2.1 Introduction

Depressive disorders are common conditions with a life prevalence of 15% in high-income countries (1) and significant economic implications for individuals and society. Major depressive disorders have negative repercussions on the overall quality of life of the people affected with an excess number of years lived with a disability (2). Although effective treatment is available, up to 65% of individuals do not fully respond or continue to experience residual symptoms, which contribute to significant disease burden (3). It is essential to improve our understanding of the neuroanatomy of depressive disorders and the functional implications to develop new targets for more efficacious treatments.

This chapter reviews the current neuroanatomical evidence for abnormalities identified in major depression by focusing on selected research findings emerging from structural neuroimaging and postmortem studies. It also gives a perspective on the role played by the hippocampus in view of the described effects of stress on this region (4).

2.2 Structural Magnetic Resonance Imaging in Major Depressive Disorders

Since the introduction of magnetic resonance imaging (MRI) in the 1990s, several hundred cross-sectional studies have investigated structural changes in depressive disorders mostly in comparison with healthy controls by adopting a region of interest (ROI) approach (which involves delineation of anatomical boundaries of selected ROIs) or by using voxel-based morphometry (VBM) (which is based on measurements of anatomical differences of volume at voxel level across the whole brain). According to structural MRI and differently from bipolar disorder, major depression is characterized not by global brain volumetric reduction but by regional morphometric changes in a number of brain regions implicated in mood regulation (5–7). A number of meta-analyses of MRI studies have reported consistent volumetric reductions in prefrontal areas, more extensively in the orbitofrontal and anterior cingulate cortices and limbic regions such as the amygdala and hippocampus. Other areas in the prefrontal cortex also commonly described include the dorsomedial and ventromedial prefrontal cortices (5, 6, 8). Recent work reporting data from twenty cohorts of patients worldwide indicated that these regions are also characterized by cortical thinning, and aside from the orbitofrontal cortex and anterior cingulate cortex, other regions include the posterior cingulate cortex, temporal lobes, and insula (9). Volumetric reduction in the temporal regions, including the insula, has also been reported in an independent study (10). Their involvement is hardly surprising considering that the temporal lobes and the insula are known to be involved in automatic responses, including multisensory recognition and participation in the processing of emotions (11, 12).

The role of subcortical regions in the neuroanatomy of depression has been of great interest in view of the role of the limbic system in the processing of emotions. Several studies have evaluated structural changes in the hippocampus complex, the most investigated limbic region in major depression (5), whereas the amygdala is the second brain region most extensively researched in this type of studies in depressive disorders (6). Findings in the literature tend to vary from no volumetric difference at all to increase or decrease in volume in comparison with healthy volunteers (13). This variability can be explained by the significant heterogeneity in the patients included, the methods applied to measure differences, and the presence of comorbidities (6, 13). Bora and others, for instance, found evidence of amygdala morphometric reduction in depression when comorbid anxiety was present (8). Overall, the most consistent evidence suggests a volumetric reduction in this region (5, 10, 14). Other regions involved in depression include the basal ganglia (ventral striatum), thalamus, pituitary region, corpus callosum, and cerebellum (5, 10, 14)).

Studies that have used diffusion tensor imaging techniques in major depression have identified white matter disconnectiviy in several areas associated with decreased fractional anisotropy. These have included bilateral frontal white matter, right fusiform gyrus, right occipital lobe and the right inferior longitudinal fasciculus, the right inferior fronto-occipital fasciculus, the right posterior thalamic radiation, and interhemispheric fibers crossing the genu and the body of the corpus callosum (15). Consistent with earlier MRI studies in affective disorders (5, 16), a recent systematic analysis of diffusion tensor imaging studies carried out by using a conservative approach indicated that the most reliable finding of decreased functional connectivity is located in the genu of the corpus callosum, crucially important in interhemispheric prefrontal connectivity relevant to affective regulation (14).

In summary, current evidence suggests that in the absence of global gray matter loss in major depression, several regions have been implicated in the neuroanatomy of this disorder. Involved brain areas include the prefrontal brain, temporal regions, and the insula. White matter disconnectivity in the genu of the corpus callosum is the most reliable finding identified by using diffusion tensor imaging. Within the limbic system the amygdala and the hippocampus are the most investigated regions. Hippocampal volumetric reduction is considered the most consistent finding.

2.3 Postmortem Studies in Depression

Posthumous evidence in major depression is based on a relatively small number of studies. Direct neuropathological abnormalities have been observed in groups of individuals with mood disorders compared with healthy controls and in some instances in relation to other conditions like schizophrenia. A reduction in glial cell number and/or density in major depression, especially in individuals with positive family history for mood disorders, has been described in the subgenual prefrontal cortex (17), supracallosal areas (18, 19), and dorsolateral and orbitofrontal regions (18, 20). In these brain areas, other abnormalities have included neurons with smaller cell body size versus larger body size found in cortical layers II, V, and VI of the prefrontal cortex, which may account for volumetric reduction in depression (18, 20, 21).

In the hippocampus complex, an increase in glial cell density with no changes in the size of glial nuclei has been associated with a decrease in the size of the soma of pyramidal and granule cells (22). Further evidence supports neuronal loss in advancing age, independent from neurocognitive degeneration and cerebrovascular pathology (23). A reduction in glial cell density has also been reported in the amygdala (24), and reduced hypothalamic volume has been described in the literature (25). The ventrolateral component of the dorsal raphe nucleus has also been demonstrated to be reduced in volume in patients with mood disorders with a smaller number of triangular neurons in this region (26). There is also indirect evidence emerging from markers of neuronal activity. For example, lower density of Immune reactive calretinin, a marker of GABAergic activity in neurons and glia, has been described in layer I of the dorsolateral prefrontal cortex (27). Further evidence of disruption in the glutamatergic function includes abnormalities in glutamate signaling genes SLC1A2, SLC1A3, and GLUL found in the locus coeruleus (28). Other abnormalities involve growth factor genes FGFR3 and TrkB, and other genes expressed in the astrocytes (28). Some of the limitations of this literature are the possible contamination of the sample with other conditions, the small sample size and the retrospective nature of diagnoses and cause of death, limited clinical and treatment history, and the possibility that tissue changes might have occurred prior to cellular analysis.

In summary, there is evidence of neuropathological abnormalities in depressive disorders affecting primarily prefrontal areas and limbic regions and including both neurons and glial cells, pointing toward a reduction in brain volume.

2.4 Evidence of Structural Changes in Depression Following Prolonged Stress, and the Role of the Hippocampus and Hypothalamic–Pituitary–Adrenal Axis

One of the etiological models of depression central to the pathophysiology of the disorder proposes that prolonged stress mediates over-activation of the hypothalamic–pituitary–adrenal (HPA) axis (29). Stress and glucocorticoids, the end product of HPA axis hyperactivity, generate a cascade of intracellular events known to downregulate brain neurotrophic factors (e.g., nerve growth factor), brain-derived neurotrophic factor (BDNF), and neurotrophin-3 (30). BDNF is essential for survival, differentiation, and functioning of neurons in the brain (31). Downregulation of BDNF demonstrated in rodents exposed to prolonged stress is believed to reduce cellular resilience and to lower the threshold for hippocampal cellular damage in animal models of depression (32). Stress and high levels of circulating corticosteroids may be responsible for the volumetric reduction, which is measured in a number of regions, including the hippocampus (33, 34), and could be the result of intracellular mechanisms leading to homeostatic modification and cellular damage (32, 35). Cellular alteration is likely to be mediated by postsynaptic mechanisms in response to receptors’ activation. Animal studies suggest that BDNF-mediated long-term synaptic potentiation could act via N-methyl d-aspartate (NMDA) receptors (36). NMDA receptors’ stimulation appears key to hippocampal toxicity and cellular stress and includes NMDA receptors’ direct interaction with the neurotransmitter glutamate facilitated by elevated glucocorticoid levels (32, 37). Increased expression of BDNF and its receptor TrkB in the dentate gyrus and the CA3 region of the hippocampus following sustained administration of antidepressant treatment in animal studies (38) might be key to treatment response in humans. This effect appears mediated by cyclic adenosine monophosphate (cAMP) dependent phosphorylation in postmortem studies (39). cAMP-response element-binding (CREB) protein mRNA levels in the CA1, CA3, and dentate gyrus regions of the hippocampus (35, 40) have been shown to be associated with hippocampal neuronal sprouting and neurogenesis and could explain recovery following treatment (35). Supporting evidence suggests low circulating levels of BDNF in unmedicated currently depressed patients correlating with hippocampal volume (41) normalization following administration of an antidepressant and clinical remission (42). Indirect evidence from MRI longitudinal studies also supports greater hippocampal volumes in medicated depressed patients that appear to be associated with remission at one year follow-up (43). Although findings in the literature are not always consistent, discrepancies could be expression of depression heterogeneity, inclusion/exclusion of comorbidities, sample size, length of pharmacological treatment, and methodological differences. Lai and others, for instance, did not find that six-week treatment with duloxetine affected gray matter in the hippocampus in a sample of depressed patients with comorbid panic disorder (44). Evidence from cross-sectional studies overall supports greater hippocampal volumes following response to treatment and remission and greater hippocampal volumes in responders to pharmacological treatment in comparison with nonresponders (45). Furthermore, lower hippocampal volume at baseline has been shown to predict non-remission to antidepressant treatment in a study in older adults (46) and first presenters (47). Other postsynaptic mechanisms linked with neurotrophic effects related to efficacious treatments of mood episodes include lithium-mediated gene expression, resulting in the induction of protein bcl-2, as shown in the frontal cortex and hippocampus (and also cerebellar granule cells and striatum) of rodents and in human neuroblastoma cells (48–50), and the modulation of protein kinase C (PKC) alpha and epsilon, as shown in rodent frontal cortex and hippocampus following administration of lithium and sodium valproate (51). Finally, lithium and antidepressants modulate adenylate cyclase systems by reducing receptor/G protein coupling (52).

In summary, a range of animal and human studies support the notion that stress can induce neuroendocrine changes occurring in depressive disorders and that a cascade of postsynaptic events is likely to increase the susceptibility of hippocampal structures to cellular damage. Although the molecular nature of volumetric loss detectable with MRI in depression is not clearly established, pharmacological treatment could be implicated in volumetric normalization associated with treatment response. Conventional pharmacological compounds associated with treatment response are believed to act on intracellular mechanisms and neurotrophic pathways.

2.5 Hippocampal Abnormalities: State versus Trait Marker for a Depressive Episode

Volumetric reduction in the hippocampus is the most replicated finding in major depression, consistent with the involvement of this region in the processing of emotions (11). Evidence from meta-analyses suggests that the effect is larger with increasing proportion of patients currently depressed (5). This notion is supported by experimental data indicating a reduction in the gray matter of the hippocampus in currently depressed subjects in comparison with healthy controls (53–60), also in unmedicated patients (44, 61–64), and not in individuals with remitted depression (64). The association between remission and full volumetric recovery is complicated by several factors impacting on the overall measured effect, such as illness duration (65, 66) or age of onset (e.g., proxy of illness duration) (67), number of episodes (68), severity of symptoms (45), the presence of treatment resistance, or a chronic illness course (which could be expression of a different biological subtype of depression, not necessarily representative of the disorder) (43). There may also be some biological characteristics intrinsic to remission including scar effects. Salvadore and others, for example, compared unmedicated remitted patients with a chronic course of illness versus currently depressed and healthy controls and reported prefrontal gray matter changes only in currently depressed patients in comparison with both the other groups (69). Another factor is the possibility that bipolar disorder can present as unipolar depression at disease onset, which could dilute morphometric differences, and greater hippocampal reduction has been associated with major depression in comparison with bipolar disorder (13, 70). Some research studies include participants with positive family history for major depressive disorders, and often no sufficient attention is given to the presence of early developmental adversity. Increasing evidence suggests that these are very important contributors to hippocampal integrity. There is in fact evidence of reduced hippocampal volume in subjects with parental history of depression prior to developing major depression (71, 72), whereas a known independent interaction exists between hippocampal volume, early developmental adversity, and risk of developing depressive symptoms in the future (71, 73). The increased risk of developing depression in the presence of early life stressors has been shown to be attributable to the excess of BDNF Val66Met variant (the BDNF transcript with the amino acid valine in position 66 instead of methionine) resulting in decreased secretion of BDNF and volumetric reduction in the hippocampus in individuals who experienced traumatic events. Another mechanism of action has been attributed to the predominance of the short allele variant of the serotonin transporter, resulting in less effective serotonergic neurotransmission and vulnerability to stress (74). In this context, volumetric reduction in the hippocampus could be the result of neurotoxicity induced by stress-related glucocorticoids release and decrease in neurotrophic factors, with evidence of reduction in the size of the soma of hippocampal pyramidal and granule glial cells (22) and decreased density of dendrites and spines in the hippocampal subregion CA3 detected in subjects with higher anxiety and depression scores (75), potentially in combination with a stress-prone less-efficient serotonergic neurotransmission.

In summary, there is compelling evidence suggesting hippocampal volumetric reduction in symptomatic individuals. The resolution of symptoms suggests a degree of amelioration supportive of a “state effect.” Moderating factors of morphometric normalization include the number of episodes, age of onset/duration of illness, severity of the disorder, family history, the experience of traumatic events, especially at young age, and cumulative effects of pharmacological and psychological treatment.

2.6 Neuroanatomical Circuitry Involved in Major Depression

Several brain regions have been implicated in major depression, which are part of circuits involved in instinctual behaviors, homeostatic mechanisms, and covert processing of emotional information mostly located in subcortical regions including the hippocampus, amygdala, hypothalamus, thalamus, with brain regions located in the forebrain, inner temporal lobes (insula), and the cingulate area (76). The overtone of emotional regulation is the result of the interaction between subcortical regions and neocortical associative areas where information regarding the “state” of the body is integrated, resulting in a range of cognitive and executive functions filtered through socially relevant norms (77). This is consistent with animal and lesion studies that suggest that pivotal brain areas participating in mood regulation are located in the prefrontal cortex and that important pathways connect the orbitofrontal cortex with the thalamus and hypothalamus/brain stem. These studies also suggest that the ventromedial prefrontal cortex is involved in appraisal of behavioral consequences in humans. Lesions in this region abolish the normal automatic visceral response to emotive stimuli, and affected individuals become unaware of long-term consequences of their behaviors (78, 79).

Brain regions where morphometric changes, often in the direction of a reduction, have been reported in major depression include the ventromedial prefrontal cortex, orbitofrontal cortex, anterior cingulate cortex, posterior cingulate cortex, temporal areas, basal nuclei, and the limbic system.

From the medial prefrontal cortex, projections reach the superior temporal gyrus, anterior cingulate cortex, entorhinal cortex, para-hippocampal gyrus (80), and the limbic system (hypothalamus, periaqueductal gray matter, and amygdala) to exercise control over visceral functions in relation to mood state (81). Direct somatotopic reciprocal cortico-thalamic connections are described to originate from medial and lateral prefrontal cortices with an indirect thalamic pathway via striato-pallido intermediate stations. In these networks, medial components of the prefrontal cortex (orbito-medial and orbitofrontal) project to the ventromedial part of the striatum, including caudate nucleus and putamen, and to the paraventricular thalamic nucleus (which also exchange connections with the ventromedial striatum) (81). The paraventricular thalamic nucleus is important in circuits of mood regulation because it is involved in autonomic responses and stress regulation. Such responses require interactions with limbic regions including amygdala, hypothalamus, and periaqueductal gray matter (81, 82).

Mayberg described the coexistence of two virtual anatomically related systems (83, 84): a system that primarily processes cognitive information in response to emotional stimuli and an interoceptive system that mediates covert responses (83, 84). In this model, subcortical areas, which include thalamus and ventral striatum, are implicated in the implicit processing of novel emotional and nonemotional information together with limbic regions such as the amygdala and the hippocampus complex (83, 84). Prefrontal regions, particularly medial cortical areas, exercise cognitive control and appraisal of emotional states (85).

In summary, the effective processing of emotional information is the result of a synergy between subcortical regions integrating information with cortical areas. Whereas subcortical regions host more automated functions, cortical areas are essential associative areas coordinating emotional information with the environmental milieu of the individual. The frontal brain, particularly the orbito-medial prefrontal areas, coordinates emotional responses with executive functions so to guide the provision of coherent behavioral responses.