Physiological and Behavioral Responses to Intermittent Starvation in C57BL/6J Mice

Physiology & Behavior 105 (2012) 376–387

Contents lists available at SciVerse ScienceDirect

Physiology & Behavior

journal homepage: www.elsevier.com/locate/phb

Physiological and behavioral responses to intermittent starvation in C57BL/6J mice

Li-Na Zhang a, Sharon E. Mitchell a, Catherine Hambly a, David G. Morgan b, John C. Clapham b, John R. Speakman a,⁎ a Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen, Scotland, UK b Research and Development (CVGI), AstraZeneca, Alderley Park, Cheshire, UK article info abstract

Article history: The dual intervention point model states that body mass is controlled by upper and lower intervention points, Received 24 May 2011 above and below which animals (and humans) intervene physiologically to bring their body mass back into the Received in revised form 8 August 2011 acceptable range. It has been further suggested that the lower intervention point may be defined by the risk of Accepted 26 August 2011 starvation, while the upper intervention point may be defined by the risk of predation. The objective of the present study was to test whether the risk of starvation determines the lower intervention point and to examine the Keywords: physiological and behavioral mechanisms that underpin the regulation of body mass, when the risk of starvation Intermittent starvation Physical activity is increased. Sixty-four mice were exposed to random days of complete fasting or 50% food restriction and their Body temperature body mass and fat mass responses were measured. Food intake, physical activity and body temperature were C57BL/6J mice measured throughout the experiment. In addition, plasma leptin and insulin, triglyceride and non-esterified fatty acids, along with hypothalamic neuropeptides gene expression in the arcuate nucleus were assessed after 13 and 42 days of treatment. We found that C57BL/6J mice increased body mass and fatness in response to a short-term (13 days) intermittent fasting, which was restored to baseline as the treatment was prolonged. In contrast, intermittently 50% food restricted mice showed no significant changes in body mass or fatness. Over the first 13 days of treatment the data were consistent with the dual intervention point model as the mice showed both increased body mass and adiposity over this period. Over the more protracted period of 42 days the effect waned and was therefore inconsistent with the model. The body mass and fat mass gains in intermittently fasted mice were mainly accounted for by increased food intake. Elevated NPY gene expression after 13 days (three 24 h fasting events) may have driven the increase in food intake. However, no changes were observed in such neuropeptides as POMC, CART, AgRP, Ob-Rb and SOCS 3 or circulating levels of leptin, insulin, NEFA and TG. Hypother- mia during fasting days may have also contributed to the increase in body mass. Over 42 days of treatment (nine 24 h fasting events) cumulative food intake was not affected by intermittent starvation. However physical activity, mainly activity during the light phase was lowered suggesting an adaptation to unpredictable starvation. Overall, mice exhibited different behavioral and physiological responses to intermittent starvation depending on the duration of treatment. © 2011 Published by Elsevier Inc.

1. Introduction known to play an important role in energy homeostasis because it contains two distinct populations of neurons: pro-opiomelanocortin 1.1. Body mass regulation (POMC) neurons that express the endogenous anorectic melanocortin receptor agonist (α-melanocyte stimulating hormone, α-MSH) The global epidemic of obesity has raised the need for a better un- along with cocaine-and-amphetamine-regulated transcript (CART), derstanding of the mechanisms that regulate body and fat mass. A and neurons that express neuropeptide Y (NPY) and also the endog- balance between energy intake and energy expenditure is necessary enous orexigenic melanocortin receptor antagonist, agouti-related to maintain a stable body mass. It is well established that the central protein (AgRP) [4–6]. These neuronal populations respond to periph- nervous system (CNS) regulates food intake and energy expenditure eral signals, such as leptin, insulin and gastrointestinal hormones by in response to neuronal, hormonal and nutrient signals [1–3]. The hy- modifying the production of the above neuropeptides that modulate pothalamus is the most studied area in CNS with respect to the regu- energy balance [7]. Moreover, leptin signaling within neurons is de- lation of energy homeostasis. The arcuate nucleus in particular is pendent on the presence of the long form of the leptin receptor (Ob-Rb) [8], which signals principally via the Janus kinase 2 (JAK2)/ signal transducer and activator of transcription 3 (STAT3) pathway ⁎ Corresponding author. Tel.: +44 1224 272879; fax: +44 1224 272396. [9]. The suppressor of cytokine signaling 3 (SOCS3) is a negative reg- E-mail address: [email protected] (J.R. Speakman). ulator of the leptin receptor which is stimulated by the JAK/STAT2

0031-9384/$ – see front matter © 2011 Published by Elsevier Inc. doi:10.1016/j.physbeh.2011.08.035 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 377 pathway [10–13]. Gene expression levels of Ob-Rb and SOCS3 are to nine random days of complete fasting or food restriction, and re- markers of leptin sensitivity [14]. cording their body mass and fat mass responses. Second we investi- There have been several theoretical models that have attempted gated the physiological and behavioral mechanisms that underpin to conceptualize the mechanisms involved in the regulation of body the regulation of body mass during intermittent fasting or food mass and fatness. Kennedy (1953) proposed the lipostatic set point restriction, by measuring food intake, physical activity and body tem- model which suggested that the size of body fat depots is sensed by perature throughout the experiment. In addition, NPY, AgRP, POMC, a ‘lipostat’, which adjusts food intake and energy metabolism to CART, Ob-Rb and SOCS3 gene expression in the arcuate nucleus of maintain the body and fat masses at a set-point [15]. Although leptin the hypothalamus were assessed after 13 days and 42 days of treat- has been often interpreted as the molecular manifestation of the lipo- ment. We also measured plasma leptin and insulin, as well as triglyc- static fat signal [16], this model is in conflict with the evidence in pat- eride (TG) and non-esterified fatty acids (NEFA). We hypothesized terns of changes in animal and human body mass [17,18].An that if the dual intervention point model is correct randomly imposed alternative interpretation emerged suggesting that body mass is not intermittent fasting and food restriction would cause body mass and regulated by a set-point, but rather is controlled by upper and lower fat mass to increase as a result of an increase in the lower interven- intervention points, above and below which animals (and humans) tion point. Secondly, we hypothesized that physiological and behav- intervene physiologically to bring their body mass back into the ac- ioral compensation mechanisms would be employed to adjust body ceptable range [19,20]. In humans the upper intervention point may mass and fatness within the range and thirdly, that leptin, insulin be located at different positions in different individuals explaining and other potential signals would change in response to intermittent why some individuals become obese when exposed to environments fasting and food restriction and neuropeptides sensitive to leptin with readily available food supplies, but others are able to regulate would be involved in the body mass response. their body weights at normal levels. Based on data from small mammals and birds, Speakman (2007) has further suggested that the 2. Material and methods lower intervention point may be defined by the risk of starvation while the upper intervention point may be defined by the risk of pre- 2.1. Animal housing and intermittent starvation regimen dation [20]. In this scenario, one would expect that an increased risk of starvation would increase the lower intervention level and animals Sixty four female C57BL/6 mice aged 6–8 weeks were purchased would gain fat and body mass. from Charles River (Charles River UK Ltd, Kent, UK) and housed in single cages (M3 cage 48×15×13 cm, NKP Cages, Kent, UK) in a tem- 1.2. Intermittent starvation perature controlled room (21±1 °C) under a 12:12-hour light:dark photoperiod with the lights coming on at 06:00 and a “dawn/dusk” Periods of negative energy balance arising from restricted feeding period of 20 min at either end of the light period. Wood shavings or total starvation are common events [21]. Throughout their lives, and shredded paper bedding were provided for enrichment. At small mammals must face periodic food shortages interspersed by pe- around age 9–10 weeks mice were implanted with transmitters riods of food abundance. that measured their body temperatures and physical activity levels Stochastically imposed intermittent periods of fasting (complete (details below) and were then fed a standard control diet absence of food) or periodic food restriction (involving reduced levels (D12450B, 10% kcal/fat, Research Diet, New Brunswick, NJ, USA) ad relative to habitual intake, but not complete absence) may mimic the libitum until 22 weeks of age when they were randomly assigned unpredictable food availability in the wild. However, few studies have into three groups: a control group AL, fed ad libitum (n=20); IF, in- investigated the physiological and behavioral responses to stochastic termittently fasted (n=22); IR, intermittently 50% food restricted food exposure. Swiss mice showed increased food intake and (n=22). All mice had free access to water throughout the study. All decreased energy expenditure on days that intervened between procedures were reviewed by a local ethical committee and per- 24 h fasting events over 4 weeks treatment (including 3 or 4 fasting formed in accordance with UK home office regulations under license days and 4 or 3 feeding days each week), and they decreased overall PPL 60/3707 held by JRS. body mass [22]. However in another study the same group found that Mice were divided into two cohorts containing 32 mice in each AL 4 weeks treatment including 3 fasting days and 4 feeding days each (n=10); IF (n=11); IR (n=11). Note the sample sizes in each group week had no effect on overall body mass [23]. Alternate day fasting and cohort were unbalanced because the number of recording pads (ADF) is a similar and widely studied regimen, however, most of the for the implanted transmitters (see below) was fixed at 64. Body studies have been focused on the beneficial effects on aging and the mass and food intake were recorded for 3 days on a daily basis prior effect of ADF on body weight is less often reported. Body mass has to the intermittent starvation regime. Animals were stochastically ex- been shown to be highly variable in response to ADF in both humans posed to 24 h fasting or restriction on 9 occasions over a period of and in animal models. In Fisher rats, when ADF regimens were 42 days. Each starvation day was always followed by a non-starvation applied in the short term, no effect on body weight was observed day. Probability of next day would be starvation was set at 0.21 and after 2 weeks [24], whereas gains in body mass were noted in was decided using random integers. During the treatment, days 4, 7, C17BL/10 mice after 8 weeks [25]. However, in other studies, when 10, 14, 16, 20, 26, 29 and 39 were assigned as fasting/restriction ADF was administered for 12 weeks, body mass decreased in both days. On each treatment day, IF mice were completely deprived of C57BL/6 mice and rats. This variability in response is confusing and food, while IR mice were given 50% of their average daily food intake its relationship to the dual-intervention model is uncertain because measured over the baseline period at the start of the experiment. in ADF the fasting days come at predictable intervals, and hence the Mice were provided with ad libitum food between fasting/restriction uncertain risk of starvation is not altered by this treatment. Even in days. Body mass and food intake were measured at 1500 h every day the studies performed by Zhao and colleagues [22,23] the occurrence when food was given or removed. The first cohort of mice was killed of fasting days is so frequent that it may be predictable — leading to on day 13 of the treatment, while the second cohort was killed on day different responses from those predicted by the dual-intervention 42. point model. In the present study we aimed firstly to test whether starvation 2.2. Physical activity and body temperature risk defines the lower intervention point for body mass and adiposity as suggested in the dual intervention point model, by experimentally Prior to experimentation the mice were implanted with a teleme- increasing starvation risk over a period of 42 days, by exposing mice try transmitter to monitor body temperature (Tb) and physical 378 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 activity (PA) (Model PDT-4000 E-Mitter, Mini-Mitter, Bend, OR, USA). linearized DNA was transcribed using T7, T3 or SP6 RNA polymerase The surgery was carried out under anesthesia induced by a mixed promoters matched to a restriction site contained within the vector. flow of isofluorane and oxygen. An incision of approximately 1.0 cm Riboprobes were labeled with 35S-UTP (Perkin Elmer, UK) and unin- was made in the ventral skin, then a second smaller incision made corporated label removed by spinning through Chromospin columns through the peritoneal wall. The transmitter was inserted intraperito- (BD Biosciences, UK). Finally riboprobes were made up to a final con- neally, and the two layers were sutured separately (Ethicon Vicryl: centration of 1×106 c.p.m ml− 1 in hybridization buffer. Antisense W9982; W&J Dunlops Ltd., Dumfries, UK). The surgical procedure and sense riboprobes were tested for specificity and background took around 15–20 min/mouse and mice were given 2 weeks with binding. minimal handling to allow them to recover from the surgery. A receiver pad (ER-4000 Receiver, Mini-Mitter, Bend, OR, USA) under 2.4.3. In situ hybridization the cage where the mouse was housed received PA and Tb information Labware was baked at 200 °C and all solutions made up in 0.1% from the transmitter which was collected every minute by a Windows diethylpyrocarbonate (DEPC) to prevent RNase contamination. Sections PC-based data acquisition system (VitalView™:Mini-Mitter,Bend, were fixed in 4% paraformaldehyde (Sigma-Aldrich, UK)/0.1 M phos-

OR, USA). PA and Tb recordings were averaged for each hour to show phate buffer (144 mM NaH2PO4 and 5.6 mM Na2HPO4, pH 7.4; PB) for daily activity and Tb patterns. Daily means, and means during dark 20 min on ice and washed in 2×0.1 M PB. To eliminate background sig- and light phases were calculated throughout the baseline and treatment nal and inactivate RNases, sections were first immersed in 0.1 mM periods. triethanolamine (TEA) (Sigma-Aldrich, UK) for 2 min before acetylation in 0.1 mM TEA/0.25% acetic anhydride (Sigma-Aldrich, UK) for 10 min 2.3. Body fatness and plasma metabolites levels at room temperature. Sections were again washed in 0.1 M PB for 4 min, dehydrated through increased concentrations of ethanol (50%, Body fat mass (FM) was determined using dual energy X-ray ab- 70%, 90% and 100%) and dried under vacuum. Labeled probe was sorptiometry (DXA; PIXImus2 Series Densimeters, GE Medical Systems pipetted onto cover slips, annealed to sections and sealed with DPX Ultrasound and BMD, Bedford, UK). Mice were anesthetized during the before an overnight hybridization at 58 °C. Post-hybridization, cover scan. Data generated from the DXA software were corrected using an slips were removed by soaking slides in 4 X saline-sodium citrate buffer equation developed specifically for our machine [26].DXAwasper- (150 mM sodium chloride and 15 mM sodium citrate, pH7; SSC). Non- formed to assess body fatness at baseline in both cohorts, and addition- specific hybridization was dissociated by increasing the temperature, ally on day 12 for cohort 1 and on days 24, 30, 37 for cohort 2. Mice were up to 60 °C for 30 min. Unhybridized RNA was digested with RNase sacrificed after fasting for 4 h. Blood samples were collected by cardiac for 30 min. Sections were desalted over a series of SSC solutions: 2 X puncture for the determination of circulating hormone and metabolite SSC, 1 X SSC, 0.5 X SSC and 0.1 X SSC and finally dehydrated through levels (leptin, insulin, TG and NEFA). Fat depots (epidydimal fat, perire- graded increased concentrations of 50%, 70%, 90% and 100% ethanol. nal fat, mesenteric fat and subcutaneous fat) were dissected and The slides were air-dried and exposed to Kodak BioMax Film (Sigma, weighed. All blood samples were placed in EDTA-treated vials and cen- UK) for time appropriate to specific riboprobe. Autoradiographs of sec- trifuged at 4 °C at 10000 rpm for 10 min. Blood plasma was stored at tions including microscale standards were scanned on Umax Power −80 °C until analysis. Look II (UMAX Data System, Fremont, CA, USA), and gene expression Plasma leptin and insulin were measured with ELISA-based was taken as the integrated optical density (IOD) in the area of interest methods using commercially available kits (mouse leptin ELISA kit, using ImageJ software system (WinZip Computing Inc., USA). The slides Millipore Billerica, MA, USA; mouse insulin ELISA kit, Crystal Chem were observed by a blinded observer. Inc, Downers Grove, IL, USA, respectively), while plasma NEFA and TG were measured using colorimetric assays (NEFA-HR (2), Wako 2.5. Statistical analysis Chemicals, Neuss, Germany; Triglyceride and Free Glycerol Kit, Sigma-Aldrich, Kent, UK, respectively). All data were expressed as means±SD. General linear modeling (GLM) with individual ID included as a random factor nested within 2.4. Gene expression of neuropeptides the treatment to account for the repeated measures was used to com-

pare changes of body mass, body fat mass, food intake, PA and Tb 2.4.1. Tissue collection and preparation throughout the experiment. One-way analysis of variance (ANOVA) Mice were killed by CO2 inhalation. Their brains were immediately re- with post hoc Tukey tests was used to test the between-group differ- − moved, frozen in isopentane over dry ice and stored at 80 °C. Coronal ences of body mass, body fat mass, food intake, cumulative food 20 μm sections of the hypothalamic region were cut on a cryostat (Leica, CM3050S, Milton Keynes, UK) and thaw mounted on poly-L- lysine-coated slides. 20 slides (with approximately 6 sections on each slide) through the hypothalamus were collected per animal, and they Table 1 Details of primers used, the fragment size and Accession numbers. were numbered as 1–10, 11–20. Two slides numbered as 1 and 11 (or 2 and 12, etc.) were used for one specific neuropeptide. Gene Primers 5′ to 3′ bp Accession number AgRP TCCTTCCCCAATCCCAATCC 205 NM_007427 2.4.2. Riboprobe synthesis CGCCCTACCTTCTACTTTATCC Template DNA, complimentary to the mRNA sequence of interest, CART CCGAGCAGCGAGGAGGTCCAGAA 372 NM_013732 GGACAGTCACACAGCTTCCCGATCC was gifted from Dr. Sharon Mitchell. Briefly fragments were amplified NPY ATGGACAAGATCCGGGACAG 227 NM_010934 from mouse brain cDNA by polymerase chain reaction (PCR). Primers TAAAAGATGGGGTTGACGCAGGTGG (Eurofins MWG Operon, London, UK), fragment size and Accession POMC GGGCAAGCGCTCCTACTCCAT 348 NM_008895 numbers are detailed in Table 1. PCR products were cloned into either CACTGGCCCTTCTTGTGCG the pCR-Script Amp SK (+) vector (Stratagene, CA, USA) or the ObRb GTCCTCTTCTTCTGGAGC 634 NM_146146 GTGTGAGCATCTCTCCTGGAG pGMET Easy Vector (Promega, Southampton, UK) and transformed SOCS-3 CGCCTCAAGACCTTCAGCTCCAA 565 NM_007707 into competent cells. Plasmid DNA was isolated using the QIAprep GGGTCACTTTCTCATAGGAGT Spin Miniprep kit (Qiagen, Germany) and linearized with appropriate Agouti-related peptide, (AgRP); cocaine- and amphetamine-regulated transcript restriction enzymes (Promega, Southampton, UK). All sequences (CART); neuropeptide Y (NPY); pro-opiomelanocortin (POMC); signaling form of were verified by Eurofins Sequencing Services (London, UK). The leptin receptor (obRb); suppressor of cytokine signaling 3 (SOCS 3). L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 379

intake, PA and Tb at different time points where group differences mice (ANOVA, PN0.05). Over the 42 days of treatment, there was a were signiﬁcant over the whole time course. The two cohorts were signiﬁcant effect of intermittent starvation on body mass within analyzed separately. Analysis of covariance (ANCOVA) was per- groups but not between groups (GLM repeated measures: day of treat- formed to examine differences body fat mass using body mass as a co- ment, F2,29=63.29, Pb0.001; day of treatment∗group, F2,29 =12.80, variate and ANCOVA was also used to test differences in plasma leptin Pb0.001; group, F2,29 =0.48, P=0.622; Fig. 1B). levels using fat mass as a covariate. Group differences in plasma insu- All mice in cohort 1 and cohort 2 had similar fatness before the lin, TG, NEFA levels and fat depots were examined with one-way treatment started (IF: 3.52±0.50 g; IR: 3.44±0.57 g; AL: 3.64±

ANOVA. Group differences in neuropeptides gene expression were 0.70 g: ANCOVA: F2,62 =0.18, P=0.836; Fig. 1C). After 3 starvation examined by comparison of the total integrated optical density days and 9 free feeding days on day 12, there was a signiﬁcant differ-

(IOD) for each gene using ANOVA. Gene expression data were pre- ence in body fat mass between three groups (ANCOVA: F2,29 =4.24, sented as percentage of the mean for AL control. P valuesb0.05 P=0.025; Fig. 1C). IF mice had higher body fat than AL control were considered statistically signiﬁcant. All data were analyzed (Post hoc Tukey test: P=0.020) while no difference was found be- using SPSS 18.0 statistical package for Windows (SPSS Inc., Chicago, tween IR mice and AL control (P=0.068) or between IF and IR IL, USA). (P=0.816)(Fig. 1C). On day 24 (after 6 starvation days and 18 free feeding days), the fat mass of IF mice were signiﬁcantly higher than

3. Results AL control (ANCOVA: F2,29 =5.02, P=0.014; Post hoc Tukey test: IF vs AL, P=0.01; Fig. 1C). No differences were found between the IR 3.1. Body mass and fatness group and the AL control or between the IR and IF groups (Post hoc Tukey test: P=0.162; P=0.384, respectively). However, the differ- 3.1.1. Cohort 1 ences in body fatness observed at days 11–13 had disappeared by Prior to the start of intermittent starvation regime, animals had day 30 (after 8 starvation days and 22 free feeding days) and day 37 similar body mass on day 0 (IF, 24.46±1.15 g; IR, 24.28±1.24 g; (after 8 starvation days and 29 free feeding days) (ANCOVA:

AL, 24.29±1.27 g; ANOVA: F2,29 =0.08, P=0.926; Fig. 1A). The inter- F2,29 =1.66, P=0.209; F2,29 =1.56, P=0.227, respectively; Fig. 1C). mittently fasted (IF) mice signiﬁcantly decreased body mass on the ﬁrst (day 4) and second (day 7) days of fasting (ANOVA: day 4, 3.2. Food intake

F2,29 =7.00, P=0.003; day 7, F2,29 =4.71, P=0.017; Fig. 1A). The IF mice did not exhibit a significant decrease in body mass on the 3.2.1. Cohort 1 third day of fasting (day 10) (ANOVA: F2,29 =1.96, P=0.159). After Before intermittent starvation treatment started, all mice had a 3 days of fasting and 7 days of free feeding, on days 11–13, IF mice similar amount of food intake (IF: 2.64±0.31 g; IR: 2.67±0.51 g; maintained a significantly higher body mass in comparison to both 2.59±0.33 g; ANOVA: F2,29 =0.12, P=0.899; Fig. 2A). The IF mice AL control and the intermittently restricted (IR) mice (ANOVA: day in cohort 1 displayed hyperphagia on the first two free feeding days

11, F2,29 =4.18, P=0.025; day 13, F2,29 =3.89, P=0.032; day 13, after the fasting days (ANOVA: days 5–6, days 8–9, day 11: F2,29 =3.78, P=0.035; Fig. 1A). pb0.001; day 12, p=0.012; Fig. 2A). By the end of this treatment (day 13), cumulative food intake differed signiﬁcantly between the

3.1.2. Cohort 2 three groups (ANOVA: F2,29 =5.73, P=0.008; Fig. 2B). Overall, IR There were no significant differences in body mass between groups mice consumed 2.79 g (9%) less food (P=0.012) than AL controls before the treatment started (ANOVA: F2,29 =0.01,P=0.989; Fig. 1B). while the amount of cumulative food intake consumed by IF mice On day 0, the average body mass of IF, IR and AL groups were 23.11± was similar to that of AL control (P=0.902) over the course of the 1.19 g, 23.12±1.21 g and 23.19±1.21 g, respectively. The responses treatment (Fig. 2B). of the second cohort over the first 13 days of treatment were almost identical to those observed in cohort 1. After the first starvation day, 3.2.2. Cohort 2 IF mice showed a significantly lowered body mass compared to AL con- Under the free feeding baseline condition, there was no difference in trol (Post hoc Tukey test: Pb0.001) and IR mice (Post hoc Tukey test: food intake between groups (ANOVA: PN0.05). On the first free feeding P=0.045), whereas no differences were detected between IR mice day following the first starvation (day 5), IF and IR mice consumed sig- and AL control (Post hoc Tukey test: PN0.05) (ANOVA: F2,29 =11.22, nificantly more food than the AF control (ANOVA: F2,29 =12.41, Pb0.001; Fig. 1B). The body mass of IF mice returned to the AL control Pb0.001; Post hoc Tukey test: Pb0.001, P=0.029, respectively; level after one-day of free feeding (P=0.605). The second starvation Fig. 2C). On the second free feeding day (day 6), the food intake of IF day resulted in a similar response in IF mice exemplified by a significant mice was still higher than AL control (ANOVA: F2,29 =11.917, reduction in body mass (ANOVA: F2,29 =11.45, Pb0.001), which was Pb0.001; Post hoc Tukey test: P=0.002; Fig. 2C) while the food intake followed by an immediate increase after refeeding (ANOVA: of the IR mice returned to the similar level of AL control (Post hoc Tukey

F2,29 =1.31, P=0.285). Moreover, after the second fasting event the IF test: P=0.858). Compared to AL controls, no increase in food intake fol- mice showed a trend for body mass to increase above that of the AL con- lowing the treatment days was found in IR mice from the second re- trol animals (0.69 g higher) although this difference was not statistical- striction day onwards (Post hoc Tukey test: PN0.05 in all cases). In ly significant (ANOVA: F2,29 =2.39, P=0.109). Different from the first contrast, IF mice showed a hyperphagic response to starvation on the two starvation days, no significant body mass loss relative to the masses first two free feeding days that followed both the second and third re- of the other groups was not observed in IF mice after the third fasting strictions (days 8–9anddays11–12) (ANOVA: day 8, F2,29 =12.58, day and no between groups difference in body mass was detected Pb0.001; day 9, F2,29 =6.42, P=0.005; day 11, F2,29 =33.83,Pb0.001; (ANOVA: F2,29 =2.70, P=0.08). Free feeding following the third starva- day 12, F2,29 =5.11, P=0.013; Post hoc Tukey test: IF vs AL, Pb0.05 tion day caused a pronounced weight increase in the IF mice above AL for all cases; Fig. 2C). The magnitude of hyperphagia was less on the sec- control levels which lasted for 3 continuous free feeding days ond refeeding day than that on the first refeeding day. The patterns in

(ANOVA: F2,29 =4.57, P=0.019; F2,29 =4.10, P=0.027; F2,29 =4.20, cohort 2 matched those of cohort 1 over the first 13 days of treatment P=0.025; respectively. Post hoc Tukey test: IF vs AF, Pb0.05; IR vs AF, almost exactly. From days 14–38, hyperphagia following fasting in the PN0.05). As the treatment progressed, this response of elevated body IF group was present only on the first free feeding days after starvation mass weakened. Significant mass losses in IF mice occurred on fasting (ANOVA: day 15, F2,29 =11.21,Pb0.001; day 17, F2,29 =16.86,Pb0.001; days 4 to 9 (ANOVA, Pb0.05) while body mass on the free feeding day 21, F2,29 =18.44, Pb0.001; day 27, F2,29 =4.56, P=0.019; day 30, days did not increase significantly above the levels of AL control or IR F2,29 =10.51, Pb0.001; Post hoc Tukey test: IF vs AF, Pb0.05 for all 380 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387

A 27 ** 26 * 25 * * 24 23 22 21 20 Body mass (g) IF IR AL 19 18 -2 0 2 4 6 8 10 12 Days of treatment

B 26 25 ** * 24 23 22 21 ** ** *

Body mass (g) 20 * *

19 IF IR AL 18 -20 2 4 6 8 1012141618202224262830323436384042 Days of treatment

C 6 AL IR IF b b 5 a ab a a 4

2 Body fat mass (g)

0 Day0 (Cohort1+2) Day12 (Cohort1) Day24 (Cohort2) Day30 (Cohort2) Day37 (Cohort2)

Fig. 1. Effects of intermittent fasting/food restriction on body mass and fatness in female C57BL/6J mice. A. Body mass over 13 days of intermittent fasting/food restriction treatment (cohort 1). B. Body mass over 42 days of intermittent fasting/food restriction treatment (cohort 2). C. Body fat mass at different time points during intermittent fasting/food restriction treatment (a combination of cohorts 1 and 2). Filled triangles represent intermittently fasted mice (IF), Filled squares represent intermittently food restricted mice (IR), open squares represent control fed ad libitum (AL). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). *, Pb0.05. Different letters over bars on the same treatment day indicate signiﬁcant difference.

cases; Fig. 2C). Unexpectedly, on the first free feeding day after the ninth measures: day 4, F2,29 =0.28, P=0.758; Fig. 3B). When the treatment starvation (day 40), IF and IR mice consumed significantly less food proceeded to the second fasting/restriction day, 24 h physical activity than AL control (ANOVA: F2,29=29.19, Pb0.001; Post hoc Tukey test: levels tended to differ between the three groups although the difference IF vs AL, Pb0.001; IR vs AL, P=0.01; Fig. 2C). There was no significant did not reach significance (GLM repeated measures: day 7, F2,29 =3.08, difference between three groups in cumulative food intake over P=0.062; Fig. 3C). From the third fasting/restriction day (day 10) and

42 days of the treatment (ANOVA: F2,29=2.56,P=0.095; Fig. 2D). onwards, there was a significant effect of fasting/restriction on 24 h physical activity (GLM repeated measures: day 7, F2,29 =9.42, 3.3. Physical activity P=0.001; Post hoc Tukey test: IF vs AL, P=0.001, IR vs AL, P =0.104; IF vs IR, P=0.086, Fig. 3D; data on days 14, 16, 20, 26, 29 and 39 Fig. 3A illustrates physical activity over a free feeding day prior to not shown). In addition to changes in 24 h activity, there was no differ- the experimental treatment starting. Mice from all groups showed a ence in daily average activity between the three groups on the first similar circadian pattern in physical activity with a low level of activity two fasting/restriction days (ANOVA: PN0.05 for both cases; Fig. 3E). during the light phase (punctuated by a brief rise when disturbed to be However, IF mice showed a significant decrease in daily average fed at 1500 h) and a high level of activity during the dark phase. On the physical activity on the third to ninth fasting/starvation days (ANOVA, first fasting/restriction day (day 4), there was no significant difference Pb0.05, Post hoc Tukey test, IF vs AL, Pb0.05) and IR mice significant in 24 h physical activity between the three groups (GLM repeated decreased daily average activity on the fifth to ninth fasting/starvation L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 381 AB 120 a b b IF IR AL 100 5 * 4.5 * 4 ** * 80 3.5 ** 3 * * 60 2.5 * 2 40 1.5 1 20 0.5

Daily food intake (g) 0 0 -2024681012 Cumulative food intake (g) AL IR IF Days of treatment

C 6 IF IR AL 5 * * * * * 4 * * * * * * * * * * 3 * * * * * * * 2 * 1 Daily food intake (g)

0 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 Days of treatment D a 40 a 35 b 30 25 20 15 10 5 0

Cumulative food intake (g) AL IR IF

Fig. 2. Effects of intermittent fasting/food restriction on food intake in female C57BK/6J mice. A. Daily food intake over 13 days of intermittent fasting/food restriction treatment (cohort 1). B. Cumulative food intake over 13 days of intermittent fasting/food restriction treatment. C. Daily food intake over 42 days of intermittent fasting/food restriction treatment (cohort 2). D. Cumulative food intake over 42 days of intermittent fasting/food restriction treatment. Filled triangles represent intermittently fasted mice (IF), Filled squares represent intermittently food restricted mice (IR), open squares represent control fed ad libitum (AL). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). *, Pb0.05. Different letters over bars on the same treatment day indicate signiﬁcant difference.

days (ANOVA, Pb0.05, Post hoc Tukey test, IR vs AL, Pb0.05) (Fig. 3E). fasting/restriction events (ANOVA: Pb0.05) except the fourth one Furthermore, day of treatment representing the number of events had on day 14 (ANOVA: P=0.081) and on the second refeeding days asignificant effect on physical activity over the 13 fasting/restriction after the first-third, fifth, seventh and ninth fasting/restriction events days (GLM repeated measures: group, F2,29 =7.27, P=0.003; group∗- (ANOVA: Pb0.05) (Fig. 3H). Moreover, day of treatment (represent- day of treatment, F2,16 =3.31, Pb0.001; Fig. 3E). ing number of events) had a significant effect on physical activity dur- The first day of refeeding (day 5) following the first fasting/restric- ing all free feeding and refeeding days (GLM repeated measures: tion (day 4) 24 h physical activity levels differed significantly be- group, F2,29 =3.74, P=0.037; group∗day of treatment F2,64 =3.39, tween groups (GLM repeated measures: F2,29 =24.62, Pb0.001). IF Pb0.001). Overall, total activity levels of IF mice were significantly and IR mice displayed lower activity levels than AL control (Post lowered by 17% in comparison with AL control over 42 days of treat- hoc Tukey: IF vs AL, Pb0.001; IR vs AL, P=0.001; IF vs IR, P=0.025; ment compared with AL controls (ANOVA, F2,29 =4.79, P=0.017; Fig. 3F). Similar responses persisted on the second day of refeeding Post hoc Tukey test, IF vs AL, P=0.017) whereas no differences on day 6 (GLM repeated measures: F2,29 =7.20, P=0.003; Post hoc were found between IR mice and AL control (P=0.82) (Fig. 3I). Con- Tukey: IF vs AL, P=0.002; IR vs AL, P=0.312; IF vs IR, P=0.076) sistent with changes in total daily activity, there were significant dif- even though the magnitude of responses became weaker. On the ferences between three groups in activity during the light phase third day of refeeding (day 13) after the third fasting/restriction (ANOVA: F2,29 =11.13, Pb0.001). However, activity levels were event (day 10), the difference in physical activity between groups to- unchanged during the dark phase across the three groups (ANOVA: tally disappeared (GLM repeated measures: F2,29 =0.80, P=0.462; F2,29 =2.22, P=0.128). In cohort 2, intermittent fasting and restric- Fig. 3G). Additionally, IF mice and IR mice displayed a decrease in tion had exactly the same impact on physical activity as during the daily physical activity on the first refeeding days following all first 13 days of treatment in cohort 1. 382 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387

ABC24h PA at baseline 24h PA on 1st starving day 24h PA on 2nd starving day 25 35 30 IF 30 25 20 IR 25 20 15 AL 20 15 10 15 10 10 5 5 5

Physical activity (AU) 0 0 0 18:00 20:00 22:00 06:00 08:00 16:00 02:00 10:00 00:00 04:00 12:00 14:00 22:00 00:00 04:00 08:00 06:00 18:00 12:00 14:00 16:00 20:00 02:00 10:00 04:00 02:00 06:00 08:00 16:00 18:00 10:00 14:00 20:00 22:00 00:00 12:00 DEF 24h PA on 3rd starving day Daily PA on 9 starving days 24h PA on 1st refeeding day 12 25 30 * 10 * * 20 25 * * 8 * * 20 15 6 15 10 10 4 5 5 2 Physical activity (AU)

Physical activity (AU) 0 0 0 123456789 20:00 08:00 22:00 00:00 06:00 12:00 14:00 16:00 02:00 04:00 18:00 10:00 08:00 12:00 06:00 20:00 22:00 00:00 02:00 04:00 18:00 Days of fasting/restriction 10:00 14:00 16:00 Daily average physical activity (AU)

G24h PA on 3rd refeeding day H Daily PA on refeeding days I Accumulated PA over 42 days 25 12 ** 450 b 20 10 ** ** ** *** ** * 400 a a 8 350 15 300 6 250 10 200 4 150 5 100

Daily activity (AU) 2 Physical activity (AU) 50 0 0 0 1357911131517192123252729313335373941 IF IR AL Days of treatment 18:00 04:00 08:00 10:00 20:00 02:00 12:00 22:00 00:00 16:00 06:00 14:00 Accumulated physical activity (AU)

Fig. 3. Effects of intermittent fasting/food restriction on physical activity in female C57BL/6J mice. A. Representative graph of 24 h physical activity at baseline. B. 24 h physical activity on the

ﬁrst fasting/food restriction day (GLM repeated measures: F2,29 =0.28, P=0.758). C. 24 h physical activity on the second fasting/food restriction day (GLM repeated measures: F2,29 =3.08,

P=0.062). D. 24 h physical activity on the third fasting/food restriction day (GLM repeated measures: F2,29 =9.42, P=0.001). E. Daily average physical activity on 9 fasting/food restriction days (GLM repeated measures: group, F2,29 =7.27, P=0.003; group∗day of treatment, F2,16 =3.31, Pb0.001). F. 24 h physical activity on the ﬁrst day of refeeding following the ﬁrst fasting/ restriction (GLM repeated measures: F2,29 =24.62, Pb0.001). G. 24 h physical activity on the third day of refeeding after the third fasting/restriction event (GLM repeated measures:

F2,29 =0.80, P=0.462). H. Daily average physical activity on refeeding days. I. Accumulated physical activity over 42 days of intermittent fasting/food restriction treatment. Filled triangles represent intermittently fasted mice (IF), Filled squares represent intermittently food restricted mice (IR), open squares represent control fed ad libitum (AL). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). *, Pb0.05. PA, physical activity. Different letters over bars on the same treatment day indicate signiﬁcant difference.

3.4. Body temperature Fig. 4D). There were no signiﬁcant differences in daily average temperature across the three groups on all refeeding days (ANOVA: All mice showed a similar circadian pattern of body temperature PN0.05 in all cases; Fig. 4E). In cohort 2, the effects of intermittent when fed ad libitum (Fig. 4A). On day 4, 24 h fasting and food restric- fasting and restriction on body temperature were exactly the same tion both led to pronounced hypothermia (GLM repeated measures: as in cohort 1 for the ﬁrst 13 days of treatment.

F2,29 =32.52, Pb0.001; Post hoc Tukey test: IF vs AL, Pb0.001; IR vs AL, P=0.002; Fig. 4B). IF mice had a signiﬁcantly lower body temper- 3.5. Circulating hormone and metabolite levels ature compared with IR mice (Post hoc Tukey test: IF vs IR, Pb0.001). The effect of fasting/restriction on daily average temperature was ob- Circulating levels of leptin were positively correlated with body fat served on each fasting/restriction day (ANOVA: Pb0.001 in all 9 mass on both day 13 and day 42 of the treatment, (Linear regression: 2 cases; Fig. 4C). Compared with control fed ad libitum, average body day 13, F1,30 =7.34, P=0.011, adjusted R =0.17; Fig. 5A; day 42, 2 temperature in IF and IR mice on fasting/restriction days was de- F1,30 =12.13, P=0.001, adjusted R =0.288; Fig. 5B). On day 13, there creased by 2.48 °C and 0.90 °C, respectively. was no signiﬁcant difference in circulating leptin levels across the

Body temperature was returned to the AL control levels after 1 day three groups (ANCOVA using fat mass as a covariate: F2,28 =0.139, of refeeding (GLM repeated measures: F2,29 =1.21, P=0.313; P=0.87; Fig. 5C). However on day 42 IF mice had a signiﬁcantly L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 383

AB39 39 37 37 35 35 33 33 31 31

29 IF IR AL 29 27 27 Body temperature (°C)

Body temperature (°C) 25 25 18:00 20:00 22:00 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 CD 39 39

(°C) 37 37 b 35 35 33 *** 33 31 31 29 29 27 27

Daily average T 25 25 123456789 Body temperature (°C)

Days of fasting/restriction 18:00 20:00 22:00 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 E 39 37 (°C)

b 35 33 31 29 27

Daily average T 25 1 3 5 7 9 11131517192123252729313335373941 Days of treatment

Fig. 4. Effects of intermittent fasting/food restriction on body temperature in female C57BK/6J mice. A. Representative graph of 24 h body temperature at baseline. B. 24 h body temperature on a typical fasting/food restriction day (GLM repeated measures: F2,29 =32.52, Pb0.001). C. Daily average body temperature on 13 fasting/food restriction days. D. 24 h body temperature on a typical refeeding day. E. Daily average body temperature on refeeding days. Filled triangles represent intermittently fasted mice (IF), Filled squares represent intermittently food restricted mice (IR), open squares represent control fed ad libitum (AL). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). ***, Pb0.001.

elevated plasma leptin levels compared with the AL control and IR mice 4. Discussion

(ANCOVA using fat mass as a covariate: F2,28 =4.92, P=0.015; ANOVA post hoc Tukey: IF vs AL, P=0.043, IF vs IR, P=0.002; Fig. 5C). Circulat- 4.1. Effects of intermittent starvation on body mass and fatness in mice ing insulin were unchanged on days 13 and 42 (ANOVA, F2,29=1.60, P=0.220; F2,29 =1.65, P=0.209; Fig. 5D). Intermittent fasting and We examined the changes in body mass and fatness in response to food restriction had no effect on TG or NEFA levels (ANOVA: PN0.05 elevated starvation risk. Our finding that after 13 days of treatment for both TG and NEFA on days 13 and 42; Fig. 5E,F). there was a significant increase in both body mass and fat mass in the group that received intermittent complete food removal supports the suggestion that starvation risk defines the lower intervention 3.6. Hypothalamic neuropeptides gene expression level, according to the dual intervention points model [20]. However, mice that were stochastically exposed to only 50% food restriction On day 13 (cohort 1) after 3 starvation days and 10 free feeding showed no difference in body mass compared to controls that were days, levels of NPY gene expression in arcuate nucleus was signifi- fed ad libitum. This lack of response does not support the suggestion cantly elevated by 49% in IF mice compared with AL control that starvation risk increases the lower intervention point. Moreover,

(ANOVA: F2,29 =3.39, P=0.047; Post hoc Tukey test: IF vs AL, when intermittent starvation was prolonged for a greater duration, P=0.039; Fig. 6A). No differences were detected between IR mice the effects on body mass and adiposity were reduced and not signiﬁ- and AL control at this time point (Post hoc Tukey test: IR vs AL, cantly greater than the AL controls. The absence of signiﬁcant eleva- P=0.257; Fig. 6A). However, the difference in NPY had disappeared tion in body mass in response to the longer-term intermittent after 42 days of treatment (cohort 2) (ANOVA: F2,29 =0.22, fasting is also at odds with the prediction from the dual intervention P=0.805; Fig. 6A). There were no effects of intermittent fasting/res- point model. It seemed that the immediate response of the mice to in- triction on the levels of expression of POMC, CART, AgRP, Ob-Rb and creased starvation risk was indeed to elevate fatness, but that as time SOCS 3 (ANOVA: PN0.05 for all on days 13 and 42; Fig. 6B–F). went on different strategies were employed. One possibility is that a 384 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387

A 20 B 16 IF IR AL 14 IF IR AL 15 12 10 10 8 6 5 4 Plasma leptin (ng/ml)

Plasma leptin (ng/ml) 2 0 0 1 1.5 2 2.5 3 3.5 1 1.5 2 2.5 3 Fat mass (g) Fat mass (g)

2 CD16 IF IR AL 14 IF IR AL 12 a 1.5 10 1 8 b b 6 4 0.5 2 Plasam leptin (ng/ml) Plasam insulin (ng/ml) 0 0 Day 13 Day 42 Day 13 Day 42

EF1.4 1.6 IF IR AL 1.2 IF IR AL 1.4 1 1.2 0.8 1 0.8 0.6 0.6 0.4 0.4 Plasam TG (ng/ml) 0.2 Plasam NEFA (ng/ml) 0.2 0 0 Day 13 Day 42 Day 13 Day 42

Fig. 5. Effects of intermittent fasting/food restriction on circulating hormones and metabolites in female C57BK/6J mice. A. Linear regression plots of plasma leptin levels against 2 body fat mass on day 13 of intermittent fasting/food restriction treatment (Linear regression: F1,30 =7.34, P=0.011, adjusted R =0.17). B. Linear regression plots of plasma leptin 2 levels against body fat mass on day 42 of intermittent fasting/food restriction treatment. (Linear regression: F1,30 =12.13, P=0.001, adjusted R =0.288). C. Plasma leptin levels on day 13 and day 42 of intermittent fasting/food restriction treatment (ANCOVA using fat mass as a covariate: F2,28 =4.92, P=0.015). D. Plasma insulin levels on days 13 and 42 of intermittent fasting/food restriction treatment. E. Plasma triglyceride (TG) levels on days 13 and 42 of intermittent fasting/food restriction treatment. F. Plasma non-esterified fatty acid (NEFA) levels on days 13 and 42 of intermittent fasting/food restriction treatment. Filled triangles represent intermittently fasted mice (IF), Filled squares represent intermittently food restricted mice (IR), open squares represent control fed ad libitum (AL). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). Different letters over bars on the same treatment day indicate significant difference. short term intermittent starvation might provoke a more efficient There clearly is variability in the capacity of animals to compensate for compensation to defend the negative energy balance, thus generating a fast day on a fed day [32] particularly if fasting days are as frequently over compensation in body mass. encountered as the fed days, and this may compromise their capacity to Our result from long-term intermittent starvation is consistent with elevate body fatness. Furthermore, there could be longitudinal changes the finding that Swiss mice and striped hamsters (C. barabensis)also in this capacity to compensate dependent on the duration of treatment. showed no change in body mass after exposure to stochastic food supplies for 4 weeks [23]. In their study, however, Swiss mice and striped 4.2. Physiological and behavioral responses to intermittent starvation hamsters showed no immediate response of body mass gain that atten- uated over time. However, in this latter study the occurrence of fasting 4.2.1. Food intake days was far more frequent than in the present study. Nevertheless, In the present study, mice experiencing intermittent fasting dis- similar findings have been seen in humans. Healthy women subjects played pronounced hyperphagia on the first one or two days of undergoing a restricted diet on 4 days each week exhibited no substan- refeeding days after the fasting day, which largely accounted for the tial weight change after 4 weeks [27]. However, in other studies where weight restoration/elevation. It seemed necessary to have at least ADF regimens were administered for 12 weeks, body weight was found 2 days between the fasting days to build up a fat reserve and hence to decrease [28,29], but, when ADF regimens were applied for 16 weeks, the difference between the current study and the previous studies no effect on body weight was observed [30,31]. Such findings imply that by Zhao and colleagues may rest primarily in the frequency of fasting. animals were unable to consume twice their daily food intake on the If this frequency is too high there may be limited capacity to compen- refeeding day for longer periods, which resulted in body weight loss. sate on the fed days. However, the magnitude of hyperphagia was L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 385 AB 250 160 IF IR AL a 140 200 ab 120 100 150 b

(% of AL control) 80 (% of AL control) 100 60 40

50 expression expression 20 0 0 NPY Day 13 Day 42 POMC Day 13 Day 42 CD

200 200 180 180 160 160 140 140 120 120

(% of AL control) 100 100 (% of AL control) 80 80 60 60

expression 40 40 expression 20 20 0 0 CART Day 13 Day 42 AgRP Day 13 Day 42 EF 200 250 180 160 200 140 120 150 100 (% of AL control) (% of AL control) 80 100 60 40 50 expression expression 20 0 0 ObRb Day 13 Day 42 Day 13 Day 42 SOCS3

Fig. 6. Effects of intermittent fasting/food restriction on gene expression of neuropeptides in the hypothalamus in female C57BK/6J mice. A. Neuropeptide Y. B. Pro-opiomelanocortin (POMC). C. Cocaine- and amphetamine-regulated transcript (CART). D. Agouti-related peptide, (AgRP). E. Signaling form of leptin receptor (Ob-Rb). F. Suppressor of cytokine signaling 3 (SOCS 3). Filled bars represent intermittently fasted mice (IF), Slashed bars represent intermittently food restricted mice (IR), open bars represent control fed ad libitum (AL). Different letters over bars on the same treatment day indicate significant difference. associated with the severity of food deprivation since IF mice con- after 24 h of food deprivation and their daily food intake was no sumed more food than IR mice on the refeeding days. So the failure greater than daily intake during baseline testing [38]. Despite a of IR mice to deposit a fat reserve was clearly not because they hyperphagic response on refeeding days, IF mice consumed a similar did not have the capacity to eat sufficient food on the days when total amount of food over the 13 days of intermittent fasting treatment, food was available. Post fasting or restriction food compensation while IR mice consumed significantly less food compared to AL con- has been observed in a number of previous studies e.g. [33–35]. For trols. However, with an even longer duration of treatment, the differ- instance, rats showed a dramatic post fast hyperphagia and conse- ence in accumulated food intake between IR mice and the AL control quently a rapid recovery of body weight lost during starvation disappeared and mice from all the three groups consumed similar [36,37]. Similarly, an initial hyperphagic response following refeeding total amount of food over 42 days of intermittent fasting/restriction was observed in C57BL/6 mice previously on 40% food restriction [33]. treatment. It has been shown that the amount of food consumed by Hambly et al. (2007) further found that the magnitude of the hyper- rodents maintained on an ADF regimen varies depending on the strain phagic response at refeeding was independent of the prior length of and, in some cases (e.g., C57BL/6 mice), the animals gorge during the caloric restriction [34]. In contrast, golden hamsters (Mesocricetus non-fasting time period to an extent that overall food intake is essen- auratus) of both sexes showed little or no post fast food compensation tially equivalent to that of animals fed ad libitum [32]. Wan et al. 386 L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387

(2003) reported that rats on ADF regimen consumed 30% less food over and refeeding immediately restored body temperature to control time compared with rats fed ad libitum [39]. The intermittent starva- levels. Caloric restriction is associated with lower mean body temper- tion regimen provides stochastic food availability together with food ature in most homeotherms that have been examined including mice, abundance allowing compensation. Our results suggested that the re- rats, monkeys and humans [49]. The use of torpor has been described sponses to refeeding varied with the extent of starvation. IF mice as one of the key strategies for coping with food shortage in rodents were able to compensate for 24 h complete food deprivation on refeed- [51,52]. In the present study, to what extent the reduced body tem- ing days and maintained cumulative food intake at the same level as perature contributed to the saving of energy and thus the prevention AL control. However the responses to refeeding in IR mice changed of weight loss remained unclear, particularly when the reduction in over time. During the first 13 days of treatment, the decrease in cumu- body temperature was restored rapidly following refeeding. lative food intake with unchanged body mass in IR mice implies a decrease in energy expenditure. On day 42, cumulative food consump- 4.3. Effects of intermittent fasting/restriction on neuroendocrine status tion in IR mice was normalized to AL control level indicating other ad- aptations e.g. decreased energy expenditure might also be restored. Leptin is primarily known for its role in regulating food intake and When intermittently 50% food restricted, mice might employed differ- energy homeostasis [53]. At the end of 12 days of intermittent fastin- ent responses dependent on the duration of treatment. g/restriction, no changes in circulating leptin and insulin levels were detected. This was unexpected given both the increased body fatness 4.2.2. Physical activity and food intake in the IF mice. Moreover, after a long-term intermit- In contradiction to previous studies demonstrating an elevation in tent starvation, IF mice exhibited elevated leptin levels while their physical activity level which occurred in several species such as le- body fatness and food intake were restored to the same level as AL murs (Lemur catta) [40], Siberian hamsters (Phodopus sungorus) [41] controls. It is possible that leptin sensitivity might have been im- and mice [42] in response to food restriction, we observed a decrease paired; however, the absence of changes in gene expression of Ob- in physical activity level in mice when fasted or 50% food restricted Rb and SOCS 3 suggested this was not the case. for 24 h, suggesting a response to decrease energy expenditure and With respect to other neuropeptides involved in eating behavior, conserve energy. This is consistent with the finding that mice exposed NPY gene expression was increased after a short-term intermittent to 20% food restriction showed decreased activity which contributed starvation. NPY and AgRP neurons of the hypothalamic arcuate nucle- to 75.5% of altered energy expenditure to compensate for reduced en- us are thought to promote feeding and consistent with an orexigenic ergy intake [43]. Similarly in female monkeys (Macaca mulatta), a role, fasting induced a 4-fold increase in the basal action potential fre- rapid decrease in physical activity was observed to counteract diet- quency of NPY/AgRP neurons [54]. It seems likely that the elevation in induced weight loss [44]. Since the energy cost of physical activity is NPY expression stimulated the increased food intake in the IF mice. an important component of energy expenditure, this physiological re- Similarly, an increase in NPY expression and a parallel decrease in sponse to reduce energy expenditure is potentially an integral factor POMC expression were observed in the hypothalamic arcuate nucleus contributing to protection against excessive weight loss during calo- of rats killed just before rebound hyperphagia following a restricted ric restriction and potentially predisposes to weight re-gain in post- feeding/repletion protocol [55]. obese individuals [45]. Further analyses of activity data showed that the decrease was mainly due to activity during the light phase coinci- 5. Conclusions dent with the occurrence of hypothermia. However, IF and IR mice had similar levels of activity during the dark phase compared to the C57BL/6J mice displayed an increased body mass and fatness in re- controls fed ad libitum. Increased activity during food restriction has sponse to 13 days of intermittent fasting, which were both restored been previously interpreted to reflect an increase in foraging behav- to baseline levels when the treatment was prolonged to 42 days. ior, which may enhance the chances of survival in a free living animal Over the first 13 days of treatment the data were consistent with [46]. This inconsistency could be explained by the short duration of the dual intervention point model as the mice showed both increased starvation (24 h), which failed to induce food foraging behavior in body mass and adiposity, however the effect disappeared over the mice. Additionally, Gutman et al. (2007) observed that desert gold more protracted period of 42 days. Whether the data are considered spiny mice (Acomys russatus) employed two strategies for coping consistent with the dual intervention point model therefore depends with food shortage, some significantly reduced activity level concen- critically on the time course of the experiment. Over 13 days the re- trating their activity around feeding time lowering energy expendi- sponse was consistent but over 42 days it was not. In the short ture while others significantly increased activity level searching for term, an increase in food intake, decreased physical activity as well food and thus losing body mass rapidly [47]. It is currently unclear as a reduction in body temperature contributed to the elevated whether and how the response in activity is correlated with the se- body mass and fatness. In contrast, the body mass of mice treated verity or duration of starvation. On refeeding days after starvation, with intermittent fasting for 42 days were not significantly elevated. our mice still maintained a lowered level of physical activity, which This finding was inconsistent with the dual intervention point also contributed to the recovery and the increased body mass of the model. Intermittent starvation resulted in unaltered cumulative IF mice. Although it has been suggested that hypoleptinemia can food intake and meanwhile a decreased physical activity level, sug- influence starvation induced hyperactivity [48,49], in our study the gesting an absence of overcompensation in energy intake but a de- decreased physical activity during intermittent starvation was ac- crease in energy expenditure in coping with stochastic food. companied by either unaltered (13 days) or elevated (42 days) leptin Overall, mice exhibited different behavioral and physiological re- levels, indicating that lowered leptin was unlikely to be the stimulus sponses to intermittent starvation which were critically dependent of the activity changes. on the duration of treatment.

4.2.3. Body temperature Acknowledgment Mice are documented as an attractive model for studying the response of body temperature to dietary restriction because body tem- This work was funded by a studentship from University of Aberdeen perature of mice is sensitive to reduced food intake [50]. Indeed mice and AstraZeneca. The authors wish to thank Dr Cathy Wyse for her from our study demonstrated a decrease in body temperature in re- patient help with cryostat brain sectioning and for her valuable com- sponse to 24 h of complete fasting or 50% food restriction. Notably, ments to the manuscript. The authors also would like to acknowledge hypothermia was only present when food was deprived or restricted Rachel Sinclair and Dr Lobke Vaanholt for their help with animal work L.-N. Zhang et al. / Physiology & Behavior 105 (2012) 376–387 387 and statistical analyses. Thanks to the staff in animal house for taking [28] Hsieh EA, Chai CM, Hellerstein MK. Effects of caloric restriction on cell proliferation in several tissues in mice: role of intermittent feeding. Am J Physiol Endocrinol Metab care of the mice. 2005;288(5):51–5. [29] Ahmet I, Wan R, Mattson MP, Lakatta EG, Talan M. Cardioprotection by intermittent fasting in rats. Circulation 2005;112(20):3115–21. References [30] Descamps O, Riondel J, Ducros V, Roussel A-. Mitochondrial production of reactive oxygen species and incidence of age-associated lymphoma in OF1 mice: effect of [1] Mercer JG, Speakman JR. Hypothalamic neuropeptide mechanisms for regulating alternate-day fasting. Mech Ageing Dev 2005;126(11):1185–91. energy balance: from rodent models to human obesity. Neurosci Biobehav Rev [31] Mager DE, Wan R, Brown M, Cheng A, Wareski P, Abernethy DR, et al. Caloric re- 2001;25(2):101–16. striction and intermittent fasting alter spectral measures of heart rate and blood [2] Morton GJ, Cummings DE, Baskin DG, Barsh GS, Schwartz MW. Central nervous pressure variability in rats. FASEB J 2006;20(6):631–7. system control of food intake and body weight. Nature 2006;443(7109):289–95. [32] Anson RM, Guo Z, de Cabo R, Iyun T, Rios M, Hagepanos A, et al. Intermittent fast- [3] Schwartz MW, Woods SC, Porte Jr D, Seeley RJ, Baskin DG. Central nervous system ing dissociates beneficial effects of dietary restriction on glucose metabolism and control of food intake. Nature 2000;404(6778):661–71. neuronal resistance to injury from calorie intake. Proc Natl Acad Sci U S A May 13 [4] Coll AP. Effects of pro-opiomelanocortin (POMC) on food intake and body weight: 2003;100(10):6216–20. mechanisms and therapeutic potential? Clin Sci 2007;113(3–4):171–82. [33] Clinthorne JF, Adams DJ, Fenton JI, Ritz BW, Gardner EM. Short-term re-feeding of [5] Ilnytska O, Argyropoulos G. The role of the agouti-related protein in energy balance previously energy-restricted C57BL/6 male mice restores body weight and body fat regulation. Cell Mol Life Sci 2008;65(17):2721–31. and attenuates the decline in natural killer cell function after primary influenza in- [6] Cone RD. Anatomy and regulation of the central melanocortin system. Nat Neurosci fection. J Nutr 2010;140(8):1495–501. 2005;8(5):571–8. [34] Hambly C, Mercer JG, Speakman JR. Hunger does not diminish over time in mice [7] Archer ZA, Mercer JG. Brain responses to obesogenic diets and diet-induced obesity. under protracted caloric restriction. Rejuvenation Res Dec 2007;10(4):533–42. Proc Nutr Soc 2007;66(1):124–30. [35] Evans SA, Messina MM, Knight WD, Parsons AD, Overton JM. Long-Evans and [8] Mercer JG, Hoggard N, Williams LM, Lawrence CB, Hannah LT, Trayhurn P. Locali- Sprague–Dawley rats exhibit divergent responses to refeeding after caloric re- zation of leptin receptor mRNA and the long form splice variant (ob-rb) in mouse striction. Am J Physiol Regul Integr Comp Physiol Jun 2005;288(6):R1468–76. hypothalamus and adjacent brain regions by in situ hybridization. FEBS Lett [36] Müller-Riemenschneider F, Reinhold T, Berghöfer A, Willich SN. Health-economic 1996;387(2–3):113–6. burden of obesity in europe. Eur J Epidemiol 2008;23(8):499–509. [9] Hübschle T, Thom E, Watson A, Roth J, Klaus S, Meyerhof W. Leptin-induced nuclear [37] Robin J-, Decrock F, Herzberg G, Mioskowski E, Le Maho Y, Bach A, et al. Restoration translocation of STAT3 immunoreactivity in hypothalamic nuclei involved in body of body energy reserves during refeeding in rats is dependent on both the intensity weight regulation. J Neurosci 2001;21(7):2413–24. of energy restriction and the metabolic status at the onset of starvation. J Nutr [10] Tups A, Ellis C, Moar KM, Logie TJ, Adam CL, Mercer JG, et al. Photoperiodic regu- 2008;138(5):861–6. lation of leptin sensitivity in the siberian hamster, phodopus sungorus, is reflected [38] Silverman HJ, Zucker I. Absence of post fast food compensation in the golden in arcuate nucleus SOCS-3 (suppressor of cytokine signaling) gene expression. hamster (Mesocricetus auratus). Physiol Behav 1976;17(2):271–85. Endocrinology 2004;145(3):1185–93. [39] Wan R, Camandola S, Mattson MP. Intermittent fasting and dietary supplementation [11] Bjørbæk C, Lavery HJ, Bates SH, Olson RK, Davis SM, Flier JS, et al. SOCS3 mediates with 2-deoxy-D-glucose improve functional and metabolic cardiovascular risk factors 985 feedback inhibition of the leptin receptor via Tyr . J Biol Chem 2000;275(51): in rats. FASEB J 2003;17(9):1133–4. 40649–57. [40] Geiser F. Metabolic rate and body temperature reduction during hibernation and [12] Howard JK, Cave BJ, Oksanen LJ, Tzameli I, Bjoørbæk C, Flier JS. Enhanced leptin sen- daily torpor; 2004 [cited 13 January 2011]. sitivity and attenuation of diet-induced obesity in mice with haploinsufficiency of [41] Ruf T, Klingenspor M, Preis H, Heldmaier G. Daily torpor in the Djungarian hamster Socs3. Nat Med 2004;10(7):734–8. (Phodopus sungorus): interactions with food intake, activity and social behaviour. J [13] Kievit P, Howard JK, Badman MK, Balthasar N, Coppari R, Mori H, et al. Enhanced Comp Physiol B 1991;160(6):609–15. leptin sensitivity and improved glucose homeostasis in mice lacking suppressor of [42] Williams TD, Chambers JB, Henderson RP, Rashotte ME, Overton JM. Cardiovascular cytokine signaling-3 in POMC-expressing cells. Cell Metab 2006;4(2):123–32. responses to caloric restriction and thermoneutrality in C57BL/6J mice. Am J Physiol [14] Król E, Duncan JS, Redman P, Morgan PJ, Mercer JG, Speakman JR. Photoperiod Regul Integr Comp Physiol 2002;282(5):51–5. regulates leptin sensitivity in field voles, microtus agrestis. J Comp Physiol B [43] Hambly C, Speakman JR. Contribution of different mechanisms to compensation 2006;176(2):153–63. for energy restriction in the mouse. Obes Res 2005;13(9):1548–57. [15] Kennedy GC. The role of depot fat in the hypothalamic control of food intake in [44] Sullivan EL, Cameron JL. A rapidly occurring compensatory decrease in physical the rat. Proc R Soc Lond B Biol Sci Jan 15 1953;140(901):578–96. activity counteracts diet-induced weight loss in female monkeys. Am J Physiol [16] Zhang Y, Proenca R, Maffei M, Barone M, Leopold L, Friedman JM. Positional cloning Regul Integr Comp Physiol 2010;298(4). of the mouse obese gene and its human homologue. Nature 1994;372(6505): [45] Leibel RL, Rosenbaum M, Hirsch J. Changes in energy expenditure resulting from 425–32. altered body weight. N Engl J Med 1995;332(10):621–8. [17] Wirtshafter D, Davis JD. Set points, settling points, and the control of body weight. [46] Cornish ER, Mrosovsky N. Activity during food deprivation and satiation of six Physiol Behav 1977;19(1):75–8. species of rodent. Anim Behav 1965;13(2–3):242–8. [18] Berthoud HR. Homeostatic and non-homeostatic pathways involved in the control of [47] Gutman R, Yosha D, Choshniak I, Kronfeld-Schor N. Two strategies for coping with food intake and energy balance. Obesity (Silver Spring) 2006;14(Suppl 5). food shortage in desert golden spiny mice. Physiol Behav 2007;90(1):95–102. [19] Levitsky DA. Putting behavior back into feeding behavior: a tribute to george collier. [48] Hillebrand JJG, Kas MJH, van Elburg AA, Hoek HW, Adan RAH. Leptin's effect on Appetite 2002;38(2):143–8. hyperactivity: potential downstream effector mechanisms. Physiol Behav [20] Speakman JR. A nonadaptive scenario explaining the genetic predisposition to 2008;94(5):689–95. obesity: the “predation release” hypothesis. Cell Metab Jul 2007;6(1):5–12. [49] Exner C, Hebebrand J, Remschmidt H, Wewetzer C, Ziegler A, Herpertz S, et al. [21] Wang T, Hung CCY, Randall DJ. The comparative physiology of food deprivation: Leptin suppresses semi-starvation induced hyperactivity in rats: implications for from feast to famine; 2006 [cited 31 October 2008]. anorexia nervosa. Mol Psychiatry 2000;5(5):476–81. [22] Cao J, Zhang L-, Zhao Z-. Trade-off between energy budget, thermogenesis and be- [50] Rikke BA, Johnson TE. Physiological genetics of dietary restriction: uncoupling the havior in swiss mice under stochastic food deprivation. J Therm Biol 2009;34(6): body temperature and body weight responses. Am J Physiol Regul Integr Comp 290–8. Physiol Oct 2007;293(4):R1522–7. [23] Zhao Z-, Cao J. Plasticity in energy budget and behavior in Swiss mice and striped [51] Gutman R, Choshniak I, Kronfeld-Schor N. Defending body mass during food re- hamsters under stochastic food deprivation and refeeding. Comp Biochem Physiol striction in Acomys russatus: a desert rodent that does not store food. Am J Physiol A Mol Integr Physiol 2009;154(1):84–91. Regul Integr Comp Physiol Apr 2006;290(4):R881–91. [24] Siegel I, Liu TL, Nepomuceno N, Gleicher N. Effects of short-term dietary restriction [52] Génin F, Perret M. Daily hypothermia in captive grey mouse lemurs (Microcebus on survival of mammary ascites tumor-bearing rats. Cancer Invest 1988;6(6): murinus): effects of photoperiod and food restriction. Comp Biochem Physiol B 677–80. Biochem Mol Biol 2003;136(1):71–81. [25] Krizova E, Simek V. Effect of intermittent feeding with high-fat diet on changes of [53] Leibel RL. The role of leptin in the control of body weight. Nutrition Reviews glycogen, protein and fat content in liver and skeletal muscle in the laboratory 2002;60(10 I). mouse. Physiol Res 1996;45(5):379–83. [54] Takahashi KA, Cone RD. Fasting induces a large, leptin-dependent increase in the [26] Johnston SL, Peacock WL, Bell LM, Lonchampt M, Speakman JR. PIXImus DXA with intrinsic action potential frequency of orexigenic arcuate nucleus neuropeptide Y/ different software needs individual calibration to accurately predict fat mass. Agouti-related protein neurons. Endocrinology 2005;146(3):1043–7. Obes Res 2005;13(9):1558–65. [55] Lauzurica N, García-García L, Pinto S, Fuentes JA, Delgado M. Changes in NPY and [27] Laessle RG, Platte P, Schweiger U, Pirke KM. Biological and psychological correlates of POMC, but not serotonin transporter, following a restricted feeding/repletion protocol intermittent dieting behavior in young women. A model for bulimia nervosa. Physiol in rats. Brain Res 2010;1313(C):103–12. Behav 1996;60(1):1–5.