Abstract
For many years, investigators have been trying to identify the neuroanatomical structures responsible for the development of neuropsychiatry disorders, specifically depression and schizophreniform disorders. The available data were based on observations made in neurological patients who developed a psychiatric comorbid disorder following the neurologic insult. With the advances in high-resolution magnetic resonance imaging and functional neuroimaging studies, we have witnessed in the last decade a wealth of new data that identify structural neuroimaging changes in mesial temporal structures, prefrontal cortex and basal ganglia in major depressive disorders. The purpose of this article is to briefly review the publshed data on neuroanatomical structural changes associated with major depressive and bipolar disorders.
Keywords
INTRODUCTION
The advent of high-resolution magnetic resonance imaging of the brain (MRI) has revolutionized our research strategies in the study of pathogenic mechanisms of neurologic and psychiatric disorders. In the last decade, there has been an increasing body of literature on structural changes of the brain identified with high-resolution MRI studies in psychiatric patients. Such studies have been carried out in patients suffering from schizophrenia, panic and mood disorders. The purpose of this article is to briefly review the published data on neuroanatomical structural changes associated with major depressive disorders (MDD) and bipolar disorders (BPD).
The determination of whether MRI changes in the brain play a pathogenic role in the development of MDD or BPD can be approached directly, by comparing volumetric measurements and descriptive studies with high-resolution MRI of various neuroanatomical structures in a cohort of psychiatric patients and age and sex-matched controls. An indirect approach to this question is through MRI studies of neurological patients who develop a secondary mood disorder, in search of structural lesions whose presence, location and volume may be significantly associated with the occurrence of the mood disorder. The relatively high comorbidity between depressive disorders and most of the major neurologic disorders facilitates this type of investigation.
In a recent review of the literature, Sheline has suggested that in patients with major depression, morphologic and volumetric changes are likely to be present in one or more neuroanatomical structures that form a “limbic-cortical-striatal-pallidal-thalamic tract.” 1 A limbic-thalamic-cortical branch has been proposed as integrating one of its arms and includes the amygdala, hippocampus, medialdorsal nucleus of the thalamus and medial and ventrolateral prefrontal cortex. A second arm has also being suggested, running in parallel and linking the caudate and putamen and globus pallidus with limbic and cortical regions.
STRUCTURAL CHANGES IN NEUROLOGIC DISORDERS ASSOCIATED WITH MDD AND BPD
We will begin by briefly reviewing the literature on the relationship between MDD and BPD and the presence and location of structural changes in neurologic patients with stroke, multiple sclerosis, Alzheimer's disease (AD) and epilepsy. It should be pointed out, however, that while we are referring to “secondary” MDD and BPD, there are ample data to suggest multifactorial pathogenic processes operant in the development of these mood disorders in patients with the above cited neurological disorders. Furthermore, a bi-directional relationship has been identified between depression and some neurologic disorders. That is, while the occurrence of a neurologic insult increases the risk of developing depression, a history of depression is also associated with a greater risk of developing a given neurologic disorder. For example, in a population-based, case-control study of patients with newly diagnosed adult onset epilepsy carried out in Sweden, Forsgren and Nystrom 2 found that a history of depression preceding the onset of the seizure disorder was six times more frequent among patients than controls, in a second population-based, case-control study of the incidence of new onset epilepsy among adults aged 55 and older, Hesdorffer et al 3 found that compared to controls, patients were 3.7 more likely to have had a history of depression preceding their initial seizure (after controlling for the impact on seizures of medical therapies for depression). Similar results were found in a population-based study in Iceland of children with epilepsy. 4 In a similar type of study carried by the same authors in Iceland, they found that depression (?MDD) (meeting DSM-IV criteria) was four-times more common among children with epilepsy than among age and gender-matched controls. Interestingly, six centuries ago, Hippocrates had recognized this bi-directional relationship when he wrote: “melancholies ordinarily become epileptics, and epileptics melancholics: what determines the preference is the direction the malady takes; if it bears upon the body, epilepsy, if upon the intelligence, melancholy”. 5 By the same token, Larson et al, 6 for example, found in a population-based study that patients with depression were 2.6 more likely to suffer from a stroke than those without, even after controlling for other risk factors of stroke (i.e., diabetes, smoking, heart disease and hypertension).
Relationship between the location of stroke and depression
Depressive disorders are a common co-morbidity of stroke. Several cross-sectional studies have identified depression in 30 to 50 percent of patients following cerebral infarction. 7 In 1984, Robinson et al 8 reported a higher occurrence of left hemispheric stroke among patients with post-stroke depression (PSD), more often involving left frontal dorsolateral cortical regions. They also found that stroke in left (but not right) basal ganglia was significantly associated with PSD. Robinson and Szetela 9 had a few years earlier observed an inverse relationship between the severity of PSD and proximity of the stroke to the frontal pole. These findings have been replicated by other authors in only left hemispheric stroke, 10 while others found the same relationship among patients with left and right-sided stroke. 11,12
Some authors have not found a relationship between location of stroke and PSD, however. 13 In a review of the PSD literature, Cummings and Mega 14 found that the relationship between location of stroke and PSD was tied to timing of the onset of depression symptoms after the stroke. Thus, there is a higher frequency of left than right hemisphere lesions associated with PSD if PSD occurs in the early weeks after the stroke, while this relationship disappears if the onset of symptoms takes place 3 months after the stroke. Furthermore, these authors observed that when symptoms appear more than 1 year after the stroke, right-sided lesions are more frequent. In addition, small subcortical lesions of the left hemisphere are associated with a higher frequency of depression than right-sided sub-cortical lesions. 15 In terms of duration of depression, patients with subcortical, cerebellar or brainstem lesions have briefer depressions than patients with middle cerebral artery infarctions. 16
Structural changes in epilepsy and depressive disorders
As with stroke, depression is the most frequent psychiatric comorbidity in epilepsy patients 17 ; prevalence ranges between 20 to 50%, with the lower rates being identified among well-controlled patients and the higher rates among patients with refractory epilepsy. There is a general consensus among neurologists and psychiatrists that depressive disorders are more likely to occur among patients with partial seizure disorders of temporal and frontal lobe origin (i.e., involving limbic structures). However, little data are available on the relationship between structural MRI changes in epilepsy patients and depression. Quiske et al 18 studied 60 patients with temporal lobe epilepsy with the Beck Depression Inventory and MRI. They found that patients with mesial temporal sclerosis had significantly higher depression scores than other patients. Gilliam 19 also found significant correlation of the extent of H1-MR Spectroscopy abnormalities in the temporal lobes with the severity of depression measured with the Profile of Mood States scores. None of these studies found an association with lateralization of the epileptogenic zone and depression.
Structural changes in multiple sclerosis and depression
As with the other two neurologic conditions, the lifetime prevalence of a MDD as determined by several studies in people with multiple sclerosis is relatively high and ranges from 10–60%. 20 –22 Point prevalence rates are also reported in the range of 27%-54%, which are significantly higher than what is expected in the general population. 23 MRI studies in multiple sclerosis patients with unipolar depression have not yielded specific anatomical localization. 24 –26
Alzheimer's Disease
Migliorelli et al 27 investigated the prevalence of depression in AD patients with the use of structured interviews and found it to be 51%, with 23% of patients presenting a MDD. In population-based studies, 20% of patients with AD were found to present symptoms of depression, 28 while another population-based study in the United Kingdom found MDD in 24% of patients with AD. 29
The diagnostic MRI signature of AD is the presence of hippocampal atrophy. 30 As discussed below, hippocampal atrophy has also been identified in patients with primary depressive disorders. Yet, whether this abnormality plays any pathogenic role in the relatively high prevalence of depression among AD patients has not been established.
Summary
In conclusion, the occurrence of depression in neurologic disorders that present with easily detectable structural changes is not associated with lesions in a single neuroanatomical site. Temporal and frontal lobe structures are clearly involved with a significantly higher frequency among patients who go on to develop a depressive disorder.
STRUCTURAL MRI CHANGES IN PRIMARY DEPRESSIVE DISORDERS
As stated in our introduction, the last decade has seen an increasing literature on structural MRI changes of the brain in patients with MDD or BPD. The neuroanatomical structures affected include the hippocampal formation, amygdala, prefrontal cortex, basal ganglia including the caudate, globus pallidus and thalamic nuclei. The most frequent findings have consisted of a decrease in the volume of hippocampus, prefrontal cortex and basal ganglia and the presence of bifrontal areas of increased signal in white matter of frontal lobes. 1 Before describing the more relevant studies, we will examine potential mechanisms that could mediate a decrease in volume of these neuroanatomical structures in depressed patients.
In 1996, Sheline et al 31 compared the hippocampal volume of 10 patients with a history of MDD in remission with those of 10 age, sex and height-matched normal controls. Depressed patients had smaller hippocampi bilaterally. These findings were reproduced in a larger cohort of 24 patients and 24 controls. 32 In that study, these authors also found that core amygdala nuclei volumes correlated with hippocampal volumes.
The proposed mechanisms for the decrease in hippocampal volume include: 1) the development of atrophy mediated by high glucocorticoid exposure, and 2) an alteration in neurotrophic factors as a result of the mood disorder. These two processes may act simultaneously. Hecimovic et al 33 reviewed these proposed mechanisms in detail in a recent article.
Hippocampal atrophy mediated by glucocorticoid overexposure
Excessive activation of the hypothalamic-pituitary-adrena axis is observed in almost half of individuals with depression and is demonstrated by a decreased ability of dexamethasone to suppress plasma levels of Cortisol, ACTH, and β-endorphin. These changes are reversible to treatment with antidepressants. 34 In experimental studies with rats and monkeys, prolonged increased concentrations of glucocorticoids have been found to damage hippocampal neurons, particularly CA3 pyramidal neurons, possibly by reduction of dendritic branching and loss of dendritic spines that are included in glutamatergic synaptic inputs. 35 Hypercortisolemia has also been found to interfere with the development of new granule cell neurons in the adult hippocampal dentate gyrus. 36 Deleterious effects of chronic glucocorticoid exposure may lead initially to a transient and reversible atrophy of the CA3 dendritic tree, to an increased vulnerability to a variety of insults, and finally result in cell death under extreme and prolonged conditions. 37,38
In a recent study, Lucassen et al 38 carried out a neuropathology study of 15 hippocampi of patients with a history of MDD and compared them to those of 16 matched controls, and 9 steroid-treated patients. In 11 of 15 depressed patients, rare but convincing apoptosis was identified in entorhinal cortex, subiculum, dentate gyrus, CA1 and CA4. Apoptosis was also found in 3 steroid-treated patients and 1 control. However, no apoptosis of pyramidal cells in CA3 was identified. 39
Abnormal Neurotrophic Processes Related to the Depressive Disorder
Involvement of neurotrophic factors has been suggested as a potential pathogenic mechanism of depression and as playing a role in its treatment. 40 Acute and chronic stress decreases levels of brain-derived neurotrophic factor (BDNF) in the dentate gyrus, pyramidal cell layer of hippocampus, amygdala and neocortex. In contrast, such treatments increase BDNF in the hypothalamus of the rat. 41 Thus, a deficiency of BDNF may contribute to structural hippocampal changes. These changes are mediated by glucocorticoids and can be overturned with antidepressant therapy, as chronic administration of antidepressant drugs increases BDNF expression and also prevents a stress-induced decrease in BDNF levels. 42 There is also evidence that antidepressant drugs can increase hippocampal BDNF levels in humans. 43 These data indicate that antide-pressant-induced up-regulation of BDNF can hypothetically repair damage to hippocampal neurons and protect vulnerable neurons from additional damage.
The role played by these potential pathogenic mechanisms may help explain some of the variability in the reported changes in hippocampal volume among studies, as shown below.
Hippocampal Atrophy in MDD
As stated above, Sheline et al 31 were among the first investigators to have reported the presence of bilateral hippocampal atrophy in a study of 10 patients with recurrent MDD in remission. They also identified large (≥ 4.5mm in diameter) hippocampal low signal foci, and their number correlated with the total number of days depressed. In addition, the authors found a significant inverse correlation between the duration of depression and left hippocampal volume. Thus, the hippocampal atrophy was more likely to be identified in patients with a more chronic and active disease. Similar findings were reproduced in a larger study of 24 patients with remitted MDD compared to matched controls for age, sex and height. 32 In this study, the amygdala core nuclei volumes correlated with hippocampal volumes. A functional consequence of hippocampal damage was evidenced by lower verbal memory scores. MacQueen et al 44 compared hippocampal volumes and hippocampal dependent memory tests between 20 patients with first episode that was never treated and normal age-matched controls. The same comparisons were carried out between a second patient group that included 17 patients with recurrent depressive episodes and matched controls and the patients with a single depressive episode. While patients with a single and multiple episodes had verbal memory deficits, only patients with multiple episodes had hippocampal atrophy. As in Sheline's studies, 31,32 there was a significant correlation between the duration of the depressive illness and the degree of hippocampal atrophy. Bell-McGinty et al 45 compared the volume of hippocampal formation and entorhinal cortex between 30 patients with MDD and 47 matched controls. They found an inverse relationship between the volumes of hippocampus and entorhinal cortex and the time since the first life-time depressive episodes.
In a study of 38 patients with MDD, Vakili et al 46 found no difference in hippocampal volumes of patients and controls. However, their data suggested a possible relationship between the hippocampal volume and disease severity (left hippocampal volumes correlated with Hamilton Depression Rating Scale at baseline), as well as with treatment response (female responders to fluoxetine therapy had significantly higher right hippocampal volume). Shah et al 47 compared hippocampal volumes of 20 patients with treatment resistant MDD to 20 patients that responded to therapy and 20 healthy controls. Patients with treatment resistant MDD were more likely to have hippocampal atrophy.
The impact of therapy with antidepressants was confirmed by Sheline et al 48 in a recent study of 38 female outpatients with a history of MDD. These authors found a significant correlation between reduction in hippocampal volume and the duration of depression that went untreated. On the other hand, there was no correlation between hippocampal volume loss and time depressed while taking antidepressant medication or with lifetime exposure to antidepressants. Thus, these findings may suggest that antidepressants can have a neuroprotective effect during depression. In contrast to MDD, BPD has not been associated with the development of hippocampal atrophy in most published studies. 1
In a recent study, Posener et al 49 suggested the need to study the shape of the hippocampus in addition to the measurement of its volume, as the former can identify structural changes even in the absence of volumetric decrements. Using the method of high dimensional brain mapping, these authors generated 10 variables or components of the hippocampal shape in a study that compared high-dimensional mapping of 27 patients with MDD and 42 healthy controls. In depressed patients, these authors identified hippocampal deformation suggestive of specific involvement of the subiculum while finding no differences in hippocampal volumes between the two groups.
Changes in Amygdala
Data on the volumetric changes of amygdala of patients with MDD are less consistent than those of hippocampal formation. This is not surprising, as measurement of the amygdala and its nuclei is technically much more difficult than that of hippocampal structures. Sheline et al 50 compared the total volume of amygdala and that of its core nuclei in 20 patients with a history of MDD, free of any neurological disorder, and 20 matched controls. The core volumes of amygdala nuclei, but not its total volume were decreased bilaterally among patients. Conversely, Frodl et al 51 found increased amygdala volumes in 30 inpatients with a first episode of MDD, compared to matched controls. The authors attributed these changes to enhanced blood flow.
In a neuropathologic study of amygdala and entorhinal cortex, Bowley et al 52 carried out a neuronal and glial cell count in brains from 7 patients with MDD, 10 with BPD, and 12 control cases. The specimens of MDD patients and those of patients with BPD not treated with lithium and valproic acid had a significant reduction of glial cells and of the glial/neurons ratio in left amygdala and to a lesser degree in left entorhinal cortex.
Structural Changes in Frontal Lobes
Involvement of frontal lobes in depression has been recognized with functional neuroimaging and neuropsychological studies. Structural changes have been investigated in various structures of the frontal lobes, including the prefrontal cortex, cingulate gyrus as well as in their white matter. Bremner et al 53 compared the volume of orbito-frontal cortex and other frontal cortical regions between 15 patients with MDD in remission and 20 controls. Patients with depression had significantly smaller orbito-frontal cortical volumes. Coffey et al 54 also found smaller frontal lobe volumes in 48 inpatients with severe depression that had been referred for electroshock therapy compared to 76 controls.
Neuropathological studies have also documented structural cortical changes in frontal lobes of depressed patients. Rajkowska et al 55 found a decrease in cortical thickness, neuronal sizes, and neuronal densities in layers II, III and IV of the rostral orbito-frontal region in the brains of depressed patients. In the caudal orbito-frontal cortex there were significant reductions in glial densities in cortical layers V and VI that were also associated with decreases in neuronal sizes. Finally, in the dorsolateral prefrontal cortex there was a decrease in neuronal and glial density and size in all cortical layers.
Structural changes have been identified in depression occurring in elderly patients and merit a brief review. Lai et al 56 found smaller bilateral orbital frontal cortex volumes in 20 elderly patients with MDD than in 20 matched controls. Taylor et al 57 also found smaller orbitofrontal cortex volumes in 41 elderly patients with MDD than in 40 controls. Furthermore, these authors found that smaller volumes were independently associated with cognitive impairment. Kumar et al 58 found that the magnitude of prefrontal volume changes was related to the severity of the depression, as elderly patients with minor depression had lesser changes than those with MDD.
The presence of white matter hyperintensities in frontal lobes has been also associated with depression in the elderly. 59 Kumar et al 60 found that decreased frontal lobe volumes and the number of white matter hyperintensities on MRI represent relatively independent pathways to late-life MDD. Tupler et al, 61 on their part, compared the number and volumes of white matter hyperintensities on MRI between 69 patients with late-onset depression, 49 with early-onset depression and 37 controls. Patients with late-onset depression had more severe hyperintensity ratings in deep white matter than early onset and controls, while both groups of depressed patients had worse ratings than controls. Of note, left-sided white matter lesions were significantly associated with an older age at the onset of depression.
Structural changes of Basal Ganglia
The volumes of caudate and putamen have been found to be reduced in three studies of patients with MDD 61 –63 and are more likely to occur in late-onset depression. 1 In patients with BPD, no differences were found with controls in 3 studies, 64,65 while another suggested an increase in the striatal and pallidum volumes. 66
CONCLUDING REMARKS
The studies reviewed in this article clearly show that MDD can result in structural changes of various neuroanatomical brain regions. Duration and severity of disease appear to be related to hippocampal and frontal lobe volumetric changes as well as frontal white matter lesions. Treatment with antidepressant medication appears to have a neuroprotective effect at the level of the hippocampus. Furthermore, the changes observed in patients with MDD may not apply to those with BPD.
The actual prevalence of these structural changes is yet to be established, as most studies were carried out in small patient samples. Yet, these findings have significant clinical implications. First, they point to the need for early recognition of depressive disorders and prompt introduction of pharmacologic therapy. Second, these data should help eradicate the simplistic view often held by neurologists and clinicians in general, that depression in neurologic disorders is a “normal” reactive process. Such attitudes account for the failure of patients and family members to report symptoms of depression and for clinicians to investigate their occurrence.
These data may also help explain the worse prognosis of neurologic disorders in the presence of a depressive disorder. For example, in patients with multiple sclerosis increasing depression has been suspected to be a marker for disease progression. 20 In stroke patients, Starkstein et al 67 found that patients with major PSD and left-sided strokes had significantly more cognitive deficits than non-depressed patients with similar location and size strokes. Furthermore, the presence of PSD has been found to have a negative impact as well in the recovery of activities of daily living. Parikh et al, 68 for example, found that in-hospital PSD was the most important variable that predicted poor recovery in activities of daily living over a 2-year period, while Chemerinski et al 69,70 found that successful treatment of PSD with nortriptyline was significantly associated with their patients' recovery. Finally, the presence of depression is associated with a significant worsening in the course of Alzheimer's disease. 71
The question that needs to be answered is whether the structural changes associated with the MDD have an additive negative effect on the functional disturbances resulting from the primary neurologic disorder. The sad reality is that depression remains under-recognized in patients with neurologic disorder, as shown by Carson et al, 72 in a study of 226 consecutive patients seen at a neurology clinic. A MDD was identified in 24%. Eight months later, 20% continued suffering from major depression. Hopefully, awareness of the potential structural changes resulting from depression will avert this type of occurrences.