# Data from: Whiskers provide time-series of toxic and essential trace elements, Se:Hg molar ratios, and stable isotope values of an apex Antarctic predator, the leopard seal

## Citation

Charapata, Patrick et al. (2022), Data from: Whiskers provide time-series of toxic and essential trace elements, Se:Hg molar ratios, and stable isotope values of an apex Antarctic predator, the leopard seal, Dryad, Dataset, https://doi.org/10.5061/dryad.ksn02v75b

## Abstract

In an era of rapid environmental change and increasing human presence, researchers need efficient tools for tracking contaminants to monitor the health of Antarctic flora and fauna. Here, we examined the utility of leopard seal whiskers as a biomonitoring tool that reconstructs time-series of significant ecological and physiological biomarkers. Leopard seals (*Hydrurga leptonyx*) are a sentinel species in the Western Antarctic Peninsula due to their apex predator status and top-down effects on several Antarctic species. However, there are few data on their contaminant loads. We analyzed leopard seal whiskers (n = 18 individuals, n = 981 segments) collected during 2018–2019 field seasons to acquire longitudinal profiles of non-essential (Hg, Pb, and Cd) and essential (Se, Cu, and Zn) trace elements, stable isotope (ẟ_{15}N and ẟ_{13}C) values and to assess Hg risk with Se:Hg molar ratios. Whiskers provided between 46 and 286 cumulative days of growth with a mean ~125 days per whisker (n = 18). Adult whiskers showed variability in non-essential trace elements over time that could partly be explained by changes in diet. Whisker Hg levels were insufficient (<20 ppm) to consider most seals being at “high” risk for Hg toxicity. Nevertheless, maximum Hg concentrations observed in this study were greater than that of leopard seal hair measured two decades ago. However, variation in the Se:Hg molar ratios over time suggest that Se may detoxify Hg burden in leopard seals. Overall, we provide evidence that the analysis of leopard seal whiskers allows for the reconstruction of time-series ecological and physiological data and can be valuable for opportunistically monitoring the health of the leopard seal population and their Antarctic ecosystem during climate change.

## Methods

**Whisker collection**

Leopard seal whiskers were collected between April–May in 2018 and 2019 during field work conducted at the U.S. Antarctic Marine Living Resources (AMLR) Program research station on Cape Shirreff, Livingstone Island, Antarctic Peninsula (National Marine Fisheries Service permit #19439 and Antarctic Conservation Act permit #2018-016) (Fig. 1). Leopard seals were chemically immobilized using a butorphanol-midazolam protocol administered with a jab stick following Pussini and Goebel (2015). While sedated, morphometric data were collected (e.g., mass, kg; length, cm; girth, cm). The longest whisker was plucked with the root intact from the muzzle of each seal and stored in a sterilized plastic Whirl-Pak® (Madison, WI, USA) at ambient temperature. The final study collection consisted of 18 whiskers from 15 females (n = 1 juvenile and n = 14 adults) and 3 males (n = 3 adults). Following shipping to Baylor University, whiskers were stored at −80 °C until analysis. Whisker length and mass were measured using digital calipers (± 0.01 mm, Neiko 01407A) and a Mettler Toledo microbalance (± 0.1 mg).

**Scaled body mass index (SBM index)**

Scaled body mass (SBM) index was calculated for each sampled leopard seal using the equation (Peig and Green, 2009):

*M _{p} = M_{i} * (L_{o}/L_{i})^b_{sma}*

Where *M _{i}* was the mass (kg) of the seal,

*L*was the seal's standard length (cm),

_{i}*L*was the mean standard length of the different age classes (juvenile n = 1 female) or adult (n = 17 [n = 3 males and n = 14 females]), and

_{o}*b*was the scaling exponent from plotting natural log transformed mass by standard length for all seals and the Mp was the predicted mass of the individual seal when standardized to

_{sma}*L*. This SBM index has been shown to scale more accurately with growth and mass compared to other condition indices used in wildlife studies, including pinnipeds (DeRango et al., 2019; Peig and Green, 2009).

_{o}**LA-ICP-MS analysis of trace elements along whiskers**

Whiskers were prepared for laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) analysis by removing visible surface contaminants (e.g., external root sheath fragments) by wiping with a Kimwipe moistened with a 2:1 chloroform methanol solution (Keogh et al., 2021; Rea et al., 2015), and drying for ≥24 h in a ventilation hood. Cleaned whiskers were shipped in sealed Whirl-Pak® bags to The University of Texas at Austin, and then stored in a desiccator until LA-ICP-MS analysis.

Continuous elemental (Hg, Pb, Cd, Se, Cu, Zn) base-to-tip whisker concentrations were measured by LA-ICP-MS, using an ESI NWR193 excimer laser ablation system (193 nm, 4 ns pulse width) coupled to an Agilent 7500ce ICP-MS. Whiskers, ranging from 48.3 to 99.4 mm in length, were mounted on double-sided stick tape. The most stable mounts were achieved by allowing whiskers to best retain their natural curvatures in 2D; whiskers mounted in a straight line were found to move over time. Coordination of whisker transects involved establishing long segmented lines with 1–2 nodes placed per mm, and each node adjusted in x-y to follow the central growth axis. The z axis was adjusted to maintain laser focus along the surface of the traverse as whiskers greatly taper from base-to-tip. The LA-ICP-MS system was optimized for sensitivity across the atomic mass unit (AMU) mass range and low oxide production (ThO/Th: 0.28 ± 0.01) by tuning on a standard (NIST 612). Final parameters where whisker ablations were obtained from trial transects on representative areas of trial whisker samples (via iterative tests of energy density, repetition rate, and gas flow) to obtain robust and consistent elemental signals free from spectral skew). Following pre-ablation (100 μm spot, 100 μm/s scan rate, 2.7 J/cm^{2} energy density [fluence]) to remove shallow superficial contaminants, a single base-to-tip transect was performed along the center of each whisker, using a 90 μm diameter spot, 100 μm/s scan rate, 2.44 ± 0.07 J/cm^{2} energy density, 10 Hz repetition rate, and carrier gas flows (L/min) of 0.85 for Ar and 0.85 for He. The quadrupole time-resolved method measured eight masses with integration times of 10 ms (34S, 63Cu, 64Zn, 83Zr), 205 ms (202Hg), and 100 ms (82Se, 114Cd, 208Pb). Measured intensities were converted to elemental concentrations (ppm) using iolite software (Paton et al., 2011), with 34S as the internal standard and a S index value of 5 wt% for whisker unknowns (Legrand et al., 2004; Noël et al., 2016; Rodushkin and Axelsson, 2003; Stadlbauer et al., 2005). Signals were converted to base-to-tip distance (μm) along the whisker based on the scan rate and duty cycle. Any data points that fell below detectable concentrations were assigned ½ the concentration of the limit of detection calculated for the analytical run of each whisker (Clark et al., 2021; Gilbert, 1987, Table S1). Outliers were defined as concentrations measured along the whisker >4 standard deviations above the mean and removed from statistical analysis (Clark et al., 2021; Tukey, 1977).

Previous researchers have used commercially available human hair as an appropriate reference standard (wt% S content, keratin matrix) for mammal hair and whisker LA-ICP-MS studies (e.g., Noël et al., 2016, Noël et al., 2014). However, these standards have only sub-ppm concentrations for several of the metals assessed (Se, Hg) during this study. Thus, we made and validated reliable standards using methods described in the Supplementary Material.

**Stable isotope analysis and timestamps**

After whiskers were analyzed for trace elements using LA-ICP-MS analysis they were returned to Baylor University for sampling of bulk carbon (ẟ^{13}C) and nitrogen (ẟ^{15}N) stable isotope analysis (Fig. 2). Lipids were removed by cleaning each whisker with a 1:1 ethanol: methanol solution, following previous leopard seal whisker stable isotope ratios studies (Botta et al., 2018; Rogers et al., 2016). After allowing whiskers to dry in a ventilation hood (≤ 24 h), whiskers were sectioned into 0.50 ± 0.01 mm lengths (using digital calipers and a hand chisel) to enable fine-scale comparison with the LA-ICP-MS trace element time-series; segment lengths were somewhat longer near the frayed tip of whiskers to obtain the minimum required mass (~0.3 mg) for stable isotope analysis.

Carbon and nitrogen stable isotope analysis was performed in the Baylor University Stable Isotope Facility, using an Elemental Analyzer (EA) Costech 4010 Elemental Combustion System (ECS) paired with a Conflow IV interphase (Thermo Scientific) and Thermo Delta V Advantage continuous flow Isotope Ratio Mass Spectrometer (EA-IRMS). Prior to combustion and isotopic analysis, whisker segments were placed into pre-weighed tin capsules (Costech 5 × 9 mm), tin capsules reweighed with a Mettler Toledo XP26 digital scale (±0.001 mg). Whisker nitrogen (ẟ^{15}N) and carbon (ẟ^{13}C) isotope values are reported as the ratio of the heavy to light isotope relative to international standards; atmospheric nitrogen and Vienna Peedee Belemnite (VPDB), respectively, using the following equation:

*ẟX = [(R _{sample}/R_{standard})-1] *1000*

where X is the targeted isotope (nitrogen or carbon) ratio expressed in delta notation (ẟ) with units per mil (‰), R_{sample} is the isotopic ratio of heavy to light isotopes (15/14 N or 13/12C) of the sample, and R_{standard} is the isotopic ratio of heavy to light isotopes measured in the standard. A two-point calibration curve for calculating nitrogen ẟ^{15}N and ẟ^{13}C values of samples was established using USGS-40 and USGS-41A international standards. The accuracy and precision of isotopic measurements was calculated based on the long-term mean and standard deviation of 105 replicates of an internal lab standard (Acetanilide, reported ẟ^{13}C = −29.53 ± 0.01 ‰, ẟ^{15}N = 1.18 ± 0.02 ‰) measured during each analytical run (n = 3 replicates/run). The replicate grand averages obtained are within (ẟ^{15}N = 1.28 ± 0.17 ‰) or very close to (ẟ^{13}C = −29.45 ± 0.05 ‰) analytical uncertainty of reported values.

Acceptable atomic C:N ratios of whisker segments ranged from approximately 3.0–3.8 based on previous leopard seal whisker isotope studies (Botta et al., 2018; Rogers et al., 2016). Nearly all whisker segments had acceptable atomic C:N ratios (3.47 ± 0.13, 3.03–4.00, mean ± standard deviation (SD), min–max, respectively (Fig. S1); one segment with anomalously low atomic C:N ratio <2.8, was excluded from statistical analysis.

Whisker segments were assigned approximate timestamps relative to date of collection based on leopard seal whisker growth characteristics and the application of a discrete Von Bertalanffy growth model (Hall-Aspland et al., 2005; Rogers et al., 2016; von Bertalanffy, 1938). The discrete Von Bertalanffy equation can be written as:

δL/δT = K(L_{a} – L_{p}–1)

where δL/δT is growth rate for sections _{p-1} to _{p}. This can be rearranged to δL = (L_{p}–L_{p–1}) and δT = (T_{p}–T_{p-1}) allowing calculation of time intervals for whisker section(s) _{p-1} to _{p} as done in Hall-Aspland et al. (2005) and Rogers et al. (2016):

(T_{p} – T_{p–1}) = (L_{p} – L_{p–1})/[K(L_{a} – L_{p–1})]

where L_{p} is the total length of the whisker, L_{p–1} is the remaining length of the whisker after sampling section _{p}, K is the growth coefficient, and L_{a} is the asymptotic length of leopard seal whisker. We used the recommended K value of 0.013 and L_{a} of 101.2 mm developed by Rogers et al. (2016) for leopard seal whiskers and applied this equation to each section to acquire approximate segment growth in days. Cumulative δT values over respective segment intervals were then subtracted from the collection date of each whisker to develop an approximate timestamp for individual segments.

**Aligning trace element and stable isotope data**

The accurate alignment of trace element and stable isotope data per whisker was standardized to the segment lengths submitted for stable isotope analysis. For example, the first section of a whisker (i.e., section “1”) with a length of 0.50 mm submitted for isotope analysis meant a number “1” was assigned to all trace element data obtained during the 0.0–0.50 mm of whisker sampled during LA-ICP-MS (Fig. 2). This was repeated for subsequent sections until trace element data were assigned a section number that directly corresponded with stable isotope data. Mean and standard deviation values were calculated for trace element data based on its respective section number to accurately be paired with stable isotope data (Fig. 2). This approach worked for all but three whiskers (n = 3), where cumulative segment lengths for stable isotope analysis were greater than the trace element whisker length data. This most likely was due to error when measuring and sampling individual segments for stable isotope analysis. To correct this error in the three whiskers, the total sampling error was calculated (i.e., total whisker length prior to sampling minus cumulative segment lengths post sampling) for these three whiskers and the total error was divided by the total number of whisker segments. The resulting values ranged from 0.06 to 0.08 mm and were subtracted from section lengths for all segments. The process of integrating trace element with stable isotope data was rerun with the corrected segment lengths and resulted in all trace element data being assigned to a corresponding stable isotope segment.

**Statistical analysis**

Type II ANOVAs (F-tests for linear models) were used to determine significant differences in mean log10 transformed whole whisker trace elements, ẟ^{15}N, and ẟ^{13}C values between sexes, standard length, and SBM index in respective linear models (Garcia-Cegarra et al., 2021). Due to only having one juvenile, age class was not assessed among these whole whisker trace element and stable isotope data. Whole whisker analyte concentrations are calculated means across all whisker segments per individual, whereas segment analyte concentrations are averaged only across segment length (Fig. 2).

Pearson correlations of log10 transformed trace elements were performed within and among individuals of pooled sexes (n = 15 females, n = 3 males) using segmented and whole whisker trace element data, respectively. Among individual correlations were calculated using average whole whisker trace element concentrations (i.e., correlations among mean trace element concentrations for each individual seal) and can provide general insight into trace element associations within leopard seal whiskers (e.g., individuals with high concentrations of element “A” tend to have low concentrations of element “B”). In contrast, within individual correlation coefficients were calculated using a smoothed time series of trace element concentrations across the whisker of each individual. The resulting Pearson correlation coefficients were then Fisher-Z transformed, averaged across all individuals, and transformed back to a mean Pearson correlation coefficient. This approach reveals correlations among trace elements through time (i.e., along the whisker) that are consistent across seals, thus are likely to represent processes or phenomena that affect most or all leopard seals in this study, which may include things like physiological, temporal, and spatial intrinsic influences on trace element intake or uptake (e.g., Clark et al., 2021).

The R package SIBER was used to calculate intraindividual standard ellipse area (SEA) corrected for small sample sizes (SEAc) across whisker segment ẟ^{13}C and ẟ^{15}N values for each seal (Jackson et al., 2011; Scholz et al., 2020). The SEAc calculates the variability among ẟ^{13}C and ẟ^{15}N values across whisker segments to provide insight into range of trophic level and foraging locations (i.e., trophic niche) of an individual seal. We used SEAc values to understand how trophic niche width related to whole whisker trace element concentrations. We also calculated a “population” SEAc using whole whisker ẟ^{15}N and ẟ^{13}C values from all whiskers (n = 18) to compare with a previous leopard seal whisker study (Botta et al., 2018). Bivariate linear models of log10 transformed trace element data and ANOVAs (F-tests for linear models) were used to determine relationships of trace elements with intraindividual SEAc values. Linear models and Type II ANOVAs (F-tests for linear models) were used to assess intraindividual SEAc with sex, standard length, and SBM index.

Linear mixed models (LMMs) were constructed to determine relationships among trace elements with changes in ẟ^{15}N and ẟ^{13}C over time while incorporating sex and biometrics (standard length and SBM index) as covariates. All trace element data were log10 transformed to approximate normal distribution to meet LMM assumptions. Trace element data from adult leopard seal whiskers were modeled (n = 17 adults [n = 3 males and n = 14 females]). Full models were constructed using the R Studio software (RStudio Team, 2020) and the package lme4 (Bates et al., 2015) based on our objectives to assess temporal relationships among trace elements with changes in diet (ẟ^{15}N and ẟ^{13}C values), that also incorporated sex and biometric data (Zuur and Ieno, 2016). A numbered “Week” of the year (1–52) was assigned to individual segments relative to the earliest segment timestamp and only retained “Weeks” that included a minimum of three unique seals (total n = 17 seals, n = 834 segments, and n = 28 consecutive weeks). The full model took the form of: log10(trace element) ~ Standard Length (numeric) + SBM index (numeric) + Sex (factor, “Male” or “Female”) + Week (numeric) + Carbon (numeric, ẟ^{13}C of segment) + Nitrogen (numeric, ẟ^{15}N of segment) + Carbon*Week + Nitrogen*Week + (1|FieldSeason) (random intercept, controlling for whisker collection year) + (1|Seal.ID) (random intercept, controlling for differences in average concentrations among individuals). Biologically relevant permutations with the fixed effects of the full model were constructed to compare with the full model (Table S2). The selected model was determined based on lowest AICc and highest AIC weight (Burnham et al., 2011; Table S3). We assessed the full and selected models for each trace element by plotting residuals with fitted values, residuals with all covariates, and assessed the distribution of residuals (Zuur and Ieno, 2016).

Selected LMMs for each trace element had fitted and 95 % confidence intervals constructed using the “bootpredictlme4” package in R using n = 500 iterations to estimate the fit of selected models with the trace element data (Clark et al., 2021; Duursma, 2021). If the interaction terms were retained in the selected model, we predicted trace element concentrations over time keeping ẟ^{15}N and ẟ^{13}C at biologically relevant “lower”, “median”, and “upper” values, while keeping the other isotope and/or main effects at their median values, if applicable. For ẟ^{15}N, the lower value was 10.15 ‰ (12.5 % range of our ẟ^{15}N values, between 0 and 1st quartile), median value was 10.83 ‰, and upper value was 12.47 ‰ (87.5 % range of data, between 3rd and 4th quartiles of our ẟ^{15}N values). For ẟ^{13}C, the lower value was −22.74 (12.5 % range of data, between 0 and 1st quartile of our ẟ^{13}C values), median value = −21.98, and upper value was = −21.41 ‰ (87.5 % range of data, between 3rd and 4th quartiles of our ẟ^{13}C values). We then visually assessed the fit of model predictions with the leopard seal whisker trace element data from the four seals that fell within those stable isotope categories (i.e., “lower”, “median”, and “upper” ẟ^{13}C and ẟ^{15}N values) (Clark et al., 2021; Zuur and Ieno, 2016).

Mercury is a relatively well-studied toxin in pinnipeds with published Hg toxicological thresholds from hair concentrations (McHuron et al., 2019; O’Hara and Hart, 2018; Rea et al., 2020), which also correlate with whisker Hg concentrations (Noël et al., 2016). Previous studies suggest different toxicity thresholds for hair Hg concentrations that are associated with deleterious effects to wildlife and humans including; 5.4 ppm [μg/g dw] in brain tissue of polar bears (*Ursus maritimus*) which correlated with a reduction in genomic DNA methylation and NMDA receptors, 10 ppm in human infants was correlated with delayed development, and ~30 ppm in mink (*Neogale vison*) hair that had resulted in acute Hg toxicity and death in some individuals (Van Hoomissen et al., 2015; Yates et al., 2005). Since hair and whisker Hg thresholds are unknown for leopard seals, we followed Rea et al. (2020) and used Hg toxicological thresholds of <10 ppm to assign whole whisker and individual segment concentrations as “Low” risk, 10 to 20 ppm as “Moderate”, and “High” risk of Hg toxicity if Hg concentrations were >20 ppm (O’Hara and Hart, 2018; Rea et al., 2020). Additionally, molar Se:Hg ratios were calculated for each segment using the formula: (Se ppm /78.96)/(Hg ppm/200.59) following McCormack et al. (2021). An ANOVA with Tukey's Post Hoc honestly significant difference (HSD) was used to determine overall differences in mean Se:Hg molar ratios among risk groups (“low”, “moderate”, and “high”). We then visually assessed how Se:Hg molar ratios patterns changed over time with respect to Hg risk classification to analyze the potential importance of Se to leopard seals as a Hg detoxicant (Rea et al., 2020). An alpha level of 0.05 was used for threshold of significance for ANOVAs (F-tests for linear models). All data presented in figures are median ± interquartile range (IQR) due to high variability among data.

## Usage notes

Microsoft Excel can open the csv files and a text editor program (e.g., Microsoft Word) for the supplementary file.

## Funding

NSF, Award: 1644004

C. Gus Glasscock, Jr. Endowed Fund for Excellence in Environmental Science at Baylor University