Keywords

1 Introduction

The etiology of anorexia nervosa (AN) is poorly understood. As with many other psychiatric disorders, a role for monoamine systems in the etiology of AN has been proposed (see elsewhere is this book), but how it is involved is currently unknown.

In the last decade, genetics has played an important role in unraveling the pathways underlying disorders. In contrast to most other methods used to unravel the neurobiological pathways underlying disorders, genetics (linkage and whole genome association studies) provides an unbiased approach with no a priori hypothesis. These studies have led to the discovery of a gene that is causally related to the disorder; however, these studies do not resolve how that gene is causing the disorder. In order to understand the relationship between the gene and the disorder, two major (complementary) approaches can be taken. One is to compare phenotypes of human subjects carrying certain (disorder-associated) alleles carefully with those of subjects not carrying such alleles. Heterogeneity not related to the gene of interest and environmental differences complicate this kind of analysis. Therefore, the other approach is to investigate the role of the gene of interest in animals, preferably in a validated disease model. Here, we first discuss some of these models before describing how these were used in our studies toward unraveling the neurobiology underlying AN. As soon as new genes are discovered that are causally related to AN, these models will be helpful to unravel how these genes are involved in increasing the susceptibility for AN.

2 Animal Models for Anorexia Nervosa

The anorexia mouse (anx/anx) is a spontaneous mouse mutant characterized by poor appetite and hyperactivity. Anx/anx mice die at the age of 3–5 weeks depending on the genetic background. Several studies have revealed alterations in the dopaminergic, serotonergic, and noradrenergic systems. In the hypothalamus, neuropeptide Y (NPY) and agouti-related protein (AgRP) show aberrant expression patterns, and pro-opiomelanocortin (POMC) expression is decreased. There is evidence for neurodegenerative changes in the mediobasal hypothalamus (Johansen et al. 2007). The gene causing the anx/anx phenotype has not been identified yet, and therefore its relevance for AN is unresolved.

Very similar to anx/anx mice are contactin-deficient mice. These mice show weight loss starting at postnatal day 10 and they die before 3 weeks of postnatal life in a malnourished state. Contactin is a neural cell adhesion/recognition molecule involved in formation of specific axon projections in the developing nervous system (Fetissov et al. 2005), but whether contactin is relevant in the etiology of AN is unknown. There have been several other mouse mutants generated displaying an anorectic phenotype. Dopamine-deficient mice, generated by deleting the tyrosine hydroxylase gene (but reexpressing it in noradrenaline neurons), have an emaciated appearance and are hypoactive. Administration of l-dopa or restoration of dopamine signaling in the dorsal striatum of these mice restores feeding (Szczypka et al. 2001). These data may underscore the importance of dopamine for motivated behaviors in general and do not prove that dopamine is specific for feeding behavior or that a deficiency in dopamine is causal to developing an eating disorder. Deficiency of either (prepro) melanin-concentrating hormone (MCH) or the MCH receptor 1 also results in a lean and hyperactive phenotype in mice (Shimada et al. 1998; Marsh et al. 2002; Zhou et al. 2005). Except for dopamine-deficient mice [in the dopamine system, genetic variation has been associated with AN (Bergen 2005)], none of the pathways in these genetic models with anorectic phenotypes have been thoroughly investigated for their involvement in AN.

Besides in these genetically altered mice, anorectic behaviors can be induced by exposing wild-type animals to restricted feeding schedules. The most widely used rodent model mimicking features of human anorexia is the activity-based anorexia (ABA) model. Here, we describe results obtained from this model.

2.1 The Activity-Based Anorexia Model

In the ABA model (also referred to as semistarvation-induced hyperactivity, activity stress, or activity anorexia) (Epling and Pierce 1983), animals have voluntary access to a running wheel in their home cage. When first exposed to running wheels, running wheel activity (RWA) increases in the first 2 weeks after which it stabilizes. Animals will increase their ad libitum food intake to compensate for their increased energy expenditure. After RWA has been stabilized, food access is restricted and given only 1–2 h per day (at fixed times) for 1 week. Upon exposure to the scheduled feeding, nocturnal animals like rats and mice will increase their RWA in the dark phase and will also start to run in the light phase. The combination of reduced food intake and increased RWA will lead to substantial body weight loss (>20%) within a week (Fig. 1).

Fig. 1
figure 1_77

The activity-based anorexia model rodents are housed in cages with running wheels. Once these animals get food once a day at a fixed time point for 1–2 h, they will increase their locomotor activity during the dark phase, but they also start to run in the light phase in the hours preceding food delivery. This latter activity is referred to as food anticipatory activity

Full size image

The development of excessive hyperactivity in the ABA model resembles hyperactivity as seen in patients with AN, which is difficult to control and has a negative effect on outcome (American Psychiatric Association 1994). Hyperactivity occurs in up to 80% of patients with AN (Hebebrand et al. 2003). In these patients, it has been regarded as a conscious attempt to lose body weight but may also be explained by a subconscious biological drive (Davis 1997). Many species display increased locomotor activity and travel large distances in periods of food scarcity to find new food sources. If at least in some animals of a group this foraging behavior is rewarding in itself, then the chance that it will occur increases. As a result, the chance of survival of that species increases as well. This may be a biological explanation of why hyperactivity may be rewarding by at least some individuals with AN.

Increased RWA upon exposure to the ABA model is displayed throughout the dark phase, and in the light phase in particular preceding food access. This latter activity is also referred to as food anticipatory activity (FAA) (Mistlberger 1994). Although in most ABA studies food is given at dark onset, the time of food access does not influence the presence of FAA. Thus, FAA will precede food access at any fixed time point of the day. The neural substrate that underlies FAA may be the food-entrainable oscillator (FEO), which has been investigated using a similar experimental setup but with milder food restriction schedules than ABA. The normal circadian rhythm, which is necessary to adapt physiological processes to time of day, can be entrained by food. Whereas noctural species are usually active in the dark phase, upon exposure to restricted feeding paradigms, locomotor activity can be shifted toward (entrained) the moment of food delivery. There are numerous reviews on the location of the FEO in the brain, many of which put forward the dorsomedial hypothalamus (DMH) as an important node, whereas others believe that the FEO is distributed over a neural network (Mistlberger et al. 2009; Gooley et al. 2006). The FEO probably contributes to FAA in the ABA model and acts to prepare the animals for the upcoming meal. In the ABA model, RWA is also increased in the dark phase (without upcoming food access). This latter increase in RWA may reflect foraging behavior, which is a goal-driven behavior that generally occurs upon food restriction. This also accounts when food is given at unpredictable times or when animals get a fixed, limited amount of food per day. Thus, the neurobiological mechanism that drives hyperactivity in the ABA model consists of at least two components: one food entrained (getting ready for eating involving the FEO) and one driving foraging behavior (searching for food).

In our attempt to unravel the brain areas involved in FAA in the ABA model, we compared wheel running in ABA rats (with 1 h food access at dark onset) with wheel running in rats that had 1 h access to food at a random time of day. The latter group of rats was not able to anticipate food access. At the end of the experiment, the random-fed rats had lower body weights because their total RWA was higher and since they could not anticipate a meal, they were not prepared to eat a lot in a short time period and therefore ate less. However, ABA rats had higher c-Fos expression levels in several hypothalamic nuclei [DMH, arcuate nucleus (Arc), lateral hypothalamus (LH)] as compared to the random-fed rats. Surprisingly, the neuronal activity in the DMH of ABA rats (measured by c-Fos expression levels) was correlated with FAA, but this was not observed in rats on the random feeding schedule. The hypothalamus thus seems to play a role in organizing the anticipatory response to upcoming food access. The specific hypothalamic neuronal populations and coupled pathways involved in food anticipation still need to be identified.

The hypothalamus, in particular the Arc, strongly responds to peripheral cues informing the brain on the status of energy balance. The Arc is in close proximity to the peripheral blood stream and senses glucose and other metabolic fuels such as free fatty acids, but also hormones that are released from the gastrointestinal system, the pancreas and adipose tissue. A recent meta-analysis on gut hormones in AN revealed increased basal plasma levels of ghrelin, PYY and CCK (Prince et al. 2009). All three hormones act in the mediobasal hypothalamus, which is also a major target site for the adipose tissue-derived hormone leptin. Plasma leptin levels are often extremely low, whereas plasma ghrelin levels are often increased in patients with AN, reflecting their low adipose tissue mass and hungry state. Since both plasma leptin and ghrelin have previously been implicated in regulating locomotor activity (Holtkamp et al. 2006; Jerlhag et al. 2007), these two hormones and their putative roles in the two components of hyperlocomotion during starvation are discussed below in more detail.

3 Leptin and Ghrelin Physiology

Leptin conveys information on adiposity, or rather starvation to the brain, whereas ghrelin is thought to be the peripheral hunger signal to the brain. Leptin injections decrease food intake by reducing meal size, decrease body weight and adipose tissue mass, and increase energy expenditure in rodents (Campfield et al. 1995; Pelleymounter et al. 1995; Grill et al. 2002). The absence of leptin or leptin receptor (lepRb) leads to hyperphagia, obesity, hypoactivity, and neuroendocrine and metabolic malfunction (Campfield et al. 1995; Pelleymounter et al. 1995; Halaas et al. 1995; Ahima et al. 1996). During starvation, plasma leptin levels fall rapidly (even faster than adipose tissue mass), and adaptive responses, e.g., reduction of energy expenditure, suppression of the gonadal and thyroid axis, and activation of the adrenal axis, are observed. These effects of starvation are blunted by leptin treatment in rodents and humans (Ahima et al. 1996; Rosenbaum et al. 2002).

Ghrelin has emerged as an important gut-brain signal in the control of energy balance (Hosoda et al. 2002; Kojima et al. 1999) and is the only orexigenic gut hormone known so far. In contrast to many endocrine signals, plasma ghrelin levels are elevated prior to meal initiation, and they decrease again during the post-prandial period (Cummings et al. 2001). Caloric restriction increases ghrelin secretion and subsequent activation of the central ghrelin signaling system via the ghrelin receptor (growth hormone secretagogue receptor 1A, GHS-R1A) in the hypothalamus [e.g., Arc, DMH, ventromedial hypothalamus (VMH) and paraventricular nucleus (PVN)] (Zigman et al. 2006; Mondal et al. 2005) increases food intake (Hosoda et al. 2002). Acute central or peripheral ghrelin injections stimulate food intake in rats (Horvath et al. 2001; Naleid et al. 2005) by increasing meal frequency but not meal size and promote fat storage leading to an increased body weight (Faulconbridge et al. 2003). Peripheral ghrelin injection also induces appetite in healthy subjects (Wren et al. 2001) and lead to similar effects on body weight and adiposity in humans (Druce et al. 2006; Wren et al. 2001). Collectively, these data are indicative of a physiological role for ghrelin in hunger and meal initiation.

Not only does plasma ghrelin levels fluctuate in relation to meal initiation (when levels are high) and meal termination (when levels are low), recent data also support the involvement of ghrelin in anticipation to meals. For instance, rhythmic ghrelin release from stomach oxyntic cells is synchronized to the time of food delivery, and GHS-R receptor knock-out mice show less FAA compared to wild-type mice when food access is restricted to 4–6 h in the light phase (LeSauter et al. 2009; Blum et al. 2009). GHS-R-deficient mice also show less c-Fos expression in the Arc, DMH, PVN, and LH following restricted feeding (Blum et al. 2009). It remains to be resolved via which neural circuit ghrelin affects FAA. Ribeiro et al. (2007) showed that the first ghrelin-responsive brain region to show c-Fos expression upon scheduled feeding is the VMH. Recent data also identified GHS-R on dopaminergic neurons of the ventral tegmental area (VTA) in the mesolimbic midbrain (Zigman et al. 2006; Guan et al. 1997; Howard et al. 1996), which have been implicated in reward-seeking behavior. Several studies have indeed demonstrated the ability of central and peripheral injected ghrelin to influence motivation and reward (Cummings et al. 2001). Moreover, Jerlhag and colleagues have demonstrated that ghrelin injection into the VTA increased dopamine release in the nucleus accumbens (NAc) and increased locomotor activity (Jerlhag et al. 2007). Direct effects of ghrelin on the electrical activity of VTA dopaminergic neurons have also been reported (Abizaid et al. 2006). Taken together, these data indicate that ghrelin may act upon the neural circuits constituting the FEO as well as at the level of the VTA to modulate the activity of dopaminergic neurons. We hypothesize that high plasma ghrelin levels in patients with AN contribute to trigger their hyperactivity.

Circulating leptin and ghrelin enter the brain at the level of the hypothalamus which has long been known as a central regulator of feeding behavior. In the Arc, two populations of neurons exist that seem to play antagonistic roles in control of feeding behavior and energy expenditure: (1) neurons expressing agouti-related protein (AgRP) and neuropeptide Y (NPY) that are both orexigenic neuropeptides and (2) neurons expressing pro-opiomelanocortin (POMC) that encodes the anorexigenic α-melanocyte-stimulating hormone (α-MSH) and cocaine- and amphetamine-regulated transcript (CART). Both types of neurons express lepRb (Munzberg and Myers 2005), whereas GHS-R are expressed only in AgRP/NPY neurons (Willesen et al. 1999). Binding of leptin to lepRb inhibits NPY/AgRP neuronal activity (Schwartz et al. 2000), whereas activation of GHS-R by ghrelin stimulates these neurons (Cowley et al. 2003). Binding of leptin to lepRb on POMC neurons directly activates POMC neurons, whereas ghrelin inhibits POMC neuronal activity indirectly via activation of GABA-ergic AgRP neurons that project to POMC neurons (Riediger et al. 2003). LepRb is also expressed in other hypothalamic nuclei, like the DMH and LH as well as in the cortex, hippocampus, midbrain, and caudal brainstem (Grill et al. 2002; Elmquist et al. 1998).

3.1 Leptin and Ghrelin in Patients with AN

Plasma and cerebrospinal fluid (csf) leptin levels are low in patients with AN and reflect reduced body weight and reduced subcutaneous fat (Hebebrand et al. 1997; Mantzoros et al. 1997; Mayo-Smith et al. 1989). Rosenbaum et al. (1997) described that body weight gain in healthy subjects leads to increased circulating leptin levels and variable leptin secretion, depending on the rate and amount of body weight gain. Plasma leptin levels were high relative to adipose tissue mass following weight gain, and low relative to adipose tissue mass following weight loss in healthy subjects (Rosenbaum et al. 1997). When the patient with AN reaches target body weight, plasma leptin levels may thus be disproportionally high (upon adjustment for BMI and adipose tissue mass) compared to healthy controls. This relative hyperleptinemia may imply a risk for renewed body weight loss since it may reduce caloric intake and increase energy expenditure resulting in poor outcome (Hebebrand et al. 2007).

Subjective measurements of motor restlessness in patients with AN axe highest during admission, when leptin levels are lowest (Exner et al. 2000). Moreover, it has been shown that hyperactivity of patients with AN (as determined by the structured inventory for anorexia and bulimia, SIAB) is negatively correlated with (lg10) leptin levels, with leptin (but not BMI) explaining 37% of the variation in hyperactivity (Holtkamp et al. 2003). Leptin levels of (adolescent) patients with AN at admission have also been found to be negatively correlated with their hyperactivity measurements using accelerometers, and with their subjective self-reports of inner restlessness (by VAS) or motor restlessness (by a 5-point Likert scale). Leptin levels have been found to predict all three hyperactivity measurements (Holtkamp et al. 2006). We found that expert ratings of hyperactivity (using a VAS examining motor restlessness and excessive exercise), which were validated with accelerometer scores, are more legitimate than patients’ self-reports of hyperactivity (Van Elburg et al. 2007a). At admission, leptin levels of adolescent patients with AN axe negatively correlated to expert ratings of hyperactivity, while during treatment this relation develops into a positive one in recovering patients with AN (Van Elburg et al. 2007b).

In healthy adolescent girls, plasma ghrelin levels decrease during late pubertal stages (Whatmore et al. 2003). However, in acute patients with AN plasma, ghrelin levels are significantly elevated. The increase in plasma ghrelin levels normalizes after (partial) body weight regain (Miljic et al. 2006; Misra et al. 2004; Nedvidkova et al. 2003; Otto et al. 2001; Soriano-Guillen et al. 2004; Tolle et al. 2003; Troisi et al. 2005). One could speculate that the relative fast normalization or even further decrease in ghrelin levels contributes to inhibition of further weight gain or even to weight loss in patients with AN, since ghrelin is thought to provide orexigenic drive.

The typical postprandial decrease in plasma ghrelin levels following eating (Cummings et al. 2001; Shiiya et al. 2002) seems to remain intact in patients with AN (Misra et al. 2004; Otto et al. 2005).

Only two studies to date have measured the effect of ghrelin injection on eating and hunger in patients with AN. The group of Hotta et al. (2009) performed a pilot study on five patients with acute AN (for which conventional refeeding therapy led to weight loss) by bi-daily injecting 3 μg/kg ghrelin intravenously (iv). This resulted in increased hunger as measured by visual analog scales (VAS) and a mean increase in eating of 20%. Food intake remained higher than pretreatment once ghrelin infusion stopped; however, body weight varied from −1.5 to +2.4 kg during treatment. Miljic et al. (2006) injected nine patients with acute AN and six partially recovered patients with 1.5 ng/kg (5 pg/kg per min for 300 min) without access to food. The VAS hunger score of patients at baseline was reduced compared to controls. During ghrelin treatment, patients showed an increase in hunger according to VAS ratings, but this increase in hunger was lower than that in healthy controls. In a third study by Broglio et al. (2004), hunger was not measured but mentioned as a side effect after injection of 1 μg/kg ghrelin in six out of nine patients with AN. Unfortunately, in none of these studies, hyperactivity was measured.

4 Leptin and Ghrelin in Activity-Based Anorexia

During ABA, RWA significantly increases while plasma leptin levels drop (Kas et al. 2003; De Rijke et al. 2005; Exner et al. 2000). This observation led us and others to hypothesize that the typical increase in RWA would be prevented by increasing circulating leptin levels in ABA rats. Indeed, leptin treatment in the ABA model suppresses hyperactivity (Hillebrand et al. 2005a; Exner et al. 2000). We showed that of hypothermia during the first 4 days of the ABA experiment. Unfortunately, reduced food intake and increased thermogenesis by treatment with leptin resulted in a rapidly worsening condition. RWA in rats fed ad libitum and locomotor activity levels in food-restricted controls were not affected by leptin treatment (Hillebrand et al. 2005a), showing that effects of leptin are dependent on the status of energy balance, with an effect of leptin on reduction of locomotor activity only during weight loss.

Using a slightly different design, it was shown before that leptin treatment (31 μg leptin/day subcutaneously via osmotic minipumps) also prevents hyperactivity in male ABA rats (body weight 230 g), both during the dark and light phase (Exner et al. 2000). Leptin-treated ABA rats in this experiment got 60% of baseline food intake/day and was not different from vehicle-treated controls (i.e., food access was not limited by time). Interestingly, the authors also showed that leptin treatment could rescue RWA when ABA had already developed.

Acute (instead of chronic) leptin treatment also reduces RWA in the ABA model. This effect is, however, only observed after a few days of exposure to the ABA model, when weight loss has already set in and the ABA rats seem to become more sensitive to exogenous leptin. The drop in endogenous leptin levels during the course of the ABA model may contribute to the higher sensitivity to exogenous leptin.

We recently found that a low dose of leptin injected directly into the VTA reduced RWA of ABA rats, suggesting that the VTA is a likely candidate region where reduced leptin signaling contributes to hyperactivity. Because leptin receptors are expressed on dopaminergic neurons and leptin has been shown to reduce the firing rate of dopamine neurons, our finding supports the hypothesis that reduced leptin signaling at the level of the VTA increases dopaminergic activity.

Other rodent models of impaired leptin signaling also show deficits in locomotor activity. For instance, mice lacking leptin or leptin receptors are hypoactive. This hypoactive behavior can be reversed by leptin treatment (Pelleymounter et al. 1995) or reexpressing leptin receptor in the Arc of leptin receptor-lacking mice (Coppari et al. 2005). Whereas we and others showed that leptin treatment (chronic 4 μg leptin in brain) in ad libitum fed rats does not influence RWA or locomotor activity (Hillebrand et al. 2005a; Surwit et al. 2000), it was also recently shown that leptin treatment increases locomotor activity in rats and that antagonizing leptin in the third ventricle reduces RWA (Choi et al. 2008). These findings seem to be in contrast with our and others findings of attenuation of hyperactivity by leptin treatment in ABA rats, and suggest that during ABA, leptin treatment affects activity levels oppositely or by different downstream targets. One explanation of these seemingly opposing results is that leptin’s effects on locomotor activity depend on the status of energy balance.

In situations of a negative energy balance, like during exposure to the ABA model, plasma ghrelin levels are increased. The high plasma ghrelin levels and hyperactivity caused by food restriction in the ABA model led us to investigate whether changes in plasma ghrelin levels were associated with the development of hyperactivity in the ABA model. Indeed, while plasma leptin levels decline, plasma ghrelin levels increase dramatically over the course of the ABA model. Interestingly, the increased plasma ghrelin levels were positively correlated with FAA but not with total RWA in female ABA rats (Verhagen et al., 2010). It was also shown that FAA can be reduced by suppressing ghrelin signaling. For example, we and others showed that ghrelin receptor knock-out mice show reduced FAA. In addition, mice chronically treated with a specific ghrelin receptor do not show FAA. Furthermore, when ABA rats were given an acute central injection of ghrelin receptor antagonist just prior to the development of increased FAA, FAA was suppressed as well. In the above described models of suppressed ghrelin signaling, food intake remained affected. Thus, there is strong evidence that increased ghrelin signaling contributes to the FAA component of hyperactivity and possibly also to the general increased hyperactivity. Others showed before that ghrelin injections in Siberian hamsters peripherally (Keen-Rhinehart and Bartness 2005) increased foraging and hoarding behavior. Jerlhag et al. (2006, 2007) injected 1 μg ghrelin ilvt and directly into the VTA of mice, and these injections caused hyperactivity, most pronounced immediately after injection (0–5 min). Although increased locomotor activity following central or peripheral ghrelin treatment was not observed by others (Wellman et al. 2008; Tang-Christensen et al. 2004), these data support a role for the VTA in ghrelins effect on locomotor activity. Thus, both low leptin levels and high ghrelin levels might contribute to the hyperactivity in ABA rats and AN, and both may do so via acting on VTA dopaminergic neurons (Fig. 2).

Fig. 2
figure 2_77

Involvement of leptin and ghrelin in running wheel activity (a) in rats exposed to the ABA model, treatment of leptin results in decreased running wheel activity during the dark phase as well as during the light phase (b) in mice lacking the ghrelin receptor (GHS-R1A KO) exposed to the ABA model, running wheel activity in the dark phase is unaffected. Running wheel in the light phase (food anticipatory activity) is reduced

Full size image

4.1 Downstream Effector Mechanisms of Leptin and Ghrelin

As described above, peripheral leptin and ghrelin signal to (amongst other) the Arc where they influence the activity of AgRP/NPY and POMC/CART expressing neurons. In order to gain more insight in the hypothalamic response upon exposure to the ABA model and in the mechanism underlying the attenuation of hyperactivity by leptin treatment or antagonizing the ghrelin system in ABA rats, we analyzed expression levels of these neuropeptides in the Arc. AgRP and NPY mRNA levels are increased and CART mRNA levels are decreased in ABA rats compared to ad libitum fed or food-restricted controls. This suggests increased orexigenic signaling (De Rijke et al. 2005), but these changes fail to stimulate food intake since ABA rats eat less than ad libitum fed and food-restricted sedentary controls (Kas et al. 2003). We found a transient up-regulation of POMC expression during the first few days of ABA, followed by a significant down-regulation of POMC, suggesting increased activity of the melanocortin (MC) system during the first days of ABA (Kas et al. 2003; Hillebrand et al. 2006). Chronic leptin treatment reduces the up-regulation of NPY and AgRP and increases POMC mRNA levels versus vehicle-treated ABA rats and thus leads to a net result of reduced orexigenic signaling (Hillebrand et al. 2005a, b). MC binding sites in the VMH of ABA rats have also been found to be increased compared to controls (Kas et al. 2003) suggesting increased sensitivity to anorexigenic α-MSH during ABA, which can be reduced by binding the endogenous inverse agonist AgRP (Nijenhuis et al. 2001). Interestingly, high ambient temperatures during exposure of rats to the ABA model reduce locomotor activity and body weight loss and also decrease MC4-R expression, suggesting that heat treatment decreases the sensitivity to melanocortins (Gutierrez et al. 2009). Thus, maladaptive changes in the hypothalamic melanocortin system may contribute to the development of anorectic behaviors.

We also investigated the involvement of neural circuits downstream of ghrelin and leptin using a pharmacological approach.

4.1.1 Melanocortins and NPY

Stimulation of melanocortin receptor activity by infusion of α-MSH into the brain enhances ABA by reducing food intake, resulting in increased body weight loss. Rats treated with α-MSH show slightly increased FAA (Hillebrand et al. 2005 b). Administration of the inverse agonist AgRP(83–132) in the brain increased food intake of ABA rats, reduced hypothermia, and reduced locomotor activity, thereby attenuating ABA (Kas et al. 2003; Hillebrand et al. 2006). Administration of the competitive melanocortin receptor antagonist SHU9119 does not influence the development of ABA, suggesting that spontaneous activity of MC receptors contributes to the development of ABA. The fact that melanocortin receptor density increases upon exposure to the ABA model supports this. Another piece of evidence supporting the role of melancortin receptors in anorexia is that we found that the Ala67Thr allele of the AgRP gene occurs more frequently in AN patients than in controls (Vink et al. 2001). Thus, altered activity in AgRP and subsequently melanocortin receptor signaling may underlie anorectic behaviors in AN patients.

The increased expression of NPY upon exposure to the ABA model fails to increase food intake during exposure to food in the ABA model. However, it may contribute to increased locomotor activity as an expression of food-seeking behavior. Centrally injected NPY increases not only food intake but also food-seeking behavior in rodents, and recently it was shown that blocking the Y1 receptor by Y1 antagonist 1229U91 reverses the stimulatory effect of food deprivation on wheel running for food in hamsters (Day et al. 2005; Ammar et al. 2000; Keen-Rhinehart and Bartness 2007).

Sodersten and colleagues showed that daily injections of NPY prior to dark onset facilitated ABA in 2 h fed female rats (Nergardh et al. 2007). It was shown that NPY-treated ABA rats reduced food intake, lost more body weight, and were more hyperactive than control-treated ABA rats. In our lab, we discovered that female ABA rats with continuous infusion of NPY showed no differences in 1-h food intake, RWA during dark or light phase, or body weight loss (Hillebrand et al. 2008). The different results might be explained by differences in the restricted feeding paradigm and/or differences in administration of NPY. Daily injections rather than chronic infusion of NPY prior to food access might better reflect the physiological fluctuations due to scheduled feeding and therefore increase RWA and may reflect food-seeking behavior.

4.1.2 Opioids and Dopamine

If voluntary wheel running is rewarding for ABA rats, this may be reflected in altered activity in the opioid and dopamine systems. Hypothalamic levels of endogenous opioid β-endorphin are increased in ABA rats that lost 25% body weight (Aravich et al. 1993). This peptide, a product of the POMC gene, seems to be mainly involved in incentive motivation to acquire food reinforcers (Hayward et al. 2002). Antagonism of the opioid system with naloxone blocks RWA in rats fed ad libitum (Sisti and Lewis 2001), and μ-opioid receptor-deficient mice (β-endorphin is an agonist for this receptor) display attenuated FAA during restricted feeding (Kas et al. 2004). We attempted to attenuate the development of hyperactivity in ABA rats by chronic treatment with opioid antagonist naltrexone (NTX). Chronic peripheral NTX treatment (0, 0.3, or 1.0 mg/kg/day) did, however, not influence RWA, food intake, or body weight loss in female ABA rats on a 1-h feeding schedule (Hillebrand et al. 2008).

Not only wheel running but also restricted feeding influences reward. For example, food restriction induces sensitization to brain stimulation reward, and this is reversed by leptin. Leptin also attenuates drug-seeking behavior after restricted feeding (Fulton et al. 2000; Shalev et al. 2001). These findings suggest the existence of a pathway between peripheral leptin that signals adiposity and the midbrain dopamine system, involved in motivational and rewarding aspects of drugs of abuse and exploration (Wise 2002). The attenuation of RWA in leptin-treated ABA rats might be explained by a direct effect of leptin on the midbrain dopamine system. As mentioned above, leptin receptors are present on dopaminergic VTA neurons projecting to the NAc (Figlewicz et al. 2003; Hommel et al. 2006), and leptin infusion into the VTA reduces firing of dopaminergic neurons, leading to decreased dopamine levels in the NAc and reduced food intake without affecting locomotor activity (Hommel et al. 2006; Fulton et al. 2006).

Long-term genetic knockdown of leptin receptor in the VTA using an adeno-associated virus containing short-hairpin leptin receptor RNA (AAV-shleptin receptor) increases food intake without affecting body weight. Furthermore, knockdown rats are more active, especially in the dark phase, compared to rats injected with the control virus and are also more sensitive to highly palatable food (Hommel et al. 2006). This suggests that decreased leptin receptor signaling increases locomotor activity followed by increased food intake (or vice versa) and, likewise, that increased leptin signaling in the VTA might underlie the attenuation of hyperactivity by central or peripheral leptin treatment in ABA rats. In fact, we showed that acute leptin injection (1 μg) bilaterally in the VTA of ABA rats suppresses RWA without affecting food intake (Verhagen et al., submitted). These studies also indicate that increased activity of the VTA dopamine system develops upon exposure to the ABA model, and future studies should therefore be aimed at direct interference with leptin signaling in the VTA in ABA rats.

Dopamine turnover in the medial basal hypothalamus is reduced during restricted feeding but seems increased or normalized in ABA rats (at 30% body weight loss) (Pirke et al. 1993). The attenuation of hyperactivity by leptin treatment compares with the effects of pimozide treatment in ABA rats, a drug with strong dopamine receptor 2 (D2) affinity (Lambert and Porter 1992). Patients with acute AN have reduced csf levels of dopamine metabolite homovanillic acid, which persist after recovery (Kaye et al. 1999). In addition, recovered patients with AN have increased D2/D3 receptor binding in the anteroventral striatum (including NAc), which may imply altered reward processing and may affect hyperactivity and eating behavior in patients with AN (Frank et al. 2005).

Blockade of dopamine signaling in the ABA model suppresses activity and increases food intake, leading to reduced body weight loss. Treatment of ABA rats with the nonselective dopaminergic antagonist cis-flupenthixol (0.1 mg/day subcutaneously via osmotic minipumps) resulted in higher plasma leptin levels indicative for a stimulation of the hypothalamic anorexigenic drive which is in contrast to the observed increased food intake (Verhagen et al. 2009a). This implies that antagonism of dopaminergic receptors stimulates food intake downstream of the hypothalamic circuits that respond to peripheral satiety signals.

Treatment with the atypical antipsychotic olanzapine (antagonizing both dopamine and serotonin systems), the nonselective dopamine antagonist cis-flupenthixol, or leptin suppresses overall hyperactive behavior in the ABA model (Verhagen et al. 2009a). Analyzing the patterns of RWA in the ABA model in more detail indicates that of these three drugs, leptin is able to reduce FAA most effectively, suggesting that changes in leptin rather than dopamine trigger FAA. Interestingly, basal dopamine release in the NAcc gradually decreased upon exposure to the ABA model, but the moment food was delivered, there was a clear peak in dopamine release (Verhagen et al. 2009b). This may reflect basic bursting activity of dopamine neurons at this time point (Stice et al. 2010). Although dopamine was not released much at the moment before food delivery (anticipation), it cannot be excluded that upon longer exposure to ABA, the dopaminergic system does play a role in FAA, since rats in the microdialysis experiments were exposed to the ABA model for only 4 days.

Taken together, the above supports the proposition that the neurobiological mechanism underlying hyperactivity (dark phase foraging behavior and FAA) in the ABA model consists of two interacting systems.

5 Discussion

We propose that hyperactivity as displayed upon exposure to the ABA model consists of two components. One component, FAA, may be mediated by activation of the FEO. Likely, ghrelin acts either directly or indirectly at the neural structures forming the FEO. But also lack of leptin signaling may contribute to FAA, since leptin strongly suppresses FAA in the ABA model (Verhagen et al. 2009a). The DMH remains a good candidate for the anticipatory hyperactivity displayed by ABA rats, since FAA correlates with c-Fos reactivity in this nucleus. The other component that drives hyperactivity in the ABA model may involve increased ghrelin and decreased leptin signaling at the level of the VTA. Ghrelin increases VTA dopamine neuronal activity, whereas leptin suppresses it. The increased activity of these neurons may mediate the increased motivation for food to frontal brain regions, such as NAcc and prefrontal cortex. The strong effect of ghrelin antagonism and leptin on RWA in ABA rats is most likely not mediated in the Arc, but in the VTA, where both can bind their respective receptors on dopaminergic neurons, consequently affecting these neurons and altering motivation to run. Future studies should investigate the precise role of these neurons during ABA, by focusing on direct effects of leptin and ghrelin as well as hypothalamic neuropeptides (e.g., orexins, CART) on these neurons in relation to hyperactivity. Molecular determinants of attenuation of hyperactivity by leptin treatment might also be further unraveled using specific mouse strains of interest, now that the mouse model of ABA is developed (Gelegen et al. 2007).

Furthermore, hyperactivity in patients with AN is also related to an anxious phenotype; therefore, neurobiological systems underlying hyperactivity and anxiety might be similar or at least connected (Holtkamp et al. 2003). Thus, studying animal models of other phenotypes, e.g., anxiety, may contribute to unraveling the molecular determinants of hyperactivity.

Leptin and ghrelin not only are related to actual eating behavior but also may have an influence on cognitive processing of food. Administration of leptin and ghrelin has been shown to alter the response of the ventral striatum to food images in humans (Malik et al. 2008; Rosenbaum et al. 2009). Interestingly, this kind of response of humans to food images is affected by genetic variation near the dopamine D2 receptor gene (Stice et al. 2008). This has also been associated with different ventral striatal responses to rewarding stimuli (Kirsch et al. 2006). It thus seems that food-related cognitive processing is strongly affected by activation of the ventral striatum and modulated by dopamine receptor activity.

The role of leptin and ghrelin and their interaction with the dopamine system may not only contribute to hyperactivity in ABA rats and patients with AN but also be a driver of cognitive processes that are disturbed in patients with AN. Food-related cognitive processing is strongly affected by activation of the ventral striatum and modulated by dopamine receptor activity (Malik et al. 2008, Rosenbaum et al. 2008; Stice et al. 2008). Interestingly AN has been associated with genetic variation in dopamine-related genes (Bergen et al. 2005) and recovered patients with AN show increased striatal dopamine D2 binding site availability (Frank et al. 2005). Thus, alterations in the dopamine system may underlie hyperactivity and food-related cognitive processing in AN, and ghrelin and leptin may affect these via their effects on the dopamine system.

Food-related cognitive processing as well as reward processing in general is disturbed in patients with AN (Wagner et al. 2007, 2008; Bailer et al. 2005, 2007; Frank et al. 2005; Kaye et al. 1999). For instance, activation of the ventral striatum by food perception is reduced in patients with AN (Wagner et al. 2008). The ventral striatum is also involved in the startle reflex which is a sort of alarm reaction (Koch et al. 1996). Exposure to food cues elicits an exaggerated startle response in patients with AN, suggesting that food primes an automatic defensive reaction (Friederich et al. 2006). Hyperactivity may be regarded as an expression of this defense reaction. Aberrations in (neuro)psychological processes like set shifting and decision making (Lopez et al. 2009; Roberts et al. 2007; Tchanturia et al. 2004, 2007) may also affect food-related cognitive processes in patients with AN. Set shifting disturbances are consistent with the perseverative obsessions with food and body shape and perfectionism that is typical for AN (Bardone-Cone et al. 2007). Interestingly, set shifting is dependent on dopaminergic activity (Avila et al. 2003). Decision making is based on the expected incentive value of an anticipated outcome and is also dependent on dopaminergic activity (Tchanturia et al. 2007; Naqvi et al. 2006). Deficits in self-regulatory control processes and frontostriatal systems are related in patients with AN (Marsh et al. 2007). The above suggests that alterations in dopaminergic signaling in patients with AN are (causally) related to their impairments in cognitive processes, including those related to food, but possibly also those related to hyperactivity. We hypothesize that impairments in cognitive processing are exaggerated by low leptin and high ghrelin levels in patients with AN.

Taken together, these data support the hypothesis that ghrelin and leptin affect the dopamine system directly via the VTA where both leptin and ghrelin receptors are expressed. Thus, the leptin and ghrelin systems are both involved in homeostatic as well as hedonic neural circuits (2). An important focus for future research is to unravel what causes the apparent leptin and ghrelin insensitivity in patients with AN, who show low and high plasma levels of leptin and ghrelin, respectively, but undereat and increase energy expenditure by physical activity.