Chronic hypoxia stimulates an enhanced response to immune challenge without evidence of an energetic tradeoff

Ecoimmunology is a growing field concerned with how variation in immune function affects organisms in different environmental and ecological contexts. Of particular interest are the potential indirect effects of the immune response on fitness via perceived tradeoffs with other physiological functions or life history traits (Lee, 2006; Lochmiller and Deerenberg, 2000; Schmid-Hempel, 2003; Schmid-Hempel and Ebert, 2003; Svensson, 1997). Metabolic energy is one of the factors that are often hypothesized to underlie physiological tradeoffs (Burness et al., 2010; Sheldon and Verhulst, 1996; Zysling et al., 2009), because if the immune system is energetically expensive to maintain and deploy, it may divert energy away from other physiological processes. Attempts to demonstrate energetic costs of the immune system have yielded conflicting results (Demas et al., 1997; Mendes et al., 2006; Nilsson et al., 2007; Norris and Evans, 2000; Raberg et al., 1998; Scantlebury et al., 2007; Schubert et al., 2008; Svensson et al., 1998; Zysling et al., 2009), so factors other than energetic costs may mediate the relationships between the immune system and other physiological processes.

An ecologically relevant method for examining physiological tradeoffs with the immune system is to compare the responses of environmentally stressed and unstressed animals (Burness et al., 2010; Ksiazek et al., 2003; Sandland and Minchella, 2003; Svensson et al., 1998). Indeed, there is much research demonstrating how thermal, nutritional and psychological stress all lead to reduced immune function (Buehler et al., 2009; Cichon et al., 2002; Ksiazek et al., 2003; Kusumoto, 2009; Svensson et al., 1998). Stress responses may directly influence the immune system via endocrine mediators (Dhabhar et al., 1995; Martin, 2009). Endocrine mediators might also be the proximate mechanisms driving energetic tradeoffs (Bourgeon et al., 2009; Demas et al., 2004; Martin et al., 2008b). Nonetheless, factors other than energetics are likely to be important in the evolution of immune function and its relationship to other physiological systems (Ricklefs and Wikelski, 2002). For example, the immune system and other physiological processes may share gene transcriptional networks, in which case transcriptional responses to abiotic stressors, such as hypoxia, might also be important in mediating immune system effects.

Chronic, systemic hypoxia is commonly associated with high altitude environments. In mammals, ascending to high altitudes elicits a variety of physiological responses to maintain oxygen homeostasis, including polycythemia (increased hematocrit), increased production of 2,3 bisphosphoglycerate, pulmonary vasoconstriction, increased lung and liver mass, increased left ventricular mass, increased tidal volume and ventilation rate, increased capillary density, and anorexia with weight loss (Appenzeller et al., 2003; Beall, 2001; Bozzini et al., 2005; Hammond et al., 1999; Powell et al., 1998; Ward et al., 2000; Yu et al., 1999). Many of these responses have potential energetic costs, especially in terms of the anorexia that hypoxia induces.

How hypoxia affects metabolic rates is a complex topic (Hayes, 1989a; Hayes, 1989b; Russell et al., 2008). Clearly, severe hypoxia reduces maximal metabolic rates compared with normoxia (Rosenmann and Morrison, 1975; Wehrlin and Hallen, 2006; Young et al., 1996). However, in this study we were not concerned with direct constraints on the upper limit to aerobic performance but rather with the potential effects on basal metabolic rates, which indicate the minimal cost of living (including basal costs of immune function) in a particular environment. During exposure of up to 2 weeks of hypoxia, some small animals downregulate their body temperature and reduce their basal metabolic rates (Gautier, 1996; Hochachka et al., 1996; Semenza et al., 1994; Steiner and Branco, 2002). But other studies suggest that hypoxia leads to increased basal metabolic rate in larger mammals or to increased rates of heat production in fasting animals native to hypoxic environments (Butterfield et al., 1992; Han et al., 2003; Hayes, 1989a). A further complication is that responses to acute and chronic hypoxia are potentially quite different, and although acute hypoxic responses are relatively well studied, responses to chronic hypoxia are much less well known. In any case, the fact that hypoxia and high altitude often are associated with changes in basal metabolic rate suggests that hypoxia is likely to be an environmental stressor that impacts energy metabolism.

If chronic hypoxia challenges small mammals energetically, then hypoxia might negatively affect immune function. In particular, one reasonable hypothesis is that under chronic, systemic hypoxia, the more expensive innate inflammatory responses might be inhibited and the less expensive adaptive humoral immune response might be maintained (Lee, 2006; Martin et al., 2006; Tieleman et al., 2005). However, there is growing evidence that this may not be the case. For instance, several studies have reported that humans traveling to high altitudes have increased circulating leukocytes and decreased circulating CD4⁺ T-cells (Chohan and Singh, 1979; Chohan et al., 1975; Facco et al., 2005). These observations are consistent with the known effects of acute hypoxia on the immune system at localized tissue levels, such as during tissue injury, inflammatory diseases and cancer (Brouwer et al., 2009; Frantz et al., 2005; Lewis and Murdoch, 2005; Taylor, 2008; Walshe and D'Amore, 2008). In these cases, localized hypoxia leads to the recruitment and increased survival of macrophages, neutrophils and other granulocytes (Bosco et al., 2006; Burke et al., 2003; Murdoch et al., 2005; Thake et al., 2004; Walmsley et al., 2005b). Conversely, acute hypoxia appears to have a regulatory, anti-inflammatory effect on mature CD4⁺ T-cells (Conforti et al., 2003).

Immune responses to hypoxic stress may be the result of regulatory effects of hypoxia on the immune system. Most of the hypoxia-induced immune responses described above can be tied to regulation by either hypoxia inducible factor (HIF) and/or nuclear factor kappa beta (NFκB). Both of these transcription factors have been identified as hypoxia sensitive, proinflammatory mediators (Michiels et al., 2002). Although HIF was initially identified as a key modulator of most physiological responses to hypoxia (Greijer et al., 2005; Semenza, 2000; Yu et al., 1999), it is now appreciated as an inflammatory mediator via its proinflammatory effect on leukocytes and its interaction with NFκB (Cramer et al., 2003; Gale and Maxwell, 2010; Hellwig-Burgel et al., 2005; Zarember and Malech, 2005; Zinkernagel et al., 2007). Conversely, NFκB is best known as the central transcription factor responsible for activating the acute phase inflammatory response via induction of key inflammatory cytokines, such as tumor necrosis factor alpha (TNF-α). Recently, NFκB has also been shown to be sensitive to hypoxic stimulation, independent of HIF (Chandel et al., 2000; Cummins et al., 2007; Rius et al., 2008; Taylor, 2008). Thus, it appears that, through the activation of HIF and NFκB, acute hypoxia may be a key signal to inflammatory responses (Taylor, 2008), and these immune responses may be part of the suite of physiological responses to hypoxia. Although the significance of these transcriptional networks to chronic, systemic hypoxia is unclear, such co-regulatory relationships deserve consideration when looking at potential physiological tradeoffs in ecological immunology.

The energetic costs of the immune system and of responding to chronic hypoxia deserve further investigation. Although there is potential for these processes to compete for limited energetic resources, these interactions are trivial if either process does not demand a significant amount of metabolic energy. Indeed, mounting an immune response may not be as energetically expensive as proposed - a topic that has been the subject of debate (Bourgeon et al., 2009; Klasing, 1998; Lochmiller and Deerenberg, 2000; Martin et al., 2008b).

To better understand how mammals may be affected by high altitude hypoxia, we investigated whether acclimating to chronic hypoxia alone alters the circulating immune system, and the effects of environmental hypoxia and immune challenges on metabolic rates. Specifically, we tested whether: (1) chronic, systemic hypoxia affects the makeup of the peripheral immune system and/or the response of the immune system to humoral and inflammatory challenges, (2) immune responses have a significant energetic cost that is reflected in resting metabolic rate, and (3) energetic costs of immune function are affected by hypoxic stress.

Three experiments were performed using the same experimental apparatus on different mice. In the first experiment (later referred to as the time-series experiment), we measured circulating white blood cells in C57BL/6 mice acclimating to hypoxia. In the second and third experiments, we measured the energy use of C57BL/6 mice experiencing hypoxia, one of two immune challenges, or both hypoxia and an immune challenge simultaneously. In all instances, mice were housed for the duration of the experiments at one of two simulated altitudes. Two chambers simulated sea level by adding appropriate levels of oxygen to Reno ambient air [barometric pressure=86.1 kPa, oxygen partial pressure (P_O2)=18 pKa]. The final oxygen concentration of these chambers was approximately 24%, or a P_O2 of 20.7 kPa. Two other chambers simulated the approximate hypoxic conditions found above 4000 m. The hypoxic chambers received Reno air mixed with nitrogen to achieve appropriate degrees of hypoxia. For the first study measuring white blood cells over time, the chambers were maintained at roughly 12% oxygen, or a P_O2 of 10.4 kPa. If one adjusts for the dilutive effects of water vapor and the effect of the barometric pressure changes over altitude, the simulated altitude more closely approximates what mice would encounter at approximately 4500 m. For the second experiment, we adjusted the P_O2 to 12.5% oxygen to match the level that was to be supplied to the respirometry chambers used to measure metabolic rates. This simulated altitude was closer to 4250 m. Details of methods for monitoring appropriate levels of oxygen and delivering oxygen to the chambers are described in Baze et al. (Baze et al., 2010). Each chamber housed up to 12 mice individually in 30×8×12 cm standard rodent cages. Mice were provided with ad libitum food and water, and kept on a 12 h:12 h light:dark cycle. Upon arrival in Reno from Charles River Laboratories (Hollister, CA, USA), the C57BL/6 female mice were allowed to acclimate to the chambers for 1 week. All mice were 14 weeks of age at the start of the experiment. All protocols were approved by the University of Nevada, Reno Institutional Animal Care and Use Committee.

To test how the immune system responded to systemic hypoxia, blood samples were taken from mice housed in hypoxic or normoxic environments on days 5, 15, 26 and 36. The total number of mice was 20, with each treatment represented by 10 mice, five in each chamber. On each of the four sampling days, we collected 140 μl of blood from the mice to track the changes in immunophenotype over the course of hypoxic exposure. One 70 μl microhematocrit tube was used to measure hematocrit. The second 70 μl tube was analyzed with flow cytometry to detect changes in the white blood cell composition of the blood. Body masses of all animals were measured at each time point. Food consumption was estimated by weighing the food remaining after the 10 day period and subtracting that amount from the initial mass of food. Data from this experiment were analyzed in SAS using a split-plot ANOVA. Environmental treatment (hypoxia versus normoxia) was the main plot effect, time was the subplot effect factor, and the chamber was a random effect.

The adaptive and the innate immune systems are thought to present different energetic demands on mammals. As such, we challenged hypoxic and normoxic mice with one of two inoculations to stimulate either the adaptive immune system or the innate immune system. The adaptive immune system was stimulated by injecting keyhole limpet hemocyanin (KLH). The innate immune system was stimulated by injecting lipopolysaccharide (LPS). LPS is found in the cell walls of Gram-negative bacteria, and it stimulates a strong acute phase inflammatory response in mammals. The KLH and LPS experimental groups were each represented by 24 mice. The 24 mice were divided into four groups of six and assigned to one of the four treatments: hypoxia/inoculation, hypoxia/sham, normoxia/inoculation and normoxia/sham.

Mice in the KLH treatment group were inoculated with 100 μg KLH or sham (saline) intraperitoneally in a volume of 200 μl at the start of the experiment and placed in their appropriate environments. Five days after the inoculation, when B-cell proliferation should be at its peak, we measured their resting metabolic rates (see below). After the metabolic rates were measured, blood samples were taken via tail nick and collected in microhematocrit tubes, and animals were placed back in their environmental treatments. The blood was then centrifuged and hematocrit was measured as the percentage of packed red blood cells relative to the total blood volume. From these samples, plasma was extracted and frozen. At a later date, plasma levels of circulating KLH antibody were measured with an ELISA test (see below). The process was repeated again 9 days later, on day 14, when antibody production was likely at its peak. Twenty-seven days after the initial immunization, the mice were immunized a second time to evaluate the metabolic demands of the ‘booster effect’, the secondary response to an antigen after having been primed by the first response. The same protocol described above was repeated again, with resting metabolic rate (RMR) and blood measurements taken 5 and 14 days after the second inoculation, or on days 32 and 41 of the experiment. All data were analyzed in SAS using a repeated measures split-plot ANOVA. Environmental treatment was the mainplot effect (hypoxia versus normoxia), injection type (KLH or sham) was the subplot factor, and chamber was a random effect. A group statement (which fits different error variances to the groups) was used to correct the problem of unequal variances between both KLH and sham data, and between data from the first injection and data from the second injection.

Mice belonging to the LPS experimental group were housed in the hypoxic or normoxic chambers for 11 days prior to their first immunological challenge. At 11 days, when in the midst of physiologically acclimating to hypoxia, the mice were given an intraperitoneal injection of either 0.2 ml of 0.5 mg kg^-1 LPS or a sham of saline, and then they were placed in the metabolic chambers containing appropriate oxygen concentrations. Two hours after the injection, blood was collected in microhematocrit tubes via tail nicks, and then mice were weighed and immediately placed back in the metabolic chambers. The metabolic measurements began 4 h after the injections, when the acute phase response should be highest. Animals were returned to their experimental environments after the metabolic measurements. After being exposed to LPS, there is a period of several days during which the animals are resistant to further inoculations, known as endotoxin tolerance. On day 27, when endotoxin tolerance was waning and animals were further acclimated to hypoxia, the entire inoculation and measurement process was repeated. From the blood samples collected at both time points, hematocrit was measured and plasma was collected and immediately frozen. At a later date, the amount of TNF-α present in the plasma was measured via a Luminex assay (see below). All data were log transformed and analyzed in SAS using a split-plot ANOVA. Environmental treatment (hypoxia versus normoxia) was the mainplot effect, injection type (LPS or sham) was the subplot factor, and chamber was a random effect.

Blood samples were collected in heparinized tubes and vials were quickly processed for analysis with flow cytometry. The focus of this assay was to count relative quantities of circulating T-cells, B-cells, natural killer (NK) cells and polymorphonuclear leukocytes (PMNs). To this effect, the fluorochrome-labeled antibodies used for the assay were FITC-anti-mouse CD45 (eBioscience, San Diego, CA, USA) to distinguish the population of CD45⁺ leukocytes from the erythrocytes, PE-anti-mouse-CD3e (eBioscience) to distinguish T-cells, PE-TxR-anti-mouse CD19 (Invitrogen, Carlsbad, CA, USA) to distinguish B-cells, APC-anti mouse NK1.1 (eBioscience) to distinguish NK cells, and PC7-anti-mouse Gr-1 (eBioscience) to distinguish PMNs. Fifty microlitres of this antibody cocktail was added to 50 μl of blood and allowed to sit for 10 min. To this solution, 575 μl of FACSLyse (BD Biosciences, Franklin Lakes, NJ, USA) was added after which the solution sat for another 10 min. Finally, 25 μl of Streptavidin beads (Polysciences, Warrington, PA, USA) were added to the solution. Antibody-bound cells were examined with a four-color Beckman Coulter XL/MCL Flow Cytometer (Beckman Coulter, Hialeah, FL, USA) in an identical manner as discussed in duPre et al. (duPre et al., 2008). Signals from low angle forward light scatter (FS), orthogonal light scatter (SS) and four colors of fluorescence (FL1, FL2, FL3 and FL4) were collected using a logarithmic amplification. The optical filters were set to collect ∼525 nm (FL1, FITC), ∼575 nm (FL2, PE), ∼670 nm (FL3, APC) and >740 nm (FL4, PC7). Data files were analyzed and median fluorescence intensities were determined with FlowJo Software (Tree Star, Inc., Ashland, OR, USA).

NUNC 96-well MaxiSorp polystyrene plates (Fisher Scientific, Fairlawn, NJ, USA) were prepared by coating the plates with 50 μl of 10 μg ml^-1 KLH solution and incubating overnight at 4°C. The following morning, the plates were washed three times with phosphate buffered saline (PBS), then blocked by adding 200 μl of PBS with 5% nonfat dry milk and allowed to incubate for 1 h at 4°C. Plates were washed three times with PBS containing 0.05% Tween (PBS-T). Fifty microliters of mouse plasma diluted in PBS-T in 1:10, 1:100, 1:1000 and 1:10,000 ratios were added to the antigen-coated, blocked plates and incubated at 4°C for 2 h. After washing the plates three times with PBS-T, 50 μl of 1:10,000 dilution of peroxidase-conjugated rabbit anti-mouse immunoglobulins (IgA, IgG and IgM) was added to the plates and incubated at 4°C for 1 h. After another three washes with PBS-T, 50 μl of freshly prepared TMB Microwell Peroxidase Substrate (KPL, Gaithersburg, MD, USA) was added to each well and incubated at room temperature for 30 min. Fifty microliters of 1 mol l^-1 hydrochloric acid was added to the solution in each well. The optical density was determined at 450 nm using Spectra-Max micro-ELISA reader (Molecular Devices, Hercules, CA, USA). For each sample, plots of mean optical density versus log plasma dilution were made in Microsoft Excel and a logarithmic curve was fitted for each line [y=a×ln(x)+b]. The y-value was chosen was an optical density in the mid portion of the linear curves (e.g. 0.5 optical density units), and titer was defined as the reciprocal of the serum dilution (x-value) corresponding to the y-value.

After extraction of blood samples from the LPS mice, plasma was stored in 25 μl aliquots in polypropylene microcentrifuge tubes at -80°C. Luminex binding assay was used to detect circulating levels of the cytokine TNF-α. The assay was carried out in duplicate for each sample according to manufacturer's directions using the Mouse TNF-α singleplex bead kit (Invitrogen) and Luminex 100 System (Luminex Corporation, Austin, TX, USA). To correct issues with unequal variance, data were log transformed and then analyzed with a split-plot ANOVA.

RMRs were measured with a flow-through respirometry system identical to that used in Wone et al. (Wone et al., 2009) with the exception of the supply air gases. Briefly, this was a 16 chamber open-circuit system, which allowed up to 12 mice to be measured simultaneously, with the remaining four chambers left empty to measure baseline controls of incurrent air. The 590 ml chambers received dry mixed air at 200 ml min^-1 STP, regulated by upstream mass flow controllers (Sensiriron, Zurich, Switzerland). Incurrent flow rates and the sampling order of the chambers were controlled with LabView software and a custom automated control system. To avoid the potential physiological response of animals responding to rapid changes in oxygen partial pressure, the metabolic rates of the mice were measured at the partial pressure in which they were acclimated. Therefore, the air supplied to the mass flow controllers came from tanks of mixed air. The normoxia groups were measured under normoxic conditions, and received mixed air of 24% oxygen and 76% nitrogen. Likewise, hypoxia groups were measured under the hypoxic conditions to which they were acclimated, and received mixed air of 12.5% oxygen and 87.5% nitrogen. Temperature of the chambers was maintained at 30°C, which is within the thermal neutral zone of Mus musculus (Gaskill et al., 2009; Gordon, 1985). After exiting the chambers, the excurrent air was scrubbed of water and CO₂ with Drierite and Ascarite II, respectively, and then monitored with two Oxilla II dual channel oxygen analyzers. Because these oxygen analyzers were capable of measuring four chambers simultaneously, excurrent air going to the analyzers was switched every 15 min, such that each mouse was measured once per hour. Between measurements of each set of animals, samples of baseline air were taken from the four empty control chambers. RMR of the mice was measured for a 4 h period, giving us four separate 15 min measurements from each mouse. These measurements were monitored at 1000 Hz, and the averages were recorded every 5 s. Oxygen consumption (i.e. RMR) was calculated using the appropriate equation (Hill, 1972). Prior to the RMR measurements, mice were fasted for 12 h and weighed to the nearest 0.01 g. Measurements took place in the early portion of the light cycle. KLH mice were allowed to adjust to the chamber environment for 1 h before measurements began. Because LPS mice were manipulated prior to the RMR measurements, they were placed in the metabolic chambers 4 h prior to measurements both to adjust to the metabolic chambers and to maintain their oxygen environmental conditions during the manipulations. At the conclusion of the RMR measurements, mice were weighed and returned to their treatment environments. For logistical reasons, normoxic mice were measured together in one group, and the hypoxic mice were measured as another group the same time the following day. The lowest 5 min metabolic rate of the four sampling periods was used as our final RMR measurement. Body mass is reported as the means of the before and after measurements.

Initially, mice assigned to the hypoxic treatment were slightly but not significantly heavier that those assigned to the normoxic treatment. Hypoxic mice lost mass during the first 4 days of exposure to hypoxia, and by day 5 they were significantly smaller than normoxic mice (P<0.001; Fig. 1A). Hypoxic mice remained significantly smaller throughout the experiment (F_1,2=357, P=0.0028). The significant differences in mass and growth rate between hypoxic and normoxic mice all took place in the first few days when hypoxic mice lost mass. Over the first 5 days, growth rate differed significantly (P=0.001) as hypoxic mice lost body mass (-0.323 g day^-1) and normoxic mice gained body mass (0.112 g day^-1). Subsequently growth rates were not significantly different, and they were more similar in magnitude than during the first 4 days.

Mean daily food consumption was not statistically significantly different between the hypoxic and normoxic mice at any point in the experiment. This result was unaffected by whether body mass was included in the model as a covariate (Fig. 1B).

Hematocrit of hypoxic mice was significantly different from normoxic mice by day 5, and remained significantly different throughout the experiment (F_1,2=1894, P<0.001). The hematocrit of hypoxic mice climbed to a mean of 76.6% by day 36 (Fig. 1C).

The numbers of circulating immune cells were quantified via flow cytometry. The total number of CD45⁺ cells (hematopoietic cells carrying the CD45⁺ antigen) was expressed as number of cells per microliter of blood. B-cells, T-cells, NK cells and PMNs were quantified as numbers per microliter, and their frequency in the total number of white cells (CD45⁺ cells) was counted. The numbers per microliter yielded somewhat erratic results. These numbers were often inflated or diminished by high or low blood volumes used for flow cytometry. Although there were a fair number of CD45⁺ cells not identified on days 15 and 36, the expression of the cell types as their frequency of the total white blood cell count was much more consistent. Whether the data were analyzed in terms of cells per microliter or as a frequency of CD45⁺ cells, no significant differences were found with regards to B-cells, NK cells or PMN. However, on day 36 of hypoxia, hypoxic mice had significantly higher frequencies of T-cells (45.6%) than normoxic mice (39.1%, P=0.019; Table 1).

At the start of the experiment, mean body mass of the mice (measured in unfasted mice) was not significantly different across the treatments. Thereafter, body mass of the mice varied with the environmental treatment. Hypoxic mice were on average smaller than normoxic mice, but these differences were only significant on day 32 (P=0.003). Body mass did not vary significantly with respect to KLH or sham injections (Fig. 2A), and there was never a significant effect of the interaction of injection and environment (P-values ranged from 0.56 to 0.95). For reasons that are unclear, all mice showed reductions in body mass during the first 14 days of living in the environmental chambers.

Inclusion or exclusion of body mass as a covariate in the analyses of metabolic rates yielded qualitatively similar patterns in the results. The lack of significant correlation between body mass and metabolic rate is consistent with previous metabolic measurements in this strain of mouse (Johnston et al., 2007) (Fig. 3). For simplicity, we report only the results from analyses that included body mass as a covariate. Days 5, 14 and 41 each had one extreme outlier with high RMRs, whereas day 32 had two extreme outliers, one high and one low. Inclusion or exclusion of these outliers did not affect the outcome of the analyses, and we report values of the analyses with these outliers excluded. The mean RMRs of mice were significantly lower in hypoxic mice than normoxic mice on days 5 (0.4706 and 0.6050 ml O₂ min^-1, respectively; F_1,2=31.24, P=0.0306) and 14 (0.4221 and 0.5793 ml O₂ min^-1, respectively; F_1,2=32.67, P=0.0293). By day 32, although hypoxic mice still had lower RMRs on average than normoxic mice (0.5799 and 0.6947 ml O₂ min^-1, respectively), the differences were no longer statistically significant. On day 41, RMRs were similar across the treatments, although hypoxic mean RMRs were still slightly lower than normoxic mean RMRs. RMR showed no consistent pattern with respect to KLH or sham treatment, and the RMRs between these groups were not significantly different on days 5, 14 or 32. Only on day 41 was there a significant difference between KLH and sham mice, with sham mice having slightly higher RMRs than KLH mice (0.6244 and 0.5766 ml O₂ min^-1 respectively; F_1,16=4.87, P=0.042), a trend opposite of what would be predicted if mounting a humoral immune response were energetically expensive (Table 2). The effect of an interaction between type of injection and the oxygen environment (hypoxia or normoxia) was never significant, with P-values ranging from 0.12 to 0.53.

By day 5, hematocrit was significantly different between hypoxic and normoxic mice (F_1,2=52.95, P=0.0184), and it remained different throughout the experiment (Fig. 2B). Mean hematocrits for normoxic mice were between 51 and 52% throughout the experiment. In hypoxic mice, the hematocrit climbed to 57% by day 5 and reached a maximum of 63% by day 32. Hematocrit did not vary significantly with respect to immunological treatments during days 5 and 41. However, during day 14, when B-cell antibody production was likely high, there was a significant interaction effect (F_1,16=7.4, P=0.015). On day 14, hypoxic mice injected with KLH had lower hematocrits (61.3%) than hypoxic mice injected with shams (63.2%; P=0.0018), although the differences were biologically modest. Hematocrits remained relatively stable until day 32 with the difference between KLH and sham mice still significantly different (P=0.0463). By day 41, although KLH mice still had slightly lower hematocrits than sham mice, the differences were no longer significant (Fig. 2B).

ELISA was used to test for the presence of KLH antibody in the mouse serum samples. This test was performed to validate the efficacy of the injections and to monitor antibody production in response to KLH. All animals injected with KLH tested positive for the presence of KLH antibodies, and none of the sham animals were positive for these antibodies. Further analysis of the data was affected by whether we included four large outliers (data points twofold or higher than the next highest data point). Inclusion of the outliers inflated error variances, reducing the ability to detect possibly significant treatment effects. The four outlier points that were excluded were one KLH and one sham mouse from day 31, and one KLH and one sham mouse from day 41, and thus were evenly distributed across treatments. The data and analyses reported hereafter exclude these four large outliers.

The presence of antibodies to KLH in mouse plasma samples was validated on day 5 (mean titer=11.1 for KLH, 4.37 for sham; P=0.016). Mean antibody production in the KLH injected mice increased by more than threefold by day 14 (mean titer=33.42 for KLH, 7.03 for sham; P=0.002). After the second KLH injection, the experimental mice showed the typical secondary immunization or 'booster' response. On day 32, 4 days after the second injection, mice showed a 100-fold increase in antibody production (mean titer=3984 for KLH, 2.33 for sham; P<<0.01) over day 14. On day 41, 14 days after the second injection, the antibody production increased by another 53% (mean titer=6088 for KLH, 2.98 for sham; P<<0.01). A significant treatment by environment interaction was detected (F_1,63=6.65, P=0.013), which was driven by the higher KLH titers of hypoxic mice. The mean ELISA titers were consistently higher in hypoxic mice injected with KLH over normoxic mice, and these differences were statistically significant across four measurements of the experiment (repeated measures split-plot ANOVA, P=0.013). In particular, these differences were significant on days 32 (P=0.0067) and 41 (P=0.002; Fig. 4; Table 3).

At the start of the experiment, the initial body masses of the mice were not significantly different between immune treatments or environments. On day 11, when LPS was first injected, hypoxic mice had a smaller mean body mass than normoxic mice (19.7 and 21.2 g, respectively), but the difference was not statistically significant (P=0.059). This pattern remained through day 27, when hypoxic mice were still significantly smaller than normoxic mice (20.2 g versus 22.6 g; P=0.0013). The body masses of mice assigned to the LPS treatment versus the sham treatment were not significantly different (Fig. 5A), nor were the interactions with injection type or environment (P=0.98 and P=0.26 for days 11 and 27, respectively).

Inclusion or exclusion of body mass as a covariate in the analyses of metabolic rates yielded qualitative similar patterns in the results (Fig. 6). For simplicity, we report only the results from analyses that included body mass as a covariate. The RMR of mice was only significantly different between normoxic and hypoxic mice at day 11. Hypoxic mice had significantly lower RMR (0.521 ml O₂ min^-1) than normoxic mice (0.638 ml O₂ min^-1; F_1,2=20.44, P=0.046). This pattern was absent by day 27, when hypoxic and normoxic mice had very similar RMRs of 0.581 and 0.585 ml O₂ min^-1, respectively (F_1,2=0.01, P=0.927). Although the mean RMRs of LPS mice were higher than those of the sham mice on both days and within both treatments, these differences were small and not statistically significant (Table 4). The interaction of injection type with oxygen environment was not significant on either day (P=0.82 and P=0.99 for days 11 and 27, respectively).

By day 11, hypoxic mice had significantly higher hematocrits than normoxic mice (63.5 and 54.3%, respectively; P=0.0112), and this pattern remained through day 27 (62.4 and 51.1%, respectively; P<0.005; Fig. 5B). Overall, LPS mice had significantly lower hematocrits than sham mice (P<0.001 and P=0.004 for days 11 and 27, respectively). However, the injection of LPS was administered to the animals shortly before hematocrit measurements were taken. Because measurable changes in red blood cell production occur over longer periods of time (days to weeks), it is unlikely that these differences reflect a response to LPS. The interaction of injection type with oxygen environment was not significant on either day (P=0.12 and P=0.244 for days 11 and 27, respectively).

To verify that the mice were mounting an inflammatory response to the LPS injections, Luminex assays were run on the plasma samples to detect the presence of the inflammatory cytokine TNF-α. All mice injected with LPS showed significantly higher values for TNF-α than sham mice, demonstrating that they all responded to the injections (P<<0.001 on both days 11 and 27). The lack of significant TNF-α levels in sham mice indicates that neither the injection nor the blood-collecting processes led to systemic inflammatory responses in control mice. The TNF-α response was stronger in hypoxic mice than in normoxic mice on both days. On day 11, hypoxic mice injected with LPS had higher mean TNF-α levels (1242.7 pg μl^-1) than normoxic mice injected with LPS (832.4 pg μl^-1). Analysis of this log-transformed data revealed a difference that was not statistically significant (P=0.065). However, when a normoxic group outlier (three times the value of the next highest value) was removed from the data set, the mean for the normoxic group changed to 509.8 pg μl^-1, and the differences between the log-transformed data were highly significant (P=0.0039). The interpretation that follows reflects the data without the outlier. On day 27, the mean TNF-α level of hypoxic mice (341.9 pg μl^-1) was again higher than that of normoxic mice (96.75 pg μl^-1). Analysis of the log-transformed data confirmed that the difference was significant (P=0.026; Table 5). The P-values associated with the overall interaction of environment and injection type were not significant on either day. However, these interaction terms include the differences between sham normoxia and sham hypoxia mice. Because the sham mice did not respond with significant TNF-α levels, inclusion of sham mice in the interaction terms likely renders the interactions for the experiment not significant. The concentrations of TNF-α at day 11 were much higher than those at day 27, and this lowered response to LPS is likely due to the waning effects of endotoxin tolerance (Sanchez-Cantu et al., 1989).

The purposes of the study were to: (1) test whether chronic hypoxia had a significant effect on the immune system, particularly the circulating milieu of leukocytes, and (2) examine the energetic costs of coping with chronic hypoxic stress, immune stress and both simultaneously. The results of these studies were somewhat surprising. Although hypoxia did not appear to have a significant effect on the peripheral blood immunophenotype, it did appear to have a positive effect on the response of the immune system to immune challenges. Chronic hypoxia led to a transient decrease in RMR, but this decreased metabolic rate did not appear to affect the immune system, nor did immune challenges appear to affect RMR. The results of this experiment suggest that hypoxia has a complicated, positive relationship with the immune system that is not mediated by energetic costs.

Mice responded to chronic hypoxia in several characteristic ways, indicating that they were indeed experiencing hypoxic stress. In all three experiments, hypoxic mice lost weight and remained smaller than normoxic mice; however, the effect was more pronounced in the time-series study without any immune challenge. Interestingly, in the time-series experiment, food consumption was never significantly different between hypoxic and normoxic mice, indicating that something other than anorexia was responsible for the loss in body mass. The starting body masses of both KLH and LPS mice were measured in unfasted mice, and are likely why initial measurements were higher than the subsequent measurements. The mice we studied were relatively young and hence still growing. The large increase in body mass from day 14 to day 32 in the KLH mice likely reflects the more than 2 weeks of growth that had taken place between measurements. Body mass also dropped slightly between days 5 and 14 and between days 32 to 41. However, these slight drops in body mass are likely related to the slightly longer fasting periods that mice experienced on days 14 and 41 for logistical reasons. Although loss in body mass is a common response to chronic hypoxia, the exact mechanisms responsible for the change in body mass and the exact body components affected are still unclear (Quintero et al., 2010).

All mice responded to chronic hypoxia with an increase in hematocrit. Again, the effect was more pronounced in the time series study, with hypoxic mice peaking at 76% red blood cells, versus 63% for the KLH experiment and 63.5% in the LPS experiment. Although mice from the time series experiment experienced slightly more extreme hypoxia (12%) than mice from the KLH and LPS experiments (12.5%), it seems unlikely that this modest environmental difference fully explains such a large difference in hematocrit. We are unaware of other experimental factors that could account for these differences.

The flow cytometry data indicate that chronic hypoxia had little effect on the composition of peripheral blood leukocytes. The total numbers of leukocytes counted (CD45⁺ cells) were never significantly different between the treatments, indicating that hypoxia does not lead to an increase in total leukocytes and that the extreme hematocrit of 76.6% did not have a negative effect on leukocyte production. The ratios of B-cells, NK-cells and PMNs across the entire experiment were within the normal range for mice, though the B-cell ratios were slightly lower than what is reported as average for 6-month-old C57BL/6 mice (http://phenome.jax.org). These ratios were never significantly different between normoxic and hypoxic mice. The only significant variation found between the treatments was on day 36 when hypoxic mice had a significantly higher ratio of T-cells than normoxic mice. This result is the opposite of what would be predicted based on studies of humans at high altitudes and from studies of the effects of HIF on inflammation, in which hypoxia usually has a negative, regulatory effect on T-cell proliferation and function (Ben-Shoshan et al., 2008; Conforti et al., 2003; Facco et al., 2005; Sitkovsky and Lukashev, 2005) The reasons for this disparity are unclear, but perhaps they are related to differences in how small animals respond to chronic hypoxia.

Although the flow cytometry data indicate that chronic hypoxia alone had little effect on the circulating immune system, hypoxia did affect how mice responded to immunological challenges. In response to both KLH immunization and LPS inoculation, hypoxic mice had more robust responses, particularly in response to the secondary injections. In KLH-injected mice, ELISA titers increased dramatically over time and with a secondary response, as expected. Overall, hypoxic mice had higher titers than normoxic mice, with these differences being highly significantly different on days 32 and 41. Data collected from the flow cytometry time series study indicated that this more robust immune response by hypoxic mice could not be explained by a positive effect of chronic hypoxia on peripheral B cell numbers.

The responses of mice in the LPS experiment followed a similar pattern to the KLH experiment. Data from the Luminex assay indicated the classic response to LPS over time, with a robust primary response in TNF-α production and a less robust secondary response, which can be explained by endotoxin tolerance (Randow et al., 1995; Sanchez-Cantu et al., 1989). Like the KLH experiment, it is curious that the hypoxic mice had significantly higher circulating TNF-α levels on the second injection than normoxic mice. Again, it is unlikely that this response can be explained by alterations in numbers of circulating immune cells.

The more robust responses of hypoxic mice to KLH immunization and LPS inoculation are unexpected but potentially very important results. One generally expects to find decreased immune responses during times of physiological stress. That chronic hypoxia affects the immune system in such a way may have important implications for the evolution of animals colonizing high altitude, or for humans who travel to high altitudes. However, we recognize that caution is warranted when interpreting these results. Large variation was inherent to these data sets, with responses to injections eliciting exponentially higher responses than their sham counterparts. Furthermore, with low sample sizes, the data were sensitive to outliers, and exclusion of extreme outliers materially affected our interpretation of the ELISA and TNF-α analyses. Despite these pitfalls, the data indicate an interesting potential effect of chronic hypoxia on the immune system, and this study suggests that larger-scale, more detailed studies are warranted.

Transcriptional ties may explain the positive effect of hypoxia on immune responses (Safronova and Morita, 2010). As mentioned previously, through the induction of the HIF and NFκB, acute hypoxia has positive effects on the innate immune system via neutrophil migration, macrophage function and the production of proinflammatory cytokines (Gale and Maxwell, 2010; Murdoch et al., 2005; Walmsley et al., 2005a). Recent evidence also suggests that acute hypoxia has a stimulatory affect on the maturation and differentiation of antigen-presenting dendritic cells, which may aid antibody production (Rama et al., 2008; Spirig et al., 2010). These studies clearly establish a regulatory link between the immune system and acute, tissue-level hypoxia. However, less is known about the effects of chronic, systemic hypoxia on the immune system. Evidence that the immune system is also affected by chronic, systemic hypoxia comes from studies at high altitude and of chronic obstructive pulmonary disease (Facco et al., 2005; McNicholas, 2009). Furthermore, despite the waning of HIF induction within hours of hypoxic exposure, studies on gene regulation and cytokine regulation suggest that the effect of hypoxia on the immune system persists over longer time scales than hours (Baze et al., 2010; Lam et al., 2008). Therefore, there is evidence to suggest that, although chronic hypoxia has little effect on circulating immune cells, it may influence gene transcription and cytokine activity, which affect immunological function.

Hypoxia-enhanced immune responses may also be related to the endocrine system. For example, hypoxia temporarily increases circulating levels of melatonin. Melatonin generally leads to increased immune function, especially in laboratory rodents (Frisch et al., 2004; Kaur et al., 2002). Similarly, hypoxia positively influences circulating levels of leptin, a hormone known to be a positive regulator of the immune system (Bernotiene et al., 2006; Fantuzzi and Faggioni, 2000; Grosfeld et al., 2002; Meissner et al., 2005; Shukla et al., 2005). Leptin is released by fat cells, and it is thought to be a signal to other physiological systems of the energy status of the animal. Leptin has been implicated as a mechanism by which energetic tradeoffs with the immune system occur. Typically, energy-consuming processes deplete fat cells, decrease leptin levels and, therefore, blunt the immune response (Demas and Sakaria, 2005). However, as in the case of hypoxia, a physiological stress that increases leptin release may actually positively influence the immune response. It is plausible that more robust immune responses in hypoxic animals occurred due to integration of regulatory relationships between hypoxia and the immune system. The effects of a physiological stress on the endocrine system may be important to consider when making predictions about relationships and tradeoffs with other physiological processes (Martin, 2009).

Initially, hypoxic mice showed a decrease in RMR compared with normoxic mice, consistent with the notion that hypoxia is a stressor that affects not only maximal aerobic output but also resting metabolism. However, for both the KLH and LPS experiments, effects of chronic hypoxia were only significantly different during the first 2 weeks of measurements. By days 27 and 34, differences in RMRs were no longer statistically significant. In small mammals, hypoxia-induced hypometabolism is often seen as an important adaptive strategy for saving energy in the face of decreased oxygen availability (Frappell et al., 1992; Frappell and Mortola, 1994; Gautier, 1996; Mortola, 2004; Singer, 2004). Previous studies on rats indicated that hypometabolic response to chronic hypoxia was transient (Olson and Dempsey, 1978), but until now, this trend has not been demonstrated in smaller animals such as mice. Our experiment suggests that the energy-saving tactic of hypometabolism in small mammals is a transitory response. Perhaps metabolism returns to normal with the increase of hematocrit and other physiological responses to hypoxia that promote oxygen delivery.

A more surprising result is that injections of either the KLH or LPS immunological stimulant did not result in increased metabolic rates. The estimated costs of the humoral immune response are expected to be relatively low (Klasing, 1998; Martin et al., 2006; Raberg et al., 2002). In accordance with this idea, our results indicate that the humoral immune response induced by KLH does not lead to increased RMRs, even in the more proliferative secondary injection. ELISA confirmed that the mice were indeed mounting a humoral immune response. All mice injected with KLH not only showed positive ELISA titers, but also demonstrated the characteristic secondary response to the second KLH injection. With that said, the most convincing evidence that the humoral immune response does not result in increased metabolic rate was the lack of correlation between the ELISA titers and RMR. If antibody production were associated with higher energy costs, we would expect the hypoxic mice on day 41 to have the highest metabolic rates (Nilsson et al., 2007). In fact, on day 41, KLH mice had significantly lower RMR than sham mice, with hypoxic KLH mice having an even lower mean RMR than normoxic KLH mice. These results are contrary to a previous report with C57Bl/6 mice, which estimated a 30% rise in the metabolic rates of KLH-injected mice (Demas et al., 1997). At this point, the reasons for our differing results are unclear, particularly given that both studies were conducted on the same strain of mice. However, studies on other species have also failed to find a significant relationship between the humoral immune response and metabolic rate (Svensson et al., 1998) or a significant correlation between metabolic rate and the strength of an immune response (Nilsson et al., 2007).

In contrast to the adaptive immune response, the innate inflammatory response is expected to be quite expensive. The acute phase response is characterized by rapid inflammation, production and immigration of leukocytes, sickness behavior and fever, and is estimated to increase metabolic rate by as much as 50% (Lochmiller and Deerenberg, 2000). Yet, surprisingly, there was no significant difference in metabolic rate in LPS-treated mice versus sham-treated mice. Again, we know that the acute phase response was present in the LPS-injected mice because all displayed the classic signs of sickness behavior, such as lethargy and piloerection (Hart, 1988; Lacosta et al., 1999), and all had high circulating levels of the inflammatory cytokine TNF-α. Possible differences in metabolic rate could have been masked if injection and blood collection also led to inflammation, thereby potentially increasing metabolic rates in all mice. However, because sham mice did not have elevated TNF-α levels, and all mice were within the normal range for basal metabolic rate in C56BL/6 mice (Dinulescu et al., 1998), there is little evidence to support this idea. Another potential reason for not detecting significant differences in RMR between sham and LPS mice is that we did not measure and account for body temperature. Raising body temperature is estimated to increase basal metabolic rate by 10-15% per 1°C (Roe and Kinney, 1965), and hyperthermia likely accounts for a significant portion of energy expenditure during the acute phase response. The febrile response of C57BL/6 mice to LPS is related to dose and ambient temperature, with the response starting to wane 4 h after injection (Rudaya et al., 2005). Hence, it is possible that by the time of our RMR measurements, the febrile response was reduced or modest (Martin et al., 2008a). A lack of fever could account for a lack of difference in metabolic rate, especially because previously referenced estimates of the metabolic cost of the innate immune response included significant increases in body temperature (Cooper et al., 1994; Fewell et al., 1991). Whatever the case, our data indicate that the inflammatory response itself is not as energetically expensive as previously predicted.

The metabolic costs of the immune system may have gone undetected because of metabolic compensation (Martin et al., 2008b). For the KLH experiment, compensation could occur by reducing energy and mass associated with other organ systems and diverting that energy to support B-cell proliferation and antibody production (Derting and Compton, 2003; Mendes et al., 2006). Although we did not measure organ masses and cannot refute this hypothesis, there is little evidence to support it. Total body masses of the mice did not vary with respect to KLH challenge. With respect to the LPS experiment, metabolic compensation could occur via altered behaviors of mice. An increase in metabolic rate induced by LPS treatment could be counteracted by lethargy (sickness behavior), anorexia and decreased energy turnover (Klasing, 2004). However, we do not think that this is a major factor in our study as metabolic rates were measured in fasting animals at rest - metabolic compensation by anorexia and inactivity are, therefore, unlikely.

One may argue that chronic hypoxia is a unique physiological stress that may be an exception to usual relationships between environmental stress and the immune system. This view warrants consideration, given that chronic systemic hypoxia is not a stress that most mammals have evolved to cope with. Indeed, the type of hypoxia that mammals are more accustomed to is that related to wounds and tissue damage. In this situation, it is quite understandable why the immune system would be part of the response, not a competing process. However, we maintain that this study is relevant towards other ecological and environmental stresses. Firstly, the relationship between hypoxia and the immune system is an excellent example of how regulatory networks may complicate simplified interpretations of tradeoffs. Progress will likely require greater attention to understanding the nature of regulatory networks. Secondly, we found no lasting energetic cost of hypoxia or of immune function. This result calls into question the putative metabolic, energetic costs of the immune system.

There is ample evidence of tradeoffs between the immune system and other life history traits, and there is strong evidence that the immune system is sometimes costly (Klasing, 2004; Lee, 2006; Martin et al., 2008b; Norris and Evans, 2000; Raberg et al., 2000). Exactly what those costs are deserves more consideration. For the energetic cost of the immune system to be significant enough to divert energy away from other systems, it also ought to be significant enough to affect the metabolic rates of animals. That neither the humoral nor the innate immune responses elicited significant changes in metabolic rates should be cause for careful consideration by ecological immunologists. Indeed, there are many other costs of the immune system that may drive the ultimate evolution of tradeoffs between other systems, including increased exposure to oxygen free radicals, depletion of particular nutrients, increased susceptibility to autoimmunity, decreased embryo and gamete survival, and behavioral modification (Klasing, 1998; Long and Nanthakumar, 2004; Raberg et al., 1998; Ricklefs and Wikelski, 2002; Zuk and Stoehr, 2002). The immune system is arguably of one the most complicated physiological systems, and immunological phenotypes and responses are likely shaped by a multitude of sources of natural selection (Schmid-Hempel and Ebert, 2003). The field of ecological immunology seems poised to benefit from rapid progress as researchers combine mechanistic analyses with broad ecological and evolution perspectives, and keep an open mind about which currencies are most important physiologically.

 
• ELISA

  enzyme-linked immunosorbent assay

 
• HIF

  hypoxia inducible factor

 
• KLH

  keyhole limpet hemocyanin

 
• LPS

  lipopolysaccharide

 
• NFκB

  nuclear factor kappa beta

 
• NK

  natural killer

 
• PMN

  polymorphonuclear leukocyte

 
• RMR

  resting metabolic rate

 
• TNF-α

  tumor necrosis factor alpha