Female Sprague-Dawley rats (Harlan-Winkelmann Co., Borchen, Germany) weighing 150–180 g upon their arrival were housed in groups under conditions of controlled illumination (12:12 h light/dark cycle, lights on/off: 06:00 a.m./06:00 p.m.) and temperature (21–23°C). Rats were fed with standard rat chow (ssniff Spezialdiäten GmbH, Soest, Germany) and tap water ad libitum unless otherwise specified. This study was carried out in accordance with the recommendation of the institutional guidelines; the protocol was approved by the state authority for animal research (#G 0117/14).

After an initial acclimatization period of 7 days, rats (total n = 44) were randomly assigned to one of four groups: (a) ad libitum group: no extra activity + ad libitum feeding schedule, (b) activity group: voluntary activity in a running wheel + ad libitum feeding schedule, (c) restricted feeding group: no extra activity + restricted feeding schedule, and (d) activity-based anorexia group: voluntary activity in a running wheel + restricted feeding schedule.

During the first week of the experiment, all rats were fed ad libitum and separated into single housing cages which were placed adjacent to each other to provide sight, acoustic and odor contact. Rats of the activity and activity-based anorexia group had access to a running wheel inside the cage for 24 h/day, while the sedentary groups (ad libitum and restricted feeding group) were housed without running wheel under otherwise identical conditions. All cages contained environmental enrichment and bedding material. Rats were acclimated to their new cages for 1 week and handled daily to become accustomed to the interaction with the investigator. This included daily removal of the rat from the cage to measure body weight. The daily routine was performed between 08:00 and 09:00 a.m.

Food restriction conditions started on day eight of the experiment. Rats of the restricted feeding as well as activity-based anorexia group received food from 09:00 to 10:30 a.m. (the 90-min feeding period during the light phase was based on Luyten et al., 2009; Wu et al., 2014), while the other two groups (ad libitum and activity group) continued to have access to food for 24 h/day. Body weight, food intake and activity were monitored over a period of 21 days. The experiment was discontinued and animals euthanized when the body weight loss exceeded 25%.

Rats were weighed daily between 08:00 and 09:00 a.m. Body weight and body weight changes were calculated for the whole 21-day experimental period (1 week of ad libitum food intake and 2 weeks of restricted feeding).

The microstructural analysis of feeding behavior was conducted using the BioDAQ episodic food intake monitoring system for rats (BioDAQ, Research Diets, Inc., New Brunswick, NJ, USA) which allows the continuous monitoring of solid chow food intake in undisturbed rats as recently reported (Teuffel et al., 2015). The system contains a food hopper placed on an electronic microbalance, both are mounted on a regular rat single housing cage. Food intake parameters are measured continuously and can be extracted from the software (BioDAQ Monitoring Software 2.3.07); periods of interest can be chosen freely afterwards for data analysis. Every interaction of the rat with the food hopper is registered as a “bout.” A meal is defined as food intake of at least 0.01 g, when feeding bouts occur after an interval of ≥15 min this is considered a new meal. Meal parameters extracted from the software include bout size, meal size, bout frequency, meal frequency, meal duration, time spent in meals and eating rate. The food intake microstructure was analyzed starting at 4 days after food restriction over a period of 4 days (expressed as mean value of 4 days/animal).

Physical activity in the running wheel was assessed electronically using the software provided by the manufacturer (Campden Instruments Ltd., Loughborough, UK) and expressed as wheel rotations per day as described before (Wu et al., 2014). Here, the activity system was combined with the cages for automated food intake monitoring. Pilot studies did not indicate any deleterious interference between the two measurements (data not shown).

The estimation of energy consumption was based on an earlier study that determined oxygen consumption of rats running at a constant speed (Shepherd and Gollnick, 1976). A respiratory exchange ratio of 1.0 was assumed based on carbohydrates as largest component in the standard rat chow used (58% of calories from carbohydrates, manufacturer's information).

At the end of the observation period, brain activation was assessed using c-Fos immunohistochemistry in ad libitum and activity-based anorexia rats (n = 3/group). In order to avoid signals from overfeeding and great distention of the stomach, food intake was restricted to 1.5 g in this last 1.5-h feeding period in the activity-based anorexia group. Directly after this feeding period, animals were perfused and brains processed for Fos immunohistochemistry as described before (Wang et al., 2011). Briefly, rats were deeply anesthetized by an intraperitoneal injection of 100 mg/kg ketamine (Ketanest™, Curamed, Karlsruhe, Germany) and 10 mg/kg xylazine (Rompun™2%, Bayer, Leverkusen, Germany). Transcardial perfusion was performed as described before (Stengel et al., 2009). After thoracotomy a cannula was inserted into the ascending aorta via the left heart ventricle. Perfusion consisted of a 1-min flush with sodium chloride (0.9%) followed by 500 ml of fixative (4% paraformaldehyde and 14% saturated picric acid in 0.1 M phosphate buffer, pH adjusted to 7.4). Afterwards, brains were removed and postfixed overnight in the same fixative at 4°C followed by a cryoprotection in 10% sucrose for 24 h. Lastly, brains were snap-frozen in dry ice-cooled 2-methylbutane (Carl Roth GmbH, Karlsruhe, Germany) and then stored at −80°C until further processing.

Rat brains from the two groups were processed in parallel to ensure similar conditions. Whole brains were cut into coronal sections (25 μm) from prefrontal forebrain to the caudal medulla using a cryostat (CryoStar NX70, Thermo Fisher Scientific, Waltham, MA, USA). Every third brain section was rinsed in phosphate-buffered saline (PBS) for 3 × 15 min. All incubations were performed using the free-floating technique at room temperature (except for the incubation with the primary antibody at 4°C) and followed by a 3 × 15 min washing step in PBS. The sections were first treated with 0.3% H₂0₂ in PBS for 30 min to block endogenous peroxidase activity. After rinsing the sections, nonspecific binding was blocked by 2% normal goat serum (NGS, Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA) for another 30 min. Sections were washed again and incubated in rabbit polyclonal anti-cFos (1:20,000, Catalog No. ABE457, Merck Millipore, Darmstadt, Germany) as primary antibody (2 h at room temperature followed by overnight at 4°C). Sections were rinsed again and incubated with biotinylated secondary goat anti-rabbit IgG (1:1000, Catalog No. 111-065-144, Jackson ImmunoResearch) for 2 h. After rinsing, this was followed by the incubation with the avidin-biotin-peroxidase complex (ABC, 1:500, Vector Laboratories, Burlingame, CA, USA) in 0.3% Triton-PBS for 1 h. Staining was visualized with diaminobenzidine tetrahydrochloride (DAB, Sigma-Aldrich, Darmstadt, Germany) and nickel ammonium sulfate (Fisher Scientific, Waltham, MA, USA). The color development was frequently checked with a light microscope and stopped after about 10 min. After staining, sections were mounted, air-dried, completely dehydrated through a gradient of ethanol, cleared in xylene and cover-slipped with Entellan™ new (Merck Millipore).

In a separate experiment, specificity of the cFos antibody was assessed by pre-absorption with synthetic SGFNADYEASSSRC (amino acids 4–17 of rat c-Fos, JPT Peptide Technologies GmbH, Berlin, Germany). The peptide (5 μg/ml) was incubated with the anti-c-Fos antibody diluted at 1:20,000 (Merck Millipore, antigen:antibody ratio of 100:1) for 2 h at room temperature followed by 22 h at 4°C. The solution was centrifuged for 15 min at 13,000 × g and the supernatant used for immunostaining as described above.

Immunoreactivity of brain sections was examined using a light microscope (Axiophot, Zeiss, Jena, Germany) and images were acquired using a connected camera (AxioCam HRc, Zeiss). The density of Fos positive cells in each brain section was determined semi-quantitatively using a 10x objective and described as −, no; +, low (~1–10 cells); ++, medium (~10–20 cells); and +++, high (>20 Fos positive cells in a 100 μm × 100 μm area of an ocular grid with a 10x objective) density of expression. Coordinates of the brain nuclei were identified according to the rat brain atlas (Paxinos and Watson, 2007). The investigator was blinded to the experimental group. The average density of Fos immunoreactive cells derived from the total number of sections analyzed for each nucleus was determined for each animal and used to calculate the mean density of expression per group.

Distribution of the data was determined by the Kolmogorov-Smirnov test. Data are expressed as mean ± SEM and analyzed by one-way analysis of variance (ANOVA) followed by Tukey post-hoc test or two-way or three-way analysis of variance followed by the Holm-Sidak method. Differences were considered significant when p < 0.05 (SigmaStat 3.1., Systat Software, San Jose, CA, USA).

During the first week of the experiment (access to the running wheel in single housing cages for activity and activity-based anorexia group; regular single housing conditions for ad libitum and restricted feeding group) no body weight differences were observed between the four groups (Figure 1). After the start of food restriction, rats of the restricted feeding group as well as activity-based anorexia rats showed a body weight decrease, while the ad libitum and activity groups continued to gain body weight (Figure 1). At the end of the 14-day food restriction period, rats of the restricted feeding group showed a body weight decrease of −13% vs. day eight (p < 0.001) and vs. AC (−22%, p < 0.001) and AL (−26%, p < 0.001) that gained body weight (+10% and +13%, respectively; p < 0.001; Figure 1). Activity-based anorexia rats showed an additional body weight loss of −9% compared to rats of the restricted feeding group (p < 0.001; Figure 1) reaching an average body weight loss of −22% during the 14-day observation period. Three-way ANOVA showed a significant influence of time [F_{(13, 504)} = 5.6, p < 0.001], activity [F_{(1, 504)} = 436.1, p < 0.001] and feeding regimen [F_{(1, 504)} = 9806.3, p < 0.001] as well as an interaction of these three factors [F_{(13, 504)} = 2.0, p = 0.02].

It is to note that three out of 14 rats subjected to activity-based anorexia failed to lose body weight (or even gained body weight during the experimental period, data not shown) and were therefore excluded from further analyses (final n = 11).

The food intake observed in the two ad libitum fed groups (ad libitum and activity group) did not differ from each other and was fairly stable over the 21-day observation period (Figure 2). While during the first 3 days of the habituation period—although during this time also fed ad libitum—the food intake was lower in the activity as well as activity-based anorexia group compared to the ad libitum and the restricted feeding groups (p < 0.001) giving rise to more time spent for physical activity and less for food intake, food intake was similar on days 6 and 7 (before food restriction) in all four groups (Figure 2). After food restriction to 1.5 h per day in the restricted feeding and activity-based anorexia group, food intake significantly dropped by −88% in these groups on the first day compared to the ad libitum group (p < 0.001) and slowly increased afterwards to reach the same level as observed in the ad libitum fed groups on the last day of the observation period (Figure 2). Overall, rats of the restricted feeding and activity-based anorexia group ate less (−38 and −41%, respectively) compared to the ad libitum group during the 14-day food restriction period (Figure 2). Three-way ANOVA indicated a significant impact of time [F_{(20, 776)} = 24.5, p < 0.001], activity [F_{(1, 776)} = 68.3, p < 0.001] and feeding regimen [F_{(1, 776)} = 626.5, p < 0.001] as well as an interaction of these three factors [F_{(20, 776)} = 1.8, p = 0.02]. Food intake during 1.5 h did not differ between the restricted feeding and activity-based anorexia group, while 24-h food intake did not differ between the ad libitum and activity group (p > 0.05; Figure 3).

Physical activity assessed using a running wheel slightly increased during the first week from ~1500 to ~2000 wheel rotations/day (Figure 4). During the food restriction period, physical activity more prominently increased reaching ~3500 wheel rotations/day in both, the activity and activity-based anorexia groups (Figure 4), corresponding to ~1300 m/day. No daily differences were observed either in the 1.5-h (data not shown) or 24-h wheel rotations between the activity and activity-based anorexia group (p > 0.05; Figure 4).

The daily energy expenditure including calculated resting energy expenditure and energy expenditure while running was ~35 kcal/200 g body weight at the beginning of the observation period (data not shown). This value slightly increased during the food restriction period in both the activity and activity-based anorexia group to ~38 kcal/200 g body weight (data not shown). Two-way ANOVA indicated a significant impact of time [F_{(20, 377)} = 3.6, p < 0.001] and feeding regimen [F_{(1, 377)} = 4.3, p = 0.04].

Caloric deficit was calculated by subtracting energy expenditure from calculated caloric intake. All four groups showed a caloric surplus during the first 7 days of the experimental period (ranging from 9.6 to 21.4 kcal/day on day 7, data not shown). While this surplus remained visible in the ad libitum fed groups over the remaining 14-day observation period (14.2 ± 4.6 kcal in the ad libitum and 17.1 ± 2.3 kcal in the activity group), in the restricted feeding groups a caloric deficit was observed from the start of the food restriction (greatest levels on day 8: restricted feeding: −30.5 ± 1.4 kcal, activity-based anorexia: −29.6 ± 0.9 kcal) with a progressive decrease of caloric deficit reaching a surplus again on day 21 (restricted feeding: 10.4 ± 3.6 kcal, activity-based anorexia: 2.0 ± 0.9 kcal, data not shown). Three-way ANOVA indicated a significant impact of time [F_{(20, 771)} = 22.3, p < 0.001], activity [F_{(1, 771)} = 125.8, p < 0.001] and feeding regimen [F_{(1, 771)} = 666.7, p < 0.001] as well as an interaction of these three factors [(F_{(20, 771)} = 2.0, p < 0.01].

After analysis of overall daily 24-h (in the ad libitum fed groups) and 1.5-h (in the restricted feeding groups) food intake, the underlying food intake microstructure was assessed using an automated food intake monitoring device. When analyzing the 1.5-h food intake microstructure no difference was observed between both restricted feeding groups (restricted feeding and activity-based anorexia, p > 0.05; Figures 5A–G). Similarly, the two ad libitum fed groups did not show a difference except for the bout size which was smaller in the activity compared to the ad libitum group (p < 0.01; Figure 5A). The two restricted feeding groups showed significantly higher levels for several parameters of the food intake microstructure such as meal size (p < 0.001; Figure 5B), bout frequency (p < 0.001; Figure 5C), meal duration (p < 0.001; Figure 5E), time spent in meals (p < 0.001; Figure 5F) and eating rate (p < 0.001; Figure 5G), while for meal frequency the difference was only observed when compared to the activity group (p < 0.05; Figure 5D).

Also when analyzing the 24-h food intake microstructure, no difference was observed between the two ad libitum fed groups (ad libitum and activity) or the two restricted feeding groups (restricted feeding and activity-based anorexia, p > 0.05; Figures 6A–G). While both restricted feeding groups showed higher levels in bout size (p < 0.001; Figure 6A), meal size (p < 0.001; Figure 6B), meal duration (p < 0.001; Figure 6E) and eating rate (p < 0.001; Figure 6G), values were lower for—as expected based on the feeding schedule—bout frequency (p < 0.001; Figure 6C), meal frequency (p < 0.001; Figure 6D) and time spent in meals (p < 0.001; Figure 6F).

To investigate neuronal activation of brain areas under conditions of activity-based anorexia we performed immunohistochemistry for the activity marker Fos. After pre-absorption of the Fos antibody with a synthetic Fos fragment no immunostaining was observed (data not shown) indicating the specificity of the antibody.

Overall, activity-based anorexia rats showed higher Fos activation levels compared to ad libitum fed rats (Table 1). In the forebrain, higher activity was observed in the piriform cortex, cingulate cortex, somatomotor cortex, lateral septal nucleus, caudate putamen and hippocampus (Figures 7A,B) of activity-based anorexia rats compared to the ad libitum group, while no activation was observed in the amygdala (Table 1). In the thalamus higher Fos activation was observed in the paraventricular thalamic nucleus of activity-based anorexia rats compared to the ad libitum group, while in the lateral habenula similar numbers of Fos positive cells were detected using semiquantitative assessment (Table 1). Evaluation of hypothalamic nuclei showed more Fos signals in the suprachiasmatic nucleus, supraoptic nucleus (Figures 7C,D), anterior hypothalamic area, both magno- and parvocellular parts of the paraventricular nucleus (Figures 7E,F), lateral hypothalamic area (Figures 7G,H), dorsomedial hypothalamic nucleus (Figures 7I,J) and the medial part of the arcuate nucleus (Figures 7K,L) of activity-based anorexia rats compared to the ad libitum group, while similar levels were observed in the ventromedial hypothalamic nucleus (Table 1). In the midbrain similar levels were detected in the Edinger-Westphal nucleus (Figures 8A,B), while higher activation was observed in the dorsal raphe nuclei (Figures 8C,D) and locus coeruleus (Figures 8E,F) of activity-based anorexia compared to the ad libitum fed rats (Table 1). Lastly, also in the medulla a higher activation was observed in activity-based anorexia rats compared to the ad libitum group, namely in the raphe pallidus nucleus (Figures 9A,B), area postrema, rostral part of the nucleus of the solitary tract (Figures 9C,D) and the dorsal motor nucleus of the vagus nerve (Table 1).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.