Measurements of body temperature and oxidative stress reveal differential costs associated with humoral immune function in a passerine bird

A fundamental concept within ecological immunology is that immune responses against pathogens is costly (Sheldon and Verhulst, 1996; Lochmiller and Deerenberg, 2000) and several arguments support this hypothesis. Firstly, an immune response is fully activated only as a response after contact with the pathogen, rather than being constantly in an activated state (Parkin and Cohen, 2001). Secondly, trade-offs exist between the immune response and other energetic demand events such as moult (Martin, 2005; Ben-Hamo et al., 2017), reproduction (Raberg et al., 2000; Verhulst et al., 2005), production of a sexual ornamentation (Sheldon and Verhulst, 1996), or flight activity in birds (Nebel et al., 2012; Eikenaar and Hegemann, 2016). Thirdly, the main components of the immune functions are upregulated during the rest phase rather than during the active phase of the day (Geiger et al., 2015), thus avoiding or minimizing competition for the availability of the resource energy. All these described trade-offs provide a strong indication that the immune function might be associated with significant energetic costs for the organism.

Metabolic rate measurements during rest phase collected from a single or multiple time points, or determination of the energy expenditure over a 24 h period have been previously employed to estimate the energetic costs of the immune response (e.g. Svensson et al., 1998; Martin et al., 2003; Amat et al., 2006). The majority of the single time point measurements of metabolic costs in wild populations revealed either increased levels or non-existent differences in resting metabolic rate after an immune challenge compared with unchallenged controls (Svensson et al., 1998; Ots et al., 2001; Hõrak et al., 2003; Eraud et al., 2005). Typically, studies measure metabolic costs close to the peak of antibody production (in the case of the humoral immune response), which may not necessarily be the most energetically demanding moment of the immune response (Abad-Gomez et al., 2013). An alternative practice/application is to repeatedly measure metabolic rate after the immune challenge. For instance, metabolic rate was measured in house sparrows (Passer domesticus) during the resting phase and during egg production (Martin et al., 2003). The energetic costs of the activation of a cell-mediated adaptive immune response seem comparable to the energetic costs required to produce half an egg (Martin et al., 2003). A third facilitated approach is the doubly labelled water technique for estimating cumulative energy expenditure within a specific period, commonly 24 h. This technique was applied on the 6th to 7th day after an immune challenge (peak antibody production) in greenfinches (Chloris chloris) to estimate daily energy expenditure and revealed a 4.7% higher daily energy expenditure in comparison to control birds, although the increase was not statistically significant (Amat et al., 2006). This method cannot exclude potential compensation within the 24 h energy budget in general and between events of the day or night that may differ in energy demand, as previously described for zebra finches (Taeniopygia guttata) (Deerenberg et al., 1998). Thus, method-specific shortcomings of previously applied techniques to quantify the energetic costs of the immune response have not enlightened our current understanding and knowledge regarding the energetic costs of an immune response.

The gold standard for estimating the metabolic rate, and thus energy expenditure, is direct calorimetry, which measures animal heat production directly (Kaiyala and Ramsay, 2011). Currently, none of the available techniques allows for continuous measurement of the energetic costs over several days without additional disturbance of animals, particularly when these animals are typically maintained in groups or flocks. A close relationship between metabolic rate and body temperature is indicated for mammals (Lovegrove, 2003; Geiser, 2004; Swoap, 2009) and despite significant evolutionary differences, the mechanisms driving this relationship seem to be similar for birds and mammals (Gilbert et al., 2008; Clarke et al., 2010). Here, we decided to use the approach of a near-continuous measurement of body temperature through implanted data loggers as an approximation of energy use to overcome the lack of adequate methods and provide measurements over several days for a flocking species. We hoped to obtain novel insight into the kinetics of energy use over the whole time of activation of the humoral immune response.

The outlined ambiguity of the energetic costs during immune response may hint that there are other costs involved in the expected energy allocation trade-offs, and oxidative stress could be such an alternative cost (Monaghan et al., 2009; Demas et al., 2012). Oxidative stress can be defined as the insufficient neutralization of produced reactive oxygen species (ROS) (Sies, 1997; Costantini et al., 2007; Monaghan et al., 2009). Production of free radicals is an unavoidable by-product of aerobic metabolism, and a significant increase in oxidative stress is associated with increased energy turnover, for example, during growth (Smith et al., 2016), physical activity (Costantini et al., 2008; Jenni-Eiermann et al., 2014) or reproduction (Alonso-Alvarez et al., 2006; Saino et al., 2011). If free radicals cannot be neutralized, oxidative stress occurs and this may damage various biomolecules (Sies, 1997; Costantini et al., 2007; Monaghan et al., 2009). Damaged biomolecules might be an additional burden to the organism, causing reduced or impaired organism functionality (Selman et al., 2012). Interestingly, the activation of the immune system significantly increased markers of oxidative stress (Costantini and Møller, 2009). Until now, activation of macrophages, heterophils etc. which aim to fight pathogens are considered as the main source of free radicals during the activation of the immune system (Costantini and Dell'Omo, 2006; Costantini and Møller, 2009). Increased energy metabolism caused by activation of the immune response was not considered up to now as a possible cost in terms of oxidative stress. Oxidative stress may serve as a link between energetic and non-energetic costs of the immune response.

In this study, we challenged one group of zebra finches (Taenopygia guttata) with a new non-pathogenic antigen [sheep red blood cells (SRBCs)] which activates components of the humoral immune response (Eraud et al., 2005; Ladics, 2007) and compared them to a sham challenged group, which served as the control. In both groups, we measured core body temperature over the entire period of the immune response to investigate the general pattern of body temperature variation, as a proxy for energy use. In addition, we used open-flow respirometry to measure aerobic metabolism just before the peak of antibody, which for zebra finches, is considered to be 6 days after immune challenge (Verhulst et al., 2005; Rutkowska et al., 2012). Antibody production is considered to represent the peak period of energetic costs. Oxidative stress caused by the activation of the immune function seems to be a non-energetic cost for the organism and we decided to quantify markers of oxidative stress at the time of peak antibody production after immune challenge. We hypothesized that immune challenge would increase oxygen consumption in immune-challenged birds compared with controls. We assumed that the humoral immune response imposes different costs at the organism level, and that these can vary over the course of activation (Svensson et al., 1998; Martin et al., 2003). We predicted that challenged birds will show generally higher core body temperature compared with the control birds, with a likely difference between daytime and night-time body temperature, as seen previously (Sköld-Chiriac et al., 2015). Since there are no precise data available on the kinetics of the humoral immune response, we were not able to precisely predict the body temperature between consecutive days or nights. In terms of markers related to oxidative stress we predicted an increase after immune challenge in comparison to the control group.

A total of 41 zebra finches [Taeniopygia guttata (Vieillot 1817)], 21 females and 20 males, ranging from 1 to 8 years old were involved in the study. The animal study was reviewed and approved by II Local Bioethical Committee in Kraków, Poland 311/2018 (22 November 2018).

Birds were housed at the Institute of Environmental Sciences, Jagiellonian University in a climatic chamber under thermoneutral conditions of 30°C (a temperature within the thermoneutral zone, Calder, 1964) in two sex-separated aviaries (L×W×H: 2×2×2.5 m), and under the photoperiod of 12 h light:12 h dark. A mix of millet species and water was provided ad libitum except for during the metabolic rate measurements. Birds were acclimatized to these housing conditions for 4 weeks before we implanted the thermosensitive data loggers or performed a sham surgery (see below details for the thermologgers' implantation, sample size distribution Tables S1 and S2). For 2 months post-surgical implantation of data loggers, birds went through six metabolic measurements (not all of which are covered in this manuscript) performed at different ambient temperatures (10–38°C). All birds experienced the same number of metabolic rate measurements before immune challenge. A minimum of 4 days passed between the last metabolic measurement and immune challenge, allowing birds to recover and to exclude any inference of metabolic measurements on our experimental treatment. Birds were divided randomly into the experimental group (SRBC immune challenge) or into the control group, which was injected with saline (see details in ‘Immune challenge and haemagglutination assay’ and Fig. 1). On the evening of day 6 after the challenge, birds were transferred to metabolic chambers for resting metabolic rate measurement (see details in: ‘Resting metabolic measurements’). In the morning after metabolic rate measurements, blood from the wing vein was collected to quantify haemagglutinin titre (see details in ‘Immune challenge and haemagglutination assay’) and markers for oxidative stress (see details in ‘Reactive oxygen metabolites and antioxidant capacity’) and then, birds were euthanized for tissue sample collection (not covered by this manuscript) and to extract thermologgers to collect records of body temperature.

All animals were implanted with thermologgers (n=27) or were exposed to anesthesia (n=14). Core body temperature was recorded at 18 min intervals by nano-data loggers (DST nano-T, Star Oddi). One logger failed to provide temperature recordings but this individual was not excluded from other analyses; however, we had to exclude from all analysis one individual because of a wound detected close to surgery site. Before implantation, the data loggers were calibrated in a water bath and a thermometer (WD-35427-52 Temp 300, Oakton Instruments, Vernon Hills, IL, USA) at four calibration temperatures from 30°C to 45°C. For the implantation, data loggers were disinfected with 95% ethanol and lightly coated with a lubricant (Vaseline 80% and paraffin wax 20%). Intraperitoneal implantation of the data loggers occurred under anaesthesia (3% of isoflurane, Aerrane Isofluranum, Baxter, Lessines, Belgium, in 0.3 l min⁻¹ airflow provided by a closed-pipe inhalation system, UNO200VAP vaporizer, Scintica instrumentation inc., Maastricht, The Netherlands). We used absorbable sutures to close the incision with 2–3 stitches (Safil 5/0, AesculapAG, Tuttlingen, Germany). The surgery room had a constant temperature of 21°C and birds were placed on a heat mat to maintain constant body temperature during anesthesia. After implantation, birds were monitored closely and were allowed a recovery period of 4 days. The immune challenge took place 2 months after the implantation. At the end of the experiment, we dissected the birds to remove the loggers and downloaded the body temperature recordings using the SeaStar program (Star-Oddi Ltd, Iceland).

The immune challenge involved intraperitoneal injection with either a novel non-pathogenic antigen (SRBCs) or saline for sham control. SRBC injection is a method widely used to assess humoral immune response in birds (Eraud et al., 2005; Verhulst et al., 2005; Ladics, 2007). On the day of the immune challenge, birds were injected with 150 µl of 40% suspension of SRBCs to activate specific antibody production. This protocol is established and confirmed to activate antibody production in the same bird population (Rutkowska et al., 2012). The highest concentration of anti-SRBC antibodies in zebra finches is expected on day 7 after immunization (our unpublished data; Verhulst et al., 2005). On the morning of the 7th day post-immunization and just after the metabolic rate measurements, we collected 200 µl blood from the wing vein into heparinized Eppendorf and then centrifuged for 15 min at 17,968 g for plasma separation (Sigma 1-15 Centrifuge; method modified from: (Hay and Westwood, 2002). Plasma was collected in capillary tubes and transferred into new Eppendorf tubes (separated for haemagglutination assay and oxidative stress markers) and later stored at −20°C until all samples were collected. After all dissections, samples were stored at −80°C (dissections took 5 days). Samples for the haemagglutination assay were removed from the freezer, were incubated at 56°C for 30 min to remove proteins that could influence the haemagglutination score. A 96-well plate (round bottom) was used for the assay and wells were filled with 20 µl saline. The first rows were filled with 20 µl plasma, followed by stepwise dilution with saline (lowest dilution of plasma 1:2, highest 1:256). Next, 10 µl of diluted SRBC was added to each well, air bubbles were removed and the plate was mixed gently then covered with parafilm and incubated for 1 h at 37°C. The haemagglutination score was determined as the negative log₂ of the last plasma dilution exhibiting agglutination (Matson et al., 2005), repeatability of the haemagglutination score in our laboratory condition was r²⁼0.88.

Oxidative damage and non-enzymatic antioxidant (OXY) capacity in plasma were assessed by d-ROM (reactive oxygen metabolites) and OXY colorimetric tests, respectively (Diacron International, Grosseto, Italy). Both biomarkers are good indicators of oxidative stress and were previously measured in many avian studies (Costantini, 2016). The d-ROM test measures mainly the organic hydroperoxides, which are reactive oxygen metabolites produced in the early phase of the oxidative cascade expressed as equivalent of mmol of H₂O₂. The d-ROMs were estimated in a duplicated sample of 4 µl plasma according to the protocol of Costantini and Dell'Omo (2006).

The OXY test measures the total non-enzymatic antioxidant capacity by quantification of the ability of plasma antioxidants to neutralize hypochlorous acid (HClO), which is an endogenously produced oxidant. Analysis was performed according to protocol by Sudyka et al. (2016); for this test we used 2×4 µl diluted plasma sample. Repeatability of the measurement within sample for dROM was r²=0.92 and for OXY was r²=0.99.

For all individuals, we quantified metabolic rate at both rest (τ-RMR) and active (α-RMR) phase, as a rate of O₂ consumption and CO₂ production using a 9-channel open-flow positive-pressure respirometry system. On the evening of day 6 after immunization, just before lights off, up to 10 individuals were transferred to a dark climate chamber and a maximum of 8 of these had resting metabolic rates (RMRs) measured. For each batch of measurements, one or two individuals were kept in the same conditions (temperature, airflow, same type of chambers) but RMR was not recorded because of multiplexer limitation (total of 4 individuals did not have RMR recorded). RMR measurement was performed in a dark climate chamber with a controlled temperature of 30°C in individual chambers (n=8). Individual chambers were built from rectangular plastic containers of 950 ml volume. Inside, we placed a metal cage to separate the bird from the inlet tube with mineral oil on the bottom. An inlet tube was inserted inside the chamber around 2 cm above the bottom and an outlet at the top which allowed a good airflow inside the chamber. The bottom of the container was filled with 50 g white mineral oil (AnVit, Szczecin, Poland) to collect faeces. A fresh sample of air at standard pressure and room temperature was dried with silica gel driers and divided into 9 streams; 8 were pumped into chambers with 8 mass flow system pumps 2 LPM (Sable System International, North Las Vegas, NV, USA), and one for baseline reference. Airflow was set at 450 ml min⁻¹ through chambers and regulated separately for each chamber. Samples from the chambers were regulated sequentially through Intelligent Multiplexer MUX (Sable System International, North Las Vegas, NV, USA). A sub-sample of the air stream was first analysed for water content and then dried with magnesium perchlorate (Anhydrone, J. T. Baker, USA) columns before passing through the CO₂ and O₂ analyser. The sample was analysed by Field Metabolic System FMS (Sable System International, North Las Vegas, NV, USA).

Both τ-RMR and α-RMR were analysed in the same way. Representative O₂ concentration value was calculated from the values recorded in the last 20 s before switching channels as described by Sadowska et al. (2015). Rest phase RMR (ml O₂ min⁻¹) was calculated based on the minimal volume of oxygen consumed from the last 8 h of the night when birds were in a post-absorptive state. Active phase RMR was calculated based on the minimal value of oxygen consumed from the last 6 cycles (each cycle 18 min) within a subjective daytime; however, lights were still off.

Markers of oxidative status (d-ROMs and OXY) were analysed with linear models, data for d-ROMs or OXY were used as dependent variables, treatment (immune challenge or control) as a categorical factor, with age and sex as a covariate. All non-significant interactions between effects were removed from the model; thus, the final model consists only of main effect of treatment.

Results of the haemagglutination assay were analysed with a linear model, with haemagglutination score used as the dependent variable, treatment (immune challenge or control) as a categorical factor, with age and sex as a covariate. All non-significant interactions between effects were removed from the model; thus, the final model consists only of main effect of treatment.

The resting metabolic rate was also analysed with a general linear mixed-effect model, separately for rest and active phase. α-RMR or τ-RMR was the dependent variable, treatment (immune challenge or control) categorical variable, body mass, age and sex as covariates and number of metabolic chambers as a random factor. All non-significant interactions between main effect of Treatment and covariates (body mass, age, sex) appeared to be not statistically significant and thus were removed from the final model.

Analysis of body temperature was performed with linear mixed-effect models, with dependent variable a 4 h average (or minimum) body temperature from the non-disturbed part of the day or night. For all models, we used 4 h mean (or minimum) body temperature derived from non-disturbance night and day before immune challenge (T_b,Pre) as a covariate to control for possible differences in individual body temperature (caused by sex, age, body mass, condition, etc.).

The first model investigated the general effect of immunization on body temperate. Combined data of the mean body temperature (for 6 days and nights) was a dependent variable, treatment (immune challenge or control), T_b,Pre as the covariate and bird ID number as a random factor to account for repeated-measures analysis.

The second analysis was used to specify the differences between time of the (daytime and night-time) body temperature. The model after reduction of non-significant interactions with continuous variables includes night mean body temperature (or amplitude change between day and night body temperature) as the dependent variable, treatment (immune challenge or control), time (day or night), interaction Treatment×Time, T_b,Pre as covariate and bird ID number as a random factor to account for the repeated-measures analysis.

The third model analysed differences between consecutive nights of the experiment. Model after reduction of non-significant interactions with continuous variables (T_b,Pre) includes meaning body temperature (or minimum) as the dependent variable, treatment (immune challenge or control), experimental night (1, 2, …, 6), interaction Treatment×Experimental night, T_b,Pre as covariate and bird ID number as a random factor to account for repeated-measures analysis. To test for the significance at a certain time, we used emmeans package (https://CRAN.R-project.org/package=emmeans); other analysis was performed using lme4 (https://CRAN.R-project.org/package=lme4) in R v.3.4.0 (https://www.r-project.org/).

Following the immune challenge with SRBCs, these birds had a significantly higher haemagglutination titre in comparison with sham controls (F_1,31=21.46, P<0.001). Out of 18 samples (3 missing plasma samples) in the immune-challenged group, 5 showed no signs of agglutination to SRBCs, while in the control group, out of 17 samples (2 missing plasma samples) only one sample showed evidence of agglutination at the lowest plasma dilution (dilution 1:2). Thus, we are confident that our manipulation successfully activated the humoral immune response to produce a specific antibody against a novel non-pathogenic antigen. Contrary to our predictions, analysis of oxidative status markers on day 7 after treatment showed that non-enzymatic antioxidant capacity was not affected by the manipulation (F_1,24=0.0075, P=0.94) (Fig. 2A). Additional support for successful activation of the immune response is derived from an increase in markers of oxidative damage in plasma for immune-challenged individuals in comparison to sham-injected individuals (F_1,30= 7.10, P=0.01) (Fig. 2B). A summary of significance levels for all covariates can be found in Table S3.

Insight into the energetic costs of the response to immune challenge can be observed by an increase in the RMR. Because of the expected circadian rhythm of the immune functions, we predicted a different response between α-RMR and τ-RMR in response to our manipulation. We confirmed a significant interaction of Sex×Treatment for of α-RMR (Fig. 3A); specifically, immune-challenged females had a higher metabolic rate in comparison to control females (P=0.03) and this was also significantly higher than that of immune-challenged males (P=0.02). In contrast, rest phase τ-RMR was not significantly different between experimental groups (F_1,31=0.56, P=0.81; Fig. 3B). Summary of significance levels for all covariates can be found in Table S3.

Continuous recordings of body temperature during all days of the experiment may be a better descriptor for the effect of our manipulation than just a one-point RMR measurement. We found a significant effect of immune challenge on body temperature (effect of treatment: F_1,23.3=7.9, P=0.009, Table 1). Immune-challenged birds had higher body temperature (LSM±s.e.=40.62±0.06°C) than controls (40.38±0.06°C). A significant interaction of Treatment×Time (F_1,282.1=6.2, P=0.01, Table 1) indicated that the experimental effect differed between day and night. Immune-challenged birds had a significantly higher night-time body temperature in comparison to controls (P=0.009), with no significant differences between the groups during the daytime (P=0.36). A significant interaction of Treatment×Experimental night (F_1,125.3=5.04, P=0.026, Table 1) revealed that experimental night differs in response to the manipulation, with immune-challenged birds having higher night-time body temperature during the first 4 nights after immune challenge in comparison to controls (Table 2). There was also a significant interaction of Treatment×Experimental night in the analysis of minimum body temperature (F_1,127.2=7.39, P=0.007, Table S4). A significantly higher minimum body temperature in immune-challenged birds was maintained until the third experimental night in comparison to controls.

The mounting of an immune response against a novel non-pathogenic antigen (SRBCs) was accompanied by higher night-time body temperature compared with control birds in the first nights following the challenge, indicating higher energetic costs during the initial stage of the immune response (Fig. 1). Seven days after the immune challenge, only the metabolic rate measurements showed a clear interaction of Sex×Treatment in differences of energy metabolism. However, markers of oxidative damage revealed a sex-independent increase of oxidative damage in plasma after immune challenge compared with controls (Fig. 2B).

Significant elevation of oxidative damage following the immune challenge is consistent with other findings in which immune challenge to birds increases the risk of oxidative stress (Costantini and Møller, 2009). The novelty of this study is the insight into the dynamics of body temperature, a proxy of energy metabolism, over the course of the immune response. Our data on core body temperature revealed that night-time, but not daytime body temperature, was affected for several days following the challenge of the humoral arm of the immune system (Table 2). The quest to maintain a higher night-time body temperature during the immune response might represent a new explanation for the observed elevated oxidative damage. Additional analysis of the oxidative damage markers with average body temperature during the first 4 nights included as covariate showed no difference between groups. This is indirect support for this claim, as including body temperature as a covariate removed this difference between groups (compare Fig. S1 with Fig. 2B). It may provide a largely overlooked proximate explanation for increased oxidative burden ultimately driven by activation of an immune response: body temperature modulation.

Elevation of body temperature is well known as a fever-like response after an organism encounters an antigen (Kluger et al., 1996, 1998), and typically will not last longer than a few hours, during which mainly the innate components of the immune system are activated (Marais et al., 2011). Longer lasting responses to immune challenge with non-pathogenic lipopolysaccharides (LPSs) are day to night differences described as ‘fever at night and hypothermia during the day’ (Sköld-Chiriac et al., 2015). However, the subcutaneous body temperature termed ‘fever at night’ is still considerably lower than the body temperature the birds achieve during the day. Birds under LPS challenge reduce the normothermic range compared with control birds from ∼2.5 to ∼0.5°C. This is largely in agreement with our present findings of body temperature modulation with higher than normothermic nocturnal body temperature, but the compensation through lower than normothermic daytime body temperature was not detectable, presumably because it is masked by the effect of activity on body temperature. In addition, maintenance of nocturnal body temperature above normothermic values was observed over several consecutive nights (Fig. 4, Table 2) and such a long-lasting response seems unlikely to be caused by the innate immune system in response to a non-pathogenic antigen (SRBCs). Even under simulation of severe infection, the innate response does not last for more than a few consecutive days (Marais et al., 2011; Sköld-Chiriac et al., 2015; Amaral-Silva et al., 2021, 2022). We consider that the birds’ body temperature modulation may be summarized as ‘maintaining a higher than normothermic body temperature during the night’ rather than terming this moderate temperature response that reoccurs over several nights as fever-like; and this more enduring modulation of body temperature must be considered to depend on the type of immune challenge, the birds' energetic state in terms of energy availability, and last but not least, the time of the challenge.

Maintenance of higher night-time body temperature in immune-challenged birds compared with control birds may be even better categorized as a less-pronounced drop of body temperature during the rest phase (less pronounced night-time hypothermia). Components of the adaptive immune response (cell-mediated and humoral response) might require the body temperature not to drop to normothermic (unchallenged) levels in order to provide an optimal operating range for activation of a specific cell. All biochemical processes are highly dependent on temperature and regular overnight reduction of animal's body temperature may negatively affect the rate of biochemical reactions and ultimately have a negative effect on the immune response and fitness (Tattersall et al., 2012; Elias et al., 2014). Therefore, maintenance of increased night-time body temperature above normothermic values may be an adaptation to reduce the negative consequences of temperature fluctuations on the activation of immune functions. Thus, a hyperthermic response appears to be a very plausible explanation. The cells most sensitive to temperature are T_h-cells (Müllbacher, 1984), which activate the B-cells to produce antibodies after SRBC challenge (Ebert, 1985; Ladics, 2007). In vitro studies confirmed that B-cells fail to produce antibodies when the temperature is 2°C lower than physiological values (Saririan and Nickerson, 1982). In our experiment, average variation in body temperature between day and night was 2.34°C for control individuals. Proximately immune-challenged birds maintained higher body temperature than controls at the same time (estimates from the model, Table 2). The observed average difference was the greatest on the first night (0.39°C) and decreased gradually until the fourth night (0.21°C). Such a mechanism may be particularly important for animals that are accustomed to saving-energy strategies of reducing body temperature during every resting period, particularly when encountering a shortage of resources or inhospitable environmental conditions (McKechnie and Lovegrove, 2002; Amaral-Silva et al., 2022).

Further support for this suggestion can be deduced from another investigation employing the same immune challenge. After an SRBC injection, female zebra finches revealed a negative relationship between the active phase metabolic rate (daytime energy metabolism measured during activation of the immune response) and the strength of immune response measured as haemagglutination titre (Bryła et al., 2021). This outcome was primarily explained by possible energetic compensation for night-time hyperthermia after the immune challenge. Currently collected data do partially confirm this claim, as we found increased night-time body temperature during the first 4 nights, but surprisingly, we also found a significant Treatment×Sex interaction for the active phase RMR. Thus, our study is the first to reveal a sex-specific response to humoral immune challenge, whereas in previous studies there was no significant effect of sex on the metabolic rate (Svensson et al., 1998), sex was not included in the analysis (Ots et al., 2001; Eraud et al., 2005) or the experiment was only performed on single-sex birds (Hõrak et al., 2003). It is generally thought that males have a weaker immune system in comparison with females (Nelson and Demas, 1996; Moore and Siopes, 2000; Demas et al., 1996). A recent meta-analysis confirmed that for birds, multiple immune functions are female biased (Vincze et al., 2022). This may explain why we observed the interaction in active phase metabolic rate, with females showing significantly higher oxygen consumption than males over the same time, and it can be explained by higher activity of the humoral immune response in females. Metabolic rate measurements in combination with body temperature recordings clearly showed that metabolic costs of the immune response appear much earlier (and maybe sex specific) than the peak of antibody production, as was used in previous studies (Svensson et al., 1998; Ots et al., 2001; Hõrak et al., 2003; Eraud et al., 2005). Further studies which will aim to quantify the costs of the immune challenge should consider measuring metabolic rate much earlier than at the peak of antibody production in case of an adaptive immune response.

Animals very often under energetically challenging conditions drop body temperature during resting, as an energy-saving strategy (McKechnie and Lovegrove, 2002; Zagkle et al., 2020). Our study shows less pronounced hypothermia during the night (birds maintained a higher than average night-time body temperature), which indicates an impairment of the energy-saving strategy, with simultaneously activated immune response defence mechanisms; thus, both mechanisms may influence oxidative stress after immune challenge. We confirmed that oxidative damage was elevated 7 days after the immune challenge in comparison to control individuals (Fig. 2). However, our methodology does not allow us to further unravel the cause of this damage. Zebra finches on day 7 after immune challenge are expected to exhibit a peak of antibody production to target foreign antigens for elimination (Verhulst et al., 2005; Rutkowska et al., 2012). An increase in oxidative damage may be caused by activation of mechanisms responsible for neutralization and degradation of the antigens (Yang et al., 2013) because some of the defence mechanisms of the immune system are based on production of reactive oxygen species: for example, oxygen burst, which is a mechanism known to be a source of oxidative stress (Sorci and Faivre, 2009). Secondly, elevated oxidative damage in immune-challenged birds compared with control birds may be a consequence of the increased energy turnover caused by less-pronounced night-time hypothermia, since ROS production is a byproduct of aerobic metabolism (Harman, 1956; 1992; Sies, 1997; Rahal et al., 2014). Consequently, animals facing an immune challenge seem to encounter a trade-off in terms of body temperature regulation: hypothermic response as an energy-saving mechanism versus hyperthermia as a consequence of the immune response activation (Schieber and Ayres, 2016; but also: Amaral-Silva et al., 2021, 2022).

Before and during the main experimental procedure, birds were exposed to several challenges: 6 metabolic measurements at different ambient temperatures, repeated capture, handling or transfer between the home cage and metabolic chambers. Even though birds were habituated to these experimental conditions, we cannot exclude the possibility of the cumulative effect of exposure to stress conditions that could have significantly affected results of the experiment. Prolonged elevated glucocorticoid exposure usually indicates a state of chronic stress; moreover, stress conditions in birds such as heat stress, low temperatures and corticosterone exposure were confirmed to have an inhibiting effect on innate (Mashaly et al., 2004; Roberts et al., 2007) or humoral (Nazar and Marin, 2011) components of the immune response (also reviewed by Martin, 2009). Conversely, the humoral immune response of barn owls (Tyto alba) nestlings was shown to be independent of corticosterone exposure (Stier et al., 2009). Chickens exposed to glucocorticoids only induce the formation of reactive oxygen species initially; during prolonged exposure, they seem to adapt to elevated concentrations of corticosterone in terms of their redox homeostasis (Lin et al., 2004). Furthermore, a metanalysis by Costantini et al. (2011) showed that glucocorticoids have a substantial impact on oxidative stress markers, although this effect varied with treatment time and was greater in long-term trials. In future studies it should be possible to exclude stress factors, meaning that measurements of body temperature, metabolic rate or even oxidative stress will be much more accurate.

The humoral immune response seems to require maintenance of a relatively higher night-time body temperature. Thus, activation of the immune response might be energetically costly for the organism. Inconclusive results from previous studies were possibly caused by inadequate time points for measurements chosen to estimate energetic costs. Especially for bird species, energetic costs can be camouflaged (within a 24 h energy budget), or even translated to other costs as shown here to oxidative stress. Our study, together with Abad-Gomez et al. (2013), suggests that the peak metabolic cost is separated in time from antibody release. We suggest that future research investigating the costs of the humoral immune response should take into account that metabolic costs may appear earlier than the peak of antibody release in plasma. Body temperature measurements should be considered in future studies, as it seems to be an easy way to collect continuous measurements for proxy energy use. It also opens an interesting field for future studies focused on the investigation of body temperature ranges for physiological reactions such as the immune response.

Thanks to Stanisław Bury for analysis of oxidative stress markers funded via Narodowe Centrum Nauki UMO-2016/21/N/NZ8/00959.