Emerging investigator series: methylmercury speciation and dimethylmercury production in sulfidic solutions

Charlotte R. Kanzler a, Peng Lian bc, Emma Leverich Trainer a, Xiaoxuan Yang§ a, Niranjan Govind d, Jerry M. Parks b and Andrew M. Graham *a
aGrinnell College Department of Chemistry, Grinnell, Iowa 50112, USA. E-mail: grahaman@grinnell.edu; Tel: +1-641-269-9813
bBiosciences Division, Oak Ridge National Laboratory, 1 Bethel Valley Road, Oak Ridge, Tennessee 37831-6309, USA
cUniversity of Tennessee, Department of Biochemistry and Cellular and Molecular Biology, Knoxville, Tennessee 37996, USA
dEnvironmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99352, USA

Received 6th November 2017 , Accepted 12th January 2018

First published on 18th January 2018


Alkylated mercury species (monomethylmercury, MeHg, and dimethylmercury, DMeHg) exhibit significant bioaccumulation, and pose significant risks to ecosystems and human health. Although decades of research have been devoted to understanding MeHg formation and degradation, little is known about the DMeHg formation in aquatic systems. Here, we combine complementary experimental and computational approaches to examine MeHg speciation and DMeHg formation in sulfidic aqueous solutions, with an emphasis on the formation and decomposition of the binuclear bis(methylmercuric(II)) sulfide complex (CH3Hg)2S. Experimental data indicate that the reaction 2CH3Hg+ + HS ⇄ (CH3Hg)2S + H+ has a log[thin space (1/6-em)]K = 26.0 ± 0.2. Thus, the binuclear (CH3Hg)2S complex is likely to be the dominant MeHg species under high MeHg concentrations typically used in experimental investigations of MeHg degradation by sulfate-reducing bacteria (SRB). Our finding of a significant abiotic removal mechanism for MeHg in sulfidic solutions through the formation of relatively insoluble (CH3Hg)2S suggests careful reexamination of reported “oxidative demethylation” of MeHg by SRB and perhaps other obligate anaerobes. We provide evidence for slow decomposition of (CH3Hg)2S to DMeHg and HgS, with a first-order rate constant k = 1.5 ± 0.4 × 10−6 h−1. Quantum chemical calculations suggest that the reaction proceeds by a novel mechanism involving rearrangement of the (CH3Hg)2S complex facilitated by strong Hg–Hg interactions that activate a methyl group for intramolecular transfer. Predictions of DMeHg formation rates under a variety of field and laboratory conditions indicate that this pathway for DMeHg formation will be significant in laboratory experiments utilizing high MeHg concentrations, favoring (CH3Hg)2S formation. In natural systems with relatively high MeHg/[H2S]T ratios (the oxic/anoxic interface, for example), DMeHg production may be observed, and warrants further investigation.



Environmental significance

Experimental and computational chemistry approaches were used to elucidate the thermodynamics, kinetics, and mechanism of a previously overlooked reaction of monomethylmercury (MeHg) with aqueous sulfide to form dimethylmercury (DMeHg). Our results indicate that formation and decomposition of bis(methylmercury) sulfide ((CH3Hg)2S) is favored under conditions of relatively high [MeHg]/[H2S], such as may be found in laboratory investigations of MeHg biodegradation and in some natural environments (e.g., oxic/anoxic interfaces). Our thermodynamic and kinetic data can be applied to better understand MeHg cycling across a variety of spatial and temporal scales, including understanding important environmental problems such as the origin of DMeHg in the subsurface oceans.

1. Introduction

Mercury (Hg) is a global pollutant. Anthropogenic sources dominate releases to the environment, with coal combustion and artisanal gold mining being the most important sources for Hg released to the atmosphere.1 Upon deposition into aquatic ecosystems, diverse species of anaerobic bacteria and archaea, including sulfate- and iron-reducing bacteria, fermentative bacteria, and methanogens,2,3 may convert inorganic Hg to bioaccumulative and neurotoxic4 methylmercury (MeHg). MeHg concentrations in aquatic ecosystems reflect the net balance between conversion of inorganic Hg to MeHg and degradation of MeHg. There are four main pathways known to degrade MeHg in situ. In oxic waters, bacterial species carrying the mer operon, which encodes the enzymes MerA and MerB, and can convert MeHg to CH4 and elemental Hg(0).5 In sunlit surface waters, MeHg may also undergo photodegradation, likely driven by either reaction of MeHg with reactive oxygen species6 or intramolecular charge transfer from thiol-coordinated excited triplet state dissolved organic matter (DOM) to the Hg–C bond.7 Recently, Lu et al.8 reported MeHg decomposition by methanotrophs carrying out C1 metabolism. Lastly, biological MeHg degradation in anoxic environments where microorganisms with the mer operon is largely absent may also occur via “oxidative demethylation,”9 and has been reported in both pure culture and field experiments. An investigation of MeHg degradation in estuarine sediments containing Desulfovibrio spp. suggested a pathway in which the methyl group of MeHg was oxidized to CO2, releasing inorganic Hg(II).9 Subsequent studies have shown that many Hg-methylating bacteria (e.g., sulfate-reducing bacteria) are capable of simultaneous Hg methylation and MeHg degradation.10–13 The biochemical pathway(s) by which MeHg is degraded by such bacteria are unknown. Complicating efforts to understand MeHg degradation by anaerobes (especially sulfate-reducing bacteria) is the possibility of chemical MeHg degradation by reaction with hydrogen sulfide/bisulfide.

Early experiments on MeHg degradation in sediments (working at exceptionally high concentrations of MeHg) reported the production of a volatile product,14 subsequently identified as dimethylmercury (DMeHg),15 following addition of MeHg to sulfidic sediments. Baldi et al.10 suggested that DMeHg production occurred via decomposition of a bis(methylmercury) sulfide intermediate, (CH3Hg)2S, (referred to as dimethylmercury sulfide in the original paper10) to DMeHg and metacinnabar (HgS). The existence of the binuclear complex (CH3Hg)2S was reported in the mid-1960s by Schwarzenbach and Schellenberg.16 However, this species is not routinely considered in modeling MeHg equilibrium speciation in natural waters (see for example, Skyllberg17 and Liem-Nguyen et al.18). The reaction for the formation of the binuclear complex:

 
2CH3Hg+ + HS ⇄ (CH3Hg)2S + H+(1)
is reported to have a log[thin space (1/6-em)]K of 23.5 (after correcting for ionic strength and the pKas of H2S and HS).16 Using this value of the formation constant, (CH3Hg)2S is predicted to be a minor species in natural waters with low MeHg concentrations. However, at nM [MeHg]T concentrations and μM [H2S]T, (CH3Hg)2S can be an appreciable or even the largest fraction of the total MeHg (Fig. SI-1). Such conditions are typical of laboratory experiments where MeHg additions are made to pure cultures of microorganisms,11–13 and possibly natural systems where sulfide concentrations are limited by metal sulfide (especially FeS(s)) solubility.19 We note that most laboratory experiments of MeHg degradation by microorganisms have not included sufficient controls to rule out the possibility of abiotic MeHg losses, such as the formation of insoluble (CH3Hg)2S.

There has also been renewed interest in MeHg degradation by reaction with sulfide in the toxicology community. For example, Yoshida et al.20 demonstrated that mammalian cells could detoxify MeHg via reaction of H2ST (produced via cystathionine β-synthase and cystathionine γ-lyase) with MeHg to produce (CH3Hg)2S. This study included direct confirmation of (CH3Hg)2S by electron impact ionization mass spectrometry.20 Other work aimed at understanding the chemical basis of selenium antagonism to MeHg toxicity has found that MeHg reacts with selenoamino acids to form a bis(methylmercury)selenide complex (as identified by 199Hg NMR) analogous to (CH3Hg)2S, which subsequently undergoes decomposition to form DMeHg and HgSe.21,22 Lastly, recent experiments by Jonsson et al.23 have shown that MeHg can react with aqueous sulfide, dithiols, and metal sulfide mineral surfaces to produce DMeHg. Up to 2% of added MeHg was converted to DMeHg on FeS(s) surfaces within 1 h, while <0.1% was converted by reaction with aqueous sulfide.23 The origin of DMeHg in the open ocean (where DMeHg can constitute up to 30–40% of total alkylated Hg24,25) is a significant knowledge gap in our understanding of the biogeochemical cycling of Hg. The work of Jonsson et al.23 raises the possibility that low-oxygen microenvironments associated with sinking particulate organic matter with higher concentrations of metal sulfide nanoparticles may contribute to in situ formation of DMeHg in the ocean.

Despite the literature reports purporting the existence of (CH3Hg)2S and its breakdown to form DMeHg and HgS, we lack sufficient information regarding the thermodynamic stability of (CH3Hg)2S, the kinetics of (CH3Hg)2S formation, and the kinetics of DMeHg formation from (CH3Hg)2S to evaluate whether reaction of MeHg with sulfide is an important pathway for loss of MeHg and formation of DMeHg in either aquatic ecosystems or laboratory experiments. In this paper, we describe the reaction of MeHg with sulfide over a wide range of [MeHg]T/[H2S]T ratios and pH conditions with the aims of (a) refining the value of the equilibrium constant for (CH3Hg)2S formation and (b) quantifying rates of DMeHg formation. We also carry out density functional theory (DFT) calculations to investigate the mechanism of DMeHg formation from (CH3Hg)2S. Development of thermodynamic and kinetic models for MeHg speciation and conversion to DMeHg in sulfidic waters is critical to improving our understanding of MeHg accumulation in the environment.

2. Experimental section

2.1. MeHg reaction with sulfide

The thermodynamics and kinetics of the reaction of MeHg with aqueous sulfide were studied by systemically varying initial MeHg concentration ([MeHg]0), sulfide concentration ([H2S]T), and pH. A summary of all experimental conditions is provided in the ESI (Table SI-1). Briefly, reaction conditions consisted of 2.5 to 75 nM MeHg with 0.01 to 1300 μM sulfide over the pH range 7.14 to 9.49. Reaction progress was observed by monitoring MeHg, total Hg (THg) and DMeHg concentrations over time. All reactions were carried out under O2-free conditions in butyl-rubber sealed borosilicate glass serum bottles with N2-sparged solutions. Reactants were added to reactor bottles inside a glovebag (atmosphere 95% N2, 5% H2; Pd catalyst for removal of O2), then crimp-sealed, covered with aluminum foil to prevent photochemical reactions, and reacted at room temperature (22 °C) with constant mixing (550 rpm).

All reactions were carried out with 50 mM NaNO3 (to control ionic strength) and 25 mM of an appropriate pH buffer (3-(N-morpholino)propane sulfonic acid (MOPS) for pH 7.14 to 7.63 experiments; 4-(2-hyroxyethyl)-1-piperazineethanesulfonic acid (HEPES) for pH 8.14 experiments; and 2-(cyclohexylamino)ethanesulfonic acid (CHES) for pH 8.84 to 9.49 experiments). The pH of the reactors was adjusted with N2-degassed NaOH and measured again at the conclusion of the experiment. Sulfide was added to reactors from freshly prepared saturated Na2S stocks (prepared by washing Na2S crystals with degassed deionized water (DDIW), blotting dry, and then dissolving in DDIW in sealed borosilicate glass serum bottles).

MeHg was added as either normal isotope abundance MeHgOH (Brooks Rand Laboratories, 1 mg L−1 stock) or as stable isotope enriched Me198HgOH (92.78% 198Hg from Oak Ridge Laboratories, synthesized by reaction with methylcobalamin26). The use of enriched Me198Hg in select experiments aided in distinguishing the DMeHg product from background Hg(0) and improved detection limits for DMeHg. The initial headspace to aqueous phase ratio in reactors was 1[thin space (1/6-em)]:[thin space (1/6-em)]5 (50 mL reactors) or 1[thin space (1/6-em)]:[thin space (1/6-em)]10 (200 mL reactors). The reactors were sampled periodically by withdrawing either aqueous samples or headspace gas with N2-flushed syringes. Samples for MeHg or DMeHg were analyzed immediately; samples for THg and sulfide were preserved with 1% BrCl and a 1[thin space (1/6-em)]:[thin space (1/6-em)]1 dilution in sulfide antioxidant buffer (SAOB),27 respectively.

As a control, we also monitored MeHg, THg, and DMeHg concentrations in a sulfide-free reactor containing 25 nM Me198HgOH, 50 mM NaCl and 25 mM MOPS (pH 7.50) prepared identically to treatments described above.

2.2. Analytical methods

MeHg was analyzed through direct ethylation: samples were added to a glass bubbler filled to approximately one-third capacity with DI water containing Me199Hg (Oak Ridge Laboratories, 91.95% 199Hg) as an internal standard, 300 μL 2 M acetate buffer (pH 4.76), and, after samples were added, 40 μL of 10 g L−1 Na-tetraethylborate (NaTEB). Bubblers were sealed and shaken, then bubbled for 15 minutes with N2 to purge sample MeHg (MeHgEt after ethylation) onto a Tenax trap (poly(2,6-diphenylphenylene oxide), a porous polymer resin). Hg isotopes and species were measured through use of isotope dilution gas chromatography inductively coupled mass spectrometry (GC-ICP-MS) with an Agilent 7500ce ICP-MS (as described by Hintelman and Evans28). Hg species were thermally desorbed from the Tenax trap using a nichrome wire, and Hg species separated on a packed column GC (60/80 mesh 15% OV-3 chromasorb WAS-DMSC; Supelco) and introduced into the ICP-MS in a stream of argon gas. Method precision was evaluated based on triplicate analysis of MeHg samples collected from a single reactor within 1 min of each other; the average relative standard deviation was 6.3 ± 6.4% (n = 6 triplicate sets). Recovery of the isotope dilution spike averaged 82.8 ± 31.5%. The absolute instrument detection limit for Me198Hg (and DMeHg discussed below) averaged 1.8 ± 2.0 fmol. In our quantification of MeHg, we assume that MeHg from the sample is derivatized with the same efficiency as MeHg from the isotope-labeled internal standard. We further assume that any (CH3Hg)2S present in the sample is not recovered by the ethylation reaction due to its lower electrophilicity and that no (CH3Hg)2S is formed from the isotope-labeled standard during the ethylation reaction given the short reaction time.

The reactor headspace was sampled to detect and quantify DMeHg. A 10 mL volume of headspace was withdrawn from each reactor using a N2-flushed syringe, and the gas was injected through a butyl rubber septum fitted directly onto a Tenax trap. The trap was thermally desorbed and Hg species analyzed by GC-ICP-MS as described above. DMeHg was identified based on retention time intermediate to elemental Hg(0) and MeHgEt (see sample chromatograms in ESI Fig. SI-2) and quantified based on a calibration curve for MeHg. Total DMeHg was quantified using Henry's Law and the respective volumes of headspace and liquid at each point in the experiment, with a dimensionless Henry's Law constant of 0.13.29 An instrument detection limit of 0.37 pg of DMe198Hg in the reactor headspace corresponded to a detection limit of 0.12 pM (50 mL reactors) or 0.05 pM (200 mL reactors) for total DMeHg.

Total Hg was measured by ICP-MS with in-line reduction of Hg(II) to Hg0 vapor using 60 g L−1 SnCl2 in 1% HCl.30 The detection limit for this method was approximately 0.9 pM for enriched 198Hg or 15 pM for ambient Hg analysis. Method precision was evaluated by analysis of standards approximately every 10 samples; mean standard recovery was 99.9 ± 3.8%. Sulfide concentrations were measured periodically using a sulfide ion-selective electrode and silver/silver chloride reference electrode, and analyzed against a standard curve a Na2S stock (standardized by titration with 0.1 M Pb(NO3)2). The approximate limit of detection for the sulfide analysis was 0.1 μM.

In select experiments, we completed mass balances for MeHg and THg at the end of experiments by desorbing MeHg and inorganic Hg from bottle walls using 5% v/v HCl (MeHg) or 5% HCl amended with 2% BrCl (THg). MeHg and THg were subsequently analyzed using the approaches described above.

2.3. Equilibrium speciation modeling and data analysis

The equilibrium speciation of MeHg was modeled using the program MINEQL+ v. 4.5 (Environmental Research Software). Equilibrium constants were taken from Schwarzenbach and Schellenberg,16 Dryssen and Wedborg,31 and the NIST Critical Database32 and are summarized in ESI Table SI-2. As discussed in the results section, the concentration of (CH3Hg)2S was inferred based on the loss of MeHg reactive with NaTEB. The equilibrium concentration of (CH3Hg)2S was estimated using the mean MeHg concentration at the end of the experiment (typically using the last three data points). The equilibrium constant for the reaction corresponding to formation of (CH3Hg)2S was determined by holding the log[thin space (1/6-em)]K values for other MeHg complexation reactions constant at values reported in Table SI-2, and varying the log[thin space (1/6-em)]K for (CH3Hg)2S formation. Speciation was computed for experimental conditions of initial MeHg concentration, pH, sulfide concentration, and ionic strength. The best-fit log[thin space (1/6-em)]K for (CH3Hg)2S formation was found by iteratively comparing the speciation model results to the experimental data, and minimizing the weighted sum of squared residuals (SSR). A 95% confidence interval on log[thin space (1/6-em)]K was estimated based on the plot of weighted SSR versus log[thin space (1/6-em)]K, as described in Kemmer and Keller (2010).33 Rates of MeHg loss and DMeHg production were determined by linear regression of the initial linear portion of the datasets (r2 generally better than 0.9).

2.4. Quantum chemical calculations

All DFT calculations were performed with a development version of NWChem.34 The B3LYP exchange–correlation functional35–37 and D3BJ empirical long-range dispersion correction model38,39 were used throughout unless otherwise noted. Calculations were performed within the restricted, closed-shell formalism. An “xfine” numerical integration grid and a “tight” screening tolerance were used. The SMD continuum solvation model40 with water as the solvent (ε = 78.4) was used for all calculations. All geometries were optimized using the Stuttgart scalar relativistic small core (RSC) effective core potential (ECP) and corresponding basis set for Hg,41 and the 6-31G(d) basis set for all other atoms. To assess the geometric effect of using a different density functional, we also optimized the geometries with the M06-L density functional. The heavy-atom root-mean-square deviations (RMSDs) were less than 0.04 Å compared to the B3LYP geometries. Thus, B3LYP geometries were used hereafter. After optimization of the reactant state (RS) and product state (PS) geometries, minimum energy path optimizations were carried out with the nudged elastic band (NEB) method42,43 and refined with the zero-temperature string method.44,45 Starting structures for intermediate states (INT) and transition states (TS) were selected from the resulting minimum energy path for final unconstrained optimization (Fig. SI-3). Vibrational frequencies were computed to confirm that RS, INT and PS were local minima and that TS1 and TS2 were true first-order saddle points. Vibrational frequencies were also used to compute thermodynamic quantities within the harmonic approximation.

Hg is a heavy metal that exhibits strong relativistic effects. These effects are often adequately described with an ECP in which core electrons are replaced with an effective potential. However, all-electron relativistic approaches can provide increased accuracy in some cases, albeit at increased computational cost. Thus, to determine whether ECP approximations are accurate for the present (CH3Hg)2S system, we compared the energies obtained with scalar ECPs to all-electron scalar zeroth-order regular approximation (ZORA) energies (ESI Table SI-3). ZORA calculations were performed using the model potential approach,46,47 as implemented in NWChem.

Besides scalar relativistic effects, inclusion of spin–orbit effects can also be important. Although complexes with unpaired spins are likely to show greater spin–orbit effects, closed-shell systems can also show these effects. In addition to relativistic core contractions, which result in screening effects, the orbitals are further split, which can influence bonding and reactivity. Thus, to quantify spin–orbit effects we compared the results obtained with scalar and spin–orbit ECP approximations. Single-point energies were computed at the optimized stationary point geometries with the following relativistic approximations: (i) the cc-pVTZ-pp scalar ECP for Hg (SC-ECP),48,49 (ii) the cc-pVTZ-pp spin–orbit ECP for Hg (SO-ECP)48,49 with spin–orbit ECP parameters accessed from http://www.tc.uni-koeln.de/PP/clickpse.en.html, and (iii) the all-electron scalar zeroth-order relativistic approximation (SC-ZORA).46,50–55 The spin–orbit calculations were performed with the two-component DFT module in NWChem. For all calculations with the cc-pVTZ-pp ECP, we used the cc-pV(T+d)Z basis set for S56 and the cc-pVTZ basis set57 for all other atoms. For the ZORA calculations, we used the ZORA-Def2-TZVPP basis set58 for all atoms.

The relative energies for RS, INT, TS and PS computed with each method are provided in Table SI-3 in the ESI. Comparing these energies with the corresponding energies calculated with the Stuttgart RSC scalar ECP reveals that the single-point energies obtained with more sophisticated relativistic approaches changed the energy profiles consistently. In all cases the single-point energies were higher at each point than the corresponding RSC energies, suggesting that the RSC ECP underestimates relative energy barriers slightly. The cc-pVTZ-pp scalar and spin–orbit ECP energies differ by less than 1 kcal mol−1, indicating that spin–orbit effects are relatively minor for this system. The differences in the relative energies computed with the cc-pVTZ-pp scalar ECP and the all-electron scalar ZORA approach are also minimal (less than 1 kcal mol−1, in general). Therefore, all energies reported hereafter were computed with the cc-pVTZ-pp spin–orbit ECP, which accounts for both scalar and spin–orbit effects. To estimate the accuracy of the DFT method used in this work, we also calculated single-point energies with the M06-L, PBE0, and BLYP density functionals at the B3LYP geometries (Table SI-3). The computed energy barriers range from 27.0 to 31.9 kcal mol−1, in good agreement with the activation energy determined experimentally (∼30.1 kcal mol−1). The error for B3LYP is only 0.1 kcal mol−1 for this system, while it is ∼3 kcal mol−1 for M06-L. Thus, B3LYP was deemed the most accurate of the DFT approximations considered here.

3. Results and discussion

3.1. Time courses of MeHg and THg

Addition of MeHg to sulfidic solutions resulted in loss of MeHg detected by direct ethylation, with most of the MeHg loss occurring within 100 h (Fig. 1a3a). No loss of MeHg was observed in sulfide-free solutions (ESI Fig. SI-4). The rate and extent of MeHg loss in sulfidic solutions depended upon solution pH, initial MeHg concentration, and sulfide concentration. Representative time courses are shown in Fig. 1a (pH effect), Fig. 2a ([MeHg]0 effect) and Fig. 3a ([H2S]T effects). At fixed pH (7.5), the extent of MeHg loss from solution increased with increasing [MeHg]0/[H2S]T, with <20% of MeHg lost at the lowest [MeHg]0/[H2S]T ratios evaluated (0.13 nmol MeHg/μmol H2ST) to >75% MeHg lost at the highest [MeHg]0/[H2S]T ratios shown in Fig. 2 and 3 (125 nmol MeHg/μmol H2ST). At fixed [MeHg]0 (9.0 nM) and nearly fixed [H2S]T (5.3 ± 2.0 μM), MeHg losses were greatest at circumneutral pH (55–60% at pH 7.14–7.63) and declined with increasing pH, with <15% of added MeHg lost at pH 9.49 (Fig. 1a).
image file: c7em00533d-f1.tif
Fig. 1 Effect of pH on MeHg loss and DMeHg production in sulfidic solutions. [Me198Hg]0 = 9.0 nM, pH fixed by addition of 25 mM pH buffer (MOPS for pH 7.14 and 7.63 experiments, HEPES for pH 8.14 experiment, and CHES for pH 8.84 and 9.49 experiments). Ionic strength adjusted with 50 mM NaNO3. Measured sulfide concentrations were 3.9 μM (pH 7.14), 2.0 μM (pH 7.63), 6.7 μM (pH 8.14), 7.3 μM (pH 8.84), and 6.6 μM (pH 9.49).

image file: c7em00533d-f2.tif
Fig. 2 Effect of initial Me198Hg concentration on loss of Me198Hg and production of DMe198Hg in solutions containing 18.5 ± 1.5 μM total sulfide (H2ST) and 50 mM NaNO3 buffered at pH 7.50 by 25 mM MOPS.

image file: c7em00533d-f3.tif
Fig. 3 Effect of sulfide concentration ([H2S]T) on MeHg loss and DMeHg production. [Me198Hg]0 = 21.2 nM, pH = 7.50 (25 mM MOPS buffer), and ionic strength adjusted with 50 mM NaNO3.

Total Hg concentrations (determined by digesting samples in 1% BrCl) generally declined less than 30% over the duration of the experiments (Fig. 1b3b). Under conditions where greatest losses of MeHg were observed (e.g., pH 7.14 and 7.63 experiments shown in Fig. 1), larger losses of THg were observed (reaching up to nearly 50% of added MeHg). To close mass balances, for a subset of experiments, we acidified emptied reaction bottles with 5% HCl and measured recovered MeHg and THg. Recovered THg was 98 ± 32% of the lost THg, indicating that sorption to bottle walls accounted for lost THg (ESI Table SI-4). Of the Hg recovered from the bottle walls by acidification, 65 ± 15% was MeHg (note, however, that MeHg speciation changes appreciably upon addition of 5% HCl).

Based on literature reports of the formation of a binuclear MeHg–sulfide complex (bis(methylmercury) sulfide; (CH3Hg)2S)15,16,20,59,60 and its selenide analogue (CH3Hg)2Se,21 we hypothesize that the formation of a (CH3Hg)2S complex, with lower electrophilicity relative to other MeHg species,20 accounts for the disappearance of MeHg amenable to quantification via direct ethylation (conversion of CH3HgX to volatile CH3HgCH2CH3). Mansfield and Black61 recently reported low recoveries of MeHg by an improved direct ethylation method when MeHg was equilibrated for 24 h with μM sulfide concentrations; a plausible explanation for their observation is the formation of (CH3Hg)2S, which is inert to reaction with the ethylating agent sodium tetraethylborate (NaTEB). The solubility of (CH3Hg)2S is not well constrained, but precipitation and adsorption of (CH3Hg)2S to bottle surfaces likely accounts for the observed losses of THg (e.g., Fig. 1b). Most of bottle wall-associated Hg was recoverable as MeHg following acidification with 5% HCl (Table SI-4), and we attribute this observation to conversion of (CH3Hg)2S to CH3HgCl complexes that are susceptible to ethylation by NaTEB. The remaining difference between HgT and MeHg following acid extraction may be (CH3Hg)2S not quantified during ethylation. As discussed in detail below, the observed dependence of MeHg losses on pH and [MeHg]0/[H2S]T ratio are consistent with loss of MeHg due to (CH3Hg)2S formation and precipitation.

3.2. DMeHg production

Concurrent with the loss of MeHg, concentrations of DMeHg increased over time (example time courses in Fig. 1c3c). Total DMeHg production was a small fraction of [MeHg]0 (<0.3%) and the loss of MeHg inferred to be (CH3Hg)2S formation. Importantly, rates of DMeHg production, spanning three orders of magnitude from 4 × 10−16 to 3 × 10−13 mol L−1 h−1 (4 × 10−6 to 1.5 × 10−3% h−1) over the experimental conditions evaluated here, increased in proportion to both the total loss of MeHg and the rate of MeHg loss (ESI Fig. SI-5 and 6), suggesting the involvement of (CH3Hg)2S as an intermediate in the formation of DMeHg from reaction of MeHg with aqueous sulfide. The slope of the regression between log of inferred [(CH3Hg)2S] and the log of the observed rate of DMeHg formation was found to be 1.36 ± 0.33, suggesting that the rate of DMeHg production is first-order with respect to [(CH3Hg)2S]:
 
image file: c7em00533d-t1.tif(2)

Others have reported DMeHg formation from reaction of MeHg with sulfide,14,15,59 but our study is the first systematic study of reaction kinetics under conditions approaching environmental relevance. The magnitude of observed DMeHg formation rates agrees well with recent reports of less than 0.1% of MeHg converted to DMeHg by aqueous sulfide in 1 h experiments at [MeHg]0/[H2S]T ratios of 4–1900 nmol MeHg/μmol H2ST.23 Observed rates and yields of DMeHg formation were substantially lower than those reported in earlier descriptions of MeHg conversion to DMeHg in sulfidic solutions. For example, working at much higher [MeHg]0 (∼0.5 mM), Baldi et al.59 reported up to 20% conversion of MeHg to DMeHg over a 16 day period in incubations of D. desulfuricans strain LS grown in a sulfidogenic medium. Notably, the observed rates of DMeHg formation are several orders of magnitude slower than that reported for reaction of MeHg with FeS surfaces at similar MeHg/reduced sulfur ratios (rates of up to 2% per h).23

Relativistic DFT calculations provide further insight into the mechanism of the reaction:

 
(CH3Hg)2S → (CH3)2Hg + HgS(3)

in aqueous solution. According to the calculations, the formation of DMeHg and HgS from (CH3Hg)2S conforms to a two-step mechanism (Fig. 4). In the first step, one of the MeHg groups is transferred to the adjacent Hg to form an unusual metastable intermediate in which one of the two Hg–S bonds in RS is replaced by a Hg–Hg bond in INT. The Hg–S bond length decreases from 2.44 Å in RS to 2.36 Å in INT, indicating double bond-like character. The reaction proceeds from RS to form INT via TS1 with a free energy barrier of 25.7 kcal mol−1 for the first step. Both TS1 and INT are characterized by a relatively strong Hg–Hg interaction, which is stabilized by anionic ligands to both Hg centers (i.e., one by CH3 and the other by both CH3 and S2−). The Hg–Hg distance in INT is 2.88 Å, which is shorter than the general mercurophilic interaction of ∼3.6 Å observed in Hg(II) compounds.62


image file: c7em00533d-f4.tif
Fig. 4 Calculated free energy profile and stationary point geometries for (CH3Hg)2S decomposition to form DMeHg in the aqueous phase. Relative energies (kcal mol−1) calculated with the cc-pVTZ-pp spin–orbit ECP for Hg are shown.

In the second step, the methyl group of the sulfur-coordinated Hg is transferred to the other Hg atom to form DMeHg in PS. The reaction proceeds from INT to the product state via TS2, which is characterized by a triangular geometry in which a CH3 group is shared by two Hg(II) centers, with one Hg center also coordinated to sulfur and the other to a methyl group (Fig. 4). The energy of TS2 relative to RS is 30.2 kcal mol−1, which corresponds to the overall rate-limiting step of the reaction. Converting this reaction barrier to a rate constant using the Eyring equation results in a value of 9.4 × 10−7 h−1 at 22 °C. In the product state, the Hg–Hg distance is 2.88 Å, the same as in INT, indicating a strong Hg–Hg interaction between DMeHg and HgS. We estimate that an additional 1.8 kcal mol−1 of free energy is required to break this Hg–Hg interaction in aqueous solution. However, the reaction will proceed further to liberate DMeHg and insoluble HgS (metacinnabar) nanoparticles. Considering that the experimentally measured cohesive energy of HgS in the zinc blende polymorph is ∼−80.0 kcal mol−1,63,64 a lower bound for metacinnabar nanoparticles, we estimate that the overall reaction free energy may be as low as −52.4 kcal mol−1. Nevertheless, the formation of DMeHg and HgS(s) from (CH3Hg)2S is thermodynamically favorable, which should drive the reaction equilibrium strongly toward the product state.

3.3. Models for MeHg speciation and DMeHg production

Equilibrium speciation of MeHg for experimental conditions of [MeHg]0, [H2S]T, and pH was modeled in MINEQL+ (v. 4.5) using published equilibrium constants for MeHg complexes (summarized Table SI-2). The value of the equilibrium constant for the formation of the binuclear complex (CH3Hg)2S (eqn (1)) was initially taken as log[thin space (1/6-em)]K = 23.5, as determined by Schwarzenbach and Schellenberg16via pH titration of MeHg and NaHS solutions. Formation of (CH3Hg)2S in experiments was estimated as one-half the loss of MeHg (based on the average of the last three data points in the MeHg time course). The equilibrium speciation model significantly underpredicted the observed losses of MeHg (ESI Fig. SI-7).

There are at least two significant uncertainties with the published equilibrium constant for (CH3Hg)2S formation. First, Schwarzenbach and Schellenberg16 neglected the contributions of CH3HgSH to the mass balance for total dissolved MeHg. Second, they formulated MeHg-sulfide complex equilibria in terms of S2− rather than HS, which introduces errors related to the pKa of HS (a value of 14.2 was assumed; the true pKa is likely greater than 18.0 (ref. 65)). We iteratively fit the data by adjusting the log[thin space (1/6-em)]K for the binuclear complex and comparing the experimental data to the model prediction. Using this approach, we obtained a best fit value of log[thin space (1/6-em)]K = 26.0 ± 0.2 (Fig. 5). That the data fall closely along the 1[thin space (1/6-em)]:[thin space (1/6-em)]1 line is strong evidence in support of our contention that the loss of MeHg from solution is due to the formation of a binuclear MeHg–sulfide complex.


image file: c7em00533d-f5.tif
Fig. 5 Measured loss of MeHg versus that predicted by equilibrium speciation model with log[thin space (1/6-em)]K = 26.0 for reaction: 2CH3Hg+ + HS = (CH3Hg)2S + H+.

Next, we used our improved equilibrium speciation model to predict [(CH3Hg)2S] across all experimental conditions and examined the relationship between predicted [(CH3Hg)2S] and the observed rate of DMeHg formation (Fig. 6). Although there is some scatter in the data, the predicted [(CH3Hg)2S] is a good predictor (r2 = 0.64) of the observed rate of DMeHg formation. The large uncertainty associated with the intercept of this regression precludes accurate estimation of the first-order rate constant in eqn (2) from Fig. 6, however. Linear regression of predicted [(CH3Hg)2S] vs. DMeHg formation rate yielded a k (equivalent to slope) equal to 1.5 ± 0.4 × 10−6 h−1. This experimental result for the rate constant is in close agreement with that obtained based on the activation energy determined from quantum chemical calculations (9.4 × 10−7 h−1), offering additional support for our proposed mechanism. With the improved equilibrium constant for (CH3Hg)2S formation and the rate constant for conversion of (CH3Hg)2S to DMeHg, we can now predict both MeHg speciation and DMeHg production rates in field and laboratory systems.


image file: c7em00533d-f6.tif
Fig. 6 Relationship between equilibrium speciation model predicted (CH3Hg)2S and observed rate of dimethylmercury (DMeHg) formation across all experiments. Circled data point corresponds to measured sulfide concentration below limit of detection (0.1 μM), and calculation of equilibrium speciation is correspondingly uncertain. Error bars represent standard errors of DMeHg formation rates determined by linear regression of DMeHg time course data.

3.4. Environmental implications

Our results call for careful reconsideration of MeHg speciation in laboratory and field experiments aimed at investigating microbial MeHg degradation. Oremland et al.9 originally proposed MeHg degradation by anaerobic bacteria not possessing the mer operon via a so-called oxidative demethylation pathway (oxidation of the methyl group to CO2). A number of subsequent laboratory studies10–13 have investigated MeHg degradation by anaerobic bacteria (principally sulfate-reducing bacteria; SRB) by amending bacterial cultures with high concentrations (5 nM to as high as 0.5 mM) of MeHg where formation of the binuclear (CH3Hg)2S species will be favored. As shown in ESI Fig. SI-8, with a [MeHg]T of 10 nM, (CH3Hg)2S is predicted to be the dominant MeHg species in solution at circumneutral pH at sulfide concentrations up to 100 μM. Notably, pure culture studies of MeHg degradation by SRB have not included abiotic controls with sulfide concentrations approaching those in the biological assays. Typical rates of MeHg disappearance in cultures of SRB have been reported as 10–40% per day.11–13 Given the similarity to MeHg losses by reaction with sulfide reported in this study, we suggest that formation of insoluble (CH3Hg)2S may be at least partially responsible for losses of MeHg observed in these pure culture experiments.

Abiotic losses of MeHg may help to explain results such as similar rates of MeHg degradation across diverse SRB strains12 and significantly greater MeHg degradation when SRB are grown in sulfidogenic (mM [H2S]T) vs. non-sulfidogenic media (typically less than 100 μM, but highly dependent upon whether degradable66 organic thiols are present in media).13 In fact, MeHg degradation was not observed for a suite of SRB under conditions ([MeHg] = 1 nM, [L-cysteine] = 500 μM, [H2S]T = 7.5 μM, pH = 7.27)67 where equilibrium speciation calculations (using our revised log[thin space (1/6-em)]K for (CH3Hg)2S) predict virtually no (CH3Hg)2S formation (<0.01% of total MeHg). Depending on the analytical approach to measuring MeHg and the recovery of (CH3Hg)2S in conventional analytical approaches (e.g., steam distillation or methylene chloride extraction), formation of (CH3Hg)2S may be misattributed to enzymatic demethylation. The low pH and addition of sulfide-binding Cu employed in steam distillation (e.g., EPA Method 1630) will likely convert (CH3Hg)2S to CH3HgClx1−x species with the result that (CH3Hg)2S is quantified as mononuclear MeHg. We recommend control experiments with spent medium (containing similar concentrations of sulfide) and careful attention to mass balances to investigate possible abiotic contributions to MeHg loss in future laboratory investigations of MeHg degradation by microorganisms.

While we have demonstrated DMeHg formation via (CH3Hg)2S in laboratory experiments, it remains questionable whether this process is important in natural environments. Using our revised equilibrium constant for (CH3Hg)2S formation, we computed MeHg speciation for a range of [MeHg]T, [H2S]T, pH, and concentrations of the competitive ligands Cl and DOM thiols (ESI, Fig. SI-9). In all cases, the fraction of MeHgT as (CH3Hg)2S increases very slightly from pH 4.0 to a maximum in the pH 6.0 to 8.0 range and then drops sharply at pH values above the maximum. Concentrations of Cl up to 10 mM (a concentration higher than that observed for most freshwaters68) have minimal impact on MeHg speciation. The significance of the (CH3Hg)2S species decreases with decreasing [MeHg]T/[H2S]T ratio. For example, at [MeHg]T equal to 1 pM (0.20 ng L−1), (CH3Hg)2S is predicted to be a significant species at [H2S]T less than 0.1 μM. At 10 pM [MeHg]T (2.0 ng L−1), concentrations observed in MeHg-contaminated sites, (CH3Hg)2S is significant at [H2S]T less than 10 μM, and the dominant MeHg species at [H2S]T less than 0.1 μM (ESI, Fig. SI-9b).

If complexation with DOM thiols is also considered in the speciation model, then the importance of (CH3Hg)2S diminishes (ESI, Fig. SI-9c and d). Thiol concentrations in the range of 0.1 to 1 μM have been reported in sediment porewaters, wetlands, and lakes,69 concentrations sufficient to significantly depress (CH3Hg)2S formation. At thiol concentrations above 10 μM, (CH3Hg)2S is predicted to be a minor species (always less than 7% of [MeHg]T for the same [MeHg]T, pH, and [H2S]T conditions considered in Fig. SI-9). Low sulfide and thiol concentrations favoring (CH3Hg)2S formation may occur at the oxic/anoxic interface of sediments or redox stratified lakes or the open ocean. In the open ocean, concentrations of 0.2 to 2.0 nM of total sulfide have been reported, with most of the sulfide complexed to trace metals.70,71 Significantly higher sulfide concentrations in the ocean may be observed in anaerobic microenvironments associated with sinking particulate organic matter72 or in coastal oxygen minimum zones.73 There are few reports of thiol concentrations in the ocean, but work by Swarr et al.,74 indicates that cysteine and glutathione may be present at 0.1–0.5 nM concentrations. Using representative data for the oceans, with total [MeHg] = 0.05–0.6 pM,25 [H2S]T = 0.1–2.0 nM,71 total [thiols] = 0.1–0.5 nM, [Cl] = 0.55 M, and pH = 8.2, we find that (CH3Hg)2S may comprise ∼5% to greater than 50% of total MeHg. This simple calculation does not include competition for sulfide and thiols by other metals, and more detailed observations of sulfide and organic thiol concentrations would help to better constrain the significance of (CH3Hg)2S to MeHg speciation in the oceans. Nevertheless, the species (CH3Hg)2S likely warrants inclusion in equilibrium speciation models for natural systems.

Having reexamined MeHg speciation in natural waters, we can now ask whether formation of (CH3Hg)2S is a significant pathway for DMeHg production. Even under situations where (CH3Hg)2S constitutes the majority of aqueous MeHg, rates of conversion to DMeHg are likely to be slow. For example, for water containing 10 pM MeHgT and 0.1 μM H2ST at pH 7.0, where (CH3Hg)2S constitutes about 50% of MeHgT, the predicted rate of DMeHg formation is 3.8 × 10−5 pM−1 h−1 (using a rate constant k in eqn (2) equal to 1.5 × 10−6 h−1). Assuming pseudo-first order conditions, over a one month period, less than 1% of MeHgT would be converted to DMeHg, suggesting that reaction with aqueous sulfide will not lead to significant DMeHg accumulation in freshwater systems with short MeHg residence times. In contrast, initial rates of methylation of MeHg to form DMeHg on FeS surfaces can reach nearly 0.1% h−1,23 suggesting a dominance for the surface-catalyzed pathway under these conditions. DMeHg concentrations are not routinely measured in monitoring of Hg speciation in freshwater systems, as acid preservation of samples leads to DMeHg decomposition to MeHg.75 Whether DMeHg accumulates in freshwater ecosystems or whether DMeHg evasion to the atmosphere is an important MeHg sink term in mass balance models for MeHg are questions deserving of further attention, whatever the mechanism for DMeHg formation. However, in the subsurface ocean, MeHg residence times (∼20 years (ref. 4)) may be sufficiently long to allow for appreciable conversion of MeHg to DMeHg via the homogeneous mechanism.

In summary, experimental and computational evidence together indicate that DMeHg forms from reaction of MeHg with sulfide in aqueous solution, a reaction first proposed 40 years ago.14,15 The kinetics and mechanism for this reaction had not been adequately characterized prior to this study, however. DFT calculations suggest that the decomposition of (CH3Hg)2S to DMeHg and HgS proceeds through a novel rearrangement of the (CH3Hg)2S complex facilitated by Hg–Hg interactions followed by intramolecular methyl transfer to form DMeHg. This reaction is likely a significant pathway for DMeHg formation in laboratory experiments where high MeHg concentrations favor formation of the binuclear (CH3Hg)2S complex. The importance of this reaction in natural systems, especially the subsurface ocean, deserves further attention.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We thank Sofi Jonsson and Scott Brooks for substantive comments on earlier versions of the manuscript, Ryne C. Johnston and Liyuan Liang for helpful discussions, and T. Andrew Mobley for insightful suggestions. This research was funded through the Grinnell College Mentored Advanced Project program. This work was also supported by the U.S. Department of Energy (DOE) Office of Science, Biological and Environmental Research, Subsurface Biogeochemical Research (SBR) Program through the Mercury Science Focus Area Program (SFA) at Oak Ridge National Laboratory (ORNL). ORNL is managed by UT-Battelle LLC for the U.S. DOE under contract number DE-AC05-00OR22725. This research also benefitted from computational resources and expertise at the Environmental Molecular Sciences Laboratory (EMSL), a U. S. DOE Office of Science User Facility sponsored by the Office of Biological and Environmental Research (BER) and located at Pacific Northwest National Laboratory (PNNL), through Rapid Access award 50011 to JMP. PNNL is a multi-program national laboratory operated for the U.S. DOE by Battelle under contract DE-AC05-76RL01830.

References

  1. D. G. Streets, H. M. Horowitz, D. J. Jacob, Z. Lu, L. Levin, A. F. H. ter Schure and E. M. Sunderland, Environ. Sci. Technol., 2017, 51, 5969–5977 CrossRef CAS PubMed .
  2. J. M. Parks, A. Johs, M. Podar, R. Bridou, R. A. Hurt, S. D. Smith, S. J. Tomanicek, Y. Qian, S. D. Brown, C. C. Brandt, A. V. Palumbo, J. C. Smith, J. D. Wall, D. A. Elias and L. Liang, Science, 2013, 339, 1332–1335 CrossRef CAS PubMed .
  3. C. C. Gilmour, M. Podar, A. L. Bullock, A. M. Graham, S. D. Brown, A. C. Somenahally, A. Johs, R. A. Hurt Jr, K. L. Bailey and D. A. Elias, Environ. Sci. Technol., 2013, 47, 11810–11820 CrossRef CAS PubMed .
  4. C. T. Driscoll, R. P. Mason, H. M. Chan, D. J. Jacob and N. Pirrone, Environ. Sci. Technol., 2013, 47, 4967–4983 CrossRef CAS PubMed .
  5. T. Barkay, S. Miller and A. Summers, FEMS Microbiol. Rev., 2003, 27, 355–384 CrossRef CAS PubMed .
  6. T. Zhang and H. Hsu-Kim, Nat. Geosci., 2010, 3, 473–476 CrossRef CAS PubMed .
  7. Y. Qian, X. Yin, H. Lin, B. Rao, S. C. Brooks, L. Liang and B. Gu, Environ. Sci. Technol. Lett., 2014, 1, 426–431 CrossRef CAS .
  8. X. Lu, W. Gu, L. Zhao, M. Farhan Ul Haque, A. A. DiSpirito, J. D. Semrau and B. Gu, Sci. Adv., 2017, 3, e1700041 CrossRef PubMed .
  9. R. Oremland, C. Culbertson and M. Winfrey, Appl. Environ. Microbiol., 1991, 57, 130–137 CAS .
  10. F. Baldi, M. Pepi and M. Filippelli, Appl. Environ. Microbiol., 1993, 59, 2479–2485 CAS .
  11. K. Pak and R. Bartha, Appl. Environ. Microbiol., 1998, 64, 1013–1017 CAS .
  12. R. Bridou, M. Monperrus, P. R. Gonzalez, R. Guyoneaud and D. Amouroux, Environ. Toxicol. Chem., 2010, 30, 337–344 CrossRef PubMed .
  13. C. C. Gilmour, D. A. Elias, A. M. Kucken, S. D. Brown, A. V. Palumbo, C. W. Schadt and J. D. Wall, Appl. Environ. Microbiol., 2011, 77, 3938–3951 CrossRef CAS PubMed .
  14. J. R. Rowland, M. J. Davies and P. Grasso, Nature, 1977, 265, 718–719 CrossRef PubMed .
  15. P. J. Craig and P. D. Bartlett, Nature, 1978, 275, 635–637 CrossRef CAS .
  16. G. Schwarzenbach and M. Schellenberg, Helv. Chim. Acta, 1965, 48, 28–46 CrossRef CAS .
  17. U. Skyllberg, J. Geophys. Res.: Biogeosci., 2008, 113, G00C03 Search PubMed .
  18. V. Liem-Nguyen, U. Skyllberg and E. Bjorn, Environ. Sci. Technol., 2017, 51, 3678–3686 CrossRef CAS PubMed .
  19. D. Rickard and J. Morse, Mar. Chem., 2005, 97, 141–197 CrossRef CAS .
  20. E. Yoshida, T. Toyama, Y. Shinkai, T. Sawa, T. Akaike and Y. Kumagai, Chem. Res. Toxicol., 2011, 24, 1633–1635 CrossRef CAS PubMed .
  21. M. A. K. Khan and F. Wang, Chem. Res. Toxicol., 2010, 23, 1202–1206 CrossRef CAS PubMed .
  22. A. M. Asaduzzaman and G. Schreckenbach, Inorg. Chem., 2011, 50, 2366–2372 CrossRef CAS PubMed .
  23. S. Jonsson, N. M. Mazrui and R. P. Mason, Sci. Rep., 2016, 6, 27958 CrossRef PubMed .
  24. C. R. Hammerschmidt and K. L. Bowman, Mar. Chem., 2012, 132–133, 77–82 CrossRef CAS .
  25. K. L. Bowman, C. R. Hammerschmidt, C. H. Lamborg and G. Swarr, Deep Sea Res., Part II, 2015, 116, 251–261 CrossRef CAS .
  26. C. Bancon-Montigny, L. Yang, R. E. Sturgeon, V. Colombini and Z. Mester, Appl. Organomet. Chem., 2004, 18, 57–64 CrossRef CAS .
  27. L. S. Clesceri, A. E. Greenberg and A. D. Eaton, in Standard Methods for the Examination of Water and Wastewater, American Public Health Association, 2000, pp. 4–170–4–183 Search PubMed .
  28. H. Hintelmann and R. Evans, Fresenius' J. Anal. Chem., 1997, 358, 378–385 CrossRef CAS .
  29. Y. Talmi and R. E. Mesmer, Water Res., 1975, 9, 547–552 CrossRef CAS .
  30. A. M. Graham, G. R. Aiken and C. C. Gilmour, Environ. Sci. Technol., 2012, 46, 2715–2723 CrossRef CAS PubMed .
  31. D. Dyrssen and M. Wedborg, Water, Air, Soil Pollut., 1991, 56, 507–519 CrossRef .
  32. National Institute of Standards and Technology NIST Critically Selected Stability Constants of Metal Complexes, Gaithersburg, MD, 2013 Search PubMed .
  33. G. Kemmer and S. Keller, Nat. Protoc., 2010, 5, 267–281 CrossRef CAS PubMed .
  34. M. Valiev, E. J. Bylaska, N. Govind, K. Kowalski, T. P. Straatsma, H. J. J. Van Dam, D. Wang, J. Nieplocha, E. Apra, T. L. Windus and W. A. de Jong, Comput. Phys. Commun., 2010, 181, 1477–1489 CrossRef CAS .
  35. C. Lee, W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 37, 785–789 CrossRef CAS .
  36. A. D. Becke, J. Chem. Phys., 1993, 98, 5648–5652 CrossRef CAS .
  37. P. J. Stephens, F. J. Devlin, C. F. Chabalowski and M. J. Frisch, J. Phys. Chem., 1994, 98, 11623–11627 CrossRef CAS .
  38. S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys., 2010, 132, 154104 CrossRef PubMed .
  39. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465 CrossRef CAS PubMed .
  40. A. V. Marenich, C. J. Cramer and D. G. Truhlar, J. Phys. Chem. B, 2009, 113, 6378–6396 CrossRef CAS PubMed .
  41. M. Dolg, H. Stoll, H. Preuss and R. M. Pitzer, J. Phys. Chem., 1993, 97, 5852–5859 CrossRef CAS .
  42. H. Jónsson, G. Mills and K. W. Jacobsen, in Classical and Quantum Dynamics in Condensed Phase Simulations, ed. B. J. Berne, G. Ciccotti and D. F. Coker, World Scientific, Singapore, 1998, pp. 385–404 Search PubMed .
  43. G. Henkelman and H. Jónsson, J. Chem. Phys., 2000, 113, 9978–9985 CrossRef CAS .
  44. W. E, W. Ren and E. Vanden-Eijnden, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 66, 052301 CrossRef .
  45. W. E, W. Ren and E. Vanden-Eijnden, J. Chem. Phys., 2007, 126, 164103 CrossRef PubMed .
  46. C. van Wüllen, J. Chem. Phys., 1998, 109, 392–399 CrossRef .
  47. C. van Wüllen and C. Michauk, J. Chem. Phys., 2005, 123, 204113 CrossRef PubMed .
  48. K. A. Peterson and C. Puzzarini, Theor. Chem. Acc., 2005, 114, 283–296 CrossRef CAS .
  49. D. Figgen, G. Rauhut, M. Dolg and H. Stoll, Chem. Phys., 2005, 311, 227–244 CrossRef CAS .
  50. C. Chang, M. Pelissier and P. Durand, Phys. Scr., 1986, 34, 394 CrossRef CAS .
  51. E. V. Lenthe, E. J. Baerends and J. G. Snijders, J. Chem. Phys., 1993, 99, 4597–4610 CrossRef .
  52. E. van Lenthe, E. J. Baerends and J. G. Snijders, J. Chem. Phys., 1994, 101, 9783–9792 CrossRef CAS .
  53. E. van Lenthe, J. G. Snijders and E. J. Baerends, J. Chem. Phys., 1996, 105, 6505–6516 CrossRef CAS .
  54. S. Faas, J. G. Snijders, J. H. van Lenthe, E. van Lenthe and E. J. Baerends, Chem. Phys. Lett., 1995, 246, 632–640 CrossRef CAS .
  55. P. Nichols, N. Govind, E. J. Bylaska and W. A. de Jong, J. Chem. Theory Comput., 2009, 5, 491–499 CrossRef CAS PubMed .
  56. T. H. Dunning Jr, K. A. Peterson and A. K. Wilson, J. Chem. Phys., 2001, 114, 9244–9253 CrossRef .
  57. R. A. Kendall, T. H. Dunning Jr and R. J. Harrison, J. Chem. Phys., 1992, 96, 6796–6806 CrossRef CAS .
  58. D. A. Pantazis, X.-Y. Chen, C. R. Landis and F. Neese, J. Chem. Theory Comput., 2008, 4, 908–919 CrossRef CAS PubMed .
  59. F. Baldi, F. Parati and M. Filippelli, Water, Air, Soil Pollut., 1995, 80, 805–815 CrossRef CAS .
  60. Y. Abiko, E. Yoshida, I. Ishii, J. M. Fukuto, T. Akaike and Y. Kumagai, Chem. Res. Toxicol., 2015, 28, 1301–1306 CrossRef CAS PubMed .
  61. C. R. Mansfield and F. J. Black, Limnol. Oceanogr.: Methods, 2015, 13, e10009 CrossRef .
  62. H. Schmidbaur and A. Schier, Organometallics, 2015, 34, 2048–2066 CrossRef CAS .
  63. II-VI and I-VII Compounds; Semimagnetic Compounds, ed. O. Madelung, U. Rossler and M. Schulz, Springer-Verlag, Berlin/Heidelberg, 1999, vol. 41B Search PubMed .
  64. S. Biering and P. Schwerdtfeger, J. Chem. Phys., 2012, 136, 034504 CrossRef CAS PubMed .
  65. D. Rickard and G. W. Luther, Chem. Rev., 2007, 107, 514–562 CrossRef CAS PubMed .
  66. S. A. Thomas and J.-F. Gaillard, Environ. Sci. Technol., 2017, 51, 4642–4651 CrossRef CAS PubMed .
  67. A. M. Graham, A. L. Bullock, A. C. Maizel, D. A. Elias and C. C. Gilmour, Appl. Environ. Microbiol., 2012, 78, 7337–7346 CrossRef CAS PubMed .
  68. W. Stumm and J. Morgan, Aquatic Chemistry, John Wiley and Sons, NY, 3rd edn, 1996 Search PubMed .
  69. V. Liem-Nguyen, S. Bouchet and E. Bjorn, Anal. Chem., 2015, 87, 1089–1096 CrossRef CAS PubMed .
  70. G. W. I. Luther and E. Tsamakis, Mar. Chem., 1989, 27, 165–177 CrossRef CAS .
  71. J. Radford-Knery and G. A. Cutter, Geochim. Cosmochim. Acta, 1994, 58, 5421–5431 CrossRef .
  72. A. L. Shanks and M. L. Reeder, Mar. Ecol.: Prog. Ser., 1993, 96, 43–47 CrossRef .
  73. H. Schunck, G. Lavik, D. K. Desai, T. Großkopf, T. Kalvelage, C. R. Löscher, A. Paulmier, S. Contreras, H. Siegel, M. Holtappels, P. Rosenstiel, M. B. Schilhabel, M. Graco, R. A. Schmitz, M. M. M. Kuypers and J. LaRoche, PLoS One, 2013, 8, e68661 CAS .
  74. G. J. Swarr, T. Kading, C. H. Lamborg, C. R. Hammerschmidt and K. L. Bowman, Deep Sea Res., Part I, 2016, 116, 77–87 CrossRef CAS .
  75. G. L. Baughman, J. A. Gordon and R. G. Zepp, Chemistry of organomercurials in aquatic systems, U.S. Environmental Protection Agency, Athens, GA, 1973 Search PubMed .

Footnotes

Electronic supplementary information (ESI) available. See DOI: 10.1039/c7em00533d
Present Address: University of Wisconsin, Department of Civil and Environmental Engineering, Madison, Wisconsin 53706, United States.
§ Present Address: Duke University, Earth and Ocean Science, Durham, North Carolina 27708, United States.

This journal is © The Royal Society of Chemistry 2018