Background

Feeding behavior is controlled by multiplex neural circuitry throughout the brain, including the hypothalamus, hindbrain, and mesolimbic dopamine circuitry (Sohn et al., 2013). While dedicated feeding circuits exist in the brain, substantial experimental evidence indicates that these are bidirectionally connected to neural circuitry controlling stress and emotion (Sweeney and Yang, 2017). These interactions are especially important when considering the biological mechanisms governing neuropsychiatric eating disorders such as anorexia nervosa (AN), since eating disorders are characterized by alterations in both feeding and emotional state (Keski-Rahkonen, 2007). As such, the underlying biological causes of eating disorders may be mediated by neural circuit alterations in brain circuits at the intersection of feeding and emotion. Thus, mouse models of stress-induced anorexia have value in dissecting the neural circuit and molecular mechanisms for normal and pathological feeding responses to stress.

Anorexia nervosa is a serious neuropsychiatric disorder characterized by reduced feeding and body weight and co-morbid neuropsychiatric traits, such as increased anxiety and depressive-related behaviors (Bulik et al., 2006). The disease occurs in approximately 0.3%–1% of women in the United States and is associated with a high rate of morbidity and mortality (Bulik et al., 2006). Since the disease is characterized by reduced feeding in the context of psychological and psychosocial stress, substantial efforts have been made to model aspects of AN in rodents (Scharner and Stengel, 2020; Zhang and Dulawa, 2021). Although AN is a complex disease with human-specific etiology, important components of the disease can be effectively modeled in rodents, such as reduced feeding in the face of stress and anxiety, compulsive exercise, and elevated anxiety and depressive behavior. These models provide an important entry point towards understanding the biological mechanisms mediating dysregulated eating and may provide a pathway towards novel therapeutic options for eating disorders.

One commonly used rodent model of AN is the activity-based anorexia (ABA) mouse model (Carrera et al., 2014). The ABA model subjects rodents (mice or rats) to restricted food access in the presence of a running wheel. In this model, a subset of mice will stop eating and exercise excessively, leading to starvation and death. This model is useful as it incorporates multiple characteristics of AN, such as excessive exercise and voluntary anorexia, and is more effective in female rodents (approximately 90% of AN cases occur in females). However, the model relays on the presence of a running wheel and restricting food access to a user-defined time window (usually between 1 and 4 h/day). Rodents in this model develop gradual hypothermia and may exercise excessively to generate heat. As such, the ABA model does not work when supplemental heat is provided to the rodents (Gutierrez et al., 2008). Thus, although hypothermia is often observed in AN, the ABA model may more closely model physiological forms of anorexia resulting secondarily to hypothermia. Further, the model can be difficult to implement in mice (as opposed to rats), and only a subset of mice are sensitive to the ABA model.

Given the limitations of the ABA model, additional attempts have been made to develop mouse models that capture the AN phenotype more accurately. One model that shows promise is the “GED” (gene, environment, deprivation) model developed in the lab of Lori Zeltser (Madra and Zeltser, 2016; Madra et al., 2015). The GED model uses a transgenic mouse line containing a BDNF-Val66Met mouse mutation that has been associated with neuropsychiatric disorders. In the GED model, the BDNF-Val66Met mice are subjected to social isolation during adolescence (environmental manipulation) and given 20%–30% calorically restricted food access during a user-defined time window of 3 h twice a day. Thus, this model incorporates a genetic component (BDNF-Val66Met mutation), an environmental stressor (social isolation), and food deprivation by both caloric restriction and limited food access to produce anorexia-related behaviors. A subset of mice exposed to the GED model voluntarily stop eating and develop severe anorexia. Like in the ABA model, in the GED model the voluntary starvation phenotype observed in AN is captured and the starvation phenotype is more common in female mice. However, this model relays on a transgenic mouse line, which limits the use of some modern behavioral neuroscience approaches, such as optogenetics and chemogenetics, which often require additional transgenic mouse lines and/or multiple breeding steps. Therefore, there is a need for simplified stress-induced anorexia models which can be implemented in all basic biology labs and transgenic mouse lines (Table 1).

Here, we describe two simple models of stress-induced anorexia that are easy to implement in mice: restraint stress–induced anorexia and social isolation–induced anorexia. These models do not require specialized equipment or transgenic mouse strains and can be implemented in most biology and neuroscience laboratories. Both of these assays may serve as useful models for disorders characterized by reduced feeding in the face of physical or psychosocial stress, such as the reduced appetite associated with stress-related mood disorders or the impaired food-seeking and consumption observed in AN.

Table 1. Strengths and weaknesses of common models of stress-induced anorexia

Subjects and housing: Adult male and female C57/BL6J littermate mice may be used for both social isolation and restraint stress–induced anorexia studies. Given that neuropsychiatric eating disorders are highly sexually dimorphic, it is important to include both sexes in the experimental design and to determine if behavioral phenotypes are differentially expressed in males and females. For restraint stress–induced anorexia assays, littermate mice should be group-housed with 3–4 mice per cage. Lights in the housing facility should be on a 12:12 h light/dark cycle. Since mice are nocturnal and consume most of their food during the dark period, it is recommended to perform feeding assays at the start of the dark cycle, although light cycle measurements may also be performed.