Brain Imaging in Mood Disorders

Godfrey D. Pearlson, MD and Thomas E. Schlaepfer, MD

The ultimate aim of imaging studies as applied to mood disorders is to describe their structural and functional neuroanatomy. An additional vital question is whether all affective disorders are associated with a single common constellation of brain changes. Alternatively, there could be multiple brain changes, each associated with one of the diverse clinical syndromes constituting the affective disorder spectrum.

Although biochemical, pharmacologic, chronobiologic and brain imaging techniques have all been used to shed light on the neurobiology of mood disorders, knowledge of the underlying pathobiology remains sketchy (41), in part because of the non-uniform clinical nature. Affective illness is clinically most diverse, with disputed boundaries, encompassing multiple subgroups, overlapping symptomatically with other psychiatric syndromes and further confounded by several comorbid conditions. These facts have made brain imaging studies of useful size and subject homogeneity hard to carry out and to interpret.

In addition, studies of affective disorder suffer by comparison to those in schizophrenia, a similarly poorly bounded clinical syndrome, which has nevertheless benefited from numerous neuropathologic investigations. These have often helped to initiate hypotheses later tested by brain imaging techniques. However, virtually no such neuropathologic investigations of affective disorders, which could have served similarly to guide neuroimaging studies, have been carried out.

The logic for using neuroimaging techniques to elucidate the pathophysiology of affective syndromes has been laid out by Cummings (18, 19). First, supportive evidence for brain differences is gathered with neuroimaging techniques by comparing affective patients to healthy controls. Similar studies are carried out in parallel in patients with mood changes occurring in the context of focal brain pathology due to traumatic injury or disease processes of known location (e.g., stroke, seizure disorders and Huntington's disease). Significant associations between these syndromes and regional deficits revealed on neuroimaging are then delineated. Last, common findings between the two types of studies are then examined, and further hypotheses generated.

Neuroimaging studies of affective disorders have been complicated by many factors affecting design and comparability, which are discussed in more detail below.

The heterogeneity of affective disorders is well recognized. Some of the more important subtypes are listed in Table 1. This diversity necessitates very careful descriptions of study populations, as recommended by Kupfer and Rush (50). Such basic descriptive details are seldom provided in neuroimaging studies, however. The population being imaged may well be crucial if the corresponding underlying pathophysiology is non-uniform. The reality, however, is that most studies to date have included mixtures of very diverse patient populations without attempts to analyze data separately. Given that many such patient groups studied are small to begin with, real findings may be hidden by the additional confound of heterogeneity. One could usefully adopt a strategy of "biotyping" mood disorders through neuroimaging, if homogeneous groups were studied and compared. Unitary markers for mood disorders could be identified showing relative specificity and potentially providing novel classification schemes. Such changes might be nonspecific, (e.g., overlap with those of schizophrenia or obsessive-compulsive disorder), or highly specific, (e.g., blood flow changes related to level of motor activity, differentiating agitated from psychomotor retarded patients).

Preliminary evidence suggests that studying patients with more homogeneous subtypes of affective disorders may indeed help clarify associated biological changes. Drevets (25), for example, studied patients with familial pure depressive disorder, a relatively homogeneous clinical subtype of unipolar disorder. Using positron emission tomography (PET) measurements, he determined that regional cerebral blood flow was increased in left prefrontal cortex, left amygdala and left medial thalamus, but reduced in left medial caudate (Fig. 1). These changes appeared fairly robust (replicated in an independent sample) but were in the opposite direction of those reported in many prior studies, which recorded functional changes in less homogeneous affective disorder populations.

The distinction between early adult and late-life onset depressive patients, reviewed in detail below, also seems to be meaningful. There is evidence for an etiology (cerebral microvascular disease) associated with late-onset cases, as well as for structural and functional differences that can distinguish both groups.

Delusionally depressed patients seem to have larger ventricles on computed axial tomography (CT) than non-delusional patients (72, 87).

Other scientists have examined the effect of duration of disorder on ventricular measurements. These measurements are confounded to some extent by age of onset. No effect of duration was noted by Dolan (24) or Nasrallah (60). The relationship between the severity of the most recent episode and ventricular enlargement was explored by Luchins (54) and Pearlson (62) in bipolars, but no clear-cut relationship was demonstrated. Several studies reported no differences in CT measurements of patients with or without family histories of affective disorders (24, 40, 60). However, Saccetti (72) demonstrated that patients with a negative family history tended to have larger ventricles.

Some, mainly elderly, depressed patients show reversible cognitive abnormalities. In a CT study, Pearlson compared cognitively normal elderly patients with depression to similarly aged patients with a reversible dementia syndrome of depression (61, 63, 81). The depressed sample as a whole was intermediate on several measures between normal controls and patients with Alzheimer's disease. Relative to normal individuals, demented depressed patients had larger ventricles and reduced CT brain attenuation numbers. In contrast, younger adult bipolar patients show more cognitive abnormalities and more white matter hyperintensities on magnetic resonance imaging (MRI) [26].

Separate from the question of state/trait markers (which will be discussed below), some researchers have investigated possible relationships between dexamethasone suppression test status and structural brain measurements in patients with major depression. One hypothesis is that high circulating levels of endogenous steroids produce cortical atrophy and ventricular enlargement similar to that found with exogenously administered cortisol (10). Kellner showed a relationship between lateral ventricular enlargement on CT and raised 24-hour free urinary cortisol measurements in depressed patients (43). Schlegel demonstrated a similar relationship between ventricular enlargement and average afternoon plasma cortisol values (78). Other investigators have shown no relationship between dexamethasone non-suppression and ventricle to brain ratio (VBR) [23, 77, 78, 81, 87].

Ill vs. well and state vs. trait issues are especially relevant to neuroimaging studies of affective disorders, which by their nature tend to be episodic, with return to normal functioning between episodes. Whether underlying abnormalities are present only during periods of illness, with return to normal between episodes, or persist independent of clinical status can be addressed only via longitudinal studies. Although state-dependence is hard to imagine for structural changes, possible reversible steroid-related structural alterations could be driven by the hyper-cortisolemia seen in major depression. A second possibility is that neuroimaging changes represent trait markers of the illness and that differences from normal are present regardless of whether an individual patient is ill or well at the time of testing. Dupont (26) reported state-independent white matter abnormalities in more than half of a sample of young bipolar individuals on MRI. Drevets (25) showed that while increased left prefrontal regional cerebral blood flow was seen only in depressed individuals and appeared to be state-related, increased flow in the left amygdala appeared in both ill and remitted patients (Fig. 1). Thus, episodic illness may have both state and trait markers. However, the data are preliminary, since few or no studies have examined the same patients in both the ill and remitted states.

Progression of structural changes has been hypothesized to occur in concert with clinical episodic progression, in an analogy with Post's model of kindling.

Medicated versus unmedicated status is linked conceptually to state/trait issues. Psychotropic medications and other treatments, such as electroconvulsive therapy (ECT), likely cause significant changes in cerebral pharmacological receptors, neurotransmitters, blood flow and metabolism. Treatment status of patients thus has the potential to confound many functional studies of affective disorders. Many studies of remitted patients examined individuals taking mood-altering medications or those who were conceivably suffering from the long-term effects of physical treatment, such as ECT. Few have attempted to resolve this issue. Also, very few human investigations have addressed the effects of long-term lithium or antidepressant treatment given to normal volunteers. Coffey (14, 15) examined depressed elderly individuals before and during ECT treatment and demonstrated no structural MRI changes between the two conditions.

Some longitudinal studies that attempted to address state/trait issues have been complicated in their interpretation by possible treatment effects (39, 54). Baxter (7) found that in three unipolar depressed patients glucose metabolism in caudate, as measured by PET, increased concomitantly with antidepressant treatment. He also showed treatment-related increases in glucose metabolism in dorsolateral prefrontal cortex. In cross-sectional studies, no effect of prior ECT (24, 62, 81) or prior medication (24, 53, 62, 79) was demonstrated on lateral ventricular enlargement in affective disorders. Future studies need to account for treatment effects more regularly than has been the case in much of the existing literature.

Much evidence suggests that the distinction between early-in-life and late-in-life (early-onset and late-onset) depressive patients is useful. The neurobiology of affective disorders seems to show some biologic and epidemiologic specificity in the two populations. For example, a proportion of late-onset cases are associated with neuroimaging evidence of cerebrovascular disease.

Using MRI, Coffey (16) and Krishnan (48) found more white matter lesions in late-onset depressive patients. Greater sulcal widening and greater severity of subcortical white matter lesions have been reported in elderly depressives with late onset depression (16, 48). These findings have recently been confirmed (68). Patients with late-onset depression have more and larger white matter hyperintensities in the caudate than age-matched patients with early-onset depression, suggesting that these caudate lesions may be involved in the etiology of late-onset depression in some patients (30). Shima (80) found that later-onset patients had greater ventricular enlargement on CT than comparably aged, early-onset subjects. Dolan (24) also showed that late- but not early-onset depressives had ventricular enlargement compared with age-matched controls. However, similar studies in young mood disorder patients did not find clear effects of age of onset on regional or global sulcal widening (23, 46, 71, 77). An MRI study that included 29 normal volunteers and 20 patients with major depression revealed particularly prominent shortened T1 relaxation times in the hippocampus of elderly depressed patients, suggesting that major depression in the elderly may be associated with tissue changes in the aging hippocampus (47).

Demented, elderly, depressed patients (also discussed below) were similar to demented, non-depressed patients with respect to a decreased centrum semiovale intensity on CT (61). Hyperintensities of deep gray matter structures (such as the basal ganglia and the thalamus) found in the elderly could not be detected in younger adult bipolar samples (30).

Coffey et al. (14) found more white matter lesions in late-onset depressed patients: Figiel (30) found basal ganglia lesions to be more common in a similar population. White matter hyperintensity rates seen in depressed elderly patients are similar to those reported in vascular dementia patients (38); the implications of this fact are reviewed in ref. 73.

Outside of differences in subtypes of affective disorder, there are important methodologic differences between studies that are responsible for the inability to compare data across different investigations; some of these are listed in Table 2.

Inherent differences in the imaging methods themselves used to study patients have made comparisons across different studies less than straightforward. Methods of analysis can differ widely between otherwise similar studies. Potential artifacts of analysis—for example, those associated with blood flow ratio methods (73, 74)—must be addressed. These considerations are especially important because of the need to study changes in neural systems as well as in localized regions.

Phenomenologically, clinical symptoms can overlap between affective disorders and schizophrenia, anxiety disorders, and obsessive compulsive disorder. By analogy, one might reasonably expect a similar overlap in associated cerebral pathophysiology, and hence the findings on structural and functional neuroimaging. Hence, diagnostic specificity is a key element.

Jeste et al. (41) stated that "in every area in which structural abnormalities have been found in neuroradiological studies of schizophrenia, abnormalities of similar magnitude have also been demonstrated in affective disorders." This statement could be true for several reasons. First, common pathophysiological mechanisms could underlie the two syndromes, as suggested, for example, by Buchsbaum et al. (12) and Cohen et al. (17). Studies have seldom been designed to address these hypotheses directly. Mood disorder patients have often been studied only as a comparison group in primary studies of schizophrenic illnesses. Nevertheless, there are numerous illustrations of Jeste's comment. Schizophrenics and manics have been shown to be similar in multiple CT studies of sulcal widening, vermian cerebellar atrophy, and lateral ventricular enlargement (59, 66, 88). These studies generally have shown similar changes in mood disorder patients and schizophrenics, as opposed to controls, but the changes are less marked in mood disorders. More recent MRI studies have shown similarities, but also clear-cut differences (65), discussed below in the temporal lobe section. Some functional neuroimaging studies have also revealed similar brain changes associated with both schizophrenia and affective disorder. Buchsbaum et al. (12) demonstrated similar hypofrontality in schizophrenic and mood disorder patient populations, compared with normals. Pearlson et al. (64) demonstrated analogous dopamine (DA) D2 elevated bmax measurements in PET neuroreceptor studies of psychotic and schizophrenic patients, compared with both normal controls and non-psychotic mood disorder patients. Berman (11) examined the issue of specificity of altered cortical blood flow in schizophrenics and mood disorder patients. They evaluated 10 patients with schizophrenia, 10 depressed patients, and 20 age- and sex-matched normal controls. Regional cerebral blood flow was separately measured at rest, during a simple number-matching task, and during performance of the Wisconsin Card Sorting Test, using inhaled radioactive xenon. Schizophrenic patients had lower prefrontal regional cerebral blood flow only during performance of the Wisconsin Card Sorting Test. No differences in either regional or global blood flow were demonstrable between depressed patients and normals during any testing condition.

A related issue is regional specificity. Studies need to determine which of a series of brain changes are uniquely associated with affective disorders, rather than reflecting a nonspecific alteration such as generalized atrophy or altered blood flow.

Overall, although there is considerable overlap with other major psychiatric syndromes, especially schizophrenia, brain imaging changes in mood disorders exhibit some putatively specific findings.

Comorbidity of differing types may be an important contributing factor to heterogeneity in affective disorders, and thus potentially an important confounding factor, especially in functional neuroimaging studies. Examples of such comorbid factors are substance abuse disorders, commonly seen in association with affective syndromes. Alcohol and other drugs, (even such ubiquitously used substances such as nicotine and caffeine) likely have important local and global effects on regional and global blood flow, as well as on neurotransmitters. It has recently been shown that opioids seem to exert their actions in parts of the limbic system (76), a circuit which has been implicated convincingly in affective disorders (44, 65). Furthermore, there is preliminary evidence for structural defects in the frontal lobe of polysubstance abusers (76). It is therefore essential to control any substance abuse in neuroimaging studies of affective disorders. Finally, cerebrovascular risk factors need to be carefully assessed in studies examining subcortical and periventricular white matter hyperintensities in bipolar disorder, in order to help separate causal from consequent factors.

A growing body of evidence speaks to both structural and functional gender differences in the human brain. Using MRI, Schlaepfer demonstrated large gray matter volume differences in the dorsolateral prefrontal cortex, a region consistently implicated in functional neuroimaging studies of affective disorders, in healthy subjects (75). In a PET study, George showed a very different pattern of regional cerebral blood flow in response to emotional stimuli in a group of healthy volunteers (34). Studies in normal volunteers examining the effects of powerful emotional stimuli (e.g., film clips of upsetting events) and of simulated emotions have begun to shed some light on functional circuits subserving emotional events.

Affective disorders are especially heterogeneous clinically, and evidence supports the contention that different subtypes are likely to be associated with different brain changes. Many apparent discrepancies among studies are likely attributable to differences in patient samples. In future research, both studying clinically homogeneous subjects and reporting their salient characteristics will reduce these confounds and allow attempts at replication.

Bearing these points in mind, we will now review the various findings reported in patients with depressive and bipolar disorders.

Biological research in affective disorders has focused traditionally on the neurochemical basis of the disease. The neuroanatomical basis of these disorders has begun to be studied only in the last decade. Currently, radiological techniques that allow investigation of the living brain in health and disease have become widely available. Structural abnormalities in affective disorders were first investigated with CT. Later the more precise methodology of MRI was applied. Compared to the many investigations of structural differences in patients with schizophrenia, relatively few data have been obtained in patients with affective disorders. When various types of structural abnormalities were first described in schizophrenia, the initial hope was that these differences might turn out to be specific. This was not realized, as similar abnormalities have often been reported in affective disorders. As discussed below, the most consistent findings of structural brain differences relatively specific to affective disorders are abnormalities of subcortical white matter and basal ganglia.

The most frequently reported structural difference in affective disorders is an increase in the lateral ventricle to brain ratio (VBR). This change is rather nonspecific, as it is also seen, for example, in normal aging, schizophrenia and Alzheimer's disease. The drive to study structural abnormalities in patients with affective disorders was first generated from imaging and neuropathological studies in schizophrenia. Patients with affective disorders were initially included in neuroimaging studies of schizophrenia to control for the specificity of the findings. CT studies of VBR showed that patients with affective disorders tended to be quite similar to schizophrenic patients and significantly different from normal control subjects (24, 41, 66, 70, 78, 79), although the differences tended to be less pronounced in patients with affective disorders (88). More recent studies, however, have been unable to confirm the finding of ventricular enlargement in affective disorder (15, 52), and a current review of the literature suggests that the correlation between VBR and clinical measures is inconclusive (22). Of 19 studies focusing on VBR, only three reported data on patients with unipolar and bipolar depression separately (4, 24, 77). One study reported that manic male subjects had significantly larger VBR than unipolar depressed patients (4). The other two studies found no VBR differences in these two subtypes of affective disorder. Studying manic patients, several groups (53, 59, 66) found the lateral cerebral ventricles to be enlarged, compared with normal controls. Others found enlargement only in the third ventricle (23). The reason for these inconsistencies is not clear (60), and it is difficult to compare studies due to differences in the methodologies of image acquisition and processing, as well as the inhomogeneity of patient populations. Important and possibly often overlooked confounding factors could be comorbidity with substance abuse, medication, and the inadequate matching of control subjects. One CT study showed that alcohol use, for instance, even in socially accepted amounts, can lead to brain atrophy (13).

Another structural difference that has been reported in patients with affective disorders is sulcal widening. Widening of the interhemispheric and sylvian fissures is characteristic of older, depressed patients (68). This finding was confirmed in younger manic patients (59). Others found sulcal widening particularly in frontal and temporal areas (24, 46, 68). Further studies, however, failed to confirm sulcal widening, either globally or regionally (23, 77). As with ventricular enlargement, the data on the existence of this structural difference in affective disorders remains inconclusive and inconsistent. Late-onset depression might be a clinically and neurobiologically unique entity. Its neuroimaging correlates are therefore difficult to compare with findings from early-onset depression.

The above mentioned observation of inconsistent findings is also true for a third structural difference that has been reported in manic patients—cerebellar vermian atrophy. Atrophy of the cerebellum in both manic and depressed patients has been reported in several studies (59, 87). In some of these studies, comorbid alcohol use was not controlled for and could therefore explain some of the atrophy. Two other studies found no alterations of the cerebellum (66, 90).

Probably the most consistent findings of structural abnormalities in patients with affective disorders are in subcortical structures. Using CT, one study demonstrated increased radiodensity of the caudate head bilaterally in elderly depressed patients (8). Smaller volumes of caudate and putamen nuclei in depression have also been reported using MRI (49).

Several groups reported subcortical hyperintensities in patients with affective disorders using MRI (5, 26, 30, 68). Dupont found these abnormalities in eight of 14 young bipolar patients; she reported later that patients with these "unidentified bright objects" tended to have more previous hospitalizations and tended to perform worse on neuropsychological tests than patients without them (26). The abnormalities consisted of areas of increased signal intensity (therefore appearing as white spots) in T2-weighted MRI images and were located a) as a periventricular rim, b) as patchy or confluent areas in subcortical white matter and c) as lesions in subcortical gray matter structures such as the pons, thalamus, and basal ganglia. Consistent with these findings in young adults, almost two-thirds of elderly patients referred for ECT were found to have changes in subcortical white matter (14). Although the authors suggested that this "leukoencephalopathy" might predispose to late-onset, treatment-resistant depression, it is important to note that higher rates of subcortical hyperintensities have been reported in both young bipolar (5, 26, 30) and elderly depressed (mainly unipolar) patient groups (14, 16, 30, 51, 68). Like the structural changes discussed previously, subcortical white matter changes currently have the status of a nonspecific finding of unknown significance.

Altshuler (2) reported reduced temporal lobe volume in affective disorder compared with controls, although Johnstone (42), who examined temporal lobe area on CT scans, did not.

Prior MRI studies reported both medial (temporolimbic) and lateral cortical temporal changes and temporal lobe asymmetries in schizophrenic patients, compared with healthy controls. The specificity of such temporal lobe (TL) changes in schizophrenia vs. affective disorder is unknown. Pearlson et al. (65) determined the occurrence and specificity of these TL changes in 46 schizophrenic patients, compared with 60 normal controls and 27 bipolar subjects. Bilateral volumes of anterior and posterior superior temporal gyri (STG), amygdala, entorhinal cortex, and multiple medial temporal structures, as well as global brain measures, were obtained. Several regional comparisons distinguished schizophrenia from bipolar disorder. Entorhinal cortex was bilaterally smaller than normal in schizophrenics but not in patients with bipolar disorder. Patients with schizophrenia, but not those with bipolar disorder, had reversed posterior STG asymmetry. Additionally, left anterior STG and right amygdala were smaller than predicted in schizophrenics, but not in patients with bipolar disorder. Left amygdala was smaller and right anterior STG larger in bipolar disorder but not in schizophrenia. Thus, temporal lobe structural changes specific to both disorders appear to exist.

Compared to primary affective disorders, considerably more structural data have been obtained for secondary affective disorders following central nervous system (CNS) lesions. In a review of affective symptoms following stroke, the incidence of depression was higher than that of mania, but manic symptoms were more frequently observed in patients with a previous personal or family history of affective disorders (82). A connection between the site of the lesion and depressive symptomatology was established in several studies. A CT study of patients with closed head injury demonstrated that major depression developed more in patients with left basal ganglia and left dorsolateral frontal lesions (29). In a second CT study, stroke patients with right temporal, right superior frontal, left inferior frontal or left parietal occipital lesions were found to have more sleep disturbance and greater dysphoria than patients with lesions elsewhere in the brain (84). Lesions associated with depression and mania seem to occur more often in the temporal and frontal lobes. Right-sided lesions tend to be associated with mania and left-sided lesions with depression (82, 83).

Affective disorders are often associated with Huntington's disease (HD). Since the disease has a fairly well documented neuropathology, and correlations between the extent of structural alterations and clinical abnormalities have been established, HD is of particular interest to psychiatry. About 40% of patients suffering from HD develop symptoms of affective disorder over the course of their illness; about 10% of these have manic episodes (31). It is interesting that the affective symptoms associated with HD often precede the motor and cognitive symptoms by many years. MRI studies report marked reduction in the volume of the putamen and the caudate nucleus in mild cases of HD, consistent with observations in primary affective disorders (49). Structural brain pathology specifically associated with the development of affective changes remains to be elucidated.

Functional abnormalities in affective disorders have been assessed with the techniques of single photon emission computed tomography (SPECT) and PET. Both modalities provide useful data about cerebral blood flow (CBF) or cerebral metabolic rate (CMR). By quantification of physiological parameters, both techniques give some insight into the biochemical bases of affective disorders. Few affective disorder patients have been investigated in the manic phase because of the difficulty of image acquisition in these poorly compliant patients in the absence of prior pharmacological treatment, which introduces inevitable confounds. Almost no studies have tracked patients longitudinally. Therefore, most data have been obtained in patients with major depression or bipolar patients in the depressive phase of the disease. Results to date are intriguing but hardly in unanimous agreement. Discrepant findings likely due to wide differences in patient selection, choice of comparison populations, stimulus conditions, image acquisition and analytic methods.

Many functional imaging studies in affective disorders initially focused on the question of whether patients in acute affective episodes have abnormal global changes in blood flow or metabolic rates. Several studies reported decreased rates in both CBF and CMR in patients with various affective disorders, compared with normal controls in a resting state (6, 7, 55). Other groups were not able to replicate these findings in patients with major depression (9, 35, 36, 37, 45), and one study even reported a global CBF elevation (69). In an activation study, global glucose CMR of depressed patients, schizophrenic patients, and normal controls was investigated while they received unpleasant electrical stimuli. No differences were found in global brain metabolism across both patient populations, but a decreased anterior/posterior ratio was found in both affective and schizophrenic patients. In a related activation study, the same group found globally increased glucose metabolism in bipolar and unipolar patients (12).

It is not clear why these global metabolism and blood flow findings are so inconsistent. As we have emphasized above, one likely source of error is the great heterogeneity of the patient groups included in the studies: unipolar, bipolar, young adult outpatients with mild depression, elderly severely depressed patients, and patients with scores on the Hamilton Depression Rating scale (HRS-D) ranging from 15–35. Furthermore, image acquisition and processing vary considerably between studies, resulting in data that are difficult to compare.

Regional reduction of metabolic activity in frontal cortical regions, particularly the dorsolateral prefrontal cortex (DLPFC), in major depression has been suggested by several PET studies (6, 7, 67) [Fig. 2]. The most consistent finding has been a decreased ratio of metabolic rate of left DLPFC relative to the total metabolic rate of the left hemisphere. This ratio correlates significantly with scores on the Hamilton depression rating scale (6). The observed hypometabolism may be more state than trait related, as the abnormality normalizes when depressed patients recover (6). A SPECT study reported significantly decreased rates of cortical blood flow in the left hemisphere in bipolar patients (21).

Anterior frontal hypometabolism mainly on the right side was reported in one sample of depressed patients (39). In an activation study using PET, depressed patients performing an attentional task showed a reduction of CMR in the medial frontal cortex (17). A recent SPECT study using 123I-labeled amphetamine (IMP) indicated that patients with major depression showed an increased IMP activity in the right temporal lobe (3). Only a few early studies reported the absence of regional abnormalities in CBF and CMR at rest (35, 36, 45). In contrast to the relatively inconsistent findings in abnormalities of global CBF and CMR, there are many and rather consistent observations of hypofrontality in major depression (12, 27).

The possibility of an abnormal caudate metabolic rate relative to hemispheric metabolism in major depression was raised by one group (6). Patients with unipolar depression showed a significantly lower ratio of the metabolic rate in the caudate nucleus divided by that of the hemisphere as a whole, when compared with normal controls and patients with bipolar depression (6, 7). Similarly, depressed patients with Parkinson's syndrome had decreased metabolism in the caudate, suggesting that a lesion at the level of the basal ganglia might be associated with depression (58). This finding is particularly interesting in light of the evidence (discussed earlier) of a high prevalence of mood disorders in patients with lesions of the basal ganglia (83).

As seen in the foregoing discussions, CBF and CMR changes in affective disorders have been found in specific brain regions including the limbic system, the basal ganglia, and both frontal and temporal cortices. Of great interest are treatment-related changes in blood flow and metabolism in these regions. A SPECT study using the tracer Tc-99m-hexamethylpropylenamineoxime (HMPAO) investigated depressed patients before and after sleep deprivation. All patients showed relative hypoperfusion in the left anterolateral prefrontal cortex under both conditions. However, only responders to the therapy showed an initial hyperperfusion in portions of the limbic system and a subsequent reduction of blood flow in these same regions following sleep deprivation (27). Using xenon SPECT, Sackeim demonstrated that ECT transiently reduced rCBF globally and regionally in depressed patients (73). When depressed subjects were scanned immediately after receiving ECT and then followed in their clinical response, the patients with the largest transient reduction in left prefrontal activity had the best treatment response. This correlation of clinical improvement with reduction in left frontal rCBF might suggest that transient reductions in left prefrontal rCBF may be a part of the mechanism of action of ECT (73, 74).

Biological hypotheses regarding the pathophysiology of affective disorders initially concentrated on abnormalities in neurochemical transmission, with a particular emphasis on possible alterations in receptor physiology. While both PET and SPECT have the potential to test specific hypotheses concerning the role of neurotransmitter systems in affective disorders, relatively few imaging studies using receptor ligands have been done, perhaps because of the inherent complexity of this research. It is important to note that no hypothesis of neurotransmitter dysfunction in affective disorders adequately explains all existing data.

A recent SPECT study of depressed patients using a ligand for the 5-HT2 serotonin receptor demonstrated increased uptake of the tracer in parietal cortex and a right/left asymmetry (R>L) in the inferior frontal region of the patients, suggesting changes in the serotonin receptor density in these regions (20). In addition, significantly lower uptake of L [11C] 5-HT across the blood brain barrier in six patients with a history of major depressive disorder was reported in a PET study (1). These differences were observed mainly in the basal ganglia and in the prefrontal cortex.

A separate PET study demonstrated that patients with strokes in the right hemisphere had higher binding of (3-N-[11C]methyl) spiperone to serotonin S2 receptors in the right parietal and temporal cortices than a similar group of patients with strokes in the left hemisphere or normal subjects (Fig. 3). The ratio of ipsilateral to contralateral binding in the cortex was negatively correlated with the severity of depression. These findings suggest a different biochemical response of the brain depending on the location of the injury. Some secondary depressions may be a consequence of insufficient up-regulation of serotonin receptors after stroke, particularly in the left hemisphere (56).

Another PET study reported lower binding of a dopamine D1 receptor ligand in the frontal cortex of bipolar patients, relative to normal controls. These findings suggest an abnormality of dopaminergic transmission as a possible etiology of bipolar disorder (85). Pearlson et al. (64) demonstrated that an increase in the Bmax of dopamine D2-type receptors was associated with the presence and severity of psychotic symptoms in both mania and depression but was not associated with severity of primary mood symptoms.

The many different (although overlapping) regions implicated in the functional neuroimaging studies reviewed above suggest that affective disorders involve dysfunction of integrated neural circuits or networks, rather than individual brain regions. Five recent reviews discuss the involvement of such neural circuits in affective disorders (22, 25, 44, 57, 86).

Ketter, for instance, summarizes the considerable literature on involvement of mesial cerebral structures in the mediation of emotional experience. He points out the importance of anterior paralimbic and prefrontal cortical activity in depression and explains the equivocal findings of different studies as resulting from disease heterogeneity among studies. He posits that baseline functional abnormalities in these structures might relate to diagnostic subtypes (44).

Drevets et al. (25) implicated two related loops: one involving the amygdala, mediodorsal thalamic nucleus, ventrolateral prefrontal cortex, and medial prefrontal cortex (PFC) and a second linking the striatum and ventral pallidum to the first circuit. Swerdlow and Koob (86) implicated the forebrain in dopaminergic circuits that affect limbic-thalamic-cortical loops, and are secondarily modulated by dopaminergic projections to the striatum, amygdala, and PFC.

Depue and Iacono (22) view mood disorder as a dysfunction of a behavioral facilitation system that normally provides motor, affective, and motivational components to achieve environmental engagement or withdrawal. They implicate circuitry influencing locomotor activity and incentive/reward behaviors in two dopaminergic projections from the ventral tegmental area. The first of these is the mesolimbic projection to the nucleus accumbens and amygdala; the second is the neocortical efferent system to the motor cortex, dorsomedial PFC, anterior cingulate, and orbital and dorsolateral PFC.

Mayberg (56) draws together evidence from neuroimaging studies of Huntington's disease, Parkinson's disease, and basal ganglia stroke, all of which are associated with depressive disorders. She feels that involvement of the paralimbic cortex (PLC) [i.e., orbital/cortical, inferior frontal, and anterior temporal cortices] is crucial. Two circuits may involve the PLC in this role; one orbitofrontal-basal ganglia-thalamic, and the second orbitofrontal-uncinate fasciculus-anterior temporal (the baso-temporal-limbic circuit) [Fig. 4].

The reviews above implicate many of the same structures and circuits, most frequently those involving portions of prefrontal cortex (especially orbital), basal ganglia, thalamus and amygdala. Many of these circuits involve neurotransmitter interactions between DA and 5-HT.

New neuroimaging technologies (e.g., functional magnetic resonance imaging [fMRI]) are being rapidly developed, as are new ways to analyze data from functional imaging modalities (e.g., statistical parametric mapping [SPM] of blood flow differences between patient populations), rather than conventional quantification of ambiguously drawn regions of interest (32, 33). These methods allow us to analyze functional circuits (rather than just regions) that may be implicated in the pathology of affective disorders. Such studies will undoubtedly enhance the quality of future study designs.

Theoretically, changes observed with neuroimaging techniques could be predictive in several senses. For example, the previously mentioned associations with biological response (e.g., dexamethasone suppression test status). Imaging changes could also predict the response to treatments of various sorts, using markers of treatment success, relapse, or recurrence and might also conceivably predict treatment side effects or complications.

Jacoby demonstrated in an outcome study that elderly depressed individuals with the greatest degree of ventricular enlargement on CT had excess mortality on follow-up (40). One question yet to be answered is whether this observation can be explained by a subpopulation of patients in whom depression appears early in the course of a neurodegenerative disease (such as Alzheimer's or multi-infarct dementia). If this is the case, it is consistent with CT studies suggesting that elderly depressed patients with reversible cognitive changes have brain alterations intermediate between those of elderly depressed and Alzheimer patients (61, 63).

Shima showed an association between ventricular enlargement on CT in depressed patients and treatment outcome (80). Coffey (14, 16) demonstrated that the presence of white matter lesions in elderly depressed patients was associated with prior poor response to antidepressant treatment, better subsequent response to ECT, and a greater likelihood of ECT-linked delirium. Wu (89) showed that specific regional cerebral metabolic changes (elevations in amygdala and cingulate) in depressives predicted a successful antidepressant response to sleep deprivation. Nasrallah (59) and Pearlson (62) found an opposite association between ventricular size and duration of hospital stay in mania. Investigators found that large ventricles in normal elderly subjects were a risk factor for the later development of a major depressive episode (40).

The area of prediction is one of great potential importance, although it has been insufficiently explored. Further follow-up studies are needed.

The generation of hypotheses from neuroimaging studies to provide guidance for the design of subsequent brain imaging studies, as envisaged by Cummings (18), is now gradually coming to pass.

Significant obstacles exist for documenting reproducible brain imaging changes associated with affective disorder, including the variable nature of the syndrome and its associations. Important lessons have been learned from earlier endeavors on how best to study the disorder and how to avoid potential pitfalls in future studies. These lessons include the need to use clear-cut hypotheses, standardized clinical assessments, instruments for adequate description and categorization of patient samples, and adequately sized, homogeneous patient and control populations. Given the diversity of the syndrome and the preliminary nature of the investigations, especially of quantitative functional studies, this whole area of investigation must be regarded as potentially fruitful but still in its early stages.

Directions for future investigations need to include quantitative dopamine and serotonin receptor PET/SPECT studies, and both structural and functional investigations of circuits (rather than just regions) implicated in the pathology of affective disorders. Once basic changes can be reliably identified, additional steps for future studies might include comparison of symptom severity with extent of brain changes, structure/function correlations, longitudinal studies examining progression vs. state dependency of brain alterations, and family studies examining those at genetic risk.

This work was supported in part by grants MH40391, MH43775, MH43326 and Johns Hopkins Outpatient Research GCRC 5MO10RR00035 (Dr. Pearlson) and grant 81BE-33483 of the Swiss National Research Foundation, grants of the CIBA and Roche Research Foundations (Dr. Schlaepfer).

published 2000