Dysregulation of adult hippocampal neuroplasticity in major depression: pathogenesis and therapeutic implications

Abstract

Major depressive disorder (MDD) was previously hypothesized to be a disease of monoamine deficiency in which low levels of monoamines in the synaptic cleft were believed to underlie depressive symptoms. More recently, however, there has been a paradigm shift towards a neuroplasticity hypothesis of depression in which downstream effects of antidepressants, such as increased neurogenesis, contribute to improvements in cognition and mood. This review takes a top-down approach to assess how changes in behavior and hippocampal-dependent circuits may be attributed to abnormalities at the molecular, structural and synaptic level. We conclude with a discussion of how antidepressant treatments share a common effect in modulating neuroplasticity and consider outstanding questions and future perspectives.

Introduction

Major depressive disorder (MDD) is one of the commonest psychiatric conditions in the United States and is the leading cause of disability worldwide¹. Historically, MDD was thought to be a disease of monoamine deficiency characterized by low levels of serotonin, norepinephrine and/or dopamine in the central nervous system ^{2, 3}. The corollary of this hypothesis was that drug classes that increase the concentration of monoamines in the synaptic cleft should have antidepressant properties⁴. Two major problems arose in relation to this antidepressant treatment model. First, although these medications are pharmacologically active within hours, their antidepressant effects are not apparent for weeks after starting treatment. Second, even when the medications are having meaningful pharmacologic effects, many patients do not improve. The most widely accepted response to both questions is to assume that the effects of antidepressants are dependent on downstream processes, such as an increase in neurogenesis, that take time to appear and are only robust in medication responders. Consequently, there has been a paradigm shift from a synaptic effect derived from the monoamine hypothesis to a more complex delayed downstream neuroplasticity hypothesis of antidepressant action that involves correcting a deficiency in neurons, processes and synapses that comprise the pathogenesis of depression.

Neuroplasticity is defined as the brain’s ability to undergo neurobiological changes in response to extrinsic stimuli such as early life adversity^5-7 and chronic exposure to stress^{8, 9} and/or intrinsic stimuli most notably genetic or epigenetic effects^{10, 11}. The responses to these changes can be visualized at a structural level (e.g., cell number, dendritic spine density and morphology or synaptic protein levels) and at a functional level (e.g., synchronous firing) that in turn determine the state of networks^{12, 13}, stress responses, mood, cognition and behavior. These changes can be adaptive and contribute to resilience in at-risk groups, or be maladaptive, resulting in neuropathology and psychiatric disorders.

The hippocampus is of particular importance in understanding depression pathogenesis^14-16. First, the hippocampus plays a critical role in mood regulation, in part, due to its connections with emotion-related brain regions such as amygdala and anterior cingulate cortex (ACC)^{17, 18} as well as its feedback role in regulating the hypothalamus-pituitary-adrenal (HPA) axis. Second, the hippocampus is one of the few brain regions thought to be capable of adult neurogenesis^{19, 20}. Third, given its role in HPA axis regulation, the high concentrations of hippocampal glucocorticoid receptors make it particularly vulnerable to allostatic load. Allostatic load is elevated in depression, which is associated with prolonged stress response, including elevated levels of cortisol associated with melancholia^{5, 21}. As such, hippocampal neuroplasticity has been implicated in MDD etiology and antidepressant action.

Learning, Memory, and Mood: Network Changes in Depression

Depression is a heterogenous disease with symptoms spanning multiple domains of emotion and behavior, including but not limited to, changes in mood, anxiety, memory, anhedonia, optimism, sleep, energy, appetite, libido and psychomotor activity. However, in this review, we will focus on the neurobiological basis of two core symptom clusters reported in MDD: cognitive and affective. While affective symptoms are routinely characterized, it has been estimated that cognitive deficits affect 20-90% of MDD patients^22-27. Furthermore, deficits in memory, attention and executive functioning have been observed in the absence of a current depressive episode^28-31 and even in individuals responding well to antidepressant treatment for affective symptoms²³, suggesting that cognitive deficits are only partly a state-dependent phenotype in MDD. Interestingly, the manifestation of these cognitive deficits is diverse. Some studies find deficits in attention and psychomotor processing that correlated with global level of functioning²⁴ while others report deficits in executive functioning in conjunction with intact memory, attention and psychomotor performance. These discrepancies may be attributed to differences in cognitive testing methodologies and patient population. More generally, however, MDD is remarkably pleomorphic in terms of mood and vegetative symptoms. Within the same patient, there is no consistency of severity of individual symptoms, symptom components or factors across successive episodes of major depression³². Therefore, different study populations may have different proportions of certain mood or cognitive subgroups of depression, explaining conflicting findings and the wide range in incidence of cognitive deficits.

The pathogenesis of cognitive and affective deficits in untreated MDD is not fully elucidated but likely involves abnormalities across multiple brain regions^33-35, and includes but is not limited to changes in functional and/or structural connectivity. The dorsolateral prefrontal cortex (DLPFC) is responsible for attention, working memory and executive function and is positively correlated with working memory load in healthy controls³⁶. Neuroimaging studies have consistently reported DLPFC hypoactivity at rest in unmedicated MDDs^{37, 38} but DLPFC hyperactivity during working memory tasks³⁹. Yet, this hyperactivity is not matched with increased memory performance⁴⁰. The association between DLPFC activity, memory and MDD has not been consistent, as others have reported DLPFC hypoactivity in conjunction with decreased working memory performance^{41, 42}. Connectivity studies of the middle frontal gyrus (part of the PFC) and hippocampus, report negative correlations between these regions to be associated with illness duration in untreated depressed patients³³. Negative correlations indicate that when one brain region is activated the other is deactivated while positive correlations indicate brain synchronicity when regions are activated or inhibited at the same time⁴³. Taken together, these studies suggest inefficient or malfunctioning prefrontal activity and connectivity, which may underlie working memory deficits in a subset of untreated depressed patients.

The PFC plays a role in memory, including working memory, but also is integral in emotion regulation through the cortico-limbic network. Top-down dysregulation of limbic structures may mediate affective symptoms and altered decision-making seen in mood disorders^44-46. In healthy controls, when asked to downregulate emotional responses in the presence of an aversive stimuli, the PFC and ACC were hyperactivated and the amygdala was hypoactivated. When participants were asked to upregulate negative emotions in response to aversive stimuli, both the PFC and amygdala were activated⁴⁷, suggesting a PFC-dependent cognitive component in emotional regulation in healthy individuals. Conversely, when patients with MDD were asked to ignore negative stimuli, they failed to recruit the DLPFC which was correlated with amygdala hyperactivity⁴⁸. Such examples of altered cognitive control of mood are characteristic of MDD during emotion-related tasks. Inability to focus on external environmental stimuli during periods of rumination can further maintain negative thought content and depressive episodes⁴⁹. Additionally, focus on negative internal stimuli can result in more strongly encoding negative experiences and sustain negative thoughts and expectations, and underly a biased recall of negative relative to other memories in depressed patients.

In addition to dysfunctional top-down regulation, altered limbic and intra-hippocampal connectivity has been observed in MDD. Connectivity between the amygdala and hippocampus is the link between declarative memory and affect-related memory and is more strongly, positively correlated in MDD patients compared with non-psychiatric controls when participants were retrieving negative emotions⁵¹. Interestingly, unmedicated adolescents with MDD showed less resting state fMRI (rsfMRI) connectivity between hippocampus and amygdala compared with healthy controls⁵². This lack of connectivity between hippocampus and amygdala correlated with severity of depression⁵². Variations between hippocampal-amygdala connectivity during resting state and emotional and cognitive tasks suggests that this network is context-dependent and modulated by environmental stimuli and individual emotional state.

Abnormal connectivity between individual hippocampal subfields with other brain regions, also contributes to MDD symptoms. Increased connectivity between hippocampal dentate gyrus (DG) (responsible for spatial encoding and memory) and ventrolateral PFC (responsible for regulation of emotions, impulsivity, and memories of anticipated rewards for each action or choice)³⁵, may contribute to strong negative emotional memories in MDD. Furthermore, in human and rodent studies, changes to hippocampal circuitry have been associated with early life adversity which suggests an epigenetic mechanism for these changes and MDD pathogenesis^{53, 54}. Taken together these studies suggest aberrant network organization and functionality may underlie dysregulated emotional perception, elaboration, cognition, and emotion regulation in depressed patients.

A Balancing Act: Cell Death and Survival

Network activity is not only dependent on firing frequency and synchrony, but also relies on the number and type of neurons available in a specific region being recruited for a given activity. As such, hippocampal cell loss and its implications for MDD are complex. The hippocampus comprises the following subregions - Cornu Ammonis (CA) regions 1-4, DG and subiculum (Figure 1). Each subregion has its own unique function, pattern of inputs and outputs, and gene expression profile. In rodents, DG and CA3 contribute primarily to intrahippocampal connectivity (working to integrate information within the hippocampus), while CA1/CA2 and subiculum are organized into global networks tasked with extrahippocampal communication¹². Likewise, the hippocampus has a range of functions along its rostral-caudal axis with the anterior/head (ventral in rodents) being responsible for emotional regulation while the posterior/tail (dorsal in rodents) plays a larger role in declarative memories. Differential volume loss within hippocampal subfields (e.g., in CA3 but not CA1) or along the hippocampal axis may underlie different aspects of MDD symptoms and psychopathology.

Figure 1: Hippocampus Anatomy.

The hippocampus is divided into multiple subregions including the dentate gyrus (DG), Cornu Ammonis (CA) regions 1-4, and subiculum (not shown). The entorhinal cortex (EC) acts as the gateway into the hippocampus via the perforant path which projects onto the DG (1). The DG sends fibers to CA3 through the mossy fiber pathway (2). CA3 pyramidal cells receive inputs from the associated/commissural fibers (not shown) and send their projections to CA1 via Schaffer collaterals (3). Lastly, neurons in CA1 project back into the entorhinal cortex and into subiculum (4).

Although some studies report no volumetric differences between MDD patients and healthy controls^55-57, these studies included patients who were either currently on antidepressants⁵⁶ or had been on antidepressants at some point in their life⁵⁵. Most studies find that untreated depressed patients have smaller hippocampal volume^{58, 59}, neuronal and glial number, and cell size⁶⁰ compared with non-psychiatric controls. Moreover, in MDD, the extent of hippocampal gray matter volume loss is related to time spent depressed^{58, 61} and smaller hippocampal volume is associated with worse depression scores. In addition to MRI, volume estimates have been conducted using unbiased stereology on postmortem human hippocampus. We have shown that DG granule neuron number and DG volume were smaller in anterior and mid, but not posterior hippocampus in unmedicated MDD postmortem⁶². Additionally, we found more granule neurons and a larger DG in resilient subjects who were exposed to childhood adversity but had no lifetime psychiatric diagnosis and died from natural causes⁶³. Interestingly, there is disagreement over which regions of the hippocampus are smaller in depression. Some find volumetric sparing of CA1⁶⁴ while others report CA1 volume loss^{61, 65, 66}. Some report volume loss selectively in the hippocampal head and others observed volume loss in the hippocampal body⁶⁷. Differences in the study sample may partially explain conflicting results because an array of environmental and internal risk factors may contribute to MDD, and with differing effects on brain regions. For example, MDD patients who reported being sexually or physically abused showed smaller left CA1 volume when compared with MDD patients who were not abused⁶⁸.

Hippocampal volume loss is not pathognomonic for depression and can occur in the presence of environmental stressors, other psychiatric conditions, and neurodegenerative diseases. Smaller ACC, DLPFC, medial prefrontal cortex (MPFC) and hippocampus were shown in non-depressed individuals with a familial history of depression (first degree relative with MDD diagnosis) exposed to emotional neglect when compared to non-depressed individuals with no familial risk factors but who had been exposed to emotional neglect, suggesting a gene-environment interaction affecting cell viability⁶⁹. Other studies disagree and report an independent environment effect on gray matter volume. For example, in the absence of familial risk, early childhood trauma severity (operationalized via the Childhood Trauma Questionnaire) correlated with amygdala responsiveness and hippocampal volume loss in a group of healthy volunteers, and these associations were not influenced by recent life stress, depression or anxiety scores⁷⁰. Interestingly, other studies suggest volumetric differences may precede environmental trauma and determine the clinical outcome of such adversity. This concept has been demonstrated in an MRI study of PTSD which showed how smaller hippocampal volume predicted risk of PTSD and did not result from the trauma that triggered PTSD⁷¹.

Currently, reported mechanisms potentially underlying cell loss include, but are not limited to, glutamate/glutamine cycling (see section on Molecular and Cellular Mechanisms of Neuroplasticity), blunted neurogenesis, decreased neurotrophic factor expression and upregulation of pro-apoptotic pathways⁷². One potential process involved in neuronal loss is decreased brain-derived neurotrophic factor (BDNF) via histone modifications to the promoter region associated with downregulation of BDNF transcripts^{73, 74}. Given its role in neuronal maturation and differentiation⁷⁵ as well as its neuroprotective effects⁷⁶, low brain levels of BDNF may affect neuronal viability in MDD. This is because BDNF can bind to tropomyosin kinase B (TrkB) receptor which activates the mechanistic target of rapamycin (mTOR) signaling pathway which promotes neuronal growth, proliferation and migration⁷⁷. In humans, BDNF polymorphisms were found in adults with young-onset depression⁷⁸. Additionally, BDNF expression was decreased in postmortem hippocampus in untreated MDDs but not in antidepressant-treated depressed individuals⁷⁹. In rodents, BDNF haploinsufficiency resulted in smaller hippocampal volume and increased anxiety-related behaviors when exposed to chronic stress^{80, 81}.

Upregulation of pro-apoptotic factors such as Bax and downregulation of anti-apoptotic factors like Bcl-2 have been shown in rodent models of depression^{82, 83}. Similarly, upregulation of genes involved in cell death and apoptosis were found in blood and postmortem PFC of patients with MDD^{84, 85}. In rodent models, exposure to a pollutant known to cause cell death or inflammation, caused upregulation of inflammatory markers and mediators of apoptosis in hippocampal neurons and resulted in depressive symptoms^{86, 87}. Elucidating mechanisms that underlie cell death and survival may be critical in harnessing effective therapies for MDD.

A role for adult hippocampal neurogenesis in depression pathology?

A significant contributor to neuroplasticity is adult hippocampal neurogenesis (AHN), the process by which new neurons are generated from adult neural stem cells. This process is regulated through epigenetic modifications of transcription factors, non-coding RNAs and metabolic pathways⁸⁸. During embryonic development in rodents, primitive dentate progenitors from dentate neuroepithelium migrate along the dentate migratory stream to establish the primitive dentate gyrus^{89, 90}. The majority of dentate granule neurons are generated during the first postnatal week⁹⁰. Although less is known about embryonic hippocampal neurogenesis in humans, some studies suggest overlap in developmental timing and molecular signatures between the rodent and human brain⁹¹. The switch between developmental neurogenesis and AHN is gradual, and some sequencing studies have demonstrated these two processes share highly similar transcriptional trajectories⁹¹. Some have demonstrated that neural stem cells in the adult hippocampal subgranular zone (SGZ) niche are remnants of dentate neuroepithelium⁹² and this same population of progenitors exclusively contributes to hippocampal neurogenesis throughout development and into adulthood, shifting out of quiescence at different time points⁹³. Others have argued that these stem cells originated in the ventral hippocampus and migrated dorsally⁹⁴.

Interestingly, if AHN exists, the extent to which it resembles its embryonic precursor on a morphologic and transcriptomic level remains a critical point of contention. Some groups have reported absent or minimal hippocampal neurogenesis in the mature brain using double immunofluorescence targeting markers expressed by neuronal cells at different maturational stages^{95, 96} and single nucleus transcriptomics⁹⁷. However, our group and others have shown evidence of neurogenesis throughout adulthood using double immunofluorescence^98-101, in situ hybridization¹⁰² and ¹⁴C decay-defined neuronal age¹⁰³. The ability to visualize newborn neurons is heavily dependent on tissue fixation procedures, experimental protocols, and subject selection, as previously described^{102, 104, 105}.

In addition to the histological concerns, skeptics of AHN also raise ideological criticisms. An influential paper from 1985, and a recent review, postulated that evolutionarily advanced brains would favor stability over plasticity, calling into question the evolutionary advantage of integrating new neurons into complex brain circuits^{106, 107}. On the other hand, it has been suggested that AHN provides cognitive adaptability to survive in a variable and ever-changing environment through the flexible integration of novel information into preexisting representations^108-110. A second ideological argument is why the hippocampus would not grow over the lifespan if neurogenesis persists. While ¹⁴C decay-defined neuronal age estimates that the human brain adds up to 700 new neurons each day (approx. 0.004% of total DG neurons), resulting in a 1.7% cell turnover annually¹⁰³, in rodents, it has been estimated that up to 30-70% of newborn neurons die in the first month and that 1300 neurons are eliminated daily from the rodent hippocampus^{111, 112}. Although equivalent studies have not been conducted in humans, it is likely that the rate of cell death offsets neurogenesis thereby preventing expansion of hippocampal volume in adult life.

Although unknown in humans, studies in adult rodents have demonstrated that new born neurons have a critical period (2-4 weeks) of hyperexcitability, when they are preferentially recruited during a wide variety of hippocampal-dependent tasks such as flexible learning^{113, 114}, spatial memory¹¹⁵ and most notably pattern separation (the ability to separate similar but different memories or experiences)^{116, 117}. Deficits in AHN may not only explain the learning and memory deficits in MDD, but also play a role in the selective engagement with negative valence memories, mood regulation and antidepressant response^118-122. A few studies provide evidence that the absence of neurogenesis elicits depressive like symptoms in rodents^{123, 124}, but most studies find that absent neurogenesis is not sufficient to induce a depressive phenotype¹²⁵ although it is required in mice for the behavioral effects of antidepressants^{118, 126}. In fact, the inability to produce more neurons appears to make the animal vulnerable to the epigenetic effects of chronic stress and subsequently the development of a depressive phenotype⁸⁸. This may be because newborn neurons confer resilience to stress by inhibiting stress-responsive mature granule neurons during anxiogenic tasks¹²⁷.

In postmortem human brain, we found fewer neural progenitor cells and mature granular neurons in unmedicated depressed subjects, selectively in anterior DG, compared with non-psychiatric sudden death controls, suggesting that neurogenesis may be blunted in MDD^{62, 63}. Humans receiving chemotherapy, which kills proliferating cells (thus blunting neurogenesis), experience cognitive deficits and are more likely to develop depression than cancer patients treated with other therapies^128-130. In rodents, administration of chemotherapy resulted in a loss of hippocampal proliferating cells that correlated with behavioral deficits^{131, 132}. Thus, it is possible that, like in rodents, the role of newborn neurons in humans is predominately for conferring resilience on the hippocampal circuitry, and possibly facilitating circuit rewiring for the antidepressant response.

Molecular and Cellular Mechanisms of Neuroplasticity

As previously discussed, patients with depression have disruptions in neurological circuits responsible for mood regulation and cognition, that may underlie MDD symptoms. Long-term potentiation (LTP) and long-term depression (LTD) are two mechanisms that impact cognitive and affective functions impaired in MDD^133-135. Increased neuronal firing in the presence of a strong, constant stimulus enhances LTP by subsequently strengthening synapses which mediate learning and memory. LTD, on the other hand, is an activity-dependent reduction in the efficacy and connection of neuronal synapses¹³⁶.

In the presence of high neuronal stimulus, α-animo-3-hydroxy-5methyla-4-isoxazolepropionic acid (AMPA) receptors and the adjacent N-methyl-D-aspartate (NMDA) receptors remain open, resulting in strong depolarization and calcium influx. Intracellular calcium activates protein kinases responsible for enhancing synaptic communication efficiency through increasing sodium conductance^{137, 138} (Figure 2). These changes are believed to be responsible for short-term memory, which can last for several hours. The late phase of LTP is dependent on transcription and translation activity for de novo gene expression that mediates structural and enduring functional circuitry changes^139-141. One such mechanism that induces these changes is the BDNF- TrkB cellular pathway. The transcription of BDNF is dependent upon activation of the cyclic adenosine monophosphate (cAMP) response element binding protein (CREB) protein which plays a critical role in LTP and synaptic plasticity^{142, 143}. BDNF binding to the TrkB receptor activates several signaling cascades (e.g. MAPK/ERK, PI3K, mTOR)¹⁴⁴ responsible for spine enlargement and glutamate sensitivity¹⁴⁵ (Figure 3). The role of BDNF-TrkB signaling is most evident in hippocampus through its role in LTP and facilitation of learning and memory¹⁴⁶.

Figure 2: Mechanisms of Synaptic Plasticity.

Left: Under basal conditions, glutamate is released from the pre-synaptic neuron. Once in the synaptic cleft, glutamate stimulates AMPA receptors on the post-synaptic neuron, triggering depolarization noted by the influx of sodium ions. With a weak stimulus, influx of sodium into the pre-synaptic neuron is minimal and infrequent, resulting in infrequent and low amplitude excitatory postsynaptic potentials (EPSPs). Right: Under high levels of stimulation, more glutamate is released from the pre-synaptic neuron resulting in stronger depolarizations on the post-synaptic neuron. Stronger depolarization increases the amount of calcium in the cell and removes magnesium from NMDA receptors, allowing more sodium to enter. High intracellular calcium levels activate kinases like Protein Kinase C (PKC) and Calcium/Calmodulin dependent protein kinase II (CaKMII) which phosphorylates AMPA receptors thereby increasing their conductance. In both scenarios, glial cells present at the synaptic cleft aid in glutamate reuptake.

Figure 3. Molecular Regulators of Neuroplasticity.

Brain-derived neurotrophic factor (BDNF) attaches to a Tropomyosin receptor kinase B (TrkB) receptor on the post-synaptic neuron. BDNF signaling causes expression of phosphoinositide 3-kinase (PI3K) which is responsible for cell proliferation and survival. PI3K activates Protein Kinase B (Akt) and Mitogen-activated protein kinase (MEK) pathways which lead to expression of cAMP response element binding protein (CREB) and mammalian target or rapamycin (mTOR), respectively. High intracellular Ca²⁺ concentrations from AMPA and NMDA activation activate CAMKII which also stimulates CREB production. CREB is a key element needed for neurite outgrowth, neurogenesis and synaptic plasticity.

LTP in MDD has most frequently been assessed using stress-induced rodent models of depression. These studies have generally observed increases in DG LTP¹⁴⁷ and reductions in LTP at Schaffer collaterals-CA1 synapses^{148, 149} in response to acute and chronic stress. Exposure to stress attenuates LTP in dorsal hippocampus but augments LTP in ventral hippocampus of rodents¹⁵⁰, which may explain memory deficits in the presence of strong emotional responses. In humans, LTP is difficult to measure directly. However, proxy measurements such as paired associative stimulation (PAS) using a transcranial magnetic stimulation protocol have shown that PAS-induced increases in motor-evoked potential amplitudes were attenuated during major depressive episodes compared with healthy controls and normalized during remission^{151, 152}. This suggests that LTP attenuation may be a state rather than trait marker in MDD.

Dendritic integration and synaptic strengthening are dependent on neuronal activity since dendritic summation of synaptic inputs are spatially and temporally dependent. Therefore, changes in glutamatergic signaling can increase or decrease excitatory postsynaptic potentials, affecting LTP. In the brain, glutamate is made from glutamine, and after glial reuptake, glutamate is converted back to glutamine. In rodents, acute exposure to stress via restraint or swimming test procedures, increases glutamate production, particularly in hippocampus and prefrontal cortex^153-155. However, sustained high levels of glutamate can be toxic to the cell due to dysregulated calcium homeostasis which is essential for maintaining neuronal integrity and long term survival¹⁵⁶. Some have reported increased glutamate levels in the plasma of untreated MDD patients^{157, 158} in line with reports of high CSF glutamate concentrations in severely depressed, hospitalized MDD patients¹⁵⁹ and in untreated elderly MDD patients when compared to healthy controls¹⁶⁰. In vivo proton magnetic resonance spectroscopy (¹H MRS) detected elevated glutamate levels in ventromedial prefrontal cortex/anterior cingulate cortex in untreated MDD¹⁶¹ consistent with a recent meta-analysis of MRS imaging studies that found higher glutamate/glutamine concentrations in medial frontal cortex of unmedicated MDD but not in antidepressant-treated patients¹⁶². As such, lower glutamate may be an antidepressant effect and not part of MDD pathogenesis. It should be noted that most MRS studies assessed Glx levels that are the sum of glutamate and glutamine, and examined CSF and/or blood which may miss brain region specific abnormalities.

Normal stress responses involve glutamate release which downregulates AMPA and NMDA receptor expression, glutamate clearance by glia, and less dendritic spine and process complexity^{163, 164}. Deficient glutamate clearance, reported in MDD, may be due to downregulation of high affinity glutamate transporters, expressed specifically in glia, but not neurons, in human hippocampus ¹⁶⁵. Levels of glutamate receptor genes GLUR1 and GLUR3 were down-regulated in postmortem DG and CA1 in MDD subjects off medication at time of death but had been previously prescribed antidepressants¹⁶⁶. Increased glutamatergic transmission in the presence of stressful situations and lack of glutamate clearance mechanisms may explain strong encoding of negative valence memories in MDD.

Impaired LTP may impact dendritic spine number/density, size, and complexity^{167, 168}. BDNF^+/− mice show less dendritic branching as well as dendritic retraction and simplification in CA3⁸¹ and DG⁸⁰ and dysregulation of BDNF/TrkB pathways has been reported in MDD and animal models of depression^{78, 169-171}. Similar reductions have been found in proteins such as CREB and its upstream effectors ERK and PKC^172-174. Synaptic strength moderates neural signaling between cells over an extended period^{175, 176}, underlying circuit functionality. Dendritic retraction may be an adaptive way to protect neurons from high glutamate transmission, resulting in less LTP and cognitive decline. Nevertheless, additional studies are needed to determine the extent to which dendritic changes occur in MDD and their regional brain distribution.

Treatment Induced Neuroplasticity

While duration or presence of major depression correlates with declines in neuroplasticity, administration of antidepressants may reverse some of the neurobiological changes observed in MDD (Figure 4). As previously discussed, traditional antidepressants target the monoaminergic system to increase levels of monoaminergic neurotransmitters in the synapse. This may upregulate LTP pathways, and downregulate LTD^{177, 178}, having benefits on network activity and consequently cognition and behavior. At a systems level, fMRI studies found that SSRI treatment impacted network functionality in MDD. Studies using rsfMRI detect DLPFC hypoactivity in untreated MDD, and SSRIs increase DLPFC activity to a level comparable with non-psychiatric controls¹⁷⁹. Furthermore, modulation of networks can be achieved through deep brain stimulation (DBS)^180-182. A double-blinded randomized trial showed enhanced working memory in MDD patients during emotional conditions¹⁸³ after transcranial stimulation of the left DLPFC.

Figure 4. Before and After Treatment.

Table shows cognitive/behavioral, network/circuit, neuron structure and number, synapse function and signaling pathways alterations in untreated and treated depression.

In addition to changes at the circuit level, MRI studies have found that SSRI treatment ameliorated hippocampal volumetric loss over time¹⁸⁴. SSRI treatment is associated with normal hippocampal volumes in postmortem human brain while untreated MDD is associated with a smaller hippocampus compared with non MDD sudden death controls⁶². In line with this, voxel-based morphometry demonstrated increased hippocampal grey matter volume after antidepressant treatment^{185, 186}. Medicated patients experiencing depressive episodes had smaller gray matter volumes than those in remission¹⁵⁷ indicating that recovery from a depressive eipsode may permit brain regrowth.

Changes in brain volume may also be due to decreased neurotoxicity from decreased glutamate concentration. Patients with MDD treated with fluoxetine for 10 days showed a reduction in plasma glutamate levels when compared to baseline¹⁵⁷. While this reduction does not lower glutamate levels to the same concentration as healthy controls, it has been shown that antidepressant-treated glutamate levels are positively correlated with Hamilton Depression Rating Scale scores¹⁸⁷. Another mechanism that may contribute to restoration of hippocampal volume after antidepressants is increased neurogenesis. We found that antidepressant treatment is associated with more neural progenitor cells and granule neurons in postmortem human DG in MDD compared with non-MDD sudden death controls. This suggests an increase in neurogenesis over normal levels that allows a catchup in mature granule neurons in MDD to regain levels seen in controls^{62, 121}.

Although SSRIs and serotonin-norepinephrine reuptake inhibitor (SNRIs) perform better than placebo at reducing symptoms of depression^{188, 189}, there are weeks-long lag time in attaining full benefit and many patients do not respond. This has increased the need for rapid acting antidepressants. Recent studies have shown that NMDA receptor antagonists like ketamine and 5-HT_2A receptor agonists like serotonergic psychedelics (the most studied include lysergic acid diethylamide/LSD and psilocybin) induce rapid and sometimes robust antidepressant effects^190-193. This makes the mechanism of action of these two types of medication of great interest.

Ketamine is a non-competitive NMDA inhibitor¹⁹⁴. At a therapeutic dose, ketamine produces activity-dependent inhibition of less than 50% of NMDA receptors and binds with greater affinity to subunits of NMDA receptors expressed in the synapses of inhibitory interneurons¹⁹⁵. The dosage and affinity are important because, to increase LTP, some NMDA receptors must be available. In fact, rodent studies have shown that infusion of ketamine onto hippocampal neurons caused dose-dependent apoptosis, which was rescued by incubation with rapamycin (an inhibitor of mTOR), suggesting context-dependent benefits of ketamine¹⁹⁶. Nevertheless, it is hypothesized that when ketamine binds to NMDA receptors, the glutamate in the synapses of excitatory pyramidal neurons in the hippocampus¹⁹⁷ and PFC^{198, 199} shift transmission to AMPA receptors, and thereby strengthens synaptic connections driving synaptogenesis²⁰⁰. At the same time ketamine can prevent hyperexcitability and subsequent neurotoxicity through antagonism of NMDA receptors. Rodent models of depression show that ketamine administration increases dendritic spine density, length, arborization and morphology in CA1 pyramidal neurons²⁰¹. This has been hypothesized to be modulated by increases in BDNF/TrkB signaling¹⁹³. Psychedelic drugs likely work via 5-HT_2A receptors but require more research to elucidate their antidepressant effects²⁰².

Precision medicine could have an especially powerful impact on diagnosis and treatment of MDD due to the heterogeneity of clinical presentations. Despite our increasing understanding of depression pathophysiology, clinicians are unable to use biomarkers in blood or cerebrospinal fluid, neuroimaging or genomics to diagnose and guide treatment for MDD²⁰³. Recent studies suggest that blood levels of sertraline, a common first-line antidepressant, can successfully be predicted through analysis of genes involved in sertraline metabolism and thus, giving an indication of drug effectiveness^204-206. Similarly, blood testing has identified genes (e.g., NRG1 which is involved in regulation of proliferation, survival, and differentiation of many cell types including neurons and epithelial cells²⁰⁷), proteins (e.g. CD47 which is implicated in neuroinflammatory cascades²⁰⁸), and proinflammatory markers (IL-1β, IL-6)^209-211 that are associated with depression. All are candidate screening biomarkers for antidepressant response. Electrophysiological characteristics and neuroanatomy findings indicated that neuroimaging can identify patient-specific targets for DBS and transcranial magnetic stimulation, increasing treatment efficacy^{213, 214}.

Concluding Remarks and Future Perspectives

Neuroplasticity is integral to healthy cognitive and affective functioning. Changes in dendritic morphology and density, neurogenesis, growth factor expression and neurotransmitter production all likely contribute to changes in functional connectivity underlying behavioral and cognitive deficits in MDD. These impairments have been targets of antidepressant action and reversal of deficits in neuroplasticity correlated with improvements in symptoms. MDD clinical heterogeneity remains poorly understood, particularly in terms of how it relates to pathogenesis and implications for treatment choices. Although a growing body of research has provided evidence that neuroplasticity is implicated in depression pathogenesis, more research is required to discern disease etiology and how pathological findings can be more specifically reversed. More rigorously characterizing patients with MDD using a combination of symptom-based, genomic, bloodwork, and brain imaging findings may prove useful in detecting depression biologic subtypes related or orthogonal to phenotypes, and better guide treatment planning.