Mercury accumulation, biomagnification, and relationships to δ 13 C, δ 15 N and δ 34 S of fishes and marine mammals in a coastal Arctic marine food web

Combining mercury and stable isotope data sets of consumers facilitates the quantification of whether contaminant variation in predators is due to diet, habitat use and/or environmental factors. We investigated inter-species variation in total Hg (THg) concentrations, trophic magnification slope between δ 15 N and THg, and relationships of THg with δ 13 C and δ 34 S in 15 fish and four marine mammal species (249 individuals in total) in coastal Arctic waters. Median THg concentration in muscle varied between species ranging from 0.08 ± 0.04 μ g g (cid:0) 1 dw in capelin to 3.10 ± 0.80 μ g g (cid:0) 1 dw in beluga whales. Both δ 15 N (r 2 = 0.26) and δ 34 S (r 2 = 0.19) best explained variation in log-THg across consumers. Higher THg concentrations occurred in higher trophic level species that consumed more pelagic-associated prey than consumers that rely on the benthic microbial-based food web. Our study illustrates the importance of using a multi-isotopic approach that includes δ 34 S when investigating trophic Hg dynamics in coastal marine systems.


Introduction
Humans are hyperkeystone species that are fundamentally changing the geological, biological and chemical components of Earth at unprecedented rates leading to a new geological epoch, The Anthropocene (Lewis and Maslin, 2015;Worm and Paine, 2016).Human industrial activity emits a wide array of chemical pollutants, such as toxic trace metals like mercury (Hg), which can pose toxicological consequences to biota across ecosystems and is of health concern to the global human population (Driscoll et al., 2013;Basu et al., 2022).Anthropogenically emitted Hg is transported to the Arctic from more southern latitudes via the atmosphere, rivers, oceans and migratory species (AMAP, 2021).Legacy mercury is also released into the environment via natural processes and as a result of climate change (e.g.increased permafrost thaw; AMAP, 2021).Furthermore, Hg concentrations in Arctic biota have been increasing over time (Morris et al., 2022) coinciding with rapidly warming temperatures in the Arctic leading to a wide suite of physical (e.g.sea ice declines) and ecological changes (e.g.inter-specific interactions and primary productivity; Kortsch et al., 2015;Yurkowski et al., 2018;Lewis et al., 2020).Of particular concern to animals is methylmercury (MeHg), which is converted from inorganic Hg by microorganisms and bioaccumulates then biomagnifies with each trophic step up the aquatic food web to longer-lived top predators and be at high risk for adverse effects (Kidd et al., 2012;Dietz et al., 2019;Dietz et al., 2022).In muscle of higher trophic level species, such as predatory fish and marine mammals, total mercury (THg) concentrations are considered equivalent to MeHg concentrations as only MeHg biomagnifies (Campbell et al., 2005;Barst et al., 2022).
Stable isotope ratios are biogeochemical tracers that provide timeintegrated information on consumer diet and habitat use (Layman et al., 2012).When measured with contaminants, stable isotopes can allow researchers to track chemical pollution through the food web and determine whether contaminant variation in top predators is a result of diet, habitat use, or other environmental factors (Loseto et al., 2008;Clayden et al., 2013).For example, nitrogen stable isotopes (δ 15 N) characterize the hierarchical structure of aquatic food webs indicating the consumer trophic position throughout the system (Minagawa and Wada, 1984).Since THg biomagnifies up the food web, relationships between THg and δ 15 N reveal the extent of Hg biomagnification via an estimate of the trophic magnification slope (TMS) -a measure of the extent to which Hg increases from lower to upper trophic levels in a food web (Lavoie et al., 2013).Other stable isotope ratios, such as δ 13 C and δ 34 S, can be effective predictors of contaminants that are associated with the habitat source of assimilated prey items (e.g., between terrestrial/ freshwater and marine resources or benthic and pelagic resources), as both of these stable isotope ratios show little variation between prey and predator compared to endmember sources of both habitat types (McCutchan Jr et al., 2003;Florin et al., 2011).
Sulfates, a form of dissolved sulfur found throughout the water column, are enriched in 34 S compared to sulfides, a form of sulfur that occurs within sediments, as a result of sulfate-reducing bacterial processes (Peterson and Fry, 1987;Connolly et al., 2004).As such, species inhabiting or foraging in benthic habitats, where sulfates are converted to sulfides, have a lower δ 34 S than their more pelagic-associated counterparts, who have a higher contribution of sulfates in their diet (Brunner and Bernasconi, 2005;Szpak and Buckley, 2020).Concomitantly, sulfate-reducing bacteria also produce MeHg from inorganic Hg but this production can potentially be mediated by the availability of sulfates throughout the water column leading to variation in Hg levels in biota (Jeremiason et al., 2006;Driscoll et al., 2013).This can result in a positive relationship between δ 34 S and THg, as shown for several seabird species and their prey (Elliott and Elliott, 2016;Góngora et al., 2018), but this relationship has not been investigated across the numerous fish and marine mammal species within Arctic food webs.
The marine waters around Southampton Island, Nunavut, Canada (Fig. 1) are considered a biological hotspot of productivity and abundance of top predators (Yurkowski et al., 2019), including a core area of Paleo-Inuit occupation for over 2500 years, suggesting this area has continued to have enhanced and reliable productivity over millennia (Clark, 1980).This area provides multiple ecosystem services by supporting Inuit subsistence harvesting (e.g.Atlantic walrus, beluga, narwhal, ringed seals) and local fisheries (i.e.Arctic char), therefore this region has recently been identified by Fisheries and Oceans Canada as an Area of Interest for marine conservation (Loewen et al., 2020).The presence of Hg across the Arctic may diminish the value of ecosystem services thus necessitating a study to increase knowledge of Hg levels in the marine waters of Southampton Island.In this study, we investigated the THg concentrations, biomagnification, and relationships to δ 13 C, δ 15 N and δ 34 S for 15 fish and four marine mammal species in coastal waters around Southampton Island.Specifically, our objectives were to: 1) quantify differences in the THg concentrations between sampled fish and marine mammal species in the Southampton Island food web, 2) determine the THg trophic magnification slope and intercept across lower and upper trophic level fish and marine mammals in the Southampton Island food web, and 3) quantify relationships between δ 13 C and δ 34 S values and THg concentrations across all sampled species in the context of variability in habitat type.

Sample collection
Fish samples were collected in August and September 2016, 2018 and 2019 on research expedition cruises in the marine waters around Southampton Island, Nunavut, Canada aboard the RV William Kennedy and the MV Nuliajuk, as well as from opportunistic collections from the breeding ledges of thick-billed murres (Uria lomvia) at Coats Island (Fig. 1).Ships towed a 3-m pelagic trawl (0.5-cm cod-end mesh) at 2-3 knots for 15 min targeting the subsurface chlorophyll maximum depth, and a 3-m benthic trawl (0.5-cm cod-end mesh) at 2-3 knots for 15 min on the bottom following approved research licenses and animal care protocols (see Acknowledgements for specific details).Ninety muscle samples were collected from 14 fish species, including Arctic cod (Boreogadus saida), polar cod (Arctogadus glacialis), fourline snakeblenny (Eumesogrammus praecisus), Arctic staghorn sculpin (Gymnocanthus tricuspis), daubed shanny (Leptoclinus maculatus), slender eelblenny (Lumpenus fabricii), capelin (Mallotus villosus), Arctic sculpin (Myoxocephalus scorpioides), shorthorn sculpin (Myoxocephalus scorpius), Greenland cod (Gadus ogac), banded gunnel (Pholis fasciata), Arctic shanny (Stichaeus punctatus), moustache sculpin (Triglops murrayi), and ribbed sculpin (Triglops pingelii).Arctic char (Salvelinus alpinus) muscle samples were collected opportunistically in August 2019 as part of Inuit subsistence harvests in association with an ongoing community-based monitoring programs with Fisheries and Oceans Canada based in Naujaat, Nunavut (Fig. 1).Atlantic walrus (Odobenus rosmarus rosmarus), ringed seal (Pusa hispida), narwhal (Monodon monoceros) and beluga (Delphinapterus leucas) muscle and liver samples were collected from June to September in 2016, 2018 and 2019 by Inuit hunters from Naujaat and Coral Harbour, Nunavut as part of their subsistence harvests and ongoing community-based monitoring programs in collaboration with Fisheries and Oceans Canada.Tissue samples were frozen at − 20 • C and shipped to the Freshwater Institute in Winnipeg, Manitoba before processing.

THg analysis
Prior to THg analysis, fish and marine mammal samples were freezedried for 48 h to remove moisture and all THg data is presented in μg g − 1 dry weight.Freeze-dried sample were weighed into a tared vessel, thermally decomposed and released THg was quantified THg using a direct Hg analyzer in accordance with USEPA7473 (NIC MA-3000, Nippon Instruments, North America) at McGill University as described in Golzadeh et al. (2020).Quality assurance steps included the analysis of DOLT-5 (dogfish liver) and S3 (Ontario chinook salmon) certified reference materials with each batch of 15 samples.Precision was indicated by relative standard deviation of sets of triplicate samples (n = 11) and ranging from 0.96 % to 10.4 % with a mean across all sets of <5 %.Accuracy was measured by THg concentrations of DOLT-5 (0.33 ± 0.077 μg g − 1 dw; certified THg value is 0.44 ± 0.18 μg g − 1 ) and S3 (0.91 ± 0.02 μg g − 1 ; certified THg value is 0.99 ± 0.06 μg g − 1 ) where the mean accuracies were 74.2 ± 3.3 % and 92.2 ± 4.7 %, respectively.THg was not detected in the blanks and all samples were above the mean detection limit of 0.00024 μg.

Stable isotope analysis
Frozen fish muscle and marine mammal muscle and liver samples were freeze-dried for 48 h and then crushed into a fine powder using a mortar and pestle.Due to the presence of lipids in Arctic marine mammal liver and muscle (Yurkowski et al., 2015), and fish muscle (Post et al., 2007), lipids were extracted with 2:1 chloroform:methanol solvent using a modified version of the Bligh and Dyer (1959) method.Stable isotope analysis was performed at the Chemical Tracers Laboratory, Great Lakes Institute for Environmental Research, at the University of Windsor using a Delta V Advantage Mass spectrometer (Thermo Finnigan, San Jose, CA, USA) coupled to a Costech 4010 Elemental Combustion system (Costech, Valencia, CA, USA) and a ConFlo IV gas interface.For δ 13 C and δ 15 N analysis, subsamples of 400-600 μg were weighed into tin capsules.For δ 34 S, 3000-6000 μg of the sample plus 300-500 μg of vanadium pentoxide was encapsulated.All stable isotope ratios are expressed in per mil (‰) in standard delta (δ) notation relative to the international standards Pee Dee Belemnite for carbon, atmospheric N 2 for nitrogen, and Vienna Cañon Diablo Triolite for sulfur using the following equation: δX = [(R sample /R standard ) -1] x 10 − 3 , where X is 13 C, 15 N or 34 S and R equals 13 C/ 12 C, 15 N/ 14 N and 34 S/ 32 S. Instrumental accuracy checked throughout the period of time that these samples were analyzed was based on U.S. National Institute of Standards and Technology (NIST) standards 8573, 8547 and 8574 for δ 15 N, 8542, 8573 and 8574 for δ 13 C (n = 50 for all), and 8555 and 8529 for δ 34 S (n = 30 for all).The mean difference from the certified values were − 0.09, 0.14, − 0.06 ‰ for δ 15 N, 0.09, 0.01 and − 0.08 ‰ for δ 13 C, and 0.25 and 0.30 ‰ for δ 34 S, respectively.Precision, assessed by the standard deviation of replicate analyses of three standards (NIST1577c, internal lab standard, tilapia muscle), USGS 40 and Urea (n = 22 for all), measured ≤0.25 ‰ for δ 15 N and ≤ 0.13 ‰ for δ 13 C for all the standards.For δ 34 S, the precision from standards USGS 42, NIST 8555 and NIST 8529 (n = 39 for all), measured ≤0.43 ‰.

Statistical analysis
All statistical analyses were performed in R v.4.1.2(R Core Team, 2021) using packages "ggplot2" (Wickham, 2016) and "AICcmodavg" (Mazerolle and Mazerolle, 2017) where α was set to 0.05.Only muscle samples of fishes and marine mammals were used for analysis due to sample size.Prior to analysis, THg concentrations were log transformed to achieve normality.THg concentrations across fish and marine mammal species were compared using one-way analysis of variance (ANOVA), followed by a Tukey's post hoc test.Pearson's correlation was used to evaluate the correlation between log transformed THg concentration and δ 13 C, δ 15 N and δ 34 S of muscle across species altogether.General linear models were used to investigate relationships between THg concentration (dry weight and log-transformed) as the dependent variable and δ 13 C, δ 15 N and δ 34 S as the independent variables across all species.Due to differences in sample size among species, we averaged log-THg concentration, δ 13 C, δ 15 N and δ 34 S values for all species except daubed shanny (n = 1) prior to correlation, partial correlation and linear regression analyses.Models were evaluated and ranked using Akaike's Information Criterion adjusted for small sample sizes where the most parsimonious model is one with the lowest ΔAIC c (Burnham and Anderson, 2004).If models were within two ΔAIC c of the lowest AIC c model, we considered the model with the fewest number of estimated parameters and with the lowest standard error in the parameter estimate to be the best model following Burnham and Anderson (2004).
Trophic magnification slope was determined as the slope value between mean log THg concentration of muscle and mean δ 15 N of muscle across all species.Relationships between mean log THg concentration of muscle and mean δ 13 C and δ 34 S of muscle across all species were considered to indicate associations with habitat type (i.e., pelagic or benthic).The effect of δ 15 N on log THg while controlling for habitat type (i.e., δ 34 S) was determined using partial correlation analysis which estimates the unique contribution of an independent variable, in this case δ 15 N and δ 34 S, to the r 2 of the model.We also used residual plots to test linearity in the relationships between log THg, δ 15 N and δ 34 S (Zuur et al., 2010).To test for differences in the slope and intercept between log THg and δ 15 N and δ 34 S, we used linear analysis of covariance.

Inter-specific variation in THg concentrations of muscle
Muscle THg concentrations varied widely among fish and marine mammal species, with median concentrations ranging from 0.08 ± 0.04 (median ± SD) μg g − 1 dw in slender eelblenny and capelin to 3.10 ± 0.80 μg g − 1 dw in beluga (Table 1).One-way analysis of variance (ANOVA) revealed significant differences in muscle THg among species (F 18,162 = 29.95,p < 0.001).Tukey's post-hoc tests revealed four species groups (Fig. 2).The first species group consisted of capelin, slender eelblenny and daubed shanny, for which the THg concentrations were significantly lower than all other species.The second species group, composed of Arctic char, polar cod, ribbed sculpin, moustache sculpin, Arctic staghorn sculpin and Arctic cod, was significantly lower than species in the third species group (i.e., shorthorn sculpin, fourline snakeblenny, ringed seal, Atlantic walrus, Arctic shanny, Greenland cod and Arctic sculpin; p < 0.05).The THg concentrations of banded gunnel D.J. Yurkowski et al. were not significantly different from any species in the second or third clusters (p > 0.05).The fourth species group consisted of narwhal and beluga, which had a significantly higher THg concentrations compared to all other species (p < 0.001).

Discussion
This study provides the first characterization of mercury concentration, its relationship with δ 13 C, δ 15 N and δ 34 S, and the level of biomagnification across numerous fish and marine mammal species representing a portion of an Arctic coastal marine food web.As such, this study can provide a reference point to assess future trends in stable isotope and mercury dynamics of fishes and marine mammals in northern Hudson Bay.Fish species showed wide variation in THg concentrations with the highest occurring in shorthorn sculpin, fourline snakeblenny, Arctic shanny and Arctic sculpin whose THg concentrations (mean of 0.56-0.77μg g − 1 dw) were comparable to their conspecifics from other locations in the Canadian Arctic (see Loseto et al., 2008;Braune et al., 2015;McMeans et al., 2015;Pedro et al., 2017;Barst et al., 2022).The mean THg concentrations of Atlantic walrus, ringed seal, beluga and narwhal around Southampton Island (mean of 0.67-3.10μg g − 1 dw) were also similar to their conspecifics at other locations across the Canadian Arctic and in west Greenland (Hansen et al., 1990;Atwell et al., 1998;Rigét et al., 2007;Loseto et al., 2008;Braune et al., 2015;Brown et al., 2016).Moreover, both δ 15 N and δ 34 S were positively associated with THg concentration across all sampled fish and marine mammal species in the Southampton Island marine food web, providing some support for the influence of trophic position and habitat type (related to the sulfate availability hypothesis), consistent with results from other studies on seabirds and their prey (Elliott and Elliott, 2016;Góngora et al., 2018).

Inter-specific variation in THg
In general, some of the known benthic-associated fish species (i.e., shorthorn sculpin, fourline snakeblenny, Arctic shanny and Arctic staghorn sculpin) showed the highest THg concentrations as well as the highest δ 15 N values and in turn, trophic position among fish species.These benthic fishes are understudied in the Arctic but have been documented to exhibit a generalist feeding strategy by consuming a wide variety of invertebrate (i.e.copepods, amphipods, decapods and polychaetes) and fish species from both the pelagic and benthic trophic pathways (Keats et al., 1993;Giraldo et al., 2016;Landry et al., 2018).Furthermore, larger-sized benthic fishes, such as shorthorn sculpin, can consume Arctic cod and cannibalize smaller conspecifics due to their relatively larger gape size which likely contributes to higher dietary exposure to Hg (Landry et al., 2018;Hilgendag et al., 2022).Thus, shorthorn sculpin, along with other sculpin species, typically have a relatively higher THg concentration among fishes, which has also been observed in the Beaufort Sea (Loseto et al., 2008), West Greenland (Rigét et al., 2007), in Frobisher Bay (Hilgendag et al., 2022) and several other areas across the Eastern Canadian Arctic (Pedro et al., 2017(Pedro et al., , 2019)).Despite the relatively higher THg concentrations observed in shorthorn sculpin, fourline snakeblenny, Arctic shanny and Arctic sculpin, these four species are at low risk for sublethal toxic effects of Hg, while all other fish species from this study are considered at no risk for Hg toxicity based on risk categories and thresholds described in Barst et al. (2022).
The mean THg concentrations of key mid-trophic level forage fish species, such as Arctic cod, polar cod and capelin were all <0.2 μg g − 1 dw, which is generally similar to observations from other Arctic systems including Arctic cod near Resolute and Clyde River, Nunavut (0.03 μg g − 1 ww; Pedro et al., 2019), Arctic cod in the Beaufort Sea (0.16 μg g − 1 dw; Loseto et al., 2008), capelin in Cumberland Sound (0.02 μg g − 1 ww; McMeans et al., 2015), western Hudson Bay (0.01 μg g − 1 ww; Pedro et al., 2019) and in west Greenland (0.07 μg g − 1 dw; Rigét et al., 2007) further highlighting the ubiquitous trophic role these species play across Arctic food webs.Though many higher trophic level Arctic predators are switching diets and now consuming more capelin, a pelagic-associated species, and less Arctic cod, an ice-associated species, than in previous decades (Gaston et al., 2003;Watt et al., 2016;Yurkowski et al., 2017;McKinney et al., 2022), similarities in THg concentrations between both fish species observed in this study and in Pedro et al. (2017) suggest a limited impact on Hg uptake by Arctic predators.The mean THg concentration of Greenland cod, a species that principally consumes capelin and larger invertebrates, was considerably higher than observed in conspecifics from west Greenland (0.03 μg g − 1 ww; Hansson et al., 2020) suggesting relatively higher consumption of fishes than invertebrates in the Southampton Island area.The THg concentration of anadromous Arctic char, which also consume a wide variety of marine forages fishes, marine invertebrates and freshwater insects, was low and indicated no risk to Hg toxicitya similar result to their conspecifics collected at numerous locations across the Arctic (see Barst et al., 2022).
The mean THg concentrations of narwhal and beluga muscle (3.10 and 3.05 μg g − 1 dw) were over four times higher compared to Atlantic walrus and ringed seals (0.67 and 0.70 μg g − 1 dw), where species   There was no relationship between mean logarithm 10transformed THg with mean δ 13 C.
differences in maximum life expectancy may be playing a role since THg concentration accumulates with age and because pinnipeds are able to offload Hg through hair growth and moulting (Dietz et al., 2006).Based on previously reported body lengthage relationships of narwhal, beluga (Kelley et al., 2015) and ringed seals (Kovacs et al., 2021), most individuals harvested from each species were adults.Narwhals have a maximum life expectancy of approximately 100 years (Garde et al., 2015), while age estimates for belugas have exceeded 70 years (Luque and Ferguson, 2010), which doubles the life expectancy estimated for ringed seals and Atlantic walrus (approximately 40 years of age; Lydersen, 2018;Hammill, 2009).The δ 15 N of narwhal and beluga were similar and suggests trophic omnivory by consuming a mix forage fishes (e.g.Arctic cod and capelin) and invertebrates, such as squid (Louis et al., 2021;Watt et al., 2013).Ringed seals had the highest δ 15 N among species in this study and suggests a high level of piscivory in the Southampton Island area.Though ringed seals had the highest δ 15 N compared to other species yet a relatively low THg concentration, it is important to note that the ringed seals were sampled near the end of their spring moult fasting period.This can lead to increased δ 15 N values that reflect catabolism of endogenous protein stores (Young and Ferguson, 2013).
Mean THg concentrations in marine mammal liver were ten to a hundred times higher than in muscle in this study likely due to the liver being an organ that detoxifies methylmercury through demethylation in mammals which results in continual accumulation of inorganic Hg (Braune et al., 2015).Overall, there is no or low risk for adverse health effects or sublethal Hg toxicity in muscle on all four marine mammal species based on risk categories and thresholds described in Dietz et al. (2022).

Biomagnification of THg
The y-intercepts (baseline THg) and slopes (trophic magnification slope) between the best (δ 15 N) and second-best (δ 15 N + δ 34 S) supported models were not significantly different suggesting no difference in baseline THg concentrations or trophic magnification slopes between the benthic and pelagic-associated species in this food web.The trophic magnification slope in this study (0.227) was similar to several other marine areas across the Arctic (average slope of 0.21 across eight systems, Lavoie et al., 2013), including the Beaufort Sea and Amundsen Gulf areas (0.232 to 0.255; Loseto et al., 2008), and Cumberland Sound (0.231;McMeans et al., 2015).The slope we report here, however, was higher than that in the marine food web of West Greenland (0.183; Rigét et al., 2007), Lancaster Sound (0.200;Atwell et al., 1998), the North Water Polynya (0.197;Campbell et al., 2005) and a polynya system near Nasaruvaalik Island, Nunavut (0.095; Clayden et al., 2015).Although some of these trophic magnification slopes were derived separately by different habitat types (e.g.benthic, pelagic and epibenthic) in those studies, pelagic-benthic habitats are highly interconnected in that benthic-associated species also acquire resources from the pelagic or sympagic carbon pathways in the form of detritus (e.g., filter-feeding clams, Sun et al., 2009;Amiraux et al., 2021).As well, the mean δ 15 N values of the pelagic-feeding Calanus hyperboreus and the benthic grazing sea urchin (Strongylocentrotus droebachiensis) were similar (9.5 ‰ versus 8.7 ‰, respectively; Amiraux et al., 2023b), suggesting that the δ 15 N baselines were generally similar between habitat types.Benthic top predators can also exert top-down trophic control on benthic-associated prey that more pelagic-or sympagic-associated consumers also use (Amiraux et al., 2023a).In addition, demersal fishes, such as several sculpin species, feed on a variety of resources from both pelagic and benthic habitats in the Arctic (Landry et al., 2018) and therefore would show some variability in δ 34 S value across species, which aligns with our results where Arctic sculpin had the highest δ 34 S value (19.1 ‰) and Arctic staghorn sculpin having the lowest δ 34 S value (13.6 ‰).Furthermore, most mobile species, especially those at a higher trophic level, are generalist consumers that exhibit habitat coupling by foraging on numerous prey items across benthic, pelagic or sympagic habitats in response to varying resource availability (Rooney et al., 2006;McMeans et al., 2013;Young and Ferguson, 2013).
Eco-physiological species traits can also influence trophic and THg dynamics within and across benthic, pelagic and benthopelagic habitats in aquatic systems (Lavoie et al., 2013;Vainio et al., 2022).Despite homeotherms (e.g., marine mammals) having a higher energy demand than poikilotherms (e.g., fishes) due to higher food consumption rates and thus a higher potential intake of Hg, their influence on estimated biomagnification factors and trophic magnification slopes across habitat type is varied from no effect on the trophic magnification slope (Lavoie et al., 2013) to a high effect (Fisk et al., 2001;Borgå et al., 2012).However, Vainio et al. (2022) found that the addition of homeothermic birds, who can consume resources across habitat types, to their trophic magnification models that included ectotherms did not show different trophic magnification factors between benthic and pelagic habitats, which was in contrast to their models excluding birds.Since Arctic ecosystems consist of numerous species with different metabolic strategies that also consume resources from multiple carbon sources or habitats, it is likely that both foraging and thermoregulatory strategies of species within and across habitat types influences trophic contaminant dynamics in an Arctic ecosystem.

Relationships between δ 13 C, δ 34 S and THg of fishes and marine mammals
When incorporating several habitat source variables, such as δ 13 C and δ 34 S, in models investigating THg variation across a food web, we found that δ 34 S had a higher range and standard deviation than δ 13 C across species (5.5 versus 4.5, and 1.1 versus 1.5, respectively) and was a better predictor of THg variation than δ 13 C, as has been observed in other aquatic studies (Evans and Crumley, 2005;Elliott and Elliott, 2016;Willacker et al., 2017;Elliott et al., 2021).In many coastal areas across the Arctic, the δ 13 C values of pelagic consumers can be higher than benthic consumers (Szpak and Buckley, 2020) or can largely overlap and the average difference in δ 13 C values can be small (e.g., within 4 ‰; Hobson et al., 2002;McMeans et al., 2013;Søreide et al., 2013;Linnebjerg et al., 2016;Amiraux et al., 2023b), likely due to the tight sympagic-pelagic-benthic coupling that drives energy flow between these habitats and is key to Arctic ecosystem functioning (Wassmann et al., 2004).This small difference in δ 13 C between benthic and pelagic consumers supports the use of a multi-isotopic approach that includes δ 34 S as opposed to solely using δ 13 C and δ 15 N when investigating Hg trophic dynamics in a food web.
Total mercury concentrations across species increased with δ 34 S.This supports the sulfate availability hypothesis which states that the supply of sulfate is mediated by sulfate-reducing bacteria in the 34 Senriched water column.Therefore, species with higher THg concentrations are likely foraging on prey in habitats richer in sulfates within the water column and on prey that are outside of the benthic microbialbased food web and which are higher in 34 S-depleted sulfides (Peterson and Fry, 1987).Similar to this study, the origin of sulfur also drove THg variation across several seabird species and their prey in the Pacific Ocean (Elliott and Elliott, 2016), as well as in small forage fishes inhabiting different estuarine wetland systems (Willacker et al., 2017).In contrast, Góngora et al. (2018) found a negative correlation between δ 34 S and THg among thick-billed murre prey species around Southampton Island, suggesting higher levels of sulfides in prey items.The difference between studies could be due to some of the different prey fish species collected and the inclusion of invertebrates in Góngora et al. (2018) that may be a part of the benthic microbial food web, whereas this study included more top predator fishes and marine mammals.
A few caveats could slightly affect our data interpretation.First, variation in protein content (e.g., cysteine) and protein quality of a predators' diet can influence their δ 13 C, δ 15 N and δ 34 S discrimination which could vary slightly among species in this study (McCutchan Jr et al., 2003;Caut et al., 2009).However, the δ 13 C and δ 15 N discrimination factors of marine fishes (Δ 13 C: 1.7 ‰; Δ 15 N: 3.7 ‰; Canseco et al., 2021), is comparable to that of piscivorous marine mammals (Δ 13 C: 1.3 ‰; Δ 15 N: 2.4 ‰; Hobson et al., 1996), and slight differences between the two species groups in Δ 13 C (0.4 ‰) and Δ 15 N (1.3 ‰) is much lower than the ranges of δ 13 C (4.5 ‰) and δ 15 N (5.5 ‰) observed across species in this study.In addition, Δ 34 S can significantly vary among consumers of a higher protein diet versus a lower protein diet by approximately 2.5 ‰ (Matthews and Ferguson, 2015), however this value is much lower than the range in δ 34 S among species in this study (5.5 ‰).Second, though fish and marine mammal samples were collected from different years, we could not examine inter-annual variability in trophic interactions in relation to environmental metrics due to low sample sizes for most species.However, the influence of interannual variability on consumer stable isotope values is likely negligible since standard deviations of δ 13 C, δ 15 N and δ 34 S for each species were generally low (<1.0 ‰).Sea surface temperatures can influence baseline δ 13 C and δ 15 N values and although surface temperature can vary by up to 8 • C over the upper 30 m around Southampton Island (Castro de la Guardia et al., 2023), water temperatures at the depth of the chlorophyll max (averaged 27 m; Kitching, 2022) and below where many of these species were collected did not vary more than a few degrees Celsius (Castro de la Guardia et al., 2023).Third, terrestrial or freshwater inputs around Southampton Island could slightly influence relationships between THg and δ 34 S across species since terrestrial and freshwater resources are typically depleted in 34 S (McCutchan Jr et al., 2003).Southampton Island is found at the confluence of water masses from the Pacific Ocean, Atlantic Ocean, and the Hudson Bay Complex, which receives large amounts of riverine input during the open-water period, but this primarily occurs in southwestern Hudson Bay and James Bay (Stewart and Barber, 2010).Summer circulation of seawater around Southampton Island originates from Foxe Basin to the north and Hudson Strait to the east where both areas are comprised of cold, saline water (~32 PSU) suggesting minimal freshwater input in this area of Hudson Bay (Stewart and Barber, 2010).Nonetheless, some of the species studied are mobile or migratory and therefore their THg and δ 34 S values could be influenced from other regions in which they feed during other seasons.

Conclusions
This study provided new insights into the mercury and trophic dynamics for part of the coastal marine food web around Southampton Island: (1) THg concentrations were highest in beluga and narwhal among marine mammals and in shorthorn sculpin, fourline snakeblenny, Arctic shanny, Greenland cod and Arctic sculpin among fishes, (2) THg concentrations for each species were generally similar to their conspecifics from other areas across the Arctic, (3) the trophic magnification slope was comparable to several other marine areas across the Arctic, and (4) δ 15 N and δ 34 S were better predictors of THg variation across the food web than δ 13 C, highlighting the importance of a multiisotopic approach that includes δ 34 S when investigating the trophic dynamics of mercury.Our study provides a marine ecosystem assessment of THg concentrations and its relationship to δ 13 C, δ 15 N and δ 34 S for 19 species inhabiting marine waters around Southampton Island that can be used to monitor future changes.With climate change continuing to have pronounced effects on sea ice dynamics, primary productivity, and species interactions in the Arctic (Post et al., 2013), and likely on mercury in the Arctic (Chételat et al., 2022;McKinney et al., 2022), long-term monitoring of mercury and food web dynamics across coastal marine systems is imperative to inform national and international conservation and management activities.
CRediT authorship contribution statement D.J.Y. and M.A.M. conceived and designed the study.D.J.Y conducted the data analysis and drafted the manuscript with input from M.A.M. E.M. conducted the THg analysis under the supervision of N.B. and M.A.M. A.T.F conducted the stable isotope analysis.All co-authors contributed samples and input at all stages, and approved the final version.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Study area around Southampton Island highlighting where fish were collected from scientific trawls (red circles) and the Inuit communities of Naujaat and Coral Harbour where marine mammals were collected as part of community-based monitoring programs (black circles).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2 .
Fig. 2. Boxplots of total mercury (THg) in muscle samples from 15 fish species and four marine mammal species collected around Southampton Island, Nunavut, 2016-2019.The box length represents the upper and lower interquartiles, the bars represent the ranges and the bold horizontal lines are the median values.Grey points represent each individual from that sampled species.Vertical lines separate the four species groupings revealed by Tukey's posthoc tests.The THg concentrations of banded gunnel were not significantly different from any species in the second or third clusters (p > 0.05).

Fig. 3 .
Fig. 3. Relationships of mean logarithm 10 -transformed total mercury (THg) with (A) mean δ 15 N, (B) mean δ 34 S, and (C) mean δ 13 C of muscle for fish and marine mammal species collected around Southampton Island, Nunavut.Error bars represent standard deviation.Only 1 sample of Daubed shanny was collected.Linear regressions (black lines) with 95 % confidence intervals (grey) were performed on mean values of each species.There was no relationship between mean logarithm 10transformed THg with mean δ 13 C.

Table 2
Candidate models of total mercury (THg) variation of muscle across fishes and marine mammals in the Southampton Island food web ranked by AIC c .

Table 3
Results from the top four candidate models explaining THg variation of muscle across fish and marine mammal species around Southampton Island.SE: standard error.