Effects of short-term multi-pollutant exposure on the oxidative stress status of captive songbirds
Abstract
- The human influence on Earth's ecosystems is omnipresent. Artificial light at night (ALAN), anthropogenic noise, and air pollution are inherent features of human activities and infrastructure and pose novel environmental challenges to urban-dwelling wildlife.
- So far, most of the studies investigating the impacts of exposure to urban pollutants on animals have either investigated the effects of urban environments per se or of single pollutants. However, urban pollutants co-occur, and interactive effects may arise when acting in combination, but we lack a deeper understanding of the effects of combined exposures.
- Here, we experimentally exposed captive zebra finches Taeniopygia guttata in a full-factorial design to increased levels of ALAN, anthropogenic noise and/or soot and measured oxidative stress status in blood before and after a 5-day exposure.
- We found that the combined exposure to ALAN and noise led to a positive synergistic effect (higher levels than the sum of individual effects) on the antioxidant glutathione and a negative synergistic effect (lower levels than the sum of individual effects) on the levels of oxidative damage, measured as the concentration of reactive oxygen metabolites. Soot had no effect on the avian oxidative stress status in the blood immediately after the exposure, neither singly nor in combination with other pollutants.
- To conclude, our results indicate that a combination of stressors can have complex non-additive interactive effects on oxidative stress status after a short-term exposure. Surprisingly, a combined exposure to ALAN and anthropogenic noise leads to a stronger antioxidant response that seems to prevent oxidative damage than exposure to only one of the stressors. Whether the increased antioxidant defence entails any long-term costs remains to be determined in future studies.
Read the free Plain Language Summary for this article on the Journal blog.
1 INTRODUCTION
The anthropogenic impact on Earth is omnipresent (Vitousek et al., 1997). The expansion of urbanized areas around the globe has led to unprecedented destruction and fragmentation of natural habitats, threatening ecosystems and biodiversity (Grimm et al., 2008; Seto et al., 2012; United Nations Department of Economic and Social Affairs Population Division, 2019). Studies investigating the impact of urban environments on wildlife have found distinct phenotypic and genotypic differences between urban and rural populations of the same species (e.g. Alberti et al., 2017; Partecke et al., 2006; Salmón et al., 2021; Sih et al., 2011; Watson et al., 2017; Weaver et al., 2018). The main drivers of such differences between urban and rural populations are thought to be increased levels of chemical pollution, anthropogenic noise and artificial light at night (ALAN), along with human presence and differences in food availability and quality. To date, studies have investigated the effects of urban environments per se on wildlife, identifying profound impacts on life-history traits including, for example, body size and condition (e.g. Biard et al., 2017; Cissé et al., 2017; Iglesias-Carrasco et al., 2017; Meillère et al., 2015), clutch size and reproductive success (e.g. Eeva & Lehikoinen, 1995; Halfwerk et al., 2011; Samuelson et al., 2018; Vaugoyeau et al., 2016), physiological traits such as immune function and oxidative stress (e.g. Bailly et al., 2016; Isaksson, 2010; Ziegler et al., 2021) and behaviours such as distress calls and general activity behaviours in response to human presence (Senar et al., 2017; Weaver et al., 2018).
The urban environment presents a complex mixture of co-occurring pollutants, yet the interactions between these have scarcely been investigated (Halfwerk & Jerem, 2021; Ross et al., 2011; Sørensen et al., 2014; Swaddle et al., 2015; Weber, 2009). A thorough understanding of the dynamics between different pollutants is particularly important when the combination of two stressors leads to non-additive effects, where synergistic interactions (interactive effects greater than the sum of individual effects) or antagonistic interactions (interactive effects that are less than the sum of individual effects) might emerge (Halfwerk & Jerem, 2021; Piggott et al., 2015; Todgham & Stillman, 2013). The combined effect of multiple pollutants may vary from the singular exposure effects for several reasons. First, different stressors are perceived through a variety of sensory modalities, such as vision, hearing or olfaction. Second, the received stimuli may then be processed via different neural and physiological pathways, which can lead to a multitude of behavioural and physiological responses (Hale et al., 2017; Halfwerk & Slabbekoorn, 2015; Munoz & Blumstein, 2012). The potential synergetic effects of stressors can have implications for conservation management strategies. For example, if one pollutant has its main impact through synergistic effects with other pollutants, a reduction of this pollutant could have an unpredicted strong positive effect. However, if antagonistic interactions are present, the reduction of one particular pollutant may be effectless or even counterproductive (Dominoni, Halfwerk, et al., 2020; McNaughton et al., 2021). The few studies investigating anthropogenic multi-stressor effects have yielded inconclusive results (reviewed in Halfwerk & Jerem, 2021). For instance, Dominoni, Smit, et al. (2020) showed that experimental exposure to either ALAN (white illumination) or traffic noise alone had opposing effects on the activity of captive adult great tits Parus major of urban origin, with ALAN increasing activity and noise (from now on noise refers to anthropogenic noise such as traffic and construction noise) decreasing activity. However, when great tits were exposed to both pollutants simultaneously, they found synergistic effects on night-time and total activity and an antagonistic interactive effect on daytime activity (Dominoni, Smit, et al., 2020). In contrast, no interactive effects of ALAN and anthropogenic noise exposure were found on great tit nestling physiology (Raap et al., 2017). In general, the studies that have tested for potential interactive effects between multiple anthropogenic stressors have revealed high context dependency and a need to further investigate multi-stressor responses.
In terms of physiological responses to a wide repertoire of different stressors, the redox system has been put forward as a key target—being involved in scavenging pro-oxidative agents and maintaining a homeostatic state between antioxidant defences and oxidative damage (Halliwell & Gutteridge, 2015; Isaksson, 2015; Isaksson & Bonier, 2020). The management and regulation of oxidative stress status has been postulated to be an important mechanism in creating phenotypic variation between different species and populations (Monaghan et al., 2009). Increased production of reactive oxygen species (ROS)/reactive nitrogen species due to increased energetic demands, disease or an external pro-oxidative assault must be met with adequate antioxidant defences to minimize oxidative damage to macromolecules. If the accumulation of damage cannot be impeded, oxidative stress occurs, which has negative impacts on survival and fitness (e.g. Costantini, 2014; Marasco et al., 2017; Wiersma et al., 2004). The aforementioned anthropogenic pollutants have been shown to either directly or indirectly affect the oxidative status of individuals (Costantini et al., 2014; Isaksson, 2010; Salmón, Stroh, et al., 2018). Air pollutants, such as nitric oxides, ozone and soot, can directly act as pro-oxidants, causing oxidative damage (Isaksson, 2010, 2015). Exposure to ALAN is also known to decrease levels of the hormone melatonin (de Jong et al., 2016; Dominoni et al., 2013; Grubisic et al., 2019; Moaraf et al., 2020; Ziegler et al., 2021), which is not only a key mediator of circadian rhythms but also an important antioxidant (reviewed in, e.g. Carrillo-Vico et al., 2005; Reiter et al., 2010). ALAN exposure could therefore directly enhance the impact of oxidative stress via the reduction of melatonin levels at night. Indirectly, exposure to ALAN, but also anthropogenic noise, has been linked to impaired immune responses (Bedrosian et al., 2011; Moore & Siopes, 2000; Oishi et al., 2006; Zheng & Ariizumi, 2007; Ziegler et al., 2021) and glucocorticoid stress responses (Bonier, 2012; Davies et al., 2017; Kleist et al., 2018), both of which are known to affect the oxidative stress status (Costantini et al., 2011; Majer et al., 2019).
In the present study, we disentangle the relative effects of three main anthropogenic pollutants—ALAN, noise and air pollution—on the oxidative stress status of captive zebra finches Taeniopygia guttata. ALAN was represented by illumination of warm white light (~13 lx), anthropogenic noise by recordings from heavy traffic close to a city road (~70 db) and air pollution by generation of soot (~100 μg/m3 daily average). We utilised a full-factorial experimental design and a multi-marker approach to gain a deeper understanding of the impacts of these pollutants on avian oxidative stress status in blood. Specifically, we measured two markers of the antioxidant defence system: the non-enzymatic antioxidant capacity (OXY) and the total concentration of glutathione (tGSH). The latter is considered one of the most important antioxidants in biological systems (Surai, 2002). As a marker for oxidative damage, reactive oxidative metabolites (ROMs) were measured. In wild populations of urban birds, the severity of oxidative stress has been found to vary, with some populations only showing a changed antioxidant level and no damage, whereas other populations show a clear increase in oxidative damage along with an impaired antioxidant response compared to their non-urban-dwelling counterparts (e.g. Amri et al., 2017; Giraudeau & McGraw, 2014; Herrera-Dueñas et al., 2017; Salmón, Stroh, et al., 2018). Such variation in responses suggests a high context dependency in terms of, for example, intensity and duration of exposure. Hence, our predictions for the directions of interactive effects depend on the effective magnitude that each pollutant individually has on oxidative stress. Here, different scenarios can be outlined as suggested by Isaksson (2020). For the single exposure groups, we predicted that all three pollutants would influence oxidative stress status but through different pathways (see Figure 1 for a simplified illustration of the mechanistic underpinnings). Furthermore, if a single pollutant exposure causes a mild physiological stress, we predicted only an increase in the antioxidant response. If a single pollutant exposure is perceived as more stressful, we predict an increase in both antioxidants and oxidative damage. Lastly, if a single pollutant exposure causes acute high stress, we predict a collapse of the antioxidant response and an increase in oxidative damage (Isaksson, 2020). Furthermore, for the exposure to multiple pollutants, we asked whether the simultaneous exposure to either two of the three pollutants or to all three pollutants had additive, synergistic or antagonistic effects on the oxidative stress status markers. However, no a priori prediction can be put forward here given the complex underpinnings of the response (see Figure 1).
2 METHODS
2.1 General procedure
This experiment was performed during April–May 2017, using captive zebra finches. During non-experimental periods, the birds were housed in same-sex groups in outdoor aviaries at the ecological station of the Department of Biology. The study was approved by the Malmö/Lund ethical committee (permit no. M108-16).
The experimental design consisted of exposure to three different urban pollutants, namely, ALAN, anthropogenic noise and soot, in a 2 × 2 × 2 full-factorial design (for details of exposure treatments see below and see Table S1). This design led to eight separate treatment groups of 39–43 birds per treatment group (approximately 50:50 female: male, Table S2), with one control group (i.e. no experimental exposures), three ‘single’ pollutant groups (ALAN, noise or soot), three groups with combined exposure to two pollutants (ALAN–noise, ALAN–soot or noise–soot) and one group where all three pollutants were combined (ALAN–noise–soot). Treatments were carried out consecutively, with a new treatment group started every week, including new birds, and the exposure lasted for 5 days for each group (procedure details are given below and order and sample sizes of treatment groups in Tables S1 and S2).
Eight days prior to the start of each exposure treatment (see below), birds were caught from the outdoor aviaries and transferred to indoor aviaries (2 × 2 × 2.2 m) to acclimatize to indoor temperatures (26 ± 1°C) and light regimes (15:9 h light:dark). Males and females were kept in separate aviaries. They were randomly distributed into same-sex groups of five to six birds among eight cages (97 × 52 × 83 cm). The day prior to the start of the exposure, the birds were transferred to the Aerosol Laboratory at the Faculty of Engineering and kept in the same group constellation as above. Food (seed mixture; Exoten, Benelux, Kinlys Group, Belgium) and water were provided ad libitum. Additionally, each cage was provided with a cuttlebone and every day with a fresh slice of cucumber. The cages were standing on a shelf in an exposure chamber (22 m3 volume, 9 m2 floor area; more details about the exposure chamber can be found in Isaxon et al., 2013). The inner walls of the chamber were equipped with 1 mm thick stainless-steel plates, which were insulated with 90 mm thick material to provide thermal insulation and soundproofing. The air supply of the chamber was passed through an active carbon filter and an ultra-low penetration air filter for the removal of gaseous and particulate air pollutants. An air-conditioning unit maintained an air temperature of 25.3 ± 2.4°C and a relative humidity of 25.1% ± 5.5%. A stand-alone fan provided further air circulation and particle distribution in the chamber. The illumination during daytime hours was provided by light emitting diode (LED) panels, with a light temperature of 4000 K, placed 20 cm above the cages. Daytime baseline illuminance between 08:00 and 17:00 was 3200 ± 690 lx (mean ± standard deviation [SD]), measured at two points for each height of 10 and 40 cm inside each cage (total of 32 measurement points). From 07:00 to 08:00 and 17:00 to 18:00, we simulated dawn and dusk by only operating half of the LED panels. During the dark phase (except for treatment groups involving ALAN), a small night-light lamp simulated moonlight of <0.1 lx. Background noise levels during weeks where no-noise treatment was applied were 42–45 dBA, mainly caused by the ventilation system and internal fan (when no birds were in the chamber). During the same weeks of no-noise treatment, but with birds in the chamber, noise levels were 68.0 ± 1.1 dBA (mean ± SD) during the day (09:00–15:00) and 61.1 ± 1.1 dBA during the nights where no ALAN treatment was applied (i.e. weeks of treatment groups: control and soot) and 63.1 ± 1.0 dBA during the weeks where ALAN exposure was applied (i.e. ALAN and ALAN–soot treatment group weeks). The difference in night-time noise levels between nights with and without additional ALAN treatment is due to the birds being more active during nights with ALAN exposure (see Section 4).
2.2 Exposure treatments
The ALAN treatment consisted of an LED lamp (2800 K, warm white; OSRAM) placed about 1 m to the side of the cages at a height of 1.6 m measured from the ground, emitting from 18:00 to 07:00, 13.7 ± 4.3 lx (mean ± SD; measured at the same measurement points as the daylight measurements described above).
The initial noise recordings took place in an urban street environment (50 km/h speed limit) on a rainy day, on a busy street in Lyngby, Denmark (Lat: 55.769816, Long: 12.500682). The measurement was made in stereo, using B&K free-field microphones and a digital recording device, which were placed 1.5 m above the ground. The excerpt was edited to assure continuous and natural change of the traffic noise level (i.e. no long silent periods) and to remove non-traffic sounds such as speech and baby cries. The edited noise file (15 min) ran in a loop during the noise treatment (the sound file is available in the Supporting Information). The analyses of the sound file gave L_Z = 75.0 dB (unweighted) and L_A = 64.2 dB (A-weighted) in the left channel, and L_Z = 74.0 dB (unweighted) and L_A = 63.2 dB (A-weighted) in the right channel. Using a time interval of 35 ms, the A-weighted result was 44.7 dBmin/76.5 dBmax for the left channel and 44.6 dBmin/75.3 dBmax for the right channel. During the experiment, two loudspeakers were positioned on the chamber floor. Recordings were running from 08:00 to 18:00, which resulted in a noise level of 70.7 ± 0.4 dBA (mean ± SD). During nights (19:00–07:00) without additional ALAN exposure, the noise level was on average 61.5 ± 0.9 dBA (mean ± SD) and during nights with additional ALAN exposure, the noise level was on average 64.9 ± 1.6 dBA (mean ± SD). An integrating sound level meter (Quest Technologies, USA, model 2500) was used for the noise measurement and the noise was measured as LAeq (A-weighted, equivalent continuous sound level). A-weight for the treatment was chosen based on birds' frequency sensitivity (Beason, 2004), which is similar to humans. To simulate the morning and evening gradual increase and decrease, respectively, of a city rush-hour noise, the computer was programmed to slowly ramp the noise recording up and down between the hours 07:00–08:00 in the morning and 18:00–19:00 in the evening.
To simulate urban air pollution, a miniature combustion aerosol standard (miniCast) soot generator (model 5201C; Jing Ltd., Zollikofen, Switzerland) was used as a source for soot particles. The soot generation operated between 8:00 and 16:00. During the remaining time, birds were exposed to filtered air, which was the same as in the experimental periods where no soot exposure was involved (see above). The soot generation set-up is described in detail in Malmborg et al. (2019). In short, the miniCast was operated at a custom setting mode OP1 (according to the manual), corresponding to using a fuel flow (C3H8) of 60 mL min−1 and a co-flow of oxidant air of 1.55 L min−1. After the combustion zone, an additional dilution of 20 L min−1 was used before the aerosol was led into the exposure chamber. The airborne particle levels were analysed on-line with a tapered element oscillating microbalance (TEOM, R & P Inc., model 1400a), a scanning mobility particle sizer (Electrostatic Classifier model 3071 TSI Inc., CPC model 3775 TSI Inc) and a seven-wavelength aethalometer (Model: AE33; Magee Scientific Corp., Berkeley, USA). Soot typically consists of primary particles building up larger aggregates (Park et al., 2003). The primary soot particles generated for this experiment were determined to be 25 nm in size (transmission electron microscopy), which is similar to that of diesel soot (Rissler et al., 2013). These primary particles build up larger soot aggregates forming a mode of particles peaking around ~130 nm with respect to number (details of the number size distribution can be found in Figure S1 and Table S3). The generated soot was mainly composed of elemental carbon, and the ratio between the organic carbon and the elemental carbon was 5%–10%, which is similar to that reported earlier for the same specific burner and operation mode (Török et al., 2018). Mass of the particles <2.5 μm in diameter (i.e. PM2.5) in the air was measured by a TEOM (R & P Inc., model 1400a) and soot mass by an aethalometre (AE33; Magee Scientific) (Drinovec et al., 2015), respectively. PM2.5, averaged over all 4 weeks during soot exposure treatments (i.e. soot, noise–soot, ALAN–soot and ALAN–noise–soot treatment groups), measured between 08:00 and16:00, was 276 ± 6 μg m−3, while the weekly average of PM2.5 (measured over 24 h) was lower at 109 μg m−3 (Table S4). This difference originates from the soot generation being turned off during the night-time. The corresponding soot mass according to the aethalometre was 288 ± 16 μg m−3 during daytime, with a weekly average of 100 μg m−3 (Table S4). This pollution level corresponds to a megacity, the size of Rio de Janeiro, with approximately 6.7 million inhabitants (Gurjar et al., 2008). The small difference between the mass measured with the TEOM (measures mass of all particles <2.5 μm regardless of composition) and the aethalometre (measures mass of soot only) is within the instrumental uncertainties, and by comparing the mass reported by the two instruments, we conclude that PM2.5 is dominated by the soot generated and that the contribution to PM2.5 from the birds themselves is insignificant. Further information on the concentrations and characteristics of the generated soot can be found in the Supporting Information.
2.3 Sampling
In the morning 5 days prior to the start of the exposure treatment (i.e. 4 days into the acclimatization phase; Table S1), we took a baseline blood sample from the jugular vein of each bird. A second blood sample was taken at the end of the fifth (and last) day of exposure (Table S1). All samples were collected in heparin-coated tubes and kept on ice for about 20 min until centrifugation. The samples were spun at 4°C at 4000 rpm for 10 min to separate plasma from red blood cells (RBCs) and immediately snap-frozen in liquid nitrogen and later stored at −80°C. On both sampling occasions, we measured body mass with a Pesola balance to the nearest 0.1 g.
2.4 Oxidative stress status assays
2.4.1 Oxidative defence
Two biomarkers of the antioxidant defence system were analysed. First, tGSH content in RBCs was measured following Baker et al. (1990), adapted for a microplate reader (Isaksson et al., 2005). About half of the samples were analysed in late 2017 and the other half in late 2020. We accounted for possible storage effects by including ‘batch’ as a factor in the statistical model (see Section 2.5). Briefly, about 8 μg of RBC were diluted (1:1) in phosphate-buffered saline. Then, 8 μL of the dilution were mixed with 16 μL of 5% 5-sulfosalicylic acid and left on ice for 10 min. After incubation, the samples were centrifuged at 4°C and 10,000 rpm for 10 min. Ten microliters of the supernatant were then diluted with 390 μL of GSH buffer (143 mM NaH2PO4, 6.3 mM EDTA, pH 7.4). Of this mixture, we transferred 20 μL to a microplate and added 170 μL of reaction solution (10 mmol L−1 5,5′-dithiobis 2-nitrobenzoic acid and 0.34 U of glutathione reductase in GSH buffer). The reaction was started by adding 34 μL of 2 mmol L−1 nicotinamide adenine dinucleotide phosphate. After shaking the plate, the absorbance was measured every 30 s for 5 min at 412 nm (FLUOstar Omega, BMG Labtech). On each plate, we ran a serial dilution standard curve with six points, ranging from 3.12 μM to 100 μM. All samples, as well as standards, were run in duplicates, and both samples from an individual (i.e. baseline and exposure sample) were run on the same plate. The intra- and inter-assay coefficients of variation were 1.9% and 7.2%, respectively.
Second, non-enzymatic antioxidant capacity was measured with the OXY-adsorbent test (Diacron International, Grosseto, Italy). This assay quantifies the in vitro reaction of the non-enzymatic antioxidant barrier in the plasma with a strong oxidant (HOCl). We followed a previously published protocol by Costantini et al. (2006). In short, we diluted 2 μL of plasma 1:100 with ddH2O. Two microliters of the diluted plasma were mixed with an HOCl solution and incubated at 37°C for 10 min while shaking at 500 rpm. Afterwards, 2 μL of chromogen (N,N-diethyl-p-phenylenediamine) was added, and the absorbance was measured at 490 nm (FLUOstar Omega, BMG Labtech). Measurements are expressed as millimoles of neutralized HOCl. All samples, including standards and control samples, were run in duplicates, and both samples from an individual (i.e. the baseline and post-exposure samples) were run on the same plate. Additionally, samples were randomized over plates across treatment groups. Due to the time sensitivity of the assay, we divided each plate into three separate runs of 10 samples each, each having a standard curve and a control sample. The intra- and inter-assay coefficients of variation were 1.5% and 12.3% (calculated from means of three control samples per plate), respectively.
2.5 Oxidative damage
The concentration of ROMs in the plasma was assessed with the dROM assay (Diacron International, Grosseto, Italy). This test measures the concentration of hydroperoxides, which are oxidative damage compounds generated in the early phase of an oxidative cascade and indicate oxidative damage to lipids, proteins and nucleic acids. We followed the protocol previously published by Costantini et al. (2006). Briefly, we diluted 4 μL of plasma in 200 μL of an acidic buffer solution (1:100 ratio of chromogen: buffer). The samples were then incubated at 37°C for 75 min. After incubation, absorbance was measured at 490 nm in a FLUOstar Omega (BMG LABTECH) microplate reader. ROM concentration was calculated using a 6-point standard curve and blank correction of the absorption values. All samples, including standards and control samples, were run in duplicates, and both samples from an individual (i.e. baseline and post-exposure sample) were run on the same plate. Additionally, samples were randomized over plates across treatment groups. The intra- and inter-assay coefficients of variation were 4.8% and 13.1%, respectively.
2.6 Statistical analysis
We analysed all data using R version 4.1.0 (R CoreTeam, 2020). We ran separate linear mixed models using the ‘lmerTest’ package (Kuznetsova et al., 2017) for all three oxidative status markers (NtGSH = 302, NOXY = 292, NROMs = 288, sample sizes vary due to plasma and RBC volume limitations; details on samples sizes per treatment group in Table S2). For all models, we used the post-exposure concentrations of the respective marker as the dependent variable and included the concentrations of the baseline samples and body mass (measured at the last day of the exposure) as covariates and sex (male/female) as a fixed factor. Additionally, we included the exposure treatments as binary fixed effects: ALAN (dark/ALAN), noise (no noise/noise) and soot (no soot/soot) and the two-way and three-way interactions between exposure treatments. We mean-centred body mass and baseline concentrations for better interpretation of the model estimates (Schielzeth, 2010). Additionally, we included batch (first/second) in the tGSH model as a fixed effect due to the later laboratory analysis of half of the samples for tGSH. We initially included experimental cage (1–8) and plate number as random effects. However, the models for ROMs and OXY indicated singularity issues when including experimental cage or plate, respectively, as a random effect. This was likely due to close to zero variance of the random effect terms. Therefore, we did not include experimental cage in the ROMs model or plate in the OXY model as random effects. We used Pearson's correlation tests to analyse the association between the three different markers of oxidative stress status within individuals.
We used a stepwise backwards model selection approach, eliminating all model terms with p > 0.05, but always retaining the three exposure treatments in the models. Statistics of the dropped terms can be found in Table S5. In the case of significant interactions, we used the package ‘emmeans’ (Lenth et al., 2020) for pairwise post hoc testing. Plots of the residuals of the full initial models along Q-Q plots for OXY can be found in the Supporting Information (Figure S2). We log-transformed the OXY post-exposure concentrations to meet model assumptions. Unless otherwise stated, values shown are estimated marginal means ± SE calculated using the package ‘emmeans’ (Lenth et al., 2020).
We classified significant interactions between the exposure treatments in the final models by comparing the single exposure additive effects with the observed effect of the combined pollutant exposure (interaction term following, Hale et al., 2017; Piggott et al., 2015; Figure 2). We used the parameter estimates of the post hoc pairwise contrasts between the control group and the respective main or interactive effects from the final model by emmeans (Lenth et al., 2020). To estimate the additive effects, we summed the two contrast estimates between the control group and the main effects that are involved in the interaction in focus. To describe the observed interactive effect, we used the contrasts between the control group and the interaction term from the post hoc pairwise comparison. We classified the interaction between two treatments as synergistic when the effect size was larger than and/or in the opposite direction compared with the additive effect. The latter case can also be termed ‘mitigating synergism’ or ‘reversal’ due to the two main effects having the same direction but their combined effect showing the opposite direction. We classified an interaction as antagonistic when the effect size was less than the additive effect. The direction of an interactive effect was classified as either positive or negative synergistic or antagonistic depending on the direction of the main and interactive effects following (see Hale et al., 2017; Piggott et al., 2015 for details; Figure 2). All raw data are archived in the Dryad Digital Repository (https://doi.org/10.5061/dryad.hx3ffbgqs; Isaksson et al., 2025).
3 RESULTS
3.1 Antioxidant defence (tGSH)
We found a significant interactive effect between ALAN and noise exposure on tGSH concentration (F1,279.54 = 3.93, p = 0.048; Table 1; Figure 3a). The combined exposure to ALAN and noise was classified as a positive synergistic effect on tGSH levels (Figure 3b), as the interactive effect was larger than the calculated additive effect. Post hoc pairwise comparisons showed that a simultaneous exposure to ALAN and noise significantly increased tGSH levels by 5.4% (17.6 ± 0.3 μmol L−1) compared to birds that were not exposed to ALAN or noise (16.7 ± 0.3 μmol L−1; p = 0.023; Figure 3a). Furthermore, post hoc pairwise comparisons showed that birds that were simultaneously exposed to noise and ALAN had 6.7% higher tGSH levels than birds that were only exposed to noise alone (noise: 16.5 ± 0.3 μmol l−1; p = 0.002, Figure 3a). All other pairwise comparisons were p ≥ 0.15. None of the other interactions between the exposure treatments were significant for tGSH (all p ≥ 0.06; Table S5). Baseline tGSH concentrations predicted post-exposure tGSH concentrations, with birds having higher concentrations before the exposure also having higher tGSH levels after the 5-day exposure (F1,290.96 = 123.38, p < 0.001; Table 1). Also, the batch of analyses influenced the tGSH (F1,12.80 = 5.67, p = 0.034), with higher levels in the second (and later) batch. We did not find an effect of either soot treatment (F1,278.17 = 1.36, p = 0.24; Table 1), body mass (χ2 = 2.07, p = 0.15; Table S5) or sex (χ2 = 1.39, p = 0.24; Table S5) on tGSH levels.
| tGSH concentration | |||||
|---|---|---|---|---|---|
| Random effects | Variance | SD | No. of groups | No. of observations | |
| Plate | 0.628 | 0.792 | 16 | 304 | |
| Experimental cage | 0.127 | 0.357 | 8 | ||
| Residuals | 3.880 | 1.970 | |||
| Fixed effects | Estimate ± SE | SS | df | F | p |
|---|---|---|---|---|---|
| Intercept | 17.39 ± 0.399 | ||||
| Centred pre-tGSH | 0.566 ± 0.051 | 478.67 | 1, 290.960 | 123.380 | <0.001 |
| Soot (soot) | −0.266 ± 0.228 | 5.28 | 1, 278.172 | 1.36 | 0.24 |
| ALAN (ALAN) | 0.245 ± 0.324 | 36.22 | 1, 279.807 | 9.34 | 0.002 |
| Noise (noise) | −0.223 ± 0.321 | 3.96 | 1, 277.731 | 1.02 | 0.31 |
| Batch (first) | −1.098 ± 0.461 | 21.99 | 1, 12.804 | 5.67 | 0.034 |
| ALAN × noise | 0.901 ± 0.457 | 15.24 | 1, 279.537 | 3.93 | 0.048 |
| Antioxidant capacity (OXYa) | |||||
|---|---|---|---|---|---|
| Random effects | Variance | SD | No. of groups | No. of observations | |
| Experimental cage | 0.001 | 0.027 | 8 | 292 | |
| Residuals | 0.029 | 0.171 | |||
| Fixed effects | Estimate ± SE (back-transformed) | SS | df | F | p |
|---|---|---|---|---|---|
| Intercept | 5.114 ± 0.022 (166.334 ± 1.023) | ||||
| Centred pre-OXY | 0.004 ± 0.0003 (1.004 ± 1.0002) | ||||
| 5.901 | 1, 283.93 | 201.95 | <0.001 | ||
| Soot (soot) | −0.031 ± 0.020 (−0.970 ± 1.020) | ||||
| 0.069 | 1, 280.25 | 2.36 | 0.13 | ||
| Noise (noise) | −0.003 ± 0.020 (−1.003 ± 1.020) | ||||
| 0.0007 | 1, 281.21 | 0.025 | 0.88 | ||
| ALAN (ALAN) | −0.015 ± 0.020 (−1.015 ± 1.020) | ||||
| 0.017 | 1, 281.07 | 0.58 | 0.45 |
| Oxidative damage (ROMs) | |||||
|---|---|---|---|---|---|
| Random effects | Variance | SD | No. of groups | No. of observations | |
| Plate | 0.0001 | 0.014 | 18 | 288 | |
| Residuals | 0.019 | 0.137 | |||
| Fixed effects | Estimate ± SE | SS | df | f | p |
|---|---|---|---|---|---|
| Intercept | 0.488 ± 0.020 | ||||
| Centred pre-ROM | 0.490 ± 0.052 | 1.648 | 1, 172.36 | 88.31 | <0.001 |
| Centred post body mass | −0.015 ± 0.006 | 0.105 | 1, 275.48 | 5.60 | 0.019 |
| Sex (male) | 0.054 ± 0.016 | 0.202 | 1. 280.00 | 10.80 | 0.001 |
| Soot (soot) | 0.003 ± 0.016 | 0.001 | 1. 271.06 | 0.04 | 0.84 |
| ALAN (ALAN) | 0.029 ± 0.023 | 0.017 | 1. 271.56 | 0.93 | 0.34 |
| Noise (noise) | 0.038 ± 0.023 | 0.003 | 1. 265.25 | 0.17 | 0.68 |
| ALAN × noise | −0.089 ± 0.032 | 0.143 | 1. 263.80 | 7.66 | 0.006 |
- Note: Final models were obtained by backwards elimination of terms with p > 0.05, but always retaining ALAN, soot and noise treatment in the model. Significant terms are highlighted in bold. For the ALAN treatment, the dark-night birds are used as the reference point; for the soot treatment, the non-soot exposed birds are the reference point and for the noise treatment, the non-noise-exposed birds are the reference point. For batch, the first batch is the reference point and for sex, females are the reference point. Continuous variables are mean centred. Estimates in brackets are back calculated from a log-transformation.
- Abbreviations: ALAN, artificial light at night; df, degrees of freedom; OXY, antioxidant capacity; ROM, reactive oxidative metabolite; SE, standard error; SS, sum of squares; StDev, standard deviation; tGSH, total glutathione levels.
- a For normality, the OXY values were log-transformed; hence, the estimate is also based on the log-transformed values.
3.2 Plasma non-enzymatic antioxidative capacity (OXY)
None of the exposure treatments or the interactions influenced the plasma antioxidant capacity (soot: F1,280.25 = 2.36, p = 0.12; ALAN: F1,281.07 = 0.58, p = 0.45; noise: F1,281.21 = 0.03, p = 0.88; Table 1). Neither sex (χ2 = 1.61, p = 0.21; Table S5) nor body mass (χ2 = 0.10, p = 0.75; Table S5) influenced the antioxidant capacity. Plasma OXY levels before the exposure treatments significantly predicted levels after the exposure, with birds having higher antioxidant capacity levels before the exposure treatment and also having higher antioxidant capacity levels post-exposure (F1,283.93 = 201.95, p < 0.001; Table 1).
3.3 Oxidative damage (ROMs)
We found a significant interaction between ALAN and noise exposure (ALAN × noise: F1,263.80 = 7.66, p = 0.006; Figure 4a; Table 1). We classified the interactive effect as a negative synergistic effect on ROM levels (Figure 4b; Piggott et al., 2015). The classification as negative synergism might be misleading, though, on a biological level, since a reduction of ROM levels is in most cases regarded as a positive effect. Post hoc pairwise comparisons showed that birds that were simultaneously exposed to ALAN and noise had 10.8% lower ROM levels (0.49 ± 0.02 mmol H2O2 L−1) than birds that were only exposed to noise (0.55 ± 0.02 mmol H2O2 L−1; p = 0.045; Figure 4a) and 9.4% lower than birds exposed to only ALAN (0.55 ± 0.2 mmol H2O2 L−1; p = 0.11). All other pairwise comparisons were non-significant (p ≥ 0.36). There was no effect of soot exposure on ROM levels (F1,271.06 = 0.04, p = 0.84; Table 1). Sex influenced ROM levels, with males having 10.5% higher ROM levels than females (males: 0.554 ± 0.012 mmol H2O2 L−1, females: 0.501 ± 0.012 mmol H2O2 L−1) after the 5-day exposure (F1,280.00 = 10.80, p = 0.001; Table 1). Body mass after the experiment was negatively associated with post-exposure ROM levels, with birds of lower mass having higher ROM levels than heavier birds (F1,275.48 = 5.60, p = 0.019; Table 1). Lastly, individuals with higher baseline ROM levels also had higher ROM levels post-exposure (F1,172.36 = 88.31, p < 0.001; Table 1).
3.4 Correlations between the oxidative stress status measures within individuals
Only the post-exposure concentrations of OXY and ROMs were correlated (r = 0.26, df = 261, p < 0.001), while neither OXY and tGSH nor ROMs and tGSH post-exposure concentrations were significantly correlated (r = 0.01, df = 268, p = 0.88; r = 0.01, df = 265, p = 0.94, respectively). When split into treatment groups, post-exposure concentrations of OXY and ROMs were significantly correlated for the control and ALAN–soot treatment groups (r = 0.47, p = 0.007 and r = 0.40, p = 0.018, respectively). All other treatment groups in the OXY–ROM comparisons had p ≥ 0.089. The ALAN treatment group was the only group that showed a significant positive correlation between post-exposure OXY and tGSH (r = 0.44, p = 0.011) and ROM and tGSH concentrations (r = 0.39, p = 0.021). All other correlations between OXY and tGSH post-exposure concentrations per treatment group had p ≥ 0.50 and for the ROM and tGSH comparison p ≥ 0.25.
4 DISCUSSION
Recently, the interactive effects between anthropogenic pollutants have gained interest, especially between ALAN and noise (Halfwerk & Jerem, 2021; Wilson et al., 2021). However, urban environments also consist of other pollutants such as air pollution that co-occur with ALAN and noise, but their potential interactive effects have been overlooked. Hence, in this study, we investigate the relative contributions of these three common anthropogenic pollutants, both individually and in combination, to avian oxidative stress status in blood. By using this full-factorial design, we were able to show that a combined exposure to ALAN and traffic noise has complex non-additive interactive effects on the oxidative stress status of captive zebra finches. In fact, the results suggest that the combined exposure leads to a stronger antioxidant response that protects individuals from accumulating oxidative damage than exposure to only one of the stressors. Soot, on the other hand, showed no single or combinatory effects.
4.1 Non-additive effects of short-term multi-pollutant exposure
For two of the three markers of oxidative stress status (tGSH and ROMs), we found synergistic effects of the combined warm white ALAN treatment and traffic-noise exposure (see also Wilson et al., 2021). In the case of tGSH, the pairwise post hoc comparisons indicated that the single exposure to either ALAN or noise did not differ in tGSH levels compared to birds not exposed to ALAN and noise, but the combination of both pollutants led to a non-additive increase of tGSH levels (i.e. positive synergistic effect). The amount of oxidative damage, measured as ROM concentration, decreased under the combined pollutant exposure compared to single exposure to noise, while (as with tGSH) no effects of single pollutant exposure to noise or ALAN were found. Surprisingly, the simultaneous increase in antioxidant defence and reduction in oxidative damage suggest a combined (short-term) exposure to noise and ALAN incurs lower oxidative stress than exposure to one of the pollutants on its own. Our results stand partly in contrast to those of a correlative physiological study of wild great tit nestlings, where they did not find a response in oxidative status (in nine measured metrics) to different levels of ALAN and/or anthropogenic noise (Casasole et al., 2017). In another study of great tit nestlings, chicks responded to higher noise levels by increasing their immune response (i.e. haptoglobin), but no effect of higher ALAN levels was detected (Raap et al., 2017). The contrasting results could be related to the different age categories used, that is, adults in the present study and nestlings in the studies above. However, species differences in sensitivity to noise and artificial light (and in combination) have also been reported, with species occupying open habitats being more sensitive/responsive than those occupying closed habitats (Wilson et al., 2021). Although our zebra finches are captive bred, they originate from open arid landscapes of Australia, which may also explain the differences to great tits that prefer woodland. In other words, species ecology and life stage are likely to influence the physiological responses to pollutants, which warrants more studies to fully understand how generalisable our results are. Another difference between our study, apart from the obvious controlled laboratory setting here, and the above studies is the magnitude of the ALAN treatment. Here we used 13 lx, which is much higher than the natural studies that ranged between almost complete darkness (0.01 lx) and 6.4 lx (Casasole et al., 2017; Raap et al., 2017). Although on the higher range, the higher ALAN here is ecologically relevant given that streetlights have an illumination of 15 lx. Furthermore, other studies on zebra finches show that constant daylight regimes for at least 30 days increase their mortality (Snyder et al., 2013), and at the physiological level, zebra finches show neuronal death and decreased melatonin levels already at 1.5 and 5 lx after 3 weeks of exposure (Moaraf et al., 2020). Given the short-term exposure here and the fact that captive breeding often involves selecting for stress resistance (Griffith et al., 2021), the higher illumination seems justified also for these reasons.
Furthermore, birds that were exposed to noise alone had a significantly different response in both tGSH and ROMs levels than birds that were exposed to the combination of ALAN and noise. Interestingly, in both cases, the addition of ALAN reversed the effect of noise alone. This could be due to the higher magnitude of the experimental ALAN (see above) compared to the noise treatment. The birds themselves were surprisingly loud in their communication; yet, it should be noted that the 3 db difference between the no-noise and noise treatment is still three times higher in terms of amplitude, and maybe of even more importance, of a different frequency compared to chirping birds. This frequency difference is not insignificant. It was recently shown that zebra finch eggs exposed to traffic noise led to developmental impairments with life-long fitness consequences, whereas embryos exposed to birdsong of a similar amplitude had no negative effects (Meillère et al., 2024). However, since the direction of the tGSH effect by ALAN alone and the combined effect go in the same direction, ALAN seems to have an overriding effect over noise in the present study (see also Dominoni, Smit, et al., 2020). ALAN is known to disrupt circadian rhythms, which have cascading effects on physiological processes (Gwinner et al., 1997; Raap et al., 2015, 2016). Our sound measurement during the night also indicates that the groups under ALAN exposure were more awake since the ‘noise’ level from the birds themselves was 2 db louder than during non-ALAN nights (see Section 2). Interestingly, in a recent study on zebra finches using much lower ALAN (0.3 lx) than ours but during a longer time (8+ weeks), their chronic effect on physiology is very similar to our short-term effect, with decreased oxidative damage and increased antioxidant levels and night-time activity (Alaasam et al., 2024).
Although there was no within-individual correlation between tGSH and ROMs in the ALAN-noise group, there was a simultaneous increase in tGSH levels and a decrease in ROMs levels post-exposure in this group. ALAN and noise may affect similar physiological pathways, such as endocrine stress responses via the hypothalamic–pituitary–adrenal axis (Davies et al., 2017; Grunst et al., 2020; Zollinger et al., 2019; see also Figure 1), which then directly or indirectly affect the oxidative status through an elevation of metabolism and production of ROS. Based on this, two alternative explanations that might explain the dynamics between these two oxidative stress markers can be put forward. On the one hand, the internal ROS production and release could be overall lower in birds exposed to these two sensory pollutants, which allows them to maintain a higher tGSH concentration and no additional damage is generated. On the other hand, and probably more likely, the higher tGSH levels in the birds that experienced the combined exposure could be a result of an up-regulation of their intracellular glutathione machinery in response to increased internal ROS production, triggered by exposure to noise and ALAN. This up-regulation of antioxidant defence, although not present for OXY, would then lead to increased scavenging of ROS and subsequent decrease in oxidative damage of macromolecules. Other experimental studies on zebra finches have shown that their oxidative stress status is indeed highly responsive to external stressors such as poor nutrition, ozone, herbicides and infection, even though they are captive-bred model species (Bertrand et al., 2006; Rodríguez-Martínez & Galván, 2020; Ziegler et al., 2024). Similarly, wild urban birds often show an up-regulation in antioxidant machinery (e.g. Herrera-Dueñas et al., 2017; Salmón, Watson, et al., 2018; Tkachenko & Kurhaluk, 2013). However, the glucocorticoid levels show no general up- or down-regulation in response to urbanization, but instead seem to be highly context dependent (Bonier, 2012).
In contrast to the direct and plastic response to the short-term exposure shown here, the responses of wild urban birds that have experienced a continuous and long-term exposure to these pollutants could be under selection. However, in a cross-fostering study between urban and rural great tit chicks, the urban environment imposed an up-regulation of enzymatic antioxidants independent of habitat of origin, supporting the role of plasticity in these responses (Salmón, Watson, et al., 2018). This makes the present result more generalizable to other species, but poses the question of whether having an increased antioxidant defence over time, such as the response to the combined ALAN and noise exposure here, could come with a cost? For example, glutathione is an internally synthesized thiol that requires essential dietary amino acids as building blocks such as methionine to build cysteine (Meister & Anderson, 1983). However, cysteine and methionine are also needed for other proteins and enzymes crucial for glucose uptake, salt and water balance among several other functions (Meister & Anderson, 1983). Resource limitation and hence possible trade-offs between different physiological processes might occur especially in wild settings. Both short-term trade-offs between, for example, oxidative defence and immune function may be possible, and long-term trade-offs between life-history traits and the oxidative status have been suggested (Costantini & Møller, 2009; Monaghan et al., 2009).
4.2 Lack of effects of soot exposure
In contrast to the two sensory pollutants, soot itself has pro-oxidant potential (Niranjan & Thakur, 2017; Sanderfoot & Holloway, 2017; see Figure 1). Hence, it was surprising to find neither an effect of single nor combined exposures to soot. Possibly, blood is not the optimal target tissue to detect short-term effects of soot exposure. However, a multitude of studies has found effects of soot exposure in plasma samples, and a systematic review has found that levels of several oxidative stress markers measured in blood correlate well with levels measured in other tissues (Margaritelis et al., 2015). An alternative source for the lack of response might be the soot type produced in our experimental set-up, which was ‘pure’ black carbon. In a natural setting, soot is a composition of particles, vaguely defined, often with a coating of, for example, organic fractions, surface oxidation and/or heavy metals. Traffic-generated soot particles that are common in urban environments contain higher amounts of coating compared to our ‘pure’ black carbon. Possibly, coated soot is more pro-oxidative than ‘pure’ black carbon (Boström et al., 2002). Therefore, the effects of soot on oxidative status in birds could be different in a wild setting, where a mixture of soot particles is inhaled. On the other hand, other researchers argue that particle size matters more for toxicity than soot type (Schwartz et al., 1996). Here, the generated particle sizes are in the range of ‘natural’ occurring soot and in the more toxic size range; therefore, the lack of effect might rather be dose, biomarker and/or species dependent, along with exposure and sampling time rather than the actual soot type. Indeed, zebra finches exposed to methylmercury during development show effects on reproductive performance later in life (Paris et al., 2018). Furthermore, rodents exposed to diesel exhaust show the highest level of oxidative stress several months after the exposure (Bendtsen et al., 2020). Here, the samples were taken immediately after the 5-day exposure, and although the avian lung is rather different from the mammalian lung, the delayed effects could be similar for birds and warrant attention in the future.
4.3 Effects of other factors
There was a strong sex effect on the amount of oxidative damage, but no effect on either OXY or tGSH. Males had overall higher levels of ROMs after the 5-day experiment than females. Sex differences in oxidative status have been found in several studies on birds, but with varying results (Alonso-Alvarez et al., 2007; Casagrande et al., 2011; Costantini, 2010; Isaksson et al., 2013). Whether increased oxidative damage in males in our study stems from, for example, hormonal differences or has other causes should be investigated in further studies. Independent of sex, a lighter body mass post-exposure was also related to higher levels of ROM. In other words, individuals in poorer physical condition were also in a poorer physiological state.
Finally, among the correlation analyses between the physiological biomarkers, post-exposure OXY levels were overall positively associated with post-exposure ROM levels, which suggests that these two markers are highly connected at the individual level but that OXY is less responsive to external sensory stressors such as ALAN and noise. In addition, there was strong within-individual repeatability in the physiological state since pre-exposure levels were strong predictors of post-exposure levels of all three physiological markers.
5 CONCLUSIONS
Here, we show that a short-term multi-stressor exposure can lead to complex non-additive effects on the oxidative stress status of a captive bird species. Our results suggest that a combined exposure to two sensory pollutants elicits a stronger response of the detoxification machinery than exposure to only one pollutant, which likely led to increased scavenging of ROS and thus a decrease in oxidatively damaged molecules. Predicting such non-additive effects is inherently difficult, and this study highlights that robust experimental designs are necessary to disentangle the relative contribution of each potential stressor to an effect when considering multiple stressors. The exact mechanistic processes that give rise to such interactions remain to be elucidated, along with repeating this study on non-captive species; only then can the potential fitness consequences and ecological implications of interactive effects be resolved.
AUTHOR CONTRIBUTIONS
Ann-Kathrin Ziegler, Jenny Rissler, Anders Gudmundsson and Caroline Isaksson designed the experiment. Ann-Kathrin Ziegler, Jenny Rissler, Anders Gudmundsson and Caroline Isaksson conducted the experiment Ann-Kathrin Ziegler performed the laboratory work. Ann-Kathrin Ziegler performed the statistical analysis. Ann-Kathrin Ziegler wrote the first draft. CI made all revisions. All authors edited and revised the manuscript. Ann-Kathrin Ziegler and Caroline Isaksson received funding.
ACKNOWLEDGEMENTS
We are very grateful to the people who provided technical assistance during this experiment: Patrik Nilsson, Jonas Jakobsson, Vilhelm Malmborg, Louise Gren, Robert Olsson, Lars Jonas Brunskog and Hillevi Hemphälä. We would like to thank Wolfgang Forstmeier for providing the zebra finches from the population of the Max-Planck Institute in Seewiesen, Germany. We are grateful to Philip Downing and Hannah Watson for providing valuable comments on previous versions of the manuscript. We thank FORMAS (2015-00526; to CI) and Stiftelsen Lunds Djurskyddsfond (67/17; to AKZ) for funding.
CONFLICT OF INTEREST STATEMENT
This study has no conflicts of interest. Caroline Isaksson is an Associate Editor for Functional Ecology, but took no part in the peer review or decision-making process for this manuscript.
STATEMENT ON INCLUSION
Our study brings together scientists (and authors) from physics to biology; all scientists were at the time based in the country where the study was carried out. All authors were engaged early on with the research and study design to ensure that the diverse sets of perspectives they represent were considered from the onset. Whenever relevant, literature published by scientists from the region was cited; efforts were made to consider relevant work published in the local language.
Open Research
DATA AVAILABILITY STATEMENT
Data deposited in the Dryad repository: https://doi.org/10.5061/dryad.hx3ffbgqs (Isaksson et al., 2025).
