Insights into Mercury Source Identification and Bioaccumulation Using Stable Isotope Approaches in the Hannibal Pool of the Ohio River, USA

Abstract

Mercury contamination in river systems due to historic and current Hg releases is a persistent concern for both wildlife and human health. In larger rivers, like the Ohio River, USA, it is difficult to directly link Hg discharges to bioaccumulation due to the existence of multiple industrial Hg sources as well as the varied dietary and migratory habits of biota. To better understand how industrial effluent influences the cycling and bioaccumulation of Hg within the Ohio River, Hg stable isotope analysis was applied to various nonbiological and biological media. High Hg concentrations in suspended particulate matter suggest this vector was the largest contributor of Hg to the water column, and distinct Hg source signatures were observed in effluent particulates from different industrial processes, such as chlor-alkali activity (δ²⁰²Hg = −0.52‰) and coal power plant discharge (δ²⁰²Hg = −1.39‰). Despite this distinction, average sediments (δ²⁰²Hg = −1.00 ± 0.23‰) showed intermediate isotopic signatures that suggest the accumulation of a mixed Hg source driven by multiple industrial discharges. Biota in the system were shown to have a conserved range of δ²⁰²Hg and estimation approaches related these signatures back to particulate matter within Hannibal Pool. Mussels were found to conserve Hg isotopes signatures independently of food web drivers and served as ideal water column indicators of bioaccumulated Hg sources. This study highlights the complexity of Hg cycling within an industrialized river and shows that an isotope tracer approach can provide insight to water column sources of Hg.

INTRODUCTION

Industrial sources of Hg to the aquatic ecosystem have been a persistent concern for wildlife due the ability of the organometallic form, methylmercury (MeHg), to bioaccumulate (Bloom 1992). The ubiquity of Hg in the environment coupled to the higher volume of releases in industrialized regions has resulted in fish consumption advisories across the United States, due to deleterious human health effects and wildlife risk (Evers et al. 2007; Kim et al. 2016). Mercury concentrations in biota have been linked to filtered total Hg and MeHg in rivers (Chasar et al. 2009) and have also been shown to respond to loads of inorganic Hg from atmospheric deposition (Orihel et al. 2007) as well as industrial point sources (Francesconi et al. 1997). Although the installation of Hg pollution controls has led to a decrease in dissolved and particulate levels of Hg in the water column, no universal standard is set across the United States to control Hg releases (USEPA 2018). In addition, nonpoint source runoff from urban regions and contaminated sites is not as well constrained as active discharges (Eckley and Branfireun 2008). Although fish Hg levels are an important concern for human and wildlife health, the existence of multiple Hg sources within an area often complicate remedial and management activities targeted to decrease concentrations.

Questions pertaining to source identification have increasingly applied Hg stable isotope techniques, which have shown success at tracking Hg sources in sediments and soils (Lepak et al. 2015; Meng et al. 2019). These tracers have been commonly applied at sites with severe Hg contamination, such as mining activity (e.g., Gehrke, Blum, Marvin-DiPasquale 2011), elemental Hg processing (e.g., Reinfelder and Janssen 2019), and paper mills (Yin et al. 2016). Mercury isotopes undergo a wide range of both mass dependent fractionation (MDF) and mass independent fractionation (MIF) (Bergquist and Blum 2007). Every Hg transformation is recorded by MDF (denoted by δ²⁰²Hg), which has been leveraged to define sources of Hg in the environment and the isotopic shifts incurred due to different processes (e.g., atmospheric transport or metallic catalyst usage) (Yin et al. 2014). Other MIF reactions such as the magnetic isotope effect can record photochemical reactions that impact odd-isotopes (¹⁹⁹Hg and ²⁰¹Hg) in the water column (Bergquist and Blum 2007) or fractionation within the even-isotopes, thus far associated with atmospheric transformations (Sun et al. 2016). The resolving power of this tool allows for the examination of Hg sources and pathways in the environment, which is not plausible using concentration analysis alone.

Although many source tracking studies using Hg isotopes have focused on sediments (Meng et al. 2019), soils (Estrade et al. 2010), and suspended particulate matter (Washburn et al. 2018), more interest has emerged in studying the bioaccumulated Hg within the food web. Early applications of Hg isotope tracers did not utilize biological tissue, mainly due to the uncertainty of signature preservation due to competing processes such as methylation (Janssen et al. 2016) and internal partitioning (Kwon et al. 2013). These issues as well as ecological factors such as niche and dietary habits can often lead to differences in fish tissue and sediment sources, commonly observed as an offset between the 2 isotopic compositions (Gehrke, Blum, Slotton et al. 2011). However, more recent surveys have shown that fish tissue and other biological matrices can preserve source signatures (Kwon et al. 2014; Cransveld et al. 2017). Recent studies have even demonstrated that isotope signatures in fish tissue are highly dependent on land use factors related to source (Janssen et al. 2019). Larger field studies utilizing fish tissue have also noted that differences in isotopic signature can be used to establish large-scale sources (Senn et al. 2010; Madenjian et al. 2019), bioaccumulation of industrial sources (Perrot et al. 2010; Cransveld et al. 2017), and long-term source shifts (Lepak et al. 2019). The measurement of Hg isotopes in biological tissue as well as typical inorganic matrices can aid in the determination of Hg sources in complex environments.

In the present study, we applied Hg stable isotope measurements to an industrialized region on the Ohio River in the United States, called Hannibal Pool (Figure 1). Fish concentrations in the Ohio River have been consistently high in prey fish when compared to similar species captured in other large North American rivers, such as the Mississippi and the Missouri (Walters et al. 2010). Throughout the river, large variations of Hg concentrations in piscivorous fish have been observed, with some samplings showing greater than 41% of collections exceeding US Environmental Protection Agency (USEPA) consumption limits (0.3 μg/g) (Emery and Spaeth 2011), whereas others show much lower estimates (Reash et al. 2015). The Hannibal Pool region is distinct due to the presence of historic legacy contamination from an inoperative chlor-alkali site as well as active coal combustion and chlor-alkali chemical production, yet the relative influence of these industries to the Hg burden in the food web is unknown. We hypothesized that Hg stable isotope measurements in Hannibal Pool would distinguish the source signatures of Hg derived from chlor-alkali, coal plant discharge, and other discharge sources of Hg. Additionally, we expected to see the preservation of one or more of these industrial sources within the food web, including mussels and fish, based on the proximity to industrial effluents. Here we present an expanded study of source tracking, not only to determine if Hg discharges influence the food web of Hannibal Pool but also to present limitations within the data set that can be used to better design biological Hg isotope studies in river systems.

Figure 1.

Sampling map of Hannibal Pool on the Ohio River,West Virginia, USA. Pie charts show the relative proportion of Hg present as particulate bound (PHgT) or filter passing Hg (FHgT) in the water column at each site or discharge location. The total unfiltered concentration of Hg in water is shown in parentheses under the site name. Site abbreviations are AX = Axiall; CAP = Captina Island; FC = Fish Creek Island; HAN= Hanlin Allied Olin; HBD = Hannibal Dam; MDS = Moundsville Sewage Treatment Plant; MIT = Mitchell Power Plant; WHL =Wheeling Island. Yellow sites denote active Hg dischargers. The base map image is the intellectual property of Ersi and is used herein under license. Copyright © Esri and its licensors. All rights reserved.

METHODS

Site description

Hannibal Pool is in the United States on the Ohio River between Wheeling, West Virginia, and the Hannibal Lock and Dam (HBD) (Figure 1, Supplemental Data Table S1). The site hosts an operational chlor-alkali plant (Axiall [AX]) that still uses Hg° catalysts as well as a coal power fire plant (Mitchell [MIT]), 1 of 49 along the main branch of the Ohio River (Reash et al. 2015). Located within the region is also the historic Hanlin-Allied-Olin chlor-alkali plant (HAN), a national priority list site with notable legacy Hg contamination in the soil and groundwater, which still has site runoff and effluents of treated groundwater entering Hannibal Pool (USEPA 2020). Additionally, the Moundsville sewage treatment plant (Moundsville [MDS]) is located upstream of these other industries. Two Ohio River Island National Wildlife Refuges, Fish Creek Island (FC) and Captina Island (CAP), are also colocated near the outflows of MIT and HAN, respectively, and are potentially influenced by industrial discharges into this section of the Ohio River. Aquatic sites across a 61-km stretch of the river were sampled, including an upstream site (Wheeling [WHL]) removed from the Hannibal Pool industrial discharges (Figure 1) but still affected by urban runoff and sewage treatment effluents.

Sample collection

Water and sediment samples were collected during a 2016 field effort (May–July) using clean techniques. Sediment cores were collected from 6 main sites, and the top 0 to 4 cm were extruded and placed into clean plastic vials. Sediments were not collected at CAP or FC because these regions were comprised of mostly flooded bank soils and not submerged bed sediments. Sediments were stored on ice for no longer than 6 h and then frozen at −20 °C until analysis.

Water samples for MeHg and total Hg (HgT) concentrations were collected at all sites and stored in the dark until filtration back at the lab. Water samples at MDS, HAN, MIT, and AX were collected as close to the direct discharge pipes as possible. At the HBD and WHL sites, samples were collected from the open water of the Ohio River. Water samples (250–500 mL) were filtered and preserved with 1% hydrochloric acid (HCl). Suspended particulate matter (SPM) was collected onto ashed and preweighed quartz fiber filters (~0.7 μm, Whatman QMA) at all sites for concentrations and isotopic analysis. Filters were stored frozen until analysis.

Collection of fish and mussel samples followed American Veterinary Medical Association guidelines for fish and aquatic invertebrates. Biological samples (log perch, bluegill, zebra mussels, and pink heelsplitter mussels) were sampled as individuals at the sites with suitable habitat (WHL, CAP, FC, and AX; Supplemental Data Table S1). Collections took place directly offshore of 2 wildlife refuge sites (CAP and FC), from the west side of Wheeling Island at the upstream WHL site, and downstream of the outflow from the Axiall chlor-alkali facility at the AX site. Fish were collected via boat electroshocking in shallow areas. For bluegill, 6 fish of the median size were composited and frozen for analysis. Each log perch sample consisted of 1 to 2 fish of comparable size. Divers collected 15 adult pink heelsplitter mussels from the river bottom and 30 to 40 adult zebra mussels from underwater structures at each location. Mussels were depurated for 24 h to remove sediment from tissues. Five heelspitter mussels and at least 10 zebra mussels were composited for each sample.

Mercury concentration analyses

Methylmercury analysis was performed on filtered waters, SPM, and sediments via distillation, aqueous phase ethylation, and gas chromatography (GC) separation coupled to inductively coupled plasma mass spectrometry; more detailed information can be found in the Supplemental Data. Total Hg for waters and particulate Hg were determined after direct oxidation with bromine monochloride (BrCl) using USEPA Method 1631 (USEPA 2002). Total Hg analysis on waters and filters were performed by tin reduction coupled to gold amalgamation and cold vapor atomic fluorescence spectroscopy (CVAFS). Sediments were processed for HgT via combustion and cold vapor atomic absorption using USEPA Method 7473 (USEPA 1998).

Biological samples were analyzed for MeHg using a 4.5M nitric (HNO₃) acid digest (Hintelmann and Nguyen 2005) followed by ethylation coupled to GC-CVAFS analysis. Following MeHg analysis, biological HNO₃ digests were further oxidized in an ultraviolet box for 1 wk, then brominated with BrCl (10% v/v) prior to HgT analysis by Au amalgamation and CVAFS.

All samples for HgT and MeHg passed quality control and assurance protocols. Quality control standards (secondary Hg standard) were run within each analysis and were between 90% to 110% recovery. For solids, analytical sample triplicates were weighed out every 10 samples, the relative percent difference (RPD) between triplicates was <15%. Certified reference materials (CRMs) were also processed within each analytical batch for solids and were chosen to match the sample matrix. All CRMs were within 10% of the certified values for Hg (International Atomic Agency [IAEA] 407 (IAEA 2003), 220 ng/g; IAEA 452 (IAEA 2013), 160 ng/g; National Research Council of Canada [NRCC] MESS-1 (Lepak et al. 2015), 90 ng/g; National Institute of Standards [NIST] 1944 (NIST 2017), 3400 ng/g. Field triplicates were taken for all samples, and RPD was <15% for biota and waters. Field blanks for waters were reported as <0.04 ng/L.

Mercury isotopes analysis

Sediments and tissues for HgT isotope analysis were prepared via hot acid digestions (80 °C) for 8 h. Sediments and soils were treated with concentrated aqua regia (3HCl: HNO₃) while biological matrices were digested in concentrated HNO₃ followed by a 10% (v/v) BrCl addition in the last 2 h of heating. All digests and extracts were diluted to 50% acid content until isotope analysis. Mercury from filters was extracted using a 25% BrCl solution and heated at 60 °C for 1 wk. Particulate matter at the WHL site had HgT concentrations that were prohibitively low for isotopic analysis, hence the Hg isotope values were not reported.

Isotope analysis was performed using a Thermo Scientific Neptune Plus multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS). Mercury solutions and tin chloride (3% SnCl₂ in 10% HCl) were mixed in line and introduced to a custom gas liquid separator (GLS) with a counter argon gas flow to release Hg° from solution for introduction into the plasma. Thallium (40 ng/mL) was also introduced to the GLS to provide mass bias correction during measurement. The MC-ICP-MS parameters (gas flow, stage position, lenses) were optimized for Hg voltage and stability (~1 V ²⁰²Hg per 1 ng/mL Hg). Samples were analyzed using standard sample bracketing with NIST 3133 such that sample solutions were matrix (<10% H⁺) and Hg concentration matched with the primary standard (Blum and Bergquist 2007). Data collected from isotope analyses are reported in delta notation according to convention, where δ represents MDF and is calculated as follows:

$${\delta }^{\mathrm{xxx}}\mathrm{Hg}\left(‰\right)=\left[\left({}^{\mathrm{xxx}}\mathrm{Hg}{/}^{198}{\mathrm{Hg}}_{\mathrm{sample}}\right)/\left({}^{\mathrm{xxx}}\mathrm{Hg}{/}^{198}{\mathrm{Hg}}_{\mathrm{NIST}3133}\right)-1\right]\times 1000,$$ (1)

where xxx represents the Hg isotope of interest and NIST 3133 was the normalizing standard. Mass independent fractionation was calculated according to

$${\Delta }^{\mathrm{xxx}}\mathrm{Hg}={\delta }^{\mathrm{xxx}}\mathrm{Hg}-\left({\delta }^{202}\mathrm{Hg}\times \beta \right),$$ (2)

where β represents a scaling factor between isotope masses.

Secondary standard (UM Almaden-NIST RM 8610) was also analyzed every 5 samples to ensure data accuracy (δ²⁰²Hg = −0.52 ± 0.10‰; Δ¹⁹⁹Hg = −0.02 ± 0.07‰; Δ²⁰⁰Hg = 0.00 ± 0.05‰; Δ²⁰¹Hg = −0.03 ± 0.08‰, 2 SD, n = 63) and agreed with accepted literature values (Blum and Bergquist 2007). Additional CRM materials IAEA 407 (fish), MESS-2 (sediment), NIST 1944 (sediment), and IAEA 452 (mussel tissue) were analyzed to determine reproducibility (Supplemental Data Table S2). Photochemical corrections were also performed on biota using the approach outlined in Gehrke, Blum, Slotton et al. (2011) to examine Hg signatures prior to photochemical demethylation using the equation below:

$${\delta }^{202}{\mathrm{Hg}}_{\mathrm{COR}}={\delta }^{202}{\mathrm{Hg}}_{\mathrm{fish}}-\left({\Delta }^{199}{\mathrm{Hg}}_{\mathrm{fish}}/4.79\right),$$ (3)

where δ²⁰²Hg_COR is δ²⁰²Hg corrected for the photodemethylation effect, and 4.79 is the slope (Δ¹⁹⁹Hg/δ²⁰²Hg) associated with photodemethylation at a DOC concentration of 10 mg/L (Bergquist and Blum 2007). Supplementary calculations to estimate the inorganic Hg(II) pool in biological tissue (Tsui et al. 2012) were also performed and outlined in the Supplemental Data.

Carbon and N isotopes

Carbon and N isotopes were analyzed for aquatic biota at the USEPA National Risk Management Research Lab in Cincinnati, Ohio. Homogenized samples were weighed into Al boats and analyzed via elemental analyzer (Carlo Erba, NC2500) coupled to an isotope ratio mass spectrometer (Thermo Finnigan Delta) for C and N isotope measurements. Based on convention, δ¹³C values refer to differences in C sources to the food web, whereas δ¹⁵N values are an indicator of relative trophic position.

All data associated with this study can be found within a companion USGS data release (Janssen and Rosera 2020).

RESULTS AND DISCUSSION

Water concentrations and Hg sources to Hannibal Pool

Concentrations of HgT in filtered waters ranged from 0.31 and 2.28 ng/L with the highest values observed at HAN (2.28 ± 1.01 ng/L, n = 3) and discharges from AX (1.01 ± 0.21 ng/L, n = 4) (Supplemental Data Figure S1). Concentration of filtered MeHg in the system were largely under 0.05 ng/L, accounting for anywhere from 2% (MIT) to 13% (HBD) of the HgT (Supplemental Data Table S3). Concentrations of HgT were significantly higher in suspended particulate (Figure 1), suggesting most of the total Hg entering and cycling in Hannibal Pool was particulate bound. At upstream sites, WHL and MDS, particulate HgT was below 1 ng/L but increased further downstream in the river, reaching a maximum concentration of 86.5 ng/L at the AX site (Supplemental Data Figure S1). Overall particulate MeHg concentrations did not greatly exceed the concentrations observed in the dissolved phase (Supplemental Data Table S3), indicating no additional MeHg production within the particulate matter. These results showed that SPM was the main vector for inorganic Hg entering Hannibal Pool and that downstream concentration increases coincided with industrial activity.

Given the prominence of industrial effluents in this system, samples were taken directly at the outflows of 3 major industrial sites (HAN, AX, and MIT) to elucidate potential isotopic source differences in comparison to water column measurements downstream (HBD) and water treatment effluent (MDS). Particulate matter from HAN had a δ²⁰²Hg = −0.58‰ whereas AX was −0.46‰ (Supplemental Data Table S4); both values fell within previous ranges reported for sediments and SPM of sites with historic elemental Hg usage (Perrot et al. 2010; Washburn et al. 2018). This aligned with the history of these sites in Hannibal Pool, which were associated with chlor-alkali production. The SPM at the MIT averaged − 1.39‰ (n = 2) for δ²⁰²Hg and was more negative in comparison to samples from AX and HAN (Supplemental Data Table S4), indicating that the power plant had a distinctive signature in comparison to the chloralkali sites. Additional samples from HBD and MDS were −0.86‰ and −0.78‰, respectively, indicative of a possible mixed industrial signature at the downstream site and a potential background signal from sewage processing at MDS (Supplemental Data Table S4). No significant photochemical (Δ¹⁹⁹Hg) signals were measured in SPM samples (Supplemental Data Table S4).

Source designation in sediments

Mercury concentrations in sediments in Hannibal Pool were mostly within the range of 100 to 200 ng/g except for sediments sampled from MDS, which were consistently lower (47.4 ± 3.8 ng/g) (Figure 2). The highest concentrations were measured at the AX outfall (1172 ng/g), but high variability was observed (average HgT = 548.2 ± 570.6 ng/g) (Figure 2). Concentrations of MeHg, the bioaccumulative form, in sediments ranged between 0.17 to 2.68 ng/g, accounting for only a small portion of the overall HgT content (<1%) at most sites (Supplemental Data Table S4).

Figure 2.

Sediment concentrations (orange) and isotopic composition of δ²⁰²Hg (light green) in Hannibal Pool on the Ohio River, West Virginia, USA. Error bars indicate environmental variability (1 SD) between triplicate field samples. Sediments from the MIT site show highly negative δ²⁰²Hg signatures in comparison to other sediments collected from the region. AX = Axiall; HAN = Hanlin Allied Olin; HBD = Hannibal Dam; HgT = total Hg; MDS = Moundsville Sewage Treatment Plant; MIT = Mitchell Power Plant; WHL = Wheeling Island.

The isotopic compositions of sediments ranged from −0.85‰ to −1.80‰ for δ²⁰²Hg (Figure 2). Sediments (n = 3) collected near the outflow of MIT were significantly (unpaired t test, p < 0.05) more negative (−1.62 ± 0.21) in comparison to all other sediments (Figure 2) and most closely resembled the SPM collected from the MIT site. This signature was similar to previous measurements of coal ash spills (Bartov et al. 2012), as well as unprocessed coal (Sun et al. 2014), and could indicate the settling of heavier coal-derived particulates within the vicinity of the MIT plant. Highly negative δ²⁰²Hg signatures have also previously been associated with terrestrial sites (Demers et al. 2013), which are dependent on the atmospheric transport of Hg emissions coupled to deposition and foliar uptake. Although the δ²⁰²Hg signatures at MIT were similar to reported values for Hg deposition, atmospheric processing also commonly imparts a negative tracer for Δ¹⁹⁹Hg (Demers et al. 2013). However, most sediments from Hannibal Pool were effectively absent of Δ¹⁹⁹Hg except for the MDS site (Δ¹⁹⁹Hg = −0.14 ± 0.07‰; Supplemental Data Table S4), which treats large amounts of watershed runoff. The absence of the negative Δ¹⁹⁹Hg tracer indicates that the negative δ²⁰²Hg signature in MIT sediments was tied to the effluent from the plant rather than direct atmospheric deposition at the site.

Sediments were also collected near other industrial sites (HAN and AX) and regions not colocated near effluent discharges (HBD and WHL) to determine Hg sources. The most downstream site within Hannibal Pool, HBD, displayed a δ²⁰²Hg composition of −1.04 ± 0.04‰ (n = 3), which suggests a potential mixing of the 2 closest upstream dischargers, AX and MIT (Figure 1). When a simple mass balance calculation (SI Methods) was applied using the δ²⁰²Hg compositions of AX and MIT SPM, approximately 60% of the signature at HBD was attributed to MIT particulate matter and 40% to AX. Sediments from the AX site were more negative than the SPM isotope measurements of the colocated effluent as well as the previous measured range of industrial Hg use (Perrot et al. 2010; Washburn et al. 2018). The lack of evidence for a chlor-alkali signature in sediments closest to the AX site indicates that the SPM settling near AX was also derived from mixed sources within Hannibal Pool, not just the effluent from the adjacent plant. The calculation of source contribution estimates that the upstream MIT discharge contributed up to 67% of the sediment signature near the AX site, similar to HBD results. Additionally, the lack of Δ¹⁹⁹Hg in these sediments suggests little to no influence of watershed runoff to these downstream sites.

Defining source contributions becomes more difficult in upstream sediments such as WHL and HAN. The WHL site had a δ²⁰²Hg signature of −1.27 ± 0.14‰ (n = 3) and no Δ¹⁹⁹Hg signature (0.04 ± 0.03‰). This indicated that another highly negative δ²⁰²Hg source, potentially similar to the signature defined at MIT, was influencing this upstream site. An important aspect to consider in this region is the high volume of coal fire power plants, coal barge traffic, and industry located on the Ohio River upstream of Wheeling, West Virginia, which could have contributed Hg to the WHL site (Reash et al. 2015). This site was further complicated by sewage effluents which were not characterized as part of the present study. The inability to define a potential upstream source also interfered with source attribution at the HAN site which, similarly to AX, had a more negative isotopic composition in sediment than the SPM from the adjacent plant. The ability to distinguish source isotopic compositions in the SPM, but not sediment profiles near the discharge sites indicated a high degree of source mixing within the sediments of Hannibal Pool. Although downstream sites (AX, MIT, HBD) seem to be heavily influenced by a potential coal-derived SPM source from MIT, upstream sites may also be influenced by unidentified industrial Hg sources similar to MIT. These results show that in Hannibal Pool, as well as other similar systems, it is vital to identify all discharges and potential sources of Hg in order to properly construct an isotope source model. The designation of the WHL site as a reference was an unsuitable choice for Hannibal Pool, due unknown Hg sources within and upstream of Wheeling, West Virginia. This highlights the necessity of using an appropriate regional reference site to establish isotopic baselines for source tracking studies as well as the need to define industrial isotope endmembers as close to the original source as possible (e.g., SPM) to avoid complications of source mixing within sediments.

Mercury bioaccumulation and sources to biota in Hannibal Pool

Hannibal Pool contains 2 protected wildlife refuges, FC and CAP, which host a variety of wildlife that can be sensitive to Hg bioaccumulation (Figure 1). Two species of mussels (zebra and heelsplitters) and 2 fish species (bluegill and log perch) were examined within Hannibal Pool. Bluegill and log perch were grouped into 1 prey fish category given there was no significant difference (p > 0.05) between the species within a site (Supplemental Data Table S5). Total Hg concentrations in biological tissues were comparable across CAP, FC, and AX sites for each species (Supplemental Data Figure S2). The only site-specific exception for HgT concentrations were log perch and bluegill collected from WHL, which were on average 30% higher than fish sampled from Hannibal Pool (Supplemental Data Figure S2). Fish in this location were theorized to gravitate toward warmer waters, nutrient plumes, and Hg discharge at the Wheeling sewage treatment plant, possibly explaining the elevated levels. Despite similarities in HgT concentrations, large variations were observed in the percent MeHg. Heelsplitters had higher HgT (90.5 ± 17.8 ng/g), but lower MeHg (23.4 ± 7.1 ng/g) in comparison to the zebra mussels (HgT = 51.1 ± 18.9 ng/g and MeHg = 30.5 ± 11.2 ng/g) (Supplemental Data Table S5). Heelsplitter mussels were the only biota from Hannibal Pool that had consistently less than 50% MeHg in their tissues. The highest MeHg was observed in prey fish representing >75% of the total Hg present in tissue, though it is noted that these species have more diverse diets and habitats than both mussel species, likely leading to concentration differences (French and Jude 2001). Variation in MeHg content between the 3 targeted groups in Hannibal Pool suggests bioaccumulation factors are highly dependent upon the organism, and measurements of MeHg in 1 organism may not be a valid indicator for all biota within the system.

Isotopic measurements performed on biota showed a range of signatures that were closely tied to photochemistry. The range of δ²⁰²Hg for all biota spanned from −0.29‰ to − 1.10‰ and was paired to an equally large Δ¹⁹⁹Hg range from −0.02‰ to 0.89‰ (Figure 3). When compared to laboratory benchmarks (Bergquist and Blum 2007), photochemical slopes (Δ¹⁹⁹Hg/Δ²⁰¹Hg) of aquatic biota indicated that MeHg undergoes photochemical demethylation in the water column prior to bioaccumulation in Hannibal Pool, driving the variation within Δ¹⁹⁹Hg (Supplemental Data Figure S3). The imprint of photodemethylation was more prominently preserved in zebra mussels and fish as denoted by higher Δ¹⁹⁹Hg values (Figure 3). The highest extent of photochemistry within the system was observed within zebra mussels from FC, which were collected within a shallower water column, allowing for more photodemethylation of Hg. The opposite is true for heelsplitters, which display lower extents of photochemistry (0.18 ± 0.13‰, n = 12), reaching near 0 values at sites like AX and CAP which have deeper water levels (Figure 3). These results suggest that habitat drives MeHg transformations prior to bioaccumulation, resulting in isotopic differences between the species. It was noted that the process of photochemical demethylation, which was observed in all biota in the present study, can also shift δ²⁰²Hg signatures (Bergquist and Blum 2007), making it difficult to compare source signatures in organisms that have different Δ¹⁹⁹Hg values. In order to compare biota sources to one another as well as to other media, such as SPM, a commonly used photochemical correction was applied to estimate the Hg signature prior to any photochemical processing, termed “δ²⁰²Hg_cor” (Gehrke, Blum, Slotton et al. 2011). Values of δ²⁰²Hg_cor were well constrained across fish and zebra mussels with a median value of −0.65 ± 0.13‰, suggesting that 1) organisms share a similar source of Hg and 2) the previous differences in the uncorrected δ²⁰²Hg values were driven by photochemistry, not by source differences. While the photochemical estimation may not be suitable for identifying sources of Hg to the food web, it can be applied to correct photochemical differences between organisms within a system.

Figure 3.

Mercury isotopic biplot for aquatic biota from Hannibal Pool on the Ohio River, West Virginia, USA. Aquatic biota results suggest a mixed source of Hg, likely representing contributions from power plant, chlor-alkali, and upstream discharges. The source of inorganic Hg to mussels (black asterisk) was estimated using linear regression models set forth in Tsui et al. 2012 (SI Methods). AX = Axiall; CAP = Captina Island; FC = Fish Creek Island; HAN = Hanlin Allied Olin; MDS = Moundsville Sewage Treatment Plant; MIT = Mitchell Power Plant; SPM= suspended particulate matter; WHL = Wheeling Island.

While δ²⁰²Hg_cor values in biota resemble mixed SPM in the system, this was not a completely fair comparison because biological tissues varied in MeHg content and were not directly comparable to SPM (Supplemental Data Tables S4 and S5). In order to confirm the validity of the photochemical correction approach, an independent estimation of the biological inorganic Hg pool was also performed (Figure 3). The inorganic pool of Hg within biological tissue was calculated for mussels using a linear regression model of MeHg concentrations and δ²⁰²Hg set forth in Tsui et al. (2012) (SI Methods). Although the correlation between δ²⁰²Hg and the fraction of MeHg in mussel tissue was weak (r² = 0.26), it produced an estimated value for the Hg(II) pool in mussels of −0.83‰ (Figure 3, black asterisk), which was slightly more negative than the photochemical correction estimate. This result may support the idea that SPM signatures were preserved in Hg(II) of biota, though more thorough investigation is needed. However, further extending this approach to calculate the MeHg pool would produce large propagations of error due to the statistically weak regression, suggesting that this estimation approach may not be completely suitable for these biological matrices. Although this calculation is useful for examining an overarching connection between biota and inorganic media, there is still a need to address whether 1 source is more readily methylated and bioavailable than others in Hannibal Pool. Future investigations may require the examination of different matrices, such as benthic insects, to improve the correlation results of MeHg estimates (Tsui et al. 2012) or the actual measurements of MeHg specific isotopes to directly connect biota to MeHg being produced in the system (Rosera et al. 2019).

Food web controls on biological Hg loads and sources

Within Hannibal Pool, it was vital to examine food web variation as a possible driving factor behind Hg sources. The biota examined in Hannibal Pool represented diverse feeding mechanisms, and these ecological factors could have influenced the preservation of Hg isotopes. The δ¹³C and δ¹⁵N isotopes, common tracers of C source and general trophic level, were used to examine the food web in Hannibal Pool and confirm that similarities in Hg isotope signatures were related to a singular Hg source rather than fractionation during ecological processes. Within industrialized rivers, specifically with wastewater treatment plants (e.g., MDS and the Wheeling, West Virginia plant), it is difficult to establish a baseline due to the quality of effluent discharges from sewage treatment plants and the processes of nitrification and denitrification (Loomer et al. 2015). Due to these complexities, δ¹³C and δ¹⁵N tracers were not used to establish trophic levels, but rather to examine correlations with Hg isotopes and concentrations.

Correlation results, using a Spearman rank approach, showed that Hg isotopes in mussels responded to Hg concentrations, but not as strongly to food web factors (Table 1). Zebra mussels and heelsplitter mussels showed significant correlations (p < 0.01) to concentrations of MeHg and HgT. Zebra mussels, specifically, displayed a relation to δ²⁰²Hg (rho = −0.79) and the estimated value δ²⁰²Hg_cor (rho = −0.85), as well as a weaker correlation to Δ¹⁹⁹Hg (rho = 0.66, p < 0.10) (Table 1, Supplemental Data Figure S4). Trends were also observed with the total Hg concentrations in zebra mussels. These correlations to Δ¹⁹⁹Hg (photochemistry) and δ²⁰²Hg_cor (aligned with SPM) suggest the Hg bioaccumulated within these organisms is directly tied to water column processing and sources, respectively. Heelsplitters also displayed a strong relation between MeHg and Δ¹⁹⁹Hg (rho= 0.87, p < 0.01) and a weaker relationship with δ²⁰²Hg (rho = 0.63, p < 0.05) (Table 1). No significant correlation was observed between δ²⁰²Hg_cor and Hg concentrations for heelsplitters, which is likely due to limitations associated with the photochemical estimation when the organism is not predominantly MeHg. The strong correlation results between mussel Hg concentrations and Hg isotope signatures suggests that changes in the Hg source profiles within the system may have a direct influence on Hg concentration within these species.

Table 1.

Spearman rank correlation results between Hg isotope tracers (δ²⁰²Hg, δ²⁰²Hg_cor, Δ¹⁹⁹Hg), Hg and MeHg concentrations, and food web tracers (δ¹³C and δ¹⁵N) within Hannibal Pool on the Ohio River, West Virginia, USA

Analyte	Correlation parameter	Fish (n = 20)			Zebra mussels (n = 9)			Heelsplitter mussels (n = 12)
		δ202Hg, ‰	Δ199Hg, ‰	δ202Hgcor, ‰	δ202Hg, ‰	Δ199Hg, ‰	δ202Hgcor, ‰	δ202Hg, ‰	Δ199Hg, ‰	δ202Hgcor, ‰
HgT, ng/g	Value	0.33	0.07	0.34	−0.80b	0.53	−0.82b	−0.09	0.08	−0.15
	p	0.15	0.76	0.14	0.01b	0.14	0.01b	0.78	0.80	0.63
MeHg ng/g	Value	0.31	0.04	0.32	−0.79b	0.66a	−0.85b	0.63a	0.87b	0.46
	p	0.18	0.88	0.17	0.01b	0.05a	0.00b	0.03a	0.00b	0.13
δ13C, ‰	Value	0.10	−0.29	0.18	0.40	0.12	0.32	0.14	0.17	0.07
	p	0.66	0.21	0.45	0.29	0.77	0.41	0.66	0.60	0.83
δ15N, ‰	Value	0.60b	0.57b	0.60b	0.67a	−0.18	0.62a	−0.12	0.14	−0.17
	p	0.00b	0.01b	0.00b	0.05a	0.64	0.08a	0.71	0.66	0.60

MeHg = methylmercury; HgT = Total mercury.

^aSignificant relationships (p < 0.10).

^bStronger correlations (p < 0.01).

Interestingly, zebra and heelsplitter mussels showed different directionality in their correlations to MeHg concentrations and Hg isotope measurements, which may be related to the life cycle and bioaccumulation differences between the 2 species. Heelsplitter mussels are a larger species that burrow into the sediments and obtain food from detritus (Levine et al. 2013), whereas zebra mussels are found higher in the water column and have been proposed to accumulate dissolved trace metals (Roditi et al. 2000). These behavioral differences likely correspond to variation in the amount of MeHg within the tissue. Given the correlation results and variations in percent MeHg (Supplemental Data Figure S4) accumulated between the organisms, bioaccumulation differences may exist between benthic and pelagic food webs within Hannibal Pool and may be identifiable with Hg isotopes in future studies. Despite the relationships observed with concentration, mussel tissues were mostly absent of trends with δ¹⁵N and δ¹³C, except for a slight correlation (p < 0.10) with δ¹⁵N, δ²⁰²Hg, and δ²⁰²Hg_cor in zebra mussels. These observations in mussels indicate that there were no strong trophic or food web controls on the preservation of Hg isotope signatures in these organisms, making them ideal for localized studies of Hg sources.

Fish did not exhibit the same trends as mussels in regard to food web and Hg concentration relationships. Fish were noticeably absent of any statistically significant trends between MeHg, HgT, δ²⁰²Hg, Δ¹⁹⁹Hg, and δ²⁰²Hg_cor (Table 1). We propose the lack of relationship between concentrations and the Hg isotope signatures relate to physiological and ecological factors. Although fish with the smallest potential home ranges were chosen for the present study, their range is still a much larger feeding habitat than that of the mussels. Log perch and other similar forage fish have been noted to inhabit nearshore zones and consume mostly macroinvertebrates (French and Jude 2001), indicating that the variety and location of foraging could introduce multiple isotope signatures to the fish that may equilibrate within the tissue at different rates or be impacted by internal cycling (Kwon et al. 2013). Further examination showed that δ²⁰²Hg, Δ¹⁹⁹Hg, and δ²⁰²Hg_cor were all positively correlated (p < 0.01) to δ¹⁵N values (Table 1). This has been observed before in fish tissue, and recent studies have attributed the trends in δ¹⁵N and Hg isotopes to dietary shifts (Lepak et al. 2019) or simply an increase of MeHg that has undergone photochemical demethylation with each trophic step (Xu et al. 2016). Given the connection between Hg isotope signals and dietary habits, prey fish may not be the best biological indicators to pinpoint sources on a finer spatial scale or identify changes in concentrations related to Hg source input in Hannibal Pool, but they do likely reflect a regional scale signature that can be compared to other stretches of the Ohio River.

CONCLUSION

The present study showed that despite the complexities of Hg cycling in regions with multiple Hg sources, isotopic and concentration studies can be utilized in tandem to understand Hg cycling. The major vector of Hg within Hannibal Pool was found to be particulate matter, which preserved the isotopic signatures of different industrial effluents. The SPM from different industries were proposed to partially contribute to the isotopic source profile observed in sediments and biota. It was also noted that Hg isotopic endmembers were best defined using water column measurements, strongly suggesting this matrix should be targeted for future investigations of sites with multiple Hg sources. Analysis of biological tissue showed that all biota accumulated a similar Hg source within Hannibal Pool, but that photodemethylation caused variations between isotopic measurements due to habitat usage and feeding preferences. Mercury concentrations in mussel tissue were found to be closely tied to Hg isotope measurements and independent of food web drivers (denoted by δ¹³C and δ¹⁵N), whereas fish displayed high dependency on relative trophic level. The present study indicated that the selection of a biological and inorganic matrices impacts the resolution of isotope source appointment, and future studies on the scale of Hannibal Pool should utilize measurements with finer spatial and temporal resolution such as waters and biota with high site fidelity.

Supplementary Material

Supplementary Information

Figure S1. Filtered and particulate bound Hg concentrations in Hannibal Pool waters.

Figure S2. Mercury concentrations for zebra mussels, heelsplitter mussels, and fish.

Figure S3. Photochemical plot for Δ¹⁹⁹Hg and Δ²⁰¹Hg in aquatic biota.

Figure S4. Relationship between Δ¹⁹⁹Hg and MeHg in aquatic biota.

Table S1. Sampling site locations and matrices sampled

Table S2. Quality control and assurance data for Hg isotope measurements

Table S3. Mercury concentrations in filtered water and suspended particulate matter (SPM) of field replicates

Table S4. Mercury concentrations and isotope results for sediment and bulk SPM

Table S5. Mercury concentrations and isotope results (Hg/C/N) for aquatic biota.