In this review, mood and anxiety disorders will be considered together for several reasons: (i) comorbidity between anxiety and depression is the rule and not the exception, with up to 90% of patients with anxiety disorders experiencing clinical depression at some point in their lifetime⁴; (ii) there is a significant problem of diagnostic classification, with highly overlapping symptom criteria; (iii) from a neuroimaging perspective the circuits involved in both sets of disorders can be difficult to distinguish; and (iv) the most powerful treatments for both disorders are the same, including antidepressants such as SSRIs and cognitive behavioral therapy (CBT). It is not that biologically meaningful subclassifications do not exist within the broad categories of emotional disorders, rather that the current clinical descriptions are probably not identifying the phenotypic clusters of disorders that may be most useful from a neurobiological and treatment perspective.

Several lines of evidence suggest that there are specific neural circuits within the limbic-cortical system that mediate stress-responsiveness, mood and emotional regulation. Disorders of mood and anxiety represent brain-based disorders that lead to dysregulation of these circuits. Traditional psychiatric medication, psychotherapy and somatic therapies converge in bringing homeostasis to these disrupted circuits. New neurostimulatory therapies based on progress in understanding emotion circuitry and new pharmacological therapies based on understanding emotional learning are likely to provide more rapid and robust methodologies for treating these debilitating and complex disorders.

The areas most reproducibly found to be dysregulated in common emotional disorders are the prefrontal cortex (PFC) and subgenual cingulate cortex (Cg25), which seem to be involved in emotion experience and processing, as well as the subcortical hippocampus and amygdala, which are involved in emotional memory formation and memory retrieval^5,11–14.

For the purposes of illustration, this review will focus primarily on data related to the role of Cg25 in emotion regulation and processing, and the role of the amygdala in emotional memory formation and expression. Cg25 is involved in the production of sad emotions and in antidepressant treatment response. It is activated during transient sadness, and after recovery from depression its activity is decreased compared with baseline after recovery from depression¹² (Fig. 1a). Cg25 decreases in activity are seen in response to chronic fluoxetine treatment for MDD (Fig. 1b), as well as during recovery from depression related to Parkinson’s disease after chronic fluoxetine treatment (Fig. 1c). Interestingly, subjects who are randomized to placebo but show a natural recovery from symptoms of depression also have decreased activity in Cg25 from baseline (Fig. 1d). Activity in Cg25 before treatment predicts treatment response with CBT¹⁵ (Fig. 1e). Additionally, a response to CBT for social phobia is accompanied by decreases in Cg25 activity, and responders have greater decreases in Cg25 activity than do nonresponders¹⁶ (Fig. 1f).

Overactivation of the amygdala is also implicated in depression and anxiety¹⁷ (Fig. 2). Amygdala activation decreases with recovery from mood symptoms. Studies that implicate Cg25 also find significant amygdala decreases with response to CBT treatment for social phobia¹⁶ (Fig. 2a) and report that sustained amygdala activity before treatment predicts antidepressant response to CBT¹⁵ (Fig. 2c).

The initial reasoning behind the use of VNS followed from its apparent effects of elevating mood in patients with epilepsy²⁰, combined with evidence that VNS affects limbic activity in neuroimaging studies²¹. Furthermore, VNS alters concentrations of serotonin, norepinephrine, GABA and glutamate within the brain^22–24, suggesting that VNS may help correct dysfunctional neurotransmitter modulatory circuits in patients with depression. In VNS, patients have a battery-powered generator implanted in the chest wall that connects to a wire wrapped around the left vagus nerve in the neck. This wire sends intermittent electrical pulses through the vagal afferent connections, which normally convey visceral and homeostatic information to the brainstem. Although several regions within the medulla receive vagal inputs, the nucleus of the tractus solitarius (NTS) receives the greatest number of vagal afferent synapses (Fig. 3a,b). These afferents carry visceral sensation from the pharynx, larynx, trachea and thoracoabdominal organs, somatic sensation from around the ear, and taste sensation. Related to its antidepressant and antiepileptic effects are the polysynaptic connections through the NTS to the noradrenergic locus coeruleus and to the serotonergic raphe nuclei²⁵. Additionally, there are vagal disynaptic projections through the NTS to the thalamus, hypothalamus, central amygdala nucleus, bed nucleus of the stria terminalis and nucleus accumbens. Finally, the vagus directly innervates the reticular thalamic formation, which is involved in coordinating a variety of activities, including sleep-wake cycle, alertness and attention^26,27, all behaviors that are frequently disrupted in depression. Patients notice improved alertness during VNS, and a study of patients without sleep disorders revealed improved diurnal alertness during VNS²⁸.

Because of the ability of rTMS to cause transient activity changes in focal brain regions, it was predicted that rTMS might have antidepressant activity similar to that of ECT, if the local prefrontal cortex depolarization effects of ECT are a primary contributing factor to its antidepressant function. Early trials of rTMS in patients with depression confirmed this prediction^39,40. A meta-analytic review, which examined 33 clinical trials performed since 1996 on the efficacy of rTMS in depression⁴¹, concluded that rTMS is more effective in the treatment of depression than sham rTMS, with a large effect size of 0.71. In contrast, another review⁴² that arguably was too conservative in the trial inclusion criteria⁴¹ found a lack of efficacy of rTMS in a much smaller number of select trials. Another recent, multisite review is promising in both treatment efficacy and tolerability⁴³. Unfortunately, there still remains limited consensus evidence regarding which are the most effective treatment parameter combinations. Thus although the data remain promising as a future treatment methodology, we need more knowledge regarding which patients are likely to receive optimal benefit and which treatment parameters are optimal, and we need larger, rigorously controlled studies.

rTMS is also potentially useful as a probe for understanding brain activity changes with response to treatment with rTMS and ECT. By using a modified TMS coil and a conventional MRI scanner, the magnetic field induced by a TMS coil can be visualized⁴⁴. With this approach, one can noninvasively and precisely determine the locations of magnetic and electrical field induction. PET in combination with rTMS indicates that TMS at 1 Hz inhibits cortical excitability, whereas high-frequency TMS (20 Hz) enhances cortical excitability^45,46, consistent with preclinical data showing that low-frequency TMS leads to cortical inhibition and quenching of seizure activity. These studies also suggest the intriguing idea that there is differential antidepressant response dependent on baseline level of cerebral glucose metabolism, with baseline hypometabolism responding better to high-frequency stimulation that seems to enhance excitability, and baseline hypermetabolism responding better to low-frequency stimulation that seems to dampen excitability⁴⁵.

In treatment-refractory depressed patients, focusing DBS on another region, prefrontal cortex area Cg25, has been examined more extensively. Limbic-cortical pathways mediate depressed mood, according to functional neuroimaging studies^11,53. Work from our group (H.S.M.) demonstrates that Cg25 is involved in acute sadness as well as in the effects of antidepressant treatment in depression^12,54 (Fig. 1). Based on data that Cg25 is metabolically overactive in treatment-resistant depression and that activity in this area correlates with response to treatment^55–57, we examined whether the application of chronic DBS to Cg25 would reduce depression symptoms in patients with severe refractory depression⁵⁸. Chronic stimulation of white matter tracts adjacent to the subgenual gyrus (Fig. 4) was associated with a robust, sustained remission of depression in four of six previously refractory patients. Furthermore, the antidepressant action was associated with marked reduction in local cerebral blood flow as well as with changes in downstream limbic and cortical sites measured with PET (Fig. 4d,f). These data indicate that disrupting the pathological limbic-cortical circuit in depression with DBS may effectively reverse treatment-resistant depression in some patients.

The mechanisms that mediate responses to chronic treatment remain incompletely understood, but it is clear that DBS produces both local and distant brain activity effects⁵⁹. The behavioral and neuroimaging changes in the above study are consistent with a DBS effect of reducing elevated baseline Cg25 activity. DBS-induced activation of inhibitory GABAergic afferents and/or synaptic or metabolic failure induced by high-frequency activity are some of the more likely mechanisms of action currently under study (Fig. 4b)^60,61. Other distal effects on remote cortical and brainstem areas could occur both through indirect trans-synaptic effects, as a direct result from Cg25 decreased activity, or through direct anterograde or retrograde activation of white matter projections that pass through the field of stimulation.

In addition to these chronic effects, some subjects noted a sudden mood change with stimulation within the operating room. This is consistent with an acute effect of deactivating area Cg25, which dynamically correlates with changes in mood (that is, activity increases with sadness and decreases with alleviation of dysphoria in depression)¹².

DBS is also being examined to treat refractory obsessive-compulsive disorder, a severely disabling disorder in which patients are overwhelmed with obsessional and repetitive thoughts and often cannot inhibit disruptive compulsive behavior. In contrast to MDD, OCD is thought to involve overactivation of cortico-striatal-limbic circuits, and there are several theoretical connections between OCD and Parkinson’s disease implying that they may involve different components of the striatal-cortical systems⁶². DBS focusing on the internal capsule extending into the ventral striatum shows promise for treatment of OCD⁶³. A long-term outcome, open-label study using 36-month chronic stimulation suggests promising long-term effects of DBS in highly treatment-resistant OCD⁶⁴. The subjects demonstrated significant decreases in OCD symptoms and improvements in overall functioning. Notably, concomitant changes in depression symptoms were also described, facilitating pilot studies of anterior internal capsule stimulation for depression.

A variety of behavioral observations support the hypothesis that extinction is a form of learning and not ‘unlearning’ or the forgetting of a conditioned association (reviewed in ref. ⁶⁷). Thus, enhancing neurotransmission at NMDA receptors, which are involved in new learning processes, might facilitate extinction⁶⁸. D-Cycloserine (DCS) is a partial NMDA agonist, acting at the strychnine-insensitive glycine recognition site of the NMDA receptor complex to enhance NMDA receptor activity⁶⁹. Both systemic and amygdala-specific administration of DCS dose-dependently enhance extinction of previously conditioned fear but do not influence fear in rats that have not received extinction training⁶⁸. These findings have been replicated in studies measuring extinction of fear with startle and freezing and with extinction of appetitive cues, such as cocaine-conditioned place preference^70–73. Collectively, data from rodent studies indicate that DCS, a drug already used in humans for treating tuberculosis, may have potential use in facilitating extinction-based therapies for human anxiety disorders.

Behavioral therapies for anxiety disorders generally involve some form of extinction training⁶⁶. This involves graded exposure to the feared object or event in the absence of any likely actual harm. This exposure may occur in the imagination, as when a narrative is read or listened to by the patient, or the patient may directly encounter the feared stimulus. Considering the similarity between extinction training in rodents and exposure therapy for anxiety disorders in humans, we were interested in finding ways to integrate pharmacotherapy with psychotherapy, which have not generally been combined into a treatment more effective than either alone. Sometimes combining pharmacotherapy with psychotherapy can even make a bad situation worse⁷⁴. However, extinction-based therapies for anxiety may be an exception to this trend.

To determine whether DCS would improve extinction of fear in human patients with fear-related disorders, we ran a double-blind, placebo-controlled trial in a controlled exposure protocol⁷⁵. Twenty-eight subjects with fear of heights (acrophobia) were treated with two sessions of behavioral therapy using virtual reality exposure to heights within a glass elevator. Single doses of placebo or DCS were taken before each therapy session. Exposure therapy combined with DCS resulted in significantly larger reductions of acrophobia symptoms on all main outcome measures, including fear within the virtual environment, acrophobia symptoms outside of the environment and height-related anxiety. Subjects receiving DCS also showed significantly greater decreases in a physiological measure of anxiety during the test exposure. These pilot data provided initial support for the use of acute dosing of DCS as an adjunct to exposure-based psychotherapy to accelerate the associative learning processes that contribute to correcting psychopathology.

DCS has also been used in a double-blind, placebo-controlled trial for the treatment of social anxiety disorder by exposure therapy. In this study, 27 subjects received four exposure therapy sessions combined with DCS or placebo. Those receiving DCS in addition to exposure therapy reported significantly less social anxiety compared with patients receiving exposure therapy plus placebo. Controlled effect sizes were in the medium to large range⁷⁶. Additionally, DCS has been used to accelerate reduction of obsession-related distress in patients with OCD undergoing extinction-based exposure therapy. DCS given before therapy decreased both the number of exposure sessions required to achieve clinical milestones and the rate of therapy dropout. After four exposure sessions, patients in the DCS group reported significantly greater decreases in obsession-related distress compared with the placebo group⁷⁷. There are two negative findings of DCS use in the augmentation of extinction in humans, one using exposure therapy for the treatment of spider phobia⁷⁸ and the second using extinction for a conditioned cue in healthy human subjects⁷⁹. However, in both studies, the placebo-controlled groups had good recovery as well. This suggests, as in the animal studies, that if a ‘floor effect’ of complete extinction is accomplished, then DCS may not augment extinction any further. DCS may be most useful in disorders that require many therapy sessions for a response or that are refractory to normal exposure-based therapies. Studies are underway on the use of DCS to augment exposure-based CBT for PTSD, panic disorder, social anxiety, OCD and depression. Based on the data that DCS enhances extinction of place preference conditioned to cocaine in rats⁷¹, at least two human studies are also examining DCS effects in exposure-based treatment of cravings in disorders of addiction.

Decades of preclinical research demonstrate that memories do not immediately become permanent at the time of the initial experience. Instead they exist in a labile state for at least hours and possibly days, during which they become consolidated into a more permanent memory (Fig. 5). There is a host of in vitro and in vivo data in animals outlining the molecular, synaptic, neurotransmitter and systems-level changes that may occur during this consolidation process⁸⁰. This literature suggests that during the immediate period after fear training in the animal model and after trauma in the human patient, it may be possible to disrupt the consolidation of new fear memories that are in the process of being formed⁶⁵.

Although there are philosophical and ethical questions raised by a potential amnestic agent that would ‘erase’ a memory fully, it is increasingly clear that differential memory systems are encoding different aspects of a memory in parallel. In other words, declarative or explicit memory systems involving hippocampal-cortical pathways are likely to be encoding the ‘what, when and where’ of an event, while in parallel, amygdala-cortical pathways are encoding the emotional salience and aversiveness of the memory. Therefore, agents that block the ‘emotional overconsolidation’ of a traumatic memory perhaps could preserve the declarative aspects of the same memory. This idealized approach would not render the patient fully amnestic, but would potentially prevent PTSD.

Many of the ideas on interruption of consolidation stem from a protocol called ‘avoidance learning’, in which an animal is conditioned in a box with two compartments. When it steps from one compartment to the next, it receives a mild footshock⁸⁰. After this training, the animal then avoids spending time in the compartment in which it was shocked, even if there are no more reinforcing shocks. Consolidation of contextual fear memory in this protocol is disrupted after training by a variety of agents, including dopaminergic, cholinergic and other modulatory drugs, but most notably those involving the adrenal stress system—epinephrine and glucocorticoids^81,82. Activation of the endogenous fight or flight adrenergic and glucocorticoid responses seems to be involved in strengthening these memories after the event. Notably, some of these manipulations do not impair consolidation of different types of memory, such as pavlovian fear conditioning, in contrast to their effects on consolidation of inhibitory avoidance learning. This difference in the animal models may have to do with the different neural circuits involved in cued fear conditioning and inhibitory avoidance learning.

Fear consolidation (at least with certain forms of fear training such as inhibitory avoidance learning) is blocked after training by an antagonist of noradrenergic activation. Propranolol, a common β-blocker used for hypertension, can block central β₁ and β₂ adrenergic receptors. Clinical trials are testing the use of propranolol to prevent the formation of PTSD after trauma. In one study, subjects who presented to an emergency room immediately after a traumatic event (primarily motor vehicle accidents)⁸³ randomly received either propranolol or placebo, orally four times daily for 10 days, followed by a taper period. The first dose of medicine was administered, on average, 4 hours after the traumatic event. One month after the trauma, PTSD measures trended lower in the 11 completers who took propranolol compared to the 20 completers who did not. Also when exposed to script-driven mental imagery of the trauma, a significant number of those who took placebo had ongoing physiological symptoms of PTSD, whereas none of the subjects who took propranolol did. Given the length of time that subjects took propranolol, and given that only 1 or 2 doses would have occurred during the expected period of early memory consolidation, it is unclear whether these effects are due to ‘blockade of consolidation’ versus ‘prevention of stress sensitization’—another hypothesis of PTSD development—but the pilot data are intriguing.

A second open-label study examined 19 completers, 11 of whom took propranolol three times daily beginning 2–20 h after the trauma for 7 days followed by a taper period⁸⁴. At 2 months after trauma, levels of PTSD symptoms were significantly different in the subjects treated with propranolol. These studies were not blinded and had small numbers of subjects. Furthermore, given that propranolol is used frequently to treat hypertension and social anxiety disorder, significant amnestic properties might have been expected to be clinically obvious by now. Larger randomized clinical trials are ongoing to examine whether these promising pilot findings are robust and reproducible in a larger controlled and blinded design.

Another consolidation-blockade approach is based on the findings that lower cortisol levels after trauma predict subsequent PTSD^85,86. Combined with a large body of data that glucocorticoids are involved in emotional memory encoding and retrieval and that high doses of systemic glucocorticoids decrease the memory-enhancing effects of catecholamines⁸⁷, this finding indicates that glucocorticoids given after trauma may enhance normal stress responsiveness and prevent overconsolidation of trauma memories. Pilot data in humans support these hypotheses. Exogenously administered stress doses of cortisol reduced the development of subsequent PTSD in medical-surgical patients after septic shock⁸⁸. A randomized, double-blinded study replicated the finding with 11 patients that received placebo and 9 that received hydrocortisone for 6 days after septic shock. Thirty-one months later, significantly more subjects who had received placebo compared with hydrocortisone had PTSD symptoms from their intensive-care hospital experience⁸⁸. These studies also have small numbers of subjects, yet provide intriguing early data.

Giving corticosterone to rats following a reminder trial 24 h after fear conditioning⁸⁹ indicates that enhanced glucocorticoid modulation may enhance the extinction component of fear, similarly to the mechanisms described above. However, the inverted ‘U’ relationship of the glucocorticoid system and the complexities of consolidation, extinction and reconsolidation make the story complicated. In a similar protocol, inactivating glucocorticoid receptors with an antagonist led to a disruption of memory retrieval when tested later⁹⁰. Although these results at first seem contradictory, they are consistent with the increasing data implying that every memory recall event is accompanied by competition between molecular events that strengthen the original memory (reconsolidation) and those that inhibit that memory (extinction)⁷⁰. By differentially enhancing or inhibiting these processes, opposing effects on the state of the existing fear memories may be obtained.