Major depressive disorder (MDD) has until recently been conceptualized as an episodic disorder associated with ‘chemical imbalances’ but no permanent brain changes. Evidence has emerged in the past decade that MDD is associated with small hippocampal volumes. This paper reviews the clinical and biological correlates of small hippocampal volumes based on literature searches of PubMed and EMBASE and discusses the ways in which these data force a re-conceptualization of MDD. Preclinical data describe the molecular and cellular effects of chronic stress and antidepressant treatment on the hippocampus, providing plausible mechanisms through which MDD might be associated with small hippocampal volumes. Small hippocampal volumes are associated with poor clinical outcome and may be a mechanism through which MDD appears to be a risk factor for Alzheimer's disease. The pathways through which stress may be linked to MDD, the emergence of chronicity or treatment resistance in MDD and the association between MDD and memory problems may be at least partially understood by dissecting the association with depression and changes in the hippocampus. MDD must be re-conceived as a complex illness, associated with persistent morphological brain changes that are detectable before illness onset and which may be modified by clinical and treatment variables.
Lively debates in the pages of psychiatric journals have recently challenged whether there is any evidence that studying biological mechanisms of psychiatric illness has changed clinical practice. We do not wish to wade into the debate. We do suggest that studies focusing on the hippocampus (HC) in major depressive disorder (MDD) enable an integration of basic experimental, neuroimaging and clinical findings into plausible biological models that are influencing the ways in which we conceptualize the nature and course of MDD.
Convergent lines of research implicate the HC in the pathogenesis of MDD. First, MDD is clinically recognized as a highly stress-sensitive illness1 and the HC is a highly stress-sensitive brain region.2 Stress, including psychological or psychosocial stress, is associated with structural changes to the HC.3, 4, 5 Conditions of chronic hypercortisolemia, such as Cushing's disease are associated with HC atrophy that is reversible following normalization of cortisol levels,6 supporting the hypothesis that stress-related changes in the HC are observable in humans as well as other species. Effective antidepressant treatments may ameliorate stress-associated changes in the HC.7, 8
Second, when the various domains of cognitive function are assessed in patients with depression, their greatest degree of impairment is on memory measures that are heavily dependent on HC function.9 Poor memory performance is associated with small volumes of HC gray matter in non-clinical samples as well.10 Furthermore, while there has been a debate about whether mood congruent memory biases are present for implicit material, such biases can be reliably demonstrated in explicit memory tasks, which are heavily dependent upon medial temporal lobe structures, including the HC.11
Third, the HC is one of several regions, including the dorsomedial and dorsolateral prefrontal cortex, the anterior cingulate cortex and the amygdala12 that connectivity studies have identified as components of a network that is dysregulated in MDD.13 Evidence from neuroimaging, neuropathological and lesion analysis studies further implicates limbic–cortical–striatal–pallidal–thalamic circuits, including prefrontal cortex, amygdala, hippocampal subiculum, ventromedial striatum, mediodorsal and midline thalamic nuclei and ventral pallidum, in the pathophysiology of mood disorders. These networks normally regulate aspects of emotional behavior and fit therefore well to pathophysiological concepts of mood disorders.14, 15 The hippocampus has a role in these networks with its connections to and from the other brain regions16 (Figure 1b).
Finally, many structural imaging studies have reported that the HC is small in patients with MDD. A recent meta-analysis of hippocampal volumes in patients with MDD confirmed that patients had hippocampal volumes that were approximately 4–6% smaller than matched control subjects in the left and right HC. The analysis included 1167 patients and 1088 control subjects, across a wide range of ages from pediatric to geriatric populations. Conclusions from this meta-analysis were consistent with the findings of earlier meta-analyses of HC volume in patients with MDD17, 18 (for anatomical location of the HC see Figure 1a). A summary of the evidence of HC involvement in MDD is provided in Figure 1.
Despite consistency in the studies implicating the HC in MDD, questions remain regarding the pathophysiological underpinnings of these findings, the clinical factors that are associated with dysregulation in the HC, the clinical significance of HC changes in MDD and the likelihood that medication can ameliorate HC-related dysfunction in patients with MDD. We review the data relevant to each of these issues.
Materials and methods
A search of the PubMed and EMBASE databases till November 2009 was conducted using the search terms hippocampus/or HC and depression and MRI. An overview of the current literature is provided.
Clinical and demographic variables influencing hippocampal volumes in MDD
The question on the clinical and demographic variables that influence HC volumes has been addressed in detail in a recent review so that we do not want to extend this section here.28 The prevalence of MDD is twice as high in women as in men,29 with this sex differential becoming apparent after girls achieve puberty. Paradoxically, perhaps, in animal models there is evidence that estrogen may protect the hippocampus from both the behavioral and structural changes observed in male animals after exposure to stress. Acute stress has different effects in male and female rats on performance of tasks that involve hippocampal function, such as classical eye-blink conditioning or Y-maze or Morris Water Maze.30 When animals are chronically stressed, male rats show a decrease in dendritic arborization, whereas females show no dendritic atrophy.31 Estrogen treatment in OVX rats protects the hippocampus from neuronal loss due to chronic stress exposure.32 The relations between stress, depression, hippocampus and estrogen levels require further investigation. In clinical studies, small HC volumes in patients compared to control subjects are apparent when study groups are comprised of women only. The average reduction in HC volumes in samples that included women only is 4.5 and 5.7 percent, respectively for the left and right HC, values that are comparable to studies in which men and women were examined as a group.28
With respect to clinical variables, small HC volumes in MDD have been linked to high levels of depressive symptoms,33, 34 early age at illness onset,35, 36, 37, 38 non-responsiveness to treatment,33, 39, 40 long duration of untreated days of illness,41 high illness burden,42, 43, 44, 45 positive history of childhood abuse,46 and high levels of anxiety.47, 48 Significant inconsistencies exist, however, particularly with respect to the relation between clinical variables (for example, age at illness onset, duration of illness, etc.) and reductions in HC volume. For example, numerous studies report no relations between HC volumes and age at onset.39, 41, 49, 50, 51, 52 Although some studies showed a correlation between small HC volumes and more severe depressive symptomatology,33, 34 the majority of structural studies that have investigated the relations with depression severity have not found a significant correlation between high ratings of depressive symptomatology and smaller HC volumes.43, 49, 50, 53 When examined in the aggregate, Mc Kinnon and colleagues reported that differences in HC volume were apparent only for patients with an illness duration of at least two and one half years or more than one episode of depression.28 Some studies examining HC volume in patients with a long history of depression find a correlation with the amount of time spent symptomatically ill, suggesting that successful treatment and symptom reduction may halt a neurodegenerative process.42, 43, 54 Another group reported that volume in the left HC showed a marginally significant relation with duration of illness in a group of drug-free patients with MDD.34 Other studies, however, have not found evidence of an effect of illness duration on HC volume.35, 36, 39, 49, 50, 55, 56, 57, 58, 59, 60, 61 These negative findings may reflect, in part, small sample sizes,51, 60 the inclusion of patients with bipolar disorder,35 or samples comprised of primarily young adult patients,51, 57, 59 and of patients with a low number of illness episodes.57
Biological mechanisms that may contribute to small hippocampal volume in MDD
Genetic influences on hippocampal volumes in MDD
Hippocampal size is highly genetically determined.62, 63 An association between heritability and brain size is supported by twin studies64 and a wide range of imaging genetic studies.65 Alternative phenotypic markers for genetic association studies that are more closely related to the underlying neurobiology of the disease are increasingly being used for genetic association studies. This partly derives from a need to address problems of clinical heterogeneity in psychiatric disorders, and partly from the need to delineate the functional consequences of identified risk variants at the level of brain structure and function. Imaging genetics facilitates an elucidation of the impact of genes at the level of the brain that can then be extended to the pathophysiology of the disease.
Elimination of 5-HT from the synaptic cleft is mediated by a single protein, the 5-HT transporter (5-HTT), which determines the size and duration of the serotonergic responses.66 The promoter region of 5-HTT has a polymorphism that results in allelic variation in functional 5-HTT expression.67 The long (l) allele is associated with production of more 5-HTT transcript than the short (s) allele and hence more functional 5-HT uptake than the (s) allele.67 The short (s) allele of the 5-HTT polymorphism is associated with anxiety, depression and aggression-related personality traits in some reports,66, 67 but a recent meta-analysis did not confirm an association between 5′HTT polymorphism and vulnerability to MDD.68 The importance of variation in 5-HTTLPR to HC volume in patients with MDD has been noted in three published reports that utilized diallelic analysis and two reports using triallelic analysis.37, 69, 70
Brain-derived neurotrophic factor (BDNF) regulates neuronal survival, migration, phenotypic differentiation, axonal and dendritic growth and synapse formation.71 BDNF is also a key regulator of synaptic plasticity and behavior72 and may be important for memory acquisition and consolidation.73 Interestingly, a reduction of BDNF expression—introduced through BDNF knockdown by RNA interference and lentiviral vectors injected into specific subregions of the hippocampus—in the dentate gyrus, but not the region CA3 of the hippocampus, reduces neurogenesis and affects behaviors associated with depression.74 A single-nucleotide polymorphism in the pro-domain of BDNF converts the 66th amino acid valine into methionine (Val66Met). This Val66Met polymorphism affects dendritic trafficking and synaptic localization of BDNF and impairs its secretion. The Val66Met SNP is associated with deficits in short-term episodic memory.75 Healthy met-BDNF carriers have relatively small HC volumes;76, 77 an effect of the Met-BDNF allele on HC volume is also apparent in patients with MDD.55 A recent study in healthy volunteers reported that subjects carrying the Met-BDNF allele have smaller HC volumes when they have more sub-threshold symptoms of depression and a higher extent of neuroticism.78
Gray matter volumes in temporal lobe regions including the HC were associated with glycogen synthase kinase-3 polymorphisms in a large sample of depressed patients and healthy controls.79 The underlying neurobiological background of this finding is unknown; it is interesting, however, that glycogen synthase kinase-3 activity might be associated with therapeutic effects of antidepressants and lithium.79 Thus, these first genetic imaging studies provide evidence that genetic polymorphisms are relevant for neurobiological mechanisms in MDD, possibly in part through their contribution to HC volumes.
Effects of stress on the hippocampus
The HC is vulnerable to stress, particularly during the early developmental period,80 but there may be species-specific and time-dependent variation in how stress affects the brain. While separation of monkeys from their siblings or mothers seems not to determine HC volumes,62, 81 chronic stress exposure in rats or tree shrews does result in reduced HC volumes.82, 83 Although transient mild stress may enhance HC function,84 chronic or severe stress disrupts HC-dependent memory in experimental animals (reviewed in Sapolsky85). Extended or high-dose treatment with glucocorticoids has a negative effect on HC-dependent memory in both animal models,86, 87 and humans.88, 89
Sustained elevations in stress or glucocorticoids impair the HC at the level of morphological neuroplasticity (reviewed in Sapolsky90). High levels of glucocorticoids91 or behavioral stress92, 93 result in atrophy and retraction of the apical dendrites of HC pyramidal cells. A reduction in the amount of neuropil without frank cell loss has also been observed, a finding that appears consistent with observations from post-mortem studies of the HC in patients with MDD.94
Glutamatergic excess may also contribute to cell damage and even to cell death.85, 90 Stress-induced increases in extrasynaptic glutamate may perturb the balance between synaptic and extrasynaptic NMDA tone and have a net inhibitory effect on the mechanisms of synaptic plasticity and neuronal growth and survival (reviewed in Pittenger et al.95).
Signaling pathways implicated in neuroplasticity target, among other downstream targets, genes for growth factors such as BDNF.96 Acute and chronic stress reduces neurogenesis in the rodent HC (reviewed in Dranovsky and Hen; Duman19, 20). As neurogenesis may be required for the behavioral response to antidepressants in rodents8 and impaired neurogenesis may represent a pathophysiological component of MDD,20 suppressed neurogenesis may represent another way that the effects of stress on mechanisms of neuroplasticity contribute to the development of MDD. The specific functional role and relevance of new neurons in the HC is not firmly established, but a link between neurogenesis and the learning-related functions of the HC is an intriguing possibility.97 Some computational theories of HC function predict a role for new neurons in HC-dependent learning.98 Figure 2 describes the putative links between the molecular, cellular and functional levels of analysis in studies of HC pathophysiology in MDD.
The effects of early childhood stress on HC volumes
A recent meta-analysis based on 14 250 participants demonstrated that early adverse life events confer significant risk of subsequent MDD.68 Youth and adults exposed to early-life adversity appear to have small HC volumes;99 for an overview on HC volume changes in post traumatic stress disorder see also Bremner et al.100 and Karl et al.101 Although a large population-based study from Australia did not find small HC volumes in subjects exposed to adverse childhood events, this study used voxel-based morphometry to conduct a whole brain analysis and the necessary corrections for multiple comparisons can decrease the probability of detecting changes in small structures like the HC.102
Vythilingam et al.46 reported 18% smaller left HC volumes in patients with early life trauma and MDD compared to non-abused patients with MDD who did not differ from healthy controls (see Table 1). Moreover, patients with MDD and childhood emotional neglect had smaller HC volumes compared to patients with MDD without emotional neglect. The small volumes were more pronounced in the left HC and in men.103 In contrast, however, HC volumes did not correlate significantly with early life events in a sample of women with remitted unipolar MDD.104 A potential explanation for the discrepancy in reports might be that standard correlational analyses of a continuous measure of early adversity might not show significant results if the underlying association between adversity and HC volumes, for example, is not described by a linear function. Comparing group differences between participants with and without abuse/neglect might then be an advantage for detecting relations that are best described by a non-linear function. Furthermore, gender effects have not been assessed in the majority of studies investigating the association between stress and HC volumes (although see Frodl et al.103), so that further investigation is necessary to understand whether there are gender-specific effects in the impact that stress has on the HC.
In summary, there is evidence from studies of healthy subjects, patients with PTSD and patients with MDD that early childhood stress is associated with small HC volumes. Whether variables such as the type of stress, the developmental period during which a child experiences the stress or the chronicity and intensity of the stress can determine HC structure and function later in life needs further investigation.
Stress × gene interactions on HC volumes
Early life stress and variations in the serotonin transporter promotor polymorphism 5-HTTLPR may interact to predict development of MDD.105 An increased risk of depression was detected in maltreated children homozygous for the S-allele.106 In a recent study, patients carrying the s-allele had smaller HC volumes when they had a history of emotional neglect compared to patients who had only one risk factor (environmental or genetic). Childhood stress further predicted HC alterations independent of genotype.107 Meta-analytic approaches suggest, however, that 5-HTTLPR may not further increase the risk for MDD in subjects who experience critical life events.68
In a recent study of 89 participants, significant interactions between BDNF genotype and early life stress were apparent in HC and amygdala volumes, heart rate and working memory. The investigators used structural equation modeling to investigate the pathways through which BDNF genotype and early life stress interact to produce effects on brain structure, body arousal, emotional stability, which in turn predict alterations in symptoms and cognition. Structural equation modeling suggested that the combination of Met carrier status of the BDNF polymorphism and exposure to early life stress predicted reduced HC volumes, associated lateral prefrontal cortex volumes and, in turn, higher depression.10 The gene × environmental interaction seems to be particularly relevant in patients with MDD. The relevance of small HC volumes as a risk factor for MDD has now been documented by both longitudinal studies and studies of subjects at high-risk for MDD as reported in the following section.
The clinical relevance of small hippocampal volumes in MDD
Alterations of hippocampal volumes in high-risk subjects
A period of high risk for MDD begins in the early teens, and risk continues to rise in a linear fashion throughout adolescence, with lifetime rates estimated to range from 15 to 25% by late adolescence.108, 109
In a recent study, small HC volumes were apparent before the manifestation of clinical symptoms of MDD in at-risk adolescents, particularly in those who experienced high levels of adversity during childhood. Both early-life adversity and smaller HC volume were associated with a higher probability of depressive episodes during prospective follow-up.110 Moreover, 23 daughters of mothers with a history of MDD had reduced HC volumes compared to 32 daughters of mothers with no history of psychopathology, indicating again that neuroanatomic anomalies associated with depression may precede the onset of a depressive episode and influence the development and course of this disorder. In another recent study, adults at-risk for MDD had HC volumes that were not significantly different from those already affected with MDD, suggesting that small volumes may precede the onset of MDD and might render subjects more vulnerable to MDD.111 Early life adversity appears to contribute to both small HC volumes and increased risk of MDD; whether changes in HC volumes represent a biological mechanism through which early life adversity is transduced into risk of MDD remains to be confirmed.
Associations between clinical response and hippocampal volume
There are a number of studies reporting associations between HC volume and probability of achieving and sustaining a clinical remission (Table 2). In one such study of patients with geriatric depression, patients with right HC volumes in the lowest quartile of the sample were less likely to achieve remission compared to those with HC volumes in the highest quartile.40 Another study reported that women who responded to 8 weeks of fluoxetine had larger right HC volumes than non-responders.33 More recently a study reported that small HC volumes predicted low rates of remission even in patients with no past treatment history, suggesting that the association between HC volumes and short-term clinical response was not simply a function of past treatment responsiveness.112 Frodl et al. reported that depressed patients who were not remitted from an episode of depression 1 year after discharge had smaller left and right HC volumes at baseline scan.39 The association between smaller HC volumes and poor clinical outcome was detectable after a 3-year follow-up.113 Kronmuller et al.114 also found associations between HC volumes and 2-year outcome, although the associations between large HC volume and good outcome were restricted to men.
The relations between small HC volumes and response to treatment suggest that the association between illness duration and HC volumes may arise from the effect that small HC volumes have on outcome. That is, there may be an iterative relation between HC volume and illness burden, with small HC volumes contributing to poor clinical outcome, which in turn may put further stress on the HC leading to structural changes that increase the likelihood of a poor clinical response. Whether such relations contribute to the development of illness chronicity remains to be determined.
Associations between cognitive function and hippocampal volume
Cognitive impairment is a core feature of MDD.115 An effect size analysis of cognitive functioning in 726 patients with MDD conducted using meta-analytic principles found that depression had the largest effect on recollection memory.9 Recollection memory tasks are highly dependent on the HC in addition to some frontal regions.116 Hippocampal volumes are correlated with verbal learning and memory performance in various neuropsychiatric conditions. Vermetten et al.117 reported that antidepressant treatment reversed HC volume reduction in patients with PTSD; declarative memory also improved in the treated group. Hippocampal volumes are also correlated with executive functioning in patients with MDD49, 118 and emotional memory in middle-aged healthy women.119
There is an epidemiological association between MDD and Alzheimer's disease120 and this association does not seem to be restricted to late onset depression that is a prodrome to the clinical symptoms of Alzheimer's disease.121 Rather, meta-analyses suggest that MDD may be a risk factor for Alzheimer's disease.121 Stress-related brain changes, decreases in neurogenesis and programmed neurodegeneration may contribute to this association122 as may inflammatory processes that are characteristic of MDD.123 Moreover, both MDD and the APOE ɛ4 allele are risk factors for dementia, suggesting that the association between the APOE ɛ4 allele and small HC volumes124 may provide a further mechanism for understanding the links between late onset depression and dementia.
Evidence that treatment for MDD may have an impact on hippocampal volumes
As data has emerged to suggest that a long history of illness is associated with small HC volumes, questions have also been raised concerning the potential role of treatments in minimizing or ameliorating the apparent decline in HC volume.
Experimental studies show that antidepressants, for example, SSRIs, clomipramine or tianeptine, suppress stress toxic effects on the HC, increase hippocampal neurogenesis and synaptic plasticity.3, 8, 125, 126, 127 Although the effect of antidepressants on neurogenesis seems to be important for some depressive-like models, there seem to also be neurogenesis-independent mechanisms of antidepressant action.128 Some results reduce the importance of neurogenesis for antidepressant response even more; when neurogenesis was blocked antidepressants retained their therapeutic efficacy perhaps by re-establishment of neuronal plasticity, like dendritic remodeling and synaptic contacts, in the hippocampus and prefrontal cortex.129 Recent research demonstrated in a mice model with high versus low 5HT1a-autoreceptor availabilities a role for the 5HT1a-autoreceptor, because 5-HT1a-autoreceptor-modulated intrinsic raphe firing rates are directly related to resilience under stress and to the response to antidepressant treatment.130 However, since tianeptine was shown not to act on the serotonergic system,131 other non-serotonergic mechanisms seem to be relevant too. For example, neurotrophic factors and the glutamatergic system need to be mentioned here; by enhancing neurotrophic factor signaling, environmental factors such as exercise and chemicals such as antidepressants may optimize glutamatergic signaling and protect against neurological and psychiatric disorders.132
An important mechanism for neurogenesis is indeed physical exercise, for example, wheel running in rodents resulted in enhanced hippocampal neurogenesis,133 improved synaptic plasticity134, 135 and increased spine density.136 Activity in experimental animal studies increased angiogenesis, accompanied by increased insulin-like growth factor137 and vascular endothelial growth factor138 and also enhanced the serum levels of the BDNF.139 In addition, activity stimulates the glutamatergic system with effects on NMDA receptors in the HC.135 Interestingly, exercise increased the blood flow in the gyrus dendatus of the HC in mice and also in healthy subjects, who took part in a 12-week aerobic programme with a frequency of 4 times a week.140 These factors have an important role in hippocampal neurogenesis and might imply that the positive effect of exercise in depression might be due to the stimulating neuroplastic effects.
Given the complexity of longitudinal studies in which treatment is controlled over long periods of time, it is not surprising that there are few studies that have examined the question of whether individual treatment modalities exert an effect on HC volume over time.
Increases in HC volume following treatment with the antidepressants sertraline and paroxetine117, 141 have been reported in patients with PTSD. In MDD, the results are less clear, with some studies failing to find an association between antidepressant therapy and HC volume.142 One study, however, found a trend toward higher HC volumes in patients who had declines in cortisol levels with pharmacotherapy,143 but see Vythilingam et al.142 (for a contradictory pattern of findings). Right HC volumes increased in patients who took their antidepressants over the full 3-year time interval, however, the study was not designed to investigate medication effects distinct from illness effects during the naturalistic 3-year follow-up.113 These studies were all carried out in relatively small samples so that slight effects might not have been detected. In addition, the difficulty to design a study with an adequate comparison group not receiving the antidepressant treatment under observation, limits research in this area.
A recent review concluded that the inclusion of bipolar patients treated with psychotropic medications in structural imaging studies was associated with either null findings or a reduction in volume loss in key regions associated with the pathophysiological mechanisms of the disorder, including the HC.144 There is evidence that lithium is associated with increases in HC volume with both short- and long-term treatment in patients with bipolar disorder.145, 146 Studies examining the impact of other medications or treatment modalities on HC volumes are lacking across diagnostic categories.
MDD has until recently been viewed as a ‘chemical imbalance’ that was not associated with permanent structural brain changes. Meta-analyses of structural imaging studies of the HC now suggest, however, that patients with MDD have small HC volumes compared to healthy subjects. Although the factors associated with small HC volumes are not fully described, there is evidence that genetic polymorphisms and early life stress may contribute to HC volumes before onset of illness. Furthermore, subjects at risk for MDD have reduced HC volumes before the clinical onset of illness. Repeated episodes of illness may further contribute to loss of HC volume, and speculatively, the changes in the HC may then contribute to treatment resistance or chronicity. Consistent with this are studies reporting that small HC volumes predict poor short- and long-term responsiveness to treatment (overview in Figure 2).
Pre-clinical and clinical studies of the HC and related structures have provided a plausible biological model for how stress may interact with genetic vulnerability to increase the risk of MDD. We now know that detecting evidence of HC disruption, as measured by small HC volumes on MRI, can allow us to identify people who are at risk of illness and patients who are at risk for treatment non-response and relapse. Clinicians should begin to consider the possibility that imaging technology, possibly also in combination to genetic data (genetic variants, family history) could be used, not for diagnosis, which may be of uncertain utility, but for matching patients with treatment intensity. If imaging can reliably identify people who are likely to require more than first line treatment to recover, then scanning might become cost effective as it is well recognized that less than 50% of patients have an adequate response to first treatment. Identifying subgroups of patients with specific morphological markers may also facilitate the development of treatment strategies that target neuroplastic processes. With few revolutionary treatment strategies on the horizon, and the cost of MDD continuing to rise across the world, it may be time to contemplate the possibility that imaging data could contribute to clinical care by matching patients to the most appropriate level of intervention.
Professor Frodl received funding for a Stokes Professorship from Science Foundation Ireland, SFI.