Binge eating disorder (BED) is characterized by excessive food intake during a short period of time and is often associated with obesity. Mouse models of binge-like eating behavior are lacking making it difficult to employ genetic models in the identification of mechanisms regulating excessive eating. We report a rapid and simple model to induce binge-like eating behavior in mice that does not require food deprivation or exogenous stressors. Weekly 24 h access to a nutritionally complete high energy diet (HED), along with continuous access to standard chow, resulted in a significant increase in HED intake following its presentation compared to mice that had continuous access to both diets. Mice exhibiting binge-like eating consumed one-third of their normal total daily caloric intake within 2.5 h of HED presentation. Moreover, total 24-h caloric intakes were increased by 50% in mice exhibiting binge-like eating. Following repeated cycles, binge-like eating of the HED was maintained over several weeks with no evidence of habituation or significant alterations in body weight and adiposity. Pharmacological evaluation of binge-like eating behavior was performed using clinically employed compounds. Interestingly, binge-like eating was dose-dependently decreased by fluoxetine, but not baclofen or topiramate. These data support clinical validation of this mouse model of binge-like eating behavior, as fluoxetine has been shown to reduce binge frequency in human subjects with BED. The availability of transgenic and knockout mice will allow for the determination of genes that are involved in the initiation and maintenance of binge-like eating behavior.
The basic feature of binge eating disorder (BED) in humans is repeated episodes of eating large amounts of food in short periods of time without compensatory behaviors such as purging or excessive exercise. The estimated prevalence of BED is roughly 1–4% in the United States. BED is often associated with being overweight. Moreover, there is a high comorbidity of BED with psychiatric disorders and substance abuse (1). Given the prevalence of BED and comorbid conditions, there is a clear need to understand its etiology and to find effective and well tolerated pharmacotherapies. Two primary treatment objectives have been utilized clinically to treat BED. First, the eating disorder is addressed through cognitive behavioral and interpersonal therapy. Second, the obesity is treated with behavioral weight loss management programs or pharmacotherapies. In clinical studies, several classes of pharmacological agents have demonstrated effectiveness in reducing the frequency of binge episodes in patients with BED including fluoxetine, sibutramine, topiramate (2), and baclofen (3).
Animal models for eating disorders typically utilize a history of food restriction, stress, and limited access to palatable diets. Overnight food deprivation will cause hyperphagia, but food deprivation causes increased locomotor activity and significantly elevates corticosterone levels. A history of caloric restriction and footshock stress can produce binge-like eating in rats (4), as can limited daily or intermittent access to a highly palatable food source (5,6,7,8,9,10,11). However, reliable induction of binge-like eating behavior in rodents is dependent on numerous factors including genetic background, postnatal rearing environment and schedule of access conditions to palatable foods (11,12,13,14). Furthermore, some of these models typically do not produce stable binge-like eating patterns until 2–8 weeks after initiation of the protocol requiring a significant time investment from the researcher (5,8,15).
Developing a mouse model of binge-like eating behavior has proven to be more difficult because minor stresses can significantly inhibit food intake in this species (16). More recently, it was reported that successive cycles of food restriction with repeated exposure to forced swim-induced stress produced binge-like eating in mice (17). However, once initiated, this behavior could not be maintained beyond three cycles. Herein we report a rapid and relatively simple model of binge-like eating behavior in mice that does not require food deprivation or the application of exogenous stressors.
Materials and Methods
All procedures were approved by the Institutional Animal Care and Use Committee of Eli Lilly and Company and were in accordance with the American Association for the Accreditation of Laboratory Animal Care-approved guidelines and the Guide for Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, 1996). Ten-week-old male C57BL/6 (Harlan Sprague Dawley, Indianapolis, IN) mice were acclimated for at least 1 week before initiating experiments. Mice were individually housed in standard plastic cages on an automatic watering rack and were on a 12 h light–dark cycle (lights on 06:00 h, lights off 18:00 h).
Induction of binge-like eating behavior in mice
Mice of similar body weights were divided into one of three experimental groups; chow, continuous or intermittent (n = 15–20 per group). Chow controls received unlimited access to a standard rodent chow diet Teklad no. 2014 (3.5% fat, ∼11% sucrose, 3.2 kcal/g, Harlan Laboratories, Indianapolis, IN). Continuous controls had ad libitum access to both the standard chow and a high energy diet (HED) Teklad 95217 (40% fat, ∼16% sucrose, 4.3 kcal/g). As shown in Figure 1, mice in the intermittent group received an initial 48-h free-choice access to both chow and the HED at the start of cycle 1. After 48 h, the HED was removed for 5 days during which time only the standard chow was available ad libitum. The HED was then presented back to the mouse (∼4 h after the start of the light cycle) and food intake was monitored 2.5- and 24-h later. During this 24-h HED access period, mice in the intermittent group always had standard chow Teklad no. 2014 available. After 24 h, the HED was removed thus completing binge cycle no. 1 (Figure 1a). For subsequent binge cycles, mice had 6 days of access to chow only followed by 24 h free-choice access to the HED and chow (Figure 1b). Food was always available ad libitum in all groups. Therefore there was no food restriction initiated by the investigator. In the continuous and intermittent groups, the two diets were separated by a divider on the metal rack top of the home cage. The location of the chow and HED were randomized to prevent any positional effects. One-half of the cages had chow on the left side with the HED on the right side of the metal cage tops.
All mice received six successive binge cycles as described above and then 7–10 mice from each of the three experimental groups (chow, continuous, and intermittent) were block randomized based on body weight into two separate groups for subsequent measurements of body composition and hormonal analysis, or behavioral analysis, 24 h after presentation of the HED during the 6th binge-like eating cycle.
Body composition analysis
Whole body composition was measured in nonanesthetized mice using NMR imaging (Whole Body Composition Analyzer; EchoMRI, Houston, TX). Fat-free mass was determined by subtracting fat mass from total body weight. Measurements were taken 24 h after the intermittent group was given free-choice access to the HED and chow during the 6th binge cycle.
Plasma collection and hormonal analysis
Mice were killed immediately after NMR imaging as described above 24 h after the 6th binge cycle. Whole trunk blood samples were collected and spun for 15 min at 3,000 × g at 4 °C. Serum was stored at −80 °C until radioimmunoassay analysis of leptin and corticosterone.
Behavioral effects of repeated exposure to binge-like eating; EPM and FST procedures
With remaining animals (n = 8–10 per group), elevated plus maze (EPM) and forced swim test (FST) procedures were run as previously described (18,19) 24 h after the intermittent group was given free-choice access to the HED and chow during the 6th binge cycle. Animals were moved to the testing area in their home cages at least 1.5 h before testing. The EPM was run first and then the FST was run 4 h later on the same day. The HED diet was still present at the time of measurement.
Pharmacological evaluation of model
A separate group of 30 C57BL/6 mice were cycled through at least two successive binge cycles as diagramed in Figure 1 before pharmacological testing began. Six mice were given continuous access to the HED, whereas the remaining 24 mice were given weekly intermittent access. Testing with a single compound occurred every other week. On weeks where no compound was given, mice in the intermittent group received the HED as scheduled. Animals were block randomized into drug and vehicle treatment groups (n = 6) based on the 2.5 h HED intake during the previous binge cycle when no drug was given. Fluoxetine, baclofen, and topiramate were obtained from Sigma-Aldrich (St Louis, MO) and were administered intraperitoneally before the HED was presented to the intermittent group. Fluoxetine and baclofen were dissolved in 0.9% saline solution and doses administered were 3, 10, and 30 mg/kg fluoxetine and 0.3, 1, and 3 mg/kg baclofen. Topiramate was dissolved in 20% captisol in sterile water and 10, 30, and 100 mg/kg doses were administered. Pretreatment times were as follows: 30 min fluoxetine and baclofen, or 1 h topiramate. Data were analyzed for each drug treatment with one-way ANOVA and Dunnett's multiple comparisons post hoc test.
Values were plotted as the mean ± s.e.m. for each group. Statistical analyses were conducted with GraphPad Prism 4 (GraphPad Software, San Diego, CA). Comparisons among groups for all behavioral and hormonal analyses were performed with one-way ANOVAs and Bonferroni post hoc tests. ANOVA values and post hoc results are reported in the respective figure legends. Successive binge cycles were analyzed with two-way repeated measures ANOVA, followed by Bonferroni post hoc tests.
Weekly palatable food access for 24-h induces binge-like eating behavior in mice in the absence of caloric restriction
In our paradigm, C57BL/6 mice that had weekly ad libitum free-choice access to both a standard chow and a highly palatable and energy dense, high-fat and high-sucrose diet (HED) for 24-h consumed significantly more calories after 2.5 h than control groups that had either unlimited access to a standard chow diet or continuous ad libitum access to both the standard chow and HED (Figure 2a). Mice in this intermittent group consumed an average of 1.2 g of HED during the initial 2.5 h which was equivalent to one-third of the total daily caloric intake of nonbingeing mice. Standard chow consumption was negligible in the continuous and intermittent groups at 0.06 ± 0.003 g vs. 0.04 ± 0.005 g, respectively. Moreover, we found that all mice in the intermittent group exhibited binge-like eating and preferred the HED over the standard chow (Figure 2b).
We also analyzed 24-h caloric intakes during the period of HED access. Mice in the intermittent group consumed almost double the amount of kilocalories per body weight compared to mice with continuous access to the HED (Figure 2c), and 100% of the calories consumed were from HED intake. Caloric intakes were similar among all groups in the 24 h prior to presentation of the HED suggesting that the binge-like eating in the intermittent group was not due to reduced caloric intake (Figure 2c). Interestingly, almost all the extra caloric intake in the intermittent group came from consumption within the initial hours of exposure to the HED during the light phase when mice are normally inactive. No compensatory reduction in nocturnal feeding was observed. Removal of the HED from the intermittent group after 24-h access caused a significant voluntary reduction in this group's ad libitum consumption of standard chow the following day. Using our protocol, the binge-like feeding pattern could be maintained over at least a 6 week period with no evidence of habituation. Furthermore, there was no evidence of entrainment of food intake behavior as levels of HED intake were consistent in the intermittent group each week (Figure 2c). Analysis of cumulative caloric intakes revealed that mice that had continuous access to the HED consumed significantly more calories on average each week compared to those that had chow only or weekly intermittent access (105.9 ± 3.2 kcal vs. 89.6 ± 5.4 kcal vs. 95.1 ± 3.9 kcal, respectively, F(2,15) = 6.66, P = 0.0009).
Both 129S6 and C57BL/6 strains of mice exhibit robust binge-like eating behavior
A majority of the existing mouse knockout strains have been derived from 129S6 embryonic stem cells. Furthermore, 129S6 mice are relatively resistant to diet-induced obesity compared to the C57BL/6 strain (20). We therefore tested 129S6 mice (Taconic Farms, Cambridge City, IN) in a separate experiment using the method outlined in Figure 1 for three successive weeks and found that this strain exhibited significant binge-like eating behavior. Similar to C57BL/6 mice, 129S6 mice in the intermittent group consumed one-third of their total daily caloric intake within the first few hours after presentation of the HED (Figure 2d). Furthermore, total 24-h caloric intake was significantly elevated in 129S6 mice in the intermittent group (Figure 2e). Analysis of cumulative caloric intakes over three successive binge-like eating cycles revealed that, unlike C57BL/6 mice, all 129S6 mice had similar weekly caloric intakes independent of group (chow 75.9 ± 4.3 kcal vs. continuous 76.3 ± 5.6 kcal vs. intermittent 76.8 ± 4.3 kcal, respectively, F(2,6) = 0.06, P = nonsignificant). Thus, expression of binge-like eating behavior was independent of the propensity for the animal to develop obesity.
Six weeks of binge-like eating behavior did not change body weight, adiposity, or serum leptin levels
We hypothesized that as a result of repeated binge-like eating behavior there might be neurochemical, hormonal or other changes that would alter body weight and body composition in the intermittent group. As expected by the increase in total caloric intake, C57BL/6 mice that had 6 weeks of continuous access to the HED had significantly increased body weight, fat mass and serum leptin levels compared to standard chow controls. However, we found no significant changes in body composition or serum leptin in the intermittent group compared to chow-fed controls after exposure to six successive binge cycles (Table 1).
Table 1. Six successive binge-like eating cycles did not alter body weight or adiposity
Repeated binge-like eating episodes did not alter behavior in EPM or FSTs, and did not change serum corticosterone levels
Given anxiety and depression are common in individuals with BED, we wanted to determine whether a history of binge-like eating in mice would alter responses in either the EPM or FSTs. These tests have been utilized in mice to screen compounds for their anxiolytic and antidepressive activities, respectively. After six successive binge cycles in C57BL/6 mice, we found no changes in the percentage of time spent in the open compartments of the EPM (Figure 3a), nor did it change immobility time in the FST (Figure 3b). Serum corticosterone levels were also not different amongst all groups (Figure 3c).
Fluoxetine reduced 2.5- and 24-h HED intake in mice with a history of binge-like eating behavior
Male C57BL/6 mice that had prior exposure to at least two binge-like eating cycles were used to evaluate the effects of fluoxetine on binge-like eating in mice. As described above, 2.5-h HED intake was significantly increased in vehicle treated mice in the intermittent group (Figure 4a). However, treatment with fluoxetine dose-dependently decreased this intake. Doses of 10 and 30 mg/kg fluoxetine (intraperitoneally) reduced 2.5 h HED intakes by 31 and 69%, respectively, compared to vehicle treated controls (Figure 4a). A dose of 30 mg/kg fluoxetine also significantly reduced 24 h HED intake (Figure 4b), but not below the intake of mice with continuous access to the HED.
Baclofen reduced HED intake 2.5 h, but not 24 h after HED access
Presentation of the HED resulted in a significant increase in its consumption in vehicle treated mice 2.5 h after receiving HED access (Figure 4c). However, a dose of 3 mg/kg reduced 2.5 h HED intake in the intermittent group compared to vehicle treated controls (Figure 4c) with no visible behavioral effects. A dose of 10 mg/kg baclofen also reduced 2.5 h HED intake (data not shown) but it produced significant sedation as previously described (21). HED consumption over the course of a 24 h access period was not significantly reduced with any dose of baclofen tested (Figure 4d).
Topiramate was not effective at reducing binge-like eating behavior in mice
Again, presentation of the HED resulted in a significant increase in consumption in vehicle treated mice in the intermittent group 2.5 h after HED access (Figure 4e). However, we did not see a reduction in HED intake at 2.5 (Figure 4e) or 24 h (Figure 4f) after treatment with 10, 30, or 100 mg/kg topiramate. No gross behavioral changes were noted at any dose tested.
Herein, we report a rapid and relatively simple model to induce binge-like eating behavior in commonly used strains of mice with weekly intermittent access to a nutritionally complete, palatable diet (HED) in the absence of calorie restriction and behavioral or biochemical evidence of significant stress. Moreover, we did not observe changes in adiposity or serum corticosterone levels. Weekly limited access to a HED resulted in a stable pattern of binge-like eating within 1 week and did not require daily experimental manipulations as do many of the currently available models of binge-like eating behavior in rodents.
Although it is difficult to mimic every feature of BED in a single rodent model, especially subjective aspects, we had four criteria we wanted the model to fulfill (22). (i) Rodents should consume more food in a brief, discrete period of time than controls under similar conditions. (ii) The behavior should be induced repeatedly over extended periods of time. (iii) Binge-like food intake should not be due to food deprivation. (iv) If compensatory behavior occurs, it should be initiated by the rodent and not the investigator. We believe we have fulfilled all four criteria with our model and thus model several of the criteria that are necessary for a clinical diagnosis of BED in humans. BED is the only eating disorder with a substantial proportion of affected males, with a ratio of females to males with BED of ∼3:2. Therefore, our use of male mice in our studies is clinically relevant. Furthermore, the HED diet utilized contained 73% more fat and 43% more sucrose per unit weight than many standard rodent chow diets and therefore is similar to the high fat and high sugar composition of foods that are commonly consumed during binge episodes in humans with BED.
We found that two parameters in our study were necessary to obtain consistent caloric intakes throughout each binge-like eating cycle; the length of initial presentation of the HED and the time interval between HED presentations. We used an initial 48-h exposure to the HED in the intermittent group to prevent neophobia. Furthermore, if the initial presentation is too short then stable binge-like eating behavior is not established until after multiple exposures in both rats and mice (6,23). We also found that at least 5 days of standard chow only access was needed for 24 h caloric intake in the intermittent group to return to control levels after removal of the HED at the end of a binge cycle and to ensure that this group was not in a negative energy balance.
More recently it was found that rats receiving intermittent access to a fat and sugar mixture three times a week consumed more of the mixture than those that had 1 h daily access to the same mixture (24) suggesting that the schedule of HED presentation is a key factor in establishing consistent binge-like eating behavior. A major benefit of our model is that with each cycle there were similar amounts of HED intake in the intermittent group during each successive binge-like eating cycle. In other models it has been reported that the use of more frequent and short HED exposures, such as once daily for 1 h, results in a progressive increase in HED intake with each subsequent access (9,23). There can be as high as a threefold increase in HED consumption over the course of a several week experiment in mice (23) and strongly suggests that binge-like eating behavior in models that use frequent and short HED access periods exhibit significant food anticipatory activity. However, it is not likely that there is such entrained intake or a learned motivational component in our model because 2.5- and 24-h HED intakes were relatively consistent with each cycle.
It is important to note that our model is an isomorphic one and mimics a behavior that is associated with BED. Although this model will allow us to study the initiation of binge-like eating behavior in rodents, it will not allow us to investigate all features that are common to BED in humans. One limitation of our model is that we did not see changes in adiposity therefore making it difficult to determine mechanisms by which binge-like eating behavior is linked to obesity. Furthermore, we found no evidence of significant stress with our model. However, human BED is often associated with anxiety and depression. We performed hormonal analysis, and EPM and FST procedures, both 24 h before and 24 h after presentation of the HED in two separate groups of C57BL/6 mice that had 6 weeks exposure to binge-like eating behavior. However, we failed to find changes in serum corticosterone levels or EPM and FST analyses. We might have observed significant increases in anxiolytic measures if we had performed our EPM study within the initial hours following removal of the HED during the final binge cycle as has been previously reported in female rats (25). Thus, we are modeling the aberrant feeding behavior but not the negative mood and emotional components that are necessary for a clinical diagnosis of BED in humans.
We wanted to test several distinct classes of pharmacological agents that have been shown to have clinical efficacy in BED to determine if binge-like eating could be reduced in our model. We tested fluoxetine, baclofen and topiramate in male C57BL/6 mice that had prior exposure to binge-like eating. Disruptions in monoaminergic pathways are thought to be involved in the cognitive and mood disturbances in humans with eating disorders. Indeed, selective serotonin reuptake inhibitors (SSRIs) and selective norepinephrine reuptake inhibitors (SNRIs) are effective in clinical studies with BED patients in reducing binge frequencies (26). We found that treatment with the selective serotonin reuptake inhibitors fluoxetine dose-dependently reduced HED intake in our model (Figure 4). However, total 24-h caloric intake was not reduced below that of the continuous access group given vehicle suggesting that fluoxetine blocked the binge-like eating without affecting spontaneous nocturnal food intake at the doses tested. Although HED intake in the continuous access group was not determined after fluoxetine administration, it is possible that at higher doses or with repeated administration that we might observe a nonspecific reduction in spontaneous food intake across all groups (27).
Although baclofen did not reduce basal food intake in nonbingeing rodents (21,28,29), baclofen reduced binge frequency in patients with BED (3). These data raise the possibility that GABAergic activity is decreased in BED. We found that 3 mg/kg baclofen reduced 2.5 h HED intake in our model. This dose was slightly higher than the 1 mg/kg dose of baclofen that was reported to reduce 2 h consumption of Crisco in a rat binge-like eating model utilizing intermittent access to the shortening three times a week (28). Although modest sedation at the 3 mg/kg dose of baclofen cannot be completely ruled out, it is unlikely because doses of 1–4 mg/kg have been reported to increase spontaneous food intake within 2 h after administration in nonbingeing rodent models (21,30). The short elimination half life (∼3.5 h in humans) most likely explains why 24 h HED intake in our intermittent access group was similar to vehicle controls (31).
Topiramate is effective in reducing binge eating frequencies in controlled clinical studies (32). However, topiramate was not active in our model at doses that have been previously shown to be effective in decreasing stress-induced alcohol intake in mice (33). Furthermore, we tested doses of topiramate that are higher than doses that are effective as an anticonvulsant (34). Topiramate has recently been shown to be effective in reducing consumption of a high fat and sugar mixture in rats exposed to stress (11). It is speculated that topiramate's clinical effectiveness might be dependent on other neuropsychiatric components such as anxiety or stress that are common in BED. However, we did not find evidence of alterations in behavior with repeated exposure to our model in the EPM or FSTs, nor did we find any significant changes in serum corticosterone levels. More recently, topiramate was shown to reduce (food) intake when rats were trained to consume based on the presence of predictive cues (14). Notably, the rats exposed to these cues also had elevated plasma corticosterone levels. Thus, the predictive validity of compounds that are effective in our model may be limited because we do not have a stress component which might be necessary for the effectiveness of topiramate in both clinical and rodent models.
Both the obesity-prone C57BL/6 and the obesity-resistant 129S6 mouse strains had equally robust binge-like eating behavior in our model and are consistent with previous studies showing that the propensity to become obese was not predictive of binge-like eating behavior in rats (7). The availability of transgenic and knockout mouse models will allow us to determine potential mechanisms that mediate both the initiation and maintenance of binge-like eating behavior.
The authors thank Dr Jeff Witkin and Xia Li for help with behavioral analysis and helpful discussions. The authors also thank Dr Dana Sindelar for helpful discussions and Joelle Dill for excellent technical assistance.
The authors are employed by Eli Lilly and Company. This research was funded by Lilly Research Laboratories.