Immune challenge-induced oxidative damage may be mitigated by biliverdin

Changes in environmental conditions pose a two-step physiological challenge to organisms. First, organisms must adjust their physiology to directly adapt to the new environmental conditions and, second, they must ensure that the response itself minimally disrupts aspects of physiology that are indirectly related to the challenge (Burness et al., 2010). For example, immune system activation can increase levels of cytokines that activate immune cells (Magnadóttir, 2006), increase leukocyte count (Schneeberger et al., 2013), promote antibody production (Zimmerman et al., 2013), and alter a host of other aspects of the immune system that increase the likelihood of eliminating a pathogenic threat. However, while the immune system responds to the pathogenic threat, other physiological parameters, such as body temperature (Skold-Chiriac et al., 2015), are shifted from their typical homeostatic range to varying degrees. As such, immune system activation frequently results in changes to metabolism-associated metrics, including metabolic rate (Ardia et al., 2012; Martin et al., 2003), food consumption rate (Skold-Chiriac et al., 2015) and circulating nutrient levels (Stevens et al., 2017). Thus, while initiating an immune response allows the individual to eliminate or ameliorate an immunological threat, mounting an immune response also perturbs other physiological systems away from their typical homeostatic ranges.

Metabolism-related variables are not the only physiological traits that are disrupted during an immune response, as oxidative physiology may also be modified in response to an immune challenge (Hõrak et al., 2007; Casasole et al., 2016; van de Crommenacker et al., 2010), although this is not always the case (Costantini et al., 2015; Cram et al., 2015; Serra et al., 2012). Those links between immune activation and oxidative physiology tend to be most apparent during inflammation, specifically after administration of pro-inflammatory molecules [e.g. administration of lipopolysaccharide (LPS) and phytohemagglutinin (PHA); Hõrak et al., 2007; Casasole et al., 2016; van de Crommenacker et al., 2010]. In addition to modifying blood flow and cytokine levels (Zelová and Hošek, 2013), inflammation is associated with the production of reactive oxygen species (ROS; Mittal et al., 2014) that chemically interact with biological molecules, changing their chemical structure and reducing or destroying their biological function (i.e. generating oxidative damage).

Because of their reactive nature, increases in ROS production can be beneficial during an immune response, as ROS can serve as an effective tool for combating and destroying pathogens (Bogdan et al., 2000). However, increased levels of ROS also interact with the host's cells, resulting in a situation wherein immunological responses are frequently associated with increases in oxidative damage to the host (van de Crommenacker et al., 2010; Schneeberger et al., 2013; but see Cram et al., 2015). This oxidative damage can negatively affect fitness, as oxidative damage is associated with degree of ornamentation (Henschen et al., 2016), reproductive investment (Guindre-Parker et al., 2013) and survival (Pap et al., 2018).

Organisms can take an active role in ameliorating the potential increases in oxidative damage that occur during an immune challenge (Marri and Richner, 2015). Specifically, they can increase their production or mobilization of antioxidants, which would mitigate the rate of oxidative damage to self-tissues during an immune response. One molecule that may play an important physiological role in minimizing such oxidative damage, especially during immune challenges, is biliverdin, which is best known for producing many of the blue–green colors in bird eggshells (Sparks, 2011; Cassey et al., 2012). This is because biliverdin (a) is putatively an antioxidant (Mölzer et al., 2012) and (b) may have immune-modulating properties (Bisht et al., 2014; Carraway et al., 1998). The production of biliverdin occurs through the heme degradation process, wherein old or damaged erythrocytes are degraded in the spleen, liver and kidney (Bonkovsky et al., 2013). During this process, heme is released and, in the presence of the enzyme heme oxygenase (HO), degraded into carbon monoxide, a free iron molecule, and biliverdin.

HO is found in relatively large quantities in the liver, spleen and several types of leukocytes (Bonkovsky et al., 2013; Thomsen et al., 2013). In particular, the HO-1 isoform is upregulated in immune cells (Carraway et al., 1998; Rushworth et al., 2005) and even the placenta (Zhang et al., 2007) during an immune response. Thus, an increase in HO-1 expression could increase biliverdin production, with biliverdin then playing an important regulatory role during the immune challenge. For example, biliverdin could function as an antioxidant (Mancuso et al., 2012) to minimize immune challenge-induced oxidative damage, or as an anti-inflammatory molecule (Wang et al., 2010) that reduces the expression of pro-inflammatory cytokines (Bisht et al., 2014) to keep the immune system from over-responding when challenged. This potential immuno-regulatory role of biliverdin has been underexplored because methods for quantifying biliverdin at physiological levels have only recently been developed (Butler et al., 2017; Berlec and Štrukelj, 2014), and because a preponderance of immunological work is conducted in mammals, which convert nearly all biliverdin to bilirubin (Bonkovsky et al., 2013; McDonagh, 2001).

In addition to an immune response directly resulting in increased levels of biliverdin due to changes in HO expression, there may also be indirect mechanisms linking an immune response with changing biliverdin levels. For example, an immune response can be associated with an increase in circulating triglyceride levels (Frisard et al., 2010), and while there are no known direct links between biliverdin and triglyceride levels, studies in fish have indicated an indirect link between diet (Avery et al., 1992) or lipid levels (Fang and Bada, 1990) and biliverdin concentration. Additionally, there are several links between the reduced form of biliverdin (i.e. bilirubin) and aspects of metabolism in humans, including negative correlations between bilirubin levels and the diagnosis of metabolic syndrome (a phenotype associated with obesity; Li et al., 2017) or hypertriglyceridemia (Wu et al., 2011). Thus, if immune challenges affect circulating triglyceride levels, then biliverdin concentration may also change proportionally via an unidentified mechanism. Given the nascent stage of research into the physiological roles of biliverdin in non-mammalian systems (Homsher et al., 2018; Butler and Ligon, 2015), documenting such potential associations may allow for development and refinement of more specific hypotheses.

For this study, we issued two types of immune challenge to northern bobwhite quail (Colinus virginianus) and quantified variation in oxidative damage, body mass, triglyceride levels and biliverdin concentration in a variety of tissues to evaluate (1) the effects of immune challenges on all metrics, (2) the correlation between biliverdin levels and oxidative damage, and (3) correlations between biliverdin levels and both triglyceride levels and body mass. We then generated several predictions to address biliverdin's association with an immune challenge, antioxidant physiology and lipid-based metrics. If biliverdin is acting in a protective, anti-inflammatory manner during immunological responses sensu bilirubin in mammals (Mölzer et al., 2017; Wu et al., 2011) then there are two alternative predictions with respect to how immune challenges would affect circulating biliverdin levels. If HO-1 is upregulated during an immune challenge (Carraway et al., 1998; Rushworth et al., 2005), then biliverdin concentration would be higher in immune-challenged groups. However, if such increases in biliverdin production are offset by an anti-inflammatory degradation of biliverdin, then we would be unable to detect such an increase in biliverdin levels in immune-challenged birds. Regardless of treatment, if biliverdin is acting as an antioxidant in vivo, then we predict a negative correlation between oxidative damage and biliverdin concentration as increases in oxidative damage decrease biliverdin concentration. Lastly, if biliverdin has similar associations with triglyceride levels in birds to those of bilirubin in mammals (Wu et al., 2011; Li et al., 2017), then biliverdin concentration in tissues should be negatively correlated with factors such as circulating triglyceride levels and body mass.

We purchased 40 adult female northern bobwhite quail, Colinus virginianus (Linnaeus 1758), from Straub's Game Farm, Herndon, PA, USA, in December 2015. We elected to use females because we did not want to reduce our power by including both males and females and having to account for a sex effect. Upon arrival of the birds, we recorded body mass (to the nearest 0.1 g) and applied two colored leg bands for individual identification. Four to five birds were then randomly placed in cages (1.86 m×0.81 m×0.53 m) in an indoor aviary with frosted windows that allowed birds to experience a natural photoperiod, and each level of treatment (see below) was randomly applied within each cage to make sure treatment was not nested within cage. Throughout the experiment, temperature was approximately 10°C higher than ambient as the greenhouse effect of the frosted windows warmed up fresh, incoming air (low: 12.4°C; high: 27.1°C). Birds were provided with Home Fresh Multi-Flock Chick N Game Starter/Grower Crumble seed (Kent Nutrition Group, Inc., Muscatine, IA, USA) twice daily with water supplied ad libitum.

Following a week-long acclimation period, we recorded pre-experiment body mass and then assigned birds to one of four treatment groups (pre-experiment body mass did not differ by treatment; F_3,36=0.34, P=0.8). To stimulate the immune system, two groups were injected intra-abdominally with 50 μl of one of two commonly used immunostimulants, both at 2 mg ml⁻¹ (Toomey et al., 2010): LPS (N=10) or PHA (N=10), suspended in sterile phosphate-buffered saline (PBS), yielding doses of approximately 0.5 mg immunostimulant kg⁻¹ body mass. The remaining two groups were control groups; one of these groups received an intra-abdominal injection of 50 μl of PBS (N=9), while the other group experienced a sham injection process (N=11), wherein individuals experienced all aspects of injection, except that a capped syringe was gently pressed to the body. We used two control groups to evaluate the extent to which physiological metrics in non-injected individuals differ from those of vehicle-injected individuals (Adler et al., 2008). Although collecting a baseline blood sample prior to injection would have provided valuable physiological information, we opted to forgo this procedure because any inadvertent erythrocyte lysis during blood collection (e.g. a hematoma) would have increased free heme, and therefore would confound any biliverdin-based data.

After approximately 24 h, which is when physiological responses to both LPS and PHA are frequently measured (Toomey et al., 2010; Martin et al., 2006), we quantified body mass and then collected approximately 400 μl of whole blood from the alar vein using heparinized capillary tubes. Blood was placed on ice until it was centrifuged for 3 min at 10,000 g to separate erythrocytes from plasma, after which plasma was transferred to 1.5 ml screw-top tubes. Birds were then euthanized in a sealed CO₂ chamber and placed on ice. Within 6 h, we collected liver samples and the entire spleen from each bird. We quantified mass of the spleen to the nearest 0.1 mg and subsequently calculated spleen–somatic index as the ratio of spleen mass to the mean mass of the bird during the experimental period (Douxfils et al., 2011; Butler and Ligon, 2015). Plasma, liver and spleen were stored at −80°C until further analysis. All procedures were conducted with IACUC approval (approval date: 12 April 2013).

To evaluate the level of oxidative damage in the plasma of birds, we used the d-ROMs test (Diacron International, Grosetto, Italy), which detects reactive oxygen metabolites such as organic hydroperoxides using Fenton's reaction. Using the kit's reagents, we combined 20 μl of plasma, calibrator or a blank of doubly distilled water (ddH₂O) in duplicate with 200 μl of reagent in a 96-well plate, incubated the plate at 37°C for 75 min, and scanned each well at 490 nm using an Infinite M200Pro fluorimeter (Tecan US, Inc., Morrisville, NC, USA). Duplicate sample values were significantly repeatable (F_1,40=198.55, P<0.0001, r=0.99) and subsequently converted to mmol l⁻¹ H₂O₂, with larger numbers associated with a greater amount of oxidative damage (Costantini, 2016).

During lipolysis, stored triglycerides in adipocytes are hydrolyzed to produce usable fatty acids and glycerol that then travel through the blood (Duncan et al., 2007). Because triglyceride levels may be correlated to d-ROMs measurements (Pérez-Rodríguez et al., 2015), and provide valuable physiological information regarding nutritional and stress states (Neuman-Lee et al., 2015), we used a sequential colorimetric end-point assay (Sigma-Aldrich, St Louis, MO, USA) to measure circulating glycerol and triglyceride concentration (Guglielmo et al., 2002; Fokidis et al., 2012, 2011). First, we quantified free glycerol concentration by combining 5 µl of plasma with 240 µl of free glycerol reagent (Sigma-Aldrich) in a 96-well plate, incubated samples for 10 min at 37°C, and read the absorbance at 540 nm using an Infinite M200Pro fluorimeter (Tecan US, Inc.). Immediately afterwards, we added 60 µl of triglyceride reagent (Sigma-Aldrich) to the same wells and again incubated for 10 min at 37°C. We read the absorbance at the same settings and subtracted the glycerol concentration from this value to provide the plasma levels of formed (or true) triglycerides. All samples were run in duplicate and were significantly repeatable for glycerol (F_1,40=5.58, P<0.0001, r=0.70), total triglyceride (F_1,40=157.17, P<0.0001, r=0.99) and true triglyceride (F_1,40=172.63, P<0.0001, r=0.99) values. Mean values were used in subsequent analyses.

To quantify biliverdin concentration in liver and spleen, we followed a protocol (Butler et al., 2017) that uses a protein that fluoresces in the presence of biliverdin (Berlec and Štrukelj, 2014). Specifically, we homogenized liver samples (0.11±0.0051 g; mean±s.d.) and spleens (0.074±0.028 g; mean±s.d.) in 1 ml of 80:20 dimethyl sulfoxide (DMSO):ddH₂O using four 1.0 mm zirconium oxide beads for 30 s at 4000 rpm in a BeadBug Microtube Homogenizer (Benchmark Scientific, Edison, NJ, USA). To 270 μl aliquots of this homogenate, we added 30 μl of either 4 μmol l⁻¹ biliverdin in 80:20 DMSO:ddH₂O or a control blank so that we could calculate recovery for each sample, vortexed the mixture, and centrifuged samples at 12,000 g for 4 min. Plasma samples were prepared similarly, except that we used 25–50 μl (depending on availability) of plasma and prepared a final solution with a 50:50 DMSO:ddH₂O ratio. For liver and spleen samples, we then placed 50 µl of homogenate supernatant in duplicate into Greiner black polystyrene 96-well plates (Sigma; M0312-32EA) followed by 50 µl of biliverdin detection reagent (Butler et al., 2017) that had been thawed on ice and centrifuged at 12,000 g for 4 min immediately prior to use. For plasma samples, we used 40 μl of solution and 60 μl of biliverdin detection reagent to increase detection efficiency. We incubated the plate in the dark at room temperature for 60 min and measured fluorescence (excitation: 680 nm, emission: 714 nm) using an Infinite M200Pro (Tecan, Zurich, Switzerland) with Z-position set at 18,500 µm, gain set at optimal and 9 reads per well. We calculated recovery for each sample, and used this recovery value and a within-assay standard curve to calculate biliverdin concentration for each sample, and averaged duplicate values. Raw values of non-spiked samples were significantly repeatable for liver (F_1,39=89.45, P<0.0001, r=0.98), spleen (F_1,40=238.26, P<0.0001, r=0.99) and plasma (F_1,33=23.28, P<0.0001, r=0.92) biliverdin levels. All tissues were run separately. We used sample mass to calculate biliverdin concentration per gram of wet tissue, or plasma volume to calculate plasma biliverdin concentration. There were three individuals for which we did not have enough plasma to run biliverdin quantification assays; hence, there was a reduction in sample size for this metric.

Statistics were performed using SAS (v.9.3; Cary, NC, USA). To test for treatment effects, we used general linear models (GLM) with treatment as a fixed, four-level factor and dependent variables of d-ROMs, biliverdin concentration in the spleen, liver and plasma, triglyceride and glycerol levels, mean body mass (mean of pre-injection and final body mass values), change in body mass (difference between final and pre-injection body mass), percentage change in body mass (change in body mass divided by mean body mass) and spleen–somatic index. While including triglyceride levels as a covariate when the dependent variable is d-ROMs can be critically important (Pérez-Rodríguez et al., 2015), we found that triglyceride concentration did not significantly contribute to the model (F_1,35=0.32, P=0.57) and inclusion of this covariate did not change any conclusions; we thus opted to test triglyceride levels and d-ROMs separately. To test for an antioxidant function of biliverdin, we ran separate simple linear regressions between d-ROMs and biliverdin concentration in plasma, liver and spleen. To investigate the role of lipid metabolism in oxidative damage, biliverdin levels and body mass, we ran separate simple linear regressions between triglyceride levels and the above physiological metrics. We also examined whether biliverdin levels were associated with body mass by running separate simple linear regressions between biliverdin levels in plasma, liver and spleen and both mean body mass and change in body mass. In all analyses, data were transformed if they were not normally distributed; glycerol, free triglycerides, biliverdin concentration in the spleen and spleen mass were log transformed, while d-ROMs values were square-root transformed.

Treatment had no effect on biliverdin concentration in the plasma, liver or spleen, circulating concentrations of glycerol, free triglycerides, spleen mass, spleen–somatic index, mean body mass or change in body mass during the 24 h experimental period (Table 1). However, there was a significant effect of treatment on oxidative damage, with both immune-challenged groups circulating significantly higher levels of d-ROMs than either the PBS-injected or non-injected groups (Table 1, Fig. 1).

Biliverdin concentration in the plasma was negatively associated with d-ROMs in the plasma (F_1,35=5.01, P=0.032; Fig. 2), but biliverdin concentration in the liver and spleen was not associated with d-ROMs (both F_1,38<1.05, both P>0.31).

Triglyceride levels were not associated with d-ROMs (F_1,38=1.63, P=0.21), mean body mass (F_1,38=0.47, P=0.50) or biliverdin levels in spleen (F_1,38=3.09, P=0.087) or liver (F_1,38=2.92, P=0.095), or with biliverdin concentration in the plasma (F_1,35=0.00, P=0.97). However, circulating triglyceride levels were positively associated with the change in body mass over the 24 h experimental period (F_1,38=13.63, P=0.0007; Fig. 3A), including percentage change in body mass (F_1,38=15.23, P=0.0004).

Biliverdin concentration in the spleen was not associated with mean body mass, change in body mass, or percentage change in body mass (all F_1,38<3.89, all P>0.056). However, the concentration of biliverdin in both the liver (F_1,38=6.06, P=0.019; Fig. 3B) and plasma (F_1,35=6.48, P=0.016; Fig. 3C) was negatively associated with the change in body mass over the 24 h experimental period, with similar patterns for percentage change in body mass (liver: F_1,38=6.19, P=0.017; plasma: F_1,35=7.41, P=0.010). Additionally, biliverdin concentration in the plasma was positively associated with mean body mass (F_1,35=6.41, P=0.016; Fig. 4), although biliverdin concentrations in the liver were not associated with mean body mass (F_1,38=0.72, P=0.40).

When exposed to an immune challenge, either LPS or PHA, quail displayed higher levels of oxidative damage 24 h post-exposure than both non-injected and vehicle-injected controls. However, there were no other treatment-based differences in quail physiology or morphology, including no effects of immune challenge on biliverdin levels. However, we did find evidence that biliverdin may locally act as an antioxidant in vivo, because of the negative correlation between oxidative damage levels in the plasma and plasma biliverdin concentration (see below). We also uncovered multiple links between biliverdin levels in several tissues and adult body mass or change in body mass during the immune challenge, highlighting a potentially underexplored physiological role for biliverdin with respect to lipid metabolism.

The increase in oxidative damage due to both LPS and PHA administration is in line with previous work highlighting that immune challenges frequently result in increases in oxidative damage (Casasole et al., 2016; Schneeberger et al., 2013; Costantini and Møller, 2009), although many studies (including the present one) used captive or young animals, which may limit our ability to generalize this pattern to adult, free-living individuals (Cram et al., 2015). However, we found that two distinct types of immune challenge (i.e. both PHA and LPS administration) result in oxidative damage, and these patterns were consistent relative to both non-injected and vehicle-injected controls. Despite the practice of using non-injected individuals to control for immune activation during vehicle-only injection (Adler et al., 2008; Humphrey and Klasing, 2005), few studies have compared how vehicle-only injections affect oxidative physiology relative to non-injected controls. Our results support the idea that either method can serve as a control, as there were no differences between non-injected and vehicle-injected individuals in any metric we captured.

We found a significant negative correlation between oxidative damage and biliverdin concentration in plasma, suggesting an antioxidant function of biliverdin may be possible in vivo. However, even within an individual, this pattern existed only in the plasma, and there were no correlations between oxidative damage in the plasma and biliverdin concentration in other tissues. This tissue-dependent correlation suggests that if biliverdin does function as an antioxidant, it would be at a local scale and biliverdin concentration in one tissue would not necessarily predict levels of oxidative damage elsewhere in the body. This finding is consistent with previous work that failed to detect a correlation between biliverdin concentration in the bile and oxidative damage in other tissues (Butler and Ligon, 2015). In other words, the degree to which biliverdin's physiological roles appear to be localized, rather than systemic, may be critical for understanding both the physiological role of biliverdin and the degree to which these physiological properties inform ecological or evolutionary hypotheses. Specifically, interest in the physiological role of biliverdin in avian species primarily originates from its potential to signal female antioxidant capacity to mates (i.e. the sexually selected egg color hypothesis) via eggshell coloration (Moreno and Osorno, 2003). Under this hypothesis, a female may signal her antioxidant capacity by depositing an antioxidant (i.e. biliverdin) into eggshells, and males would adjust their paternal care accordingly. While our data are consistent with the assumption that biliverdin has an antioxidant function in vivo, there is less information on the extent to which biliverdin derived from one tissue (the oviduct) is associated with systemic antioxidant capacity.

We predicted that if biliverdin displayed similar patterns of triglyceride-associated metrics to bilirubin (i.e. high levels of bilirubin are associated with lower triglyceride levels; Cho et al., 2016), individuals with more biliverdin-rich tissues should display traits associated with reduced levels of circulating triglycerides, including having a lower mean body mass, and potentially losing more mass during an immune challenge. Our data provide partial support for this prediction, as individuals that circulated greater amounts of biliverdin did lose more body mass during the 24 h experimental period; however, these individuals also had a larger mean body mass. There are two important considerations when evaluating this result, which was opposite to our predictions. First, we did not quantify individual food consumption prior to or during the 24 h experimental period, which could have important explanatory power regarding body mass. Second, our predictions were derived from work with humans exhibiting metabolic syndrome, which is characterized by factors such as obesity, inflammation and hypertension (Li et al., 2017; Vanella et al., 2014). These human individuals exhibit both lower levels of bilirubin (the reduced form of biliverdin; Li et al., 2017) and lower activity of HO (Vanella et al., 2014), which is why we predicted that lower levels of biliverdin might be associated with higher levels of triglyceride-associated metrics. However, metabolic syndrome is also associated with overweight individuals circulating higher levels of triglycerides (Wu et al., 2011), and because we found that the largest birds did not have significantly higher levels of circulating triglycerides, humans exhibiting metabolic syndrome may not be an appropriate group from which to generate predictions between body mass and circulating biliverdin levels.

The correlations between biliverdin, triglycerides and body mass did align with our predictions with respect to change in body mass over the 24 h experimental period. During this time, birds that increased in body mass also circulated higher levels of triglycerides and had lower concentrations of biliverdin in both the liver and plasma. While these patterns generally fit the overall predictions derived from the relationship between bilirubin and metabolic syndrome (e.g. overweight individuals have lower levels of bilirubin), there are several important distinctions. First, the metabolic syndrome literature is with regard to mean body mass (e.g. Li et al., 2017), not change in body mass over a short time frame, including time periods that are shorter or longer than 24 h. Second, many of the associations between bilirubin and body mass are linked to triglyceride levels (Cho et al., 2016), yet our results demonstrate that while change in body mass is related to both triglyceride and biliverdin levels, triglyceride concentration does not directly correlate with biliverdin concentration in either the liver or plasma. Thus, it is likely that some other physiological pathway is the causal mechanism driving change in body mass, HO expression (and therefore biliverdin levels) and triglyceride mobilization or transport. For example, hormones such as leptin have been associated with changes in bilirubin levels (Liu et al., 2015) and HO activity (Kupai et al., 2015), which may affect both circulating triglyceride and biliverdin levels. While the putative mechanisms underlying the relationships between body mass, triglyceride levels and biliverdin concentration are poorly described in non-mammalian taxa, we report several correlations that suggest that further direct investigations (e.g. controlled food studies over various lengths of time) into the physiological interplay between biliverdin and triglyceride-associated metrics in non-mammalian species could be an area of fruitful research.

We would like to thank E. Reynolds and A. Thevinin for helpful input during experiment design and analysis, D. Husic, J. Hines and J. Cannon for logistical support, and A. Little, J. Minnick, M. Rossi and S. Bociulis for assistance with sample analysis. We would also like to thank three anonymous reviewers for their comments and suggestions.