A Hypothesized Role for Dendritic Remodeling in the Etiology of Mood and Anxiety Disorders

Given that the neurogenesis hypothesis for anxiety disorders and depression is supported by considerable evidence, but nevertheless appears insufficient on its own to explain these disorders, it is reasonable to consider additional neuronal factors. What other processes may also be involved in the brain of a depressed or pathologically anxious individual? The complementary hypothesis proposed here is that widespread dendritic remodeling in key regions of the CNS plays a key role in human mood and anxiety disorders. In the brain, dendrites are constantly expanding and contracting in length, often in an experience- or activity-dependent manner. On these dendrites, small projections called spines emerge or disappear. Spines express postsynaptic receptors for most neurotransmitters and seek out synaptic connections with boutons on axons of presynaptic neurons, again in an activity- and experience-dependent way. 42 Changes in the size and shape of the dendritic arbor and in the density of dendritic spines are central elements in activity-dependent synaptic plasticity. 43 A recent study, for example, showed that dendritic spine number is increased in the PFC of marmoset males who father young compared to those who are childless. 44 Many neurotrophic factors, including brain-derived neurotrophic factor (BDNF), have been shown to enhance dendritic arborization and increase dendritic spine number, thus increasing the number and strength of synaptic connections in the brain. 45 Dendrites have been shown to grow in a mature vertebrate cortex via active neuronal pathways. 46 Recent studies have also shown that dendritic spine shrinkage and enlargement are key components of two major molecular memory processes, long-term potentiation and long-term depression. 47 – 49

It has been known for more than a decade that psychosocial stress causes atrophy of apical dendrites in the CA3 region of the hippocampus. 50 In preclinical studies in which animal models are used as replicas of human psychiatric illness, the response to experimental stress is often used as a proxy for the symptoms of anxiety disorders and depression. Given the current emphasis on the role of stress in the generation of anxiety and depression in humans (e.g., Kendler et al. 19 ), it is reasonable to use such preclinical studies to consider the role of dendritic remodeling in these disorders. However, given that stress has been implicated in a number of other psychiatric disorders, it is also possible that the hypothesis proposed here applies more generally. During chronic stress, dendritic arborization, length, and spine number have all been shown to decrease in the PFC as well as the hippocampus. 32, 51 Excessive excitatory neurotransmission supported by glutamatergic activation of NMDA receptors and diminished by GABA-ergic neurotransmission decreases dendritic length and spine number. Of note, Hashimoto et al. 52 reported increased glutamate levels in the frontal cortex of postmortem brains from patients with major depression and bipolar disorder. Similarly, the stress hormone corticotropin releasing factor (CRF), known to play an important role in anxiety and depression, adversely affects dendritic morphology during brain development 53, 54 and causes rapid loss of dendritic spines after experimentally induced stress in laboratory animals. 55

Thus, the complementary hypotheses to the neurogenesis theory are as follows:

The above hypothesis accounts for several features of both preclinical and clinical studies in depression and anxiety. First, it overcomes the problem of placing too much emphasis on changes in neurogenesis, which, while of great interest, are limited in scope and distribution throughout the adult mammalian brain. 59 Second, it retains the idea that chronic stress affects the expression of neurotrophic factors and that this is a key element in the etiology of these disorders. Third, it hints at a cause for the inability of depressed and anxious patients to use logic to diminish their out-of-control negative and catastrophic thoughts, their worries and fears, and their sense of hopelessness and helplessness. That is, with fewer synaptic connections, normal inhibitory functions of the PFC over the amygdala 60 are disrupted so that fear cannot be regulated, positive memories from the hippocampus cannot be recognized by the PFC, nor can negative memories be modified, and rational thinking is compromised. Of course, the data that form the foundation for this dendritic remodeling hypothesis are gleaned largely from preclinical studies, in the spirit of the current emphasis on “translational neuroscience.” Therefore, the extrapolation to human psychiatric illness is only as solid as the validity of current animal models of fear, learning, and memory to the clinical arena.

Finally, the theory proposed here would help to explain how antidepressants; glutamate antagonists; and positive environmental manipulations, including psychotherapy, can increase dendritic arborization and spine density in multiple relevant areas of the brain, leading to antistress and antidepressant effects. Deletion of the TrkB receptor, the receptor to which BDNF binds, has been shown to cause reduction in spine density in the CA1 region of the hippocampus. 61 Recently, Chen and colleagues 15 created a transgenic mouse with the same polymorphism in the gene encoding BDNF (Val66Met) that has been associated in humans with decreased hippocampal volume. Not only did the mice with the Val/Met mutation have reduced hippocampal volume and BDNF levels compared to wild-type mice, they also had decreased dendritic arbor complexity and a depressed-anxious phenotype. This suggests that research implicating BDNF as a target for antidepressant action may actually have disclosed a BDNF effect on dendritic morphology rather than on hippocampal neurogenesis. On the other hand, short-term treatment with fluoxetine increases dendritic spine and synapse formation in rat hippocampus. 62 Norrholm and Ouimet 63 similarly showed altered dendritic spine density in animal models of depression that responded to antidepressant treatment. Enriched environment caused an increased arborization of mossy fibers in mouse hippocampus in one study, 64 and in another, 65 enriched environment improved spatial learning and decreased anxiety-like behavior in mice even when the hippocampus had been irradiated and neurogenesis consequently blocked. The latter study again suggests that neurogenesis may not be necessary for antidepressant-like manipulations to work.

The clearest experimental evidence for the theory proposed here was recently provided by Bessa et al. 66 These investigators showed that blocking hippocampal neurogenesis in rats with the cytostatic agent methylazoxymethanol (MAM) did not prevent antidepressants with a range of mechanisms of action from reversing the depression-like behavior invoked by exposure to chronic mild stress. On the other hand, the antidepressants did reverse damage to dendrites and synaptic connections caused by stress exposure, both in the hippocampus and the PFC. These findings allowed the authors to conclude that it is more likely that antidepressants work by promoting dendritic remodeling rather than hippocampal neurogenesis.

Although stimulation of glutamate receptors is necessary for the normal development of the dendritic arbor, 67 the speculation here is that excessive glutamate neurotransmission caused by exposure to stress is responsible for diminution of dendritic length and loss of spines. If antidepressant therapies promote dendritic remodeling, it could be argued that this would only expose them to the toxic effects of hyperactive glutamatergic neurotransmission and hence be counterproductive. Thus, it would appear critical that antidepressant therapies also reduce excessive glutamatergic release or block glutamate receptors, an idea long championed by Scolnick. 68 Preclinical evidence is emerging that this is the case. 69, 70 Boyce-Rustay and Holmes 71 recently showed that inactivating the NR2A subunit of the NMDA receptor in genetically altered mice produced anxiolytic and antidepressant effects. In one study, fluoxetine and desipramine exerted moderate but selective effects while reboxetine, an antidepressant available outside of the United States that blocks norepinephrine reuptake, showed strong effects in decreasing glutamate receptor expression in both the hippocampus and PFC. 72 Riluzole, a drug that decreases glutamate neurotransmission, increased both BDNF and neurogenesis in the CA3 region of the rat hippocampus. 73 Johnson and Shekhar 74 showed that NMDA receptor antagonists blocked lactate-induced panic-like responses in their rodent model of panic disorder. Baskys et al. 75 showed that stimulation of group I metabotropic glutamate receptors, which decreases presynaptic release of glutamate via a negative feedback mechanism, reduces excitotoxic injury and promotes neurogenesis.

Clinically, drugs that inhibit glutamatergic neurotransmission, including lamotrigine, riluzole, metabotropic glutamate receptor agonists, memantine, and acamprosate, are in various stages of testing for their efficacy in mood and anxiety disorders. Mathew et al. 76 reported positive effects of riluzole in generalized anxiety disorder, and Zarate et al. 77, 78 also reported efficacy for riluzole, which is marketed for the treatment of amyotrophic lateral sclerosis, in depression. Dramatic effects of the NMDA receptor blocking drug ketamine have also been reported in patients with depression. 79 Lamotrigine is a drug with multiple actions including decreasing glutamatergic neurotransmission and is already widely used for the treatment of bipolar disorder, for which it seems to have particularly strong efficacy for depression. The antibiotic ceftriaxone, which increases presynaptic glutamate reuptake, was found to have antidepressant properties in mice 80 and will surely be studied in human clinical trials in the very near future.

There is particular interest in the role that antidepressant drugs play at the level of the NR2B subunit of the NMDA receptor. Preclinical studies have shown that the NR2B subunit plays a critical role both in the development of dendritic arbor morphology 81 and in synaptic plasticity and learning. 82 The NR2B subunit has also been found to be central in the effect of acute stress on hippocampal long-term potentiation and depression of memory. 83 Burghardt et al. 84 found that chronic, but not acute, administration of the antidepressant citalopram decreased conditioned fear responses in association with down-regulation of the NR2B subunit of the NMDA receptor. The group found similar effects with acute administration of the glutamate blocking agent memantine, which is already marketed for the treatment of Alzheimer’s disease. 85 An NR2B selective antagonist, Ro25-6981, exerted antidepressant effects in mice. 86 There is a report of another NR2B subunit specific NMDA antagonist, CP-101,606, having antidepressant effects in a placebo-controlled trial involving depressed patients. 87

Therefore, there is clearly widespread interest in the possibility that blocking excessive glutamatergic neurotransmission may be an effective antidepressant and antianxiety maneuver. The speculation here is that this works by blocking the excitotoxic effects of excessive glutamate on dendrites in multiple brain areas, including prefrontal cortex and hippocampus, that are an integral part of exposure to repeated stressful life events and ultimately to mood and anxiety disorders. In addition, molecules that stabilize or increase dendritic branching could be candidate antidepressant and antianxiety drugs. For example, cypin, a protein that binds to the protein postsynaptic density-95 (PSD-95), was recently shown to promote the formation of stable dendritic branches. 88 Perhaps cypin, or a similar molecule, might have efficacy in human psychiatric illness. Other molecules recently discovered to promote dendritic spine protrusion are netrin-G ligand (NGL) 89 and synaptic adhesion-like molecule 2 (SALM2). 90 Also, a microRNA, miR-134, has recently been shown to be brain-specific and to regulate the size of dendritic spines. 91 There is currently great excitement that microRNAs may be involved in gene transcription patterns that are relevant to CNS disorders, and in this case a relationship to anxiety and depression is also a possibility. Finally, in keeping with current emphasis on the role of epigenetic factors in human psychiatric illness, it has been found that overexpression of histone deacetylase-2 (HDAC2) decreased spine density and memory formation. 92 An inhibitor of HDAC, vorinostat, is already approved for cancer therapy.

There are several aspects to the hypotheses proposed here that will require further clarification before they can be accepted. These include:

The theory proposed here can be tested experimentally in a number of ways. First, further studies documenting the effects of stress on neurotropin activity; glutamatergic neurotransmission; and dendritic length and spine number in areas outside of the hippocampus, particularly the PFC, should be done. Second, studies investigating whether standard antidepressant therapies block such stress-induced changes in PFC dendritic morphology are needed. These should be followed by preclinical experimental interventions such as the application of drugs that enhance neurotropin expression and/or reduce glutamatergic activity to the PFC during chronic stress to see if these too block dendritic retraction. Finally, further studies of “antiglutamatergic” drugs and of drugs that promote dendritic growth in patients with depression and anxiety disorders may demonstrate the clinical significance of the hypothesis. As noted above, although most emphasis has recently been placed on the role that stress has in the etiology of mood and anxiety disorders, to the extent that psychiatric illnesses in general are stress related, it is plausible that the dendritic remodeling hypothesis we have proposed will apply more generally across psychiatric diagnoses. As neuroimaging techniques that allow assessment of dendritic morphology in living human brain become available, it will be possible to test this hypothesis directly in clinical subjects. Until then, the usefulness of this hypothesis to clinical practice rests in its ability to predict therapeutic interventions, such as drugs that decrease glutamatergic neurotransmission or promote dendritic arborization, that are effective in treating mood and anxiety disorders.

The authors wish to thank Dr. Emanuel Landau of the Mount Sinai School of Medicine for his help and guidance with this manuscript and Peter Lam of Comprehensive Neuroscience for preparing the manuscript for publication. Dr. Gorman has no conflicts to report. Dr. Docherty reports having consulting, advisory, or speaking associations with Bristol-Myers Squibb, AstraZeneca, Janssen, and Otsuka.