Both appropriate metabolic rates and sufficient immune function are essential for survival. Consequently, eco-immunologists have hypothesized that animals may experience trade-offs between metabolic rates and immune function. Previous work has focused on how basal metabolic rate (BMR) may trade-off with immune function, but maximal metabolic rate (MMR), the upper limit to aerobic activity, might also trade-off with immune function. We used mice artificially selected for high mass-independent MMR to test for trade-offs with immune function. We assessed (i) innate immune function by quantifying cytokine production in response to injection with lipopolysaccharide and (ii) adaptive immune function by measuring antibody production in response to injection with keyhole limpet haemocyanin. Selection for high mass-independent MMR suppressed innate immune function, but not adaptive immune function. However, analyses at the individual level also indicate a negative correlation between MMR and adaptive immune function. By contrast BMR did not affect immune function. Evolutionarily, natural selection may favour increasing MMR to enhance aerobic performance and endurance, but the benefits of high MMR may be offset by impaired immune function. This result could be important in understanding the selective factors acting on the evolution of metabolic rates.
Both metabolism and immune function are key factors for survival in nature. If metabolic rates are too high, animals may exhaust energy reserves and thus decrease their survival . If metabolic rates are too low, animals may be unable to fuel essential physiological activities [2,3]. Accordingly, which level of metabolic rate (e.g. low, intermediate or high) is best may depend on ecological conditions, sex and body size [2–7]. Likewise, the optimal intensity of immune function may depend on trade-offs [8,9]. Vigorous immune function might be best for fending off disease and parasites, but it might also lead to damaging inflammatory responses or wasting of resources that would be better invested elsewhere [10,11]. On the other hand, if immune function is not vigorous enough, animals might succumb to disease and parasites [10,12–14]. While individual traits, such as a particular aspect of immune function or a particular measure of metabolic rate, may affect survival, ultimately survival requires that the combination of traits an animal has are sufficient to cope with the ecological circumstances it encounters.
Metabolic rates and immune function might trade-off for at least two reasons. First, both systems share some of the same signalling molecules (e.g. corticosterone) [10,15–18]. Second, both traits may compete for limited resources, such as energy or other nutrients [17,19]. Some studies report that immune function is energetically costly [20–23], but others fail to find a significant energetic cost [24–26]. The ambiguity of these findings might be explained, at least in part, by the complexity of the immune system, and because some immune functions may be more costly than others [27–29]. Likewise, there is more than one measure of metabolic rate, and some aspects of immune function may trade-off with some measures of metabolic rates but not with others.
Most work on metabolic rate and immune function trade-offs has focused on how immune function correlates with basal metabolic rate (BMR) or field metabolic rate (FMR) [16,30]. BMR is the minimum energy required for a resting post-absorptive animal that is not growing or reproducing and that is in a thermoneutral environment . FMR is the energy used for aerobic metabolism by free-living animals in the wild, and it includes the costs of BMR, locomotion, digestion, thermoregulation, reproduction and growth . A third important measure of aerobic metabolic performance is maximal metabolic rate (MMR). MMR sets the upper limit for sustained, vigorous activity and sustained thermogenesis (i.e. heat production). In our study, MMR was activity induced by treadmill exercise. A high MMR is indicative of a high capacity to mobilize and use energy. If upregulated immune responses are energetically expensive [10,20–23], then a high MMR might be associated with high immune function. Alternatively, high MMR might be associated with low immune function if immune response and MMR compete for limited, energetic resources. In this study, we expand beyond what has previously been investigated by looking for associations between MMR and immune function. In addition, we tested for association between BMR and immune function.
The immune response is complex and multifaceted [27,29,33]. Recent literature has hypothesized that immune response should vary with the energetic demands of an individual, such that individuals coping with higher energy demands should favour immune responses that are less costly. Thus, acute phase inflammatory responses (i.e. energetically expensive immune responses; hereafter inflammatory response) should be downregulated in individuals with higher demands on their resources, and these individuals should invest more in non-inflammatory responses, such as antibody production [27,28,34]. MMR is often associated with energetically demanding activities, such as defending territories, foraging, escaping from predation and running endurance [35–37]. Thus, if MMR is a proxy for an energetically demanding lifestyle and inflammatory responses are downregulated in individuals with energetically demanding lifestyles, then inflammatory response should be greatest in individuals with low MMRs, and antibody production should be highest in individuals with high MMRs [27,28,34].
Herein, we tested whether two measures of metabolic rate, MMR and BMR, trade-off with either adaptive or innate immune function. Because previous work in eco-immunology has mostly ignored MMR (however, see Schwanz  who investigated how parasite infections affect metabolic capacity), we focused our study on trade-offs with MMR. Artificial selection experiments are a powerful tool for studying trade-offs . Accordingly, to test the nature of the relationship between MMR and immune function, we manipulated metabolic rates genetically via artificial selection and then tested for effects on both inflammatory response and antibody production.
2. Material and methods
(a) Study animal and selection treatments
We studied laboratory house mice derived from an artificial selection experiment on metabolic rates. Mice from the HS/IBG (heterogeneous stock/Institute of Behavioral Genetics) stock were obtained from the University of Colorado, Boulder, CO, USA and assigned to one of three selection treatments. The treatments were randomly bred control (CONT), directional selection for high mass-independent MMR (high-MMR) and antagonistic selection (ANTAG) for simultaneously high mass-independent MMR and low mass-independent BMR. In each generation, there were 12 lines of mice divided into four blocks of mice. Each block had three lines of mice (one CONT line, one high-MMR line and one ANTAG line); hence each selection treatment was replicated four times. A detailed description of the selection protocol is provided in the electronic supplementary material, M1. Briefly, mice were weaned at 21 days of age and housed up to five per cage. Cages contained a layer of corncob bedding and paper towels for nesting material. Food and water were available ad libitum. Mice were maintained on a 12 L : 12 D photoperiod and kept at ambient building temperatures (21.0°C–25.5°C).
We measured MMR once by forced exercise on a motorized treadmill with an incremental step test [40,41]. Each treadmill was housed within an open-flow respirometry chamber and had a grid that provided a shock to motivate the mouse to run. Each mouse was run until it either refused to run or showed no increase in oxygen consumption with increased treadmill speed. MMR was defined as metabolic rate during the highest 1-min period of oxygen consumption during forced exercise. Oxygen consumption during treadmill measurements was calculated using the instantaneous correction for chamber washout . BMR was measured using open-flow respirometry. Over a period of 6 h, each mouse was monitored for 16 min per hour. BMR was defined as the lowest, consecutive 5 min of metabolism during any of the six 16-min periods. Details of the metabolic rate measurements are provided by Wone et al. .
Typically, metabolic rates correlate strongly with mass, hence to minimize confounding metabolic effects with mass effects, we selected on the mass-independent metabolism (i.e. on residuals from regressions of metabolism on body mass). Within each line, we avoided sibling pairs, and we generally created 13 pairings to attempt to ensure 10 successful litters from each line. For the CONT lines, breeders were chosen randomly (i.e. no artificial selection). For the high-MMR lines, we bred the 13 females and 13 males with the highest mass-independent MMR. For the ANTAG lines, we simultaneously selected for high (+) residual MMR and low (−) residual BMR. To do this, we calculated the mass-independent BMR and mass-independent MMR of each mouse and then the cross product of the residuals. The mice were ranked based on the cross product with the most negative cross product ranked highest. Because mice with large negative residual MMR may not have run at their maximal capacity, we only selected breeders with a positive residual for MMR.
For the experiments contained herein, we used female mice from generation eight of the selection experiment. In general, high-MMR mice had elevated mass-independent MMR when compared with the randomly bred CONT mice, and ANTAG lines had a mass-independent MMR between the high-MMR and CONT lines (J. P. Hayes 2009, unpublished data). Overall, there were no significant differences in mass-independent BMRs among the selection treatments for all of the mice from generation eight (J. P. Hayes 2009, unpublished data).
We used only virgin, female mice to control for differences in immune function caused by sex hormones and to eliminate possible effects of pre- versus post-reproductive conditions. Because of the complicated design and duration of the experiments presented herein, we did not control for oestrus cycle in our experimental design.
(b) Inflammatory response
To induce an innate immune response, we injected mice with lipopolysaccharide (LPS), an endotoxin found in the outer membrane of Gram-negative bacteria. LPS mimics a bacterial infection and causes an innate immune response . In addition, LPS elicits the physiological indicators associated with an inflammatory response, such as elevated circulating pro-inflammatory cytokines .
Mice used to assess inflammatory response were 144 days old on average (range: 119–168 days old). Four mice from each selection treatment of each block were assigned randomly to the LPS and SHAM treatments (96 mice total = 4 mice × 3 selection treatments × 4 blocks × 2 immune treatments). Mice in the LPS treatment were injected intraperitoneally with 20 μg of LPS (from Escherichia coli 0111:B4, product no. L4391, Sigma-Aldrich, St. Louis, MO, USA) in 0.1 ml phosphate-buffered saline (PBS, dose determined in a preliminary study). SHAM treatment mice were injected with an equivalent volume (i.e. 0.1 ml) of PBS. Blood samples (approx. 70 μl from the tip of the tail) were collected after approximately 2 h to correspond with expected peak production of cytokines . Blood serum was centrifuged and frozen at −80°C and later analysed.
We used Luminex technology (Luminex 200, Invitrogen Corporation, Carlsbad, CA, USA) to quantify blood serum levels of four cytokines involved in initiating the immune response: tumour necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), interleukin-1β (IL-1β) and granulocyte macrophage colony-stimulating factor (GM-CSF)  (details in electronic supplementary material, M2).
(c) Humoral response
Mice used to assess humoral response were 168.5 days old on average (range: 171–211 days old) at the start of the experiment. Four mice from each selection treatment of each block were assigned randomly to the KLH and SHAM treatments. Mice in the KLH treatment were injected intraperitoneally with 100 μg of KLH (Sigma-Aldrich catalogue no. H7017, Sigma-Aldrich, St. Louis, MO, USA) in 0.1 ml PBS. SHAM treatment mice were injected with an equivalent volume (i.e. 0.1 ml) of PBS. Mice received primary inoculation on day 0 and secondary, booster inoculation on day 21 of the experiment. Blood samples (approx. 70 μl of blood from the tip of the tail) were taken 14 and 15 days after the booster inoculations (i.e. 35 and 36 days after primary injection) to correspond with peak antibody production [20,47]. We focused on the secondary response because it is typically stronger then the primary response , and we would most probably see differences when the humoral response was greatest. Blood serum was frozen at −80°C and saved for later analysis. Enzyme linked immunosorbent assay (ELISA) was used to quantify antibody levels in blood serum samples  (details in electronic supplementary material, M3).
(d) Data processing and statistical analyses
The response variables to quantify inflammatory response were TNF-α, IL-6 and GM-CSF. Five mice (three SHAM and two LPS) were removed from all of the cytokine analyses because they did not have large enough blood serum samples for the Luminex assay. While we also tried to quantify concentrations of IL-1β, the concentrations for all but 19 (of 91) of our blood samples fell below the concentrations for the standardized curve. This suggests that IL-1β were not concentrated enough in our diluted serum sample to accurately quantify so we do not include IL-1β in our statistical analyses or our discussion.
Humoral response was quantified with the average optical density of the diluted serum samples. Average optical density is a measure of number of circulating antibodies . Owing to an error while injecting one set of replicate lines during the booster inoculation and one mouse dying of natural causes, the final analysis with KLH and SHAM mice was of 71 mice. One mouse from the SHAM treatment was removed from the antibody analyses because there was not enough serum to run an ELISA.
There are many ways to analyse these data and each analysis addresses a different question. First, we looked for differences between the SHAM and immune-challenged (LPS or KLH) mice to verify that the mice had been injected properly and to determine if there was a significant effect of the immune treatment. After verifying the presence of an immune treatment effect, we performed a second set of analyses to look for differences among selection treatment means for immune response using only the immune-challenged mice (KLH or LPS mice). We opted to address this question with only immune-challenged mice because the power to detect the interaction term is low when the unit of replication is line—not individual—and the additional statistical noise from including the SHAM mice may have been problematic .
Finally, mice within selection treatments exhibited considerable variation in metabolic rates and immune function. Consequently, if selection treatment was not supported as a significant effect, we tested for a trade-off at the individual level by determining how immune response varied with metabolic rates across all of the mice. Individual-level analyses with MMR and BMR were conducted with and without body mass in the model.
Our immune results at the level of selection treatment cannot be interpreted without information about the MMR and BMR for the mice used in the immune treatments. Thus, we looked for differences among selection treatment means for mass-independent MMR and mass-independent BMR. These results were combined with the immune results from the selection-level analyses to determine if the pattern suggested a correlation between the metabolic rates and immune function.
Owing to constraints imposed by the overall selection experiment, we could not measure all mice at the same age, and mice differed in age by as much as 49 days for the innate response experiment and by as much as 40 days for humoral response experiments. Age was included as a covariate in preliminary analyses but it was not significant so it was dropped from subsequent analyses. Furthermore, within each block for each experiment, the mice were close in age (greatest difference in ages within a block was 6 days); thus, by including block as a potential random effect in our analyses, we could statistically control for the variation in age.
The raw data for all response variables exhibited large differences in variances between the SHAM and LPS or KLH-treated mice and among selection treatments. Accordingly, the data were analysed with generalized least-squares (GLS) models. GLS models, as opposed to general linear models, allowed us to relax the assumption that the variances in response for all of the treatment groups were similar . Hence, we fitted variance covariates and/or random effects, resulting in mixed effects models, to accommodate the heterogeneity in variance and the nested data structure. The values of IL-6 were square root transformed to help further improve homogeneity in variance. When MMR and BMR were response variables, the predictor variable of mass was transformed so it was centred around 0 but not scaled.
For each response variable, models were first fit with the ‘beyond optimal’ combination of fixed effects (i.e. the full model of potential fixed effects), and the optimal random effect structure was determined by likelihood ratio tests (LRTs) of nested models and visual examination of diagnostic plots of standardized residuals. After finding the optimal random effects structure, appropriate fixed effects were selected by sequentially dropping model terms and comparing nested models with LRT . Significance of pairwise contrasts between factor levels was determined post-hoc with Tukey all-pair contrasts. All analyses were performed in R v. 2.14.1 , with the packages nlme , contrast  and multcomp .
Details of ‘best fit’ models for each response variable are provided in the electronic supplementary material, M4. We followed the GLS approach recommended by Zuur et al. . For brevity, we illustrate the approach with a single example—the effect of selection for MMR on antibody concentration in SHAM and KLH treatment groups (for annotated code see the electronic supplementary material, M5). First, we fitted a restricted maximum-likelihood (REML) model containing all potential fixed effects (KLH treatment, selection treatment and the interaction KLH treatment by selection treatment) but with no random effects and without allowing for heterogeneous variances. We extracted the standardized residuals: (i) to check for heterogeneity in variances among the potential grouping structures (line, KLH treatment, selection treatment, block and combinations of these variables); and (ii) to inform the building of random effects models. Then we built models allowing for differences in variances between grouping structures (i.e. variance covariates) and differences in variances of random effects, and then compared the fit of alternative models. The model with the best-fitting random structure included different variance covariates for both KLH treatment and selection treatment.
Then we determined if KLH treatment and the interaction between KLH treatment and selection treatment were significant fixed effects. To do this, we used maximum-likelihood (ML), not REML, to refit the GLS model with the best random structure and which included the full set of fixed effects: KLH treatment, selection treatment and the interaction of KLH treatment with selection treatment. We then fit a reduced model with a fixed effect of selection treatment but without the fixed effects for KLH treatment or the interaction of KLH treatment with selection treatment. These two models were then compared using an LRT to determine if the KLH treatment and the KLH treatment–selection treatment interaction contributed significantly to model fit. Finally, the model with the best set of fixed effects (in this case, the reduced model) was then refit using REML to obtain parameter estimates. REML was used to compare random structures and generate parameter estimates because this model fitting method results in unbiased variance and other parameter estimates . In comparison, ML was used to compare models differing by their fixed effects because the restricted log likelihood estimates from REML are not comparable in the same manner as log likelihoods when fixed effects are changed [50,55].
(a) Inflammatory response
For all inflammatory response variables (TNF-α, GM-CSF and IL-6), LPS mice had higher levels than SHAM mice (see the electronic supplementary material, figure S1a–c). All LPS mice also exhibited sickness behaviour , although this was not assessed quantitatively.
When the mice injected with LPS were considered, selection treatment had a significant effect on GM-CSF (LRT between model with selection treatment as fixed effect and null model, LR1 = 8.11, p = 0.017; figure 1a), TNF-α (LR1 = 9.45, p = 0.009; figure 1b) and IL-6 (LR1 = 10.24, p = 0.006; figure 1c; electronic supplementary material, table S1). GM-CSF, TNF-α and IL-6 were highest in CONT mice and lowest in high-MMR mice (Tukey contrasts; z = −2.99, p = 0.008; z = −3.15, p = 0.004; z = −3.25, p = 0.003; electronic supplementary material, table S2). GMC-SF and TNF-α levels were also lower for ANTAG mice than for CONT mice (z = −2.57, p = 0.027; z = −2.45, p = 0.035; electronic supplementary material, table S2).
In the subset of mice used for LPS studies, selection treatment affected both mass-independent BMR (LR1 = 12.31, p = 0.002; figure 1d) and mass-independent MMR (LR1 = 12.96, p = 0.002; figure 1e, electronic supplementary material, table S1). Recall that mass-independent models include mass as a covariate in the model. High-MMR mice had higher mass-independent BMR than either CONT (z = 2.88, p = 0.011) or ANTAG mice (z = 4.83, p < 0.001; electronic supplementary material, table S2). Both high-MMR and ANTAG mice had higher mass-independent MMR than CONT mice (z = 4.10, p < 0.001; z = 2.55, p = 0.029; electronic supplementary material, table S2).
(b) Humoral response
KLH mice had higher antibody levels than SHAM mice (see the electronic supplementary material, figure S1d). When only the mice injected with KLH were considered, selection treatment was not a significant predictor of antibody production (LRT between model with mass, MMR and selection treatment and model with only mass and MMR; LR = 4.70, p = 0.096). However, both mass and metabolic rate were associated with antibody concentrations (table 1 and figure 2). Antibody concentration was negatively associated with MMR (β = −0.22 ± 0.08) and positively associated with mass (β = 0.09 ± 0.02). The relationship between antibody concentration and MMR only held when mass was included in the model (table 1). There was no support for a relationship between antibody production and BMR, regardless of whether mass was included in the model (table 1).
Mass-independent BMR did not significantly differ among selection treatments in the subset of mice from generation eight used in the KLH treatment (LR1 = 2.26, p = 0.323), but mass-independent MMR did significantly differ among selection treatments in the KLH mice (LR1 = 21.181, p < 0.001). High-MMR mice had higher mass-independent MMR than ANTAG (z = 3.09, p = 0.006) or CONT mice (z = 4.79, p < 0.001).
Whether increased immune function trades off with metabolic rate is a complex question because there are many aspects of immune function [27,29,33] and many measures of metabolic rate. Trade-offs, if they exist, may be general and obvious, or they may be subtle and specific to particular measures of immune function or metabolic rate. Because MMR sets the upper limit to sustained aerobic processes, we hypothesized that trade-offs with immune function might be most easily detectable in animals selected for greater maximal capacities. Artificial selection for increased mass-independent MMR resulted in trade-offs with innate immune function. Similarly, at the individual level MMR was negatively associated with antibody concentrations (β = −0.23 ± 0.10), but only when mass was included in the model. By contrast, BMR, a measure of the lower limit of metabolic rate, did not trade-off with either adaptive or innate immune function. Our study suggests that there is a trade-off between evolving a high mass-independent MMR and evolving a robust immune response. Indeed, trade-offs with immune function may be one factor opposing the evolution of high MMR. Hence, it may be fruitful for those interested in trade-offs between immune function and energy metabolism to expand their investigations beyond BMR and FMR and to consider the role that may be played by MMR.
The LPS experiment supports the idea that there is trade- off between mass-independent MMR and innate immune function and that selection for high mass-independent MMR led to impaired innate immune function. An intriguing hypothesis proposed by evolutionary physiologists is that in vertebrates MMR and BMR may be strongly linked [37,43,57–60]. While some studies are beginning to cast doubt on that hypothesis [36,61,62], selection for high MMR may have led to correlated responses in BMR which in turn led to the reduced innate immune function. Our data do not support this idea. First, for all three cytokines the concentrations were highest in CONT mice and lowest in high MMR (figure 1; electronic supplementary material, table S3 and S4). This matches the variation in mass-independent MMR but not the variation in mass-independent BMR. Second, we asked whether significant differences in metabolic rates corresponded with significant differences in cytokine levels. In total, this allows nine comparisons for each measure of metabolic rate, because there are three pairwise comparisons between lines (high MMR versus ANTAG, high MMR versus CONT, and ANTAG versus CONT) and three cytokines. For mass-independent MMR, eight out of nine comparisons were concordant, but one was not. The one comparison that was not was that ANTAG and CONT had significantly different mass-independent MMR but their IL-6 concentrations were not significantly different. For mass-independent BMR, only four of nine comparisons were concordant. For example, high-MMR mice and ANTAG have significantly different mass-independent BMR but their concentrations of cytokine levels were not significantly different for GM-CSF, TNF-α or IL-6. Hence, we conclude that the reduction in innate immune response associated with selection for high mass-independent MMR was not mediated via BMR.
In contrast to the results for innate immunity, concentrations of circulating antibodies (i.e. optical density) did not vary significantly (p = 0.096) across selection treatments. This result suggests that selection did not affect adaptive immune function. Nonetheless, the p-value for the test of a selection effect is suggestive so we pursued the analysis further by looking at individual-level variation. This individual-level analysis takes into account the substantial individual variation within selection lines.
The individual-level analyses support the hypothesis that a trade-off exists between MMR and adaptive immune function. Circulating antibodies were significantly negatively correlated with MMR (p = 0.023; figure 2), but only when mass was also included in the model. The inclusion of body mass is perhaps not surprising because body mass is an important predictor of many ecological and life-history traits [63–66], including MMR and some immune parameters [67–69]. There was no evidence that immune function was related to BMR alone or when an adjustment for mass was also included in the model (table 1).
Why might increased MMR inhibit immune function? MMR is the upper limit of an individual's ability to use energy through aerobic pathways. Costly immune function is often suppressed during stressful or energetically expensive activities, and resources saved by immunosuppression may be adaptively reallocated to other costly activities . The mechanism by which this trade-off occurs is unclear, however signalling molecules shared across physiological systems are a possibility [10,15,18,19,71]. For example, corticosterone is a hormone involved in both energy metabolism and immune function. At chronically high levels, corticosterone suppresses immune function and causes an increase in blood glucose [72,73]. Our selection protocol required mice to run on a treadmill, and circulating corticosterone increases during activity [74,75]. Additionally, mice selected for high voluntary wheel running have increased baseline corticosterone, presumably because the elevated corticosterone mobilized energy needed for wheel running [76,77]. Thus, we may have inadvertently selected for increased baseline corticosterone in the mice selected for high MMR. If so, corticosterone might link increased MMR with depressed immune function.
Exercise physiology also provides another context within which to examine our results. Endurance training increases mass-specific MMR in healthy humans  and mass-independent MMR in mice . Exercise also decreases inflammatory biomarkers, such as TNF-α [79,80]. For example, single bouts of moderate exercise can suppress endotoxin-stimulated increases in TNF-α in healthy humans , and exercise normalizes TNF-α expression in knockout mice that typically show increased TNF-α levels . We ran mice on treadmills, a common procedure for simulating acute bouts of exercise. If by selecting for high MMR, we also selected for an ‘exercise’ phenotype, this may explain the reduced inflammatory markers associated with LPS injections.
Most previous work investigating how metabolic rates affect immune function has focused on BMR, which has sometimes been found to correlate with immune function [16,30,83]. By contrast, our study suggests that MMR, not BMR, trades off with some aspects of immune function. Trade-offs may vary across species and environments, and as in any study, our results might not be representative of a general pattern for all populations or species. We studied laboratory mice that did not face the same constraints and selective pressures as wild mice, particularly when it comes to encounter rates with diseases . Furthermore, wild mice tend to have greater and more varied immune responses then laboratory mice . Thus, because the selected mice presumably underwent minimal selection for immune function and may have minimal immune response relative to wild mice, the trade-offs that we found may not also be found in natural populations.
Finally, as is typical in many eco-immunology studies, we used molecular markers associated with the adaptive and innate responses as our response variables. These markers are parts of the cascades of physiological responses leading to complex immune responses. While markers of immune function sometimes correlate directly with fitness [14,67], most studies do not test fitness directly and extrapolate to fitness differences suggested by physiological trade-offs [23,86]. Thus, it unclear whether the physiological trade-off we observed between MMR and immune function would lead to differences in fitness in a natural environment.
In summary, innate immune function was suppressed by artificial selection on mass-independent MMR. In addition, at the individual level, MMR was negatively associated with adaptive immune function when mass was included in the model but not in models without mass. BMR was not associated with immune function. These results are important for at least two reasons. First, MMR may be a key to understanding potential trade-offs between metabolic rates and immune function. Hence, as physiologists and ecologists explore the physiological trade-offs underlying life-history variation, more attention should be focused on MMR, the upper limit to aerobic performance. Second, the selective factors leading to the evolution of metabolic rates remain poorly understood , but the positive benefits of increased MMR for enhancing exercise performance may be offset by negative effects on immune function. In conclusion, our results indicate that eco-immunologists looking at trade-offs should not only consider multiple aspects of immune function, they should also consider different measures of metabolic rate.
All procedures conducted in this study are in accordance with the UNR Institutional Animal Care and Use Committee and US laws.
We thank M. Labocha, M. Sears and many undergraduates in the Hayes laboratory for their help with the selection experiment and V. Lombardi for help with immune techniques. This manuscript was significantly improved by the comments of L. Martin and several anonymous reviewers. This study was supported by NSF IOS-0344994 to J.P.H. and a research grant from the Ecology, Evolution and Conservation Biology Program at University to Nevada, Reno to C.J.D.
- Received November 7, 2012.
- Accepted December 11, 2012.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.