The Role of Dopamine in Contributing to Vulnerable and Resilient Phenotypes in a Mouse Model of Anorexia Nervosa

In this issue of Biological Psychiatry, Beeler et al. [1] provide a compelling study of potential risk and resilience to activity-based anorexia (ABA), a rodent model of anorexia nervosa (AN). Traditionally, ABA combines restricted intake with unlimited running-wheel access, which ultimately leads rodents to increase their activity, self-starve, and expire unless removed from the test paradigm. Working to extend the applicability of this model to risk for AN, the authors aimed to determine whether genetically modifying dopamine would have an impact on resilience to the expected course of ABA. Overall findings identified two phenotypes - one group who adapted to ABA and weight stabilized (resilient mice), and another who were notably more vulnerable. Specifically, hyperdopaminergic mice (i.e., bred to have elevated extracellular dopamine) experienced accelerated effects of ABA and lower rates of survival, supporting the possibility that dopamine plays a key modulatory role in the interaction between food restriction and physical activity in AN.

Theoretical underpinnings of ABA

From an evolutionary perspective, animals must search for food when supply is limited, and typically display increased pre-meal ‘food-anticipatory activity’ [2]. The ABA paradigm demonstrates this interplay between restricted intake and increased activity, which in turn informs understanding of AN [3]. However, to date, the underlying pathophysiology of AN remains elusive. While the ABA model mimics some core features of the disorder, no translational biomarkers have been identified that indicate differential risk for ABA in humans. Of note, recent data suggest adolescent female mice who are more prone to ABA-induced weight loss might have a decreased hedonic response to sucrose [4]. Relative to reward response in AN, dopamine has been studied extensively for its role in behavioral motivation processes but much of this study has centered around appetitive behavior [5], not in its potential contribution to activity. As a consequence, much less is known about the role that dopamine may play in specifically impacting onset and maintenance of exercise in AN.

Differences in ABA response relative to age and condition

Beeler et al. [1] provide a series of investigations that varied aspects of the ABA paradigm. In the first set of experiments, the authors studied adolescent and adult mice separately, distributing each of the age groups into four conditions: (i) standard ABA (2h/day food; unlimited wheel access); (ii) unlimited food and wheel access; (iii) food-restricted control (FR, 2h/ day food; unlimited access to a locked wheel); and (iv) unlimited access to food with a locked wheel.

Among adolescent mice, findings indicated that compared to those in the food-restricted condition with a locked wheel (i.e., FR), mice in the ABA condition demonstrated similar steep levels of weight loss, but a significantly decreased chance of survival. During food restriction, there was also an abrupt elevation in light-cycle running (but not dark) for ABA mice. Though light-cycle running is typically considered food-anticipatory activity (i.e., just prior to a shift to the dark cycle), this increase started earlier near the onset of the light cycle, a pattern which increased across days of food restriction. There was no difference between ABA and FR mice in overall food intake; however, there was a significant drop in intake in the ABA mice on the day they were removed from the paradigm. Taken together, increases in light-cycle running seem to be most implicated as contributing to poor ABA outcomes for adolescent mice.

Adult mice participated in a longer food-restriction protocol (10 v. 5 days), allowing for extended evaluation of the nature of ABA-induced weight loss. Overall, adult mice were less vulnerable to developing AN in the ABA paradigm compared to younger mice. In the first few days, both ABA and FR lost weight, and over time, ABA accelerated weight loss compared to FR. However, some mice in each group were able to stabilize weight in days 6–10 and were considered ‘resilient,’ in that they survived until the end of the paradigm. Of note, some FR mice also required removal from the experiment, indicating that within both an exercise (ABA), or non-exercise (FR) condition, phenotypes were evidenced that could be considered either resilient or vulnerable.

There were additional differences noted between ABA and FR mice, and also between ABA-resilient v. ABA-vulnerable mice. Across both ABA and FR groups, resilient mice consumed more food than vulnerable mice. Interestingly, ABA-resilient mice gained more weight than FR-resilient mice in the last few days of the paradigm, suggesting that access to exercise promoted an adaptive response in appetitive behavior for the resilient subgroup. While food restriction increased light-cycle running across paradigms, the increase was modest in ABA-resilient mice but dramatic in ABA-vulnerable mice. Specifically, this study suggests that ABA evokes two distinct changes in running, one that appears to be food-anticipatory activity (associated with resilience), and one that reflects disrupted circadian cycling (associated with vulnerability).

Increased dopamine and impact on ABA vulnerability

The next set of experiments compared two genotypes, adult female dopamine transporter knockdown (KD) mice and wild-type (WT) littermates, in 3 conditions: ABA, FR, or unlimited food and wheel access. In the ABA condition, KD mice demonstrated accelerated weight loss and poorer survival rates than WT. Moreover, compared to some WT mice who were deemed resilient, all KD mice exhibited the vulnerable phenotype. Also in the ABA condition, both vulnerable KD and WT mice showed increases in light-cycle running (but not dark), but this maladaptive behavior change occurred earlier in KD mice.

Similar to the earlier findings in younger mice, there was a positive association between light-cycle running and weight loss, a relation that was stronger in KD than WT mice. Running towards the end of the light cycle typically anticipates food intake but vulnerable mice ran more in response to caloric restriction rather than food anticipation. This activity was enhanced by increased dopamine (as evidenced by KD mice), and only demonstrated in the ABA condition, not in the unlimited food and exercise condition. Together, these findings suggest that dopamine signaling has a unique effect on the interaction of exercise with food restriction (ABA), and not when there is exercise allowed with adequate nutrition.

While the authors did not examine KD adolescents, they did test whether differences in striatal dopamine between adolescents and adults could account for differences in ABA vulnerability. Results did not indicate that basal function impacted response. Instead, it is possible that smaller animals who have higher metabolic rates (adolescents) are less apt to survive in the ABA model. Additionally, greater initial body weight in adult mice may slow decomposition sufficiently for a resilient phenotype to emerge within the ABA paradigm among adults, but not among younger mice. The authors also posit that there may be age-related differences in how caloric restriction and exercise impact dopamine levels during the progression of AN that differentially impact ABA vulnerability.

Extending ABA findings to AN

Rationale for this work includes its potential to enhance the translational utility of ABA as a model of AN. The authors note that the majority of individuals who combine diet and exercise do not develop AN, and suggest that activity provides a supporting role in AN onset for some individuals and not others. Counter to this supposition, evidence for associations between premorbid activity and AN onset in humans is minimal, and limited in its cross-sectional nature [6], and potential for recall bias [7]. Therefore, we might consider that it is not the activity itself that poses risk for human AN, but instead, a more dynamic interaction of activity with other biobehavioral features.

Individuals with AN often present with elevated trait-level anxiety and anhedonic overcontrol [8] that together suggest striatal dysregulation plays a key role in modulating AN behavior [9]. Data from the current study [1] of KD mice suggests that premorbid (i.e., trait) differences in dopamine can modulate vulnerability to AN/ABA. However, it is also possible that the dopamine system also changes over the course of AN. Resilient mice only emerged within the adult paradigm, in part due to the necessity of starting ABA with mice who were older, and more robust to weight loss. As a consequence, we do not learn from this study whether ABA resilience can be determined in adolescence similarly to adults, particularly relative to dopamine function. Given that typical age of AN onset falls in adolescence [10], this may be a particularly important path of inquiry.

The ABA model is limited in that hyperactivity is common in many with AN but not all, and it is not a diagnostic criterion. However, vulnerability and resilience to ABA, as reported by Beeler et al. [1], provides a compelling suggestion that dopamine function may prove to be a translational mechanism by which we will ultimately better understand risk for AN. Open questions remain as to the nature of exactly how dopamine may shift in its function over the course of illness. Further, a priority should be placed on better understanding of how dopamine may differentially impact risk and resilience to ABA across age and vulnerable periods in development.