Factors affecting MeHg bioaccumulation in stream biota: the role of dissolved organic carbon and diet

Abstract

The bioaccumulation of the neurotoxin methylmercury (MeHg) in freshwater ecosystems is thought to be mediated by both water chemistry (e.g., dissolved organic carbon [DOC] and dissolved mercury [Hg]) and diet (e.g., trophic position and diet composition). Hg in small streams is of particular interest given their role as a link between terrestrial and aquatic processes. Terrestrial processes determine the quantity and quality of streamwater DOC, which in turn influence the quantity and bioavailability of dissolved MeHg. To better understand the effects of water chemistry and diet on Hg bioaccumulation in stream biota, we measured DOC and dissolved Hg in stream water and mercury concentration in three benthic invertebrate taxa and three fish species across up to 12 tributary streams in a forested watershed in New Hampshire, USA. As expected, dissolved total mercury (THg) and MeHg concentrations increased linearly with DOC. However, mercury concentrations in fish and invertebrates varied non-linearly, with maximum bioaccumulation at intermediate DOC concentrations, which suggests that MeHg bioavailability may be reduced at high levels of DOC. Further, MeHg and THg concentrations in invertebrates and fish, respectively, increased with δ¹⁵N (suggesting trophic position) but were not associated with δ¹³C. These results show that even though MeHg in water is strongly determined by DOC concentrations, mercury bioaccumulation in stream food webs is the result of both MeHg availability in stream water and trophic position.

1. Introduction

Mercury (Hg) is a global pollutant that bioaccumulates in biota even in remote lakes and streams in its biologically available form, methylmercury (MeHg) (Fitzgerald et al. 1998). MeHg is a potent neurotoxin to both humans (Clarkson et al. 2003; Mergler et al. 2007) and wildlife (Caron et al. 2008; Evers et al. 2008; Scheulhammer et al. 2007). Fish consumption is the main source of human exposure to MeHg (Hightower and Moore 2003; Mahaffey et al. 2009; Sunderland 2007) and all 50 states in the US have fish consumption advisories related to Hg for freshwater lakes, rivers, and streams (USEPA 2013). Although atmospheric sources provide major Hg inputs to watersheds, there is considerable variability in fish MeHg concentrations among systems within a region (Chen et al. 2012b; Kenney et al. 2014; Simonin et al. 2008; Ward et al. 2010). As such, a better understanding of the factors that influence Hg bioavailability to food webs will enhance our ability to identify and potentially manage biological Hg hotspots within the landscape (Burns and Riva-Murray 2018; Chen et al. 2012a; Evers et al. 2007).

Mercury bioaccumulation in aquatic food webs is influenced by complex biogeochemical interactions. For example, increased MeHg bioaccumulation in freshwater biota is positively associated with wetlands, including proximity and the percentage of wetland area upstream, because the anoxic conditions and abundance of organic carbon to fuel microbial activity make wetlands optimal environments for microbially-mediated methylation of Hg (Burns et al. 2013) (Chasar et al. 2009; Driscoll et al. 2007; Ward et al. 2010). Both inorganic Hg and MeHg bind with dissolved organic carbon (DOC) and DOC is strongly correlated with aqueous Hg concentration and transport in lotic systems (Chasar et al. 2009; French et al. 2014; Tsui et al. 2010; Vidon et al. 2014a; Vidon et al. 2014b). However, the effect of DOC on the bioavailability and bioaccumulation of MeHg is complex. Different studies have reported at least three associations between streamwater DOC and biotic MeHg: biotic MeHg increases with increased DOC (Burns and Riva-Murray 2018; Driscoll et al. 2007; Pickhardt and Fisher 2007), decreases with increased DOC (French et al. 2014; Gorski et al. 2008; Tsui and Finlay 2011), and varies nonlinearly with DOC, increasing from low to moderate concentrations but then declining at very high DOC concentrations (Chaves-Ulloa et al. 2016; Driscoll et al. 1994). Although the reasons for these complex DOC-Hg relationships are not entirely clear, DOC quality, especially its molecular structure and complexity, has been implicated (French et al. 2014; Vidon et al. 2014b). Further, the quantity and quality of streamwater DOC, which are mostly determined by processes that happen on land, in turn influence the quantity and bioavailability of dissolved MeHg (Brigham et al. 2009; St. Louis et al. 1994). These relationships between DOC and biotic Hg, where there are closely associations with the surrounding terrestrial environment (Brett et al. 2017; Ward et al. 2010) have not been extensively studied in small streams (Chaves-Ulloa et al. 2016; Jardine et al. 2012; Vidon et al. 2014b).

Given the inconsistent associations between DOC and MeHg bioaccumulation across different studies and systems and lack of such studies in small tributary streams, we sought to evaluate the relationship between streamwater DOC and MeHg bioaccumulation in low-order forested streams within a single watershed to control for regional variation such as deposition rates, weather patterns, and underlying geology (Chen et al. 2018; Evers et al. 2007; Julian 2012; Selin and Jacob 2008). Here we studied 12 first and second-order streams in the Lake Sunapee watershed in New Hampshire, which had previously been shown to have a five-fold gradient in DOC concentrations (2.9 to 14.8 mg/L; (Chaves-Ulloa et al. 2016; Roebuck 2009). Low-order streams are typically dominated by omnivorous primary and secondary consumers that consume both allochthonous (from the watershed) and autochthonous (produced within the stream) sources of carbon, with an overall greater reliance on terrestrially derived carbon sources relative to larger lotic systems (Ward et al. 2010). These habitats typically experience greater in situ MeHg production and less photodegradation of the MeHg and DOC present than occurs in open waters (Burns et al. 2013; Minor and Stephens 2008; Tsui et al. 2013). While there has been extensive research on the factors that predict concentrations of Hg in fish tissue across all types of aquatic systems, as a function of age and size (Clayden et al. 2014; Driscoll et al. 1995; Eagles-Smith et al. 2016; Kamman et al. 2005; Kidd et al. 2012), less is known about the factors determining invertebrate Hg concentrations in lotic systems (Chaves-Ulloa et al. 2016; Jardine et al. 2013; Riva-Murray et al. 2011).

MeHg bioaccumulation in aquatic systems is also related to trophic factors such as diet composition (e.g., trophic position; degree of reliance on autochthonous versus allochthonous diet) and food chain length and complexity (Campbell et al. 2005; Chen and Folt 2005; Chumchal and Hambright 2009; Chumchal et al. 2011; Jardine et al. 2013; Karimi et al. 2016; Karimi et al. 2007; Pickhardt et al. 2002; Ward et al. 2010). For example, organisms at higher trophic levels typically have higher MeHg concentrations than those at lower trophic levels due to biomagnification (Campbell et al. 2005; Chumchal and Hambright 2009; Chumchal et al. 2011; Jardine et al. 2013). Greater food quantity and higher food quality (e.g. higher energy and more nutritious) can result in biomass dilution and growth dilution, respectively, reducing MeHg bioaccumulation in consumers (Chen and Folt 2005; Karimi et al. 2016; Karimi et al. 2007; Pickhardt et al. 2002; Ward et al. 2010). On the other hand, autochthonous sources of food have been associated with higher MeHg bioaccumulation than allochthonous carbon sources in lake (Power et al. 2002), estuarine (Chen et al. 2014; Chen et al. 2009) and stream systems (Jardine et al. 2012; Riva-Murray et al. 2013; Tsui et al. 2009; Ward et al. 2010). This may be due to the lower quality and lower MeHg concentrations in allochthonous food sources and higher Hg methylation in autochthonous algal biofilms within the stream (Buckman et al. 2015; Klaus et al. 2016; Tsui et al. 2009).

Our research evaluates the effects of biogeochemistry and diet on the bioaccumulation of Hg and MeHg in primary and secondary consumers in a network of low-order tributary streams from a single watershed. We sampled water chemistry and invertebrates and small fish representative of the biotic community in each of the streams. Based on prior studies, we predicted that 1) dissolved MeHg and total mercury (THg) concentrations would be positively related to DOC concentration in streamwater, 2) the relationship between biotic Hg and streamwater DOC concentration would be either positive and linear or non-linear with a maximum at intermediate DOC concentrations, 3) biotic Hg would increase with trophic level, and 4) biotic Hg would vary with dietary carbon source. For the last prediction, we used δ¹³C signatures as an indicator of diet, as has been used in numerous other studies (e.g. (Chen et al. 2014; Jardine et al. 2012; Riva-Murray et al. 2013; Rounick et al. 1982; Tsui and Finlay 2011).

2. MATERIALS AND METHODS

2.1. Site Description

This work was conducted in 12 low-order tributary streams within the watershed of Lake Sunapee, NH, USA (43°23′3.40″N; 72° 3′4.39″W; Fig. S1, Table S1). The lake’s watershed contains no known point sources of Hg and has minimal urban development; yet, there is a range of land cover types (primarily hardwood, coniferous, and mixed forest but also developed land and wetlands) and variability in stream chemistry (but not pH) across subwatersheds (Table S1). Together, these features create an ideal study region in which to investigate differences in Hg bioaccumulation in lotic invertebrates and fish due to streamwater and food web characteristics across the five-fold range in stream DOC concentrations. During summer 2011, we sampled 12 tributary streams from among the 22 major tributary streams that drain into Lake Sunapee (Table S1). Water and selected biota, including one to three fish taxa and up to three benthic invertebrate taxa, were collected from similar habitat conditions (shallow riffle and pool section of the tributaries, above any road or dam) in each of the 12 study streams.

2.2. Water sampling

Water samples for analysis of DOC and dissolved Hg were collected from all 12 streams 27-29 June and 22-24 August 2011. In both sampling periods, samples for DOC and Hg were collected at least five days after a rain event to minimize effects of rainfall on our measurements of THg (Schuster et al. 2008) and DOC (Mullane et al. 2015; Romani et al. 2006). The water sampling represents two time points for each site, providing some information about temporal dynamics, but does not capture the full range of seasonal variation in each of these systems (Babiarz et al. 1998a; Lawson and Mason 2001; Zhang et al. 2012). We collected one DOC sample per stream per time period using a 60 mL syringe and field-filtering water through Whatman 934-AH glass microfilter (1.4 μm) into 125 mL Nalgene bottles that had been pre-soaked and rinsed in deionized water as described by Findlay et al. (2010). Samples were kept on ice then frozen until analysis at the Cary Institute of Ecosystem Studies in Millbrook, NY, on a Shimadzu® 5050 TC Analyzer using the spectrophotometric method also described in Findlay et al. (2010). The June water sample results have previously been published in Chaves-Ulloa et al. (2016), but the August water sample results and all of the other results presented here are new.

Water samples for THg and MeHg analysis were collected in certified clean PETG bottles (500 mL, Nalgene™, Fisher Scientific); sample bottles were rinsed three times with the site water before filling just below the water surface at the widest section of the stream in a run. Duplicate samples were collected, double-bagged in acid-rinsed plastic bags, and stored on ice in the dark until filtering within 6 hours of collection through pre-combusted (4 hrs at 550°C) quartz fiber filters (Whatman Grade QMA with particle retention of 0.3 μm). Between samples from different streams, the filter set up and filters were rinsed with double deionized (DDI) water and 1% Optima HC1 acid (Fisher Chemical). Procedural blanks were taken by filtering 125 mL DDI water before each duplicate set of samples. The filtrate was preserved with 0.4% Optima grade HC1 and stored in certified clean (Thermo Scientific) 250 mL amber glass bottles in cold (4°C), dark conditions before analysis (Jackson et al. 2009b; Taylor et al. 2011).

2.3. Biotic sampling

Larval insects were collected using D-nets and kick nets in riffles, pools and runs in June and August 2011. Organisms were kept alive in a 9 °C cool room and sorted within 24 hours of collection. Insects were identified to family (Merritt et al. 2008; Wiggins 2000) and all sorting was done using trace metal clean techniques to prevent contamination (USEPA 1996). Sorting trays and forceps were acid-cleaned sequentially in a dilute Citronox soap solution, 1 N nitric acid, and 1.4 N hydrochloric acid.

Predatory Aeshnidae dragonfly nymphs, herbivorous and detritivorous Tipulidae cranefly larvae, and four families of mayfly nymphs (Heptageniidae, Ephemereliidae, Baetidae, and Leptophlebiidae) were selected for Hg speciation (inorganic and MeHg) analysis because of their importance to the food web and because they had the most representation across streams. For each site, regardless of collection period, individuals of the same family and size class (small, medium, or large based on body mass, see definition of size classes in Table S2) were combined and homogenized together to ensure enough biomass (>0.05g dry weight) for accurate Hg speciation analysis. For the dragonflies and mayflies, THg and MeHg concentration values from all size classes found in a particular stream were used in our analysis (Table S3a); however, only the small size class of cranefly larvae (0.001 - 0.049 g DW) was included in our data analyses given the distinctly different feeding modes represented by different size classes. For dragonfly and mayflies, not all size classes were represented in all streams (Table S3a). Selected specimens were placed in trace metal clean vessels (25 mL Polystyrene vials, Sarstedt) and stored in the freezer (−20°C) until freeze-drying using a Labconco FreeZone 4.5L System. The desiccated samples were homogenized using acid-cleaned utensils.

In July 2011, an electrofisher (Smith-Root LR-24) was used to collect fish from each stream, with nets placed at both the lower and upper ends of a 75 m reach. The fish community structure was not known in advance of the study. All fish encountered in one pass along the reach were collected and stored in stream water until identification in the field. Representative samples of each species were selected in the field, euthanized according to protocol (Dartmouth College IACUC Protocol #10-04-06), placed in an acid-cleaned plastic bag, and preserved on dry ice until freezing at −20°C back at the lab. In the laboratory, each individual was again weighed and measured for total length.

We selected samples for mercury and stable isotope analysis from the three fish species that were both found at the majority of sites and were most abundant at each stream site: brook trout (Salvelimis fontinalis), creek chub (Semotilus atromaculatiis), and yellow perch (Perea flavescens). For these three species, samples were divided into three size classes (Table S2). Where possible, three individuals of the same size class from each species were analyzed from each site (Table S3b). From each fish, a small fillet sample (<2% of the whole body weight) was set aside and freeze dried for stable isotope analysis. While other studies have found that there can be a measurable difference in stable isotope signatures between whole body and muscle tissue, other studies show that any differences are biologically insignificant (Curtis et al. 2017). The rest of the fish was homogenized and a subsample of the homogenate (0.10 to 0.12 g) was weighed and acid-digested for THg analysis at 180°C for 20 minutes in a CEM Xpress microwave system using 1.8 mL Optima™ HNO₃ (Fisher Chemical), 0.2 mL Optima™ HCl (Fisher Chemical), and 2.0 mL E-pure water (Barnstead).

2.4. Hg and Hg-speciation analysis

All THg and MeHg analyses were performed by the Dartmouth Trace Element Analysis Core Facility using isotope dilution gas chromatography-inductively coupled plasma mass spectrometry (GC-ICPMS) for water, invertebrate and fish samples. Filtrates from water samples were spiked with an appropriate amount of enriched inorganic ¹⁹⁹Hg (HgI) and enriched methyl²⁰¹Hg (MeHg) and analyzed for Hg speciation by extraction with 10 mL 4M HNO₃ and heating overnight at 60°C. Ionic Hg and MeHg were determined in the water sample filtrate following the ultra-low level methods described in Jackson et al. (2009). For determination of THg and MeHg in invertebrates, freeze-dried samples were spiked with enriched inorganic ¹⁹⁹Hg (HgI) and enriched Me²⁰¹Hg (MeHg) and then extracted in 2–3 ml of TMAOH (tetramethyl ammonium hydroxide, 25% w/v) (Jackson et al. 2009a). MeHg and inorganic Hg in the invertebrate samples were determined by species-specific isotope dilution purge-and-trap ICP-MS (Element 2 ICP-MS, Thermo Scientific). For the samples analyzed in this way, THg was calculated as the sum of MeHg and inorganic Hg (Point et al. 2007; Taylor et al. 2008). Digested homogenate from individual whole fish was analyzed for THg using the GC-ICPMS. We assumed that THg in whole fish was representative of MeHg concentration, given that numerous previous studies have shown that the majority (>90%) of Hg in fish is MeHg (Bloom 1994; Driscoll et al. 2007; Wiener et al. 2003). Analyzing MeHg in fish would have been prohibitively expensive in this study.

For each sample type, quality control was evaluated by analyzing a blank, duplicate, spiked sample and standard reference material (SRM) for every batch of 20 samples. Quality assurance and quality control (QA/QC) measures were all reviewed and were deemed adequate (low method detection limits, blanks below the detection limit, high percent metal recovery from spiked samples and standard samples). The method detection limit (MDL) for water filtrates averaged 0.0072 ng/L (SD ±0.0003) for MeHg and 0.07 ng/L (±0.02) for inorganic Hg. All blanks were below detection limits (BDL). Percent recovery for stream water samples spiked in the range of sample concentrations averaged 107.7% (±1.4) for MeHg and 88.0% (±2.7) for inorganic Hg. The MDL for invertebrates averaged 19 ng/g (±2) for MeHg and 20 ng/g (±6) for inorganic Hg. The MDL for fish was 28 ng/g for THg. As standard reference materials (SRMs), DORM 3 (NRC, Ottawa, Canada; 355 ng/g MeHg; 382 ng/g THg) was used for the invertebrate analyses and DORM 2 (DORM 2, NRC, Ottawa; 4470 ng/g MeHg; 4640 ng/g THg) was used for the fish analyses. The invertebrate analyses had an average recovery of 103% (±4%) for MeHg and 97% (±12%) for inorganic Hg. The fish analysis had a 72.5% recovery of THg injected into a control sample.

2.5. Stable isotope analysis

We measured δ¹⁵N and δ¹³C stable isotopes (Chen et al. 2014) for a 1 mg subsample of most (61 of 91 samples) of the homogenized fish and invertebrate samples. Due to lack of sufficient mass following metal analysis, we were only able to analyze one mayfly sample from one stream for stable isotopes, thus no mayfly isotope data are included here. Additionally, not all mayfly and cranefly samples contained sufficient mass for both MeHg and stable isotope analyses; when mass was limiting, we prioritized measuring MeHg. Table S6 indicates the samples for which both Hg and stable isotope analyses were possible. Duplicates were included once every 10 samples. All stable isotope analyses were run by the University of California Davis Stable Isotope Facility. External standards, duplicates, and blanks were run for QAQC.

Four external standards were used: G-9 Glutamic Acid; G-13 Bovine Liver, G-17 USGS-41 Glutamic Acid; and G-18 Nylon 5 (for invertebrates) and G-11 Nylon (for fish). For the invertebrate analysis, G-9 had an average recovery of 100.3% for C and 97.9% for N, G-13 of 99.9% for C and 99.1% for N, G-17 of 100.0% for C and 100.0% for N, and G-18 of 100.0% for C and 100.0% for N. For the fish analysis, G-9 had an average recovery of 100.4% for C and 98.1% for N, G-13 99.8% for C and 97.8 for N, G-17 100.0% for C and 100.0% for N, and G-11 of 100.0% for C and 100.0% for N. For invertebrates, duplicate samples had an average relative percent difference of 1.3% for C and 6.2% for N. For fish, the average relative percent difference of our duplicate samples was 0.71% for C and 3.0% for N. Food source samples were not collected for this study, so the assumptions were made that individuals of the same taxa had similar dietary habits across sites and differences in C-isotopes of the consumers reflected differences in food sources rather than in-stream variability in the C isotopes of autochthonous producers. Similar to previous papers (Chen et al. 2014; Ward et al. 2012), data for δ¹⁵N were adjusted for site-level differences (those due to N availability, not trophic structure) prior to further analysis by subtracting the corresponding site effects estimated from a two-way ANOVA with δ¹⁵N as the response variable and site and species as the treatment factors, then adding back the grand mean δ¹⁵N value for interpretation as in Gilbert-Diamond et al. (2011).

2.6. Statistical analyses

We conducted separate statistical analyses to test each of our hypotheses. First, we evaluated the associations between dissolved MeHg and THg concentrations in streamwater (response variables) and DOC (predictor variable). Prior to this analysis, the duplicate measurements for dissolved MeHg and THg concentrations were averaged to create one estimate per stream per timepoint. Then we used a general linear model (GLM) that included a linear effect of DOC and an indicator (or dummy) variable (set to 0 or 1) that was used to sequentially assess whether the slope and intercept of this relationship differed between June and August (Gram et al. 2004; Kutner et al. 2005). Similar models were used to evaluate the association between streamwater percent MeHg and DOC.

Before testing our specific hypotheses about associations between Hg concentrations within the tissues of invertebrates and fish (i.e., biotic Hg) and stream chemistry and diet, we evaluated differences across the six taxa (three invertebrates, three fish) for each of the three biological response variables – Hg concentrations in biota, trophic position (as measured by site-adjusted nitrogen stable isotopes [δ¹⁵N]), and reliance on autochthonous sources of carbon (as measured by carbon isotope signature [δ¹³C]) using the non-parametric Kruskal Wallis test to accommodate outliers. These analyses were conducted with the site-specific mean for each taxon and variable due to the unequal number of samples for each taxon in different streams (0 to 4). Tissue Hg concentrations did not meet the assumptions of normality and were log₁₀-transformed prior to analysis. As described above and outlined in Table S2, fish samples were selected to span the representative size classes within each site. MeHg concentrations were used as the focal response variable for the invertebrate taxa, while THg was used as a proxy for MeHg in fish as has been done in previous studies, since most of the fish Hg would be MeHg.

With these comparisons in mind, we evaluated associations between log₁₀-transformed biotic Hg concentrations (response variables) and average streamwater DOC and dissolved MeHg (potential predictors). We used separate GLMs with both linear and quadratic terms for the predictors when the relationship appeared to be curvilinear (Table S4) and invertebrates and fish were analyzed separately. When the quadratic term was significantly different from zero, we estimated the DOC concentration at the point at which biotic Hg peaked by setting the derivative of the quadratic function to zero and solving for DOC (Dodson et al. 2000). Similar to the analysis for stream chemistry described above, we included dummy variables to indicate the taxon (cranefly, dragonfly, or mayfly for the invertebrate analysis and brook trout, creek chub, or yellow perch for the fish analysis) and then tested for interactions with the predictor(s) to evaluate whether the relationships between biotic Hg and streamwater chemistry differed among taxa (Kutner et al. 2005; Pickhardt et al. 2002).

Lastly, we examined whether trophic position (site-adjusted δ¹⁵N) or variability in dietary sources of carbon (δ¹³C) explained the variation in logio-transformed biotic Hg concentrations that was not explained by streamwater MeHg. The ideal approach would have been to simply add δ¹⁵N and δ¹³C as additional predictors in the GLMs, then evaluate if the new parameters were statistically significant and improved model fit. Unfortunately, however, we did not have stable isotope data for all biotic samples (31 of 91 samples had mercury concentration data but were missing isotope data, Table S6), and when we tried this approach, the estimated parameters for the association between biotic Hg and streamwater MeHg were highly sensitive to these missing data. As such, we instead used a two-step process: first, we calculated the residuals from the models previously fit for biotic Hg as a function of streamwater MeHg for the full biotic Hg dataset (i.e., Fig. 3b,d; Table S4) and added the grand mean logio-transformed biotic Hg concentration to the residuals of each model to keep the response variable in more interpretable units as in Gilbert-Diamond et al. (2011). Then, we used this stream water-adjusted biotic Hg concentration as a response variable in new GLMs with taxon (again as an indicator variable), either N or C isotope signature, and the interaction of taxon and isotopic signature as predictors (Chen et al. 2014; Gram et al. 2004) (Table S5). For the δ¹³C dataset, we excluded two visual outliers in order to meet the assumptions of the statistical analysis but did include them in the figures. Inclusion of these points has a strong influence on model fit.

Fig. 3.

Log₁₀-transformed tissue Hg concentrations in invertebrates (top, as MeHg) and fish (bottom, THg assumed to be nearly all MeHg) from tributaries of the Lake Sunapee, NH watershed as a function of dissolved organic carbon (DOC, left) and dissolved MeHg (right). Lines included on each panel are drawn based on general linear model that began with an indicator variable for the taxon name, linear and quadratic terms for the predictor variables, and interactions between the indicator and predictor variables. All models are significant at α=0.05 (P<0.0002), with R² as indicated on the corresponding panel. Parameter values for the individual panels are summarized in Table S4.

All statistical analyses were conducted using R 3.4 (Team 2013) including the packages readxl, tidyverse, stringr, stats, emmeans, and ggpubr (Kassambara 2018; Lenth et al. 2019; Wickham 2017). All analyses met the assumptions of homoscedasticity and approximately normally distributed residuals (after transformation where appropriate).

3. Results and Discussion

3.1. DOC and mercury in streams

DOC and dissolved Hg concentrations differed five-fold and 10-fold, respectively, across the 12 Lake Sunapee streams (Table S1). Dissolved organic carbon concentrations ranged from 2.91-14.8 mg/Lin June 2011 and 1.93-16.0 mg/L in August 2011. Dissolved THg concentrations ranged from 0.24-2.4 ng/L in June and 0.24-3.4 ng/L in August 2011, while dissolved MeHg concentrations ranged from 0.039-0.82 ng/L in June and 0.020-1.8 ng/L in August (Table S1). There was within-site seasonal variability in dissolved Hg concentrations; some streams had higher concentrations in June while others had higher concentrations in August with no evident pattern among streams (Fig. S2). In the Lake Sunapee watershed, dissolved THg and MeHg concentrations were both less variable across streams in June than in August (77.9% coefficient of variation for MeHg and 50.2% for THg in June, as compared to 92.8% for MeHg and 58.5% for THg in August). Methylation efficiency, MeHg/THg or percent MeHg, ranged from 7% to 42% in June and 5% to 53% across streams. The increase in temperature in August may impact sites differently depending on site by site conditions (e.g. shade, water depth, flow) resulting in more variable methylation rates (Babiarz et al. 1998b). Further, methylation efficiency has been shown to vary greatly (~1-80%) in Canadian shield streams (Roy et al. 2009) and streams in the Northeast US, but on average is low for the latter, with a median of 3.9% (Shanley et al. 2005). In this study methylation efficiency also increased significantly with DOC (Fig. S4), suggesting that MeHg is bound to and transported with DOC.

As hypothesized, dissolved MeHg and THg, as well as percent MeHg, increased linearly with DOC in both sampling periods (Fig. 1); these tight associations were not significantly different between the two sampling periods and are consistent with results from other studies in lakes (Braaten et al. 2014; Dittman et al. 2010; Driscoll et al. 1995; Watras et al. 1998), rivers (Chasar et al. 2009), and streams (Dittman et al. 2009; Riva-Murray et al. 2011; Schuster et al. 2008; Shanley et al. 2008). They are likely a result of upstream processing and lateral transport from riparian wetlands of carbon and mercury that results in both high DOC and DOC-bound mercury downstream (Brigham et al. 2009; Grigal 2002; Schuster et al. 2008; Tsui et al. 2009). In this study, the streams with the highest DOC concentrations were also the ones with the highest fraction of the Hg in the methylated form (percent MeHg, Fig. S4), consistent with a scenario in which DOC production and Hg methylation are co-occurring in the same habitats, typically wetlands (Fig. S3) (Bloom 1994; Brigham et al. 2009; St. Louis et al. 1994). This pattern was detected irrespective of the sampling date (Fig. S4). There was a hint of a unimodal relationship between the dissolved percent MeHg and DOC in the August sampling, but the quadratic term was not significant (Fig. 1).

Fig. 1.

Dissolved MeHg, THg, and percent MeHg increased with DOC from tributaries in the Lake Sunapee, NH watershed, consistent with predictions, in both of the sampled periods (a, c. June and b, d. August). There is an apparent quadratic relationship between %MeHg and DOC (dotted line in panel d), but the quadratic term was not statistically significant at α=0.05 (t₉ =−1.99, P=0.078).

To put these concentrations into context, dissolved THg, MeHg, and DOC concentrations in the Lake Sunapee tributary streams were within ranges of past work in other northeastern U.S. study systems (Bradley et al. 2011; Burns et al. 2012; Chasar et al. 2009; Schuster et al. 2008). However, observed concentrations of dissolved THg (maximum 3.2 ng L⁻¹) relative to DOC in this study were lower than those in two previous studies (Chasar et al. 2009; Dittman et al. 2009), while MeHg concentrations were twice as high across the equivalent DOC range. This difference indicates that MeHg increased more sharply per unit DOC in the Lake Sunapee system than in other systems regionally, possibly due to greater methylating efficiency in this watershed. This also suggests that if DOC continues to increase, as has been shown in North America and Europe (Brown et al. 2017; Clark et al. 2010; Monteith et al. 2007), the rate of MeHg concentration increase in these streams may be greater than in other stream ecosystems.

3.2. Differences in biotic response variables across taxa

In Lake Sunapee streams, average invertebrate MeHg concentrations across streams ranged from ~100 ng/g to nearly 1,200 ng/g dry weight (Fig. 2; Table S6). There were no significant differences among invertebrate taxa in MeHg concentration, but the dragonflies showed the most variation (not shown), had a higher median concentration than other taxa (Fig. 2a), and had a higher percent MeHg (Fig. S5) that was comparable to published findings for fish (Driscoll et al. 2007; Haro et al. 2013; Jeremiason et al. 2016; Point et al. 2007; Riva-Murray et al. 2011; Taylor et al. 2008).

Fig. 2.

Comparison of (a) biotic Hg concentrations, (b) trophic position (as estimated from the δ¹⁵N signature, after correcting for site as described in the text), and (c) diet allochthony (as estimated from the δ¹³C signature), among the six taxa sampled from tributaries of the Lake Sunapee, NH watershed. Each point indicates the mean value within a tributary for that taxon (N=l-4, depending on taxon and tributary). Symbol types and colors indicate taxon (and are the same as are used subsequently in Figures 3 and 4). Solid lines are the median for all samples of that species for each metric. The biotic Hg concentrations are MeHg for invertebrates and THg for fish (assumed to be nearly all MeHg). Stable isotope data are not available for mayflies due to limited sample mass.

A priori we expected higher MeHg concentrations in organisms at higher trophic levels, and thus, higher (estimated) MeHg in fish than in invertebrates (Chasar et al. 2009; Chumchal et al. 2011). However, in this study the Hg concentrations in invertebrates and fish in the Sunapee streams were not significantly different (Kruskal-Wallis test; χ²=4.6, df=5, P=0.47), in part due to high variability across streams (Fig. 2a). In fact, the highest observed concentrations were in craneflies and dragonflies, not fish, and Hg concentrations in the fish fell within the observed values for invertebrates in this (Fig. 2a) and previous studies (Chasar et al. 2009; Chumchal et al. 2011; Tremblay et al. 1998; Tsui et al. 2009). The small sizes (Tables S2, S3) (and likely ages) of the fish representative of these low-order streams may partially explain this observation, since fish of this size are likely to feed lower in the food chain (Browne and Rasmussen 2009; Kusnierz et al. 2014; Post and McQueen 1994). Although the site-corrected δ¹⁵N signatures suggest that the fish were feeding at a slightly higher trophic level than the sampled invertebrates (Fig. 2b; Kruskal-Wallis test; χ²=19.2, df=4, P=0.0007), the differences across taxa were less than 3%o (with medians that ranged from 7 to 9.5† across taxa), which likely does not denote a meaningful biological difference in trophic position (Minagawa and Wada 1984; Vander Zanden and Rasmussen 2001). The N isotope data therefore support the conclusion that the fish in this study were feeding at roughly the same trophic level as the invertebrates, and thus it is not surprising that the fish did not have higher biotic Hg concentrations than the invertebrates.

We also observed statistically significant differences in the δ¹³C data, which provide insight into diet composition (Fig. 2c; Kruskal-Wallis test; χ²=13.3, df=4, P=0.01). Dragonfly larvae had more negative δ¹³C signatures than the other invertebrate larvae and fish. Based on previous papers that have interpreted more negative δ¹³C signatures as being indicative of autochthonous diet items (Jardine et al. 2012; Riva-Murray et al. 2013; Rounick et al. 1982), this suggests that dragonfly larvae may have eaten more autochthonous diet items, which might explain their relatively high MeHg concentrations (Minagawa and Wada 1984; Peterson and Fry 1987; Peterson et al. 1985; Vander Zanden and Rasmussen 2001). Moreover, our samples comprised late instar dragonfly larvae. Since dragonflies are long-lived invertebrates that can live as larvae for several years (Pennak 1953; Thorp and Covich 2010), they can bioaccumulate MeHg over multiple years. Further, as they grow and age, they feed higher in the food web and are exposed to higher Hg concentrations in their food.

3.3. Bioaccumulation as a function of DOC and dissolved Hg

For both invertebrates and fish, we observed a quadratic relationship between the log₁₀-transformed concentration of MeHg or THg in biota and streamwater DOC (Fig. 3). These relationships were highly statistically significant, and typically explained about a third of the considerable variability in biotic Hg concentrations within and among streams (Fig. 3). The intercept of the relationships with stream chemistry changed among the three invertebrate taxa (higher for dragonflies than for mayflies and craneflies) but was the same for the three fish taxa (Fig. 3). Based on the fitted relationships, invertebrate MeHg increased with DOC to ~8.6 mg/L (Fig. 3a), while fish THg increased with DOC to ~11.3 mg/L before declining (Fig. 2c).

Thus, our findings support the hypothesis that bioaccumulation of Hg is non-linearly associated with DOC across sites, consistent with recent studies that have also detected a curvilinear relationship (Chaves-Ulloa et al. 2016; Chiasson-Gould et al. 2014; Driscoll et al. 2007; French et al. 2014). Moreover, fish THg exhibited a non-linear relationship with dissolved MeHg in streamwater. The nonlinear relationship of fish Hg with dissolved MeHg also indicates that the supply of MeHg does not entirely determine biotic Hg concentrations. By contrast, most previous work in lakes, rivers, and streams observed positive relationships between biotic Hg concentrations and both DOC and dissolved MeHg (Chasar et al. 2009; Driscoll et al. 2007; Driscoll et al. 1994; Watras et al. 1998). Further, other studies have related a Bioconcentration Factor (BCF) to DOC concentrations (e.g., Tsui and Finlay 2011) and found a negative relationship. However, this was not done here because dividing biotic concentrations by aqueous concentrations that are already positively correlated with DOC can result in a false negative relationship between Hg and DOC (similar to fitting Y/X vs. X).

The curvilinear relationship with DOC seen in both invertebrates and fish suggests that there may be different mechanisms affecting Hg bioavailability at low versus high concentrations of DOC. Increases in fish Hg have previously been associated with increasing lake water DOC, particularly at levels above 4.0 mg C/L (Driscoll et al. 2007; Mierle and Ingram 1991). In addition, some past studies have found a strong positive relationship between Hg in stream invertebrates (MeHg), forage and predatory fish (THg) and DOC or dissolved MeHg across a broad geographic range, but do not observe a decrease in tissue MeHg at high DOC levels (i.e. 27-37 mg/L) (Burns and Riva-Murray 2018; Chasar et al. 2009). However in other studies, decreased Hg bioaccumulation in both fish and amphipods has been observed at higher DOC concentrations, particularly above a threshold of 8 mg C/L (Driscoll et al. 1994; French et al. 2014), and our previously published study of the same stream tributaries showed a threshold as low as 5 mg/L DOC for MeHg concentrations in predatory spiders feeding on emergent insects in these stream food webs (Chaves-Ulloa et al. 2016). The presence of such a threshold could be due to several factors: reduced binding to fulvic and increased binding to humic acids (French et al. 2014), the effect of DOC on cell physiology (Chiasson-Gould et al. 2014), or a change in DOC quality with increasing concentrations where high DOC streams may have higher allochthonous carbon inputs that may reduce Hg bioavailability to methylating bacteria more than autochthonous carbon (Burns et al. 2013; Vidon et al. 2008; Wang et al. 2014). The trend between increased allochthonous carbon and reduced MeHg may be due to reduced methylation, quality, uptake efficiency, or assimilation (Jonsson et al. 2017; Klaus et al. 2016; Riva-Murray et al. 2013; Tsui et al. 2009).

While there was a quadratic relationship between log10-transformed fish THg and dissolved MeHg (Fig. 3d), the relationship between log₁₀-transformed invertebrate MeHg and dissolved MeHg was linear (Fig. 3b). Our results are similar to other studies which have documented differences in patterns of mercury bioaccumulation between primary consumers and secondary consumers in aquatic food webs (Chen et al. 2000; Chen et al. 2005). The linear relationship of invertebrates with dissolved MeHg may be due to their closer linkage with primary producers and bioconcentration of MeHg from water. The curvilinear relationship for fish suggests that the food availability and food quality for fish may differ across the range of DOC and MeHg in streams. Food quality and growth of fish may diminish fish bioaccumulation at higher MeHg and DOC levels. Or it may mean that there is a growth dilution of Hg in fish in higher DOC streams resulting in lower fish concentrations (Karimi et al. 2007; Ward et al. 2010). For example, Ward et al. (2010) found that even though Atlantic salmon Hg concentrations tracked their prey, fast-growing fish had lower Hg concentrations than slow-growing fish and growth rate accounted for 38% of the variability in Hg concentrations in salmon (Ward et al. 2010; Ward et al. 2012).

On the other hand, the positive linear relationship of invertebrate MeHg and dissolved MeHg that we detected contrasts with the quadratic relationship with DOC. This was somewhat unexpected since DOC and dissolved MeHg were positively related in our results as well as prior results (Dittman et al. 2009; Riva-Murray et al. 2011; Schuster et al. 2008; Shanley et al. 2008) as described above. The linear trend with invertebrates suggests that dissolved MeHg drives the concentration in aquatic invertebrates but is mediated by the quality and source of DOC (Dittman et al. 2010). At high DOC concentration sites, more of the MeHg may be bound to DOC making it unavailable for uptake by the food web (Chaves-Ulloa et al. 2016). Future research should focus on DOC quality and mercury dynamics.

3.4. Bioaccumulation as a function of diet

We tested whether mercury concentrations in biota increased with δ¹⁵N (suggesting trophic position) and changed with δ¹³C (indicating altered diet composition) within the major taxonomic groups of invertebrates and fish, after accounting for the relationships with dissolved streamwater MeHg (Fig. 4). Based on previous studies, we expected more MeHg in individuals that fed at higher trophic levels, as indexed by δ¹⁵N (Chasar et al. 2009; Jardine et al. 2013). For both the invertebrates and fish, differences in Hg concentrations in tissue were indeed related to trophic position, though the relationship was much stronger for invertebrates than for fish (R²=0.31 vs. R²=0.09). Even after accounting for the extent to which biotic Hg was associated with streamwater MeHg, individuals and taxonomic groups with higher δ¹⁵N signatures had significantly higher concentrations of MeHg (Fig. 4a,c) showing that the relationship holds even after accounting for biogeochemical factors.

Fig. 4.

Log₁₀-transformed tissue Hg concentrations in invertebrates (top, as MeHg) and fish (bottom, THg as an estimate MeHg) from tributaries of the Lake Sunapee, NH, watershed as a function of trophic position (adjusted δ¹⁵N, left) and diet composition (δ¹³C, right), after adjusting for the observed associations with streamwater MeHg shown in Figure 3 (see text for details). All data points are shown, but the δ¹³C outliers (marked with “X”s) were excluded from statistical analysis. Lines indicate predictions from the statistically significant components of a general linear model that began with an indicator variable for the taxon name, a linear term for the predictor variable, and an interaction between the indicator and predictor variable. Only the models for trophic position were significant at α=0.05 (P=0.007 for the invertebrates and P=0.03 for the fish), with R² as indicated. Parameter values are summarized in Table S5.

Further, because of the presumed importance of upstream and in-channel methylation of Hg in producing MeHg found within the streams, we expected that organisms deriving larger proportions of their diet from within the stream (autochthonous) would have higher MeHg concentrations than organisms that rely on allochthonous diet items subsidized from outside (Driscoll et al. 2007; Jardine et al. 2012; Ward et al. 2012). However, we did not find that dietary carbon source, based on δ¹³C as an indicator, predicted Hg concentration for either invertebrates (Fig. 4b) or fish (Fig. 4d), as has been found in previous studies in estuarine and lake food webs (Chen et al. 2014; Chen et al. 2009; Chen et al. 2012a; Power et al. 2002). In addition to being in different systems, our contrasting result may be because we examined the relationship within taxonomic groups (i.e. invertebrates or fish), whereas some studies look across invertebrate and vertebrate taxa (Chasar et al. 2009; Chen et al. 2014; Chen et al. 2009; Kidd et al. 2012), potentially spanning greater variability in dietary carbon sources. Further, we tested for this association only after adjusting for the relationship between biotic Hg concentrations and streamwater MeHg concentrations, because site conditions may impact the δ¹³C signature of the biota more than expected (Leggett et al. 1999; Vuorio et al. 2006). Alternatively, for invertebrates, our null result may be due to the pooling of organisms collected at different time points (June and August) and in different habitats (riffles, pools and runs); because biofilm and invertebrate δ¹³C values both vary temporally and spatially (between riffles and pools), we may not have been able to detect relationships. Given the differences in methodologies and systems, the relationship of Hg bioaccumulation to dietary carbon source in streams requires further investigation.

4. Conclusions

Overall, our results from low-order tributary stream support hypotheses derived from previous research conducted primarily in lakes, rivers, and larger streams that bioaccumulation of Hg varies in complex ways with streamwater DOC. This study shows that, as anticipated, watershed-derived DOC was strongly positively correlated with both THg and MeHg. However, unlike most previous studies, both fish and invertebrate Hg concentrations showed a curvilinear relationship with DOC, suggesting that bioaccumulation may be inhibited at high streamwater DOC levels relative to accumulation at intermediate DOC concentrations (Burns and Riva-Murray 2018; Chasar et al. 2009; Tsui and Finlay 2011). Although concentrations of Hg and trophic positions in fish and invertebrates (as indexed by δ¹⁵N) in this study were broadly similar, after accounting for streamwater MeHg concentrations observed trophic position was associated with biotic Hg concentrations. Both the non-linear relationships and complexities stemming from food web structure highlight the need for further study of the biogeochemical factors controlling MeHg fate in stream food webs.