A transdiagnostic and translational framework for delineating the neuronal mechanisms of compulsive exercise in anorexia nervosa

Abstract

Objective:

The development of novel treatments for anorexia nervosa (AN) requires a detailed understanding of the biological underpinnings of specific, commonly occurring symptoms, including compulsive exercise. There is considerable bio-behavioral overlap between AN and obsessive-compulsive disorder (OCD), therefore it is plausible that similar mechanisms underlie compulsive behavior in both populations. While the association between these conditions is widely acknowledged, defining the shared mechanisms for compulsive behavior in AN and OCD requires a novel approach.

Method:

We present an argument that a better understanding of the neurobiological mechanisms that underpin compulsive exercise in AN can be achieved in two critical ways. Firstly, by applying a framework of the neuronal control of OCD to exercise behavior in AN, and secondly, by taking better advantage of the activity-based anorexia (ABA) rodent model to directly test this framework in the context of feeding pathology.

Results:

A cross-disciplinary approach that spans preclinical, neuroimaging, and clinical research as well as compulsive neurocircuitry and behavior can advance our understanding of how, when and why compulsive exercise develops in the context of AN and provide targets for novel treatment strategies.

Discussion:

In this article, we (i) link the expression of compulsive behavior in AN and OCD via a transition between goal-directed and habitual behavior, (ii) present disrupted cortico-striatal circuitry as a key substrate for the development of compulsive behavior in both conditions, and (iii) highlight the utility of the ABA rodent model to better understand the mechanisms of compulsive behavior relevant to AN.

Introduction

Up to 80% of individuals experiencing anorexia nervosa (AN) engage in compulsive exercise, activity which is undertaken to alleviate or prevent negative affect and/or in the service of weight control (Dalle Grave et al., 2008; Sawyer et al., 2016). Recent trends in this behavior suggest increased incidence, with reports that young females demonstrated a 46.8% increase in compulsive exercise during the COVID-19 pandemic (Miskovic-Wheatley et al., 2022). The occurrence of compulsive exercise in those with AN is associated with increased suicide risk, worse symptoms, and higher rates of relapse (Dalle Grave et al., 2008; Gorrell et al., 2021; Smith et al., 2013) and these individuals are specifically at risk of poorer response to standard treatment (Calogero & Pedrotty, 2004; Dalle Grave et al., 2008). Despite these known vulnerabilities, we have limited understanding of how compulsive exercise arises and can be effectively treated in the context of AN, partly because physical activity has been discouraged within eating disorder treatment and research for some time (Quesnel et al., 2018). Although cessation of physical activity is appropriate in the context of medical acuity, treatments that target this behavior are critically needed, and their development depends on a thorough understanding of when, why, and how compulsive exercise develops in some patients with AN, and not in others. Current treatments for AN have low success rates, in part, because they focus on other eating disorder features without specific attention to exercise, even though exercise is emerging as a fundamental reinforcer of eating pathology (Coniglio et al., 2022). It is in this context that approaches to symptom-specific treatment targets have taken on new significance, with a focus on understanding the causes of compulsive behavior not only beneficial for the treatment of AN, but for a range of disorders characterised by compulsivity e.g., obsessive-compulsive disorder (OCD) and, substance use disorders. Although AN is characterized by certain hallmark features that define its diagnostic criterion, the degree to which obsessive thinking (e.g. fear of weight gain, calorie counting) and compulsive behaviors (e.g. cutting food into small bites, rigid exercise routines) are present varies across individuals. Given this known heterogeneity in the presentation of relative compulsivity among those with AN, different biological mechanisms may underlie the development of specific phenotypes of this disorder in more “compulsive” individuals. This distinction would inform our current understanding of the (elusive) aetiology of AN (Walsh, 2013), and a better recognition of what drives compulsivity-based psychopathology could bridge the translational gap between preclinical and clinical research to aid in the development of novel treatments.

Looking to conceptual models of compulsive behavior in the context of AN, a majority of literature supports considering exercise to be compulsive (Meyer et al., 2011) – defined by repeated selection of a consolidated action (i.e., a habit) that is insensitive to devaluation and importantly, resistant to punishment. In line with this, maladaptive behavior in AN persists at extreme levels despite the negative consequences of deadly weight loss (Lichtenstein et al., 2017). AN has strong genetic links with OCD (Yilmaz et al., 2022), indicating that obsessive thinking and compulsive behavior in both conditions are driven by shared underlying biological mechanisms. This commonality is observed both in terms of high genetic correlation and a sizeable SNP (single nucleotide polymorphism) heritability in the cross-disorder phenotype (Yilmaz et al., 2020) (see Figure 1). The neural circuit most frequently implicated in the pathogenesis of OCD links the orbitofrontal cortex (OFC) to the dorsal striatum (Menzies et al., 2008) and there is emerging evidence that activity in the OFC differentiates individuals with AN that are more prone to habits (Steding et al., 2019). Moreover, both human and animal studies suggest that striatal dopamine (DA) and acetylcholine function are critical for compulsive behavior in the context of eating pathology (Beeler & Burghardt, 2021; Favier et al., 2020; Foldi et al., 2017; Welch et al., 2021), but their influence on exercise in AN has not been addressed. Initial drivers of dietary restriction and exercise in AN are often motivated by various goals for thinness, control, and/or perceived health – that is, these behaviors are classically “goal-directed” (Hare et al., 2009). At some point then, for some individuals, a shift from goal-directed exercise to compulsive exercise (i.e., habit-driven, rather than goal-directed) occurs (Dalle Grave et al., 2008). This process could be intervened upon and targeted with early intervention strategies, if only we knew how to identify it.

Figure 1. Genetic and phenotypic overlap between anorexia nervosa (AN) and obsessive-compulsive disorder (OCD).

Both AN and OCD are hallmarked by repetitive and inflexible patterns of thoughts and behaviors with shared pathological features including increased compulsivity and habit reliance, indicating a clear similarity in presentation across disorders. There are also substantial rates of co-morbidity, with genetic studies revealing high genetic correlations and sizeable single nucleotide polymorphism (SNP) heritability. Articles referenced are specifically in relation to the AN population (Frank, 2021; Kaye et al., 2013; Salbach-Andrae et al., 2008; Steding et al., 2019; Yilmaz et al., 2020; Yilmaz et al., 2022).

We propose that a better understanding of the biological bases of compulsive exercise in AN is possible with the updated use of the activity-based anorexia (ABA) model in combination with incisive techniques to interrogate the striatal control of habit formation in rodents. ABA is uniquely suited to elucidate the neural changes that underlie a shift between goal-directed and compulsive exercise, because rats exposed to ABA conditions are Susceptible or Resistant to weight loss based on the development of compulsive wheel running. The existence of these subpopulations enables us to directly observe alterations in neuronal activation, neurochemical dynamics and behavioral repertoires that occur in animals that develop compulsive exercise compared to those that maintain healthy levels of activity under tightly controlled experimental conditions. We propose directions for future research that build upon the neurobiological underpinnings of OCD, with a view to capitalise on the knowledge gained in this adjacent field with high genetic and phenotypic overlap with AN. Our goal is to apply this conceptual and empirical approach to the study, and ultimately treatment, of compulsive exercise in individuals with AN.

Linking the neurobiological underpinnings of OCD to compulsive behavior in AN

Compulsive exercise in AN persists despite the clear negative consequences of rapidly declining body weight and inappropriate energy balance. Patients with AN often prioritize exercise, experience distressed feelings when unable to exercise or the exercise schedule is disturbed, and exercise behavior persists despite cardiovascular complications and physical injury (Dittmer et al., 2018; Noetel et al., 2017). Those living with AN find exercise more rewarding than controls and have a stronger preference to exercise even at low body weight (Giel et al., 2013). These features of compulsive exercise are in line with a traditional definition of compulsive behavior, which is automated and repetitive, and persists despite adverse consequences (Lipton et al., 2019). What remains unknown is whether a predisposition to compulsive exercise precedes weight loss in some individuals with AN or whether a tendency toward compulsive exercise is triggered by rapidly declining body weight. In support of the former is the genetic overlap between AN and OCD (Cederlöf et al., 2015; Watson et al., 2019), which suggests that individuals with AN have a predisposition to more activity and are more likely to develop OCD-like tendencies regardless of current weight loss. In support of the latter is the suggestion that compulsive behaviors develop when a particular goal-directed behavior (in the case of AN, exercise in pursuit of the goal of weight loss) becomes habitual, but are continued even when the consequences become harmful (i.e., compulsive). This transition from goal-directed behaviors to habits has been identified in patients with substance use disorder and OCD (Sjoerds et al., 2013). Yet, there is no current consensus around a clinical definition or standardized assessment of compulsive exercise in AN (Dittmer et al., 2018). This is despite high prevalence of this behavior across AN and atypical AN samples, with 80% rates of incidence for inpatients (Dalle Grave et al., 2008) and 59–73% for outpatients (Sawyer et al., 2016). This is in addition to the relatively high rates of AN diagnoses (~17–37%) presenting with an OCD comorbidity (Lennkh et al., 1998; Salbach-Andrae et al., 2008), and the presence of OCD traits in childhood being a risk factor for developing AN (Micali et al., 2011). Further, compulsive exercise is an intermediate phenotype predicted by polygenic risk scores for both OCD and AN (Yilmaz et al., 2022), suggesting that exercise behavior may be particularly relevant for probing overlapping biological mechanisms. These conditions are also both characterised by obsessive thinking (Bastiani et al., 1996), although obsessions in AN are generally disorder-specific (i.e. related to body weight, shape and food) (see Figure 1). With these similarities in mind, an understanding of why compulsive exercise develops in AN could be informed by the considerable body of literature concerning the neuronal control of compulsive behavior in the context of OCD; i.e., the fundamental neuroanatomical and neurochemical mechanisms that drive compulsivity.

Compulsive behaviors in OCD have been suggested to derive from an imbalance between goal-directed and habitual control resulting in excessive habit formation with distinct neurobiological underpinnings (Gillan et al., 2011). Goal-directed behavior is flexible and dependent on action-outcome associations, and is thus specific to the outcome of the action (Balleine, 2011). In contrast, habitual behavior is controlled by learned stimulus-response associations and is therefore less sensitive to outcome of the action and so, less flexible. Yet, when habitual control becomes excessive and behavior persists even when an action results in punishment, maladaptive compulsive behavior develops and becomes much more challenging to extinguish (Lipton et al., 2019). Although the biological mechanisms underlying habit formation have been extensively investigated, how these may relate to AN symptomology needs to be further elucidated. Individuals with AN perform on behavioral tasks in such a way that defines subgroups of those who appear to be more habit-driven vs. goal-directed (Steding et al., 2019), highlighting a need to further explore this distinction in order to help tailor therapeutic approaches to individual characteristics. In this forum, we focus on compulsive exercise in AN, but note that this line of inquiry could also be relevant to food-related behaviour and across other eating disorder diagnoses (Lampe et al., 2023). With respect to feeding, individuals with AN report food restriction as a habit that they “do without thinking” (Coniglio et al., 2017; Davis et al., 2020) and binge-eating and purging may both be acts that form a cycle of repetitive and rigid behaviors (Dougherty et al., 2023). Notably, increased reliance on habitual processes in AN is not only shown in feeding-related symptoms, but also demonstrated across other modalities like excessive handwashing (Gupta et al., 1987), suggesting patients with AN broadly tend to rely on habitual control over their behavior, which may lead to compulsive tendencies (Seidel et al., 2022). Hence, it is possible that high reliance on habitual behavior is an adaptive strategy for some patients with AN, considering the challenges these individuals have with other aspects of executive control. Specifically, patients with AN demonstrate deficient goal-directed control despite excessive goal pursuit, impaired cognitive flexibility, and excessive self-control (Foerde, Daw, et al., 2021; Steinglass et al., 2012; Tchanturia et al., 2011). Considering the genetic and phenotypic overlap between OCD and AN described above, a better understanding of the neural circuits and transmitter systems that underpin anorectic behaviors in the context of OCD will shed light on whether this overreliance of habitual control of behavior drives the development and maintenance of compulsive exercise in individuals with AN.

Neuronal control of compulsive behavior: evidence from human and animal studies

Fronto-striatal circuit dysfunction links OCD with AN

Disturbances in the neurobiological drivers that balance goal-directed and habitual control have been well-established in OCD, whereby dysfunction in fronto-striatal circuitry and the persistent reliance on the habitual system leads to compulsive behavior, with the most notable disruptions originating in the dorsolateral prefrontal cortex (dlPFC), a region associated with cognitive control, and the orbitofrontal cortex (OFC), which plays a role in inhibitory cognitive processing (Menzies et al., 2008; Remijnse et al., 2006). Likewise, there is emerging evidence for the roles of dlPFC circuits in specific symptoms of AN (Kaye et al., 2013) (Eddy et al., 2023) suggesting a common impairment in control and the regulation of valence in executive processing across both disorders (Nejati et al., 2021). Moreover, reduced OFC activation during reward anticipation is observed in patients who are more habit-driven (Steding et al., 2019), suggesting a link between OFC dysfunction and accelerated habitual behaviors in AN. Considering both these cortical regions project into the striatum, a region implicated in value-based decision-making, changes in the activity of these fronto-striatal projections might lead to failure in updating the current value of the outcomes of actions, resulting in maladaptive decision-making and compulsive behavior.

Striatum and the basal ganglia

While altered functions in these cortical regions that project to striatum are evidenced both in OCD (Calza et al., 2019) and in AN (Kaye et al., 2013), what remains unclear is whether and how they might drive compulsivity. The striatum is a major input structure of the basal ganglia, where it integrates signals from both midbrain and cortex to regulate learning and motor- and action-planning (Yin & Knowlton, 2006). Like for those with OCD, where hyperactivity in dorsomedial striatum (DMS) is known to drive OCD phenotypes and contribute to excessive habit formation (Gillan et al., 2015; Saxena et al., 2018), increased dorsal striatal activation has also been reported in AN. Specifically, patients with AN engage the dorsal striatum more when choosing what to eat and demonstrate a stronger association between anterior caudate (i.e., the DMS) activity and decisions about food (Foerde et al., 2015; Foerde, Walsh, et al., 2021). This suggests that maladaptive food choices in AN might be a result of an impairment in the goal-directed action system. In addition, increased activation in the caudate is reported in both a monetary rewarded guessing task (Wagner et al., 2007), and in response to aversive stimuli (Cowdrey et al., 2011). Seemingly, this exaggerated caudate function is observed even in patients who have weight-restored from AN, suggesting it may constitute an attempt at more goal-directed (i.e., dorsal striatal-driven) than hedonic (i.e., ventral striatal-driven) means of responding to rewarding stimuli (Wagner et al., 2007) (see Figure 2). Therefore, understanding how basal ganglia dysfunction arises in AN may provide insight into the underpinnings of the loss of homeostatic control of energy balance between exercise and food intake.

Figure 2. Temporal and regional changes in striatal activation throughout the development of compulsive exercise.

The progression of hyperactivity in anorexia nervosa (AN) and activity-based anorexia (ABA) (A) is hallmarked by a decrease in body weight and fat mass (B) as food restriction and exercise becomes increasingly excessive or “compulsive.” These behaviors that seemingly start as goal-oriented, eventually transition to habitual, and then subsequently become impaired (compulsive). This change in the rewarding properties of these behaviors such as food intake may be due to a slow decline in ventral striatal dysfunction (C – as food reward is altered during illness progression) while behaviors that are required to be more goal-driven like exercise may be impeded by alterations in dorsal striatal function. For example, the typical transition from dorsomedial to dorsolateral striatal activation that underlies the consolidation of a skill (D) may develop abnormally, with the dorsomedial striatum becoming over-engaged (E – additional peak when goal directed control is impaired) to exert excessive cognitive control and/or elicit compulsive behavior.

Medial and lateral sub compartments of the dorsal striatum

The transition from goal-directed to habitual behavior has been studied in detail in animal models, mostly in the context of changes that occur within the dorsal striatum, which receives dopaminergic (i.e., DA) input from the substantia nigra (pars compacta) and is controlled by differential cortical inputs to the medial (i.e., DMS) and lateral (i.e., DLS) compartments. Whereas the DLS primarily receives input from somatosensory and motor regions of the cortex, the DMS receives more cortical input from ‘associative’ cortices, i.e., the medial prefrontal regions (prelimbic and infralimbic) and the OFC (Hintiryan et al., 2016) making it the clear target for disturbed functioning in OCD and AN. Moreover, while activity in the DMS supports outcome-based learning that is goal-directed in its initial stages, the DLS is preferentially engaged in the automation of actions when experience has accumulated, also known as habit formation. Indeed, these regions are so functionally distinct in terms of learning that silencing DLS prevents the transition to habitual responding, while silencing DMS results in early emergence of habitual behavior (Lipton et al., 2019). A reorganization from primarily DMS-driven to DLS-driven activation occurs during normal motor learning (Badreddine et al., 2022), and is an important part of establishing behavioral patterns. More recently, the DMS has been implicated in compulsive behavior since dopamine signaling in the DMS but not DLS predicts punishment-resistance in mice (Seiler et al., 2022). This transition between DMS and DLS activation has been proposed to be a dynamic process, in which the activity of the two subregions develop simultaneously and change dynamically to control behaviors (Badreddine et al., 2022; Vandaele et al., 2019). Overall, the balance between DMS and DLS activation is essential in controlling learning and decision-making, however, mechanisms controlling this balance remain unknown.

Moreover, the dorsal striatum is composed of medium spiny neurons, which account for about 95% of striatal neurons, and are GABAergic in nature, as well as cholinergic interneurons (Matamales et al., 2009). These GABAergic medium spiny neurons are divided into two types based on their projection target. The direct pathway expresses D1 receptors while the indirect pathway expresses D2 receptors; each of these projections have opposing roles in the thalamus, which regulates cortical activity in turn to drive behavior. There exists an even further functional division between these pathways based on the dorsal striatal sub compartment in which they are located (Chuhma et al., 2011). For example, in the DMS, the direct pathway is known to be integral to the initial formation of action-outcome associations and for invigorating learned behavior, while the indirect pathway is important for supporting the updating of learning and behavioral flexibility (Peak et al., 2020). In contrast, the direct pathway in the DLS regulates the consolidation of learned actions, while the indirect pathway instead promotes the expression of previously learned habitual actions (Smith et al., 2021). As such, the dorsal striatum is ideally positioned to both integrate input from the cortex and modulate output to the basal ganglia nuclei in order to facilitate motor control, learning, behavioral selection, cognition, and emotion (Yin & Knowlton, 2006). Intriguingly, enrichment of overlapping genes related to both AN and OCD is greatest in the medium spiny neurons of the basal ganglia (Yilmaz et al., 2020), providing a cellular target for mechanistic investigation into the direct role of these neurons in compulsive exercise. Whether or not DA signalling via D1 and/or D2 receptors on medium spiny neurons drives differential activation of DMS and DLS pathways to generate compulsive exercise in AN is unknown.

Dopamine signalling in AN: a two-stage process

There is robust evidence supporting a role for DA dysfunction in AN with decreased DA metabolism (Kaye et al., 1999) and elevated D2/D3 receptor binding suggesting decreased DA transmission or upregulated DA receptor abundance (Frank et al., 2005) both of which persist in individuals weight recovered from AN. Whether DA dysregulation precedes illness onset or how it contributes to the establishment of AN behaviors, including exercise, remains entirely unknown. However, it is well-established that the DA system is engaged in the development and maintenance of compulsive behavior, and a two-stage model has been proposed in which food restriction and exercise may be initiated through positive reinforcement and increased DA release in individuals with AN, thereby promoting further food restriction and exercise (Beeler & Burghardt, 2021). However, when these same (goal-directed) behaviors are maintained over time, they persist not because they elicit rewarding outcomes but because they alleviate negative affect associated with their absence (negative reinforcement) and hypofunction of the DA system (Beeler & Burghardt, 2021; Gorrell et al., 2020). If it is true that DA dysfunction occurs in AN as initial hypersensitivity to the rewarding effects of DA and subsequent blunting of these effects, this could explain why strategies to increase or decrease DA function have been unsuccessful in treating AN (Beeler & Burghardt, 2021). Perhaps this DA hypothesis requires further refinement, whereby the DA dysfunction is driven by temporally- and regionally-specific DA changes, that might be reflected in changes in activation patterns in dorsal versus ventral striatum throughout illness progression (Figure 2). It is essential to differentiate the two stages in AN (initiation or establishment, and maintenance or entrenchment) in terms of their neurochemical basis if a targeted pharmacological treatment option is to be uncovered. This is especially true considering the differential effects of DA transmission on D1 and D2 receptors in the DMS and DLS on reward and behavioral control (i.e., the learning and expression of goal-directed or habitual actions) (Gerfen, 2022).

Dopamine and exercise in humans

Our current understanding of the direction of DA dysregulation over the course of AN is inadequate, and the correlation between such changes and the different symptoms of AN (including exercise) is extremely limited. In the few studies that have specifically examined DA and exercise in individuals with AN (Gorrell et al., 2020), preliminary evidence suggests increased neural response to exercise stimuli in the pre-frontal cortex and cerebellum (Tricomi et al., 2009), greater visual attention to exercise stimuli (Giel et al., 2013), and greater willingness to work for exercise reward among those acutely ill with AN, compared to controls. Among those recovered from AN, mixed findings suggest that response to in-laboratory exercise cues are potentially DA – mediated (O’Hara, Keyes, Renwick, Giel, et al., 2016), or DA- independent (O’Hara, Keyes, Renwick, Leyton, et al., 2016), underscoring a need for further research. Since for many, exercise likely begins as a goal-directed behavior and then transitions to become compulsive even before the diagnosis of AN, it is challenging to track this process in patients in order to determine when this transition to compulsivity occurs. Therefore, to understand why compulsive exercise develops in AN, and the role that DA plays in the shift from positive to negative reinforcement as exercise behavior becomes entrenched, it is necessary to elucidate the neurobiology of this transition with the use of preclinical rodent models with translational relevance and using modern neuroscience tools.

The utility of the activity-based anorexia (ABA) model for understanding the biological underpinnings of compulsive exercise

ABA recapitulates AN behavioral phenotypes

Activity-based anorexia (ABA) is a biobehavioral rodent model that has been used since the 1960’s to elucidate biological factors that might be causal to the development of anorexia nervosa (Gutierrez, 2013). When rats or mice are given unlimited access to a running wheel paired with time-limited access to food, animals exhibit paradoxical hyperactivity despite energy deficit, which precipitates rapid weight loss that will lead to death if left unchecked (Adan et al., 2011). There are multiple similarities between ABA in rats and human AN patients, including a predominance of the phenotype in young females, altered hormonal and neuropeptide function and compulsive exercise (Miletta et al., 2020). It is important to note that only the combination of time-limited access to food and free access to a running wheel causes ABA, leading to debilitating weight loss. If exposed to either feature alone, animals quickly learn compensatory behaviors that allow them to maintain weight, and when exposed to both features a proportion of animals remain resistant to weight loss allowing for more profound investigations into the drivers of inter-individual variability (Milton et al., 2018; Milton et al., 2022). Understanding precisely how ABA rats develop compulsive exercise behavior in the context of limited access to food will allow detailed insight into how this behavior arises in humans.

Although the psychosocial features of AN that motivate food restriction and exercise in human patients cannot be captured in this model, rodents develop many other hallmark features of AN, including cognitive deficits and altered neurotransmitter signalling (Scharner & Stengel, 2020). ABA-exposed but weight recovered animals demonstrate impaired cognitive flexibility, suggesting that experience with time-limited food restriction with unlimited running wheel access can have profound effects on higher-order cognitive functioning and overall decision-making (Huang et al., 2023) that is linked to activity in corticostriatal circuits (Milton et al., 2021). Perhaps the generation of compulsivity that might occur throughout the development of ABA may have a direct impact on these higher-order cognitive functions, with increased compulsivity being linked with poorer behavioral inhibition, insensitivity to positive-feedback, impulsivity and risky decision-making (Chudasama et al., 2003; Dalley et al., 2011). However, despite the wide use of the ABA model in the investigation of underlying mechanisms of AN over recent decades, the underlying neurobiological causes of excessive wheel running evoked by food-restriction in ABA remain inadequately understood. It is possible that wheel running in ABA begins as rewarding and goal-directed, but then shifts to habitual over time (Greenwood & Fleshner, 2019), which would be mirrored by a shift in DLS activation over DMS activation. It is also plausible that excessive exercise in ABA is not habitual per se but represents a similar impairment in goal-directed control as seen in the presentation of OCD, where the “goal” is relief of anxiety or negative affect but both the actions (compulsions) and DMS activation (too much) are inappropriate. Since both goal-directed action and motor activity are modulated by dopaminergic transmission (Conn, Alexander, et al., 2023; Conn, Zou, et al., 2023) using the ABA model could provide mechanistic insight into the relationship between these processes and their relevance to symptoms in AN.

Compulsivity in ABA: The interacting roles of dopamine and acetylcholine

There is accumulating evidence that altered dopaminergic signalling is involved in the development of ABA, although there remains little clarity regarding the direction of effects or mechanisms of action, considering the majority of studies have used blunt tools such as systemic administration of antagonist compounds to investigate dopaminergic contributions to the ABA phenotype. For example, selective blockade of D2/3 receptors is shown to reduce weight loss in ABA (Frank et al., 2005; Klenotich et al., 2015). Consistent with this, mice that increase running wheel activity in ABA show an increase in D2 receptor expression in the dorsal striatum, whereas overexpression of D2 receptor in the ventral striatum induced rapid weight loss and hyperactivity in ABA mice (Gelegen et al., 2008; Welch et al., 2021). In addition, mice with hyperdopaminergia induced by a DA transporter knockout model demonstrate greater susceptibility to ABA caused by an accelerated increase in running wheel activity (Beeler et al., 2020), which implies the role of upregulated DA signalling in hyperactivity in ABA. However, increased DA release in the ventral striatum was found in ABA rats during food access but not during running in the anticipation of food (Verhagen et al., 2009). Moreover, chemogenetic activation of the mesolimbic pathway (i.e., midbrain neurons that are primarily dopaminergic and project to the ventral striatum) prevents weight loss in ABA rats by increasing food intake and food-anticipatory activity without affecting general running activity (Foldi et al., 2017). Taken together, these studies support the notion that dysregulated DA signalling in striatum is involved in the development of hyperactivity in ABA, although there is still much to learn about the direction of effects and the contributing roles of glutamatergic cortical inputs to the striatum, and the microcircuitry that exists within it including interneuron activity and interactions between DA and other signalling molecules (Adan et al., 2011; Cai & Ford, 2018; Favier et al., 2020; Threlfell et al., 2012).

Based on the interacting roles of DA and acetylcholine in striatal plasticity and learning (Chantranupong et al., 2023; Reynolds et al., 2022), it is likely that a transition from DMS to DLS activation during motor learning depends on these transmitter systems. For instance, decreased cholinergic release induced by depletion of acetylcholine vesicular transporters on cholinergic interneurons in the DMS, specifically reduces DA release, which can lead to impaired behavioral flexibility, insensitivity to reward value and increased susceptibility to ABA in mice (Favier et al., 2020). The relationship between DA and acetylcholine is complex, whereby acetylcholine cooperates and competes with DA to modulate striatal activity. Notably, DA input also differentially regulates activation of cholinergic interneurons in the DMS vs. the DLS. While it exerts an inhibitory effect on cholinergic interneurons in the DMS via D2 receptors to inhibit subsequent DA release, it also activates them in the DLS to promote further DA release (Cai & Ford, 2018). This is crucial in the context of ABA because acetylcholine efflux in the DMS is shown to have a role in behavioral flexibility (Ragozzino et al., 2009). While cholinergic outputs in the DMS are selectively increased in rats during reversal learning, blocking these cholinergic receptors leads to decreased DMS cholinergic output and impaired reversal learning (Ragozzino et al., 2009). Similarly, ablating DMS cholinergic interneurons leads to impaired sensitivity to alterations in action-outcome associations, indicating its role in goal-directed learning, particularly in updating new action-outcome associations (Bradfield et al., 2013). Considering that the balance between dopaminergic and cholinergic signalling in the dorsal striatum is essential for the balance between goal-directed and habitual control, it could be crucial for the generation of compulsive wheel running in ABA.

Conclusions and recommendations for improving translational research

Treating individuals who do not respond to standard treatment requires precise understanding of the mechanisms of action that underpin their symptom phenotypes. To this end, although the similarities between the expression of compulsive behaviors in OCD and AN have been demonstrated in numerous studies (Dittmer et al., 2018; Noetel et al., 2017), and the genetic overlap between these conditions is considerable (Yilmaz et al., 2020), this association has not yet been exploited to translate existing treatments for OCD, such as transcranial magnetic stimulation, to target exercise behavior in AN (Steuber & McGuire, 2023). The rapid evolution of tools to precisely manipulate and measure neural circuit function and neurochemical release in rodents, together with a more nuanced view of using the ABA model to study specific aspects of AN pathology (i.e., the development of excessive exercise induced by weight loss), will enable this transdiagnostic link to be systematically tested. We hypothesise that distinct cortico-striatal activation patterns differentiate compulsive running from goal-directed running in ABA, which align with orbitofrontal-striatal circuit dysfunction involved in persistent habitual responding seen in models of OCD. We also anticipate that opposing actions of DA and acetylcholine will balance goal-directed and. compulsive running and underpin excessive habit selection. This understanding could not only aid in the development of treatments that target changes occurring at different stages of AN progression, but could also result in a more successful approach to translation: both across diagnostic categories that involve a variable level of compulsivity, and between preclinical and clinical research. Such an approach would involve rapid testing of hypotheses in animal models, applying this knowledge within a human clinical context, and continuously back-translating respective findings in an iterative manner (see Figure 3). This approach has great potential to determine whether the expression of compulsivity in AN should be considered clinically distinct from other compulsive behaviors and to tailor therapeutic approaches accordingly in line with the burgeoning movement toward personalized medicine (Anderson et al., 2023; Scala et al., 2023). Moreover, identifying the involvement of multiple neurochemical systems and their interactions in the development and maintenance of compulsive exercise will pave the way for generating novel pharmacotherapeutics with multiple targets (i.e., both dopaminergic and cholinergic actions). The delivery of such compounds directly to specific regions of the brain may even be achievable in the near future, using techniques like directed ultrasound, which could bypass off-target effects that often reduce medical compliance. Closing the significant gap between basic and clinical research will contribute to an improved understanding of the development and maintenance of compulsive exercise in AN and enable more rapid advances in treatments that specifically target this symptom.

Figure 3. Bridging the translational gap between preclinical and clinical research in anorexia nervosa (AN).

Using the framework of neural circuitry involved in the generation of compulsive behavior in obsessive-compulsive disorder (i.e. the neural circuits that make up the basal ganglia, including subregions of the cortex, striatum and ventral midbrain) and applying this to wheel running in activity-based anorexia (ABA) rats (A) will inform the targeted treatment of compulsive exercise in anorexia nervosa, that has shown differential activation in the human equivalents of these brain regions (D). Advanced genetic tools in awake, freely-moving animals allow the measurement (i.e. using in vivo fiber photometry; see (B) and manipulation (i.e. chemogenetics and optogenetics) of specific circuitries in order to assess their casual role in executive functions relevant to compulsivity (i.e., behavioral inhibition, impulsivity, reward processing and cognitive flexibility), preferably using touchscreen-based cognitive tasks in rodents (C) that are equivalent to human cognitive tasks (E). The translation of these findings could aid in the development of precision medicines including multisite pharmacotherapy, or the refinement of neuromodulatory treatments such as transcranial magnetic stimulation, both of which can be tailored to individual patients based on clinical subtypes (F). Brain regions of interest (see A for the analogous regions in the rat brain relevant to those regions that have been shown to have altered function in human anorexia nervosa in D): mPFC; medial prefrontal cortex, OFC; orbitofrontal cortex, dlPFC; dorsolateral prefrontal cortex, DMS; dorsomedial striatum, DLS; dorsolateral striatum, NAc; nucleus accumbens, VS; ventral striatum, SN; substantia nigra, VTA; ventral tegmental area.

Public significance statement:

Individuals with anorexia nervosa (AN) who exercise compulsively are at risk of worse health outcomes and have poorer responses to standard treatments. However, when, why, and how compulsive exercise develops in AN remains inadequately understood. Identifying whether the neural circuitry underlying compulsive behavior in OCD also controls hyperactivity in the activity-based anorexia model will aid in the development of novel eating disorder treatment strategies for this high-risk population.