50 years of hurdles and hope in anxiolytic drug discovery

Preclinical measures of anxiety

Numerous preclinical tests for anxiety have been developed, and the specifics of these tests have been described in many comprehensive reviews^12–14. Here, we only briefly introduce the most frequently used tests (FIG. 2; TABLE 2) to illustrate the strengths and weaknesses of current approaches.

One general consideration from the outset is validity. The validity of a test for anxiety in an animal rests on three criteria: face validity (does it measure something analogous to one or more human anxiety symptoms?), predictive validity (is it reliably sensitive to clinically efficacious anxiolytics?) and construct validity (does it involve some of the same pathophysiological mechanisms found in human anxiety disorders?)¹⁵. None of the currently available tests or models of anxiety (see below) can be said to unequivocally meet these criteria.

Approach-avoidance conflict tests — a group of tests that have been a mainstay of preclinical anxiety research for many years¹⁶ — assay anxiety-like behaviour in rodents by generating a conflict between a drive to approach novel areas and, simultaneously, to avoid potential threat therein. These simple tests, which include the well-known novel open-field test, the elevated plus-maze test and the light/dark exploration test, were invented in the 1980s to exploit the natural tendency of rats¹⁷ and mice^18,19 to prefer enclosed areas over exposed and/or elevated places. Among the different anxiety disorders, the tests are thought to most closely model GAD and specific phobias, largely based on their perceived face validity and sensitivity to benzodiazepine anxiolytics. These tests have been used in nearly 4,000 drug discovery experiments and continue to be very popular. Indeed, well over half of the rodent-based experiments on anxiety-related drugs have used one or more of these tests; among them, by far the most commonly used ones have been the elevated plus-maze test and the light/dark exploration test.

The term ‘conflict-based test’ is also often used to describe measures of behaviour suppression by mild electric shock. This group includes the Vogel conflict test²⁰ and the Geller-Seifter²¹ test, which measure anxiolytic-like activity via the maintenance of a behavioural response (for example, licking or bar pressing) despite the receipt of a shock. These putative GAD-related tests were part of many drug discovery programmes in the 1980s and 1990s but have since fallen out of favour, perhaps because they require animals to be trained over multiple days and are more labour-intensive and time-consuming than the approach-avoidance tests.

Some anxiety tests have been designed to tap into the fundamental defensive responses shown by animals in the face of immediate danger. Such defensive or ‘fear’ behaviours can be conceptually distinguished from the anxiety states produced by less imminent and more ambiguous threats¹⁶, and may be most relevant to anxiety disorders such as panic disorder and PTSD. For example, the ‘Mouse Defense Test Battery’ (MDTB) was designed to provide multiple measures related to fear and anxiety, based on observations of how wild rodents respond to danger²². In this task, mice are placed in an oval runway and tested for their responses (fight, flight, freeze, vocalize or scan) to an approaching anaesthetized rat (a natural predator). Specific behavioural measures in the MDTB are sensitive to specific classes of anxiolytic medication. For example, benzodiazepines that are effective in GAD reduce mouse risk assessment, whereas serotonergic agents that are efficacious in panic disorder and PTSD attenuate fight and flight behaviours²³. In spite of these promising results, however, the MDTB has not been widely adopted, again probably owing to the training and technical demands involved. As a practical compromise, researchers have incorporated measures derived from the analysis of defensive behaviours, such as risk assessment, into anxiety-related tests such as the elevated plus-maze test; in some cases, this has resulted in improved sensitivity to certain anxiolytic drug classes²⁴.

Another set of fear-based tests that are relevant to PTSD and specific phobias involve variations on classical Pavlovian fear conditioning. Here, an animal learns to associate a context or a specific environmental stimulus (for example, a light or a sound) with electric shock to produce a conditioned fear response that can be quantified in various ways (for example, freezing, escape, avoidance or startle). Studies of Pavlovian fear conditioning have contributed greatly to our understanding of the basic neural circuitry and molecular mechanisms of memory, but they have not been traditionally considered to be among the ‘classical’ tests in anxiolytic drug discovery. This may be changing, however, with the recent focus on devising ways to pharmacologically attenuate fearful memories through the process of reconsolidation or extinction (see below)²⁵ and, more generally, through a growing appreciation of abnormal learning and cognition in anxiety.

Preclinical anxiety models and endophenotypes

Tests or assays for anxiety, in which the animal is placed in an experimental situation to evoke an acute anxiety-like response, can be distinguished from models of anxiety, in which an animal has been manipulated in some way to produce a more lasting or permanent increase in anxiety. The goal of anxiety models is to produce a form of abnormally elevated anxiety that more closely resembles, by definition, the pathological nature of human anxiety disorders. This can be achieved, for instance, by acutely or chronically subjecting animals to stressors before testing^26,27. Another approach involves identifying genetic populations (inbred and selectively bred strains)^28,29 or engineering mutant mice with innate anxiety-like phenotypes (TABLE 3). This approach has proven to be valuable for screening novel anxiolytics^30–32 and testing the pharmacoselectivity of putative anxiolytics³³; emerging genetic technologies such as optogenetics³⁴ will be integral to future basic anxiety research^35,36.

The term endophenotype (an immediate phenotype or biomarker) describes a premorbid or symptomatic behavioural, neural or biological feature of an anxiety disorder that, in principle, is more easily quantified than the disorder as a whole. An example of a neural intermediate phenotype in panic disorder and certain other anxiety disorders is the exaggerated blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) amygdala response to threatening stimuli³⁷. In rodents, specific behavioural measures can also be viewed as endophenotypes of anxiety symptoms; for example, risk assessment and flight in the MDTB task may relate to threat avoidance and hypervigilance in GAD and panic disorder, respectively²².

There is growing interest in identifying anxiety endo-phenotypes that are comparable across rodents and humans with a view to foster translation (TABLE 4). A good example that has grown in popularity is the extinction of fear memories. Extinction is an extension of the aforementioned conditioned fear paradigms and is typically assessed by measuring the decrease in a fear-related behaviour (for example, freezing or a startle response) following repeated presentation of an environmental cue or context that is associated with an aversive event (for example, electric shock). Given that extinction has a close therapeutic analogue in the form of exposure therapy, preclinical studies have applied extinction to test for drugs that function as adjuncts to strengthen extinction and reduce intrusive fear memories in PTSD and specific phobias^38,39. There has been encouraging progress in the development of anxiolytics (for example, D-cycloserine) based on preclinical findings that have used extinction as a paradigm⁴⁰.

Below, we assess the preclinical evidence that has accrued — using these and other preclinical approaches — on the neurotransmitter systems that have been the main targets of anxiolytic drug discovery.

GABA–benzodiazepine system

Benzodiazepines such as chlordiazepoxide and diazepam have been reference anxiolytics for over 50 years. These drugs exert their effects by allosterically activating specific GABA_A receptor subtypes to promote inhibitory neurotransmission in the brain. Benzodiazepines are efficacious in the acute treatment of GAD, SAD and panic disorder but have limited to no efficacy in other anxiety conditions^41,42. In addition, the long-term use of benzodiazepines is hampered by the occurrence of troublesome side effects, including sedation, memory disturbances, tolerance and dependence liability⁴³ (TABLE 1).

The inherent therapeutic limitations of benzodiazepine anxiolytics led to the search for compounds that were chemically unrelated to the benzodiazepines, with more specific therapeutic actions and without their concomitant unwanted effects. As a result, novel compounds were developed to preferentially bind to specific GABA_A receptor subtypes, to combine preferential affinity and differential intrinsic activity at these receptors or to display low efficacies at each GABA_A receptor subtype⁴⁴. A comprehensive programme of preclinical research provided very encouraging results and led to clinical studies of partial agonists of GABA_A receptors or agonists of GABA_A receptor α2 or α3 subunits for GAD⁴⁴. However, none of these drugs has reached the market.

The development of some compounds, such as the benzodiazepine receptor partial agonist bretazenil (a benzodiazepine derivative)^45,46 and the GABA_A receptor α2 and α3 subunit agonist SL651498 (REF. 47), was discontinued owing to unexpected sedative and/or amnestic effects. Ocinaplon⁴⁸, which combines preferential affinity and differential intrinsic activity at GABA_A receptors, failed clinically owing to toxicity, as did the GABA_A receptor α2 and α3 subunit agonist TPA023 (REF. 49), despite exhibiting anxioselective activity in GAD. The mitochondrial benzodiazepine receptor agonist XBD-173 (REF. 50) also failed in a Phase II trial for GAD, although this may have been attributable to the choice of outcome measure (CCK-induced panic) and because the trial was not controlled for the presence of a genetic polymorphism moderating the binding of the drug⁵¹. Nonetheless, these disappointments have been a major reason why pharmaceutical companies seem to have abandoned the development of drugs targeting the GABA–benzodiazepine system for anxiety; to our knowledge there are no drugs targeting this system currently under development.

5-HT

The serotonin (5-HT) system has long been implicated in the mediation of anxiety⁵². For example, genetic variation in the human 5-HT transporter and in the 5-HT_1A receptor influences anxiety traits^53,54, and knockout mice lacking the genes encoding the 5-HT transporter and the 5-HT_1A receptor show increased anxiety-related behaviour^30,55,56. 5-HT is also a primary target of existing anxiolytic medications. Indeed, the 5-HT_1A receptor partial agonist buspirone was the first pharmacotherapeutic alternative to benzodiazepines for the treatment of GAD. It was first described by Goldberg et al.⁵⁷ and later shown to have anxiolytic efficacy in controlled clinical studies⁵⁸ before being launched in 1985 by Kwizda Pharma. Buspirone and other partial agonists of the 5-HT_1A receptor may exert anxiolytic activity via the activation of 5-HT_1A heteroreceptors in forebrain areas^59–61. However, drugs targeting the 5-HT_1A receptor have failed to demonstrate efficacy in other anxiety disorders, such as panic disorder or OCD⁶², and their utility is further limited by extensive first-pass hepatic metabolism⁶³.

The serendipitous observation that antidepressants such as tricyclic antidepressants or monoamine oxidase inhibitors have anxiolytic properties⁶⁴ stimulated research on the anxiolytic properties of newer-generation, better-tolerated antidepressants such as SSRIs^65,66. SSRIs are thought to exert their therapeutic effects by increasing extracellular 5-HT levels⁶⁷. This class has proven to have efficacy across a range of anxiety disorders, and fluoxetine was the first SSRI to be approved for GAD in 1999 (REFS 41,42). Today, SSRIs are a first-line treatment for many anxiety disorders and are some of the most commonly prescribed medications in the field of psychiatry. However, many patients do not respond to SSRIs, and adverse effects such as sexual dysfunction and a delayed onset of action — sometimes associated with a transient period of increased anxiety — have reduced the acceptability of SSRIs in clinical practice⁶⁸.

A vast amount of preclinical pharmacological data has been accumulated on the effects of 5-HT-interacting drugs in anxiety-related procedures (Supplementary information S1 (box)). FIGURE 1 shows that the number of experiments focusing on 5-HT was the highest during the 1990s and, despite a decrease since the early 2000s, 5-HT remains the primary focus of drug testing in pre-clinical anxiety research. Not surprisingly, given their clinical success, studies on the anxiety-modulating actions of 5-HT-targeting drugs predominantly examined 5-HT_1A receptor agonists and SSRIs, typified by buspirone and fluoxetine, respectively. The former (buspirone) has been, by far, the most studied anxiolytic outside the benzodiazepine class. Anxiolytic-like properties of buspirone and other 5-HT_1A receptor agonists have been reported in about two-thirds of experiments. However, there are also reports that 5-HT_1A receptor agonists induce pro-anxiety effects, and several studies did not reveal any modification of anxiety-like behaviours by these drugs (FIG. 3).

The effects of SSRIs are also inconsistent. Anxiolytic-like actions were observed in approximately 40% of the experiments conducted, whereas 20% reported anxiogenic-like effects and the remainder failed to detect any behavioural changes. 5-HT₂ and 5-HT₃ receptors have also been proposed as potential targets for anxiolytics but, again, compounds with high affinity and selectivity at these receptors produced equivocal results in preclinical experiments (FIG. 3).

What might account for these inconsistencies? Variability in experimental conditions across laboratories has often been cited as a potential influence on rodent anxiety-like behaviour⁶⁹. In the case of 5-HT_1A receptor agonists, increasing lighting levels in the elevated plus-maze test can switch an anxiogenic-like effect to anxiolytic-like activity⁷⁰, and manipulating shock associations in a conditioned suppression task can transform an inactive drug profile into an anxiolytic-like profile⁷¹. Findings such as these raise the possibility that certain key procedural factors, notably those affecting stress, determine the magnitude and direction of the anxiety-related effects of drugs acting on the 5-HT system. Although it remains to be thoroughly investigated, this attractive hypothesis is in line with the known, complex, stress-modulating role of the 5-HT system⁷² and could have important implications for the design and choice of the animal model used in studies of 5-HT-targeting anxiolytics.

Overall, despite the intense focus on 5-HT receptor ligands, only the 5-HT_1A receptor agonist tandospirone has made it to the market, and only in Japan and China. Agomelatine is an agonist of melatonin MT₁ and MT₂ receptors and an antagonist of the 5-HT_2C receptor that has been developed and launched in Europe as an antidepressant; it has also demonstrated efficacy in a Phase II trial in GAD⁷³, but the exact contribution of 5-HT_2C receptor antagonism to these anxiolytic effects is unclear. Some other drugs that either selectively or non-selectively target 5-HT receptor subtypes or modulate 5-HT reuptake are in active clinical development for anxiety disorders (TABLE 5). However, the anxiolytic effects of these drugs in preclinical settings have not been reported in the published literature.

Neuropeptides

The field of neuropeptide research has seen considerable progress in the past two decades, with the identification of new centrally expressed peptides and the elucidation of their functions using genetic manipulations and newly developed specific receptor ligands^74–76. Almost 20 different peptide systems have been suggested to have a role in the modulation of anxiety (FIG. 4). This line of research was driven by the finding that these neurotransmitters and neuromodulators, as well as their receptors, are found in areas of the brain that are implicated in the control of anxiety. Further support emerged from studies showing that the central infusion or genetic manipulation of neuropeptides modified anxiety-related behaviours^77,78. A detailed review of the vast preclinical literature in this area is beyond the scope of this article (for a full summary of experiments, see Supplementary information S1 (box)). Below, we consider three of the most intensively studied neuropeptides — CCK, CRF and tachykinins — and we also mention some other promising neuropeptides such as neuropeptide Y (NPY).

CCK was the first peptide to be discovered in the central nervous system (CNS)⁷⁹, where it is abundantly distributed and binds to two receptor subtypes, the CCK₁ receptor and the CCK₂ receptor, with the latter having a much broader distribution pattern⁸⁰. Initial research focused mainly on CCK and the development of selective CCK₂ receptor antagonists as potential anxiolytics; this generated much interest in the late 1980s through to the 1990s. These compounds produced anxiolytic-like effects in less than two-thirds of the experiments; the remainder of experiments failed to detect any behavioural changes, and some even showed pro-anxiety effects (FIG. 3). The results of CCK₂ receptor deletion in animal models of anxiety are similarly discrepant, with both anxiolytic- or anxiogenic-like effects being reported, and about half of studies have shown no clear change in anxiety-like behaviour (FIG. 3). This questioned the idea that CCK represents a valid target for anti-anxiety medications. Clinical trials undertaken with CCK₂ receptor antagonists in anxiety disorders, including GAD and panic disorder, have also been unsuccessful^77,81 and no CCK-based drugs have yet been approved.

CRF is the major physiological regulator of the stress response⁸² and has been one of the most studied neuropeptides in anxiety (FIG. 4). CRF binds to at least two receptors: the CRF₁ receptor and the CRF₂ receptor. Most preclinical studies have focused on the CRF₁ receptor because it is expressed at high density in corticotropic cells in regions of the brain that mediate anxiety⁸³. Anxiolytic-like effects of CRF₁ receptor antagonists have been reported in the majority of preclinical experiments (one-third of these experiments failed to detect effects), which is consistent with the anxiolytic-like phenotype of Crfr1-null mutant mice (FIG. 3). Echoing the literature data on 5-HT, stress might have a strong influence on the effects of these drugs. CRF₁ receptor antagonists most reliably produce anxiolytic-like effects under conditions of elevated stress (for example, in tests involving predator or shock exposure) or in animals displaying excessive CRF–CRF₁ receptor signalling (for example, CRF-overexpressing mice)^78,81. As in the case of drugs that target 5-HT, the choice of experimental models is therefore critical for the accurate assessment of the anxiolytic potential of targeting CRF. This issue could also extend to clinical studies. Controlled trials with CRF₁ receptor antagonists in anxiety disorders such as GAD and SAD have yielded negative results, but these studies were carried out in heterogeneous patient groups⁸¹, raising the question of whether effects would be more readily detected in patient subpopulations with the highest levels of anxiety.

The tachykinins substance P (also known as neurokinin 1), neurokinin A (NKA; also known as tachykinin precursor 1) and neurokinin B (NKB; also known as tachykinin 3) are widely distributed in the CNS⁸⁴. Substance P and NKA, along with their respective receptors tachykinin receptor 1 (TACR1; also known as the NK₁ receptor) and TACR2 (also known as the NK₂ receptor), are especially well expressed in structures of the brain that are implicated in anxiety, including the amygdala and septum⁸⁵. Several non-peptide antagonists at NK₁ and NK₂ receptors produced anxiolytic-like effects in a little more than half of the experiments (FIG. 3; Supplementary information S1 (box)). Variability in the outcomes of these studies seems to be highest when certain behavioural assays are utilized, in particular the elevated plus-maze test and social interaction tests. NK₂ receptor antagonists seem to have more reliable anxiolytic effects in tests involving strong or explicit stressors, such as in the MDTB. However, late-stage clinical trials with NK₂ receptor blockers have shown either negative or inconclusive results in GAD, SAD and PTSD⁸¹. Thus, selective blockade of tachykinin receptors may be insufficient to achieve therapeutic efficacy^81,86. Differences in tachykinin receptor physiology between rodents and humans have also been suggested to account for at least some of the failure to translate preclinical data on this target to the clinic⁸¹.

Other neuropeptides that have been studied for their anxiolytic potential include NPY⁸⁷, nociceptin⁸⁸, galanin^75,89, melanin-concentrating hormone (MCH)^90,91 and neuropeptide S (NPS)⁹² (FIG. 4; Supplementary information S1 (box)). These peptides and their receptors are densely expressed in various regions of the brain that mediate anxiety. Preclinical experiments have investigated the administration of nociceptin, galanin, NPS or non-peptide ligands of their receptors either directly into the brain or — in the case of putatively brain-penetrant compounds — systemically; however, these experiments have not produced consistent effects on anxiety-related behaviours. Perhaps more promising are the results from the administration of MCH receptor antagonists, which have demonstrated anxiolytic-like effects in about three-quarters of preclinical experiments conducted to date (FIG. 3). The literature data on NPY is also encouraging⁹³. Based on around 100 pharmacological and gene mutant experiments, many of which have been conducted in recent years (FIG. 4), the preclinical evidence supports the potent anxiolytic actions of NPY (BOX 1). However, although there are some promising lead compounds, there are no drugs targeting NPY, or any other neuropeptide, currently undergoing clinical evaluation for anxiety disorders (TABLE 4).

Glutamate

Multiple lines of evidence strongly implicate glutamate — the major excitatory neurotransmitter system in the brain — in anxiety disorders. There are abnormal levels of glutamate and various glutamate receptor classes in the brains of patients with anxiety disorders, and glutamate levels are altered in rodents by stressors⁹⁴. However, delineating the contribution of the glutamate system to anxiety is a formidable task, given the large number of signalling receptors involved in glutamate neurotransmission. The glutamate system has nonetheless emerged as an increasingly active area of preclinical research within the past decade, with around 100 experiments conducted in 2012 alone (FIG. 1).

Metabotropic glutamate receptors (mGluRs), particularly mGluR1, mGluR2, mGlu3 and mGluR5, have been well studied preclinically and shown to have a role in anxiety behaviour. Orthosteric agonists, negative allosteric modulators or antagonists at mGluR1 (for example, JNJ16259685 and LY456236), at mGluR2 and mGluR3 (for example, LY354740) or at mGluR5 (for example, 2-methyl-6-(phenylethynyl)pyridine (MPEP)) have shown anti-anxiety effects across various rodent assays⁹⁴. Although there have been some negative results, around 80% of studies have been positive, with MPEP being particularly notable for its robust anxiolytic-like activity (FIG. 3; Supplementary information S1 (box)). MPEP was in preclinical development by Merz Pharmaceuticals but the drug was discontinued (for as yet undisclosed reasons) before entering clinical trials. Drugs acting at other mGluRs, including mGluR7 agonists (AMN082), have not been studied in as much depth and their effects still need to be clarified⁹⁴. Clinically, some mGluR compounds, such as the mGluR2 and mGluR3 orthosteric agonist LY354740 (or its pro-drug LY544344), have produced encouraging preliminary results in GAD⁹⁵ (but not in panic disorder)⁹⁶, which have been somewhat tempered in some cases by pro-convulsant activity in animals⁹⁵. Clinical trials are currently underway for the mGluR2 positive allosteric modulator ADX-71149 and for the mGluR1 and mGluR5 antagonist RGH-618 in anxiety disorders (TABLE 5).

The NMDA (N-methyl-D-aspartate) receptor antagonist ketamine was recently found to exert rapid antidepressant effects in treatment-resistant major depression⁹⁷. This has generated considerable interest in NMDA and AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) receptors as targets for depression and is likely to provide insights into the anxiety-related effects of these compounds, for example, based on the effects observed in patients with comorbid depression and anxiety who receive ketamine⁹⁷. In addition, the preclinical literature on the anxiolytic-like effects of NMDA and AMPA receptor antagonists has substantially grown in recent years. For example, the non-selective NMDA receptor channel blocker MK-801 has shown anti-anxiety effects across several assays, and NMDA receptor blockers have shown anxiolytic effects in around three-quarters of studies (FIG. 3; Supplementary information S1 (box)). Because indiscriminate blockade of NMDA receptors is unlikely to be a well-tolerated option for an anxiolytic, compounds that target specific NMDA receptor subunits (for example, the NMDA receptor subunit NR2B antagonist ifenprodil) have been studied but they do not produce comparably robust effects (Supplementary information S1 (box)). Similarly, the anxiety-related preclinical effects of AMPA receptor antagonists such as NBQX (2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f] quinoxaline-2,3-dione) have overall proven to be inconsistent (Supplementary information S1 (box)).

Out of the other potential glutamate-acting targets for anxiety, D-cycloserine — which potentiates NMDA receptor signalling via the glycine co-agonist site — has, as already noted, shown efficacy as a therapeutic adjunct in various anxiety disorders^98,99. Another glycine-acting drug, bitopertin (RG1678), which inhibits glycine reuptake by glycine transporter 1, is currently being investigated for efficacy in OCD (TABLE 5), but the class of glycine transporter 1 inhibitors has produced mixed preclinical data. Last, pregabalin, riluzole and topiramate are three drugs that exert glutamatergic effects as part of a complex pharmacological profile; pregabalin is approved (in Europe) for GAD, whereas all three are undergoing proof-of-concept studies for PTSD and SAD, with the caveat that the precise contribution of glutamate to their anxiolytic actions remains unclear.

Endocannabinoids

Endocannabinoids represent another system that has attracted attention in recent years as a potential target for novel anxiolytics (FIG. 1). The endocannabinoids anandamide (also known as N-arachidonoylethanolamide) and 2-arachidonoylglycerol, and their principal CNS receptor (the cannabinoid 1 (CB₁) receptor), are densely expressed in the brain, particularly in regions mediating anxiety¹⁰⁰. Further implicating this system as a relevant translational target, there is growing evidence that abnormalities in the CB₁ receptor and other endocannabinoid systems are implicated in anxiety disorders such as PTSD^101,102. The effects of CB₁ receptor agonists, inverse agonists and antagonists on anxiety-related behaviours have been intensively studied across a range of preclinical assays and models, with mixed results. There are examples of CB₁receptor ligands and gene mutations producing either anxiolytic-¹⁰³ or anxiogenic-like^104,105 effects in rodents¹⁰⁶ (FIG. 3).

Part of the complexity of the anxiety-related effects associated with manipulating CB₁ receptors is very likely to stem from the ubiquitous expression of CB₁ receptors in different anxiety-mediating regions and circuits of the brain, some of which may have opposing roles in anxiety (for example, cortical regions versus the amygdala, and GABAergic circuits versus glutamatergic circuits)¹⁰⁷. In addition, the enthusiasm for developing agents that target the CB₁ receptor was tempered by the withdrawal of the CB₁ receptor antagonist rimonabant (also known as SR141716) from the market as an anti-obesity medication owing to depression, suicidal ideation and anxiety symptoms in the patient populations receiving the drug¹⁰⁸.

An alternative approach for pharmacologically modulating endocannabinoids is to target their post-release reuptake and degradation. Endocannabinoids are thought to be primarily released ‘on demand’ as a function of physiological requirements. Therefore, pharmacologically inhibiting their reuptake or degradation could augment functionally relevant recruitment of endocannabinoids and produce more selective effects on anxiety than CB₁ receptor agonists. Although this is an attractive hypothesis, preclinical studies have not shown robust anxiety-related effects of, for example, compounds that augment anandamide via inhibition of the catabolic enzyme fatty acid amide hydrolase (FIG. 3; Supplementary information S1 (box)).

More promising are the recent findings that both anandamide transporter blockers (such as AM404) and fatty acid amide hydrolase inhibitors (such as AM3506 and JNJ-5003) promoted the extinction of rodent fear¹⁰¹ and prevented stress-induced anxiety-like behaviour¹⁰⁹. These preliminary observations suggest that this class of compounds may be preferentially active under conditions of high stress and abnormal endocannabinoid tone¹¹⁰. The anxiolytic potential of fatty acid amide hydrolase inhibitors is currently being investigated in early-phase clinical trials, and it remains to be confirmed whether this or other approaches to targeting endocannabinoids¹¹¹ will prove to be an effective translational strategy.

Lessons learned and future perspectives

Taking stock of half a century of intensive research, where does the effort to find effective medications for anxiety disorders now stand? Clearly, there are promising targets in the various neurotransmitter systems discussed above, and there is reason to be optimistic that one or more of these will yield a novel, safe and clinically efficacious anxiolytic. Considering the huge amount of data that has amassed, however, the drug discovery efforts in this field can be, and often have been, viewed as a failure.

However, this conclusion is not unique to the anxiety field; in fact, it has been levelled at most of the drug discovery efforts in psychiatry¹¹². It is worth reiterating the point that finding medications for psychiatric illnesses is made all the more daunting by fluid diagnostic end points that are based almost entirely on behavioural symptomatology rather than on a deep mechanistic understanding of the underlying biology. Indeed, this and various other issues have been offered as explanations for why the search for new anxiolytics has stalled. Some of the issues were reinforced by our systematic analysis of the literature, and below we expand upon three issues that came to the fore.

Concluding remarks

Anxiolytic disorders are serious medical problems that are commonplace and becoming more prevalent in many parts of the world. The growing burden of anxiety disorders demands better treatments but, although the field has promising leads, the efforts to identify new anxiolytics seem to have reached an impasse. Here, we have offered a comprehensive analysis of the published preclinical research conducted to date with the aim of providing an objective analysis of the major trends, biases and limitations within the field in order to help direct a more effective translational approach in the future. We are optimistic that a new generation of preclinical studies that are built around circuit-informed, pathogenic rodent models and strong, bi-directional translational links to clinical research can move us out of the age of anxiety and into the age of discovery.

Supplementary Material