Methylmercury (MeHg) is a toxic form of mercury that bioaccumulates in organisms and biomagnifies through food webs. MeHg concentrations can be high in aquatic environments, and this puts high trophic-level predators who derive energy originating from aquatic environments at risk of toxic effects. Due to the potential for bioaccumulation of MeHg over an individual’s life, the risk of MeHg toxicity may increase as animals age, and this risk may be especially high in species with relatively high metabolic rates. Total mercury (THg) concentrations were measured from the fur of adult female little brown bats (Myotis lucifugus) collected between 2012–2017 in Salmonier Nature Park, Newfoundland and Labrador. Using linear mixed-effects models, the effects of age, year, and day of capture on THg concentrations were evaluated and interpreted with AICc and multi-model inference. We expected that THg concentrations would increase with age, and that due to annual summer moulting, individuals captured earlier in the season would have lower THg concentrations than individuals captured later in the season. Contrary to expectations, THg concentrations decreased with age, and date of capture did not explain any variation in concentration. Among individuals, there was a negative relationship between the initial THg concentration of an individual and the rate of change in THg concentrations with age. Using a regression analysis, we found evidence of a population-level decline in THg concentrations in fur over the 6-year study period. Overall, the results indicate that adult female bats eliminate enough MeHg from their tissues to affect a decrease in THg concentrations in their fur over time and that young adults are potentially at the greatest risk of experiencing toxic effects from high MeHg concentrations; this could result in reduced reproductive output, and warrants further research.

Study area

The fur specimens used in this study were collected between 2012 and 2017 in and adjacent to SNP, NL, Canada (Lat: 47.3º, Long: -53.3º). Salmonier Nature Park is a 1455-ha nature reserve that is part of the Avalon Forest Ecoregion which is characterized by its ribbed moraines covered in forests that are interspersed with lakes and bogs. The atmospheric deposition rate of Hg in Newfoundland is among the highest in Canada (Chételat et al. 2020) and surface waters in SNP exhibit low pH due to the low acid neutralization ability of the soil and rock (Clair et al. 2007). Since low pH is correlated with increased methylation of Hg (Zhu et al. 2018), high mercury deposition and low surface water pH make wildlife in SNP susceptible to accumulating harmful concentrations of MeHg (Little et al. 2015a).

Sample collection

Myotis lucifugus were captured in SNP using mist nets (Avinet, Dryden, New York, USA) and harp traps (Austbat Research Equipment, Lower Plenty, Victoria, Australia) set near maternity roosts. Adult M. lucifugus are sexually segregated during their summer active season, and targeted efforts near maternity roosts thus resulted in captures comprised almost exclusively of adult females and juveniles. Bats were caught throughout their active period in SNP, with the earliest capture on May 15 and the latest capture on August 13. Age class (adult or juvenile) was determined based on the degree of ossification of the third metacarpal-phalangeal joint (Kunz & Anthony 1982). Captured bats had a passive integrated transponder (PIT) tag (0.09 g; EID-ID100 implantable transponders, EIDAPInc, Sherwood Park, Alberta, Canada and Trovan Electronic Identification Systems, UK) subdermally injected between the scapulae. Recaptured bats already carrying a PIT tag were identified using a handheld PIT tag reader, providing data on time between captures. Individuals first captured as adults were assigned a ‘minimum age' of 1. In the subsequent captures of these individuals, they were assigned a ‘minimum age’ which corresponded to the number of years since their first capture. Individuals initially captured as juveniles were assigned ‘exact’ ages for each capture. Fur samples were taken from between the scapulae using cuticle scissors. To prevent cross-contamination of fur, scissors were cleaned between individuals by washing in bleach, then water, followed by ethanol. Fur was placed in polypropylene vials and frozen at -20 or -80^oC. Differences in storage temperature were a result of space limitations; however, given the stability of Hg in fur, these differences are not expected to change the THg content (Phelps et al. 1980). Forearm length and mass were measured, and sex was determined. Reproductive status (pregnant, lactating, post-lactating, not obviously pregnant, or non-reproductive) was determined by palpitation of the abdomen and visual inspection of the nipples for signs of lactation (Racey & Swift 1985). Following processing, all animals were released at the site of capture. All animal handling and sample transport was done in accordance with the protocols approved by St. Mary’s University Animal Care Committee and the Government of Newfoundland and Labrador (SMU-MSVU protocols 11-18, 12-17, 13-15, 14-10A, 15-12, 16-12; Government of Newfoundland and Labrador protocols IW2011-14, IW2012- 26, IW2013-10, IW2014-31, IW2015-52, IW-2015-11, WLR2016-12, WLR2017-16, WLR2017-17).

Mercury analysis

Due to a limited sample size of male bats, only fur from adult female bats was analysed. Prior to Hg analysis, fur samples were visually examined for external sources of contamination such as dirt. Fur with visible contamination was cleaned with a Kimwipe (Kimberly-Clark, Mississauga, ON). To avoid loss of sample mass, fur was not washed to remove other potential sources of exogenous contamination. Chételat et al. (2018) and Little et al. (2015b) both found that washing then drying bat fur does not alter THg content. Thus, we assumed any exogenous contamination was negligible for THg and MeHg analyses.

Fur samples were weighed (± 0.001 mg) and analysed for THg content using a direct mercury analyser in accordance with Environmental Protection Agency (EPA) method 7473 in the Biotron Centre for Experimental Climate Change Research at Western University, London, Ontario (Milestone Tricell DMA 80, Milestone, Inc., Shelton, CT). The mean weight of fur samples analysed for THg was 1.192 mg (SD = 0.626). For quality control, IAEA-086 human hair was used as an external certified reference material (CRM) to determine total mercury recovery (average relative percent recovery = 101 %). Analytical precision was determined using sample duplicates (n = 10; average relative percent difference = 5%). As determined using the definition and procedure outlined in EPA (2016), the method detection limit for THg was 0.07 ng and no samples fell below this limit. THg concentrations are reported in µg/g fresh weight (fw).

Methylmercury was analysed using a modified version of EPA method 1630 in an ISO 17025 accredited laboratory in the Biotron Centre for Experimental Climate Change Research at Western University using a Tekran^® 2700 automated methylmercury analyser. The mean weight of fur samples analysed for MeHg was 1.696 mg (SD = 0.265). Percent MeHg recovery was determined using a human hair CRM (IAEA-086) (average relative percent recovery = 110 %). Analytical precision was determined using a sample duplicate (n = 1; relative percent difference = 2%). As determined using the definition and procedure outlined in EPA (2016), the method detection limit for MeHg was 0.051 ng/g and no samples fell below this limit. MeHg concentrations are reported in µg/g fw.

Statistical analysis

We used THg concentrations in 159 fur samples (representing 104 individuals), 9 of which were subsampled and analysed for MeHg concentration. A regression analysis was used to estimate the proportion of THg in the fur comprised of MeHg. An alpha value of 0.05 was used, and residuals of analyses were assessed to check assumptions. All data analysis was performed in R version 4.0.3 (R Core Team 2020).

To assess trends in THg concentrations among calendar years, a regression analysis was performed on THg data collected from the fur of 60 randomly selected (10 samples/calendar year) adult females. To avoid sample bias towards older individuals, only bats that had a ‘minimum age’ of 1, and thus of unknown age, were selected. To minimize within-year variability that may result from factors such as moulting and/or insect availability, the analysis was limited to samples collected from adult females captured between July 29 and August 7 (2012–2017). These dates were selected based on the coincidence of sampling periods in all years.

To characterize how THg concentrations in fur vary with bat age, we used fur samples from 47 adult female bats that were captured at least twice over the study period (mean = 2.21). Using the R package lme4 (Bates et al. 2015), we created multiple linear mixed-effects models from three a priori selected parameters (Burnham & Anderson 2002) including minimum age, Julian day, and year. Julian day is the day-of-year the fur sample was taken (day 1 to day 365) and was included to account for annual moulting, which has been suggested to occur during summer months (Fraser et al. 2013). If moulting occurs in the summer months, individuals captured earlier in the year may not have moulted at the time of capture. If THg increases with age, we would thus expect earlier captures to have a slightly lower THg relative to captures later in the year. Year was included to account for variability among calendar years. To account for differing intra-annual trends among years, an interaction term including year and Julian day was included. There was no evidence of multicollinearity among the parameters, minimum age, year, and Julian day (variance inflation factor < 5) as determined using the vif function in the R package, car (Fox & Weisberg 2019).

Results from nine a priori linear mixed-effects models and one null random effects model (Table 2) were evaluated using Akaike’s information criterion corrected for small sample size (AIC_c) (Burnham & Anderson 2002) with the MuMIn package in R (Bartoń 2019). Following the procedure in Zuur et al. (2007), and using AIC values, the optimal random effect structure was identified as a random intercept and slope structure (Table 3). Thus, each mixed-effects model included random intercepts and slopes, with individual as the random effect. Using multimodel inference, we calculated unconditional estimates for parameters included in models that comprised ≥ 95 % of the confidence set.