Methylmercury produced in upper oceans accumulates in deep Mariana Trench fauna

Monomethylmercury (MMHg) is a potent toxin that bioaccumulates and magnifies in marine food webs. Recent studies show abundant methylated Hg in deep oceans (>1000 m), yet its origin remains uncertain. Here we measured Hg isotope compositions in fauna and surface sediments from the Mariana Trench. The trench fauna at 7000–11000 m depth all have substantially positive mass-independent fractionation of odd Hg isotopes (odd-MIF), which can be generated only in the photic zone via MMHg photo-degradation. Given the identical odd-MIF in trench fauna and North Pacific upper ocean (<1000 m) biota MMHg, we suggest that the accumulated Hg in trench fauna originates exclusively from MMHg produced in upper oceans, which penetrates to depth by sorption to sinking particles. Our findings reveal little in-situ MMHg production in deep oceans and imply that anthropogenic Hg released at the Earth’s surface is much more pervasive across deep oceans than was previously thought. Monomethylmercury is a toxin that humans can be exposed to after consumption of seafood in which it has bioaccumulated. Here the authors show that amphipods in the deepest point of the global ocean contain monomethylmercury with surface origins, suggesting rapid sinking of this toxin on particles.

H umans are exposed to toxic monomethylmercury (MMHg) by the consumption of marine fish 1 . Ocean circulation, biogeochemistry, and marine ecology play a crucial role in MMHg exposure to humans. Natural and anthropogenic releases of Hg to air, land, and water eventually enter the oceans by atmospheric deposition and terrestrial discharge 2,3 . Within the oceans, inorganic Hg(II) (IHg(II)) is reduced to gaseous Hg(0) through biotic and abiotic processes, but also transformed to its methylated forms, MMHg and dimethyl-Hg (DMHg), potentially by anaerobic bacteria 4,5 . Particulate organic matter (POM) scavenges IHg(II) in surface oceans and releases it at depth 6,7 . This so-called biological pump results in a macronutrient-like vertical distribution of total Hg concentrations, with low values in epipelagic waters of the surface ocean (0-100 m), increasing through oxygen-depleted mesopelagic waters of the intermediate ocean (100-1000 m), and remaining high in deep ocean waters (>1000 m) 2,[8][9][10][11] .
MMHg makes up a large fraction of methylated Hg in pelagic marine waters, and its levels and chemical speciation determine the burden available to marine food webs [10][11][12][13] . Investigating where and how MMHg is produced in oceans is critically important to understand global Hg cycling and human exposure. At present, MMHg is thought to be produced in situ in the oxygen-depleted zone of mesopelagic waters during microbial remineralization of sinking POM derived from phytoplankton primary production in the photic zone 9,12,[14][15][16] . A fraction of mesopelagic MMHg can be transported to surface oceans via diffusion and upwelling, and in situ MMHg production in oxygenated, epipelagic waters has also been demonstrated 8,10,[17][18][19][20] . Because MMHg in the epipelagic waters is readily photo-degraded, a depletion of MMHg concentrations is observed in surface oceans 21,22 . Mesopelagic MMHg can penetrate into deep oceans via downwelling or via complexation to or incorporation in sinking POM (i.e., biological pump). However, field data and models indicate that the downward transport of mesopelagic MMHg via downwelling and sinking POM is possibly limited, if not negligible 12,23,24 . Yet, available data, show abundant MMHg and DMHg in deep oceans 8,12,13 , implying that in situ Hg methylation and/or di-methylation in deep oceans is plausible. Due to sparse incubation and measurement of deep waters, the origin (surface or deep), lifetime and fate of deep ocean MMHg and DMHg are therefore largely unknown.
Hg stable isotopes exhibit mass-dependent fractionation (MDF, represented by δ 202 Hg) in all investigated abiotic and biotic Hg transformations, and large mass-independent fractionation of odd-mass number isotopes (odd-MIF, represented by Δ 199 Hg and Δ 201 Hg) in abiotic, photochemical Hg transformations 25,26 . Significant even-mass number MIF (even-MIF, represented by Δ 200 Hg and Δ 204 Hg) has also been observed primarily in atmospheric precipitations, which is proposed to be related to photochemical oxidation of gaseous Hg(0) in the upper atmosphere 27,28 . Due to lack of significant Hg isotope fractionation during accumulation and trophic transfer of MMHg in aquatic food chains 29,30 , Hg isotope compositions of marine biota have demonstrated great potential in tracing the sources, production and degradation of MMHg in oceans 16,31,32 .
The hadal zone of oceans hosts a great diversity of endemic fauna, potentially serving as useful vectors to constrain MMHg dynamics and sources in deep oceans. Here, we measured concentrations and isotope compositions of Hg in endemic amphipods (mostly Hirondellea gigas) at water depths of 7000-11,000 m below the surface (mbs) and in benthic sediments at water depths of 5500-9200 mbs from the oligotrophic Mariana Trench in the North Pacific Ocean (NPO, 11.5°N, 142.5°E). Two amphipod samples (Hirondellea gigas) and one snailfish sample at water depths of~8000 mbs from the nearby Yap Trench (9.5°N, 138.5°E) were also collected and measured ( Fig. 1 and Supplementary  Fig. 1). We find that both MDF and MIF of trench fauna are comparable to those of upper NPO biota MMHg, suggesting that the accumulated Hg in trench fauna originates from MMHg produced in upper oceans.

Results
Total Hg and MMHg concentrations. Total Hg concentrations in trench amphipods have an average of 547 ± 230 ng g −1 (1 SD, 235 to 1070 ng g −1 , dry weight, Supplementary Table 1), which is comparable to their counterparts from the abyssal Arctic Ocean (250-1000 ng g −1 ) 33 , but considerably higher than benthic organisms in freshwater and coastal ecosystems (mostly <400 ng g −1 on average) 34,35 . The relatively high Hg concentrations in trench amphipods likely relate to their low tissue turnover rates and long life spans 36,37 . Total Hg concentrations of amphipods show no correlation with their depths of occurrence (R 2 = 0.00, P = 0.85), and an insignificantly positive correlation with their body length, a surrogate of life span (R 2 = 0.12, P = 0.08, Supplementary  Fig. 2a). MMHg concentrations were measured in nearly half of amphipods, and vary from 8 to 403 ng g −1 . The fraction of MMHg (MMHg%) is highly variable, from 2 to 59% of total Hg. A significantly negative correlation is observed between MMHg% and body length of amphipods (R 2 = 0.48, P = 0.02, Supplementary Fig. 2b). The snailfish that feeds on amphipods has a total Hg concentration of 970 ng g −1 and MMHg concentration of 809 ng g −1 , which indicates that its MMHg% (83%) is higher than the corresponding values of amphipods (Supplementary  Tables 1 and 2).

Discussion
The similar MIF signatures in deep fauna and surface fishes reveal important insight into hadal Hg provenance. First, Δ 200 Hg has thus far been used as a conservative tracer at the Earth's surface, and is thought to be generated during net Hg(0) oxidation in the upper atmosphere 28,38 . The identical, positive Δ 200 Hg in deep and surface fauna ( Supplementary Fig. 4) suggests that the IHg (II) precursor to MMHg in the deep and surface fauna is similar and likely of atmospheric Hg origin. Second, odd-MIF mainly accompanies photoreduction of IHg(II) and photodegradation of MMHg in aqueous environments 39,40 . The Δ 199 Hg/Δ 201 Hg ratio is a diagnostic of photochemical reaction types, with an experimental value of~1.3 for MMHg photodegradation and~1.0 for IHg(II) photoreduction 39 . The deep fauna all have positive odd-MIF values, and Δ 199 Hg and Δ 201 Hg are highly correlated along a linear slope of 1.15 ± 0.08 (1SE, Fig. 4a). This Δ 199 Hg/Δ 201 Hg slope is comparable to slopes (~1.2) reported for fishes from upper NPO and other coastal and open marine waters 16,31,32,41 , suggesting that the odd-MIF in the trench fauna is produced by similar surface ocean MMHg photodegradation processes. Experimental photodegradation of aqueous MMHg has also been suggested to induce a characteristic Δ 199 Hg/δ 202 Hg ratio of 2.43 ± 0.10 39 . A similar slope is observed in NPO fishes (2.13 ± 0.19, 1SE, Fig. 4b) 16,32 . Most of our trench fauna samples lie on the NPO fish slope, but do not show sufficient variation in both Δ 199 Hg and δ 202 Hg to confirm the Δ 199 Hg/δ 202 Hg slope as a diagnostic. Minor δ 202 Hg deviations in some fauna samples ( Fig. 4b) are likely caused by small amounts of non-photochemical, microbial degradation of MMHg, which results in no MIF but increases the δ 202 Hg of residual MMHg accumulated in fauna 42 .
The isotope compositions of total Hg in trench amphipods are representative of the isotope signatures of MMHg in the same amphipods, despite the relatively low MMHg% of 2-59%. The reason for this is that most of the IHg(II) in the amphipods is likely produced by in vivo degradation of MMHg rather than assimilation of IHg(II) from the ambient environments. This is supported by multiple lines of evidence.  (Fig. 3). The snailfish with 83% of MMHg % has nearly the same Hg isotope composition as trench amphipods which have varying and low MMHg%. If significant fraction of IHg(II) was to be assimilated from the surrounding environments, we would expect a more scattered Hg isotope distribution due to the mixing of IHg(II) and MMHg. We would expect that the trench fauna with high IHg(II)% to have low δ 202 Hg and Δ 199 Hg, because IHg(II) in the oceans has relatively low δ 202 Hg and Δ 199 Hg, as seen from our measured seafloor sediments and previously reported open ocean particles 32 and sediments 46,47 , and coastal seawater 48,49 . (4) Insignificant differences (<0.1‰) in Hg isotope compositions are observed between the whole tissues and the separated tissues (muscle, lipid, and gut contents) of amphipods, although their Hg concentrations and speciation vary significantly (Supplementary Table 2). This suggests that the internal distribution or in vivo transformation of Hg species do not significantly fractionate Hg isotopes between MMHg and IHg(II) in amphipods. In the following we discuss the origin of fauna MMHg, based on all available open ocean Hg isotope data, and make a number of deductions that constrain the deep ocean MMHg cycling.
IHg(II) in sinking marine particles of different sizes and depths in the NPO has a constant Δ 199 Hg of 0.16 ± 0.09‰ (1 SD) 32 , which is found back at depth in Mariana sediments (0.20 ± 0.07‰, 1 SD) (Fig. 2). Potential MMHg produced in situ in trench sediments or deep water column should therefore inherit the low Δ 199 Hg of~0.2‰. MMHg produced from hydrothermal vent fluids is expected to have near-zero Δ 199 Hg, because Hg of geological origin generally has near-zero MIF 50 . The trench amphipods are benthic scavengers, which mainly reside in the top of the sediments surrounding the hydrothermal vent fields, and feed on sinking particles. If MMHg is produced in the sediments, diffusing to the water column, or in hydrothermal fluids or in the deep water column, it would transfer the near-zero to very small Δ 199 Hg to the amphipods. The occurrence of significant, 1.47 ± 0.13‰ (1 SD), Δ 199 Hg in the deepest marine fauna on Earth suggests that this Δ 199 Hg is not produced in situ in the deep ocean environments. Instead, elevated MMHg Δ 199 Hg can only be acquired during photodegradation of MMHg in the euphotic zone (Fig. 2), since potential microbial methylation of IHg(II) and non-photodegradation of MMHg in deep waters would not induce MIF 42,51-53 . We therefore suggest that the trench MMHg is produced in the surface ocean and subsequently transport to depth before entering the deep ocean food web.
IHg(II) and MMHg can be transported from surface to intermediate and even deep oceans by sinking particles, and by subducting water masses during deep water formation 7,54 . The latter transport pathway is unlikely in the Mariana/Yap trench system because there is little deep water formation in the NPO 55 . The lifetime of MMHg with a surface ocean Δ 199 Hg of~1.5‰ would have to be thousands of years for the MMHg to arrive in the Mariana Trench by diffusion and advection, which is also unlikely 4 . The mixing of MMHg produced in epipelagic and mesopelagic waters has been previously proposed to account for the observed declines of δ 202 Hg and Δ 199 Hg values in NPO fishes with their feeding depths (Fig. 2) 16 . We observe that Δ 199 Hg values in trench fauna at 7000-11000 mbs are comparable to those of NPO fishes at 300-600 mbs (Fig. 2) 6,7 . A study on amphipods from the Mariana Trench and two other Pacific trenches revealed that the bomb 14 C signals of the surface ocean could rapidly penetrate into the deepest ocean via fast sinking of surface POM 56 . It is thus likely that, similar to IHg(II), the MMHg produced in the surface and intermediate waters above the trenches transfers to the deep ocean through sinking of surface POM as well. A conceptual model of Hg cycling in the Mariana/Yap trench system is shown in Fig. 5, illustrating the methylation and degradation of Hg, and how the surface MMHg odd-MIF propagates to deep food webs.
The δ 202 Hg values of trench fauna at 7000-11000 mbs (0.27 ± 0.14‰, 1 SD) are~0.3‰ higher than the weighted mean δ 202 Hg (−0.05 ± 0.30‰, 1 SD) of NPO surface/intermediate MMHg (see Methods). This magnitude of the δ 202 Hg shift is comparable to the~0.4‰ shift observed during microbial degradation of MMHg in culture, which increases the δ 202 Hg value of residual MMHg 42 . Thus, microbial degradation of MMHg likely occur during MMHg transport from the upper ocean to depth. The ensuing question is why MMHg is degraded, but not produced in the deep ocean that occupies more than 70% of the whole Ocean's volume. Both non-photochemical degradation and production of MMHg are predominantly mediated by microbes. The microbes that produce MMHg are thought to be mostly anaerobic bacteria with hgcAB genes, while the microbes that degrade MMHg could be either aerobic or anaerobic bacteria 57,58 . The microbes that are capable of MMHg degradation are thus more widespread in nature, including oceans. The deep marine waters are cold, nutrient-limited, and oxygen-rich 59 (Fig. 5), which might inhibit the methylation ability of anaerobic bacteria 60 . The bacteria in deep oceans that are capable to demethylate MMHg also likely exist in the guts of trench amphipods 44,45 , and cause the in vivo demethylation of MMHg we infer above.
The Hg isotope compositions of trench fauna clearly indicate that bioavailable MMHg in the planet's deepest ocean is produced in the upper ocean, and that consequently little MMHg is produced in the deep ocean. We use a binary mixing model based on Δ 199 Hg to estimate that 37-48% and 52-63% of MMHg in the trench fauna is derived from the surface and the intermediate ocean, respectively (see Methods). Several centuries of human activities have severely perturbed natural Hg cycling, elevating atmospheric Hg levels at the Earth's surface by a factor of 3-7 2,61 . Currently,~300 Mmol of anthropogenic Hg is estimated to reside in oceans, with two thirds in the upper ocean and the remainder subducting in North Atlantic Deep Water and Antarctic Bottom Water 2 . The tight MMHg linkage between trench fauna and the upper ocean suggest that anthropogenic Hg is much more pervasive across the world's oceans than was previously thought, and may reach the deepest ocean by fast particle sinking. Ongoing surface ocean warming was recently suggested to increase MMHg levels in the upper ocean 62 , which would likely rapidly propagate to the hadal ecosystems.
Our findings not only inform on the source of MMHg in the deep ocean, but also allow us to speculate about the production of DMHg in the deep ocean. Mean oceanic dissolved MMHg and DMHg profiles show similar concentration trends at depth throughout the world's oceans [10][11][12][13]24 . The limited global variation in MMHg and DMHg concentrations indicates slow transformations, supported by our finding that little IHg(II) methylation takes place in the deep NPO. The exact link between MMHg and DMHg in marine waters is presently unclear. State of the art marine Hg models exclude direct DMHg formation from IHg(II) 24,63 , supported by slow observed rates 17  Here, we find that deep ocean MMHg is supplied from the upper ocean by particle sinking rather than in situ microbial methylation. DMHg is a dissolved gas and is unlikely to be transported on particles. If in situ microbial MMHg production in the deep waters is limited, then microbial methylation of MMHg to form DMHg is likely also limited. This leads us to speculate that deep ocean DMHg is abiotically produced, possibly on particle surfaces as supported by recent laboratory experiments 64 .

Methods
Sampling. Endemic amphipods were captured by bait traps installed on the deepsea lander vehicles (@Tianya, @Yuanwei, @Wanquan) that were operated by the Chinese Academy of Sciences during several cruises between July 2016 and March 2017. The lander vehicles were deployed to the seafloor of Mariana (11.5°N, 142.5°E ) and Yap trenches (9.5°N, 138.5°E) in the NPO at depths of 7000-11,000 mbs for 12-24 h and monitored by cameras ( Fig. 1 and Supplementary Table 1). The bait was enclosed in an isolated mesh bag of a funnel trap, which allowed for odor plume diffusion and prevented consumption of the bait. The captured amphipods were frozen immediately at −80°C upon loading on deck. The amphipods mainly comprised of Hirondellea gigas with body lengths of 1-4.5 cm except one amphipod Alicella gigantea with a body length of 12 cm (Supplementary Fig. 1). After returning to laboratory, amphipods were rinsed with Milli-Q water before tissue separation, freeze-drying and homogenization. The amphipods mostly had empty gut contents, and were processed as whole tissues after removing the heads to avoid potential contamination by bait. To increase the sampling representativeness, several amphipod individuals of similar sizes were usually combined to one sample. Three large-size amphipod samples were separated for muscle, lipid and gut contents using pre-cleaned surgical scissors and tweezers to assess whether internal distribution or in vivo transformation of Hg species induce Hg isotope fractionation (Supplementary Table 2). We also trapped one snailfish that primarily fed on amphipods in the Yap Trench, for which we separated its muscle and processed it in the same way as amphipods. Ocean floor sediments at depths of 5500-9200 mbs were collected with a box corer that was launched by a geological winch during one cruise operated by the Chinese Academy of Sciences in 2016. Immediately after recovery, short sediment cores were collected by pushing PVC tubes (L = 80 cm, ID = 9 cm) into the sediment blocks. The top 0-6 cm layers of sediments from all the short cores were then subsampled and stored at −20°C before freeze-drying and homogenization.
Total Hg and MMHg concentration measurement. Total Hg concentrations in the homogenized freeze-dried samples were measured by a Lumex RA-915F Mercury Analyzer. The samples were combusted in the PYRO-915+ unit, and the evaporated Hg(0) was then purified and transferred to an analytical cell before measurement by differential Zeeman atomic absorption spectrometry (AAS). Total Hg concentrations in the digested sample solutions used for Hg isotope ratio analysis were measured by a Tekran 2600 cold vapor atomic fluorescence spectrometer (CV-AFS) according to US-EPA method 1631E. MMHg was determined in amphipod and snailfish samples by a Tekran 2700 gas chromatograph-CV-AFS after solvent extraction using KOH/CH 3 OH solution, and ethylation by NaBEt 4 in closed purge vessels 35 . The combined 2 SD analytic uncertainties of total Hg concentrations, as evaluated by periodically measured certified reference materials (CRMs, GBW07405, soil; DORM-4, dogfish muscle), were <5% for both Lumex RA-915F and Tekran 2600. The combined 2 SD analytic uncertainties for MMHg concentrations of DORM-4, were <15%. The relative standard deviations of sample/CRM duplicates were <5% for THg and <8% for MMHg concentrations.
Hg isotope ratio measurement. According to the measured total Hg concentrations in freeze-dried samples,~0.2-0.6 g of each fauna sample was digested using 4 mL double-distilled HNO 3 and 4 mL super-pure H 2 O 2 in a 30 mL Teflon lined vessel heated by a programmed microwave digestion system. Sediments were pre-concentrated for Hg using the combustion-trapping method 65 , and 10 mL of 40% double-distilled acid (2HNO 3 /1HCl, v/v) was used to trap the volatilized Hg (0). Procedural blanks and CRMs were processed with the samples in a same manner. The procedural blanks accounted for <1% of Hg mass in the samples, and the Hg digestion and preconcentration recoveries were in the range of 88-110% for both the samples and procedural CRMs. The processed sample solutions were diluted with Milli-Q water to Hg concentrations of 0.5-2 ng g −1 , and were measured for Hg isotope ratios by coupling a customized cold vapor generation system to multi-collector inductively coupled plasma mass spectrometry (MC-ICPMS, Nu Plasma 3D at Tianjin University, China). The typical sensitivity for 202 Hg was~2 V per ng g −1 Hg at a solution uptake rate of 0.8 mL min −1 . Instrumental mass bias was corrected by both an internal NIST 997 Tl standard solution (supplied via Aridus II desolvation nebulizer system) using the exponential fractionation law and NIST 3133 Hg standard-sample bracketing method. The bracketed NIST 3133 solutions were matched to the sample solutions within 5% in both acid matrix and Hg concentrations. The Faraday cups were positioned to simultaneously collect all seven Hg isotopes and two Tl isotopes. Acquisition time was 7 min (5 blocks, 20 cycles, 4.2 s of integration time) with 3 min of initial uptake time. Between samples,    (6) where microbial methylation produces MMHg that inherits the small odd-MIF of IHg(II) (7). The mixing of MMHg produced in epipelagic and mesopelagic waters in the intermediate ocean (8)  MIF value is denoted as Δ xxx Hg (‰, xxx = 199, 200, 201, 204), representing the difference between the measured δ xxx Hg value and that predicted from δ 202 Hg using a kinetic MDF law 66 : The mass-dependent scaling factor xxx β is 0.2520 for 199 Hg, 0.5024 for 200 Hg, 0.7520 for 201 Hg, and 1.4930 for 204 Hg.
Hg isotope ratios of secondary standard NIST 3177 solution and procedural CRMs (GBW07310, DORM-4) analyzed during different analytic sessions (Supplementary Table 4) were in agreement with those reported in previous studies 35,65,66 . The typical 2 SD analytic uncertainties of samples were estimated as the larger 2 SD uncertainties of Hg isotope ratios in NIST 3177 and GBW07310 (sediment) or DORM-4 (amphipod and snaifish). The 2 SD uncertainties of Hg isotope ratios in samples with replicate analyses were applied as the analytic uncertainties only when they were larger than the typical 2 SD analytic uncertainties.
Description of statistical analysis. All the statistical analyses were performed using OriginPro 9 except for mixing models by MatLab R2016b. Nonlinear exponential fitting was applied for Fig. 2 and linear fitting for other figures. The fitting curves are bounded by 95% confidence bands, and Pearson's R-Square and P-values are calculated by algorithms of the software. Two-side analysis of variance (ANOVA) was used to asses if the regression slope is significantly different from zero at the 0.05 level.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.