Female mice respond differently to costly foraging versus food restriction

Energy balance is a key mediator of individual survival and reproductive output in free-living animals, which makes energy a useful currency for understanding life history trade-offs(Drent and Daan, 1980; Ricklefs and Wikelski, 2002; Speakman, 1997; Stearns, 1992). Several environmental, physiological and behavioral factors may influence patterns of energy intake and expenditure in free-living animals(Speakman, 2000). Among these,there has been limited but intriguing evidence suggesting that environmental quality influences both resting metabolic rate (RMR) and daily energy expenditure [DEE (Speakman et al.,2003)]. One attractive hypothesis is that poor environments may force animals to expend additional energy as a result of the increased cost of foraging (Wiersma et al.,2005). However, such a functional link is far from being established. An experimental approach is useful to gain further insight into the manner in which foraging costs influence patterns of energy allocation.

Several studies have manipulated foraging costs per reward in the laboratory using a work-for-food design. In rodents, wheel running is typically used as a proxy for foraging activity(Perrigo, 1987; Perrigo and Bronson, 1983; Perrigo and Bronson, 1985; Vaanholt et al., 2007),whereas studies in avian species have required animals to fly short distances,hop on perches or sift through chaff to receive food (reviewed by Wiersma and Verhulst, 2005). Animals facing elevated foraging costs increase their activity or keep activity constant (Vaanholt et al.,2007; Wiersma and Verhulst,2005). Food intake and DEE are decreased when a fixed reward schedule is used (Bautista et al.,1998; Day and Bartness,2001; Deerenberg et al.,1998; Perrigo,1987; Vaanholt et al.,2007) but may be elevated when rewards are unpredictable(Bautista et al., 1998; Wiersma et al., 2005).

Various types of changes occur in response to manipulation of foraging costs. Animals typically alter their time–activity budgets(Deerenberg et al., 1998; Perrigo, 1987), utilize stored energy (Day and Bartness, 2001; Vaanholt et al., 2007) and reduce whole body mass and organ size(Vaanholt et al., 2007). Reductions in resting metabolic rate are typical (reviewed by Wiersma et al., 2005). Even more extreme savings may occur through hypothermia, which has been observed in starving animals maintained on low levels of nutrition (e.g. Daan et al., 1989; Gelegen et al., 2006; Hudson and Scott, 1979) and may also be possible under elevated foraging costs(Perrigo and Bronson, 1983; Vaanholt et al., 2007).

In addition to changes in patterns of energy use, high foraging costs have secondary physiological consequences which could compromise fitness and survival prospects of free-living animals. Most small mammals are sensitive to food availability in scheduling their reproductive effort(Wade and Schneider, 1992),and alterations in energy balance may reduce fertility and breeding success(Johnston et al., 2006). Perrigo (Perrigo, 1987) found that high foraging costs are detrimental to breeding success of two species of wild mice (Mus musculus and Peromyscus maniculatus). Along with reductions in total expenditure, animals faced with high foraging costs reallocate energy away from somatic repair [e.g. feather re-growth(Wiersma and Verhulst, 2005)],protection from cellular damage by free oxygen radicals(Wiersma et al., 2004), and immune function (Deerenberg et al.,1997). Trade-offs of this type can affect mortality risk and rates of senescence.

Work-for-food experiments are a tool which has been used to mimic natural variation in foraging costs. This design simultaneously exposes animals to the dual energetic challenges of forced activity and reduced food intake. It has thus far remained unclear to what extent changes in physiological indicators of energy balance (e.g. body composition and metabolic rate) are the combined result of increased rates of energy turnover on the one hand and of food restriction on the other. To address this problem, we set out to explicitly separate these effects. We manipulated female laboratory mice by exposing them to low and high foraging costs and compared the latter to a group of inactive animals receiving a food ration matched to the intake of animals with high foraging costs. We measured activity, food intake, energy metabolism, body mass and composition, as well as secondary responses in estrous cyclicity and in primary immune response to a novel antigen challenge. Our hypothesis was that effects of high foraging costs are due to the combined effect of food restriction and high energy turnover. Therefore, we predicted that effects such as decreased immune response, and loss of ovarian cyclicity would be more dramatic in animals facing high foraging costs than either other group.

A total of 24 female mice (Mus musculus domesticus L.) of the outbred Hsd:ICR(CD-1) strain, bred at our facilities at the Biological Center in Haren, The Netherlands, were used in the experiment; founders of this population were obtained from Harlan, France in 2005. At the age of 4 months animals were individually housed in standard cages (Macrolon type II long 33×15×13 cm, UNO Roestvaststaal BV, Zevenaar, The Netherlands)without running wheels. All mice were kept in the same room at 21±1°C on a 12 h:12 h light:dark cycle (lights on 04:00 GMT+1 h)and had ad libitum access to water and food (standard rodent chow:RMH-B/2181, Arie Block BV, Woerden, The Netherlands). Mice were abdominally implanted with 1.2 g temperature transmitters (Series 3000 XM-FH Mini Mitter,Bend, OR, USA) under surgical anesthetic and were allowed to recover for one week in their home cages before the start of the experiment. Temperature data are not presented here, and body mass data were corrected for the mass of implanted transmitters. All procedures concerning animal care and treatment were in accordance with the regulations of the University of Groningen's ethical committee for the use of experimental animals (license DEC 4321).

At the start of the experiment, animals were assigned to three experimental groups (N=8 per group): (group 1) L, mice that faced low foraging costs; (group 2) H, mice that faced high foraging costs; (group 3) R, mice that did not have to forage for food, but were each pair-fed with a mouse from the H group. At this time L and H mice were housed in Plexiglas cages(20×20×30 cm) fitted with plastic running wheels (diameter 14 cm;code 0131 Savic®, Kortrijk, Belgium). R mice remained in standard cages without wheels. The study was divided into three phases: baseline, training and workload phases. The start of the training phase was termed experimental day 0. Time notation followed this convention throughout. Body mass and food intake (to the nearest 0.1 g) were measured daily throughout the study,starting 1–2 h prior to lights off so that the disturbance coincided roughly with activity onset. This was also the time at which the R mice received their food ration based on the food intake of the H mice over the previous day. The weighing sequence was rotated daily. After weighing, we hand-sifted bedding to count uneaten food blocks or pellets, which were always removed. Running-wheel activity [RWA; in number of revolutions (revs)] was logged in 2 min time intervals (bins) using a computerized event recording system. We calculated travel distances in km day^–1 and estimated time spent running (h day^–1) by counting the number of 2-min bins with a value of >0. Maximum and mean running speeds were estimated from the highest number of revolutions in a 2-min bin and from the mean number of revolutions across all non-zero bins, respectively. Activity of R mice housed without wheels was not measured.

In the baseline phase (days –27 to –1), all mice were kept on ad libitum food. Individual baseline reward rates (in revs pellet^–1) were calculated by dividing mean RWA (revs day^–1) by gross energy intake (kJ day^–1) for each mouse over days –14 to –1 of this phase. This measure was then adjusted for the manufacturer's energy content of the rodent blocks given during the baseline phase and the precision pellets which would later be used during the training and workload phases(Table 1).

In the training and workload phases, all animals were switched to a diet of 45mg precision food pellets (TestDiet 5TUM/PJAI, Sandown Chemicals, Hampton,Surrey, UK). Cages of L and H mice were connected to food dispensers (Med Associates Pellet Dispenser ENV-203, Sandown Scientific, Hampton, UK) that delivered food pellets on a fixed reward ratio (as described above) and were linked to a steering computer (Series 3 Programmable Controller, General Electric). On average, mice had to run 130±10 revs pellet^–1 (∼17000revsday^–1) to obtain their baseline food intake. Foraging costs of the L mice remained constant at their individual reward rate during the training and workload phases (days 0–40). Thus, when running the same amount as during the Baseline phase,they would have received an amount of food similar to their mean intake during that phase. In practice, L mice ran more during the training and workload phases than during the baseline phase and received more pellets than they ate each day. In the H group, foraging costs per reward increased over the training phase: foraging costs were raised by 10% of the individual baseline every other day, until they had doubled. Foraging costs then remained constant at 200% of baseline, marking the start of the workload phase (day 21). Mice in the R group did not forage to obtain food, but were pair-fed with animals from the H group: each R-group mouse received a daily ration of precision pellets matched to the previous day's intake of an individual H mouse (i.e. they were`yoked controls').

Vaginal cytology was scored after daily weighing from day –27 to 28. A vaginal swab was taken from each animal (while gently holding it by the tail) using a paintbrush dampened with distilled water. Cells were smeared onto a clean glass slide and stained with a drop of Methyl Blue dye (S. Johnston, personal communication). A single observer blind to time and treatment visually scored slides with a light microscope at 100×magnification. Following Miller et al.(Miller et al., 2004), the cycle phase was identified based on the proportion of leukocytes, nucleated epithelial cells, and cornified epithelial cells counted in >100 cells per smear. An animal was considered fertile in the baseline phase if it entered estrous (⩾50% cornified cells per smear) on at least three occasions between day –27 and –1, and at least once between day –7 and–1. For statistical analyses, the day with the highest percentage of cornified cells measured in a 1-week period (see Data analysis below) was used as an estrous score.

In the workload phase [days 25–28 (±1 day)] daily energy expenditure (DEE; kJ day^–1) of animals in their home cages was estimated using the doubly labeled water method [DLW(Lifson and McClintock, 1966; Speakman, 1997)]. The protocol followed Vaanholt et al. (Vaanholt et al.,2007). Mice were weighed to the nearest 0.1 g, briefly restrained by the scruff of the neck, and injected i.p. with ∼0.1 g of enriched water(37.6% ²H and 60.6% ¹⁸O). The precise dose injected was quantified by weighing syringes to the nearest 0.0001 g before and after injection. After a 1 h equilibration period(Król and Speakman,1999) animals were bled at the tail tip, and an initial blood sample (15 μl) was collected in duplicate glass capillary tubes, which were immediately flame-sealed with a propane torch. Mice were then returned to their home cages. Forty-eight hours after the initial sample(Speakman and Racey, 1988), a final blood sample was collected in triplicate following the same procedure.

Resting metabolic rate (RMR; kJ h^–1) was estimated as the lowest value of a 20 min running mean of MR (typically three measurement points) and later expressed as kJday^–1. We also estimated the average daily metabolic rate (ADMR; kJ day^–1) for animals in the metabolic chambers.

Determinations of ²H:¹H and ¹⁸O:¹⁶O ratios in blood samples were performed by mass spectrometry at the University of Groningen's Center for Isotope Research(Visser and Schekkerman,1999). Blood samples were prepared by microdistillation in a vacuum line, first heating the broken tubes and then cryogenically trapping the emerging water vapor with liquid nitrogen(Nagy, 1983). Water samples were stored and then automatically injected into a Hekatech high temperature pyrolysis unit (Gehre et al.,2004), in which the injected water was reacted with glassy carbon. The resultant H₂ and CO gases, emerging into a continuous He flow-through system, were then led through a GC column to separate the two gases in time and finally fed into a GVI Isoprime isotope ratio mass spectrometer for the analysis of δ¹⁸O andδ ²H. Measurements were corrected for memory effects using an algorithm similar to the one described by Olsen et al.(Olsen et al., 2006). At least three internal water standards chosen to cover the entire enrichment range of the blood samples were prepared and analyzed following the same methods. We measured samples in duplicate unless a flaw was detected in the flame-sealing step. Typical relative duplo differences were below 2.5% forδ ²H, and 1% for δ¹⁸O. If differences exceeded 3% we critically examined the data and omitted the aberrant value,performing further calculations on a single replicate. Otherwise, duplicate values were averaged.

On day 39 (±1 day), primary immune responsiveness was assayed by challenging animals with a novel antigen, keyhole limpet hemocyanin (KLH). After weighing, each mouse was injected s.c. with 0.1 ml of 0.9% sterile saline containing 150 μg KLH (Calbiochem, Merck KGaA, Darmstadt, Germany;#374811, lot #B304050). Blood samples were taken from the tail tip at 5 and 10 days after injection to measure anti-KLH immunoglobulin (IgG) production. Samples of ∼100 μl were collected in unheparinized glass capillary tubes and allowed to clot on ice. After removing clots, samples were centrifuged at 2500 g for 1 h at 4°C. Serum was aspirated and stored it at –80°C until analysis with an enzyme-linked immunosorbant assay for anti-KLH IgG [analyzed at Indiana University by G. D. following Demas et al. (Demas et al.,1997)]. Briefly, 96-well microtiter plates were coated with antigen by incubating overnight at 4°C with 0.5 mg ml^–1KLH, washed, and then blocked with 5% nonfat dry milk overnight at 4°C to reduce nonspecific binding. Plasma samples were diluted 1:20. Pilot data showed that peak immune response occurred between day 10 and 15 post-injection; therefore, we only analyzed samples taken on day 10. Positive control samples (pooled plasma from mice previously determined to have high anti-KLH antibodies) and negative control samples (pooled sera from KLH-naive mice) were also added in duplicate to each plate. Secondary antibody [alkaline phosphatase-conjugated anti-mouse IgG diluted 1:2000 with phosphate-buffered saline plus 0.1% Tween 20 (PBS-T; Cappel, Durham, NC, USA)] was added, and plates were read with a 409 nm filter following the addition of the enzyme substrate p-nitrophenyl phosphate (Sigma Chemical, St Louis, MO,USA). Mean optical densities (OD) are expressed as a percentage of the plate positive control OD for statistical analyses.

Ten days after KLH injection, animals were sacrificed for carcass analysis between 13:00 h and 17:00 h (GMT+1 h). Animals were killed by CO₂inhalation and immediately decapitated. Bodies were dissected to separate organs, skin and the musculoskeletal system, and weighed to the nearest 0.0001 g. Samples were stored at –20°C until analysis. We determined dry and lean dry organ masses by drying to constant weight at 103°C [European Standard Protocol ISO 6496-1983(E)] followed by fat extraction with petroleum ether (Boom BV, Meppel, The Netherlands) in a Soxhlet apparatus.

We tested two experimental hypotheses: that high foraging costs per se alter behavior and physiology of female mice (H_A1: H females differ from L females), and that high foraging costs induce quantitatively different responses than food restriction alone(H_A2: H females differ from R females). Because L and R mice differed in two respects – both foraging activity and feeding regime– we could not make a priori hypotheses about the differences between them and did not statistically compare these groups. We tested responses in the workload phase using unpaired t-tests. Some baseline-phase parameters differed between groups, but including baseline phase parameters as model covariates never yielded different results in t-tests (ANCOVA; results not presented). Some additional comparisons used general linear models (GLM).

Three animals died or became ill and were retrospectively excluded from the study. Sample sizes for most analyses were therefore L=8, H=6 and R=7. Because of technical problems, we further restricted analyses of DLW data to L=7, H=6 and R=5; for consistency, we used the same subset of mice for other analyses related to energy metabolism. Data were analyzed using Statistica v. 6.1(StatSoft, Inc., Tulsa, OK, USA). We checked data for normality and arcsin-transformed proportional measures to sin^–1 (√y)before analysis (arcsin-sqrt). Two-tailed P-values of ⩽0.05 were considered statistically significant. Because most organ masses were correlated within individuals, we analyzed body composition using a principal components analysis [PCA; after Selman et al.(Selman et al., 2002)]. This procedure addresses the problem of multiple statistical comparisons on correlated data by creating a smaller number of uncorrelated response variables (principal components) which account for the maximum amount of variation in the data. We applied Varimax normalized rotation to factors with Eigenvalues ⩾1 and performed statistical comparisons on these rotated factors.

Our experimental manipulation affected activity, energy intake and body mass over the course of the experiment(Fig. 1). Chance differences between groups in the baseline phase (Table 2) arose after some animals died or were retrospectively excluded from the study, even though we initially weight-matched the groups [overall mean mass (± s.e.m.)=32.6±0.6]. In the baseline phase, R animals also ate significantly less than L and H animals(Table 2). This is most likely because they did not have running wheels in their cages and had lower energy requirements.

In order to obtain a constant amount of food, H mice would have needed to double their activity, but they did not do this(Fig. 2). Total activity in workload phase did not differ significantly between L and H females(Table 2), nor did their activity rhythms differ (Fig. 3). L and H mice spent 7.1±0.5 and 8.0±0.3 h per day running, respectively, and ran at maximum speeds of 4.4±0.3 vs4.4±0.4 km h^–1 (mean speeds: L=2.6±0.2,H=2.4±0.3). The two groups did not significantly differ in the amount of time spent running or in their running speeds (P>0.1). Although total foraging effort was similar for females facing low and high foraging costs, energy intake was significantly lower for H females(Fig. 1, Table 2). H females (and their R yoked controls) consumed ∼25% less energy than L females in the workload phase. H mice did not maintain their body mass and were significantly lighter than both L and R females in the workload phase(Fig. 1 and Table 2).

H mice used significantly less energy than L mice on a daily basis(Table 3). They also used significantly less energy than R mice when removed from the foraging task(RMR, ADMR; Table 3), even though they received the same amount of food at the start of the overnight measurement. We estimated gross energy intake (GEI) during the workload phase from food consumption and food energy content (measured using bomb calorimetry; Table 1). The ratio of DEE:GEI ranged from 0.72 to 0.80 across experimental groups and was significantly higher in H than R mice(Table 3). Estimated surplus energy (S) was significantly lower in H than in R groups(Table 3).

Workload-phase metabolism – both RMR and DEE – was significantly predicted by whole body mass(Fig. 4). Experimental group had an additional, significant effect on DEE (GLM, mass F_1,14=46.4, P<0.0001; group F_2,14=15.0, P=0.0003;model R²=0.88; non-significant interaction removed). There was also an effect of group on RMR (GLM, group F_2,15=6.9, P<0.01, R²=0.48; non-significant effects removed). These effects can be best understood from calculations of mass-specific energy expenditure. Most mass-specific metabolic parameters were virtually identical for H females and both other groups. The exception to this pattern was DEE, which was equal in L and H females, and significantly lower in R than in H mice (Table 3). Overall, H and R mice both saved energy by reducing body mass.

We compared the body composition of H females to each of the other two groups on day 50 (Tables 4, 5, 6). At this time, H mice had experienced 3 weeks of increasing foraging costs and four weeks of foraging costs at 200% of baseline. The wet masses of all organs were less in H mice than in L mice (Table 4). Wet organ masses were also lower in H than in R mice, but the differences were not as large (Table 4). Organ masses were highly positively correlated within individuals. To analyze differences in overall body composition, we performed a principal components analysis on the correlations between wet organ masses(Table 5). The first three components had Eigenvalues of ⩾1, together explaining 74.7% of the total variance in the data. After normalized Varimax rotation, PC1 had high positive loadings for the major metabolically active tissues: the heart, liver, kidneys and brain (Eigenvalue=5.86, percentage variance=53.3). PC1 also had positive loadings for the muscles/skeleton and pelage. PC2 had a high negative loading for the wet empty stomach mass (Eigenvalue=1.27, percentage variance=11.6),and PC3 had high positive loadings for the lungs and intestine(Eigenvalue=1.07, percentage variance=9.8). Comparing PC scores between groups, we found that PC1 and PC3 were significantly reduced in H vsL females (Table 6). High foraging costs, therefore, reduced the size of metabolic organs,musculoskeletal systems, lungs and emptied guts.

All major body components were dramatically lighter in H mice compared to L mice (Table 6). Whole body mass at sacrifice, fresh mass, dry mass, lean dry mass and body fat were all significantly reduced. Body fat was ∼77% lower in H females than in L females. The fresh mass of the carcass – which was measured after animals had been exsanguinated and the guts had been emptied – was about 30% lower in H mice than in L mice. Several body components were somewhat lighter in H females than in R animals, but only whole body mass and lean dry mass differed significantly between these groups. H females had 55% less body fat on average than R females, but this difference was not statistically significant. Intestine lengths (measured to the nearest 0.1 cm) did not differ between the groups (mean ± s.e.m.: L=51.9±3.2,H=52.6±3.0, R=55.3±3.0).

Our experimental manipulation of foraging costs and food intake markedly influenced estrous cyclicity, a measure of reproductive readiness. We restricted all comparisons to females that came into estrus regularly during the baseline phase (L: 6/8, H: 6/6, R: 5/7). In the workload phase, females facing high foraging costs all ceased to show signs of estrus. H females had a significantly lower proportion of cornified cells than L females (independent t-test on arcsin-sqrt transformed data; H vs L: t₁₀=–5.0, P=0.0006; H vs R:t₉=–1.9, P=0.09). They were therefore less likely to come into estrus at least once during the final week of the workload phase.

Estrous cyclicity was related to body mass, such that H and R groups had similar mean masses (30.7 and 30.8 g, respectively) on the last day they were observed in estrus. Since the decline in body mass was more rapid in H mice,this similarity in body mass suggests a body-mass threshold to maintaining fertility in the face of energy limitations. On average, H females were last observed in estrus on day 5, whereas the average last estrous day for R mice was day 19 (the estrous cycle was monitored until day 28). Most H females stopped cycling soon after they were subjected to foraging costs, while food restriction had a more gradual effect on fertility (R group). Treatment group significantly predicted last estrous day overall (Kruskal–Wallis test:H_2,17=7.0, P=0.03), and H females continued cycling longer than L females (Kruskal–Wallis multiple comparisons H vs L: P=0.04).

An assay of primary immune response performed at the end of the study suggested that the experimental groups differed in antibody production(Fig. 5). Plasma anti-KLH IgG levels at day 10 post-injection did not differ between H and L animals (power analysis: partial η²=0.32, f=0.68, power=0.73), but were lower in both than in R animals (ANOVA, F_2,18=4.2, P=0.03; LSD post-hoc test significant for both L and H vs R). Food restricted mice without running wheels (R), therefore,showed a more robust humoral immune response to a novel antigen than highly active animals receiving sufficient food (L) or experiencing restricted feeding (H). The strength of the immune response could not be attributed to body mass or food intake on the day of injection (GLM; mass: F_1,19=0.1, P=0.79; food: F_1,19=1.5, P=0.24).

Female mice responded to our experimental manipulations with alterations in energy intake and expenditure. Animals subjected to high and low foraging costs obtained different amounts of food, despite showing similar activity levels. Thus, high foraging costs reduced daily gross energy intake and also DEE. Earlier studies manipulating foraging costs per reward with a work-for-food design (Perrigo,1987; Perrigo and Bronson,1983; Perrigo and Bronson,1985; Vaanholt et al.,2007) found similar results but lacked a design feature essential to understanding the underlying process: a control group of food-restricted animals which did not face foraging costs per se. The present design allowed us to decouple the most important aspects of high foraging costs:foraging activity and food intake. Our results demonstrate that both the food restriction and the increased activity which result from high foraging costs on a fixed reward schedule contribute to the metabolic and physiological changes observed in those animals.

Our finding that female mice reduced energy expenditure when faced with high foraging costs is similar to the results of other studies using a fixed reward schedule (reviewed by Vaanholt et al., 2007; Wiersma and Verhulst, 2005). However, it does not match the prediction that foraging costs should increase energy expended on food acquisition [see schematic in Wiersma et al. (Wiersma et al., 2005)]. A negative relationship between foraging costs and energy expenditure also differs from observations made by Speakman et al.(Speakman et al., 2003) in free-living short-tailed field voles, which had higher metabolic rates when wintering in poor quality habitat patches. There may be differences in the way animals respond to fixed versus variable reward schedules, and thus far, only variable rewards have been shown to elevate DEE(Wiersma et al., 2005; Wiersma and Verhulst, 2005). Disparate effects of fixed and variable reward ratios on energy intake have been attributed to differences in motivation(Fotheringham, 1998). Particularly in rodents, it remains unclear to what extent limitations on the intensity and duration of activity influence responses to fixed-reward designs.

The female mice in our study ran between 20 000 and 25 000 revolutions per day (∼9–11 km day^–1) in the workload phase, but activity was not statistically different between groups. Foraging costs did not affect total wheel-running activity. Together with the results of earlier studies, our study provides evidence for limits on sustainable activity levels. Perrigo and Bronson (Perrigo and Bronson, 1983) observed a response qualitatively equivalent to ours in female CF-1 laboratory mice faced with foraging costs between 60 and 225 revs pellet^–1 (in the same size running wheels). From 60 to 135 revs pellet^–1, wheel-running activity gradually increased with rising foraging costs. Above 135 revspellet^–1,however, mice seemed to reach an activity limit (of just over 15 000 revs day^–1), regardless of further elevations in foraging costs. There may be sex or strain differences in running behavior. Independent of ad libitum activity levels (see Vaanholt et al., 2007), limits to total daily activity may constrain an animal's ability to cope with high foraging costs.

Our manipulation had clear effects on energy balance, which were most dramatic in female mice facing high foraging costs. The ratio of DEE:GEI approximates the proportion of consumed energy which would have been assimilated given a net energy balance of zero. The actual assimilation efficiency was probably higher [e.g. 79.1%(Hambly and Speakman, 2005)]. If we assume that animals actually absorbed 79.1% of gross energy intake, we can compare estimated energy balance or energy surplus (S) between groups. Based on this calculation, mice facing high foraging costs would have had no surplus energy or have been in a negative energy balance. By contrast, females facing low foraging costs or food restriction alone were at least `breaking even'. This conclusion is supported by the fact that the high-cost group lost body mass and showed almost totally depleted fat stores. By contrast, animals in the other groups maintained or increased body mass during the experiment. Similar to Vaanholt et al. (Vaanholt et al., 2007), we found that body composition was dramatically affected by high-cost foraging, and females facing high foraging costs had relatively smaller metabolic organs, muscles/skeleton, and skin/pelage. The effect on body mass was less pronounced in food-restricted animals.

Changes in body mass over time were associated with reductions in energy expenditure. Resting metabolic rate was positively correlated with whole body mass in all groups, and female mice faced with both low and high foraging costs had similar mass-specific metabolic rates in the workload phase. Overall, our RMR estimates of 1.12 kJ g^–1day^–1 for all groups in the workload phase were consistent with Vaanholt's (Vaanholt et al.,2007) estimates of 0.98 kJ g^–1day^–1 for male mice under workload conditions. We did not observe additional reductions in mass-specific energy requirements, however. Vaanholt et al. (Vaanholt et al.,2007) found that male mice on a high workload had reduced mass-specific RMR and DEE compared with baseline conditions. One possible explanation is that although our L group females were not food deprived, they were `trained' and thereby decreased mass-specific energy requirements. Our data also suggest that high foraging costs force mice to the limit of minimum energy expenditure.

In principle, reducing body mass could have been one component of an energy-saving strategy to prolong survival. Smaller muscles and metabolic organs require less energy, and the cost of transport is reduced with lower body mass (Rezende et al.,2006). Wiersma et al. (Wiersma et al., 2005) suggested that physiological changes made by European starlings (Sturnus vulgaris) facing a flight-for-food paradigm may be energy-saving `adjustments'. If body mass were not reduced in this context, birds would have been faced with high flight costs,theoretically reducing both food intake and foraging efficiency;alternatively, they would have needed to drastically elevate DEE. Other studies in birds have tested the hypothesis that loss of mass during nestling care can reduce the energetic cost of foraging, thereby potentially increasing reproductive output (e.g. Norbert,1981). Compensatory mechanisms reduce the impact of nutritional stress on free-living organisms [i.e. small mammals(King and Murphy, 1985)].

Under home cage conditions, minimum energy expenditure may actually have been lower than we estimated from respirometry measurements. Prior studies have suggested that animals facing high foraging costs can reduce resting body temperature (Perrigo and Bronson,1983; Vaanholt et al.,2007). In a pilot study using the same design as the present one,we found that rectal temperatures of H mice (35.6±0.3°C) were less than both L (37.2±0.2°C) and R (36.4±0.2°C) animals(K.A.S., unpublished data). These differences were small but statistically significant (N=8 per group). Measurements of core body temperature in the present study also suggest that animals facing high foraging costs employ daily heterothermy during periods of inactivity.

Female mice experiencing high foraging costs ceased to show estrous cyclicity. This response also occurred in some food-restricted females at a later time. Delaying or suppressing the estrous cycle during food restriction(Bronson and Marsteller, 1985)is a strategy for curtailing reproduction at an early stage, and Perrigo and Bronson (Perrigo and Bronson,1983) found that peripubertal laboratory mice on a workload schedule showed fewer ovulatory cycles. Wild-type female mice (Mus musculus and Peromyscus maniculatus) faced with high foraging costs were also less likely to become pregnant(Perrigo, 1987). It is interesting that in our study, females on a low workload continued to cycle. This demonstrates that foraging effort does not limit reproductive readiness,but that high foraging costs per reward have a negative influence on fitness prospects.

The results of our immunocompetence assay were intriguing. Female mice experiencing high foraging costs did not show compromised antibody production relative to animals on low foraging costs. The fact that inactive, food restricted mice actually showed higher antibody titers is more difficult to interpret (these data match pilot findings using the same experimental design;K.A.S. unpublished data). From an energetic perspective, one would have expected anti-KLH antibody titers to be lowest in H group females. Primary antibody production after KLH challenge has been shown to be metabolically costly [although it dose not induce fever or sickness and does not affect morphological traits such as body mass, adipose tissues, reproductive masses or lymphoid tissues in house mice (Demas et al., 1997)]. Food restriction also reduces immunological memory(Martin et al., 2007) and depresses acute phase response to LPS (a cell-mediated response) in hamsters(Conn et al., 1995) and mice(Matsuzaki et al., 2001). The fact that primary immune response in our experiment was not dependent on energy balance or body fat stores, therefore, is contrary to general findings on the energetics of immunity (reviewed by Demas, 2004). Overall, these results suggest that other factors may be responsible for the effects we observed.

Although voluntary exercise typically enhances immune response [i.e. delayed-type hypersensitivity (Bilbo and Nelson, 2004)], Moraska et al.(Moraska et al., 2000) showed that rats forced to run on a treadmill mount a compromised primary immune response to KLH. Although L and H females in our experiment could modulate their own activity levels, they had to run in order to receive food pellets. Thus, activity in this setting was not purely voluntary. As Vaanholt et al.(Vaanholt et al., 2007)showed, high wheel running activity under work-for-food conditions increases plasma corticosterone levels, which could be immunosuppressive. Such an effect is probably responsible for the lower immune response of our two foraging groups compared to the inactive, food-restricted group. Our results suggest that immune function may be affected differently in an experimental context linking activity with food intake (i.e. at different foraging costs), then under simple manipulations of energy balance or activity.

Our results show that energy balance is dramatically altered in small mammals faced with high foraging costs. These extreme changes in energy use,body composition, and other physiological parameters are not stimulated to the same extent by food restriction alone. Manipulations of foraging costs may also have long-term effects. Under poor environmental conditions, animals may obtain less food, while at the same time reallocating energy to activity at the expense of other processes. This may eventually affect rates of senescence(Speakman et al., 2002),possibly yielding results different from studies manipulating food intake alone (i.e. caloric restriction). Ultimately, therefore, studies manipulating the yield of foraging effort can have a broad impact. Research on how animals respond to variation in environmental conditions not only demonstrates ecologically relevant strategies for coping with food scarcity, but it may also lead to a better understanding of physiological processes underlying life history trade-offs.

LIST OF ABBREVIATIONS

 
• ADMR

  average daily metabolic rate

 
• DEE

  daily energy expenditure

 
• DLW

  doubly labeled water

 
• GEI

  gross energy intake

 
• KLH

  keyhole limpet hemocyanin

 
• RMR

  resting metabolic rate

 
• RQ

  respiratory quotient

 
• RWA

  running wheel activity

 
• S

  surplus energy

We thank Paulien Barf and Gonda de Vries for collecting pilot data, Sarah Johnston and Anne de Ruiter for advice on estrous cycle determination, and Berber de Jong and Jan-Albert Manenschijn for help with practical work. Gerard Overkamp provided expert technical assistance, creating and maintaining the workload setup and the respirometer. Harro Meijer provided useful discussions regarding analysis of the DLW samples. Simon Verhulst and John Speakman provided comments which greatly improved the quality of this manuscript. Other colleagues from the Behavioral Biology Department also assisted with practical work and made many helpful suggestions. K.A.S. and L.M.V. were supported by grants from the University of Groningen, S.D. by EC grant LSHM-CT-2006-018741(EUCLOCK).