Ecotoxicology (2020) 29:275 – 285 https://doi.org/10.1007/s10646-020-02171-x Food stress, but not experimental exposure to mercury, affects songbird preen oil composition L. A. Grieves 1 ● C. L. J. Bottini 1 ● B. A. Bran fi reun 1 ● M. A. Bernards 1 ● S. A. MacDougall-Shackleton 2 ● E. A. MacDougall-Shackleton 1 Accepted: 28 January 2020 / Published online: 8 February 2020 © Springer Science + Business Media, LLC, part of Springer Nature 2020 Abstract Mercury is a global pollutant and potent neurotoxic metal. Its most toxic and bioavailable form, methylmercury, can have both lethal and sublethal effects on wildlife. In birds, methylmercury exposure can disrupt behavior, hormones, the neuroendocrine system, and feather integrity. Lipid-rich tissues and secretions may be particularly susceptible to disruption by lipophilic contaminants such as methylmercury. One such substance is feather preen oil, a waxy secretion of the uropygial gland that serves multiple functions including feather maintenance, anti-parasitic defense, and chemical signaling. If methylmercury exposure alters preen oil composition, it could have cascading effects on feather quality, susceptibility to ectoparasites, and mate choice and other social behaviors. We investigated whether exposure to methylmercury, either alone or in association with other stressors, affects preen oil chemical composition. We used a two-factor design to expose adult song sparrows ( Melospiza melodia ) to an environmentally relevant dietary dose of methylmercury and/or to another stressor (unpredictable food supply) for eight weeks. The wax ester composition of preen oil changed signi fi cantly over the 8-week experimental period. This change was more pronounced in the unpredictable food treatment, regardless of dietary methylmercury. Contrary to our prediction, we found no main effect of methylmercury exposure on preen oil composition, nor did methylmercury interact with unpredictable food supply in predicting the magnitude of chemical shifts in preen oil. While it remains critical to study sublethal effects of methylmercury on wildlife, our fi ndings suggest that the wax ester composition of preen oil is robust to environmentally relevant doses of this contaminant. Keywords Mercury ● Methylmercury ● Preen oil ● Songbird ● Stress ● Uropygial gland Introduction Mercury is a global pollutant and potent neurotoxic metal produced by both natural and anthropogenic processes (Evers 2018). Human activities over the past 160 years (1850 – 2010 CE) have resulted in a 2 to 3-fold increase in the atmospheric burden of mercury (Streets et al. 2017). Current environmental mercury levels are mainly due to a combination of new inputs, continued re-emission and deposition, and the persistence of mercury in ecosystems (Munthe et al. 2007). In water and fl oodplain soil, inor- ganic mercury can be converted to methylmercury, the most toxic and bioavailable form of mercury. Methyl- mercury can then enter the terrestrial food chain via aquatic and soil invertebrates (Newman et al. 2011; Mahbub et al. 2017; Paranjape and Hall 2017). Thus, methylmercury contamination is not only a concern in aquatic ecosystems, but also in apex terrestrial predators and animals that ingest a large proportion of invertebrates as part of their diet (e.g., bats and many bird species, including songbirds; Cristol et al. 2008; Evers 2018). Enhanced methylation processes are predicted under future climate conditions, increasing the environmental levels of methylmercury (Krabbenhoft and Sunderland 2013). It is therefore important to study the effects of These authors contributed equally: L. A. Grieves, C. L. J. Bottini * L. A. Grieves lgrieves@uwo.ca 1 Department of Biology, The University of Western Ontario, 1151 Richmond St., London, ON N6A 5B7, Canada 2 Department of Psychology, The University of Western Ontario, 1151 Richmond St., London, ON N6A 5C2, Canada Supplementary information The online version of this article (https:// doi.org/10.1007/s10646-020-02171-x) contains supplementary material, which is available to authorized users. 1234567890();,: 1234567890();,: methylmercury on wildlife (Scheuhammer et al. 2015; Fuchsman et al. 2017; Whitney and Cristol 2017). In birds, sublethal methylmercury exposure can have profound effects on the neuroendocrine system (Tan et al. 2009; Scheuhammer et al. 2015), the immune system (Kenow et al. 2007; Fallacara et al. 2011; Whitney and Cristol 2017), and on behavior (Frederick and Jayasena 2010; Whitney and Cristol 2017; Swaddle et al. 2017). Methylmercury exposure can induce oxidative stress (Gibson et al. 2014; Henry et al. 2015; Espín et al. 2016a), affect androgen hormone levels (Tan et al. 2009; Jayasena et al. 2011; Tartu et al. 2013), and has been associated with decreased reproductive success (Frederick and Jayasena 2010; Fallacara et al. 2011; Braune et al. 2012; Tartu et al. 2013). Methylmercury exposure has also been linked to reduction in feather quality, including altered re fl ectance (White and Cristol 2014) and decreased brightness, hue, and chroma (Giraudeau et al. 2015), with implications for communication, fl ight, and migration (Klaassen et al. 2012; Carlson et al. 2014; Ma et al. 2018a, b; Pryke and Grif fi th 2007; White and Cristol 2014; McCullagh et al. 2015; Scheuhammer et al. 2015; Roeder et al. 2019). Preen oil secretions from the uropygial gland are important for feather maintenance in most bird species (Salibian and Montalti 2009). Because methylmercury is fat soluble (Halbach 1990), it can accumulate in sebac- eous tissues like the uropygial gland. Indeed, mercury has been found in the uropygial gland of several waterbird species (e.g., common redshanks, Tringa tetanus , Goede and De Bruin 1984; common loons, Gavia immer , Frank et al. 1983; herring gulls, Larus argentatus , Leonzio et al. 1986; black-headed gulls, Larus ridibundus , Leonzio et al. 1986; Scopoli ’ s shearwater, Calonectris diomedeo , Renzoni et al. 1986; and common cormorant, Phalacro- corax carbo , Saeki et al. 2000). Preen oil serves multiple non-mutually exclusive functions in birds, ranging from waterproo fi ng and feather maintenance to social commu- nication via chemical cues in preen oil (Moreno-Rueda 2017). The chemical composition of preen oil is affected by many factors, such as sex (Whittaker et al. 2010), age (Shaw et al. 2011), time of year (Bhattacharyya and Chowdhury 1995), circulating androgen levels (Whittaker et al. 2011), diet (Thomas et al. 2010), food stress (Reneerkens et al. 2007), genotype at the major histo- compatibility complex (MHC; Leclaire et al. 2014), parasitic infection status (Grieves et al. 2018), and skin and preen gland microbiota (Jacob et al. 2014; Whittaker et al. 2019). Because preen oil chemical composition is dynamic and can be affected by diverse factors such as those listed above, and because methylmercury can accumulate in sebaceous tissues such as the uropygial gland, there is potential for mercury exposure to alter the chemical composition of preen oil. We propose three potential mechanisms by which this could be achieved. First, mercury may accumulate in the uropygial gland (e.g., Leonzio et al. 1986) and thereby alter preen oil directly. Second, mercury in the uropygial gland may alter the preen gland microbial community, as has been found with other heavy metals (Chatelain et al. 2016), resulting in different fermentation patterns and thus the production of different preen oil compounds (Whittaker and Theis 2016). Third, mercury may alter circulating hormone pro fi les (Tan et al. 2009; Tartu et al. 2013), which could then indirectly affect the chemical composition of preen oil (Whittaker et al. 2011). Given the importance of preen oil in chemical communication (Caro et al. 2015) and in maintaining feather quality (Salibian and Montalti 2009), if mercury does alter preen oil, it may have implications for songbird survival and reproductive success. The effects of pollutants like methylmercury may be modulated by stressors such as unpredictable food supply. Methylmercury and stress can have additive effects when their combined effects are greater than the effects of one challenge alone. For example, in wild birds, co-exposure to heat stress and methylmercury during the early nestling period reduces fl edging success more than does exposure to either condition in isolation (Hallinger and Cristol 2011). Alternatively, the effects of methylmercury may be non-additive with those of other stressors, either syner- gistic or compensatory (e.g., Hoffman and Heinz 1998; Coors and De Meester 2008; Chatelain et al. 2016). Songbirds have been proposed as sentinel species for assessing environmental mercury contamination (Jackson et al. 2015). We chose song sparrows ( Melospiza melodia ) as a model species for this study because they are asso- ciated with wetland habitats and consume primarily invertebrates during breeding (Arcese et al. 2002; Wing- fi eld et al. 2012); thus, they may be frequently exposed to environmental and/or dietary methylmercury in the wild (Newman et al. 2011; Jackson et al. 2015). We tested whether dietary methylmercury exposure, unpredictable food stress, and combined exposure to both methylmer- cury and food stress would alter the chemical composition of preen oil in songbirds. We experimentally exposed song sparrows to an environmentally relevant dose of methylmercury, unpredictable food stress, and both food stress and methylmercury, and compared the chemical composition of preen oil before exposure and after eight weeks of exposure. We predicted that exposure to methylmercury, stress, and combined exposure would alter preen oil composition. To our knowledge, this is the fi rst experimental study to investigate the effects of methylmercury exposure on preen oil chemistry. 276 L. A. Grieves et al. Methods Study subjects and housing We used 54 adult song sparrows in this study. Of these, 36 (27 male, 9 female) were captured on their breeding terri- tories in London, Ontario, Canada (42.9849° N, 81.2453° W) between 8 August and 1 September 2017 and held overwinter. The remaining 18 (16 male, 2 female) were captured in London between 9 and 11 April 2018. We captured birds via mist nets, using playback of adult song and juvenile distress calls to attract song sparrows to the nets. The male-biased capture of birds was a result of fi eld limitations: males are more likely to fl y into mist nets in response to targeted playback than females (Grieves, Bot- tini, MacDougall-Shackleton, MacDougall-Shackleton, pers. obs.). The 36 birds caught in 2017 were part of unrelated experiments (Grieves et al. 2019a, b) and were donated to this study in accordance with guidelines to reduce the number of animals used in research wherever possible (Canadian Council on Animal Care (CCAC 2020)). All 54 birds were also part of a larger study exploring the effects of dietary methylmercury exposure on songbirds (Bottini et al., unpubl. data), and this sample size of 54 was selected as the minimum number of birds needed for that work, following the CCAC ’ s reduction principle. We housed birds in individual cages at 20 – 22 °C with relative humidity of 30 – 70%. Birds were kept under a simulated natural photoperiod (approximately 13L:11D in April – May to 15L:9D in June – July) and with ad libitum access to water and food (Living World Premium Mix for Budgies parakeet seed mixed with ground Mazuri small bird diet) until 16 April 2018, when we began transitioning the birds to a nutritionally complete agar-based synthetic diet. We prepared the agar diet so that it contained 60% carbohydrate, 13.4% protein, and 10.6% lipid (dry mass basis; instructions and details in Supplementary Materials and Table S1). This diet was the birds ’ major food as of 30 April 2018, except that a small quantity of uncontaminated blended eggs and bread (mean ± SE = 6.3 ± 0.1 g) or 2 – 4 mealworms was supplied once a week. Methylmercury and stress exposure We assigned birds to one of four treatment groups such that sex and capture date were balanced across groups: methylmercury ( n = 17: 13 males, 4 females; 12 captured in 2017, 5 captured in 2018), food stress ( n = 12: 10 males, 2 females; 7 captured in 2017, 5 captured in 2018), combined- exposure to food stress and methylmercury ( n = 13: 10 males, 3 females; 9 captured in 2017, 4 captured in 2018), and control ( n = 12: 10 males, 2 females; 8 captured in 2017, 4 captured in 2018). We staggered the start of food stress and methylmercury exposure by 24 h such that half of the birds in each of the four groups started treatments on day 1 and the other half started treatments on day 2, with exposures beginning on 15 and 16 May 2018 and con- tinuing daily until 9 and 10 July 2018 (i.e., for 8 weeks; this study, but see “ Bird fate ” , below). Contaminated food was provided in disposable plastic food cups at the bottom of the cage and cups were changed daily. Uncontaminated food was provided in reusable plastic cups at the bottom of the cage and these were changed weekly. Each day, regardless of treatment, we removed the leftover food and replaced it with 20 g of fresh agar food. Birds in the methylmercury-exposed groups (methyl- mercury; combined exposure to methylmercury and food stress) were fed the agar-based diet formulated with 0.25 ppm methylmercury chloride (0.25 mg/kg wet weight; Alfa Aesar, #33553). This level was selected to mimic levels reported for lepidopteran, coleopteran, and arachnid invertebrates in mercury-contaminated areas in the United States (Cristol et al. 2008; Newman et al. 2011; Ortiz et al. 2015), and thus represents an environmentally relevant dose. We chose methylmercury chloride (MeHgCl) because it is commercially available and commonly used in experi- mental studies (e.g., Hill and Shaffner 1976; Hoffman and Heinz 1998; Franceschini et al. 2017). Tissue accumulation of MeHgCl is similar to the more environmentally wide- spread methylmercury cysteine (MeHgCys; Rutkiewicz and Basu 2013; Varian-Ramos et al. 2017); further, MeHgCl is relatively weakly chemically bound compared to more strongly associated complexes like MeHgCys (binding constant of MeHgCys > MeHgCl; Hughes 1957; Simpson 1961). Thus, we expected that MeHgCl would dissociate rapidly in solution and form other MeHg-organic ligand compounds, more closely mimicking the way MeHg would be incorporated into diet in the natural environment. For each of the 24 batches of food made (methylmercury diet = 13 batches, control diet = 11 batches), we collected a 4 g sample and froze it at − 80 °C until analyses of its total mercury (THg) content could be completed. Con fi rmatory analysis of these diet samples (details below) indicated that the methylmercury diet had a THg level of 0.18 ± 0.02 ppm (mean ± s.d.). Although the concentration of methylmercury we obtained in the birds ’ diet was lower than our target (0.25 ppm), this value still falls within the range of dietary mercury levels associated with harmful effects on avian reproduction (i.e., 0.16 – 0.75 mg/kg; Fuchsman et al. 2017), and thus is still an environmentally relevant dose. Birds in the non-methylmercury groups (food stress; control) received the same diet but with no added methylmercury (THg level 0.001 ± 0.0007 ppm) (mean ± s.d.). Birds undergoing food stress (food stress; combined exposure to Food stress, but not experimental exposure to mercury, affects songbird preen oil composition 277 methylmercury and food stress) had all food removed from their cages for 3 h daily at randomly selected times (fol- lowing Schmidt et al. 2010) during the 13 – 15 h light period. Blood collection and mercury analyses On 1 – 2 May (pre-treatment) and 11 – 12 July (post-treat- ment), we collected 50 – 200 μL blood samples from the brachial vein into heparinized capillary tubes, then transferred whole blood into microcentrifuge tubes within 5 – 60 min of collection and stored at − 80 °C. Because the mean ratio of total blood mercury to blood methylmercury is close to 1:1 (Rimmer et al. 2005), we used total mercury (THg) as an estimate of the methylmercury burden in blood. We measured THg levels in 20 – 40 μL of whole blood and 0.12 g of the birds ’ diet by thawing samples at room temperature, vortexing for 5 s, then quantifying THg in nickel boats using a Direct Mercury Analyzer (DMA- 80, Milestone Inc., Shelton, USA) following U.S. EPA Method 7473 (U.S. EPA 1998). THg analyses were con- ducted at the Biotron (an ISO 17025 accredited facility) at Western University, London, Ontario, Canada. The DMA analysis sequence includes drying samples at 200 °C for 60 s, thermal decomposition at 650 °C for 720 s, catalytic conversion and amalgamation on gold for 12 s, then measuring THg emissions by atomic absorption spectro- scopy at 253.65 nm. Before sample analysis, we checked DMA calibration with two methods blanks (no-boat), one sample blank (empty nickel boat), a protein-certi fi ed reference (DORM-4; National Research Council Canada), an aqueous certi fi ed concentration standard (CCS; 100 ng), then another sample blank. Following this, each batch of 10 samples included a blood standard certi fi ed reference (CRM3; 50 μL of metal level 3 whole blood control; UTAK create control #44523), a spiked CRM3 (CRM3 plus 25 ng of CCS) for blood samples or a spiked food sample (food plus 50 ng of CCS), a duplicate of each spiked sample, a blank sample, and a duplicate blood sample. We report wet weight concentration (ww.) and quality assurance mean ± SE. Mean percent recovery for the reference samples was 101.58 ± 1.31% (CRM3; n = 16), 95.68 ± 0.91% (DORM- 4; n = 11) and 101.63 ± 1.06% (CCS; n = 9). For all sam- ples with THg concentrations greater than 10 times the minimum detection limit (0.08 ng), the relative percent difference between duplicates of blood and food samples ( n = 14 pairs) was 3.26 ± 1.69%, and 11.62 ± 3.82% ( n = 4 pairs) respectively. The relative percent difference for spiked food sample duplicates was 1.16 ± 0.66% ( n = 4 pairs) and 1.08 ± 1.31% ( n = 16 pairs) for duplicated spiked CRM3. We did not test the supplementary eggs, bread, and mealworms for mercury contamination. However, we took reasonable precautions to avoid contamination: mercury-contaminated foods were prepared in a fume hood away from uncontaminated foods and food cups containing contaminated food were replaced daily. We infer that these measures were suf fi cient to avoid mercury- contamination of supplemental and control foods because we found no evidence of contamination in the control birds ’ THg blood measurements (see “ Results ” ). Preen oil collection and analysis On 14 May 2018 (15 days after beginning the uncontami- nated agar-based diet and 1 – 2 days before beginning the unpredictable food stress and methylmercury exposure treatments), we collected a preen oil sample from each individual (pre-treatment sample). We gently probed the uropygial gland to express 1 – 5 mg of oil into an unhepar- inized capillary tube. We then snapped the capillary tube into a 1.6 mL microcentrifuge tube and stored the samples at − 20 °C pending analysis. We used the same procedure to collect a second sample of preen oil from each subject on 9 – 10 July 2018, after 8 weeks of experimental treatment (post-treatment sample). We did not test for THg in preen oil samples. We used gas chromatography with fl ame ionization detection (GC-FID) to separate and quantify compounds present in preen oil. We dissolved preen oil samples in 1 – 5 mL of chloroform (CHCl 3 ; scaled for the volume of preen oil collected for a fi nal concentration of 1 mg preen oil/mL CHCl 3 ) inside borosilicate-capped glass vials. Following a previously established protocol (Slade et al. 2016), we injected 1 μL of each sample onto a 5% phenyl methyl siloxane column (Agilent Technologies DB-5, 30 m × 0.32 μm ID × 0.25 μm fi lm thickness) on an Agi- lent 6890N instrument. We injected samples at 70 °C and held for one minute, ramped to 130 °C at 20 °C per minute, ramped to 320 °C at 4 °C per minute, then held at 320 °C for 10 min. We used hydrogen as a carrier gas at 2.5 mL/min. Injector and FID were constant at 200 and 310 °C, respectively. Each batch of 20 samples included a blank containing only solvent (CHCl 3 ) as a negative control and a sample of known composition previously analyzed by both GC-FID and gas chromatography-mass spectrometry (GC-MS; Slade et al. 2016) to ensure con- sistency between runs. Because the volume of preen oil collected varied across samples, we quanti fi ed the rela- tive, rather than absolute, size of each peak based on peak area relative to that of the full chromatogram. Only peaks that comprised at least 0.1% of the total chromatogram area were retained for analysis (Leclaire et al. 2012), resulting in 37 unique peaks. Peaks were standardized by total area such that within each sample all peaks totaled to 100% (Stoffel et al. 2015). 278 L. A. Grieves et al. Bird fate This study was part of a separate experiment in which birds continued to be exposed to methylmercury and stress treatments until 13 and 14 August 2018. On 13 and 14 August 2018, 14 birds (3 control, 4 stress only, 2 methyl- mercury only, 5 both stress and methylmercury) were euthanized via iso fl urane inhalation (Fresenius Kabi), in accordance with our animal use protocols (see “ Compliance with ethical standards ” ) and CCAC re fi nement guidelines to minimize pain or distress to animals. The remaining 37 birds were provided with uncontaminated agar diet and ad libitum access to water, with no additional treatments, until completion of the larger study. From 31 October to 4 November 2018 all 37 birds were euthanized as described above and the bodies were stored at − 80 °C pending further tissue collection for THg and MeHg content analyses required for the other experiment. Data analysis Song sparrow preen oil is comprised of wax monoesters arranged in a series of different chain length fatty alcohols and fatty acids in different combinations (Grieves et al. 2019c). To infer the major and minor acid:alcohol esters found in each GC-FID peak in the present study, we com- pared our GC-FID data with a previously published GC-MS analysis of song sparrow preen oil (Grieves et al. 2019c). To test for group differences in preen oil composition, we log (x + 1) transformed relative peak areas and, because large chromatogram peaks can disproportionately affect distance measures, we normalized the data using the “ range ” method in the decostand function in the R package vegan (Dixon and Palmer 2003). We then constructed a Bray – Curtis dissimilarity matrix for all pairwise combina- tions of the 104 samples ( n = 54 and 50 for pre- and post- treatment, respectively). Variation in sample sizes re fl ects the fact that three individuals (2 methylmercury group, 1 control group) died or were euthanized during the course of the experiment, and post-treatment samples could not be collected from another individual (methylmercury group). To assess the statistical signi fi cance of differences in preen oil chemical composition between the four groups post-treatment, we conducted permutational multivariate analysis of variance on the pairwise Bray – Curtis dissim- ilarity matrices. This permutation-based approach, analo- gous to a nonparametric MANOVA, does not make assumptions about the data ’ s distribution and may be less sensitive to group differences in the dispersion of points compared to other methods (Anderson 2001; Anderson and Walsh 2013). We used the same method to test for differ- ences in preen oil composition between pre-treatment and post-treatment. These analyses were performed in R version 3.3.3 (R Development Core Team 2017) using the adonis command in the package vegan (Dixon and Palmer 2003). To visualize pairwise chemical distances between sam- ples, we used nonmetric multidimensional scaling in the R package vegan (Dixon and Palmer 2003). This approach places each sample on a two-dimensional scatter plot, pre- serving ranked pairwise distances such that two points close together represent two samples with relatively similar che- mical composition whereas points further apart represent samples that are more dissimilar (Clarke 1999; Stoffel et al. 2015). For each bird, we calculated the change in blood THg over the course of the experiment (i.e., post-treatment – pre- treatment values) and the post-treatment – pre-treatment change in preen oil chemical composition. To calculate the change in preen oil chemical composition we calculated the Euclidean distance between the nonmetric multidimensional scaling coordinates for each individual ’ s post- and pre- treatment preen oil samples. We ran unpaired t -tests in GraphPad Prism version 6.01 to test whether the change in blood THg was greater in mercury-exposed vs unexposed birds and between mercury-exposed vs mercury and food stressed birds. To test whether mercury-exposed, food stressed, or co-exposed birds differed in the magnitude of change in preen oil chemical composition, we performed a bifactorial ANOVA in base R (R Development Core Team 2017). Finally, as a complementary, individual-level ana- lysis we also calculated each individual ’ s change in blood THg (post-treatment – pre-treatment values) and used simple linear regression (implemented in GraphPad Prism version 6.01) to test whether change in blood THg predicted the magnitude of change in preen oil composition. All statistical analyses were considered signi fi cant at α = 0.05. Results Pre-treatment levels of blood THg for all birds combined were 0.007 ± 0.005 (mean ± SE) ppm. Post-treatment blood THg levels were 5.22 ± 0.21 ppm for birds exposed to methylmercury (methylmercury; combined exposure to methylmercury and food stress) and 0.004 ± 0.001 ppm for birds not exposed to methylmercury (food stress; control). Blood THg levels were signi fi cantly higher for the methylmercury-exposed birds than for non-exposed birds (t 1,47 = 31.70, P = < 0.0001). There was no difference in blood THg levels between methylmercury-exposed birds and both methylmercury-exposed and food stressed birds (t 1,25 = 0.36, P = 0.73). We were able to infer the major and minor wax monoesters found in song sparrow preen oil for 20 of the 37 peaks retained in our analyses. The molecular weight of these peaks ranged from 438 to 578, with a total of 29 – 39 Food stress, but not experimental exposure to mercury, affects songbird preen oil composition 279 carbons. Peaks were made up of C12:C17 through C19:C22 acid:alcohol monoesters and peaks contained from 2 to 8 monoesters each, with the same total number of carbons in each peak. For example, a peak at 38.6 ′ with a combined m / z of 480 contains 5 monoesters, all with a total of 32 carbons. The major component is a C12 acid esteri fi ed to a C20 alcohol, with lesser amounts of C13/C19, C14/C18, C15/C17, and C16/C16 monoesters. The 17 peaks from which we could not infer the monoester composition were predominantly very low abundance compounds (Supple- mentary Materials, Table S2). Preen oil composition was signi fi cantly different between the sexes during both the pre-exposure and 8- weeks exposure (post-exposure) sampling periods, though the effect was greater in the pre-exposure sample collected in May 2018 (Pre-exposure: F 1,53 = 51.08, R = 0.50, P < 0.0001; Post-exposure: F 1,49 = 3.78, R = 0.07, P = 0.03; Supplementary Materials, Fig. S1). There was a signi fi cant difference in preen oil composition between pre-treatment and post-treatment sampling periods (Fig. 1 and Table 1) and we found a signi fi cant sex × time interaction (Table 1). In contrast to our prediction, the chemical composition of preen oil did not differ among the four groups post- treatment (Fig. 2 and Table 1). However, the pre- vs post- treatment shift in preen oil composition we observed was more pronounced in the two groups that experienced unpredictable food stress compared to the two groups that received predictable food (F 1,49 = 13.31, P < 0.001; Fig. 3). By contrast, the observed shifts in preen oil composition were not more pronounced in the two groups exposed to methylmercury (F 1,49 = 1.72, P = 0.20; Fig. 3), nor did we detect an interaction between methylmercury and stress exposure (F 1,49 = 0.52, P = 0.48; Fig. 3). As a complementary analysis, we tested whether indi- viduals with greater increases in blood THg underwent greater shifts in preen oil composition. We found no rela- tionship between blood THg and change in preen oil che- mical composition for any of the treatment groups (control: F 1,9 = 0.01, R 2 = 0.001, P = 0.95; unpredictable food stress: F 1,9 = 0.08, R 2 = 0.01, P = 0.79; methylmercury: F 1,13 = 0.19, R 2 = 0.01, P = 0.67; combined exposure: F 1,10 = 1.11, R 2 = 0.1, P = 0.32), nor for the full dataset (F 1,47 = 2.40, R 2 = 0.05, P = 0.13, Supplementary Materials, Fig. S2). Discussion We did not fi nd evidence that methylmercury exposure alters preen oil composition. However, the change in preen oil chemical composition from pre-exposure to post- exposure was greater in birds exposed to unpredictable food stress compared to unstressed birds. Further, food stress did not interact with methylmercury exposure in its effect on preen oil composition. We found signi fi cant effects of time of sampling (i.e., pre-treatment vs post- treatment) and sex on preen oil composition, consistent with previous fi ndings (Grieves et al. 2018, 2019c), as well as an interaction between sex and time of sampling. Because preen oil composition changed in all groups, not only in birds exposed to methylmercury or food stress, we interpret the effect of time of sampling on change in preen oil composition as a seasonal effect. This pattern is consistent with seasonal changes observed in the preen oil of free-living song sparrows during roughly the same time of year (April through August, i.e., from early breeding to post breeding; Grieves et al. 2019c). Birds in our study experienced a light schedule designed to match the natural photoperiod throughout the treatment period. Seasonal variation in preen oil has been demon- strated in several other species (red knot, Calidris canu- tus , Reneerkens et al. 2007; dark-eyed junco, Junco hyemalis , Soini et al. 2007; gray catbird, Dumetella carolinensis , Shaw et al. 2011; white-throated sparrow, Zonotrichia albicollis , Tuttle et al. 2014; herring gull, L. argentatus , Fischer et al. 2017). Song sparrows switch from a seed-based diet in the nonbreeding season to an invertebrate-based diet during the breeding season (Arcese et al. 2002), and seasonal changes in preen oil chemical composition may be due at least in part to seasonal changes in diet (Thomas et al. 2010; but see Azzani et al. 2016). However, in this study birds were maintained on a constant food type and therefore we do Fig. 1 Two-dimensional nonmetric multidimensional scaling plot of preen oil composition in captive song sparrows. Bray – Curtis dissim- ilarity values were calculated from standardized and log (x + 1) transformed abundance data. Axis scales are arbitrary. The closer the symbols appear on the plot, the more similar the samples are. 2D stress = 0.06. 2D stress represents the amount of disagreement between the 2D con fi guration and predicted values from the multi- variate regression (values closer to zero are better). Groups sig- ni fi cantly differed between pre-exposure sampling and when sampled after 8 weeks of exposure (post-exposure) to methylmercury and/or unpredictable food stress (F 1,49 = 23.56, R = 0.13, P > 0.0001) 280 L. A. Grieves et al. not think it likely that diet played a role in the time of sampling effect we observed here. Sex differences in preen oil chemical composition have previously been found in this and other bird species (e.g., Whittaker et al. 2010; Amo et al. 2012; Grieves et al. 2019a, c). We found a signi fi cant interaction between time of sampling (pre-treatment vs post-treatment) and bird sex. This is likely explained by our fi nding that, while sex dif- ferences were signi fi cant in both treatment periods, they were more pronounced during the pre-treatment period compared to the post-treatment period. This is consistent with prior fi ndings in this species that sex differences in preen oil chemical composition are observed during the early part of the breeding season (April – May) but are no longer apparent by the start of the nonbreeding season (i.e., by August; Grieves et al. 2019c). Our treatment period began in mid-May, consistent with the peak breeding period of wild song sparrows, and continued until mid-July, which is approaching the postbreeding period for song sparrows in this study area (Grieves et al. 2019c). It is intriguing to observe such pronounced changes in the preen oil chemical composition of captive song sparrows held under natural photoperiodic conditions over this short eight week time scale. Such seasonal changes in preen oil composition may be due to endogenous mechanisms such as fatty acid reduction/oxidation reactions (Reneerkens et al. 2007; Bonadonna et al. 2007) and/or changes in hormonal state (Whittaker et al. 2011; Azzani et al. 2016). Odorous secretions have been shown to differ between stressed and unstressed groups across taxa (e.g., chemical secretions in fi sh, Barcellos et al. 2011; fecal odor in chickens and rats, Bombail et al. 2018), but we are aware of only one other study that explored the effects of food stress on preen oil secretions (Reneerkens et al. 2007), which are often used as a proxy for body odor in birds (Caro et al. 2015). Red knots exhibit seasonal shifts in preen oil Table 1 Results of permutational multivariate analysis of variance using Bray – Curtis dissimilarity matrices to test for effects of treatment and sex (Test 1) and sampling-time and sex (Test 2) on preen oil composition in captive song sparrows Group df Sum of squares Mean sum of squares F R 2 P Test 1 Treatment (post-exposure: control, methylmercury, stress, co-exposure) 3 0.11 0.04 0.85 0.05 0.52 Sex 1 0.17 0.17 3.78 0.07 0.03 Treatment × sex 3 0.18 0.06 1.35 0.08 0.23 Residuals 42 1.86 0.04 – 0.80 – Test 2 Time (pre-exposure/post-exposure) 1 0.56 0.56 23.42 0.13 >0.0001 Sex 1 1.00 1.00 41.62 0.23 >0.0001 Time × sex 1 0.37 0.37 15.39 0.09 >0.0001 Residuals 99 2.37 0.02 – 0.55 – Fig. 2 Two-dimensional nonmetric multidimensional scaling plot of preen oil composition in captive song sparrows following 8 weeks ’ exposure to methylmercury and/or unpredictable food stress. Bray – Curtis dissimilarity values were calculated from standardized and log (x + 1) transformed abundance data. Axis scales are arbitrary. 2D stress = 0.08 (see Fig. 1 for details). Treatment groups did not signi fi cantly differ after 8 weeks of exposure (F 1,49 = 0.92, R = 0.02, P = 0.46) Fig. 3 Change in song sparrow preen oil chemical composition grouped by ( a ) 8-weeks ’ exposure to methylmercury (F 1,49 = 1.72, P = 0.20) or ( b ) 8-weeks ’ exposure to unpredictable food stress (F 1,49 = 13.31, P = 0.001). Horizontal line shows group means, error bars represent SEM. Asterisk denotes signi fi cance at α = 0.05. Note that the same birds were included in both a and b: a The unexposed group includes control birds and food stress only birds ( n = 22) while the methylmercury group includes methylmercury-exposed and co- exposed birds ( n = 27). b The unstressed group includes control birds and methylmercury-exposed birds ( n = 26); the food stress group includes food stressed birds and co-exposed birds ( n = 23) Food stress, but not experimental exposure to mercury, affects songbird preen oil composition 281 composition from monoester to diester secretions; however, when stressed via food restriction fewer birds switch to diester production and the switch back to monoester pro- duction occurs earlier in food restricted compared to unstressed birds (Reneerkens et al. 2007). Though we did not identify which compounds changed over the study period, only monoesters have been found in the preen oil of song sparrows (Grieves et al. 2019c) and the closely related white-throated sparrow (Thomas et al. 2010). In our study, the change in preen oil chemical composition over eight weeks was greater in song sparrows subjected to unpre- dictable food stress compared to song sparrows with ad libitum access to food. Oily secretions of the uropygial gland can provide a means of pollutant depuration and these secretions are increasingly used to measure environmental contaminants such as DDT, polychlorinated biphenyls (PCBs), and persistent organic pol- lutants in birds (e.g., Yamashita et al. 2007; Jaspers et al. 2008; Espín et al. 2016b; Solheim et al. 2016; López-Perea and Mateo 2019). While these studies measured the amount of contaminant in preen oil, we are aware of only three prior studies that measured the effects of contaminants on the che- mical composition of preen oil itself (Gutiérrez et al. 1998; Leclaire et al. 2019; López-Perea and Mateo 2019). First, the preen oil volatiles in feral pigeons were altered by dietary exposure to the essential trace element zinc but not to the toxic metal lead (Leclaire et al. 2019), despite previous evidence that, like methylmercury, lead negatively affects feather quality and reproductive success in birds (Chatelain et al. 2017; Sample et al. 2019). Second, intra-muscular injection of lin- dane organochlorine pesticides (toxic contaminants that can accumulate in the uropygial gland) did not affect the preen oil of pigeons (Gutiérrez et al. 1998). Third, PCB exposure altered the preen oil composition of common moorhens, Gallinula chloropus . Birds exposed to higher levels of PCBs had shorter wax esters and fatty acids compared to birds exposed to lower levels of PCBs (López-Perea and Mateo 2019). Dietary supplementation with zinc or other trace ele- ments may stimulate the production of costly volatiles or other preen oil chemicals, thus leading to changes in preen oil composition (Leclaire et al. 2019). Indeed, at least one other study has shown effects of dietary supplementation (as opposed to contamination) on preen oil composition; captive white-throated sparrows supplemented with sesame oil had longer chain monoesters than birds sup- plemented with a fi sh oil diet (Thomas et al. 2010). In contrast, toxic, non-essential elements like methylmercury (this study), lead (Leclaire et al. 2019), or lindane (Gutiérrez et al. 1998) do not appear to affect preen oil composition, suggesting that the production and main- tenance of preen oil is relatively robust to these con- taminants. However, we note that PCB contaminants have recently been shown to alter preen oil wax ester compo- sition in free living birds (López-Perea and Mateo 2019). We cannot exclude the possibility that a higher level of methylmercury exposure may affect preen oil composition. That said, the mean blood THg level of our post-treatment birds was 5.2 ppm, and only 4% of wild birds sampled in western North America exceeded a blood-equivalent THg content of 4.0 ppm (reviewed in Ackerman et al. 2016), so it seems unlikely that most wild birds would be exposed to such high levels of methylmercury under natural conditions. Similarly, we cannot exclude the possibility that exposure to methylmercury or other contaminants may alter volatile compounds present in preen oil, which were not measured in our study. Similarly, accumulation of methylmercury in the preen gland or in preen oil itself might affect microbial communities of the preen gland and/or feathers, as has been found with other trace metal pollutants (Chatelain et al. 2016). Further work is needed to determine whether preen oil volatile compounds and microbial communities are affected by exposure to contaminants such as methylmercury. Acknowledgements We thank T. Kelly and R. W