differentiate patients with bipolar disorder from healthy subjects and people with other psychiatric disorders. Although structural measurements may not have clear functional correlates, careful description of structural abnormalities in bipolar disorder may define a neuroanatomic substrate to guide neurophysiologic studies (eg, fMRI). Notably, overall brain volumes appear to be normal in bipolar disorder, as few studies have found global decreases in gray or white matter.4,5,6,7 However, regional differences have been observed in prefrontal cortex, and subcortical and medial temporal structures, which are all components of anterior limbic networks that modulate the behaviors affected in bipolar disorder.
Studies of prefrontal cortex have often defined regions of interest that lack any specific functional meaning (eg, all matter anterior to the genu of the corpus callosum). This approach ignores the inherent complexity of this brain area. In fact, the prefrontal cortex consists of several histologically and functionally discrete brain regions that are not well delineated at the level of anatomic resolution available with current imaging methods (eg, 1 mm3). Therefore, these distinct prefrontal subregions are difficult to demarcate from each other using structural imaging.
Consequently, most imaging studies of prefrontal cortex in bipolar disorder have examined large anterior brain regions and have typically not observed differences between bipolar and healthy subjects.4,5,6 A study by Sax et al8 is one informative exception to this. The investigators found decreased prefrontal volumes in bipolar patients compared with healthy subjects, and, in the patients, prefrontal cortical volume inversely correlated with performance on a measure of attention (CPT). By measuring a behavioral correlate associated with bipolar disorder (namely, inattention), the validity of this finding was strengthened. However, this finding has not been replicated.
An alternative to studying large, indistinct regions of prefrontal cortex is instead to delineate specific prefrontal subregions based on neuroanatomic markers. Although these markers do not guarantee that a region is functionally homogeneous, they do presumably identify prefrontal regions that are less functionally heterogeneous than those defined more grossly. Lopez-Larson et al9 divided the prefrontal cortex into five subregions (superior, middle, inferior, orbital and cingulate) based upon anatomic landmarks that roughly correspond to known functionally discrete brain regions. The investigators also separately examined both gray and white matter compartments of these subregions. They found that bipolar patients exhibited no differences from healthy subjects in prefrontal volumes overall, but the patients did have smaller gray matter volumes in the left superior and middle and right prefrontal subregions. Longer duration of affective illness was associated with smaller left inferior prefrontal gray volumes. This finding is consistent with work by Brambilla et al,10 who found that increased age in middle-aged bipolar patients was significantly associated with smaller gray matter volumes, which was not observed in healthy subjects. Therefore, patients with bipolar disorder may show increased gray matter loss as they age as a consequence of recurrent affective episodes. These results highlight the importance of identifying discrete prefrontal subregions, as subtle differences may not be evidenced in larger, less discrete prefrontal measurements. Additionally, correlations with clinical measures suggest that illness course may influence anatomical measurements.
Drevets et al11 focused on an even more discretely defined prefrontal region as they studied the subgenual prefrontal cortex (SGPFC). The SGPFC is the portion of the anterior cingulate inferior to the genu of the corpus callosum. The SGPFC appears to modulate human mood states11 and because of its connection within the anterior cingulate and with other anterior limbic brain regions, helps to integrate cognitive and emotional information. The investigators found that patients with bipolar disorder and a family history of affective illness exhibited smaller left SGPFC volumes than healthy subjects. Hirayasu et al12 replicated this finding in patients with first-episode psychotic affective (bipolar and unipolar) illness. Specifically, they found that these patients had smaller left SGPFC volumes than healthy subjects, patients with schizophrenia, and patients with affective illness without a family history of affective illness. However, Brambilla et al13 failed to find decreased SGPFC volumes in familial affective disorder, most likely reflecting differences in patient sampling from the previous studies. Nonetheless, these studies suggest that specific SGPFC abnormalities may occur in patients with familial affective disorders.
Together, structural imaging studies suggest that subregions of the prefrontal cortex may be structurally different in bipolar as compared to healthy individuals. There have been few studies comparing bipolar disorder with other mental illnesses, so the specificity of these findings is uncertain. Nonetheless, the presence of subtle neuroanatomic abnormalities in prefrontal cortex is consistent with recent reports of prefrontal histological abnormalities in bipolar patients. Specifically, Rajkowska et al found decreased glial density in prefrontal regions of bipolar patients,14 although also observed similar abnormalities in unipolar patients.15 Decreased neuron density and glial enlargement were also present in dorsolateral and orbital prefrontal cortical regions. Öngür et al16 examined the SGPFC in two sets of brain specimens from patients with affective disorders (unipolar and bipolar combined). The number of glial cells was reduced, particularly in cases with a family history of affective disorder. In contrast, this abnormality was not found in somatosensory cortex, suggesting regional specificity within the brain. These histological findings may underlie the abnormalities observed with structural imaging, although this needs to be verified directly.
Distinct regions of the prefrontal cortex project to corresponding regions in the striatum to initiate the prefrontal–striatal–thalamic loops that modulate human emotional, cognitive and social behavior; as noted, these are the behaviors that are abnormal in bipolar disorder. These networks appear to iteratively sample incoming information to modulate human behavior. Therefore, a number of investigators have examined striatum in bipolar patients and have reported increased striatal size in patients with bipolar disorder compared with healthy subjects,17,18,19,20 although this has not been a universal finding.6,21 Aylward et al17 first reported caudate enlargement, which they observed in male bipolar patients compared to healthy subjects. Subsequently, Strakowski et al18 examined multiple structures within the anterior limbic network and observed enlargement of the striatum in men and women with bipolar disorder. Recently, these investigators19 extended this finding by comparing first- and multiple-episode bipolar patients and healthy subjects. The goal of this study was to determine whether striatal enlargement (and other periventricular changes) occurred early in the course of bipolar illness or instead was a consequence of illness progression. Putamen volumes were enlarged in both patient groups compared with healthy subjects, but the patient groups did not differ, indicating that striatal enlargement is not a result of illness chronicity.19 DelBello et al22 also observed putamen enlargement in a study of adolescent bipolar patients, further supporting the notion that this abnormality occurs early in the course of illness. Consistent with these findings, Noga et al20 found caudate enlargement in both affected and unaffected monozygotic twins discordant for bipolar disorder, as compared to the healthy subjects. These results suggest that striatal enlargement may be a heritable vulnerability factor for developing bipolar disorder, which may be necessary but is not sufficient to cause the condition. Additional studies in at-risk subjects (eg those with bipolar parents who do not express the illness) followed longitudinally would best address the relationships between striatal structural abnormalities and illness onset.
The thalamus is the next component of the prefrontal–striatal–thalamic networks that has been studied in bipolar disorder. Dupont et al23 and Strakowski et al18 both observed increased thalamic volumes in bipolar compared with healthy subjects. However, this finding has not been observed in a number of other studies.6,8,19,24
Medial temporal structures
Amygdala and hippocampus project into prefrontal–striatal–thalamic networks and are involved in the modulation of emotional and cognitive behavior. Hippocampal enlargement has been reported in one study of bipolar patients, but this finding has not been replicated. Partial volume effects may have contributed to this finding since relatively thick image slices of 1 cm were used in this study. In the same study, no differences in amygdala volumes between bipolar and healthy subjects were identified.25 One study reported smaller left amygdala volumes in bipolar subjects compared with healthy controls, however, measurements were taken from only two slices.26 In contrast, amygdala enlargement in bipolar patients compared to healthy subjects has now been reported in three recent, independent studies from three different research groups.18,27,28 Amygdala enlargement has also been reported in bipolar patients compared to patients with schizophrenia.27 In these same reports, hippocampal volumes in bipolar patients did not differ from healthy subjects, although the patients with schizophrenia exhibited smaller hippocampal volumes. The finding of normal hippocampal and increased amygdala volumes, therefore, may be a relatively specific neuroanatomic abnormality in bipolar patients, as studies of patients with schizophrenia report decreased hippocampal and amygdala size.29 However, a recent study reported that right hippocampal volume was smaller in affected monozygotic bipolar twins as compared to the nonaffected cotwins.20 In contrast to amygdala enlargement found in adults, DelBello et al22 reported decreased amygdala volumes in adolescents with bipolar disorder. These findings suggest that amygdala enlargement may therefore be a consequence of abnormal development of this structure during adolescence and early adulthood. However, in the absence of prospective studies, this suggestion is speculative. Nevertheless, since the amygdala is central to emotional regulation, which is abnormal in bipolar patients, there is considerable face validity to the notion that amygdala development is abnormal in bipolar disorder. However, none of the investigations have identified specific clinical, functional or histological correlates of amygdala enlargement to further clarify the neurophysiological meaning of this finding. Additional work identifying these correlates is a critical next step to extend previous findings.
Although historically considered to be involved only in motor control, more recently investigators have recognized that the midline cerebellum (ie, cerebellar vermis) is strongly interconnected with limbic brain regions.30 Interestingly, several older CT studies noted decreased cerebellar volumes in bipolar disorder,6 but these were not replicated or extended for many years and were limited in sample selection and imaging methods. DelBello et al31 revisited the potential role of the cerebellum in bipolar disorder by measuring the size of the cerebellar vermis in bipolar patients with multiple prior affective episodes compared with first-episode patients and healthy subjects. The multiple-episode patients exhibited decreased vermal size (particularly in area III) compared with first-episode patients and healthy subjects. These investigators have recently replicated this finding and extended it to include vermal area II (Mills et al, under review). Vermal size appears to be associated with the number of prior affective episodes, as increasing episodes predict a smaller vermis. Therefore, this finding may represent a consequence of recurrent mood episodes.
Lateral ventriculomegaly is a well-recognized finding in bipolar disorder.6 Since ventricles are a space rather than a structure, decreased tissue volumes in periventricular brain regions should accompany ventricular enlargement, although this has been minimally studied. Moreover, it has been unclear whether ventricular enlargement is present at the onset of bipolar disorder or instead develops over time during the course of illness. The former suggests the presence of brain underdevelopment, whereas the latter suggests tissue loss. To investigate this, Strakowski et al19 compared ventricular volumes among first- and multiple-episode bipolar patients, along with healthy subjects. Additionally, they examined whether differences in volumes of periventricular structures (eg, striatum, hippocampus) explained differences in ventricular size. They found that multiple-episode patients exhibited significantly greater ventricular volumes than first-episode patients and healthy subjects; the latter two groups did not differ. The multiple-episode group did not exhibit decreased volumes in periventricular structures to explain the larger ventricles. None of the first-episode patients exhibited lateral ventricles larger than the mean (or median) of the multiple-episode group. Additionally, ventricular size was strongly correlated with the number of previous manic episodes. Brambilla et al32 also found that ventricular volume was associated with the number of prior affective episodes in bipolar patients. These results suggest that ventriculomegaly occurs in bipolar patients due to ventricular expansion as a consequence of repeated affective episodes. The specific mechanism by which this occurs does not appear to be a loss of gray matter in periventricular structures. A next step would be to determine whether white matter loss occurs over time and is reflected in ventricular expansion. Although this study has not been done, Strakowski et al33 observed an increased gray/white matter ratio in bipolar patients (particularly women), consistent with this notion that decreased white matter might explain the ventriculomegaly of bipolar disorder. Additionally, increased rates of white matter abnormalities, in the form of T2-signal hyperintensities, in bipolar patients are commonly reported.34 Finally, Adler et al35 observed abnormalities in frontal white matter tracts in bipolar disorder, assessed using diffusion tensor imaging. This study suggested disruption of prefrontal white matter bundle coherence, which may underlie the ventriculomegaly observed in other studies.
The symptoms of bipolar disorder suggest dysfunction in anterior limbic networks that subsume these behavioral functions. Neuroanatomic studies have therefore focused on components of these networks that include prefrontal–striatal–thalamic circuits and closely interconnected areas such as the amygdala and midline cerebellum. These studies reported neuroanatomic abnormalities in these brain networks that included prefrontal subregional decreased volumes, amygdala and striatal enlargement, and midline cerebellar atrophy. Some of these abnormalities, specifically decreased gray matter in left inferior prefrontal cortex, decreased size of cerebellar vermis and ventriculomegaly have been associated with the number of affective episodes, suggesting that brain changes occur in response to illness course. The mechanisms underlying these changes are unknown, but warrant further investigation as these changes may represent increasing vulnerability either to additional affective episodes or to disability. Amygdala and striatal enlargement have not been associated with clinical course and appear to be present at illness onset. Moreover, the dyad of increased amygdala size coupled with normal hippocampal volume appears to be relatively specific for bipolar disorder. More comparisons between bipolar and other patient groups are needed to verify this supposition, but if present it could identify a specific vulnerability for bipolar illness. Together these neuroanatomic findings provide a substrate to inform functional and spectroscopic imaging studies, as well as histological investigations, in order to clarify the neuropathophysiology of bipolar disorder.
Top of page
Structural imaging studies are useful for identifying possible neuroanatomic substrates for bipolar disorder, but they do not provide direct measures of neurophysiological abnormalities. One approach to extend these studies is through magnetic resonance spectroscopy (MRS), which allows in vivo study of brain chemistry. By adjusting imaging parameters, different types of neurochemicals can be studied. Perhaps, the most commonly used spectroscopic approach is proton (1H) MRS, which can be used to measure chemical concentrations for a variety of substances including N-acetyl aspartate (NAA), creatine (Cr), choline-containing compounds (Cho) and myo-inositol (Ino). Additionally, certain amino-acid neurotransmitters such as GABA, glutamate and glutamine may be identifiable. NAA may be a marker of functional and structural neuronal integrity and has been found to be reduced in neurodegenerative disorders.36 The Cr peak contains both creatine and phosphocreatine, which are markers of cellular energy. The Cho peak contains phosphocholine, glycerophosphocholine, phosphatidylcholine, sphingomyeline, choline and acetylcholine and may be a marker of cell membrane integrity. The Ino peak contains myo-inositol, which is of particular interest for bipolar disorder, since lithium strongly inhibits inositol-monophosphatase and via the phosphoinositide cycle this results in a decrease of myo-inositol concentrations. Other chemicals, which maybe detected using proton spectroscopy, include lactate, lipids and amino acids (glutamate/glutamine and GABA).37 Another technique, phosphorous (31P) MRS, provides a means to measure products of cellular metabolism such as adenosine triphosphate (ATP), phosphocreatine (PCr), phosphomonoesters (PME) and phosphodiesters (PDE). Of the imaging modalities, MRS may have the greatest potential for testing specific mechanistic hypotheses. A few studies have now been completed in bipolar disorder with these techniques that build upon the anatomic studies already discussed.
1H (proton) MRS studies
Proton MRS can be used to investigate specific neurochemical abnormalities in brain regions that might underlie the expression of bipolar disorder. As recently reviewed,4,5 a number of MRS studies have reported elevated choline concentrations in the striatum of bipolar patients that are independent of mood state or lithium therapy. This abnormality has generally not been observed in other brain regions,4,5 although there has been a recent report of choline elevation in cingulate.38 Given the previously reviewed evidence of structural abnormalities in the striatum in bipolar disorder, these data provide additional evidence for a subcortical basis of the expression of bipolar illness. However, elevated striatal choline may not be specific for bipolar disorder since it has also been observed in unipolar depression, albeit inconsistently (Strakowski5 for a review).
In addition to subcortical findings, recent MRS studies have revealed neurochemical abnormalities in several subregions of the prefrontal cortex.38,39,40,41,42,43,44,45 Specifically, decreased NAA in the dorsolateral prefrontal cortex was found in bipolar children and adolescents with parental bipolar disorder42 and in bipolar adults.45 Cecil et al40 examined metabolic abnormalities in orbital and prefrontal white and gray matter of bipolar patients experiencing a manic episode compared with healthy subjects. Consistent with other studies, they found reductions in NAA concentrations in the patients. Additionally, these patients exhibited decreased choline in gray matter and elevated amino acids (Glx) in white matter as compared to healthy subjects. Moore et al38 used serial MRS assessments of the anterior cingulate cortex to identify relationships among mood state, medication exposure, and choline and inositol concentrations. Choline (expressed as a Cho/Cr-PCr ratio) was elevated in the right cingulate in bipolar compared with healthy subjects. In the left cingulate of bipolar subjects, depression ratings positively correlated with choline concentrations, suggesting that choline elevation was affective state dependent in cingulate, in contrast to what has been reported in striatum, in which choline changes appear to be independent of affective state.4,5 Moreover, bipolar subjects who were not taking antidepressants had elevated choline concentrations compared with healthy subjects and bipolar patients taking antidepressants. This finding indirectly suggests that effective antidepressant therapy may normalize elevated cingulate choline levels. Consequently, the authors concluded that abnormalities in cingulate choline might represent impaired intraneuronal signaling in bipolar patients that is reflected in depressive states.38 In this same study, no abnormalities in inositol concentrations were noted. In contrast to the work by Moore et al38 Davanzo et al43 found a nonsignificant elevation in myo-inositol concentration in bipolar children compared with healthy subjects. In a follow-up study, these investigators reported that the elevated anterior cingulate myo-inositol might be specific to patients with bipolar disorder, since it was not present in patients with intermittent explosive disorder.44 Additionally, the bipolar adolescents in this study exhibited elevated myo-inositol concentrations compared to healthy subjects and adolescents with intermittent explosive disorder.44 A major confound of this study, though, was that patients were on a number of medications. Similarly, Cecil et al41 found elevated myo-inositol in frontal gray matter of children with a mood disorder and a parent with bipolar disorder, suggesting that elevated myo-inositol may be a biological marker for early onset affective disorders. These investigators also reported decreased NAA in the cerebellar vermis of these children.41 In contrast, Castillo et al39 did not find abnormalities in myo-inositol in dorsolateral prefrontal and anterior cingulate subregions of bipolar patients.39 However, they did report elevated glutamine/glutamate and lipids in the prefrontal cortex in bipolar children compared with healthy subjects.39 Together, these studies suggest that abnormalities in prefrontal metabolism, particularly in inositol pathways, which are altered by lithium treatment, may occur in bipolar disorder even early in the course of illness. Therefore, additional studies that examine this abnormality in at-risk samples and unaffected family members are needed to determine whether elevated prefrontal myo-inositol is a vulnerability marker for bipolar disorder. Additionally, decreased NAA levels may be a marker of impaired neuronal function. Indeed, Dager et al46 found elevations of lactate that coincided with decreases in NAA using proton spectroscopy to study medication-free bipolar patients. Both NAA and lactate are produced during mitochondrial metabolism and this finding of decreased NAA/increased lactate suggests that, in bipolar disorder, there may be a shift in energy redox state from oxidative phosphorylation toward glycolysis. It will be important to learn whether this abnormality predates the onset of bipolar illness, or instead is a consequence of treatment or recurrent affective episodes.
In addition to studies in prefrontal cortex, there are two published reports of MRS-identified neurochemical abnormalities in the hippocampus of bipolar subjects, both of which reported bilateral NAA reduction in bipolar compared to healthy subjects.47,48 Since NAA concentration is often regarded as a measure of neuronal health, with higher levels of NAA associated with a healthier neuropil, these findings suggest injury to or dysfunction of neurons in these brain areas. However, most of the patients in these studies were taking medications, which may have confounded measurements of NAA. Nonetheless, Deicken et al48 found a significant negative correlation between right hippocampal NAA and years of illness in bipolar subjects, suggesting that there may be worsening of neuronal pathology as the disease progresses.
In addition to hippocampal metabolite concentrations, Bertolino et al47 examined other regions of interest including the dorsolateral prefrontal cortex, superior temporal gyrus, inferior frontal gyrus, occipital cortex, anterior and posterior cingulate, centrum semiovale, prefrontal white matter, thalamus and putamen. No additional group differences in metabolite concentrations were found.47 Similarly, Ohara et al49 used proton MRS to assess metabolite concentrations in the basal ganglia (primarily lenticular nuclei), and they found no statistically significant differences between euthymic bipolar and healthy subjects. In contrast, Deicken et al50 observed elevated NAA and Cr in the right and left thalamus. The authors suggest that the elevation might represent neuronal hypertrophy, reduced glial density, or abnormal synaptic or dendritic pruning. In the absence of replications of either group's findings, these studies remain difficult to interpret.
31P (phosphorus) MRS studies
Phosphorus MRS provides a means to measure energy metabolism in the intact human brain. In a recent meta-analysis of 31P MRS studies in patients with bipolar disorder, Yildiz et al51 assessed the effect of mood state on the results of 31P MRS studies and found that PME values of euthymic patients were significantly lower than PME values of healthy controls. However, depressed bipolar patients had significantly elevated PME values compared with euthymic bipolar patients. No statistically significant differences in PDE concentrations were found between bipolar subjects and healthy controls. The results of this meta-analysis support the hypothesis that bipolar patients exhibit alterations in membrane phospholipid metabolism, which may reflect both trait and state dependent abnormalities in signal transduction pathways. However, further studies assessing the effects of medications on PME and PDE concentrations in bipolar patients are necessary51 to verify this notion.
Using MRS to examine medication effects
In addition to identifying specific regional neurochemical abnormalities, several studies have recently used 1H MRS to examine the in vitro neurochemical effects of medications in patients with bipolar disorder. Moore et al52 examined the effects of lithium on myo-inositol levels in 12 depressed bipolar adults. Significant decreases in myo-inositol levels were found in the right frontal lobe after short-term (5–7 days) treatment with lithium, which persisted with chronic treatment (3–4 weeks). However, the decrease in myo-inositol was observed prior to any change in the patients' clinical state. The authors concluded that inositol depletion is not associated with the direct therapeutic effects of lithium, although lithium treatment may initiate a cascade of cellular events, first evidenced in changes in inositol, that lead to therapeutic benefit.52 In a similar study examining the neurochemical effects of lithium in bipolar youth, Davanzo et al43 reported a decrease in anterior cingulate myo-inositol following 1-week of lithium treatment. However, in this study, patients were receiving concomitant medications, which significantly confounded the interpretation of this study. Lithium has also been shown to increase NAA in frontal, temporal, parietal and occipital lobes of bipolar adults,53,54 which has been interpreted to suggest that lithium may be neuroprotective (ie, by increasing evidence of neuronal health, namely levels of NAA). In contrast, there were no significant increases in NAA in bipolar patients who were taking sodium valproate.54 Silverstone et al55 used 1H and 31P MRS to examine the effects of lithium and sodium valproate on metabolite concentrations of the frontal and temporal cortex of bipolar patients. They reported no difference in myo-inositol or PME concentrations between bipolar patients taking divalproex or lithium as compared to healthy controls, suggesting that administration of these agents may normalize the phosphoinositol cycle in bipolar patients. However, there were no baseline MRS evaluations directly comparing the neurochemical changes before and after treatment, so these findings are at best indirect.
In a novel study, Lyoo et al56 administered choline as an adjunctive treatment for bipolar disorder and assessed brain purine, choline and lithium levels using 1H and 7Li MRS in order to identify the role of oral choline supplementation in modifying energy metabolism in bipolar subjects. The results of this study suggested no significant group differences in change from baseline to end point in clinical ratings, brain choline/Cr and brain lithium levels between those administered choline as compared to placebo. However, the group that was treated with choline showed a significant decrease in purine metabolite ratios from baseline (purine/NAA) and purine/choline) compared to placebo, suggesting that choline supplementation might correct the mitochondrial dysfunction that may be present in bipolar patients.56 However, the lack of clinical changes mitigates against the clinical relevance of this observation.
Recently, Soares et al57 used 7Li MRS to measure brain lithium concentration in bipolar patients and found no statistically significant correlation between serum and brain lithium levels, although patients on single daily dosing of lithium had higher brain-serum lithium ratios than those on twice a day dosing. Together, these studies demonstrated that MRS provides a means to evaluate treatment effects in bipolar disorder. As the number of studies using MRS increase and findings are replicated, MRS may help clarify neurochemical mechanisms underlying treatment response.
As noted, studies have fairly consistently observed elevations in striatal choline concentration in bipolar disorder, although this finding has also been observed in unipolar depression.58 However, Hamakawa et al59 compared patients with unipolar depression to those with bipolar disorder and found that the increased choline resonance observed in bipolar patients was greater than in the unipolar patients. Moreover, although the unipolar patients exhibited higher choline concentrations than healthy subjects, this difference did not meet statistical significance. Further confounding the story for unipolar depression, Renshaw et al60 reported lower choline concentrations in unipolar depressed compared to healthy subjects. Therefore, the specificity of elevated choline in bipolar disorder remains unclear and, instead, this may simply be a marker for affective disorders more generally. However, coupled with the previously reviewed structural MRI studies, these results suggest that striatal pathology may be a hallmark of bipolar disorder.
The specific meaning of elevated striatal choline in bipolar disorder is not known. The MRS choline peak consists primarily of glycerophosphocholine and phosphocholine,61 which are components of cellular membranes. Therefore, the MRS abnormalities that have been observed suggest abnormalities in membrane metabolism. Lithium has been found to inhibit choline membrane transport,62 which supports this assertion. However, this is a relatively minor lithium effect, as its effects on the inositol second-messenger system are more pronounced. Several studies have reported elevated myo-inositol concentrations, albeit inconsistently, in the brains of bipolar patients, which could represent the therapeutic target for lithium. Through its inhibition of inositol monophosphatase, lithium increases inositol monophosphate and decreases brain inositol levels.63 However, as noted, Moore et al52 found that lithium's effects on inositol levels occurred prior to any clinical effect. Therefore, these changes likely represented an initial component of a cellular chemical cascade that leads to lithium's therapeutic actions. The elevated myo-inositol levels observed in some studies may therefore be the end result of other abnormalities in second-messenger systems that have not yet been identified. Nonetheless, this finding may serve as a useful neurochemical marker for future MRS studies.
Several studies have also now reported decreased NAA in prefrontal cortical regions of bipolar patients. This nonspecific finding is unlikely to reveal details of the pathogenesis of bipolar disorder. However, NAA may represent a measure of mitochondrial metabolism and the decreases could suggest a shift to inefficient glycolysis (when coupled with lactate findings46) that may be useful as a target of treatment response or to predict the beginning of illness progression. Considerable more study is needed to determine whether any of these speculations prove to be useful. Additionally, it is difficult to compare studies since there is considerable variability in affective state, medication status and imaging methods. Standardizing research approaches could help in this regard.
Top of page
As noted previously, a limitation of morphometric imaging is that normal neural structure may not guarantee normal function. Functional brain imaging permits direct examination of the functional neuroanatomy of bipolar disorder. Several approaches, including positron emission tomography (PET), single-photon emission computed tomography (SPECT) and, more recently, functional magnetic resonance imaging (fMRI), provide in vivo methods for defining the anatomy of brain function. These technologies can be used to produce brain maps that correspond to changes in brain metabolism or blood flow (representing brain 'activation'), which are obtained at rest or during cognitive tasks designed to study specific neural networks, including anterior limbic regions that may underlie bipolar disorder.
Prefrontal cortical activation
As we have reviewed previously,4,5 functional imaging studies of prefrontal cortex generally find decreased metabolism and perfusion during depression in bipolar compared with healthy subjects. Since the prefrontal cortex is functionally complex, functional imaging provides a means to examine discrete prefrontal subregions that are difficult to delineate anatomically. With this in mind, Drevets et al11 examined the SGPFC specifically and found that this brain region exhibited decreased blood flow and metabolism in bipolar depression, although during mania (in a small subsample), metabolism increased to above normal. More recently, Blumberg et al64 similarly observed increased blood flow in the anterior cingulate (more generally) during mania as compared to euthymia, suggesting an affective state-related effect. In a second, smaller study they found decreased blood flow in the orbitofrontal cortex of manic bipolar patients at rest, and decreased right prefrontal cortical flow during performance of a word generation task compared with euthymic patients and healthy subjects.65 Other investigators have found both increased and decreased prefrontal glucose metabolism or cerebral blood flow during mania (reviewed in Strakowski et al4 and Strakowski5). One explanation for these apparent discrepancies is that different subregions of the prefrontal cortex exhibit different changes in perfusion or metabolism even during the same task. Consistent with this notion, Rubinsztein et al66 observed increased cerebral blood flow in the anterior cingulate, but decreased cerebral blood flow in the inferior frontal gyrus and right fronto-polar cortex during a decision-making task in manic bipolar patients compared with healthy subjects.
Functional MRI provides advantages in both temporal and spatial resolution relative to other imaging techniques. However, measures of activation are always relative as fMRI measures do not directly assess absolute changes in blood flow. Nonetheless, by integrating cognitive tasks, which serve as 'probes' to activate specific neural networks, with fMRI, it is possible to clarify the functional neuroanatomy of bipolar disorder. With this in mind, Curtis et al67 used fMRI during a verbal fluency task and found that bipolar patients had increased prefrontal cortical activation compared with healthy controls. The same patients performing a semantic decision task showed no activation differences. The mood states of the patients were not clearly delineated in this study, although average Young Mania Rating Scale and Hamilton Depression Scale ratings were low. Yurgelun-Todd et al68 found that bipolar patients performing a facial affect discrimination task showed decreased dorsolateral prefrontal cortical activation, when presented with fearful faces. Mood ratings in this study appeared to vary widely, complicating study interpretation. In contrast, in a fMRI study by Blumberg et al,69 mood states were clearly reported. In this study, patients performed a Stroop task. Bipolar patients showed a distinct area of decreased ventral prefrontal cortex activation, compared with healthy subjects that appeared to be independent of mood state. Manic patients showed decreased right ventral prefrontal cortical activation, whereas depressed patients showed an area of increased activation, compared with euthymic patients. Importantly, comparisons of fMRI studies are inevitably hampered by the wide variability of cognitive tasks employed. Tasks that serve as cognitive probes must be selected with careful consideration of the brain regions and functions of interest. In none of the studies described were 'in-scanner' performance data collected, raising questions as to the extent that the tasks described were actually performed as planned. Moreover, as demonstrated by the study of Blumberg et al,69 fMRI-assessed brain activation may vary during specific mood states. It is problematic, therefore, when studies do not report the mood states of participants.
To address this, we recently completed a study comparing euthymic, unmedicated patients with bipolar disorder to healthy subjects, studied with fMRI while performing a continuous performance task (CPT).70 By studying euthymic patients, the confound of mood state is removed, potentially permitting identification of state-independent (ie, trait) brain activation abnormalities that may represent risk factors for developing bipolar symptoms. Moreover, in this study, patients had been off of medication for at least a month, minimizing the confound of medications on brain activation. We found that, although the bipolar and healthy subjects performed the CPT virtually identically in the scanner, the brain activation patterns associated with this task differed significantly. Specifically, the patients exhibited significantly more activation in emotional brain regions (eg, ventrolateral prefrontal cortex, parahippocampus/amygdala) and task-related compensatory visual associational areas that suggested that the patients had a baseline over-reactivity of emotional regions, that during euthymia, were compensated for, in order to perform the task, by recruiting compensatory cortical areas. This inappropriate activation of emotional brain areas during a nonemotional task may underlie the vulnerability to mood episodes present in bipolar patients.
Subcortical and medial temporal activation
In addition to studies of prefrontal cortex, functional abnormalities in striatum and other subcortical structures have been reported in bipolar disorder. Specifically, compared with healthy subjects, Baxter et al71 observed decreased caudate metabolism in depressed bipolar patients. However, after adjusting for hemispheric metabolism, which was also decreased in depression, the specific regional specificity in the striatum was less clear. O'Connell et al72 reported increased blood flow in the basal ganglia, with right-sided flow greater than left, in bipolar patients during mania. Blumberg et al64 partially replicated this finding as they observed increased blood flow in the left head of the caudate in manic compared with healthy subjects. These latter investigators also compared a group of bipolar adolescents with healthy controls and found increased signal in the thalamus and putamen.73 However, the mood state of patients in this latter study was not described. Together, these findings suggested that state-dependent changes in activation occurred in striatum during the course of bipolar disorder that mirrored those in the prefrontal cortex.
Ketter et al74 examined cerebral metabolism in depressed, treatment-resistant rapid-cycling patients, and reported somewhat contrasting findings. They found decreased prefrontal and paralimbic cortical metabolism, which correlated inversely with ratings of depression. In contrast, ventral striatum, thalamus and amygdala demonstrated increased metabolism, positively correlated with depression ratings. These findings suggested an affective state-dependent loss of prefrontal activation that produced dysinhibition of limbic subcortical structures, leading to affective symptoms. Additionally, they observed increased cerebellar metabolism that appeared independent of mood state, disorder subtype or cycle frequency, which might be a trait or vulnerability factor. Yurgelun-Todd et al68 used a facial affect discrimination task to activate amygdala and found that, in bipolar patients relative to healthy subjects, amygdala was activated by the task with a corresponding reduction in dorsolateral prefrontal cortex activation. Consistent with the work by Ketter et al,74 this study suggests that a loss of prefrontal cortical control might be associated with increased amygdala activation.
A recent fMRI study of 24 bipolar patients and 13 healthy comparison subjects reported increases in cortical and subcortical activation during motor performance. The authors also reported that manic bipolar patients had significantly higher activation in right globus pallidus compared with depressed bipolar patients. Additionally, patients not treated with antipsychotics or mood stabilizers exhibited significantly higher activation throughout the motor cortex, basal ganglia and thalamus compared with patients who were receiving these medications, suggesting that mood stabilizers and antipsychotics may normalize cortical and subcoritcal hyperactivity associated with bipolar disorder.75
In addition to examining functional neuroanatomy, functional imaging permits direct evaluation of neurotransmitter activity through studies of receptor binding. These studies explore how changes in neurotransmitter function may lead to changes in functional neuroanatomy, and ultimately, symptoms of bipolar disorder. Several studies examined dopamine and serotonin density and binding in bipolar disorder. Ichimiya et al76 found increased serotonin transporter binding in the thalamus in a group of mood-disordered patients, including bipolar patients, in a variety of mood states. However, the increase in specifically bipolar patients was not significant. Serotonin 5HT1A binding potential was increased in the raphe and mesiotemporal cortex, in a group of patients with affective illness. Although bipolar patients were a minority of subjects, differences were greatest in bipolar patients and depressed patients with bipolar relatives. More studies have used PET to examine dopamine binding in patients with bipolar disorder. Using 11C-SCH2330 PET, Suhara et al77 demonstrated that dopamine D2 receptor binding potentials are lower in the frontal cortex of bipolar patients than in healthy comparison subjects, but no differences were observed between these groups in the caudate. Similarly, Wong et al78 observed no differences in the caudate/cerebellum dopamine D2-binding ratio between bipolar patients and healthy subjects. Studies employing [123I]iodobenzamide and [11C]raclopride similarly found no increase in baseline dopamine binding in striatum in euthymic and manic patients, respectively.79,80 Moreover, there was no difference in amphetamine-induced striatal dopamine release between patients and healthy controls, although bipolar patients exhibited greater behavioral changes, suggesting increased postsynaptic neural responsivity.79 Yatham et al81 did not find any increase in [18F]6-fluoro-L-Dopa uptake in bipolar patients, consistent with these other studies. Pearlson et al82 also found that bipolar patients without psychosis showed no increase in dopamine receptor density. However, patients with psychotic symptoms demonstrated increased density of dopamine D2 receptors in the caudate compared with healthy controls. Moreover, D2 receptor density correlated with scores on psychotic items of the Present State Examination. To date, then, receptor-binding studies do not commonly observe differences in D2 receptor binding, except perhaps in bipolar patients with psychotic features. Findings of potential serotonin abnormalities need replication and extension.
Functional imaging provides direct measures of the functional neuroanatomy of bipolar disorder. To date, functional imaging studies suggest that several prefrontal subregions are abnormally activated at rest and during cognitive tasks in bipolar compared with healthy subjects. In general, during depression most investigators have reported a generalized decrease in prefrontal activation, although studies in unipolar depression have suggested that some subregions, such as the subgenual cingulate and orbital frontal cortex, may exhibit increased activation in this mood state.83 Similarly, during mania several prefrontal regions exhibit decreased activation, suggesting that this response may occur as a result of mood episodes in general. However, during mania increased activation has been reported in the anterior cingulate, which, as noted, has also been observed in unipolar depression (although not bipolar depression11). Therefore, increased activation of the anterior cingulate may be associated with different abnormal mood states. These data suggest that activation of 'limbic' (emotional) prefrontal areas may disrupt (and decrease activation) of 'cognitive' prefrontal regions. Further supporting this notion, several investigators have observed increased activation in other components of the anterior limbic network, including striatum, thalamus and amygdala, in patients with bipolar disorder during mood episodes, particularly mania, but also depression.64,68,74 Many of these abnormalities appear to be mood state dependent and relatively few studies of carefully defined euthymic patients have been reported that might clarify potential 'trait' functional abnormalities. As noted, we have observed activation in cortical areas during euthymia that might serve to compensate for a baseline over-reactive brain networks that subserve emotion,84,85,70 and which could represent a trait characteristic if validated in larger studies. Regardless, functional imaging studies support the notion that abnormalities in anterior limbic networks, regions in which structural and spectroscopic abnormalities have also been reported, appear to underlie the expression of bipolar disorder. Moreover, the dysfunction of these anterior limbic networks may disrupt function of cognitive networks,83 consistent with clinical observations of patients with mood disorders experiencing cognitive disabilities during affective episodes.
Top of page
Neuroimaging studies of bipolar disorder have provided a number of important clues toward the functional neuroanatomy of this condition. Structural imaging suggests abnormalities in prefrontal cortical areas (SGPFC), striatum and amygdala that appear to exist early in the course of illness and, potentially, predate illness onset. Other abnormalities in midline cerebellum, lateral ventricles and other prefrontal regions (eg, left inferior) appear to develop with repeated affective episodes so may represent neuroanatomic sequellae of the illness. Spectroscopic studies identify abnormalities of membrane and second messenger metabolism, as well as bioenergetics, in striatum and prefrontal cortex as well, further establishing striatal–thalamic–prefrontal circuits as an integral part of the expression of bipolar disorder. Finally, functional imaging studies report activation differences between bipolar and healthy subjects in these same anterior limbic regions. Together, neuroimaging studies support a model of bipolar disorder that involves dysfunction within subcortical–prefrontal networks and the associated limbic modulating regions (amygdala, midline cerebellum). There are a number of different prefrontal–striatal–thalamic circuits that appear to iteratively sample input from sensory, cognitive and emotional cortical regions to modulate complex human responses.86,87 Those iterative networks that modulate socioemotional behaviors (eg, orbitofrontal–medical striatal–thalamic) appear to be dysfunctional in bipolar disorder.87 Indeed, several investigators have proposed that dysfunction within selected prefrontal–striatal–thalamic networks underlies the expression of affective illness4,5,6,74,88 based on imaging data as well as reports that injury to structures within these networks produces affective symptoms.86 For example, Starkstein et al89 found that specific injury to the right caudate head may precipitate mood cycling, similar to that observed in bipolar disorder.86 Therefore, a relative loss of prefrontal modulation of subcortical and medial temporal limbic structures may underlie the symptoms of bipolar disorder. Conversely, dysfunction within these limbic areas may disrupt the function of cortical regions associated with mood and cognitive function (eg, anterior cingulate, dorsolateral prefrontal cortex) resulting in bipolar symptoms.74,83 Importantly, several of these structures (eg, SGPFC and amygdala) project extensively to the hypothalamus, which is likely to produce neurovegetative symptoms and contribute to the abnormal 'feelings' of affective disorders. These considerations are portrayed schematically in Figure 1.86,90
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact email@example.com or the author
Schematic of the anterior limbic network as a model of the expression of bipolar symptoms. (a) 'Iterative' prefrontal subcortical network that samples inputs from other brain regions to modulate appropriate behavioral responses. (b) Limbic areas that modulate (eg amygdala) or express the response (eg hypothalamus).
Full figure and legend (23K)
Imaging studies taking together highlight that it is unlikely that bipolar disorder will localize to abnormalities within a single neuroanatomic structure. The brain is not organized into discrete independent functional packets, but, instead, consists of complex, interconnected neural networks. Therefore, dysfunction in any part of a network will echo throughout the brain in complex and, currently, unpredictable activations. Nonetheless, taken together, these studies suggest that, in bipolar disorder, there may be relatively diminished prefrontal modulation of subcortical and medial temporal structures within the anterior limbic network (eg, amygdala, anterior striatum and thalamus), which results in dysregulation of mood. Future studies focused on these specific relationships will clarify the functional neuroanatomy of bipolar disorder.