Molecular Fates of Organometallic Mercury in Human Brain

Abstract

Mercury is ubiquitous in the environment, with rising levels due to pollution and climate change being a current global concern. Many mercury compounds are notorious for their toxicity, with the potential of organometallic mercury compounds for devastating effects on the structures and functions of the central nervous system being of particular concern. Chronic exposure of human populations to low levels of methylmercury compounds occurs through consumption of fish and other seafood, although the health consequences, if any, from this exposure remain controversial. We have used high energy resolution fluorescence detected X-ray absorption spectroscopy (HERFD-XAS) to determine the speciation of mercury and selenium in human brain tissue. We show that the molecular fate of mercury differs dramatically between individuals who suffered acute organometallic mercury exposure (poisoning) and individuals with chronic low-level exposure from a diet rich in marine fish. For long-term low-level methylmercury exposure from fish consumption, mercury speciation in brain tissue shows methylmercury coordinated to an aliphatic thiolate, resembling the coordination environment observed in marine fish. In marked contrast, for short-term high-level exposure we observe the presence of biologically less-available mercuric selenide deposits, confirmed by X-ray fluorescence imaging, as well as mercury(II)-bis-thiolate complexes, which may be signatures of severe poisoning in humans. These differences between low-level and high-level exposures challenge the relevance of studies involving acute exposure as a proxy for low-level chronic exposure.

Graphical Abstract

Introduction

Many of the chemical compounds of mercury are poisons, with some being more toxic than those of any other non-radioactive heavy element. Since biogeochemical cycling of mercury is a natural part of our environment, human exposure to mercury is inevitable at some level. Recent studies suggest that mercury pollution contributes significantly to ongoing environmental increases.^1,2 Mercury is listed by the World Health Organization amongst its top ten chemical health concerns³ and more than 100 nations have signed the Minamata Convention which is aimed at reducing anthropogenic mercury emissions.⁴

The toxic properties of inorganic and organometallic forms of mercury are remarkably different^5–7 with the detailed toxicology of each depending upon the exact chemical speciation.⁸ The serious consequences of poisoning with organometallic mercury compounds were demonstrated by large-scale acute poisonings that occurred in Japan and Iraq.⁹ These incidents revealed a debilitating neurotoxicity, which was particularly pronounced during brain development in the fetus and young child.^5,9–12 Organometallic mercury species readily cross both the placental and the blood brain barrier resulting in brain mercury levels equivalent to those found in adults,¹³ and can result in microcephaly, cerebral palsy, seizures, intellectual disability, blindness, paralysis and other severe consequences in children exposed in utero.^5,6,9,10 The toxicity of organometallic mercury to the developing vertebrate brain was first recognized during the Minamata poisoning, where severely handicapped children were born to asymptomatic mothers.¹² The Minamata tragedy was long held as an example of methylmercury exposure, but recently the exact environmental species responsible has been questioned.¹⁴ It may not have been methylmercury, but a different species of organometallic mercury compound, containing the α-mercuriacetaldehyde group.¹⁴ This mercury species is not expected to occur in fish under normal environmental conditions. One of the most perplexing phenomena in the toxicology of organometallic mercury species is that there can be significant latency between administration and onset of symptoms in adults,¹⁵ which in humans can be as long as 150 days.¹⁶ The cause of this latency remains a major unanswered question in this field. A number of possible mechanisms have been suggested¹⁵ with recent work proposing a role for thyroid metabolism.¹⁷

Selenium is increasingly recognized as essential to the pathophysiology of mercury toxicity, since the biochemistry of selenium and mercury are inextricably linked.^17–24 Ralston and co-workers have proposed a selenium depletion hypothesis to explain mercury’s toxic effects.²⁵ The hypothesis is that inhibition of essential selenoenzymes is the primary mechanism of methylmercury toxicity. Ralston et al. proposed that consuming fish is not only safe but beneficial irrespective of the mercury levels, providing that the molar selenium content of the fish exceeds that of the mercury.²⁶ While methylmercury has recently been shown to bind selenocysteine of thioredoxin reductase in vitro,²⁷ some caution about the hypothesis of Ralston et al.²⁶ may be justified because the relationship between mercury and selenium is complex. Selenium can present either an antagonistic relationship, negating mercury toxicity,^17,18 or a synergistic relationship magnifying mercury’s toxic effects.¹⁷ Moreover, in experimental animals the effect has been shown to change not only with the chemical form of both elements but also the order in which mercury and selenium exposure occurs.¹⁷

In previous work we have used conventional X-ray absorption spectroscopy (XAS) to examine brain tissue from individuals having different types of mercury exposure, including tissues from chronic low-level and short-term high-level organometallic mercury poisoning.²⁰ Recently, high energy resolution X-ray fluorescence detected XAS (HERFD-XAS) has been exploited at both the Se K edge^14,28,29 and the Hg L_III edge,^{14,23,24,27,30} in order to provide significantly improved speciation in dilute systems. HERFD-XAS depends upon measurement of the X-ray fluorescence with narrower energy resolution than the natural fluorescence linewidth, the fluorescence lines of choice being the Se Kα1 and the Hg Lα1 for the Se K and Hg L_III edges, respectively. While sources of spectroscopic broadening in XAS experiments include the resolving power of the X-ray monochromator, with conventional hard-X-ray XAS measurements the most substantial source of broadening typically arises from the short lifetime of the Se 1s or Hg 2p_3/2 core-hole which is created by the primary X-ray photoexcitation. The energy resolution of HERFD-XAS is instead partly governed by the substantially longer lifetime of the hole created upon decay of an outer electron to fill the initial core hole, with concomitant emission of an X-ray fluorescent photon. HERFD-XAS probes the diagonal of what is known as the resonant inelastic X-ray scattering (RIXS) plane, and shows dramatically improved energy resolution over conventional XAS. It has been called high-resolution XAS (HR-XAS), although strictly speaking this is inaccurate unless there are no off-diagonal RIXS contributions.³¹ Figure 1 shows an example of Hg Lα1 HERFD-XAS in comparison with conventional Hg L_III-edge XAS for a 1 mM frozen solution of dimethylmercury in propane-2-ol. The dramatic improvement in spectroscopic resolution beyond conventional XAS that is afforded by HERFD-XAS is immediately apparent. In particular, the lowest energy transition presents as a poorly resolved shoulder in the XAS, but becomes a well-resolved prominent peak in the HERFD-XAS. Here we use the explicit fluorescence line nomenclature Se Kα1 and Hg Lα1 HERFD-XAS, rather than Se K-edge and Hg L_III-edge. HERFD-XAS of a given absorption edge can be recorded using different fluorescence lines, such as Swarbrick et al.³² recording Pb L_III HERFD-XAS using the minor Pb Lβ5 rather than the major Pb Lα1, whereas Lα1 can only be used to record the Hg Lα1 HERFD-XAS.

Figure 1.

Hg Lα1 HERFD-XAS (red line) and Hg L_III XAS (blue line) of dimethylmercury. The insets show the structure of dimethylmercury (left) and Hg Lα1 emission (right) with the experimental spectrum (blue points) measured at the non-resonant excitation energy of 12800 eV, together with a Lorentzian peak fit (gray line) showing the centroid energy selected for recording the HERFD-XAS.

Here we use Hg Lα1 and Se Kα1 HERFD-XAS to compare mercury and selenium speciation in human brain tissue samples for individuals exposed to mercury through a lifetime of fish consumption with those for individuals poisoned with organometallic mercury compounds. Our results give insights into the molecular fates of mercury under conditions of short-term high-level versus chronic low-level exposure, and have relevance to guidelines on the human consumption of marine fish.

Materials and Methods

Sample Preparation.

Chemicals and reagents were purchased from Sigma-Aldrich (Oakville, Ontario, Canada) or Alfa Aesar (Ward Hill, Massachusetts, USA) and were of the highest quality available. Solutions of standard compounds were prepared as previously described.^14,33,34 Tissue samples were loaded into 2 mm thick poly-acetal cuvettes closed with a metal-free polyimide adhesive tape window, and then flash-frozen by immersion into a partly-frozen isopentane at approximately 120 K. Samples were transported and stored at liquid nitrogen temperatures, or in a −80°C freezer, until data acquisition. Solid standards, prepared as finely ground powders mixed with boron nitride at <1 wt.% mercury or selenium, were packed into 1 mm-thick aluminum sample holders sealed on either side with polyimide adhesive tape.

X-ray Spectroscopy.

X-ray spectroscopy experiments were carried out at Stanford Synchrotron Radiation Lightsource (SSRL) with the SPEAR3 storage ring containing 500 mA at 3.0 GeV. Standard X-ray absorption spectroscopy (XAS) measurements used beamline 7–3 while high energy resolution fluorescence detected XAS (HERFD-XAS) experiments used beamline 6–2. On both beamlines in-hutch photon shutters were employed to prevent exposure of the sample to the X-ray beam when data were not being actively recorded. Samples were maintained at a temperature of 10 K using a helium flow cryostat (Oxford instruments, Abingdon, UK) and were mounted at an angle of 45° to the incident X-ray beam. The incident beam energies were calibrated with reference to the lowest-energy inflection of the transmittance signal of Hg-Sn amalgam foil for the Hg L_III edge, or a grey hexagonal selenium foil for the Se K-edge, values for which were assumed to be 12284.0 eV and 12658.0 eV, respectively.

On 7–3 a Si(220) double crystal monochromator was used for the incident beam, with harmonic rejection accomplished by detuning the monochromator to 60% of peak intensity. Incident and transmitted X-rays were monitored using nitrogen-filled gas ionization chambers, while X-ray absorption was measured as the Hg Lα1,2 or Se Kα1,2 fluorescence excitation spectrum using an array of 30 germanium detectors (Canberra Ltd. Meriden, Connecticut, USA).³⁵ In order to maintain detector count-rates in the pseudo-linear regime, Ga₂O₃ and elemental arsenic X-ray filters were employed for Hg L_III and Se K-edge XAS respectively, to preferentially absorb scattered radiation, together with silver Soller slits (EXAFS Co., Pioche, Nevada, USA) to reject filter fluorescence. Data were collected using the XAS Collect data acquisition software.³⁶

On 6–2 a Si(311) double crystal monochromator was used for the incident beam with harmonic rejection achieved through the cut-off energy (ca. 18 keV) of an upstream Rh-coated mirror. Incident and transmitted X-rays were monitored using gas ionization chambers filled with helium and nitrogen, respectively. High resolution X-ray fluorescence was measured by means of a 7-element array of spherically bent crystal analyzers, using Si(555) and Si(844) crystals³⁷ to select a very narrow energy band of the Hg Lα1 or Se Kα1 emission, respectively. The emission intensity was measured using a single-element silicon drift detector (Hitachi High-Technologies Science America Inc., Northridge, CA, USA). Aluminum filters upstream of the incident ion chamber were used to adjust the incident X-ray flux and to maintain the detector count-rates in the pseudo-linear regime for the most concentrated samples. Data were collected using the SPEC data acquisition software (Certified Scientific Software, Cambridge, Massachusetts, USA).

Data reduction and analysis were carried out as previously described^14,38 using the EXAFSPAK suite of computer programs.³⁹ Extended X-ray absorption fine structure (EXAFS) phase and amplitude functions were calculated using the program FEFF.^40,41 Data normalization to the absorption edge jump was carried out using the EXAFSPAK program BACKSUB,¹⁴ which employs tabulated X-ray cross sections.⁴² Linear combination analyses used the EXAFSPAK program DATFITCL, which provides speciation information through Levenburg-Marquardt refinement of fractions of a linear combination of standard spectra to experimental spectra. The program approximates the variance-covariance matrix as C = (J^TJ)⁻¹ where J is the Jacobian matrix, ignoring higher order terms. Estimated standard deviations (σ) and covariances were respectively computed from the diagonals and off-diagonals of C. DATFITCL maps standard spectra onto the same energy grid as the experimental spectrum using piecewise cubic spline interpolation, and provides refinable energy shifts for each component; unless otherwise stated these were fixed at zero eV in the present study. Quality of fit was estimated using the function $F={\sum }_{i=1}^n\left({\mu }_i-{\sum }_j^m{f}_j{\mu }_{j,i}\right)/\left(n-m\right)$, where μ_i are the normalized experimental data with a total of n incident energy points, μ_j,i are the normalized standard spectra with fraction f_j, and m is the number of components standard spectra included in the refinement. Convergence was said to be achieved when differences in F for consecutive iterations approached the machine precision. Upon achieving convergence, a component rejection algorithm was employed for components with σ_i⁄f_j > 1. If no components had σ_i⁄f_j > 1 then the refinement was judged to be complete, and the program exited. Up to one component was excluded in each cycle, that with the largest σ_i⁄f_j, and the refinement repeated until no more components were rejected. Fitting of Hg Lα1 HERFD-XAS spectra employed five initial standards; methylmercury-L-cysteineate, mercury(II)-bis-L-cysteineate, nano-HgSe or crystalline HgSe, β-HgS and methylmercury-L-selenocysteineate in the form of methylmercury-inhibited thioredoxin reductase. Inclusion of β-HgS did not improve any of the fits, and this was eliminated. For the Se Kα1 HERFD-XAS initial fits included L-selenomethionine, seleno-bis-S-gluthathione, selenocysteine-cysteine selanylsulfide, and nano-HgSe or crystalline HgSe. Because of high mutual correlations in the refinements for both Hg Lα1 and Se Kα1 HERFD-XAS, nano-HgSe and crystalline HgSe were not allowed to co-refine, but instead two different optimizations were carried out, with either nano-HgSe or crystalline HgSe, with the choice for the final refinement depending upon which gave the lower values for F in the final fit.

X-ray Fluorescence Imaging (XFI).

XFI experiments⁴³ were conducted on using the Advanced Photon Source Beamline 2-ID-D employing a Si(111) double crystal monochromator and using Fresnel zone plates (Xradia, Pleasanton, CA) to generate a micro-focused X-ray beam. An incident X-ray energy of 13450 eV was chosen to avoid excitation of Br Kα1,2 X-ray fluorescence from plastic components of the experimental setup. X-ray fluorescence was monitored using silicon-drift Vortex detectors (Hitachi High-Technologies Science America Inc., Northridge, CA, USA). XFI data reduction and analysis followed established practices.⁴³

Density Functional Theory (DFT) Calculations.

DFT geometry optimizations and energy calculations were carried out using DMol³ and Biovia Materials Studio Version 2018 R2^44,45 employing the meta-GGA approximation using the M11-L functional both for the potential during the self-consistent field procedure, and for the energy.⁴⁶ DMol³ double numerical basis sets included polarization functions for all atoms with all-electron relativistic core treatments. The effects of solvent were approximated using a COSMO field⁴⁷ with a dielectric constant of 78.54.

Summary of Tissue Samples.

Table 1 summarizes background information on the tissue samples examined in the present study. Samples A–D originated from individuals who were lifetime residents of the Republic of Seychelles, where marine fish forms a large part of the normal diet. Seychellois have more than ten-fold higher brain mercury than populations with lower dietary mercury,^13,48 with considerable (ca. 7-fold) observed variation between different individuals.^13,48 Samples A and B, and C and D were from the same individuals. None of these Seychelles subjects suffered any known adverse effects attributable to the mercury in their diets. Samples E and F were from a 48 year old professor at Dartmouth College who died 10 months after accidental skin contact with an unknown volume of dimethylmercury in her laboratory.^16,49 Sample G came from an adult who as an 8-year old had consumed pork from an animal that had inadvertently been fed grain treated with an organometallic mercury species.^50,51 She survived 21 years after the poisoning in a severely debilitated state before dying.⁵² Samples H and I are controls from subjects in Rochester, New York who had no known exposures to mercury.

Table 1.

Background Information on Brain Tissue Samples

Sample	Tissue	Gender	Age	Origin	Condition	Hg a	Se a
A	Cerebellum b	M	67	Seychelles	frozen	0.51	5.0
B	Cerebellum b	M	67	Seychelles	formalin	0.79	1.8
C	Cerebellum c	M	67	Seychelles	frozen	0.42	5.7
D	Cerebellum c	M	67	Seychelles	formalin	0.23	2.9
E	Cerebellum	F	48	Dartmouth	formalin	30.9	10.8–37.0
F	Cerebral Cortex	F	48	Dartmouth	paraffin	59.0	9.5–44.0
G	Cerebral Cortex	F	29	New Mexico	paraffin	19.7	23.0–70.0
H	Cerebellum	F	57	Rochester	formalin	—	3.6
I	Cerebellum	F	51	Rochester	paraffin	—	6.0

^a.Local levels of areas that were scanned spectroscopically, given in units of effective μM concentration, and estimated from the measured edge-jump.¹⁴

^b.Samples were from the same individual.

^c.Samples were from the same individual.

Results and Discussion

Molecular Nature of Organometallic Mercury Exposure.

The data presented in the current study derive from samples representing three different organometallic mercury exposure conditions, to related chemical compounds as elucidated below. These exposures represent chronic long-term low-level exposure from fish consumption, short-term high-level exposure with long-term survival and short-term high-level exposure with shorter-term survival, along with controls with no known exposure.

The chronic long-term low-level exposure from fish consumption (samples A – D) are derived from Seychellois who were exposed to methylmercury, probably in the form of methylmercury l-cysteinate,^53,54 from lifetime consumption of marine fish which form the primary dietary protein source for this population. Most marine fish typically contain small quantities of methylmercury bound to a thiolate donor.^53,54 Additional clinical and neuropathological information on these cases has been recently reviewed by O’Donoghue et al.¹³

The short-term high-level exposure with long-term survival occurred in the New Mexico poisoning (sample G), in which there was high level exposure from consuming pork contaminated with organometallic mercury compounds. She consumed pork from hogs inadvertently fed grain dusted with the fungicide Panogen^®, or possibly Ceresan^®. Panogen^® is methylmercury cyanoguanidine (also known as methylmercury dicyanamide) while Ceresan^® corresponds to a range of products containing either methylmercury or ethylmercury bound to p-toluenesulfonanilide.⁵² Methylmercury cyanoguanidine has never been structurally characterized, but likely structures are those in which the CH₃Hg–group is bound to one of the nitrogens of the guanidinium moiety, or to the nitrogen of the cyanide group. In either case the cyanoguanidine would be expected to dissociate from the methylmercury in vivo, and to be rapidly replaced by a thiolate donor from abundant extracellular or intracellular thiols, most probably l-cysteine, perhaps in the form of glutathione, to form compounds containing the CH₃Hg–S(l-Cys) linkage. For Ceresan^®, structural information on related compounds is available,⁵⁵ and as expected the metal is observed to be bound to the sulfonanilide nitrogen. Like methylmercury cyanoguanidine, the Hg–N linkage would also be expected to be replaced by available biological thiols in vivo, with the organomercurial forming covalent bonds with the thiolates. Given the possibility that an ethylmercury derivative was involved, it is possible that this case is due to poisoning by ethylmercury and not methylmercury. As we have previously discussed,⁵⁶ methylmercury and ethylmercury derivatives show comparable localizations in the larval zebrafish model,⁵⁷ with reasonably similar toxicological profiles⁵⁸ but with some differences.⁵⁹

The short-term high-level exposure with shorter-term survival (samples E and F) was a researcher preparing a ¹⁹⁹Hg NMR standard, following approved safety guidelines. She inadvertently spilled dimethylmercury upon the dorsum of her latex-gloved hand. An unknown volume of dimethylmercury passed through the latex and was absorbed. The spill was thought to comprise only a few drops,⁶⁰ but the high mercury content of the tissues¹⁶ suggests a higher dose. She received aggressive treatment with meso-dimercaptosuccininc acid, which acts to sequester Hg(II) as a monothiolate,⁶¹ but died approximately 10 months after exposure.¹⁶ Dimethylmercury is a highly volatile compound, and is also a hydrophobic entity with an octanol-water partition coefficient K_OW of 200,⁶² (compare K_OW for CH₃HgOH of 0.07⁶³). Because of this, once internalized, it would be expected to cross cell membranes with ease and to partition into lipid-rich tissues. Dimethylmercury is much more stable in water than other group 12 organometallics, but readily undergoes protonolysis in aqueous media to form mono-methylmercury derivatives and methane.^64,65 Experimental studies reported by Östlund⁶⁶ of mice injected with dimethylmercury labelled with ²⁰³Hg found that 80–90% of the dose administered was rapidly exhaled and there was no detectable dimethylmercury present after 16 hours. However, mono-methylmercury species were distributed throughout the mouse tissues.⁶⁶ Östlund used autoradiography to investigate ²⁰³Hg spatial distributions⁶⁶ and found that shortly after dimethylmercury injection the majority of the ²⁰³Hg was present in fatty tissues, with very little in the central nervous systems or in the placenta of pregnant mice. However, at four days after injection there was an increase of ²⁰³Hg in the central nervous system.⁶⁶ Interestingly, at four days, the fetal mice showed a marked uptake in the eye lens,⁶⁶ reminiscent of that observed for larval stage zebrafish.^8,57,67,68 Östlund concluded that “dimethyl mercury behaves as a chemically inert substance towards animal tissues” with toxic effects arising from degradation into other compounds.⁶⁶ This conclusion is chemically reasonable, as dimethylmercury itself is essentially coordinatively satisfied. It can form weak bonds with hard ligands,⁶⁹ but lacks the distinguishing affinity for chalcogenides that defines much of mercury chemistry. Overall, the evidence clearly indicates that protonolysis of one of the two C–Hg bonds of dimethylmercury had occurred in Östlund’s mice mostly between 4 and 16 hours after injection. The same process of protonoylsis to methylmercury species seems probable to have occurred in the researcher’s poisoning.

Molecular Forms of Mercury in Brain Following Different Exposures.

The two short-term high-level exposures we studied were initially to different organometallic mercury compounds, and differences in localization might be expected in the hours following exposure, and before dimethylmercury protonolysis. However, it seems probable that mono-methylmercury intoxication was ultimately responsible for the deaths in both high-level exposure cases. Figure 2 compares the Hg Lα1 and Se Kα1 HERFD-XAS of samples A+B, E, F and G. Figure 3 shows an example of linear combination analyses using HERFD-XAS spectra of well-characterized standard species. Linear-combination fitting of the data from the Seychelles samples A and B shows that the mercury in the chronic low-level exposure from fish consumption is present almost exclusively in an organometallic form, with the CH₃Hg moiety coordinated to sulfur and most likely a cysteine thiolate as a CH₃Hg–S(l-Cys). In contrast to samples A and B, the poisoning samples G, E and F show more complex mixtures of Hg species (Table 2) including, in addition to the species found in the Seychelles samples, species of mercuric selenide and inorganic mercury bound to two sulfurs. The significance of these additional species is discussed below. Dimethylmercury shows a highly distinctive spectrum (Figure 1); as expected, both samples E and F gave zero fractions of this species when it was included as a possible component in linear combination fits (not illustrated).

Figure 2.

Hg Lα1 (left) and Se Kα1 (right) HERFD-XAS of brain tissue samples. Sample AB is the weighted average of the two data sets A and B. Sample details are given in Table 1. The black lines show experimental data, and the red lines show linear combination analyses according to the details given in Table 2. Green vertical broken lines are included to emphasize differences in the spectral peak positions.

Figure 3.

Example linear combination analyses for brain tissue sample F (Table 1), showing Hg Lα1 (top) and Se Kα1 (bottom) HERFD-XAS. Points indicate experimental measurements and the black lines the best fits. The colored lines beneath indicate the spectra of individual components, scaled according to their contributions. In both Hg and Se plots the red lines indicate HgSe. In the Hg plot the green and blue lines indicate mercury(II)-bis-thiolate and methylmercury-thiolate, respectively (Table 2). In the Se plot the green, blue, orange and gray indicate selenium-bis-S-glutathione, l-selenomethionine, selenocysteine-cysteine selanylsulfide, and oxidized Se, respectively (Table 2).

Table 2.

Hg and Se Linear Combination Speciation Analysis

	% mercury species
Sample	HgSe	RS–Hg–SR	RS–Hg–CH3	F×103
A+B	–	–	100 ± 1	6.23
E	45 ± 2 nano	29 ± 3	26 ± 2	0.40
F	45 ± 2 solid	49 ± 2	33 ± 2	0.31
G	61 ± 5 nano	34 ± 7	5 ± 4	1.55
	% selenium species
Sample	HgSe	R–Se–R’	RS–Se–SR	RS–Se–SR	oxidized Se	F×103
A	–	48 ± 3	34 ± 4	15 ± 8	3 ± 2	2.24
B	–	44 ± 1	25 ± 1	21 ± 3	10 ± 1	3.56
C	–	32 ± 2	24 ± 2	37 ± 4	7 ± 1	7.13
D	–	26 ± 2	7 ± 2	58 ± 3	9 ± 1	6.36
E	57 ± 2 cryst.	17 ± 1	11 ± 1	9 ± 3	6 ± 1	1.82
F	59 ± 1 nano	20 ± 1	11 ± 1	7 ± 1	3 ± 1	0.46
G	49 ± 2 cryst.	26 ± 1	21 ± 1	–	4 ± 1	2.73
H	–	67 ± 3	16 ± 2	–	17 ± 2	9.73
I	–	73 ± 4	6 ± 4	14 ± 8	7 ± 2	7.12

Percent contributions of each component in the fit, ± estimated standard deviations obtained from the diagonal of the variance-covariance matrix. For HgSe, both crystalline (cryst.) and nano-particulate (nano) forms were separately tested in fits of each sample, with the form giving the best fit presented in the table, as indicated adjacent to the value. Additional standard spectra used were as follows: RS–Hg–SR, mercury(II)-bis-l-cysteinate; RS–Hg–CH₃, methylmercury-l-cysteinate; R–Se–R’, l-selenomethionine; RS–Se–SR, selenium-bis-S-glutathione; R–Se–SR, selenocysteine-cysteine selanylsulfide; oxidized Se, [(CH₃)Se]⁺ as a generic oxidized selenium form, with floating energy shift in refinement.

Mercuric selenide species.

For the high-level exposure samples G, E and F, both the Hg Lα1 and Se Kα1 HERFD-XAS of the tissues examined clearly showed the presence of different proportions of mercuric selenide (HgSe) which has distinctive Hg Lα1 and Se Kα1 HERFD-XAS. We have previously reported conventional XAS spectra for both crystalline and nano-particulate mercuric selenide (nano-HgSe), which show distinct though related XAS spectra.^18,20 Manceau and co-workers²³ have also recently reported the Hg Lα1 HERFD-XAS of nano-HgSe, which is in reasonable agreement with our data. Figure 4 compares the Se Kα1 and Hg Lα1 HERFD-XAS spectra of nano-HgSe with the conventional XAS spectra of the same samples. As expected, the HERFD-XAS spectra appear as a better-resolved version of the conventional XAS, with distinctly similar spectra for crystalline and nano-particulate HgSe. However, they differ in the amplitudes of the post-edge structure, which is more pronounced for the crystalline material. HgSe occurs mineralogically as tiemannite,⁷⁰ which has the zincblende structure. The wurtzite structure may plausibly occur in small particles,⁷¹ although this has never been experimentally observed.⁷² Both zincblende and wurtzite show four-coordinate Hg and four-coordinate Se, and are expected to be difficult to distinguish by EXAFS.¹⁸ The core HgSe zincblende structure, with fused adamantane-type rings comprised of alternating Hg and Se separated by 2.63 Å,⁷⁰ has been duplicated in a number of small molecule clusters with pendant selenolates, such as Hg₁₀Se₄(SeR)₁₂, which can coalesce into clusters with larger cores.⁷³ Naturally-occurring tiemannite frequently occurs in combination with sulfur, as HgS_xSe_1–x, with nearly the entire range of x from β-HgS (x=1) to HgSe (x=0) having been observed.⁷⁴ Crystalline HgSe deposits have previously been identified in whale liver,⁷⁵ while we have demonstrated mixed chalcogenide HgS_xSe_1–x nano-deposits in larval stage zebrafish treated with mercury compounds.^17,21 Our standard solution preparations of nano-HgSe stabilized with external glutathione linkages comprise black solutions.¹⁸ They are approximately 100 atoms each of Hg and Se, with a diameter of about 20 Å,¹⁸ and are similar to the preparations of others.⁷²

Figure 4.

Comparison of HERFD-XAS and conventional XAS of nano-particulate and crystalline HgSe, for mercury (left) and selenium (right). The top pairs of spectra are the HERFD-XAS, and the bottom pairs of spectra the conventional XAS.

Figure 5 shows high-resolution XFI data on an 8 μm thick section of cerebellum from the New Mexico poisoning case. The molecular layer to the top right of the section is a region essentially devoid of cells, consisting predominantly of dendrites and axons derived from cells in other layers. The granular layer is to the bottom left and contains surviving granule cells with the cell nuclei clearly visible via their high phosphorus content. No Purkinje cells are apparent in this section. The correlation plot (Figure 5, inset) shows a 1.08:1.0 molar correspondence between concentrations of mercury and selenium, close to equimolar, indicating the presence of HgSe in the sample. The slight excess of mercury over selenium might suggest the presence of HgS_xSe_1–x with x~0.07, in which sulfur structurally substitutes for selenium, and which may form when selenium is limiting.²¹ With an XFI spatial resolution of 500 nm, the observation of isolated hot-spots of Hg and Se that are approximately one pixel across suggests that clusters are smaller than this dimension. While Figure 5 shows clearly discernable cell nuclei, the types of cells involved in the HgSe cannot be determined. For example, astrocytes are likely to phagocytose HgSe particles, which would then accumulate in lysosomes within these cells. Hence, many of the larger HgSe co-accumulations of Figure 5 may represent a number of discrete, smaller, HgSe particles, rather than large single isolated particles. Figure 5 has discernable pixels containing both Hg and Se with areal densities⁴³ of 2 to 5 pmol·cm⁻² in both Hg and Se. If these are due to a single HgSe particle then this would correspond to a nano-particle comprised of approximately 3,000 to 7,000 atoms of both Hg and Se, which would be 60 to 80 Å in diameter.

Figure 5.

X-ray fluorescence imaging of an 8μm thick section of brain tissue corresponding to sample G. Maps of selected elements are shown, with maximum areal densities for P, S, Zn, Se and Hg of 42, 35, 13, 35 and 35 pmol·cm⁻², respectively. The minima in all the maps correspond to zero areal densities. The inset shows a correlation plot between the Hg and Se content of the image, with the best fit line shown (slope = 1.08).

Our previous conventional XAS EXAFS analysis²⁰ also gives clues as to the size of the HgSe clusters in brain tissues. The major structural differences between the core of a HgSe nano-particle compared with the bulk crystalline material, are expected to be related to disorder in both short and long-range interatomic distances. We have carried out density functional theory (DFT) geometry optimizations of nano-particulate HgSe structures. These calculations were necessarily restricted to smaller, computationally-tractable clusters of HgSe, [Hg₃₅Se₃₅], [Hg₇₉Se₇₉] and [Hg₁₃₅Se₁₄₀]¹⁰⁻, corresponding to approximately 13, 19 and 25 Å in diameter, starting with the crystal structure of bulk HgSe, and with no symmetry constraints or restraints imposed. Since the computational burden of DFT scales non-linearly with the size of the cluster (approximately as the cube of the number of basis functions), for our systems [Hg₁₃₅Se₁₄₀]¹⁰⁻ is close to the reasonable upper size-limit for a DFT geometry optimization calculation. We find that the central core of the geometry-optimized energy-minimized DFT computational model resembles the structure of the bulk phase, with a greater fraction of four-coordinate Hg and Se near the centre, and distortions in both Hg–Se bond-length and geometry increasing towards the surface of the cluster, where many Hg and Se are present in two-coordinate environments. Thus, for smaller particles, the disorder is predicted to be greater and the amplitudes of long-range EXAFS contributions would be damped due to the presence of substantial static Debye-Waller factors. With our computational models we calculate ${\sigma }_{\mathit{\text{stat}}}^2$ values for directly coordinated Hg–Se of 0.0154, 0.0141 and 0.0106 Ų for [Hg₃₅Se₃₅], [Hg₇₉Se₇₉] and [Hg₁₃₅Se₁₄₀]¹⁰⁻, respectively. All of these are very large, but the trend, of smaller values as cluster size increases, is expected to continue with larger clusters. For the vibrational component of disorder we compute a Hg–Se ${\sigma }_{vib}^2$ value of 0.0023 Ų from FEFF⁷⁶ and employing force constants derived from the DFT Hessian matrix.⁷⁷ Subtracting this from our previous value for σ² of 0.008 Ų for both Hg L_III and Se K-edge EXAFS,²⁰ gives an estimated ${\sigma }_{\mathit{\text{stat}}}^2$ values of ~0.006 Ų, suggesting a cluster size substantially larger than those of our DFT geometry optimizations. In addition to the first-shell EXAFS from Hg–Se bonds at ~2.61 Å, both crystalline and nano-particulate HgSe show outer shell features giving rise to peaks in the Fourier transform at ~4.3 Å.²⁰ These outer shell amplitudes are more sensitive to long-range order than the first shell EXAFS; our previous measurements of brain tissue samples²⁰ show outer shells that are more pronounced than for the approximately 20 Å [Hg₁₀₀Se₁₀₀] nano-particulate HgSe previously investigated,¹⁸ and are much less intense than those observed with crystalline HgSe. Thus, the likely size of the HgSe nano-particles in the brain tissue is between 80 and 20 Å, the upper boundary being calculated from the XFI areal density, see above. In agreement with this, we find that in some cases the nano-particulate HgSe standard fits the data better, whereas in others crystalline HgSe shows better fits (Table 2). The implications of HgSe presence in brain are discussed below.

Inorganic mercury bound to two thiolate donors.

The second major component revealed by linear combination analysis of the samples derived from acute exposures is Hg(II) bound to two thiolate donors (e.g. Hg(Cys)₂) (Table 2, RS–Hg–SR). This component was also detected using conventional XAS,²⁰ and indeed, it almost appears that it may be a signature of serious organometallic mercury intoxication that results in neuronal death and obvious neuropathological changes. The predominant modes of Hg(II) binding to thiolate donors are linear digonal, trigonal planar or pseudo tetrahedral.⁷⁸ A search of the Cambridge Structural Database,⁷⁸ excluding complexes with bridging thiolates, shows that four-coordinate Hg–S complexes are the most abundant (66 hits) followed by two-coordinate (56 hits) and then three-coordinate (9 hits), with characteristic Hg–S bond-lengths for two, three and four-coordination of 2.35, 2.47 and 2.54 Å, respectively. At neutral or acidic pH values the preferred geometry of Hg(II) binding by cysteine thiolates is linear digonal, whereas four coordinated pseudo-tetrahedral is preferred at higher pH.⁷⁹ There are thus a large number of possibilities for this pool of bis-thiolate bound Hg(II). We have previously shown that methylmercury compounds can coordinate the selenolate of the selenoenzyme thioredoxin reductase,⁸⁰ but thioredoxin itself has previously been suggested as a toxicological target of both inorganic and organometallic forms of mercury.^81–83 Both thioredoxin 1 and its mitochondrial counterpart thioredoxin 2 are abundant in human brain, with their active sites both containing a CXXC motif (CPYC and CGPC in thioredoxin 1 and 2, respectively). Vertebrate tissues also contain a number of other proteins important in redox homeostasis that share similar CXXC active sites, such as the glutaredoxins.⁸⁴ The configuration of such active sites can readily accommodate inorganic mercury as a linear digonal bis-thiolate complex. Consequently, it seems plausible that the Hg(SR)₂ present in the brain tissue samples from acute exposures might be derived from inhibited thioredoxin and related systems. Other proteins such as metallothioneins contain multiple cysteine residues,⁸⁵ which may also bind mercury with linear digonal coordination. The abundant intra and extra cellular thiol glutathione (γ-l-glutamyl-l-cysteinylglycine), may also play a role. Our EXAFS data of solutions of Hg(II)-bis-glutathione (not illustrated) clearly show binding with two sulfur donors with Hg–S bond-lengths of 2.36 Å, which is characteristic of linear digonal coordination.

Organometallic mercury bound to chalcogenide donors.

Of the two acute cases, linear combination analysis (Table 2) indicated that both samples from the high-level exposure with shorter-term survival case (E and F) contained thiolate-bound methylmercury, but at slightly different levels, 26 and 33% for cerebellum and cerebral cortex, respectively. In contrast, the sample from the higher-level exposure with long-term survival (case G) showed only relatively small amounts of this species, with fractions of the same order as the estimated standard deviation, and we conclude that the sample has no detectible thiolate-bound methylmercury. By contrast, the low-level, long-term exposure showed essentially all the mercury detected was present in this form.

We also examined the possibility of other mercury chemical species, in particular methylmercury l-selenocysteinate,⁸⁰ which when incorporated into the linear combination fitting schemes did show subtly improved fits in some cases. As we have previously noted,^14,30 the Hg Lα1 HERFD-XAS spectra of methylmercury l-selenocysteinate and methylmercury l-cysteinate are very similar, but the features of the Hg Lα1 HERFD-XAS for the selenium congener are shifted slightly to lower energy.^30,80 Such coordination might originate through inhibition of essential selenoenzymes. We have recently shown that methylmercury inhibits the essential selenoenzyme thioredoxin reductase through covalent binding of the selenolate of the active site selenocysteine residue.⁸⁰ Significantly, no improvements in the fitting of the low-level exposure samples were obtained when methylmercury l-selenocysteinate was incorporated in the linear combination analysis, and the improvements in fits for high-level exposure samples tended to be marginal (not illustrated), with the selenium congener replacing the methylmercury l-cysteinate. However, little correspondence was observed between fits of the Hg Lα1 and Se Kα1 HERFD-XAS, and we conclude that the presence of methylmercury l-selenocysteinate cannot be demonstrated with confidence.

Molecular Form of Selenium in Brain Tissues.

The Se Kα1 HERFD-XAS show distinctive spectra for the low-level mercury exposure from fish consumption, compared with the samples from the high-level short-term organometallic mercury exposures. In all cases the samples from the high-level short-term exposures showed significant HgSe, as described above, whereas this was not found to be a component for the low-level exposures from fish consumption. No other species of selenium bound to mercury could be identified with confidence in any of the samples, and the presence of HgSe may be another indicator of human poisoning. The remainder of the selenium was modelled as a range of organic selenium species which may represent essential selenium not bound to mercury, such as in selenoenzymes. However, as we note below, such organic selenium is expected to be more strongly affected by the preservation method.

Estimated selenium concentrations in the brain samples (Table 1) were substantially elevated for the high-level short-term exposures compared with the low-level exposures, the latter selenium levels being comparable with those of the controls with no known mercury exposure. It seems plausible that the excess selenium is largely attributable to biologically inactive HgSe species, while a pool of lower-level essential biologically active selenium is depleted. In addition, even though the selenium levels are comparable, the selenium species distribution changes between the controls and the Seychelles samples. In the former R–Se–R’ is predominant, whereas in the latter this is less prominent, with increased proportions of RS–Se–SR and R–Se–SR. These higher proportions of formally more oxidized S-coordinated Se relative to the reduced R–Se–R’ are also seen in the acute samples. This may suggest that selenium is perturbed in the Seychelles case, even though it is not directly binding mercury. Whether this perturbation is directly or indirectly a consequence of mercury exposure remains unclear.

The final linear combination analyses for both Hg Lα1 and Se Kα1 HERFD-XAS are shown in Figure 2, and summarized in Table 2. We have noted that the presence of Hg(II) bound to two thiolate donors may be a signal of Hg poisoning. A second major form that was detected is HgSe, but this should be benign because of its very low solubility (solubility product ~10^–59).⁸⁶ Any mercury present as HgSe will therefore be rendered bio-unavailable and effectively excluded from any involvement in biological processes. The third mercury form, methylmercury bound to one thiolate donor, was present in the shorter-term survival poisoning case, and less-pronounced in the long-term survival case, and comprised the only detectable mercury form in the low-level exposure case.

In-Vivo Origins of Mercuric Selenide.

As noted above, the relationship between selenium and mercury is complex. The toxicological antagonism between selenium and inorganic mercury in vertebrates is well-established at doses corresponding to acute exposures.^17,19 With inorganic mercury exposure, nano-particulate HgS_xSe_1–x forms with endogenous selenium.²¹ Mercury chloride’s toxic effects are dramatically reduced when selenium is pre-administered (as l-selenomethionine) to zebrafish larvae and there is formation of nano-particulate HgSe. This is not true if the selenium is administered after the mercury.¹⁷ However, with methylmercury chloride the administration of selenium (as l-selenomethionine) magnifies the methylmercury toxicity, irrespective of the order. The mechanism is unknown.¹⁷ Both inorganic mercury and methylmercury inhibit selenium transport. The mechanism may be by binding to the selenoprotein P transporter and thus depleting tissue selenium. If so, levels are not restored by subsequent l-selenomethionine treatment.¹⁷ Thyroid metabolism depends upon selenoenzymes and is severely disrupted by organometallic and inorganic mercury forms.¹⁷ The studies of Dolgova et al.¹⁷ are insightful but spanned only a few days. It is clear that conversion of methylmercury to inorganic forms occurs in the vertebrate brain, this process is relatively slow,⁸⁷ and the mechanisms by which it occurs are unclear. It is also known that methylmercury complexes of selenium can spontaneously decompose to form HgSe. In our hands, solutions of methylmercury l-selenocysteinate⁸⁸ are stable at high pH but at neutral pH rapidly precipitate HgSe, as identified by powder X-ray diffraction.²⁰ Even the crystalline solid of methylmercury l-selenocysteinate slowly decomposes to form HgSe on storage at room temperature. The mechanism of this reaction is thought to involve two methylmercury l-selenocysteinate molecules combining to form bis-methylmercuric selenide (CH₃Hg)₂Se,⁸⁹ and bis-l-alanylselenide.⁹⁰ (CH₃Hg)₂Se will spontaneously eliminate HgSe forming dimethylmercury.⁸⁹ As discussed above, in aqueous solution dimethylmercury will undergo spontaneous protonolysis to form methane and mono-methylmercury, which can then participate in further cycles of the reaction.⁹⁰ A related mechanism can operate with a mixture of methylmercury l-selenocysteinate and methylmercury l-cysteinate, which can form (CH₃Hg)₂Se and bis-l-alanylsulfide.⁹⁰ A similar mechanism has been suggested to occur with sulfur,¹⁴ forming (CH₃Hg)₂S,^91,92 which has been shown to form in the gut and in the tissues of mice.⁹² For this to occur in vivo, two molecules of methylmercury l-(seleno)cysteinate would need to come into close enough proximity to react chemically. In our studies of thioredoxin reductase and methylmercury l-selenocysteineate we have observed that the protein, unlike free methylmercury l-selenocysteineate, is stable in solution.⁸⁰ This is probably because the protein prevents individual methylmercury l-selenocysteineate groups from coming into sufficiently close proximity to react; hence in vivo the inhibited species would be expected to be stable until the protein was recycled via the proteasome. This bimolecular requirement may explain the slow observed rates of demethylation of methylmercury in vivo, especially when the levels of exposure are low.

Manceau and co-workers have suggested that methylmercury binding to multiple selenocysteine residues at the C-terminal domain of selenoprotein P may cause the mercury to demethylate.²³ Selenoprotein P is the most abundant source of selenium in humans and is a selenium transporter.⁹³ It exhibits multiple selenocysteine residues, 10 in the human protein, 9 of which are in the C-terminal one third of the protein.⁹⁴ Most of these selenocysteine residues are in proximity to either a cysteine residue or another selenocysteine, so that under the relatively oxidizing conditions present in plasma, formation of selenylsulfide (–Se–S–) or diselenides (–Se–Se–) seems likely. The protein is highly glycosylated, complicating structural studies. The suggestion of Manceau and co-workers of an adventitious function in demethylating methylmercury is chemically plausible. We would predict that multiple selenolate donors would increase the lability of the Hg–C bond, effectively catalyzing protonolysis. A similar process functions with the organomercury lyases, in which two cysteine residues coordinate the metal, and act to redistribute the electron density onto the methyl. This would facilitate protonolysis,⁹⁵ with the proton originating from an active site aspartate or serine residue. Selenium donors should be more effective in catalyzing such chemistry than sulfur,⁹⁶ although the reaction would need to rely on adventitious proton donors, and so might still be slow. The resulting inorganic complex could form a four-coordinate [Hg(SeCys)₄]²⁻ core within the protein, which Manceau and co-workers suggest. This might deplete essential selenium, and with time could result in HgSe deposits.^23,24 Ideally, our analyses should have included a [Hg(SeCys)₄]²⁻ standard, but as previously noted,³⁰ no structurally characterized mononuclear four coordinate mercury selenolate complexes have been reported to date. Some initial refinements did include both two and three-coordinate Hg(II) selenolate complexes, bis-(benzeneselenolato)mercury(II) and the tris(benzeneselenolato)mercury(II) anion, respectively,³⁰ but no convincing improvements in the quality of fit were obtained, possibly because these aryl complexes show different spectra from the biologically relevant alkyl species.³⁰ We note that the lack of a bona-fide [Hg(SeCys)₄]²⁻ standard is a limitation of the present study.

Mercury and Selenium Speciation Comparing Fixed and Cryopreserved Brain Tissues.

Our study depends upon the availability of human tissue samples, some of which are from irreproducible historical acute poisoning events and only available as fixed samples. Our previous speciation studies¹⁴ have been criticized as lacking relevance because of possible chemical changes caused by fixation, and in particular that the mercury content will be lost to the formalin fixative.⁹⁷ We have previously extensively discussed this,^14,98 but here we directly compare frozen and formalin fixed tissues taken from the same individual. Figure 6 compares Hg Lα1 and Se Kα1 HERFD-XAS of two pairs of cerebellum samples taken from two different individuals; one sample from each individual was fixed in formalin as per normal medical practice, while the other was frozen and stored at −80°C. Concentrations are low, particularly for mercury, but within the noise the observed spectra show no appreciable speciation differences for mercury. A conventional chemical analysis of tissue adjacent to that used for samples A and B supports the conclusion that total mercury is not depleted by fixation, with A and B giving 193 and 200 dry weight parts per billion, respectively. Our results are also in agreement with those of Davis et al.⁵² and Moffet et al.⁹⁹ who experimentally showed that fixation of rat brains in formalin did not result in a significant loss of mercury. The selenium Se Kα1 HERFD-XAS, however, shows distinctive differences between fixed and frozen samples. This finding is in agreement with our earlier discussions.^14,28,98 Commercial formalin fixative solutions consist of aqueous formaldehyde (HCHO) with some methanol (CH₃OH) which serves to inhibit formaldehyde polymerization that would otherwise precipitate as paraformaldehyde over time. Because of its low molecular weight and solubility in both lipid and aqueous phases, formaldehyde will diffuse rapidly throughout tissue samples, and will serve to fix the sample through cross-linking of proteins. Previous work has shown that fixation in formalin will dramatically alter the distributions of soluble tissue components such as Cl⁻ and K⁺.¹⁰⁰ As we have previously discussed²⁸ compounds of mercury are likely to be chemically stable to exposure to formaldehyde. Similarly, selenium present as the insoluble and inert HgSe is unlikely to be affected in either localization or chemical form by exposure to formaldehyde. However, organo-selenium compounds such as selenocysteine are expected to react with formaldehyde to form selanylmethanol derivatives, which in turn may cyclize by reacting with amino groups.²⁸ Using the edge jump of the HERFD-XAS as an indicator of concentration we find similar mercury levels in the fresh frozen and fixed samples, with the fixed tissue at 0.79 μM and the frozen slightly lower at 0.51 μM. The fact that the frozen sample showed lower concentrations than the fixed sample may be due to the fact that different brain regions (similar regions in different hemispheres) were sampled. This result is consistent both with our earlier discussions,^28,98 and with the findings of others indicating that mercury is not lost from tissues into formalin-based fixative solutions.^{52, 99}

Figure 6.

Comparison of Hg Lα1 (top; samples A and B) and Se Kα1 (bottom; samples C and D) HERFD-XAS for frozen (red) and fixed (blue) tissue. Green lines in both panels show difference spectra (fixed minus frozen). For mercury, but not selenium, the difference spectra show no spectral differences beyond the noise. This indicates detectible speciation differences for selenium but not for mercury.

The work presented here shows that the molecular-level fates of methylmercury compounds differ remarkably between chronic low-level long-term exposure from fish consumption and short-term high-level clinically-relevant poisoning. The chemical form of mercury in chronic low-level methylmercury exposure, from marine fish consumption with no evident adverse effects, remains essentially unchanged from that observed in marine fish.^53,54 With short-term high-level exposure, however, we find that the relationship with selenium is critical with formation of likely biologically inert nano-particulate HgSe deposits, and the occurrence of Hg(II)-bis-thiolate complexes, which may be a marker for actual methylmercury poisoning. One implication of these findings is that mechanistic studies involving acute exposures of animal models using higher levels of methylmercury species different than those that occur naturally in the normal human diet, may have little or no relevance to the vast majority of human exposures which are chronic low-level in nature, and derive from consumption of marine fish.

SYNOPSIS:

The molecular-level fates of organometallic mercury in brain substantially differ between poisoning and environmental exposure from a high-fish diet.