Challenges with modelling anxiety disorders: a possible hindrance for drug discovery

1. Introduction

The previous decade has witnessed a dramatic withdrawal of pharmaceutical companies from psychiatric-based drug discovery predominantly due to the lack of novel-acting therapies [1]. There are numerous reasons for this including a lack of understanding of the underlying etiology of the disorders, as well as the necessity for novel and more translationally relevant animal models of anxiety disorders. It is important to state from the outset that animal models, particularly rodent models, can only model specific aspects of anxiety disorders and not their entirety. In this opinion piece, we discuss the limitations of current tests and avenues that should be explored when planning novel anxiety models and tests.

2. Traditional tests of anxiety

For any tests used for assessment of anxiety-related behavior, it is crucial to establish face validity, construct validity, and predictive validity [2], but none of the existing anxiety tests satisfactorily fulfill all three criteria. Currently available rodent anxiety paradigms are broadly classified into two groups based on their reliance on unconditioned or conditioned responses to aversive stimuli. A comprehensive review of the tests is out of scope here, but can be found elsewhere [3].

The most commonly used unconditioned tests of anxiety – open field (OF), elevated plus maze (EPM), and light/dark exploration (LDB) – are thought to exploit the conflict between the tendency to avoid potentially dangerous spaces and the motivation to explore novel areas. The readout of anxious behavior in these tests is decreased time or entries into the aversive (open/lit) zones.

Conditioned anxiety paradigms assess animals’ conditioned responses to neutral or unpleasant stimuli. In Geller-Seifter and Vogel conflict tests, deprived animals are trained to earn food or water reward, respectively, which in a test session is simultaneously punished with a foot shock. The conflict between hunger/thirst and receiving foot shocks decreases eating/drinking behavior, which is reversed by anxiolytics. In Pavlovian fear conditioning (FC), a neutral unconditioned stimulus (tone or light) is paired with a foot shock, and the conditioned fear response to reexposure to the unconditioned stimulus is measured as an indication of the animal’s anxiety/fear levels. Pavlovian FC is able to model certain aspects of post-traumatic stress disorder (PTSD). In addition, a number of models employing repeated or intense foot shocks and/or other strong stressors (e.g. exposure to predators) have been shown to induce PTSD- and trauma-like states in rodents. Such models have the benefit of having similar factors involved in the induction and maintenance of anxiety/fear across species making them highly attractive models to study the mechanisms of PTSD.

While the aforementioned paradigms are frequently used to assess anxiety in rodents, there are several drawbacks associated with these paradigms, particularly those with unconditioned responses as their readout, which hamper our understanding of the neurobiological and genetic causes of anxiety as well as aggravate drug discovery. Thus, many of the commonly used paradigms are strongly dependent on locomotor activity, which can lead to false-negative or false-positive findings. Furthermore, the tests are sensitive to mild changes in variables (e.g. lighting, experimenter, and microbiome), which hampers their replicability in both an intra- and inter-laboratory manner [4]. Of note, the microbiome appears to play a role in anxiety disorders in humans. Thus, its assessment in animal models may help to uncover a better understanding of the microbiome in anxiety disorders. Treatments commonly show different outcomes in these tests, which may be related to their sensitivity to changes in variables, differences in their underlying circuitry, or genetic makeup. In conclusion, while there are many tests available assessing anxiety-related behaviors in rodents, these have many limitations that hinder their translation into the clinic. Below, we make a number of suggestions on ways to improve anxiety testing in rodents.

3. Expert opinion

While thorough experimental planning clearly helps to address some of these problems (e.g. establishing optimal conditions in pilot studies), we, and others [1,3], believe that a shift in focus on more appropriate paradigms is necessary to avoid or at least diminish the above-mentioned drawbacks. Paradigms that assess unconditioned responses typically measure ‘normal’ anxiety (e.g. induced by novelty and/or bright open spaces), the cause of which is fundamentally different from pathological anxiety in humans [5]. Most of the paradigms assess unspecific/generalized anxiety or fear. This seems inadequate given the complex and diverse clinical presentation of these disorders [6]. Moreover, studies performed in animals ‘straight off the shelf’ simply injected with the compound(s) of interest in such tests are not likely to address the underlying research question. Studies often employ only one or two similar tests; ideally, a number of different tests that utilize a variety of different outcomes as their main readouts should be performed [7].

However, very few tests have been established that are advantageous over the ‘traditional’ paradigms. One example is the novelty-suppressed feeding (or hyponeophagia) paradigm, which does not (heavily) rely on locomotor activity [8]. Additionally, the focus should go beyond models for generalized anxiety. Instead, improved paradigms would investigate a specific anxiety domain, in a manner akin to that proposed in the NIH’s Research Domain Criteria (RDoC) framework [9]. Another example is the recently established social FC (SFC) paradigm, which is a test for generalized social anxiety [10]. However, these tests can still be influenced by the limitations mentioned above, such as microbiome, experimenter, and (to a limited extent) changes in locomotor activity.

As witnessed by these two examples, there are several considerations to make when choosing optimal models including minimization of potential confounds (described above). Additionally, a high translational value, such as predictivity for classical anxiolytics, but also chronic antidepressant treatments is desirable, as is the case for the hyponeophagia and SFC paradigms, particularly given the high comorbidity between anxiety disorders and depression [6,10,11], though it is also possible that the efficacy of antidepressants may be due to a lack of selectivity of these models. It should be cautioned that tests should not be built solely based on their response to currently available drugs. This can lead to a simple ‘me-too’ approach, whereby novel tests may only be sensitive to GABAergic or selective serotonin reuptake inhibitor-type drugs and, therefore, not aid the discovery of novel treatments. Furthermore, a paradigm with high predictive validity for anxiolytics may not necessarily be an appropriate test for assessing the effect of certain manipulations (e.g. gene knockout) on anxiety.

The use of inbred strains in modeling anxiety has the advantage of minimizing genetic variability, allowing precise investigation of the effects of discrete manipulations (e.g. stress, gene knockout, and drug treatment). However, anxiety disorders are likely to have multigenetic etiology, which can be studied using genetically more diverse rodent lines (e.g. outbred and recombinant inbred), with the drawback of increased cost and use of animals. Ideally, identification of similar underlying mechanisms or therapeutic responses using a combination of both approaches would greatly enhance drug development.

Another factor that hampers drug discovery is that the majority of studies tend to be performed in adult male rodents, neglecting two important aspects, namely sex differences and early- and late-age-of-onset [6]. Although it is well documented that anxiety disorders are more prevalent in women [1,6], this fact is largely ignored in preclinical research [12]. In support of this argument, sex-related differences to drug administration have been observed in rodents and humans [12,13]. Moreover, sex-related strain differences in anxiety-related behavior are notable – e.g. fewer female C57BL/6, but not Balb/c, mice show an anxious-like phenotype compared to their respective males [14]. Another consideration is sex differences in pain sensitivity, which are likely to influence paradigms that rely on foot shocks (e.g. FC and the Vogel test) [15]. Further, it is well reported that male and female rodents respond differently to stressors and that not all stressors are effective in both sexes. Such differences are likely to lead to differences in the underlying neural changes and drug responses [4,13], assessment of which will increase our understanding of the stress-related alterations. Several factors are responsible for these sex-dependent differences including hormones and genetic factors and, as has been more recently described, composition of the gut microbiome [16]. Thus, it is important to highlight the necessity to include both sexes in preclinical research, which will allow to uncover mechanisms involved in the higher susceptibility of women to anxiety disorders. This has started to be addressed by a number of funding agencies, such as the NIH and European Commission, but needs to be embraced by the research community.

Last, but not least, dedicated behavioral laboratories and experts to perform these paradigms and address the indicated issues are relatively sparse. Thus, it is important that appropriate resources are allocated to ensure that testing of novel compounds/targets is performed in laboratories with such resources and expertise.

In summary, while tests such as the OF, EPM, and LDB still have their merits and requirement in future studies, optimized paradigms that attempt to better mimic aspects of the human disorders that they are trying to model need to be established. Despite the limitations of rodents’ behaviors in comparison to complexity of human behavior, utilizing tests that cover a range of specific domains relevant to anxiety will improve translation of animal findings into the clinical setting. Moreover, by including both sexes, as well as younger and older subjects in our studies, we feel that additional insights will be gained, which will also aid the way that the clinical trials for potential novel therapies are designed. Therefore, this is a time of great potential in anxiety-based research, and while we face challenges, these are starting to be overcome.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

F Freudenberg is supported by the Deutsche Forschungsgemeinschaft (grant No. FR3420/2–1, FR3420/2–2), while D Slattery and D Aguiar are supported by the Capes-Humboldt Research Fellowship Programme (88881.136136/2017-01). The authors are also supported by the EU Seventh Framework Programme (602805).