**Oxidative Stress Biomarkers in Erythrocytes of Captive Pre-Juvenile Loggerhead Turtles Following Acute Exposure to Methylmercury**

**Javier Hernández-Fernández 1,2,\*, Ellie Anne López-Barrera 3, Leonardo Mariño-Ramírez 4, Pilar Rodríguez-Becerra <sup>1</sup> and Andrés Pinzón-Velasco <sup>5</sup>**


Received: 21 April 2020; Accepted: 19 May 2020; Published: 22 May 2020

**Abstract:** This study describes the use of erythrocytes (RBCs) of loggerhead turtles as in vitro models for evaluating their toxicity to methylmercury. Blood samples of loggerhead turtles that were born in the Colombian Caribbean were used. The LC50 of RBCs to methylmercury was determined at 96 h using methylmercury concentrations of 0.5–100 mg L−1. Next, the viability of the RBCs and the activity of the enzymes superoxide dismutase (SOD), glutathione S-transferase (GST), and lipid peroxidation by malondialdehyde (MDA) at 6 and 12 h of exposure to acute concentrations of 0, 1, and 5 mg L−<sup>1</sup> were evaluated. The LC50 for loggerhead turtle RBCs was 8.32 mg L−1. The cell viability bioassay of RBCs exposed for 12 h only showed 100% cell viability. Increasing in vitro MeHg concentrations caused a corresponding increase in MDA concentration as well as decreases in the activities of SOD and GST. The RBCs represent an excellent model for ecotoxicological studies and SOD, GST, and MDA are biomarkers of environmental pollution and oxidative stress in loggerhead turtles. This was the first study conducted on loggerhead turtle where the response of RBCs to MeHg-induced oxidative stress is evaluated.

**Keywords:** Cytotoxicity; erythrocytes; methylmercury; malondialdehyde; in vitro; superoxide dismutase

#### **1. Introduction**

The loggerhead turtle (*Caretta caretta*) is a marine reptile that is distributed throughout the world; however, its population is in decline and in danger of extinction due to anthropic and environmental causes. The loggerhead turtle is considered to be a sentinel species of environmental pollution due to its longevity and high rate of biomagnification of contaminants [1]. These conditions allow for loggerhead turtles to be used as monitoring units for environmental pollutants. Loggerhead turtles reside in the Mediterranean Sea, and the Atlantic, Indian, and Pacific oceans [2]. In Colombia, the loggerhead turtle is distributed in the departments of Atlántico, San Andrés Archipelago, Providencia and Santa Catalina, La Guajira, Magdalena, and Sucre. It has been reported that some juvenile individuals of this species use the Colombian Caribbean as foraging area, where they feed on algae,

vascular plants, and invertebrates, such as crabs, mollusks, shrimps, starfish, and jellyfish [3]. It has been estimated that sea turtles can survive for very long periods (more than 50 years), time during which they store tremendously contaminating heavy metals in their tissues. Among them, mercury enters through the digestive tract and it is accumulated [4]. These sea turtles are sensitive to mercury storage due to their life history (prolonged anoxia, longevity, omnivore feeding, aquatic habitat) [4]. In Colombia, the loggerhead turtle is categorized as critically endangered according to the criteria of the IUCN. A 97% reduction of individuals was reported, from 200 nesting females that arrived per season to the Colombian Caribbean (in 1972) to six per year in the most recent assessment [5]. For this reason, it is important to preserve this specie in order to carry out studies that increase the survival, birth rates, reproductive success, and allow for collecting samples to monitor their health, and support future research [6].

Mercury produces lethal and sublethal effects in wildlife, causing different alterations (immunopathological, carcinogenic, teratogenic, neural, cardiovascular, renal, reproductive, and endocrine) [7–10]. To date, it is known that mercury, interacting with the sulfhydryl groups of cells, generates oxidative stress and alters calcium levels. However, the reaction that it experiences in its methylated state (MeHg), with hundreds of proteins and the changes that they cause, need to be investigated in greater depth in order to understand the adverse effects on embryonic and fetal development, and how it interferes with the normal functioning of the neurological, cardiovascular, and immune systems [11]. Mercury (Hg) has potential for biomagnification in the environment. Its high ecotoxicological risk is due to its biogeochemical behavior, allowing for it to associate with solids, suspended matter, and sediments [10]. In oceans and seas, mercury is present in its inorganic and organic forms, with the methylated form, being methylmercury, the most toxic one (CH3-Hg<sup>1</sup>+) (MeHg) [12]. MeHg is more toxic than inorganic mercury, and it prevails in the marine food chain. It typically comprises 75–99% of the total mercury (Hg-T) in organisms [13].

In the Colombian Caribbean, sediment Hg-T concentrations range between 20.6–552.7 μg L−1, with averages around 180–300 μg L-1 and peaks of up to 1300 μg L−<sup>1</sup> [14,15]. The threshold effect level of Hg-T is 130 μg L−<sup>1</sup> [15]. It is estimated that the global release of mercury from natural sources, anthropogenic sources, and reemission from past activities is 6000–11,000 tons/year [11]. In particular, anthropogenic sources are believed to account for approximately 80% of the annual Hg emission to the environment [16,17].

Methylmercury induces the production of reactive oxygen species (ROS), generating oxidative stress and altering the antioxidant defense system of the cells when it joins the sulfhydryl and selenotiol groups [18,19]. It also affects the permeability of phospholipid membranes, by producing porosity, stiffness, and lipid peroxidation [19]. Organisms have an antioxidant metabolic machinery made up of the enzymes glutathione-s-transferase (GST), glutathione peroxidase (GPX), superoxide dismutase (SOD), glutathione reductase (GR), and others, which neutralize free radicals and other ROS and prevent cell damage, to neutralize ROS that is produced by oxidative stress induced by the action of MeHg [20]. In red blood cells, glutathione in its reduced and oxidized states, together with the coenzyme NADPH, help to restore the Fe+<sup>2</sup> state in hemoglobin [21].

Erythrocytes (RBCs), unlike other cell types, have a high affinity for methylmercury [22]. In addition, they express a large amount of enzymatic and non-enzymatic antioxidants allowing for the monitoring of the antioxidant response [23]. RBCs are nucleated in birds, reptiles, fish, and amphibians. These blood cells have been used in environmental studies stress [24–30], and to evaluate the response of xenobiotics in mammals [31–33]. However, hardly any studies have been published where environmental pollutants have been quantified in sea turtles [34,35]. To date, there are no reports of the use of RBCs in specific bioassays in which oxidative stress markers were evaluated for exposure to environmental pollutants in sea turtles.

The present study aims to evaluate the use of erythrocytes (RBCs) of captive loggerhead turtles as a predictive in vitro model of oxidative stress that is produced by acute exposure to different concentrations of MeHg. The Median Lethal Concentration (LC50) and the viability of the RBCs were determined and, with this result, the concentrations of the bioassay were selected, evaluating SOD, GST, and Malonaldehyde (MDA) activity. These results provide a novel, non-invasive, cost-effective method for ecototoxicological studies on the negative effects of contamination by methylmercury and other molecules on loggerhead turtle populations.

#### **2. Materials and Methods**

In this study, loggerhead RBCs were used as an in vitro model for determining the oxidative stress of acute MeHg concentrations. The RBCs were subjected to a pre-adaptation process to the supplemented MEM culture medium. The LC50 of RBCs to methylmercury was determined at concentrations from 0.5 to 100 mg L−<sup>1</sup> for 96 h of incubation. The determined LC50 was taken into account to assess oxidative stress. The RBCs were subjected to acute concentrations of MeHg (0, 1, and 5 mg L−1) and incubated at 30 ◦C. It was evaluated whether the SOD, GST, and MDA activities at 6 and 12 h after MeHg exposure can be used as biomarkers of environmental contamination and oxidative stress in loggerhead turtles (Figure 1).

**Figure 1.** Experimental design. The loggerhead sea turtle erythrocytes were subjected to 1. Pre-adaptation in supplemented Minimum Essential Medium (MEM) medium, 2. Determination of LC50 in nine doses of methyl mercury, 3. Cell viability determination and 4. The erythrocytes were subjected to acute doses of methyl mercury and the activities of superoxide dismutase (SOD) and glutathione-s-transferase (GST) enzymes, and the Malonaldehyde (MDA) concentration were measured for spectrophotometry.

#### *2.1. Study Area, Collection and Analysis of Samples*

The peripheral blood samples were collected from five inmature, apparently healthy, captive-reared pre-juvenile loggerhead sea turtles. The turtles were maintained at ambient temperature (average 30 ◦C, minimum 27 ◦C) in an outdoor seawater pool at the El Rodadero Aquarium in Santa Marta, Colombia (11◦13 1.09" N, 74◦14 13.75" W), which was characterized by a semi-arid climate (Figure 2). The samples were obtained under a research permit that was granted by the Ministry of Environment and Territorial Development (#24 of 22 June 2012) and Contract for Access to Genetic Resources (#64 of 23 April 2013). We also have a permit for the collection of samples of Colombian biodiversity, issued to the Jorge Tadeo Lozano University (Resolution 1271 of 23 October 2014, IDB040I File). The turtles were visually assessed for any injuries to the flippers and carapace. Turtles were considered pre-juveniles due to their development status and Straight Carapace Length (SCL) <34 cm [36,37]. Turtles were marked, measured, and weighed (Table 1). Blood samples were extracted from the dorsal region of the cervical sinus, in accordance to previously described methodology by Dutton [38]. For the collection of peripheral blood, 4 mL sterile syringes and 2 mL tubes with sodium heparin were used. Standard blood tests were performed by Microvet Laboratories, Bogotá D. C., prior to the bioassay in this study, in order to assess the health status of the turtles (Table 1). The total mercury (Hg-T) in RBCs and seawater was analyzed by atomic absorption spectrophotometry at 254 nm (DMA-80 evo, Milestone-24010 Sorisole (BG), Italy). The 100 μL of sample, without any pre-treatment, were introduced into a sample boat, dried through a prescribed method in the DMA-80, and burned at 800 ◦C. Devolatilized elementary Hg (Hg vapors) was collected on a gold amalgamation trap and subsequently desorbed for quantification. DMA-80 reaches a detection limit as low as 0.001 nanograms of mercury. Analyses of duplicates and standard reference material were performed to control accuracy and precision. This quantification was carried out at Toxicology Laboratory, COLCAN, and Bogotá D. C.

**Figure 2.** Location of El Rodadero Aquarium (11◦13 01 N 74◦14 14 W), Santa Marta (Colombia), site for peripheral blood sampling site of the five individuals of loggerhead sea turtles.

#### *2.2. Experimental Procedures*

#### 2.2.1. Isolation and Pre-Adaptation of Loggerhead Sea Turtles Erythrocytes

All of the experiments were carried out in the laboratory of genetics and molecular biology of the Jorge Tadeo Lozano University. The RBCs of peripheral blood were isolated by centrifugation at 1500 rpm for 2 min at 4 ◦C. Plasma and buffy coat were removed. The RBCs packed in pellet were washed five times with phosphate buffered saline (PBS, 0.9% NaCl in 10 mM sodium phosphate buffer, pH 7.4) and suspended to prepare a 2% (v/v) cell suspension in Minimum Essential Medium (MEM) (Gibco, Grand Island, NY, USA), supplemented with sodium carbonate, 13% fetal bovine serum (FBS), and 1% antibiotics and antimycotics (100 mg/L penicillin, 10 mg/L streptomycin) (MEM-S). In this culture medium, the RBCs were pre-adapted for 96 h, incubated at 30 ◦C, and then used in the following bioassays.

#### 2.2.2. Determination of LC50 of Loggerhead Sea Turtles Erythrocytes

Solutions of methylmercury (II) chloride CH3HgCl (Sigma–Aldrich, St. Louis, MO, USA) were prepared in PBS buffer at concentrations of 0.5, 0.7, 1, 5, 10, 20, 50 y 100 mg L−<sup>1</sup> to determine the LC50. An experimental unit was used for each concentration, plus an additional control group, each having three replicates. Each replica was served in 1.5 mL Eppendorf tubes containing 2.6 <sup>×</sup> 106 RBCs/mL dissolved in 180 μL of MEM-S and 5 μL of the appropriate MeHg solution. The cells were incubated at 30 ◦C for 96 h. Distilled water was added to the control group and the bioassay was repeated three times. At 96 h, aliquots of 10 μL of RBCs were mixed with 10 μL of trypan blue solution (Sigma–Aldrich; Merck KGaA). Living and dead cells were counted in a Neubauer counting chamber (Zeiss, Oberkochen, Germany) while using the Leica DM750 microscope (Leica Microsystems, Heerbrugg, Switzerland) with 40× magnification. LC50 was calculated through a Probit analysis [39] using the "MASS" package [40] that was included within the free RWizard software [41].

#### 2.2.3. Determination of Cell Viability of Loggerhead Sea Turtles RBCs

Cell viability was evaluated at 0, 1, and 5 mg L−<sup>1</sup> of MeHg. These doses were selected based on the results of the LC50 test, which were also supported by previously published results [42]. The cell viability test was performed in triplicate, then, 2.6 <sup>×</sup> <sup>10</sup><sup>6</sup> RBCs/mL dissolved in 180 <sup>μ</sup>L of MEM-S and 5 μL of the appropriate MeHg were incubated at 30 ◦C for 6 and 12 h of exposure. Subsequently, 10 μL of 0.4% (v/v) trypan blue solution (Sigma-Aldrich; Merck KGaA) and 10 μL of RBCs suspension subjected to each of the MeHg concentrations were mixed in a polypropylene microtube and was incubated for 5 min at room temperature. 10 μL of the Trypan blue/RBC cell mixture were then placed on a Neubauer counting chamber (Zeiss, Oberkochen, Germany) and, subsequently, stained (non-viable) and unstained cells (viable) were counted while using a 40× magnification microscope (Leica DM750 microscope, Leica Microsystems, Heerbrugg, Switzerland).

#### *2.3. Oxidative Stress of Loggerhead Sea Turtles Erythrocytes*

SOD and GST, 2 <sup>×</sup> 106 RBCs/mL of loggerhead turtles exposed and not exposed to MeHg (0, 1 and 5 mg L<sup>−</sup>1) were used to assess oxidative stress markers. These were lysed with four volumes of ice-cold HPLC, and then centrifugated at 10,000 *g* for 15 min at 4 ◦C. The supernatant (RBCs lysate) was collected for assaying and stored at <sup>−</sup><sup>80</sup> ◦C. For the MDA assay, 2 <sup>×</sup> 107 RBCs/mL of loggerhead turtles exposed and not exposed to MeHg (0, 1, and 5 mg L<sup>−</sup>1) were used. The samples were centrifugated at 2500 rpm for 5 min at 4 ◦C, and then resuspended in 1 mL of MEM. RBCs were collected and stored at −80 ◦C until use.

The activity of Superoxide Dismutase (SOD) and Glutathione S-Transferases (GST), and the levels of Malonaldehyde (MDA) were determined in RBCs lysate using commercial kits (Superoxide Dismutase Assay Kit [Item No. 706002], Glutathione S-Transferase Assay Kit [Item No. 703302], and TBARS assay kit [Item No. 10009055], Cayman Chemical Co, Ann Arbor, MI, USA), following the manufacturer's protocol for each Kit.

The Superoxide Dismutase (SOD) activity measuring kit is based on a method in which xanthine and xanthine oxidase react with tetezolium salt (2-(-4-iodophenyl)-3-(4-nitrophenol)-5-phenyl tetrazolium chloride), generating superoxide radicals, which then react and form formazan red. The SOD activity was then measured by the degree of inhibition of this reaction. In a 96-well microplate, 10 μL of the Standard SOD enzyme (included in the kit, in AG wells) were served. In the rest of the microplate, 10 μL of previously exposed and unexposed lysed RBCs (0, 1, and 5 mg L−<sup>1</sup> of MeHg) were served (in triplicate), then, 200 μL of the diluted radical detector were added to all wells. To start the reaction, 20 μL of diluted xanthine oxidase were added to all wells, both standard and samples. Xanthine oxidase enzyme was not added for the control and the recorded readings were used to make the absorbance correction once the measurements were made. The plate was shaken for 30 min at room temperature and then the absorbance at 440–460 nm was measured in a plate reader (Elisa reader Expert 96 ASYS Hitech, Eugendorf, Austria). The average absorbance of each standard and sample was calculated and then corrected, subtracting the absorbance recorded in the control. The linearized rate (LR) of all standards was calculated, first dividing the absorbance of standard A by itself and then that of the other BG standards dividing the value recorded in A, by those that were recorded in each of the BG standards. This means LR for Std A = Abs Std A/Abs Std A; LR for Std B = Abs Std A/Abs Std B, and so on. The linear standard SOD rate (LR) was plotted against the final SOD activity (U/mL).

The SOD activity of the samples was calculated while using the equation that was obtained from the linear regression of the standard curve, substituting the linearized rate (LR) for each sample. A unit of SOD is defined as the amount of enzyme that is required to exhibit a 50% dismutation of the superoxide radical. SOD activity is standardized while using the coupled cytochrome c and xanthine oxidase assay. The results were presented in international units per milliliter (U/mL).

$$OD\left(\frac{\text{UI}}{mL}\right) = \left| \left( \text{sample } LR - y - \frac{intercept}{slope} \right) \times \left( \frac{0.23 \text{ mL}}{0.01 \text{ mL}} \right) \right| \times \text{Sample } dilution \tag{1}$$

The GST enzyme was quantified in triplicate in RBCs lysate. 150 μL of assay buffer, 20 μL of glutathione, and 20 μL of each of the previously exposed and unexposed lysed RBC samples (0, 1, and 5 mg L−<sup>1</sup> MeHg) were served in a 96-well microplate. As a negative control, 170 μL of test buffer and 20 μL of glutathione were used, and as a positive control, 150 μL of test buffer, 20 μL of glutathione, and 20 μL of reconstituted GST. The reaction was started by adding 10 μL of CDNB (included in the kit) to all wells and gently shaking the microplate. GST activity was read at 340 nm on a plate reader (Elisa reader Expert 96 ASYS Hitech, Eugendorf, Austria). The addition of the GST produced an increase of between 0.012 and 0.064/min. The change in absorbance over time was calculated while using the following equation, (where A represents the absorbance at 340 nm):

$$\frac{\Delta A}{\min} = \frac{A\left(Time2\right) - A\left(Time1\right)}{Time2\left(min\right) - Time1\left(min\right)}\tag{2}$$

GST activity was calculated using the following formula:

$$SSTActivity = \frac{\frac{\Delta A}{\text{mm}}}{\frac{1}{0.00803 \text{ }\mu\text{M}}} \times \left(\frac{0.2 \text{ mL}}{0.02 \text{ mL}}\right) \times \text{Sample dilution} = \frac{\frac{\text{mm} \text{L}}{\text{mm}}}{\text{mL}} \tag{3}$$

The reaction rate can be determined while using the CDNB extinction coefficient of 0.00503 μM<sup>−</sup>1. One unit of the enzyme will be conjugated to 1.0 mMol of CDNB with reduced glutathione for one minute at 25 ◦C. The actual extinction coefficient para CDNB at 340 nm is 0.0096 μM−<sup>1</sup> cm−<sup>1</sup> [43]. This value has been adjusted for the pathlenght of the solution of the well (0.524 cm).

The MDA concentration in RBCs was quantified in duplicate using the RBCs collected previously and stored at −80 ◦C. RBCs were sonicated at 36% amplitude with three pulses of 20 s each, using a sonicator (Sonics Vibra Cell, 750 W, Newtown, CT 06470, USA). RBCs thatwere resuspended in MEM and malonaldehyde standards (100 μL MDA, included in the kit) were mixed in 5 mL falcon tubes with 100 μL of SDS and 4 mL of the color reagent (included in the kit). All tubes were placed in a serological bath for 1 h at 100 ◦C, then incubated on ice for 10 min to stop the reaction. They were then spun at 1600 rpm for 10 min at 4 ◦C. 160 μL aliquots of each of the treatments and standards were added to a 96-well microplate. MDA measurement was performed spectrophotometrically at 530 nm, on a plate reader (Elisa Expert 96 reader (ASYS Hitech, Eugendorf, Austria). The measurement of MDA concentration was reported in μM.

Absorbances in all cases were corrected, as follows: (1) the average absorbance of each sample and the standard sample was calculated, (2) the absorbance value was subtracted from the standard A (0 μM) of itself and the other values (both standards and samples, and the corrected absorbance were found), (3) the corrected absorbance of the standard samples was graphed as a function of the MDA concentration, and (4) the MDA values for each sample of the standard curve were calculated using the equation:

$$\text{MDA}(\mu M) = \frac{(corrected\,\,above\,\,line\,) - (\,\,y-intercept\,\,)}{Slope\,\,of\,\,the\,\,line} \tag{4}$$

#### *2.4. Data Analysis*

Statistical analyzes for biochemical biomarker data (SOD, MDA, and GST concentration) in controls and treatments were subjected to a normality test and the homogeneity variance test using the Shapiro Wilk, Kolmogorov–Smirnov, and Levene tests, respectively. Significant differences between treatments and control were determined by an ANOVA test and then by multiple comparisons between treatments using the Tukey test. The sample size did not allow for stipulating significant differences between the treatments. The correlation coefficients of Pearson, Spearman, and Kendal were calculated to determine whether there is a correlation between the concentrations of MeHg to which the erythrocytes were exposed and the observed levels of GST, SOD and MDA at 6 and 12 h of exposure. To carry out the analyzes and figures, the R program was used [44] with RWizard software [41].

#### **3. Results**

#### *3.1. Blood Count and Health Status of Loggerhead Turtles*

The sampled individuals' sex could not be determined because of their immature development state. They had an average weight of 2.78 ± 0.29 kg, an average straight carapace length (SCL) of 29.4 ± 2.41 cm, and an average curve carapace length (CCL) of 26.2 ± 2.71 cm (Table 1). With these SCL, turtles are considered to be pre-juveniles [36,37]. The blood count was conducted prior to the MeHg bioassay in order to determine the health status of the turtles. All of the specimens presented an apparently good state of health that was corroborated by all of the obtained blood parameters (Table 2). The percentages of lymphocyte count, mean corpuscular volume, and the percentage of lymphocytes, eosinophils, and monocytes were similar to those previously reported [34]. The variations in concentration of the different parameters were within the accepted levels [34,35]. Nonetheless, the blood parameters between the different species of sea turtles, and between free-living or captive individuals of the same species are different and, in some cases, highly variable, depending on their geographical distribution, [4,45,46], as numerous additional intrinsic and extrinsic factors can impact blood composition.


**Table 1.** Body size of the f specimens of loggerhead sea turtles sampled for erythrocytes, identification number, live weight, Straight Carapace Length (SCL), and Curved Cap Width (ACL).

**Table 2.** Blood parameters of captive loggerhead sea turtles from El Rodadero Aquarium, Santa Marta. SD = standard deviation (n = 5). Reference values taken from Yang et al., and Rouselet et al. [47,48]. M.C.V.: medium corpuscular volume; H.C.M.: amount of hemoglobin per red blood cell; C.M.H.C.: mean corpuscular hemoglobin concentration.


#### *3.2. Quantification of Hg in Erythrocytes of Loggerhead Turtles*

The concentration of Hg in seawater samples and in 100 μL samples of RBCs was determined by atomic adsorption (see details above). The background Hg-T concentrations in control seawater was 0.0032 <sup>±</sup> 0.001 mg L−<sup>1</sup> (0.00323 ppm), whereas the background Hg-T concentrations in RBCs (before the bioassay) were 0.031 <sup>±</sup> 0.011 mg L−<sup>1</sup> on average. After the bioassay, at 12 h, the RBCs exposed to concentration of 1 mg L−<sup>1</sup> (1 ppm) had an average Hg-T concentration of 0.31 <sup>±</sup> 0.10 mg L−<sup>1</sup> and those exposed to 5 mg L−<sup>1</sup> (5 ppm) contained 2.18 <sup>±</sup> 0.77 mg L<sup>−</sup>1. These results demonstrate the high affinity of RBCs for Hg accumulation, due to their great capacity to retain it in cell membranes [49].

#### *3.3. Lethal Concentration (LC50) and Cell Viability of Erythrocytes of Loggerhead Turtles*

Loggerhead RBCs that were exposed to 50 and 100 mg L−<sup>1</sup> of MeHg for 96 h had 100% cell mortality (Figure 3). For all other exposure concentrations, cell viability ranged from 90% to 20%. From these data, the LC50 was obtained with a value of 8.32 mg L−<sup>1</sup> (8.32 ppm). Based on these results, it was decided to perform the bioassays for the determination of oxidative stress in the RBCs of loggerhead turtles at 0, 1, and 5 mg L−<sup>1</sup> of MeHg, and for incubation durations of 6 and 12 h.

In the bioassay of cell viability, the RBCs that were exposed to 0, 1, and 5 mg L−<sup>1</sup> of MeHg for 12 h showed 100% viability.

**Figure 3.** Cell viability of erythrocytes of loggerhead sea turtle against different concentrations of methylmercury, used for calculating lethal concentration (LC50). Error bars indicate mean and standard deviation. Treatments with significant differences (ANOVA, *p* < 0.05) indicated by asterisk (\*). Treatments with non-significant differences indicated by same letter (a, b, c, d, e).

#### *3.4. Bioassay and Oxidative Stress*

The SOD activity presented variations in each of the different treatments. At 6 h, the SOD activity of RBCs exposed to 1 mg L−<sup>1</sup> of MeHg was slightly more than double that in controls (1.56 <sup>±</sup> 0.58 vs. 0.72 <sup>±</sup> 0.24 U/mL of RBCs), and slightly greater than at 5 mg L-1 MeHg (1.20 <sup>±</sup> 0.52 U/mL), which, in turn, was higher than in the control, demonstrating that RBCs at 1 and 5 mg L−<sup>1</sup> of MeHg experienced oxidative stress (Figure 4A). After 12 h of exposure, SOD activity was higher in RBCs that were exposed to 5 mg L−<sup>1</sup> MeHg (1.39 <sup>±</sup> 0.45 U/mL) than in controls (0.82 <sup>±</sup> 0.24 U/mL) and those RBCs exposed to 1 mg L−<sup>1</sup> of MeHg (1.08 <sup>±</sup> 0.25 U/mL), again indicating oxidative stress. There were no statistically significant differences between RBCs' SOD activity for the two MeHg exposure concentrations (1 and 5 mg L<sup>−</sup>1) for either of the two exposure durations (6 and 12 h), no correlation between the SOD activity and MeHg concentration to which the RBCs were exposed after 6 h of exposure (in all cases *p* > 0.05).

The glutathione transferase activity of loggerhead turtle RBCs at 6 and 12 h of exposure showed variations in cells that were exposed to 1 and 5 mg L−<sup>1</sup> of MeHg with respect to the control. At 6 h, the GST activity was higher for the 1 and 5 mg L−<sup>1</sup> MeHg treatments as compared to the control (12.72 ± 5.41 and 11.93 ± 9.54 vs. 6.76 ± 4.93 nmol/min/mL of RBCs, respectively, demonstrating high response to the concentration and exposure time to MeHg. However, after 12 h of MeHg exposure, the GST activity was lower in RBCs exposed to 5 mg L−<sup>1</sup> MeHg when compared to those exposed to 1 mg L-1 MeHg and to the control (5.96 <sup>±</sup> 6.36 vs. 9.54 <sup>±</sup> 8.91 and 6.92 <sup>±</sup> 5.16 nmol/min/mL of RBCs), respectively. Nevertheless, the observed differences between MeHg treatments and the control were not statistically significant for any exposure concentrations or time point (in all cases *p* > 0.05) (Figure 4B).

**Figure 4.** Activities of (A) Superoxide dismutase (SOD) and (B) glutathione S-transferase (GST) in erythrocytes of loggerhead turtles exposed to methylmercury (MeHg) at 0, 1 and 5 mg L−<sup>1</sup> for 6 and 12 h (mean ± standard error of the mean, (n = 5). Differences between treatments were statistically non-significant in all cases (*p* > 0.05).

The malondialdehyde (MDA) levels increased with respect to the control with increasing MeHg concentration and exposure time. After 6 h, the MDA concentrations were 38.31 ± 20.92 and 46.65 <sup>±</sup> 13.42 <sup>μ</sup>M for 1 and 5 mg L−<sup>1</sup> MeHg, respectively, when compared to 27.58 <sup>±</sup> 0.42 <sup>μ</sup>M in the controls. After 12 h, the MDA concentrations increased to 49.15 ± 15.75 and 56.38 ± 13.97 μM of MDA for 1 and 5 mgL−<sup>1</sup> MeHg, respectively, whereas MDA in controls remained virtually unchanged (28.91 ± 9.35 μM, Figure 5). While the observed differences between treatments were not considered to be statistically significant (ANOVA, *p* > 0.05), a positive correlation was observed between the MDA concentrations and MeHg concentrations after 12 h of exposure time (r Spearman = 0.56, *p* < 0.05, Figure 6). This result is preliminary; therefore, it is necessary to carry out evaluations with a larger sample to improve the statistical significance of this analysis and verify this possible correlation.

**Figure 5.** Malondialdehyde (MDA) levels in control (untreated erythrocytes) and in erythrocytes exposed to MeHg at 1 and 5 mg L−<sup>1</sup> for 6 and 12 h (mean <sup>±</sup> standard error of the mean, n = 5). Differences between treatments were non-significant in all cases (*p* > 0.05).

**Figure 6.** Spearman's correlation between the variables MeHg concentration, exposure time and MDA levels.

#### **4. Discussion**

The present study is the first report on oxidative stress of RBCs of captive loggerhead turtles (*Caretta caretta*) that were exposed to methylmercury in bioassays. RBCs of sea turtles have not been used to perform in vitro bioassays until today. This study showed RBCs to be a sensitive, simple, and inexpensive model for performing bioassays for ecotoxicological evaluation of heavy metals, such as mercury and, potentially, other environmental pollutants. Furthermore, SOD, GST, and MDA were sensitive biomarkers of environmental pollution and oxidative stress in loggerhead turtles.

The blood count was conducted prior to the MeHg bioassay, in order to show the health status of the turtles. Plasma and hematological biochemistry analytes vary significantly between loggerhead turtle populations in the wild, due to environmental factors, temperature, type of diet, sex, size, pH, salinity, water quality, population density, development stage, maturity, weight, and season [50,51]. Additionally, the hematological response to man-made influences must be taken in consideration when reviewing previously published data. For these reasons, it is important to establish population-specific reference information [52]. In this study, all of the specimens presented an apparently good health showed by the blood parameters obtained (Table 2). The percentages of all white blood cells were within the previously described ranges for loggerhead turtles (Figure 7, [47]). Additionally, mean corpuscular volume, albumin, and plasma protein were similar to those reported previously [47,48,51–53]. The variations in concentration of the different parameters were within the accepted levels [53,54]. A simple analysis that is presented in Figure 7 shows that the percentages of white cells: heterophile, eosinophils, limphocythes, monocytes, and azurophiles are within the reference ranges [47] that are present in loggerhead turtles with an apparent good health. However, as we have seen, the size and weight of turtles are important variables that produce large variations in the concentrations of the different hematological analytes. In the study that was carried out by Yang et al. (2019) [47], the 191 analyzed individuals are 2.5 times larger than the individuals in our study according to the SCL range (54.8–100.8 cm vs. 25–32 cm). For this reason, the percentages of the heterophile and limphocythes in this study are approximately half of the maximum values (91.6% and 90%, respectively). In the same way, the eosinophils, monocytes, and azurophiles found are four, five, and three times under the maximum value (13.7%, 16.1%, and 8%, respectively). All of the cell types exceed the minimum range to present them within the normal ranges. These differences are apparently a result of the weight and size of the turtles, but, as described above, it might be due to other variables, such as captivity, feeding, and/or stress [50,51].

**Figure 7.** Comparison between the percentages relative to white cells in the blood counts for the loggerhead turtles (*Caretta caretta*) (ours) and the minimum and maximum ranges published by Yang et al. (2019).

This study used captive loggerhead turtles. For this reason, the background Hg-T in water and in turtle RBCs were determined prior to bioassays (0.0032 <sup>±</sup> 0.001 mg L−<sup>1</sup> and 0.031 <sup>±</sup> 0.011 mg L−1, respectively). The 10-fold concentration of baseline Hg-T in RBCs demonstrates how feeding is an important factor for the bioaccumulation and biomagnification of this metal. The diet of loggerhead turtles in their natural habitat is composed of cnidarians, gastropods, copepods, and other crustaceans, polychaetes, cephalopods, fish, algae, plant remains, fish eggs, and hydrozoans [55]. In this study, captive donor loggerhead turtles from Santa Martha Aquarium were fed with fish of species as machuelo (*Opisthonema oglinum*), macabí (*Elops saurus*) and picuda (*Sphyraena ensis*). It is presumed that fish diet had Hg concentrations like those that were previously reported for machuelo (*Opisthonema oglinum*) (0.11 μg L<sup>−</sup>1) [56].

In the present study, the average total mercury (Hg-T in RBCs) concentration of 0.031 <sup>±</sup> 0.011 mg L−<sup>1</sup> was obtained prior to bioassay and, after 12 h of bioassay with erythrocytes at 1 mg L−<sup>1</sup> MeHg, the concentration of Hg-T increased approximately 10 times (0.309 <sup>±</sup> 0.10 mg L−1). At 5 mg L−<sup>1</sup> of MeHg, the Hg-T concentrations increased to 2.183 <sup>±</sup> 0.77 mg L<sup>−</sup>1, being 70 times higher than controls and seven times higher than RBCs that were subjected to 1 mg L−<sup>1</sup> of MeHg.

Various studies have shown that the different dissolved forms of mercury have a high affinity for the sulfhydryl groups present in the cells, and that in their methylated state it has an affinity of over 90% for human RBCs, where it adheres in less than an hour [22,57,58]. Studies of RBCs of mammalian species have also shown that MeHg is evenly distributed between intra and extracellular hemoglobin [59,60]. In this study, these results demonstrate the high affinity of RBCs for Hg and their great capacity to retain it inside of the cell.

Moreover, the MeHg median lethal concentration (LC50) of RBCs for loggerhead turtles exposed in vitro was determined in this study, eight MeHg doses ranging from 0.5–100 mg L−1, in addition to seawater controls. At the two highest concentrations, there was 100% cell mortality (Figure 3). In the control group, mortality was 13% and it increased with the MeHg concentration. The LC50 for loggerhead turtle RBCs was 8.32 mg L−<sup>1</sup> after 96 h of incubation. To date, there are no published studies

on LC50 of MeHg on loggerhead sea turtle RBCs, so there are no literature values to compare these results with. Recently, the use of RBCs was reported in turtle studies of the red eared sliders species (*Trachemys callirostris*) to assess the genotoxic effect of environmental pollutants [61]. Cocci et al. [62] described the use of primary erythrocyte cell cultures as in vitro models for evaluating the effects of different endocrine disrupting chemicals. Finlayson et al. [55] established primary cell cultures from the intestine, heart, liver, ovary, and skin of green sea turtles (*Chelonia mydas*), and skin for loggerhead turtles and determined the mean effective concentration (EC50) with an average for all cell types of 18,4 μM for Hg<sup>+</sup>2, being skin tissue the most sensible. In this study, LC50 was found to be almost double this value (33.05 μM). They found significant differences between the different cell types evaluated. In this study, other cell types, RBCs, were evaluated in loggerhead turtles, also showing great differences. The differences in response to Hg contamination between cells and between different species are appreciable [63–65]. Cytotoxicity is generally due to non-specific molecular and cellular effects [55]. Some pollutants, such as Hg, have higher accumulation values in loggerhead turtles [66,67]. Loggerhead sea turtles' RBCs appear to have a higher sensitivity and affinity for Hg than skin cells and other tissues (as described above), which would produce the big difference between these two results. RBCs could represent a better in vitro model for toxicological studies than primary cell cultures according to this result, and due to the high sensitivity of the RBCs presented here. However, primary cell cultures can be used to specifically study the response of organs to environmental pollutants. Viability studies that were conducted in RBCs in marine fish species as *Sparus aurata* and *Dicentrarchus labrax* have reported LC50 of 5.25 and 7.25 mg L<sup>−</sup>1, respectively, in RBCs that were exposed to MeHg for 24 h [49]. This result is similar to the present study, although the exposure times are different, showing that the RBCs of the turtles are more resistant. We found that loggerhead turtle RBCs can survive at 4 ◦C for about eight weeks when stored with plasma. The determination of the LC50 allowed for selecting the appropriate concentrations of MeHg to be used in the bioassay (0, 1, and 5 mg L<sup>−</sup>1), so that the majority of RBCs remained alive until the end of the bioassay, as corroborated by cell viability close to 100%.

Studies on oxidative damage to wild animals from an ecological perspective have been conducted over the past 20 years [68], but knowledge on oxidative stress in sea turtles is scarce [1,55,69,70]. Further, the toxic level of Hg and others metals and their short- and long-term adverse effects are still unknown for sea turtles [1,55,71].

RBCs have an antioxidant defense system that is composed of the enzymes superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx) and molecules, such as reduced glutathione (GSH), tierrodoxins, alpha tocopherol, beta carotene, and vitamin C [72]. The exposure of RBCs to methylmercury induces oxidative stress, which destabilizes the lipid bilayer of the cell membrane and causes morphological and physiological changes in the cell [63,73]. RBCs increase or decrease the expression levels of catalase, superoxide dismutase, and glutathione enzymes in order to counteract stress. These enzymes have been monitored in sea turtles in blood and liver tissues [74,75] to determine the correlation between the bioaccumulation of metals and oxidative stress and, in this way, understand the possible mechanisms that sea turtles use to balance themselves. There are no reports of in vitro bioassays of RBCs of sea turtles that were exposed to MeHg where oxidative stress is evaluated.

In this study, where loggerhead RBCs exposed in concentrations of 5 mg L−<sup>1</sup> of MeHg, cells presented lipid peroxidation and reduction in the enzymatic activity of SOD and GST after 6 h of exposure. It seems that high concentrations of MeHg inhibit the expression of enzymes and, therefore, their activity [76–78]. These results are similar to those that were reported by Berntssen et al. [76] in brain, liver, and kidney tissues of fish that were exposed to 5 and 10 mg L−<sup>1</sup> MeHg. Greater susceptibility of brain tissue to oxidative stress was reported, which was represented by increased levels of lipid peroxidation (MDA) and ROS formation, and parallel reduction in the activity of SOD and GST enzymes in brain cells with increasing MeHg concentration (up to 10 mg L−<sup>1</sup> MeHg).

The increase of the enzymes Cu, Zn-SOD, and Mn-SOD or γ-glutamylcysteine synthetase is the first adaptive response that cells present when exposed to mercury compounds [79]. However, studies in mice that were exposed to Hg and MeHg have shown that doses of these compounds reduce glutation and SOD levels, respectively, in liver and brain tissues [80,81]. These experimental results agree with the results shown in this study, and they would indicate that loggerhead turtle RBCs are as susceptible to oxidative stress that is caused by MeHg exposure as the brain cells of the fish [76], mouse, and insect neuronal and liver cells [79], since these same trends were observed for the activity of the SOD and GST enzymes, and lipid peroxidation.

The SOD and GST enzymes make up the first line of cellular defense against oxidative stress, neutralizing oxygen and nitrogen reactive species and preventing lipid peroxidation [20]. When the levels of these enzymes and other non-enzymatic antioxidants, such as cysteine, which participates in the uptake and release of mercury [79], are not intracellularly sufficient or their production levels reach maximum ceilings, their activity is reduced and metals and free radicals that could not be extracellularly expelled cause the degradation of PUFAS membrane and, therefore, lipid peroxidation [60].

MeHg is known to generate lipid peroxidation [82]. Studies have shown that long-lived species have erythrocyte cell membranes that are more resistant to lipid peroxidation, which might vary with the age of the individual and, therefore, be a direct predictor of survival [83]. In the present study, it was observed that the mean levels of MDA increased at 6 and 12 h of exposure, even though observed differences between exposure concentrations were not statistically significant due to the high variance of results. Apparently, the low number of individuals (n = 5) could explain this result. A positive correlation was observed between the concentration of MeHg and the levels of MDA, r = 0.56, *p* < 0.05 after 12 h of exposure (Spearman). Studies with a higher n should be performed to verify this possible correlation due to the low number of samples used.

In this study, we found that an in vitro increase in MeHg concentration caused a corresponding increase in MDA concentration (LPO), as well as decreases in the activities of SOD and GST, result which is in agreement with previous studies [76–78,84]. Nonetheless, it is important to mention that there are other sensitive enzymes that respond to MeHg that should be studied as biomarkers of toxicity to mercury compounds. Peroxiredoxin and thioredoxins are found in all tissues and are involved in multiple functions that are associated with cellular maintenance and survival, including protein repair, regulation of the cell cycle, and cell signaling [85]. These enzymes could function in host defense against oxidative stress, modulate immune response, and control intracellular peroxide levels [86,87], as well as participate in the antioxidant defense system [88,89].

Furthermore, it was reported for Chinese soft-shelled turtle (*Pelodiscus sinensis*) that peroxiredoxins was implicated defense against microbial pathogens and oxidants and would provide information of functional mechanism of immunity [90]. Studies that were carried out with these enzymes in human cell lines have reported that mercury and MeHg bind to their thiol groups (–SH), decreasing activity and causing inactivation in most cases [91–93]. Peroxiredoxins and thioredoxins are present in blood and, thus, are very important for living beings. They are very sensitive to oxidation [94,95], but, thanks to their plasticity, they have the ability of making post-translational modifications that prevent them from oxidizing, maintaining the homeostatic balance of cells [96]. Thioredoxins and peroxiredoxins function structurally and they have been barely studied in turtles [90]. Some of these studies have focused on evaluating how brevotoxins affect thioredoxin, GST, and SOD levels [97]. Moreover, turtles' regulation of intracellular levels of GST and peroxiredoxins has been studied under oxidative stress caused by anoxia, freezing, and exposure to lipopolysaccharides [98–100]. Overall, studies to date conclude that peroxiredoxins and thioredoxins are important targets for the compounds of mercury and their inhibition is likely to explain the results beyond Hg-induced toxicity [101]. For this reason, the evaluation in future studies of TrxR and Trx in blood and plasma could provide a useful tool for the prediction of adverse effects that are induced by mercury compounds.

Additionally, the results indicate that mercury can inhibit the activities of many proteins/enzymes that are involved in combating oxidative stress in cells, including thioredoxin (Trx) (MeHg and HgCl2), glutaredoxin (Grx) (MeHg) [91,102], glutathione peroxidase (Gpx) (MeHg) [103], glutathione reductase (GR) (MeHg and HgCl2), thioredoxin reductase (TrxR) (MeHg and HgCl2) [91,104], superoxide

dismutase (SOD) (MeHg and HgCl2) [105], neuronal nitric oxide synthase (nNOS) (MeHg) [105], and Keap1 [106]. In addition, mercury can also alter the ratio of GSH/glutathione disulfide (GSSH) either by directly binding to GSH, depleting GSH, or decreasing the levels of Gpx [102,107,108].

For species, such as sea turtles, where obtaining in vivo effect values is an impossibility, in vitro data provides a useful opportunity for conduct research into chemical risk to these species. As the field of sea turtle toxicology continues to expand, the results from studies such as this provide crucial information for the future experiments and guiding management and conservation plans [55].

It is necessary to improve the state of knowledge about the variability of indicators of oxidative stress in loggerhead turtles and their relationship to concentrations of MeHg. For this purpose, the use of erythrocytes as biomarkers of environmental pollution is a promising, non-invasive, new tool for assessing the health status of this endangered species. This versatile and non-lethal monitoring tool might increase the usefulness of sea turtles as sentinel species, as indicators of the environmental health of their marine-coastal habitats in the light of spatial-temporal changes, and environmental degradation [69].

#### **5. Conclusions**

This is the first study that was conducted on loggerhead sea turtles, where the response of erythrocytes to MeHg-induced oxidative stress is evaluated. Erythrocytes resulted in being an excellent, simple, and inexpensive model to perform bioassays for ecotoxicological evaluation with heavy metals and other environmental pollutant molecules, specifically for MeHg in this study. Being loggerhead turtles considered a threatened or endangered species, RBCs represent an excellent ethical experimental in vitro model, allowing ecotoxicological studies to be carried out without the need to sacrifice animals. In addition, SOD, GST, and MDA were sensitive biomarkers of environmental pollution and oxidative stress in loggerhead turtles. However, these biomarkers presented high standard deviations due to the low number of individuals (n = 5) used. Future studies are suggested for carrying out further evaluations with a larger sample to improve the statistical significance of these results and confirm the benefits of these biomarkers as indicators of pollution and oxidative stress. It was observed that exposure to methylmercury in low concentrations induce protective redox defenses in RBCs, as seen from the induction of anti-oxidant enzyme SOD and GST activity. However, when the methylmercury concentration increases, these defenses are overcome and lipid peroxidation injury increases. Erythrocytes of loggerhead turtles in this study survived when kept in plasma in vacutainer tubes with Na-heparin at 4 ◦C for about eight weeks. This important characteristic allows for the use of these cells in sequential experiments.

**Author Contributions:** Conceptualization, J.H.-F. and E.A.L.-B.; methodology, J.H.-F., P.R.-B. and E.A.L.-B.; software, J.H.-F. and P.R.-B.; validation, J.H.-F., P.R.-B., L.M.-R., A.P.-V. and E.A.L.-B.; formal analysis, J.H.-F. and E.A.L.-B.; investigation, J.H.-F. and P.R.-B.; resources, J.H.-F., P.R.-B. and E.A.L.-B.; data curation, J.H.-F., P.R.-B., L.M.-R., A.P.-V. and E.A.L.-B.; writing—original draft preparation, J.H.-F. writing—review and editing, J.H.-F., P.R.-B., L.M.-R., A.P.-V. and E.A.L.-B.; supervision, E.A.L.-B., L.M.-R. and A.P.-V.; project administration, J.H.-F.; funding acquisition, J.H.-F., L.M.-R. and E.A.L.-B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Office of Research, Creation and Innovation of the Universidad Jorge Tadeo Lozano (340-07-10). Additional funding came from the Intramural Research Program of the National Institutes of Health (NIH, USA), National Library of Medicine (USA), National Center for Biotechnology Information (NCBI) ZIA LM082713–06. There was no additional external funding received for this study.

**Acknowledgments:** We are grateful to Rodadero Aquarium in Santa Marta for their collaboration in obtaining and providing samples of loggerhead for the development of this study. Samples were obtained under a research permit that was granted by the Ministry of Environment and Territorial Development (#24 of 22 June 2012) Contract for Access to Genetic Resources (#64 of 23 April 2013), and permit for the collection of samples of Colombian biodiversity, issued to the Jorge Tadeo Lozano University (Resolution 1271 of 23 October 2014, IDB040I File). The authors thank Michel Arhens for critical review of the manuscript.

**Conflicts of Interest:** The authors declare that they have no known competing financial interests or personal relationships that could have influenced the results reported in this paper.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Total Mercury in Soils and Sediments in the Vicinity of Abandoned Mercury Mine Area in Puerto Princesa City, Philippines**

#### **Jessie Samaniego \*, Cris Reven Gibaga, Alexandria Tanciongco and Rasty Rastrullo**

Department of Science and Technology-Philippine Nuclear Research Institute, Commonwealth Ave., Diliman, Quezon City 1101, Philippines; crlgibaga@pnri.dost.gov.ph (C.R.G.); amtanciongco@up.edu.ph (A.T.); rmrastrullo@gmail.com (R.R.)

**\*** Correspondence: josamaniego@pnri.dost.gov.ph

Received: 28 October 2019; Accepted: 18 December 2019; Published: 2 July 2020

**Abstract:** The abandoned mercury (Hg) mine area in Puerto Princesa City, Palawan was included in the list of abandoned and inactive mines in the Philippines which pose a high risk to human health and the environment, and require rehabilitation. The mine site, operated by Palawan Quicksilver Mines, Inc. (PQMI) from 1953 to 1976, is located approximately 3 km inland from Honda Bay coast and within the catchment of the Tagburos River, which is a local fishery and recreational area. In this study, total Hg levels in soils and sediments were measured to assess the possible release of Hg from the site into the surrounding natural environment. Results showed that total Hg (THg) concentrations found in soils (0.04–67.5 mg kg−1), mine waste calcines (52.7–924.2 mg kg−1), river sediments (1.8–119 mg kg<sup>−</sup>1), and marine sediments (0.04–12.7 mg kg−1) were elevated compared to the global background of 0.045–0.16 mg kg<sup>−</sup>1. The high concentrations of Hg in soils and river sediments were influenced by the different pathways for the release of Hg from its mine operations; while Hg in marine sediments was caused by the erosion of mine waste calcines near the pit lake, and calcines used to construct a wharf at the nearby Honda Bay. Mine wastes represent the largest source of Hg contamination in the area, due to the low efficiency of the recovery process during calcination. This work corresponds with the Minamata Convention on Mercury on the updated environmental assessments of abandoned Hg mines as potential source sites of mercury contamination.

**Keywords:** mercury; PQMI; Palawan; abandoned mines; mine wastes; sediments

#### **1. Introduction**

Mining has been an important economic activity in the Philippines, which started centuries ago. With a lack of environmental regulations until recent times, old mining operations have led to many abandoned mine sites, and consequently resulted in high heavy metal concentrations in soils and sediments in the vicinity. In the Philippines, there are more than 20 mine sites that are classified as inactive and abandoned, one of which is the Palawan Quicksilver Mines, Inc. (PQMI) in Puerto Princesa City, Palawan [1]. This former mine site is considered as one of the Hg waste hot spots of the world, because the Hg pollution is affecting the marine environment and the people living nearby [2].

Most of the mercury mining areas around the world have been investigated in relation to their Hg pollution in their local ecosystems, which pose potential risks to humans living nearby. In the past decades, risk assessments on soils and sediments contaminated with Hg from the mining activities have been studied in Hg mining sites in Spain [3,4], China [5,6], New Zealand [7], United States of America [8], and others. Similar patterns of high to extremely high (46,000 mg kg<sup>−</sup>1) Hg concentrations in mine waste calcines and runoff sediments were reported from these Hg mining sites, which pose potential contamination on soil, water, and air in the area [7].

In the Philippines, a naturally occurring deposit of Hg can be found only in Puerto Princesa City on the island of Palawan. It mostly occurs in cinnabar (HgS) ores hosted by opalite bodies deposited along fractures and weak zones in the bedrock [9,10]. From 1953 to 1976, PQMI mined the Hg deposit in the village areas of Sta. Lourdes and produced around 2900 metric tons of mercury [11]. Today, the abandoned open-pit mine is filled with water, and transformed into a pit lake located about 3 km west from Honda Bay coast (Figure 1). During the mining operations of PQMI, around 2,000,000 metric tons of cinnabar mine waste calcines were produced [11]. Part of those mine waste calcines were used to construct a 600 m by 50 m and 3 m thick wharf at the nearby Honda Bay to serve as port for mining operations from 1953 to 1976. The wharf is now used as port for tourist and fishing boats. The rest of the mine waste calcines were stocked northeast near the pit lake shore (Figure 2), which was bare and had little vegetation until now. For the past four decades, remediation work has not been carried out at this site. Today, the abandoned Hg mine area is in the list of abandoned and inactive mines in the Philippines which pose a high risk to human health and the environment, and which require rehabilitation [1].

Mercury is a highly toxic element that is found both naturally, and as an artificially introduced contaminant, in the environment. Mercury mine sites remain as significant sources of Hg pollution in the environment, due to the unconverted HgS ores left in the mine waste calcines. The calcines contain soluble Hg minerals such as montroydite, eglestonite, corderoite, and elemental Hg. Leaching of these calcines, which has been reported from several abandoned Hg mines, becomes a significant source of Hg if released into the environment [7,11–14]. These anthropogenic activities can create widespread Hg contamination in soils of abandoned mine areas [15], while bare and unvegetated soils can produce dust with particulate-bound Hg during summer, that may be inhaled by the residents. Exposed HgS and mine waste calcines pose a very serious health problem for the residents, as the absorbed Hg from HgS and calcines is mainly accumulated in the kidneys, resembling the disposition pattern of inorganic Hg [16]. Inorganic Hg found in HgS and calcines, may be moved along with the sediments and through water runoff. It can undergo methylation by microorganisms and convert it to methylmercury (MeHg) [17]. Methylmercury may bioaccumulate, and concentrate in the marine and freshwater food chain that might end up in human body [17,18]. Excess concentrations of Hg in the human body can cause serious health threats, like damage to the central nervous system, and is detrimental to fetal development [17–19].

Studies conducted in the abandoned PQMI mine area since 1994 show that the population in the affected areas have relatively high Hg concentrations in their hair and blood samples [20–22]. In 1994, health reports revealed significant elevation of blood Hg levels exceeding the previous recommended exposure level of 20 ppb in 12 out of the 43 residents examined, after their complaints of unusual symptoms (e.g., miscarriages, tooth loss, muscle weakness, paralysis, anemia, tremors, etc.) [21]. Recent reports from the Department of Health (DOH) and Department of Environment and Natural Resources (DENR) say that 33–40% of the 10,000 combined Tagburos and Sta. Lourdes residents have 'chronic mercury poisoning', from the exposure to mine tailings and ingestion of marine products with high Hg content [23]. However, a recent survey of the surface water quality of different water bodies in the vicinity of the PQMI mined area showed undetected Hg concentrations in the pit lake and river, and low Hg concentrations in coastal water near the wharf (0.001 mg L−1) and in hot spring water (≤0.0004 mg L<sup>−</sup>1) [1]. The very low Hg concentrations measured in the surface water suggest that there is low solubility and bioavailability of inorganic Hg from HgS, and mine waste calcines. For the past 20 years after the last known study on Hg contamination in the area, Hg from HgS and calcines might be carried to the river by erosion and surface runoff, and eventually deposited in Honda Bay where they become a source of contamination and enter the food chain.

The purpose of this study is to document the current status of Hg pollution in different sites surrounding the abandoned PQMI mine area, describe how legacy mining still affects the area, and how it has evolved since the last studies performed in the 1990s. To evaluate the potential Hg contamination in this area, Hg concentrations were measured in the PQMI pit lake, Tagburos River, Honda Bay, Honda Bay Wharf, mine waste calcines, soils (including agricultural soils) in the vicinity, and other rivers which are not tributaries of the Tagburos River. Assessments on the Hg concentration in fish and the extent of human exposure to Hg were not included in the study. The output of this study will serve as guide to the rehabilitation prioritization of the abandoned PQMI mine area. This study will also conform with the requirements of the Minamata Convention on Mercury, where Philippines is one of the signatories, and give updates on the current environmental conditions of possible point sources of Hg, such as abandoned Hg mining sites.

**Figure 1.** Map of the study site showing the rivers, pit lake, and Honda Bay, with a map of the Philippines and Palawan island (inset).

**Figure 2.** Photo of mine waste calcines stockpiled at the shore of the pit lake.

#### **2. Materials and Methods**

Analyses of total Hg concentrations in different soils and sediments in the vicinity of the abandoned PQMI mine area were conducted to determine the current Hg concentrations in different sub-areas, especially in Tagburos River and Honda Bay. Characterization studies on the Hg contamination in this area is very limited, with the last known published article in 1996 [21]. In 1999, sedimentation rate

in Honda Bay was estimated at 1 cm yr−<sup>1</sup> based on the 210Pb dating of core samples collected from Honda Bay [10]. With a wider scope of sampling area considered in this study, including other rivers and mountainous terrain nearby, assessment of the extent of Hg contamination in the abandoned mine area can be drawn.

#### *2.1. Study Area and Sampling Sites*

The study area, Puerto Princesa City, is located in the midsection of Palawan Island. It is bound to the east by the Sulu Sea, and to the west by the South China Sea (Figure 1). The area in Puerto Princesa City with reported human Hg risk encompasses two villages: Sta. Lourdes and Tagburos. The study site, located around 14 km north of Puerto Princesa City center (118◦43 N, 09◦50 E), is approximately 3 km inland from the Honda Bay coast, and situated within the catchment of the Tagburos River, which is a local fishery and tourist recreational area. In this area, a naturally occurring deposit of Hg ore (cinnabar), which is hosted by the Tagburos Opalite formation, can be found. Cinnabar was mined from 1953 to 1976, and the open pit mine site was naturally filled with water, forming a lake.

#### *2.2. Sampling*

A total of 102 soil and sediment samples were collected from the study area, between September 2018 to October 2019. Soil samples were collected from the vicinity of former mining sites; while surface sediments were collected from the bottom of the pit lake, Honda Bay, Tagburos River, and other rivers not traversing the pit lake in the area. Mine waste calcine samples were collected from the stockpile near the pit lake and from the scattered calcines situated 500 m northeast of the pit lake. All samples were classified into seven groups: PQMI pit lake, Tagburos River, Honda Bay, Honda Bay Wharf, mine waste calcines, soils (including agricultural soils) in the vicinity, and other rivers which are not tributaries of Tagburos River. Samples were collected using a plastic trowel, from the top 10 cm of the soil surface and surface sediment of rivers. Sediments from Honda Bay and the pit lake were collected using a sediment grab sampler (Wildco Petite Ponar® Grab Sampler, Yulee, FL, USA), collecting the top 5–10 cm of surface sediments. All collected soil and sediment samples were placed in polyethylene resealable plastic and transported to the laboratory at the Philippine Nuclear Research Institute in Manila.

#### *2.3. Total Mercury Analysis*

Analyses of THg in the soil and sediment samples were performed at the laboratory of the Philippine Nuclear Research Institute. Samples were prepared by air-drying and sieving up to US Standard 230 mesh (63 μm), to separate the silt-clay (<63 μm) size particles from the sand fraction. Only the <63 μm fractions were analyzed for THg concentrations. Total Hg analyses in the soil and sediment samples were determined using a direct mercury analyzer (Milestone DMA-80 evo, Sorisole, BG, Italy) following the U.S. EPA Method 7473 (mercury in solids and solutions by thermal decomposition, amalgamation, and atomic absorption spectrometry). This Hg analyzer is capable of determining THg concentrations in solid, liquid, and gas samples. Soil samples were weighed (0.10 g) and placed in a nickel boat without any pre-treatment. The boat is then introduced to the direct mercury analyzer where it is initially dried at up to 250 ◦C, and then thermally decomposed at 900 ◦C in a continuous flow of oxygen. The THg content is determined using atomic absorption spectrometry at 254 nm. All reported THg concentrations of soil and sediments analyzed were the average of three replicates. The THg analysis results from DMA-80 were validated using certified reference material for stream sediments (GSD-1) prepared by the Institute of Geophysical and Geochemical Exploration of China, with a mean (S.D.) THg value of 0.018 (0.004) mg kg<sup>−</sup>1. The mean (S.D.) value of the analyzed results using DMA-80 was 0.019 (0.002) mg kg<sup>−</sup>1, which is in good agreement with the certified value.

#### **3. Results and Discussion**

#### *3.1. THg in Mine Waste Calcines and Honda Bay Wharf*

From all the surface sediments collected in the vicinity of the abandoned mining site, there are two noticeable hotspots of very high Hg concentration relative to their surroundings (Figure 3). These are the mine waste calcines near the pit lake, and the calcine used in the construction of Honda Bay wharf.

**Figure 3.** Map showing the total Hg concentration in soils and sediments in the study area.

Mine waste calcines were found to be stocked northeast of the abandoned open pit lake (Figure 2). The mine waste calcines have poor sorting, angular grains, and can reach around 6 to 8 m in thickness. The calcine exhibits a reddish to orangey brown color due to the weathered iron bearing minerals and residual cinnabar. It mostly contains crushed bedrock material hosting the Hg deposit, along with few grains of silica from the opalite, which brought on the mineralization.

In this study, the measured THg in mine waste calcines have concentrations ranging from 52.7–924.2 mg kg−1. These calcines were stocked at the shore of the lake and scattered up to 500 m northeast of the lake. Total Hg concentrations found in the mine waste calcines are largely similar to the concentrations measured in the same study area in 2002–2003, but this time, a higher concentration (924.2 mg kg<sup>−</sup>1) was measured at one sampling point. A possible cause of the high THg concentration in mine waste calcines, is the inefficient and incomplete calcination or roasting of cinnabar ores. The range of THg concentrations in mine waste calcines measured in this study were generally within the range of THg concentrations on calcines from China, New Zealand, and the USA, as presented in Table 1.

Total Hg concentrations measured in soil of Honda Bay wharf and stockpiled mine wastes exhibit similar ranges of concentration. Samples collected in Honda Bay wharf display the same physical characteristics as the calcines stockpiled near the pit lake. This confirms that the soil materials used in the construction of the wharf were the mine waste calcines from PQMI mine site. High THg concentration (731.6 mg kg<sup>−</sup>1) was also measured in the calcines stocked near the wharf. It was higher than the estimated THg concentration (570 mg kg−1) of all the calcine materials used in Honda Bay wharf in 1996 [24]. Sediment samples collected 10 m north of the wharf shore have Hg concentrations

of >100 mg kg<sup>−</sup>1, which is in agreement with the concentration reported by Benoit and co-workers [24] in 1996. As stated in previous studies [20,21,24], Honda Bay wharf is considered one of the possible sources of Hg contamination in Honda Bay. The removal of Honda Bay wharf as well as the dredging of contaminated tailings and sediments in Honda Bay, were suggested in the study of Williams and co-workers in 1996 [21], but these recommendations were not followed. Today, the inhabitants of the villages located in the vicinity of the abandoned mine area including Honda Bay, which is home to around 10,000 inhabitants, are exposed to potentially serious health issues due to Hg contamination.


**Table 1.** Total Hg concentrations in soils and mine waste calcines collected from selected Hg mines worldwide.

\* number of samples was not provided in the literature.

#### *3.2. THg in Sediment in Tagburos River and Pit Lake*

Tagburos River, which is the main tributary in the abandoned mine area, has measured THg concentrations in the sediments ranging from 1.8–119 mg kg−1. Sediments found in the river were influenced by the geology of the surrounding area, as the exposed soils with Hg can be transferred into the river and eventually deposited in Honda Bay. High Hg concentration in the river sediments suggest that exposed mine waste calcines with elevated Hg concentration in the abandoned mine area are being carried to the river by erosion and surface runoff, and are eventually deposited in Honda Bay where they become a source of Hg contamination and enter the food chain. Rehabilitation programs, such as revegetation, are recommended in this abandoned mine area, to avoid the erosion of soil particles with high Hg concentrations going to the river and depositing in Honda Bay.

High THg concentration measured in the pit lake sediments ranged from 37.5–454.1 mg kg−1. The highest concentration was measured near the exposed bedrock on the lake shore, as shown in the inset of Figure 3. These values are within the range of the measured THg in pit lake sediments (6.9–400 mg kg<sup>−</sup>1) in 2003 [11]. The high THg concentration at the lake bottom may be due to naturally occurring cinnabar in the bedrock, and deposition of eroded calcine stocked near the lakeshore.

#### *3.3. THg in Soils*

Natural concentrations of Hg in soil are dependent on the source parent rock's mercury concentration, which becomes concentrated through the processes of weathering and natural erosion. At the vicinity of abandoned mine area, measured THg concentrations (0.04–67.5 mg kg<sup>−</sup>1) are 70 times higher than the global background for pristine soils (Table 2). The soil THg concentrations measured in this study were lower than the range of concentrations (0.01–168.7 mg kg<sup>−</sup>1) analyzed in the same area in 2002–2003. Though the lowest Hg concentration measured was 0.04 mg kg<sup>−</sup>1, corresponding to the background concentration; the other parts of the area might have been influenced by the movement of Hg-laden soil particles from the stocked mine waste calcines and other mine debris scattered in the abandoned mine area during operations. With consideration of the elevated sedimentation rate at Honda Bay (1 cm/year) [10], annual soil erosion to the bay via Tagburos River is likely a major source of Hg contamination for downstream ecosystems. Another human exposure pathway of Hg in the area is the dust coming from the bare soil during summer, which produces particulate-bound Hg that may be inhaled by the residents in the area.


**Table 2.** Mean total Hg concentration of different soil and sediment samples collected from different areas.

#### *3.4. THg in Honda Bay Sediment*

For the past four decades after the abandonment of the Hg mine operations by PQMI, there were very limited studies and monitoring conducted to determine the Hg concentrations in Honda Bay marine sediments. In the present study, analytical results of the THg concentrations in sediments collected from Honda Bay ranged from 0.04–12.7 mg kg<sup>−</sup>1. This range of data shows a greater range of THg concentration, compared to the previous measurements conducted by different researchers in the same area some 20 years ago [20,21,24]. In the study by Benoit and co-workers in 1994 [21], the estimated Hg concentration in Honda Bay sediments (0.003–3 mg kg−1) came from the wharf, which transported seaward or along the coast; while in this study, the identified possible sources of Hg contamination in Honda Bay marine sediments are the wharf (up to 12.7 mg kg−1, 250 m from the wharf) and Tagburos River (up to 9.6 mg kg−<sup>1</sup> in river mouth). The present THg concentration in the marine sediment also higher than the global Hg background concentration for sediment. Furthermore, the values gathered in the present study are in the same range of sediment THg concentration in the dredged area of Minamata Bay in 2006, as shown in Figure 4 [26]. Finally, such an elevated range of Hg concentrations at Honda Bay could also be biomethylated inside the Bay, becoming a significant source of methylmercury for both the marine and freshwater food chains that might be consumed by local inhabitants.

**Figure 4.** Mean and range of THg concentrations in surface marine sediments of Honda Bay, from 1982 to present. Data from Honda Bay were compared with data from the dredged Minamata Bay (2006) and global background. \* Kapauan et al. [20], \*\* Benoit et al. [24], \*\*\* Williams et al. [21], \*\*\*\* This study.

#### *3.5. THg of Sediment from Other Rivers*

Collection of sediment samples were even extended to other rivers that flow through the abandoned mine area, to determine the trend of Hg concentration surrounding the study area. The measured THg concentrations across several rivers in the area ranged from 5.3–11 mg kg−1. These values are more than 60 times higher than the global background, indicating Hg accumulation in the river caused by the erosion of the soil in the area with a naturally higher Hg background concentration. In comparison, sediments in Tagburos River show significantly higher Hg concentration than the other rivers since it traverses the abandoned mine site and the stockpiled calcines. Mercury may be moved along with the sediments and through water runoff, thereby increasing Hg values in the sediment deposited in Tagburos River.

#### **4. Conclusions**

In order to determine the pollution status and availability of Hg from the abandoned Hg mine in Puerto Princesa City, this study investigated the soil and sediment samples collected from the surroundings of the abandoned PQMI mine area. Results of the Hg analyses of these samples followed similar contamination patterns from other reported Hg mine sites worldwide. Mercury concentration in mine waste calcines ranged from 52.7–924.2 mg kg<sup>−</sup>1, of which some values were higher than those analyzed in 2003. Mercury levels measured in Honda Bay wharf were found to be similar in Hg concentration and characteristic with the mine waste calcines, confirming that the construction of wharf utilized the calcines from the PQMI mine site. Sediment samples obtained from Tagburos River and other tributaries, Honda Bay, and soils also exhibited Hg values that exceeded the global background, due to naturally high Hg concentrations in the mineralized area, and from the erosion and deposition of mine waste calcines into the surrounding areas. A greater range of Hg concentrations were measured at Honda Bay compared to the previous measurements conducted by several researchers in the same area some 20 years ago. These increased Hg concentrations in Honda Bay sediments may undergo methylation and enter the food chain, which was observed in Hg mining sites around the world [3,7,12]. This study concludes that one of the pathways of the reported Hg contamination in the area is caused by sediment transport from two point sources: (1) the calcines stocked at the shore of the pit lake and nearby area with scattered calcine, and (2) Honda Bay wharf. Sediment transport from the first point source was facilitated by Tagburos River and its tributaries, while the second point source was due to coastal erosion. Further studies are needed to evaluate the methylmercury levels in biota and particle-bound Hg in the air to evaluate health issues. Mining rehabilitation activities, such as

revegetation to reduce soil erosion and soil particle suspension, need to be implemented based on the results of this study.

**Author Contributions:** Conception and design of study and supervision, J.S. and C.R.G.; acquisition, analysis, interpretation of data, drafting the manuscript, and approval of the version of the manuscript to be published, J.S., C.R.G., A.T., and R.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Philippines' Department of Science and Technology's Grants-in-Aid Program (DOST-GIA).

**Acknowledgments:** The authors acknowledge the Department of Science and Technology; the Philippine Council for Industry, Energy, and Emerging Technology Research and Development (DOST-PCIEERD); and the Provincial Science and Technology Center of Palawan.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Particulate Mercury and Particulate Organic Matter in the Itenez Basin (Bolivia)**

### **Fabiola Guzmán-Uria 1, Isabel Morales-Belpaire 2,\*, Dario Achá 1,\* and Marc Pouilly <sup>3</sup>**


Received: 31 October 2020; Accepted: 23 November 2020; Published: 26 November 2020

**Abstract:** In rivers and other freshwater bodies, the presence of mercury can be due to direct contamination by anthropic activities such as gold mining. However, it can also be attributed to atmospheric deposition and erosion, runoff, or lixiviation from surrounding soils. In the case of the Amazon rainforest, high mercury contents have been reported for litter and topsoil, which could affect the mercury concentrations in water bodies. Samples of suspended particulate matter were obtained from a transect of the Itenez River, associated lakes, and some of its tributaries. The aim was to obtain information on particulate mercury's origin in the study area and determine the relationship between particulate mercury and particulate organic carbon. The concentration of mercury, organic matter, and the C:N ratio of the suspended matter was determined. The concentration of particulate mercury by water volume depended on changes in suspended matter loads, which in turn were mostly affected by the nature of the watershed or sediment resuspension. The observed values for the percentage of organic matter and the C:N ratio suggest that most of the mercury content in rivers and lakes originated from soils. A positive correlation was found between mercury concentration by weight of particulate matter and organic carbon content in particles. This correlation might be due to the direct binding of mercury to organic matter through functional groups like thiols or to an indirect effect of oxyhydroxides that can adsorb mercury and are associated with organic matter.

**Keywords:** particulatemercury; suspended particulatematter; particulate organic carbon; Amazon rainforest

#### **1. Introduction**

Mercury (Hg) is a significant environmental and health issue in the Amazon basin. Mercury's toxicity is particularly elevated due to the property of its organic form, methylmercury, to penetrate biological organisms and to generate accumulation during the individual's life and magnification along the trophic chain. In tropical forests, typical vegetation with broad leaves facilitates the capture of mercury from atmospheric deposition. Consequently, high contents of this metal have been reported for forest litter and topsoil [1]. In water bodies, mercury content can be related to atmospheric deposition, but most originates from the runoff or lixiviation of superficial and subterraneous water from mercury-laden soils [2].

The influx of mercury to water bodies from runoff can be exacerbated by natural erosion and anthropic activities such as deforestation, agriculture, mining, or urbanization. In rivers that drain deforested watersheds, an increase in total mercury in the water column and surface sediments has been observed and attributed to the erosion of fine particles from soils [3,4]. Additionally, forest fire for land conversion can give rise to an increase in Hg mobility [5]. However, the impact of deforestation on mercury concentrations in rivers and lakes varies in different watersheds [6].

Gold mining activities are also responsible for the discharge of mercury to the environment [6]. Exogenous (industrial) mercury used for gold mining can be emitted to the atmosphere or released into water bodies in the form of droplets and mercury-coated sediments [7]. Pollution may be worst when cyanide is used as part of the process because cyanomercuric complexes are more easily transported [7]. Schudel et al. [8], using stable isotopes of mercury, showed that mercury from mines where cyanide was used could be traced 120 km downstream. Although gold mining causes deforestation in smaller areas than cattle rising or agriculture, it is one of the leading causes of deforestation in South America [9]. Moreover, informal gold mining can cause more substantial soil perturbation than other activities due to the removal of the organic layer under the trees, which increases the loads of sediments containing high mercury concentrations [10].

Pouilly and coworkers [11] reported data on mercury concentrations in fish populations from different trophic levels in the Itenez basin (Bolivia). They also reported average values for particulate mercury in Blanco, San Martín, and Itenez rivers and their associated floodplain lakes. The authors highlighted that no clear relationship between particulate mercury concentrations in water and mercury concentrations in fish muscle was observed, indicating that bioavailability must be taken into account for interpreting the results. A second study by Pouilly et al. [12] suggested that periphyton's contribution to the food web could explain higher mercury bioavailability since it is a site of mercury methylation [13]. We report additional data to those that were presented by Pouilly et al. [11] about suspended particulate matter, total particulate mercury, particulate organic carbon, and the C/N ratio. We took into account some additional sampling points directly influenced by deforestation and mining and performed a seasonal analysis (early and late dry season). These additional data were analyzed to obtain information on the origin of particulate mercury in the study area and determine the relationship between particulate mercury and particulate organic carbon.

The data that we report and analyze correspond to samples that were taken from the Itenez basin in June and September 2007. No previous or more recent data on particulate mercury in the Itenez basin has been reported to our knowledge, although this river is one of the main tributaries of the upper Madeira River and its basin covers 303,300 km<sup>2</sup> shared between Brazilian and Bolivian territories [11]. The punctual values on particulate mercury that we report are not expected to reflect the current situation. However, the information on its relationship with anthropic activities and with organic matter contributes to the characterization of mercury pollution in that particular region. Moreover, this characterization is valuable for newer studies on mercury pollution in the region, particularly given recent deforestation, which has been aggravated by the 2019 and 2020 forest burnings.

#### **2. Materials and Methods**

Water samples were taken in June (early dry season) and September (late dry season) 2007 from a longitudinal transect in the Itenez River, which is located at the frontier between Bolivia and Brasil and from the tributaries Blanco and San Martín. The sampled transect in Itenez river is 660 km long, starting at the "Piso Firme" locality (S 13◦31 36.9 W 61◦49 40.7) and finishing 100 km upstream of the confluence with Mamoré River (S 12◦11 43.0 W 64◦40 00.5). Floodplain lakes related to the Itenez, San Martin, and Blanco Rivers were also sampled. Additional samples were obtained from three streams that cross Bolivian territory influenced by the "San Simon" goldmine and from three tributaries of Itenez that cross highly deforested Brazilian territory. The location of the sampling points can be observed in Figure 1. A more detailed description of the study area can be found in [11].

Five-liter plastic containers were used for sampling. These were conditioned by following a slight modification of the protocol proposed by Montgomery et al. [14]. The containers were first washed with tap water and metal-free detergent and rinsed with MilliQ water. After this, they soaked overnight with a 7% NaOH solution. They were then rinsed with MilliQ water and soaked with 10% HCl for an additional 12 h. Finally, they were rinsed three times with MilliQ water. Once clean, the containers were kept in hermetic plastic bags until their use. For the sampling of SPM, GF/F (0.7 μm) filters were used. Before sampling, the filters were placed in packages wrapped with aluminum foil and heated up to 500 ◦C for 8 h. Upon cooling, the filters were weighted and individually placed in Petri dishes.

**Figure 1.** Location of sampling points.

Water was collected from the mainstream of rivers and the center of lakes by submerging the containers around 1 m deep. Following [13], the person in charge of sampling wore clean gloves throughout the procedure. The containers were rinsed three times with the river or lake water before being filled. Additionally, in each sampling point, temperature (◦C), electric conductivity, and pH were measured with CONSORT C353 multimeter (Belgium). The depth of water was also registered.

SPM was obtained in the field by vacuum filtration of the water sample. The volume of water that caused saturation of the filter was annotated. Each filter was then placed in a Petri dish and stored at −17 ◦C until analysis.

Mercury content of SPM was determined at the Laboratorio de Calidad Ambiental of the University of San Andrés (La Paz, Bolivia). The filters that contained SPM were oven-dried at 40 ◦C for 24 h and weighted. Each filter was then introduced in a thermoresistant screw-capped 50 mL glass tube. 1 mL of nitric acid and 100 μL of hydrochloride acid were added to the tubes, and these were sealed with Teflon and capped. All tubes were heated at 100 ◦C for two hours. After cooling, 200 μL of hydrogen peroxide was added, and the tubes were again heated at 100 ◦C for two hours. 5 mL of ultrapure water was finally added.

Mercury of the digested samples was quantified with Cold Vapor Atomic Fluorescence Spectrometer (CVAFS, Brooks Rand Model III, Seattle, WA, USA) using the EPA 1631 Revision E: (EPA 2002) protocol. Reference samples were obtained from the National Research Council of Canadá: TORT-2 lobster hepatopancreas and MESS-3 Marine Sediment Reference Materials for Trace Metals (NRC-CNRC 2016). Total mercury concentration in SPM was expressed by water volume ([Hg]v) and SPM weight ([Hg]w). SPM from filters was additionally analyzed for particulate organic carbon (POC) and nitrogen in the Laboratorio de Medios Analíticos Cayenne- IRD in French Guyana with the CHNS THERMOQUEST AS2100 analyzer.

Wilcoxon test was applied to compare data obtained among sites (rivers and lakes) and seasons (early and late dry season). Spearman correlation coefficients between variables were calculated. All analyzes were carried out with R statistical computing freeware program (http://www.r-project.org).

#### **3. Results and Discussion**

A significant decrease in water depth was observed in September (late dry season, Tables S1 and S2) for most sampling points, except IR 1 and D3 (Tables S1 and S2). Most physicochemical characteristics did not vary among sites or seasons (Tables S1 and S2). The Blanco River showed a higher conductivity than the other water bodies (50 <sup>μ</sup>S cm−<sup>1</sup> as compared to a mean of 27 <sup>±</sup> <sup>4</sup> <sup>μ</sup>S cm−<sup>1</sup> for the Itenez River and 20 μS cm−<sup>1</sup> for San Martin), and this value increased during the late dry season (80 μS cm<sup>−</sup>1).

Quantities of SPM showed a mean value of 5.6 ± 1.1 mg/L with minor variations along the Itenez river in June (Table S1). In September, the quantities of SPM along the Itenez River were significantly higher (*p* = 0.005) than in June (Figure 2A). For point IR6 (Vuelta Grande), the value obtained in September more than doubled the value measured in June (12.4 mg/L in September compared to 5.7 mg/L in June, Table S1). Such an increase can be related to the resuspension of sediments by physical factors or bioturbation since the water column at this point became very shallow (0.9 m) [15].

**Figure 2.** Comparison of suspended particulate matter (**A**) and particulate mercury concentration by volume (**B**) for the Itenez River and associated lakes in the two seasons.

In June, SPM values for the Itenez lakes were similar to those obtained for the river (Figure 2A); this is in agreement with the fact that, during the high water level season, rivers and lakes are connected. No significant difference for SPM in the Itenez lakes was observed between September and June, except for lake IL9, which showed a strong increase of suspended matter (from 4.8 mg/L in June to 100.2 mg/L in September, Table S1). In September, the connections between the river and the lakes were interrupted, and most floodplain lakes of Itenez showed slightly lower concentrations of SPM than the river, probably because of a sedimentation process and the fact that they did not receive new particulate material. Lake IL9 became very shallow in September (1.70 m), and resuspension phenomena can explain the very high value of SPM (Table S1).

Two of the sampled tributaries of Itenez, a river from Brazilian territory that crosses highly deforested areas and the Colorado stream, which is influenced by gold mining showed in June the highest values for SPM (20.7 mg/L and 15.2 mg/L, respectively) (Table S2). However, the other tributaries directly influenced by anthropic activities showed values similar to those of the Itenez river. In September, the Asunta Stream showed a very high value (36.7 mg/L) for SPM, but the other two streams influenced by the "San Simon" goldmine had low values (6.4 and 9.7 mg/L). On the other hand, in June, Blanco river, which drains white waters, contained higher amounts of particulate matter than Itenez river (27.7 mg/L), while San Martín, a clear water river, and its lakes, showed the lowest values (2.9 mg/L, Table S2). Both rivers cross territory with little anthropic influence [11]. In September, however, concentrations of SPM in San Martin increased strongly (up to 19.9 mg/L), probably due to

the resuspension of sediments [15], while this was not the case for Blanco. Although deforestation and mining give rise to disturbance of the upper horizons of soil and cause an increase in the loads of solid particles that reach water bodies [9,10], factors not related to anthropic intervention must also be taken into account in order to determine the causes of high SPM.

The concentration of particulate mercury by water volume ([Hg]v) followed the same pattern as particulate suspended matter, showing little variations along the Itenez river (mean value 1.10 ± 0.19 ng/L) and its associated lakes in June (mean value 1.26 ± 0.23 ng/L, see Table S1). In September, the Itenez river had significantly higher [Hg]v than in June (*p* = 0.0129) (Figure 2B), as was observed for SPM.

Although high values for [Hg]v were measured in June for one of the rivers coming from deforested Brazilian territory and for the Colorado stream (3.83 ng/mL and 2.45 ng/L), the other tributaries influenced by anthropic activities showed concentrations of particulate mercury similar to those of Itenez river (Table S2). The lowest values for [Hg]v were observed in June for the San Martín River and its floodplain lakes (around 0.70 ng/L), while the highest value of all samples was obtained in the Blanco River (3.96 ng/L, Table S2). In September, the Asunta stream showed a substantial increase in [Hg]v (7.00 ng/L), but this was not observed for the other tributaries influenced by anthropic activities (Table S2).

The samples were taken from a very large area with different degrees of anthropic intervention, which goes from no intervention to gold-mining and deforestation. This area includes rivers and floodplain lakes, most of which contain clear water while some contain white water. No data are available for comparisons in this region, and there are few data on particulate mercury under similar conditions in the Amazonian basin. One exception is data from Roulet et al. [16] for suspended particulate matter and particulate mercury in the Tapajos river. The Tapajos river and the Itenez river are considered clear water rivers [16]. The values reported by Roulet et al. [17] for fine particulate matter and particulate mercury in the early dry season are similar to our data for the Itenez river and its lakes. The increase of SPM that we observed in the dry season (September) was also observed at Tapajos and was attributed to the decrease of water volumes. However, in our work, mercury associated with the coarse and fine particulate matter has been pooled, while Roulet et al. [17] did not report mercury concentration associated with coarse particulate matter. Comparisons must, therefore, be interpreted carefully.

More than the punctual values on particulate mercury by water volume, we were interested in the relationships between mercury and the degree and type of anthropic interventions. The relative contribution of mining activities and deforestation to mercury pollution in water bodies is still controversial and might depend on regional characteristics. Obritz et al. [18] indicate that the accumulation of mercury in soils depends on soil properties. Moreover, the accumulation in upper organic layers is different from the accumulation in mineral horizons. This implies that the contribution of soil mercury to water bodies can vary between regions. They also point out that runoff from contaminated sites such as gold mines can dominate Hg loads in water bodies. Horbe et al. [19] indicate that there are large variations in geochemical values of mercury in western Amazonia, and, by then, the contribution of different soils to Hg loads in rivers and lakes is variable. In our case, no systematic increase in [Hg]v was observed in sample points of Itenez river nearest to the mining sites or in streams affected by the "San Simon" mine. Furthermore, the obtained values for Hg loads in rivers that crossed highly deforested Brazilian territory and for rivers that crossed territories with little anthropic intervention showed no clear relationship between anthropic activities and particulate mercury loads. Phenomena such as resuspension, soil properties, and the water properties that the river drains (clear/white) must be taken into account when interpreting Hg loads.

When the concentration of total particulate mercury was evaluated by weight of SPM ([Hg]w), the values obtained in June for the complete set of samples fluctuated between 133 and 327 ng g−<sup>1</sup> (Figure 3A, Tables S3 and S4). No significant variation between June and September was observed for the Itenez river or its lakes. In samples from tributaries affected by mining activities and rivers that

cross highly deforested Brazilian territory, the values for [Hg]w were similar to those of Itenez river. Such was also the case for samples from Blanco and San Martin rivers.

**Figure 3.** Characteristics of suspended particulate matter including the concentration of mercury by weight ([Hg]w) (**A**), organic carbon (**B**), and C/N ratio (**C**) for the Itenez River and associated lakes in June (early dry season) and September (late dry season).

The percentage of particulate organic carbon (POC) in SPM in all samples from the Itenez basin varied from 6 to 26% in June (Tables S3 and S4). Samples from Itenez lakes tended to contain higher POC percentages than those from the Itenez river. POC decreased significantly (*p* = 0.02) in September in Itenez River, but no significant variation was observed for its lakes. Lower percentages of organic carbon associated with suspended matter can be due to lower organic matter inputs to water bodies in dry months and carbon turnover. Although particulate organic carbon is less bioavailable than dissolved carbon, some carbon turnover takes place [20].

In June, the C:N ratio for particulate matter from the Itenez river, lakes, and tributaries showed values between 9.9 and 14.9. In September, the C:N ratio values were lower, fluctuating between 7.4 and 11.2 for all samples (Tables S3 and S4). This decrease was significant for both the Itenez river (*p* = 0.004) and its lakes (*p* = 0.009), Figure 3C, and it can be related to preferential sorption of nitrogen compounds, as observed by Aufdenkampe et al. [21]. For both June and September, and for both habitats (lakes and rivers), the obtained C:N values are characteristic of suspended matter originated from soil [17,22]. This finding implies that the particulate organic carbon and the particulate mercury are mostly terrigenous in our study area. Terrigenous POC has a lower tendency to drive mercury methylation than autochthonous material [23], suggesting relatively low bioavailability of mercury.

A significant negative correlation was observed between SPM and POC for June (ρ = −0.739, *p* < 0.001) and September ((ρ = −0.739, *p* < 0.001). Likewise, SPM and [Hg]w were negatively correlated in June (ρ = −0.640, *p* < 0.001) and September (ρ = −0.709, *p* < 0.001, see also Figure 4A,B). Maurice-Bourgoin et al. [24] have also observed a negative correlation between SPM and [Hg]w for samples from the Negro, Solimoes, Amazon, and Madeira rivers. This "dilution effect" can be due to several processes. Higher loads of SPM generally imply that water also contains higher concentrations of colloids, which are not retained by filters but could absorb part of the available mercury [25]. Hsu-Kim et al. [6] indicate that there can be a "growth dilution" of particulate mercury caused by an increase in biomass attached to particles. On the other hand, for sampling points where strong resuspension of sediments was observed, POC's values and [Hg]w were much lower than in the other points. Sediments generally contain less organic carbon and mercury than suspended matter in the water column [3,23]. This difference can be related to high heterotrophic activity in sediments at the benthic and hyporheic zone [26]. High heterotrophic activity causes carbon turnover and can also cause mercury transformations, which leads to its separation from particles. A combination of the "dilution effect" and sediment resuspension could explain the behavior observed in Figure 4A,B.

**Figure 4.** Relationships between particulate organic carbon (POC) and suspended particulate matter (SPM) (**A**), [Hg]w and SPM (**B**) and [Hg]w and POC (**C**) for samples taken in June (closed squares) and September (open squares).

Particulate organic carbon was positively correlated with particulate mercury (see Figure 4C) both in June (ρ = 0.595, *p* = 0.004) and September (ρ = 0.595, *p* = 0.004). Such a positive relationship was also reported by other authors [23,24], although in some other cases, no significant correlation was found [15,17]. The nature of the samples of the particulate matter probably influences this correlation. Roulet et al. [17] found no correlation between [Hg]w and the POC percentage in fine particulate matter. The difference between this result and ours might be due to the fact that, in our case, we included coarse material in the analysis; this coarse material could contain higher percentages of organic matter. The characteristics of the water that is drained by the river could also play a role in the relationship between [Hg]w and POC. Maia et al. [15] found no significant correlation between [Hg]w and POC for floodplain lakes of the Amazonian river. However, these lakes contained water classified as white or black water, while the Itenez river is considered a clear water river. The fact that not all authors find a positive correlation between particulate mercury and particulate organic matter can also be due to regional differences and types of soil. Therefore, our data can contribute to a wider analysis of which conditions give rise to such a positive relationship.

Samples were taken in two seasons of the same year, at the beginning of the low water season and by the end of the low water season. The relationships between suspended particulate matter, particulate organic matter, and particulate mercury by weight [Hg]w did not show seasonal variations. The possible terrigenous nature of particles could explain the seasonal invariance of these relationships

The nature of the bonds between mercury and particulate suspended matter could affect mercury's fate and the possibilities of its methylation. Our results suggest that particulate mercury binds to organic carbon. However, mercury adsorption has also been related to aluminum and iron oxyhydroxides [3,27], which, in turn, are closely associated with organic matter [17]. Therefore, further studies are necessary to establish if the correlation of particulate mercury with POC is due to direct binding of mercury with organic matter or binding with the oxyhydroxides associated with this organic matter. Moreover, several studies have shown that mercury binding to organic matter is mostly through thiols, sulfides, and disulfides [25,28]. Consequently, the nature of the particulate organic matter might influence the amount of captured mercury. A study on sediment cores taken from the Titicaca lake has shown that capturing mercury on solids depends more on quality than quantity of organic matter. There is a positive correlation between bound mercury and reduced sulfur content [29]. Quantification of reduced sulfur in SPM could help establish the nature of the bonds between mercury and organic matter.

#### **4. Conclusions**

The concentration of particulate mercury by water volume [Hg]v in the studied water bodies depended on phenomena that caused changes in suspended matter loads. The effect of anthropic activities on these loads could not be established since the nature of the watershed or resuspension had a more direct effect on concentrations of SPM.

The POC values and the C:N ratio observed in suspended particulate matter suggest that most of this material in Itenez, Blanco, and San Martin rivers and lakes is terrigenous for the two sampling dates considered (early and late dry seasons), which implies that particulate mercury in these water bodies mostly comes from soils. At the same time, particulate organic matter (SPM) and [Hg]w were negatively correlated with the quantity of suspended matter, which may suggest a dilution effect.

A positive correlation was found between [Hg]w and POC. Although this suggests that mercury is bound to organic matter, further studies are needed to clarify this since organic matter could be associated with oxyhydroxides, and these minerals contain binding sites for mercury.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3417/10/23/8407/s1, Table S1. Characteristics of samples from Itenez river and its floodplain lakes, Table S2. Characteristics of samples from Blanco, San Martín, and tributaries of Itenez, Table S3. Contents of mercury ([Hg]w), organic matter (POC), and C/N ratio in suspended particulate matter of Itenez river and floodplain lakes, Table S4. Contents of mercury ([Hg]w), organic matter (POC) and C/N ratio in suspended particulate matter of Blanco, San Martin, and other tributaries of Itenez.

**Author Contributions:** Investigation, F.G.-U.; writing—review and editing, I.M.-B., review and editing, D.A., methodology, review, and editing, M.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the IRD (http://www.ird.fr) project Trophic Ecology of Amazonian Aquatic Systems (JEAI-EMAA). It also received financial support from a WWF Bolivia (http://bolivia.panda.org) funded project (KN10 WWF-IRD agreement). Published with the support of IRD through the (JEAI-TITICACA) project.

**Acknowledgments:** We are particularly grateful for all the advice and assistance given by Jean Louis Duprey for all the laboratory work. We thank the Trinidad WWF team, Itenez departmental park (PD-AMNI Ite nez), and Beni Prefecture (Bolivia) for the logistical support and all the people who helped in the field work, especially the fishermen and boat pilots: H. Ribero Rodriguez, T. Suarez, and J. Vasquez. We greatly appreciated the warm welcome in the Itenez communities of Bella Vista, Remanso, Mategua, and Versalles.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Impact of Old and Recent Gold Mining Sites on Mercury Fluxes in Suspended Particulate Matter, Water and Sediment in French Guiana**

**Jennifer Hellal 1,\*, Jörg Schäfer 2, Régis Vigouroux 3, Laurent Lanceleur <sup>2</sup> and Valérie Laperche <sup>1</sup>**


Received: 15 October 2020; Accepted: 30 October 2020; Published: 4 November 2020

**Abstract:** In 2006 the use of mercury (Hg) was banned for gold mining in French Guiana. However, mining of old placers could mobilize Hg accumulated in soils and sediment. This study aimed to measure the current impact of a mining concession (Boulanger site) on the Hg load in the watershed. Turbidity, Total Mercury (THg), and Monomethylmercury (MMHg) were measured in water, suspended particulate matter (SPM), river sediment and sediments from old tailing ponds along a river section of 30 km up and downstream from a mining concession in French Guiana during a dry and a rainy season. Total dissolved Hg (THgD) concentrations varied little from up- to down-stream but were all higher (fourfold on average) during the rainy season (3.2 to 4.4 ng L<sup>−</sup>1), than during the dry season and consistent with previous data known for the Amazonian area. Dissolved MMHg (MMHgD) represented up to 30% of THgD during the dry season, which is higher than previous results (typically around 2%). Mercury concentrations in sediments were highest in the vicinity of areas affected by old (before 2006) rather than new gold mining practices. Even though Hg was banned in 2006, present gold mining practices still release natural Hg and Hg inherited from older mining practices into the watershed.

**Keywords:** mercury; sediment; water; SPM; gold mining; French Guiana

#### **1. Introduction**

Mercury (Hg) is a toxic metal and a global pollutant, historically used in alluvial gold-mining to amalgamate and recover gold. In 2005, small-scale gold mining was estimated to contribute to more than 10% of the annual global anthropic Hg load to the atmosphere [1]. Use of Hg for gold mining was officially banned in French Guiana in 2006. However, "inherited" Hg in old gold mined areas is still present and remains a potential source of Hg to the environment. It is estimated that for 1 g of extracted gold, ~1.3 g of Hg are lost, despite its partial recycling [2,3]. In addition to Hg inputs from gold mining, Amazonian soils naturally contain high Hg concentrations due to the equatorial pedoclimatic situation (on average 10 times higher than temperate soils) [4]. During gold mining and generally during all deforesting operations [5–8], Hg adsorbed to colloids (clay, organic matter, and iron or aluminum oxides) is exported to the watershed, contaminating downstream ecosystems. In anaerobic environments (stagnant lagoons and sediment/water interface), Hg can be methylated, mainly by sulfate and iron-reducing bacteria, to form monomethylmercury (MMHg) [9,10]. MMHg is a neurotoxin which readily accumulates in the food chain up to concentrations considered dangerous for human health (around 1 mg kg−<sup>1</sup> fresh weight in carnivorous fish such as Aymara, which is twice the

World Health Organization (WHO) recommendation and 105–106 times higher than surrounding water concentrations) [11]. In local Amerindian populations, whose diet traditionally consists of freshwater fish, Hg concentrations can reach values clearly exceeding World Health Organization (WHO) safety recommendations. Indeed, concentrations of more than 10 μg g−<sup>1</sup> hair have been measured in these populations [12].

French Guiana has known several gold rushes towards the end of the 19th century and especially at the beginning of the 20th. The gold mining techniques at that time, mainly imported from Anglo-Saxon methods developed in the Rocky Mountains, already used Hg to amalgamate gold. During these periods, gold-mined sites in French Guiana were mainly placers in small riverbeds or lateritic colluvium exploitations. The use of river barges came later, during the second half of the 20th century. It is considered that the activity at that time was much higher than presently, considering the number of placers and the volume of mineral treated. Evidently, this gold mining activity left traces of Hg contamination both on-site and downstream from the placers in the hydric network and the long-term stability of these Hg stocks is unknown. Many of the old placers are now covered by secondary rain forest, where little erosion occurs. However, these areas are potential sources of Hg pollution and need to be investigated to estimate their potential contribution to Hg dispersion in the Guianese hydric network, especially as many still contain economically relevant amounts of gold and are being re-exploited. Although 35% of the Hg used for goldmining is estimated to be released to the atmosphere, the remaining 65% are dumped on-site as metallic Hg (0) [13]. According to Laperche et al. [14], Hg concentrations in Guiana river sediments are generally close to background soil concentrations (100 <sup>±</sup> 50 <sup>μ</sup>g kg−1) and strongly increase close to gold mining sites. Concentrations up to 12 mg kg−<sup>1</sup> have been measured in the River Approuague sediments and up to 15 mg kg-1 along the Eddie Creek (unpublished data). Spatial distribution in sediments along French Guiana Rivers indicates that metallic Hg lost by gold miners globally remains in the rejection zones, as contamination "hot spots" [14].

At gold-mined sites, Hg in soils either occurs naturally or is left over from past gold mining in the case of reopened old sites [8]. Due to mining, this Hg is mobilized together with soil particles and is dispersed downstream in the hydric network, where it can contaminate different compartments of the ecosystem. The water rejected from legal gold mining sites is controlled by measuring water turbidity, as high water turbidity can damage aquatic ecosystems due to light extinction [15]. Guédron et al. [16] showed that exported Hg is mainly associated to the fine suspended particulate matter (SPM) and colloids, prone to be transported over longer distances than coarse particles. However, despite this knowledge and the controls, the actual quantities of Hg rejected remain unknown, whether in pristine new placers or in re-opened old gold-mining sites that potentially contain more Hg.

The objective of the present work was to evaluate and compare the amounts of Hg exported from different gold mining areas, pristine or not, to the hydric network in contrasting hydrological situations (dry and rainy season) and to identify potential changes in water quality.

#### **2. Materials and Methods**

#### *2.1. Study Area and Sampling Strategy*

The Comté-Orapu basin covers 3359 km2. This study focuses on the sub basin of the Boulanger Creek, a tributary of the Orapu River (Figure 1), and downstream rivers including the Comté River, which joins the Orapu, forming the Oyack River. Sampling was carried out along a section of ~30 km, from the Boulanger Creek's spring, accessed along the Molokoï path, to the Oyack River, along which water and sediment samples were collected, during a dry season in September 2012 and the following rainy season in March 2013 as described below.

**Figure 1.** Map of the study area in French Guiana and the sampling points.

During each sampling campaign, 14 water (~15 cm depth) and sediment samples were collected along the Boulanger Creek (Table 1 and Figure 1) and downstream Rivers (Orapu, Comté and Oyack).


**Table 1.** Description of sampling points.

Suspended Particulate Matter (SPM) were recovered by filtering between 250 and 500 mL water on-site through pre-weighed cellulose nitrate filters (0.45 μm; Ø47 mm; Sartorius®, Dourdan, France). The filters alone (blank) and the filters with SPM were transported to the laboratory in closed petrislides, then oven dried (50 ◦C, ~36 h) to constant weight. The mass of SPM sampled was determined on a precision balance (0.01 mg scales) from the difference with the initial dry filter weight. The filters were then conserved in sealed polypropylene boxes in the dark until further analysis.

Water samples were obtained by filtering 125 mL water (<0.45 μm, polyvinylidene fluoride filters (PVDF)) into pre-cleaned Teflon fluorinated ethylene propylene (FEP) bottles. Samples were acidified with suprapur HCl 0.5% *v*/*v* (Merck, Darmstadt, Germany) and stored in double polyethylene bags until analysis. All materials (PP, Teflon) in contact with samples were acid-washed (5 days in 20% HNO3 *v*/*v* followed by 5 days in HCl 10% *v*/*v*), rinsed several times with deionized water (Milli-Q®, Merck Millipore, Darmstadt, Germany) and dried under a laminar flow hood and stored in sealed polyethylene bags pending sampling. Polyethylene gloves were used for all handling operations.

River sediment samples were collected along the river section at the same spots as the water and SPM samples. At each sampling site, sediments were collected from just above the low water line during low tide on river banks where finer sediment tends to accumulate. It is assumed that they provide geochemical information that reflects the previous season's conditions [14]. There was no chemical (saltwater) impact of the tide on the water samples. Sediment samples were stored in polyethylene bottles and sent to the laboratory in Petit Saut (Hydreco, French Guiana, French) for treatment and analysis.

During the dry season (September 2012), sediment samples were also collected from the edges of 14 old tailing ponds (Figure 1). Some of these tailing ponds were filled more than 10 years ago and they are now covered by vegetation without having been properly rehabilitated, which makes it too dangerous to walk on the tailing pond's surface. No tailing samples were collected during the rainy season in March 2013. Indeed, as shown in the results section, Hg contents measured in the 2012 samples were similar to previous analyses carried out in 2006 [14], suggesting little evolution of the Hg concentrations in the tailing ponds.

#### *2.2. Mercury Analysis in Tailing Ponds and Sediment Samples*

Tailing and sediment samples were oven dried at 38 ◦C to constant weight and sieved (2 mm) before analysis. Total Hg (THg) concentrations in solids were determined by Atomic Absorption Spectrophotometry after dry mineralization and gold amalgamation with an Automatic Mercury Analyzer (Model AMA 254 Altec, Uherske Hradiste, Czech Republic). All particulate THg concentrations were expressed in ng g−<sup>1</sup> dry sediment. Triplicates were performed for each analysis for quality assurance/quality control (QA/QC). The relative error was routinely ±5% (Relative Standard Deviation) and always under±10%, consistently with typical relative errors reported for this method [17]. Detection limit (defined as 3 times the standard deviation (SD) of the blanks) was maximum 5 ng g<sup>−</sup>1. Concentrations obtained for repeated analyses of the certified reference material (CRM) MESS-3 (Marine Sediment Reference Materials, National Research Council Canada, Ottawa, Ontario were consistently within the published concentration range (91 <sup>±</sup> 8 ng g<sup>−</sup>1).

#### *2.3. Dissolved Mercury (HgD) and Methylmercury (MMHg) Analyses*

All analytical procedures were conducted using ultra clean sample handling to avoid laboratory contamination of low-level sample extracts and a cross-contamination of high-level samples [18]. Calibration was performed by species-specific isotope-dilution by adding known amounts of 201Hg-enriched Hg(II) solution (Oak Ridge National Laboratory, isotopic purity 96%) and 202Hg-enriched MMHg solution (CH3 202HgCl; ERM-AE670; Institute for Reference Materials and Measurements, isotopic purity of 98%) directly to the sample before the extraction procedure [19,20]. Analysis of Hg species in the organic extract was performed by Gas Chromatography (Thermo-Fischer Scientific, Focus GC, France) coupled with ICP-MS (Inductively coupled plasma mass spectrometry, Thermo-Fischer Scientific; X7, France) as previously described [21].

Detection limit (3 sigma of the blank values including reagent blank and sample handling) was 0.008 ng L−1. The analytical results were continuously quality checked by analyzing an internal standard solution prepared by digesting a known amount of international certified reference material (International Atomic Energy Agency, IAEA 405) with 6 M HNO3 [20]. Accuracy was 94 ± 4% and precision based on replicate analyses (n = 12) of the standard solution was better than 5% (RSD).

#### *2.4. Mercury Concentrations in Suspended Particulate Matter*

Total Hg in SPM was measured by Atomic Absorption Spectrometry with a Direct Mercury Analyzer 80 (MILESTONE, Sorisole, Italy), i.e., analytical equipment using procedures similar to those used for THg analyses in sediment (see above). Quality control was performed by measuring blanks and

certified materials of different origins (estuary, sea or river sediments) and concentrations (from 30 to 1600 ng g−1; NCS DC 70317 (Tibet sediment) from Laboratory Standards (Budapest, Hungary), IAEA-433 (Marine Sediment) from IAEA Reference Products for Environment and Trade, BCR-320R (Channel sediment) and BCR-280R (Lake sediment) from the Joint Research Centre (Geel, Belgium), after each set of 5 samples. Accuracy was evaluated to ~90% and precision was better than 5% (RSD).

#### *2.5. Turbidity*

Turbidity is water cloudiness caused by suspended and dissolved particles. As water turbidity is mainly caused by the presence of SPM [22], turbidity has often been used to estimate SPM concentrations. Turbidity is a function not only of SPM but also of the sizes, shapes, and composition/colour of the particles. These variables may introduce bias to SPM estimation. In this study, turbidity was assessed by the value of Nephelometric Turbidity Unit (NTU). Turbidity was measured with a microprocessor turbidity meter (Eutech Instrument Turbidimeter TN-100, Fisher Scientific, Illkirch, France) calibrated by using Formazin solutions of 0.02, 20, 100, and 800 NTU, according to the manufacturer's operating instructions.

Parallel measurements of SPM concentrations by filtration and weighing were performed (see above) to establish potential relationships between turbidity and SPM concentration under the different hydrological situations studied.

#### *2.6. Water Flow and Hg Fluxes*

Water flow was measured with an acoustic Doppler velocimeter ([ADV] 10-MHz Ocean ADV, Sontek Inc., San Diego, CA, USA), in 6 different places along the rivers: (1) upstream and (2) downstream the Boulanger Creek; (3) in the Orapu River, just upstream from its confluence with the Boulanger Creek; (4) in the Orapu River, just downstream from its confluence with the Boulanger Creek; (5) in the Comté River just before it flows into the Orapu; and (6) in the Oyack, just after its formation by the joining of the Comté and Orapu Rivers.

Fluxes were estimated for the SPM in g s−1, for total particulate Hg adsorbed on the SPM (THgp, μg s<sup>−</sup>1) and for total dissolved Hg (THgD, μg s<sup>−</sup>1).

#### **3. Results**

#### *3.1. Hg in Tailing Ponds*

Total Hg concentrations in the 14 tailing pond samples ranged between 70 and 1020 ng g−<sup>1</sup> dry sediment (Figure 2). The Hg concentrations in samples (n = 8) from the pond zones exploited before 2006 (from Creek St Roch to Creek Diable, points 17–24, Figure 2 covered the whole concentration range, whereas the recently exploited zone showed concentrations between 110 and 470 ng g−<sup>1</sup> (n = 6). Except for point 4 (1020 ng g−1), the average Hg concentration in pond sediments was 190 ng g−1, which is very similar to values obtained in 2006 for the same ponds (180 ng g−1, 26 samples from 10 ponds) [14].

**Figure 2.** Localization of samples taken from old tailing ponds in the Boulanger concession in the dry season of September 2012 and total mercury (THg) contents (ng g<sup>−</sup>1). Points 17–24 are tailing ponds dating before the mercury (Hg) ban for gold mining in 2006, whereas points 25–30 are more recent.

#### *3.2. Total Hg in River Sediments and Suspended Particulate Matter in Relation to Turbidity*

Total Hg in river sediments (THg) from both sampling campaigns were compared (Figure 3a). Overall, there was little difference between the dry season and the rainy season with THg concentrations varying from 40 to 380 ng g<sup>−</sup>1. For both campaigns, the highest concentrations occurred at points B2 and B3 in the sector exploited prior to 2006 (B2) and in the more recently exploited area (B3).

The lowest THg in river sediment was measured at point B5, where the Boulanger Creek was deviated from its original course. The sediment at this point contained 42 ng kg<sup>−</sup>1, close to the lowest value (31 ng g<sup>−</sup>1) found in this area [14]. Moreover, THg concentrations at B5 were even lower in 2013 (34 ng g−<sup>1</sup> dry sediment), when the creek was positioned to flow through the middle of the flats in a rehabilitated zone.

Downstream from B5, total Hg concentrations in river sediments increased regularly from 70 to maximum 150 ng g−<sup>1</sup> along the Boulanger Creek to its confluence with the Orapu River (point B9) and were then constant until the Orapu and the Comté Rivers merge and form the Oyack River (points Or1 to Or3).

Total Hg concentrations in SPM collected during the dry season are comparable to those in the sediment and vary from 80 to 505 ng g−<sup>1</sup> (Figure 3b). However, during the rainy season Hg concentrations in SPM increased by two- to three-fold, except in the most upstream point where values were the same (~150 ng g<sup>−</sup>1) for both seasons. Maximum Hg concentrations in SPM, i.e., ~600 ng g−<sup>1</sup> were higher than maximum values in sediments (~380 ng g<sup>−</sup>1).

**Figure 3.** (**a**) Hg associated to river sediments, (**b**) Hg associated to suspended particulate matter (SPM), (**c**) suspended particulate matter (SPM) and (**d**) Turbidity at all sampling points for the dry season in September 2012 and the rainy season in March 2013. B1 to B9: Boulanger Creek, Or1 to Or3: Orapu River, C1: Comté River and Oy1: Oyack River.

**Figure 4.** Relation between SPM and Nephelometric Turbidity Unit (NTU) in the dry season in September 2012 and the rainy season in March 2013.

Although Hg concentrations in SPM increased during the rainy season, the amounts of SPM in river water were globally similar for both sampling campaigns (Figure 3c) and ranged between 2 and 18 mg L<sup>−</sup>1. Turbidity on the other hand was 2–9-fold higher during the rainy season with an increase in NTU along the Boulanger Creek to 38 NTU that then dropped in the Orapu and Oyak to around 10–15 NTU (Figure 3d). The lowest value (<5NTU) corresponded to an isolated spot upstream from the mining district and uninfluenced by human activities.

Accordingly, amounts of SPM did not follow the same pattern as the turbidity values. However, a linear relation was found between turbidity and SPM concentrations, that differed between the dry and rainy seasons (Figure 4).

#### *3.3. Hg and MMHg in River Water*

Filtered water THgD concentrations varied little from up to downstream during the rainy season, from 3.2 to 4.4 ng L−<sup>1</sup> (Figure 5) but were all higher than during the dry season (fourfold on average). During the dry season, an increase of MMHg in waters from upstream Boulanger down to the mined areas occurred. Concentrations then decreased downstream in the Orapu, probably because of the increase of the flow rate of the Orapu River (80 m3 s−1, SM-1) compared to the discharge of the Boulanger Creek (2.2 m3 s<sup>−</sup>1, SM-1). In contrast, during the rainy season, no clear pattern was observed but higher concentrations (160 pg L−1) were measured at points B2, B5, B8, and C1. During the dry season, the percentage of MMHg (MMHg/THgD) reached up to 30% downstream from the gold-mining district (Figure 5a–c), with an overall average of 13%, whereas during the rainy season it was of 3% in the same zone with a maximum of 4%.

**Figure 5.** Dissolved Hg and monomethylmercury (MMHg) in water samples from the Oyack River catchment. (**a**): Total dissolved Hg (THgD) (ng L<sup>−</sup>1), (**b**): MMHg (pg L<sup>−</sup>1) and (**c**): %MMHg/THgD. In all three graphs, stars give data for the dry season in September 2012 and triangles for the rainy season in March 2013. B1 to B9: Boulanger Creek, Or1 to Or3: Orapu River, C1: Comté River and Oy1: Oyack River.

#### *3.4. SPM and Hg Global Export*

Based on the water discharges measured at different points during the dry and rainy seasons (SM-1), estimations of the amount of exported SPM, particulate Hg or HgP and THgD were made for both seasons (Figure 6). Between the dry and the rainy season, water discharges increased by a factor of 3.4 to 4.1 at all sampling points (SM-1). There was also a similar increase (between 5.6 and 5.7) from up- to downstream along the Boulanger Creek. During the dry season, the area around the Boulanger Creek and the Comté River showed less gold mining activity than was expected. This induced a decrease in SPM export. During the dry season also, SPM exportations did not increase from upstream to downstream probably due to a balance between sedimentation and flux that maintained the SPM charge stable. On the opposite, during the rainy season, increasing river discharges resulted in higher erosion and transport, leading to an increase in SPM charge from upstream to downstream. Depending on the sampling points, the increase in HgP varied from 2 (upstream Orapu River) to 15 (Oyack River) between the dry and rainy season, reflecting the related soil erosion, whereas the increase in HgD was constant (~20 fold) whatever the sampling point.


**Figure 6.** Comparison of water flux in the Orapu-Comté watershed and associated estimations of SPM and Hg export during the dry and rainy sampling campaigns (September 2012 and March 2013), (dry season (D); rainy season (R); particulate Hg (HgP); dissolved Hg (HgD).

#### **4. Discussion**

#### *4.1. Hg in Tailing Ponds Remains Constant over Time*

As the concentrations measured in 2012 were very similar to those measured in 2006, it appears that the Hg in the tailing ponds remains stable over time, at least at the pluri-annual time scale. The range of values found here are much lower than Hg contents reported in Venezuela [23], where Hg in pond sediments ranged from 160 ng g−<sup>1</sup> to 542 μg g<sup>−</sup>1, but comparable to those found in abandoned ponds in Colombia [24]. Here the authors measured THg in sediments from 27 tailing ponds, abandoned since 1997 and found between 39 and 1271 ng g−<sup>1</sup> dry sediment.

Although it was expected to find higher amounts of Hg in the ponds dating from before the Hg ban for goldmining in 2006, some relatively high concentrations were also found in the more recently exploited areas. The typical background sediment concentration is around 100 ng g−<sup>1</sup> sediment [14] and

all the samples, except for point 17 (Figure 2), where THg was 70 ng g<sup>−</sup>1, were above this value. In the zone exploited after 2006, high Hg in the tailing ponds could be a legacy from older exploitation periods.

#### *4.2. Recent Goldmining Activity Increased THg in River Sediment Despite the Hg Ban in 2006*

Overall values of THg in river sediments varied between 40 and 380 ng g−<sup>1</sup> and were similar in magnitude to previous results reported for French Guiana [14]. Spadini et al., [25] found 510 ng g<sup>−</sup>1, 295 ng g−<sup>1</sup> and 261 ng g−<sup>1</sup> dry sediment at distances of respectively 0, 35, and 50 km downstream from mining areas. Goix et al. [26], measured between 32 and 358 ng g−<sup>1</sup> with significantly higher values in artisanal and small-scale gold-mining (ASGM) areas along the River Oyapock. These results are also close to values of 20 to 53 ng g−<sup>1</sup> reported for river sediments along the Malinoski–Tambopata river system in Peru, a river affected by artisanal and small-scale gold-mining (ASGM) [27]. In another Peruvian river system, the Madre de Dios, also impacted by ASGM [28], authors reported Hg levels in sediments ranging from 9 ng g−<sup>1</sup> upstream up to 95.3 ng g−<sup>1</sup> downstream from the gold mined areas. More recently in Brazil, [29], measured between 19 and 155 ng g−<sup>1</sup> along the River Tapajos. Again, highest values were observed in the vicinity of gold mining areas. However, all these Hg concentrations in goldmining-affected sediments of the Amazonian water basin were clearly lower than those reported for River Gambia in Senegal, where Hg in sediments reached 1160 +/- 800 ng g−<sup>1</sup> in an area affected by ASGM [30].

In the present study, the highest concentrations of Hg in sediment occurred at points B2, B3, and B4 along the Boulanger Creek. These sites are both located directly downstream from the old tailing ponds, from where Hg-rich sediment can be exported and from the most recently exploited mining area (2011–2012), where bare soils are likely to favor soil erosion. These re-exploited old placers probably are sources of both Hg naturally present in soil and Hg leftovers from old mining practices, which would explain the increase at point B4 between the dry and rainy season (THg doubled from 100 to 200 ng g−1) in relation to the onset of the exploitation next to the nearby Fourmi Creek in 2013. Previously, in a study conducted on the Tapajos River in Brazil, Telmer et al., [31] concluded that mining operations were at the origin of Hg anomalies, not from the amalgamation process itself, but from the increased erosion during the mining operations. This re-mobilizes the mercury contained naturally in soils and mercury resulting from historical mining activities that then migrates to the river. These results are coherent with our results and are supported by recent research conducted by Goix et al., [26] using an isotopic approach. These authors showed that up to 78% of Hg in sediments downstream from the gold mining site was of anthropogenic origin, of which more than half derived from the release of liquid Hg (0) and the rest from erosion of surrounding Hg-contaminated soils.

Outside the mining zones, THg concentrations in sediment were of the same order as those measured previously in this area [14]. Accordingly, the relatively low concentrations measured upstream from the mining areas (point B1) were similar to previously reported values for uncontaminated Amazonian sediments [32,33].

The decrease in the Hg load in sediments downstream from the Boulanger Creek and especially in the larger rivers such as the Orapu and the Comté is supposedly due to the huge dilution with waters draining the whole watershed and highlight the locally high values of creeks such as the Boulanger, impacted by goldmining.

#### *4.3. Greater SPM Hg Load during the Rainy Season All along the Water Shed*

Suspended Particulate Matter (SPM) increased along the Boulanger Creek, from 2 mg L−<sup>1</sup> at point B1 to a maximum of 18 mg L−<sup>1</sup> at B8. Downstream from there, values were between 4 and 8 mg L−<sup>1</sup> during the rainy season and 8–12 mg L−<sup>1</sup> during the dry season. Mining activity was low that year on the Boulanger Creek, which could explain the small difference between the two seasons, but also a natural dilution effect can reduce SPM concentrations. Due to the low erosion rates of the Precambrian Guiana Shield, French Guiana Rivers naturally have very low SPM values [34]. Amounts of SPM observed in the present work are similar to average values reported by these authors along other Guiana Rivers, i.e., the Oyapock and the Maroni Rivers.

During the rainy season, the Hg load in SPM increased by 2–3-fold at all sampling points except B1. During the dry season, concentrations were between 100 and 250 ng g−<sup>1</sup> while during the rainy season they were between 400 and 600 ng g−1. It would seem that the SPM mobilized during the rainy season contains more Hg than during the dry season. Roulet et al. [33,35,36] reported THg concentrations from 167 to 425 ng g−<sup>1</sup> (dry weight) in SPM from clear waters (such as those in French Guiana), on a commercial sector of the Tapajos River, and from 154 to 355 ng g−<sup>1</sup> (dry weight) in a zone 50 km from the mining district. The THg concentrations in the present study were slightly higher than this, but clearly lower than those reported for the Rio Négro, where THg reached values up to 2070 ng g−<sup>1</sup> [32]. However, in 2006 [16], during the rainy season at point B5 (down-stream Leoncé; Figure 1), THg concentrations were 1420 ng g−<sup>1</sup> (dry weight) i.e., three times higher than in the present study, which may be due to the fact that during this period there was active mining upstream from site B5. In other Amazonian watersheds, similar or higher values have been reported, e.g., between 400 and 4000 ng g−<sup>1</sup> were found along the ASGM-affected Malinoski–Tambopata river system in Peru [27]. Furthermore, Hg concentration ranges of 3.2 to 133 ng g−<sup>1</sup> in SPM during the dry season, and 13.3 to 42 ng g−<sup>1</sup> during the rainy season were found along the Madre de Dios River [28]. In Brazil, along the Tapajos River, Hg in unfiltered water varied between 0.65 ng L−<sup>1</sup> and 23.8 ng L−<sup>1</sup> in tributaries closer to gold mining activities. Given that Hg is mainly transported in the particulate fraction, this maximum value corresponds to around 86 ng Hg g−<sup>1</sup> SPM [29].

#### *4.4. Relation between Turbidity and SPM Was Di*ff*erent According to the Season*

Highest turbidity values occurred along the Boulanger Creek in areas flowing through active mining zones and downstream from these zones. Turbidity values decreased sharply in the Orapu River and then increased steadily down towards the Oyack River but not to values as high as in the Boulanger Creek. Turbidity must not be regarded as proportional to SPM concentration but as one effect of SPM among others on water physical properties such as transparency. Thus, although SPM concentrations during the rainy season are comparable to those of the dry season, turbidity can be higher. This is partly explained by the nature of SPM, which can vary from one river to another depending on the organic and mineral load and influence light reflection. Indeed, changes in turbidity per unit of SPM are likely associated with changes in particles morphology and also in changes in size as large particles scatter light less efficiently than smaller ones [37,38]. The fact that, for a similar SPM concentration, turbidity in the studied rivers was clearly higher during the rainy season, may also suggest the presence of high amounts of colloidal/dissolved components, such as humic and/or fulvic acids due to increased soil leaching [38]. Two distinct linear relationships were obtained for the rainy and dry season between SPM and turbidity, possibly linked to the different nature of SPM during the two seasons. Further work to characterize these differences could also contribute to explaining the Hg load. Turbidity could provide an easy way to control gold mining sites and might be an efficient monitor of correct use of tailing ponds, although it appears here that for a same turbidity measure, the SPM load is not the same depending on the season, and thus the amounts of transported Hg are also different.

#### *4.5. River Water THgD and MMHgD*

As in previous studies, the Hg fraction in filtered water was very low compared to the particulate fraction. Indeed, SPM remains the major pathway for river borne Hg as Hg bound to SPM [31]. Concentrations of THgD averaged 0.25 ng L−<sup>1</sup> in the dry season of 2012 and 3.70 ng L−<sup>1</sup> in the rainy season of 2013. These concentrations are comparable to those found in unpolluted waters [39] and are similar to those measured previously in the Boulanger Creek in 2005–2006 [16].

During the dry season, concentrations of MMHgD were globally higher along the Boulanger Creek than further downstream, whereas during the rainy season no clear pattern was found and concentrations fluctuated between low (20 pg L−1) to high concentrations (160 pg L−1). In comparison, [16] found concentrations of 24–25 pg L−<sup>1</sup> in the Boulanger Creek in 2005–2006. Although highest concentrations of MMHgD were measured during the rainy season, during this period it represented a very small (<3%) fraction of THgD, whereas during the dry season this ratio went up to 30% at points B7 and B8. Such increase could be due to conditions that are more stagnant during the dry season favoring anaerobic conditions and Hg methylation.

#### *4.6. SPM and Hg Global Export*

Exported SPM from the Boulanger Creek was low compared to the Orapu River; however, it is possibly the multitude of such small creeks impacted by goldmining activities which results in the concentrations observed downstream. An approach such as the one carried out by Gallay et al. [34], who estimated the impact of land degradation on sediment fluxes in the Maroni and Oyapock River catchments, coupled to Hg measurements would yield further information on the impact of modern goldmining on Hg fluxes.

#### **5. Conclusions**

Regardless of the process used for the extraction of alluvial gold in French Guiana, the activity consists first of the resuspension of the clays and loose materials of the placers using pressurized water jets. Without carefully sized settling basins, this considerably increases river turbidity and Hg mobility. The hypothesis of the present study was that this phenomenon is more intense when older sites are exploited and that "inherited" Hg can also be remobilized. Based on the two sampling campaigns carried out in the dry season and in the rainy season downstream of the Boulanger mining sector, this study gives a first assessment of the spatial and seasonal variability of Hg exported from a re-exploited site. Results showed that, even in the absence of gold mining activity (during the dry season in September 2012, activity on the CMB concession was low), Hg was exported to the rivers downstream. Furthermore, when previously mined areas (with the use of Hg for amalgamation) are re-exploited, exports tended to be greater and were amplified during the rainy season.

The present work confirms previous observations reporting that soil erosion, whether due to deforestation for agricultural purposes, or soil excavation for goldmining is likely to contribute to Hg mobilization from soil to the water shed. Although the ban of Hg for amalgamation in 2006 put a stop to additional Hg input in legal mining exploitations, recent mining exploitations on sites where gold was mined during the last century by amalgamation imply a higher risk of re-mobilizing "inherited" mercury.

**Author Contributions:** Conceptualization, J.H., J.S., and V.L.; funding acquisition, J.H. and V.L.; investigation, J.H., R.V., L.L., and V.L.; project administration, V.L.; writing—original draft, J.H. and V.L.; writing—review and editing, J.S. and V.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the French Guiana Water Office and DEAL in the frame of the PRSE2 Guyane (Second programme for the promotion of Health and environment).

**Acknowledgments:** We thank Philippe Matheus, manager of the Boulanger Mining Company (CMB) for giving us access and authorization to collect samples from the CMB mining concession. We also thank Erwan Manach from CMB for helping with access to the sampling spots; Manuel Moisan and Bernard Joseph for their help with the fieldwork; and Cécile Bossy for her help in data acquisition.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

## **Accumulation of Methylmercury in the High-Altitude Lake Uru Uru (3686 m a.s.l, Bolivia) Controlled by Sediment E**ffl**ux and Photodegradation**

**Stéphane Guédron 1,2,\*, Dario Achá 3,\*, Sylvain Bouchet 4, David Point 3,5, Emmanuel Tessier 4, Carlos Heredia 3, Stéfany Rocha-Lupa 3, Pablo Fernandez-Saavedra 3, Marizol Flores 2,5, Sarah Bureau 1, Israel Quino-Lima <sup>2</sup> and David Amouroux 4,\***


Received: 14 October 2020; Accepted: 4 November 2020; Published: 9 November 2020

**Abstract:** In shallow aquatic environments, sediment is a significant source of monomethylmercury (MMHg) for surface water (SW). High-altitude aquatic ecosystems are characterized by extreme hydro-climatic constraints (e.g., low oxygen and high UV radiation). We studied, during two seasons, the diel cycles of MMHg in SW and sediment porewaters (PW) of Lake Uru Uru (3686 m a.s.l, Bolivia) contaminated by urban and mining activities. Our results show that diel changes in SW MMHg concentrations (up to 1.8 ng L<sup>−</sup>1) overwhelm seasonal ones, with higher MMHg accumulation during the night-time and the dry season. The calculation of MMHg diffusive fluxes demonstrates that the sediment compartment was the primary source of MMHg to the SW. Most MMHg efflux occurred during the dry season (35.7 <sup>±</sup> 17.4 ng m−<sup>2</sup> day−1), when the lake was relatively shallow, more eutrophicated, and with the redoxcline located above the sediment–water interface (SWI). Changes in MMHg accumulation in the PWs were attributed to diel redox oscillations around the SWI driving both the bacterial sulfate reduction and bio-methylation. Finally, we highlight that although MMHg loading from the PW to the SW is large, MMHg photodegradation and demethylation by microorganisms control the net MMHg accumulation in the water column.

**Keywords:** monomethylmercury; water–sediment interface; diel and seasonal cycles; photodegradation

#### **1. Introduction**

Lake sediment is known to be an important source of mercury (Hg) species, including the neurotoxic monomethylmercury (MMHg) for overlying waters [1], through their diffusion and advection under dissolved and colloidal phases [2–6]. Hg methylation mainly occurs in sub to anoxic surficial sediments or biofilms developed at the sediment–water interface (SWI), implying various microorganisms such as

methanogens, sulfate-reducing (SRB), and iron-reducing (IRB) bacteria [7–10]. In shallow environments (e.g., shallow lakes, lagoons, ponds), both the shear stress and the bio-irrigation driven by tides, wind, and benthic organisms enhance the upward release of elements contained in the sediment porewater towards the water column [2–6,11–13]. Such MMHg exchange is mainly regulated by the position of the redox front around the SWI where it is mainly produced, the sediment structure (e.g., porosity, tortuosity), and the abundance and type of (in)organic ligands [14,15] and colloids [16].

In high-altitude shallow lakes, extreme hydro-climatic constraints (e.g., high UV radiation and low oxygen) result in large diel variations of the water column's biogeochemical and physical parameters that can overwhelm seasonal amplitude [17]. This is the case in Lake Uru Uru (3686 m above sea level (a.s.l.)), one of the most Hg-contaminated lakes of the Bolivian Altiplano [18,19]. Previous studies [18,19] have shown that urban and mining discharges to the lake, combined with the strong diurnal biogeochemical gradients, have resulted in large diel variability in MMHg levels (0.2 to 4.5 ng L<sup>−</sup>1) in the water column. The quantification of methylation potential during incubation experiments with enriched isotopes has suggested that sediments could be the major source of MMHg for the water column [18]. Nevertheless, the sediment–water exchange has not yet been assessed.

In this work, we studied Hg and MMHg sediment–water column exchanges in Lake Uru Uru during two 36-h cycles performed at the end of the wet (May) and at the end of the dry (November) seasons of 2014. Hg speciation was measured in the water column (i.e., filtered, particulate, and dissolved gaseous Hg) and in the surficial sediment porewater. We also calculated Hg and MMHg diffusive fluxes at the SWI to assess the sediment contribution in MMHg effluxes toward the water column. The final objective was to identify the biogeochemical and physicochemical factors influencing these sediment–water exchanges and to evaluate those regulating the accumulation of MMHg in the water column.

#### **2. Materials and Methods**

#### *2.1. Lake Uru Uru General Settings*

Lake Uru Uru (3686 m a.s.l.) is located in the central part of the Bolivian Altiplano region (Figure 1), south of Oruro city. This shallow aquatic ecosystem is a human-made reservoir mainly supplied by the Rio Desaguadero waters. Depending on the season, the lake surface and depth vary between 120 and 350 km<sup>2</sup> and from 0.25 to 1 m, respectively, representing a change in lake volume up to 12 times between the two seasons.

Historical and current mining (e.g., Au, Ag, and Sn) and smelting activities in the region have led to the contamination of the lake by various metals, including Hg [19,20]. In addition, the recent urban development has resulted in the almost permanent eutrophication of this shallow aquatic ecosystem [21]. A sharp contrast exists between the northern and southern parts of the lake, with a higher density of sedge (*Schoenoplectus totora*), grass (*Ruppia* spp.), and algae (Characeae) in the southern part [22]. Previous investigations demonstrated that the diel variability (e.g., temperature, oxygen) in the lake could overwhelm seasonal variations [18]. Furthermore, the synergistic effect of mining (i.e., acid mine drainage (AMD)) and urban effluents from both the Oruro and Huanuni areas (Figure 1) to this shallow lake were proven to enhance Hg methylation, with the sediment being the main source of MMHg during the dry season [19].

**Figure 1.** (**a**) Study site location: location of Lake Uru Uru in the Bolivian Altiplano and (**b**) sampling site (UU12) in the southern part of the lake.

#### *2.2. Sampling Location, Strategy, and Elemental Analysis*

Surface water and sediment porewater samples were collected during two campaigns performed in May (end of the wet season) and November (end of the dry season) 2014. Surface water was sampled with a Teflon-coated Go-Flo trace metals sampler. Samples were then stored unfiltered or filtered (0.22 μm PVDF) in acid-washed FEP Teflon containers and acidified (except for dissolved gaseous mercury samples) with HCl (0.5%, *v*/*v*, Ultrex grade—Baker). Surface sediment porewaters (0–5 cm) were sampled with 5 cm long microporous polymer tube samplers (RhizonCSS®, 0.2 μm porosity; Rhizosphere Research Products, the Netherlands) over ~4 h, following published protocols [3,23].

Dissolved gaseous mercury (DGM) was purged from unfiltered water samples on-site within ~0.5 h after sampling and collected on gold-coated sand traps [24]. DGM was analyzed by thermodesorption (600 ◦C) and cold vapor atomic fluorescence spectrometry (CV-AFS) [24]. Filtered (F) and unfiltered (UNF) total mercury concentrations (THg) in surface and pore waters were determined by cold vapor atomic fluorescence spectrometry (CV-AFS) after conversion of all mercury species into Hg0 followed by detection using a Tekran® (Model 2500, Canada) [25,26]. Filtered monomethylmercury (MMHgF) concentrations were analyzed using an ethylation purge and trap-gas chromatograph-AFS analyzer (MERX System, Brooks Rand®, Seattle, WA, USA) following published protocols [27,28]. All samples were run in duplicate and quantified using the standard addition technique. Unfiltered monomethylmercury (MMHgUNF) concentrations were analyzed by propylation ID-GC–ICPMS (S.I.2.2) following published protocols [4,29,30].

Dissolved organic carbon concentrations (DOC) were determined using a Non-Dispersive Infra-Red (NDIR) CO2 Shimadzu® (Model VCSN, Japan) spectrometer after humid oxidation in a sodium persulfate solution at 100 ◦C. Major elements (Ca, Si, K, Na, Mg, and S) in filtered water (SW and PW) were measured by ICP-AES (Varian model 720ES) within the analytical chemistry platform of ISTerre (OSUG-France) and trace elements (Fe, Al, Mn, V, Sb) by ICP-MS (Nexion, Perkin Elmer) within the analytical chemistry platform of Univ. de Pau et des Pays de l'Adour (UPPA). Anions were measured by ionic chromatography using a 332 Metrohm apparatus and its accuracy was evaluated with a Carl Roth 2668.1 standard. Hydrogen sulfide (H2S) was determined using a

modification of the diamine method described elsewhere [31,32] and analyzed with HPLC (Agilent ) after calibration with a Radiello® solution (Code 171, Italy) for H2S [28].

A submersible multiparameter probe (HI-9828, Hanna Instruments, Woonsocket, RI, USA) was used to characterize the water physico-chemistry (pH, conductivity, dissolved oxygen concentration and saturation, and temperature). Data collection was conducted every three seconds for two minutes at the surface (0–10 cm) and above the sediment–water interface (SWI) at each time of sampling [28].

#### *2.3. Sediment–Water Flux Measurement and Calculation*

Molecular diffusion from sediment to overlying water was estimated through the calculation of diffusive flux for Hg(II) (= THg − MMHg) and MMHg at the SWI using the measured concentration gradient and Fick's first law [33,34]:

$$\text{Jsend} = \text{-}\mathcal{O} / \text{D} \text{sed} \ (\partial \mathbf{C} / \partial \mathbf{z})$$

where Jsed (ng.m−<sup>2</sup> d<sup>−</sup>1) is the diffusive flux from the sediment, Ø is the porosity (Ø = 0.78 [20]), Dsed is the diffusive coefficient in sediment (2 <sup>×</sup> <sup>10</sup>−<sup>6</sup> cm<sup>2</sup> <sup>s</sup>−<sup>1</sup> and 1.2 <sup>×</sup> <sup>10</sup>−<sup>5</sup> cm<sup>2</sup> <sup>s</sup>−<sup>1</sup> for Hg(II) and MMHg ionic/molecular diffusion coefficient in water [35]), and ∂C/∂z is the linear concentration gradient through the SWI. Positive Jsed indicates an upward-directed flux (efflux from the sediment into the overlying water column), and negative Jsed indicates a downward-directed flux (influx from the water column into the sediment).

#### **3. Results**

#### *3.1. Diel and Seasonal Variations of Mercury Species in Surface Water*

The water column was oxic during both seasons, with the lowest oxygen levels (O2 = 76 ± 23%saturation) and redox potential values [i.e., oxydo-reduction potential (ORP) = −8 to −72 mV] found during the dry season compared to the wet one (O2 = 93 ± 17%sat), where redox (i.e., ORP = 192 to −115 mV) was only negative during the night, resulting from algal respiration and sulfate reduction (H2S <sup>=</sup> <sup>51</sup> <sup>±</sup> <sup>42</sup> <sup>μ</sup>g L<sup>−</sup>1, Figure 2a,b).

Dissolved gaseous mercury (DGM = 11.2 <sup>±</sup> 9.9 pg L<sup>−</sup>1) in surface water (SW) was mainly in the form of Hg(0) [19] and was overall low and never exceeded 1% of THgUNF for both seasons (Figure 2a), similar to previous studies in Lake Uru Uru [19] and Lake Titicaca [24] but in the lowest range of values reported for temperate freshwater lakes [36,37]. Higher DGM levels were found during the dry season (21.0 <sup>±</sup> 7.5 pg L<sup>−</sup>1), which were around four times higher than during the wet one (5.6 <sup>±</sup> 5.5 pg L<sup>−</sup>1).

Average unfiltered (UNF) concentrations of THg (3.3 <sup>±</sup> 1.0 ng L<sup>−</sup>1) and MMHg (0.6 <sup>±</sup> 0.2 ng L<sup>−</sup>1) were 3 to 10 times higher than those measured in Lake Titicaca [2,24] reflecting the local mining contamination and high MMHg production in this ecosystem [18,19]. Both the average filtered (F) and unfiltered (UNF) THg and MMHg levels in surface water (Figure 2c,d) were similar between the two seasons. In contrast, filtered alkali and alkaline-earth metal concentrations (e.g., Li, Na, K, Mg, Ca, and Sr) and DOC increased by two to three times during the dry season due to the evaporation and concentration of ions in water (Figure 2b and Figure S1).

The diel variability of F and UNF MMHg concentrations overwhelmed the seasonal one, with the highest values measured during the night-time when oxygen levels dropped (Figure 2a), resulting from the aquatic ecotopes' respiration. MMHg/THg percentages were high in surface waters for both seasons. However, they presented larger diel variations, with lower average values during the dry season (i.e., 35 ± 20% and 26 ± 7% for F and UNF, respectively) compared to the wet one (i.e., 42 ± 4% and 35 ± 6%, Figure 2e).

**Figure 2.** (**a**) Dissolved gaseous Hg (DGM), dissolved oxygen saturation percentage (%sat), and temperature (T ◦C); (**b**) dissolved organic carbon (DOC), hydrogen sulfide (H2S), and redox potential (ORP); (**c**) filtered (F, empty symbols) and unfiltered (UNF, black symbols) THg concentrations; (**d**) filtered (F, empty symbols) and unfiltered (UNF, dark cyan symbols) MMHg concentrations; and (**e**) percentage MMHg over the THg concentrations in the F (empty symbols) and UNF (dark cyan symbols) in surface water during the wet (left panels, 20–21 May 2014) and the dry (right panel, 18th to 19th November 2014) seasons. Lines represent average concentrations for F (dotted line) and UNF (dashed line) species and ratios.

#### *3.2. Diel and Seasonal Variations of Mercury Species in Surface Sediment Porewater*

Filtered THg (8.8 <sup>±</sup> 3.3 ng L−1) and MMHg (1.7 <sup>±</sup> 1.1 ng L−1) levels in surface sediment PW (Figure 3a) were 2 to 8 times higher than those of SW. In contrast to SW, large seasonal differences were observed in the PW, with higher MMHg concentrations during the dry season (3.0 <sup>±</sup> 1.2 ng L<sup>−</sup>1) compared to the wet one (1.0 <sup>±</sup> 0.4 ng L−1, Figure 3b). Percentage MMHg/THg (Figure 3b) was also

higher during the dry season (28.1 ± 11.0%) compared to the wet one (15.5 ± 8.8%), highlighting a higher accumulation of MMHg in surface sediment PW during the dry season.

**Figure 3.** (**a**) Filtered (F) THg (empty circles) and MMHg concentrations (dark cyan triangles up) and (**b**) percentage MMHg over the THg concentrations (dark cyan triangle down), dissolved organic carbon (DOC, semi-filled dark yellow circles), and hydrogen sulfide (H2S, grey circles) concentrations in sediment porewater during the wet (left panels, 20–21 May 2014) and the dry (right panel, 18th to 19th November 2014) seasons. Lines represent average concentrations for F (dotted line) and MMHg/THg (dashed line) concentrations and ratios.

Diel variations were large for THg, but no specific trend was identified except for a rising trend in concentrations during the day for the dry season. In contrast, MMHg exhibited similar trends as for SW, with rising concentration during the night for the dry season, while they did not vary significantly during the wet season.

DOC concentrations in PWs (26 <sup>±</sup> 5 mg L−1, only measured during the dry season) were, on average, 1.5 times higher than those of SW but exhibited a similar diel trend, suggesting that equilibrium was established between these two compartments. Sulfide concentrations were an order of magnitude higher than the one in SW, with at least three times higher average concentrations during the dry season (39.9 <sup>±</sup> 7.2 mg L<sup>−</sup>1) compared to the wet one (12.1 <sup>±</sup> 9.0 mg L<sup>−</sup>1, Figure 3b).

#### **4. Discussions**

#### *4.1. DOC and Sulfides Drive MMHg Accumulation in Sediment PW and E*ffl*uxes Towards the SW*

The high MMHg concentrations found in PW are consistent with a previous study in Lake Uru Uru, which showed evidence that methylation was mediated mostly by SRB in the absence of sunlight [18]. Such nutrient-rich, alkaline waters with elevated levels in DOC and sulfates were also reported to favor the activities of anaerobic bacterial communities [4,23,24,38]. During the wet season, the gradual rise in MMHg concentration in PW during the night illustrates the progressive installation of anoxia in surface sediment PW due to the rising respiration of the aquatic ecotopes, which is also reflected by the drop in dissolved oxygen in overlying waters (Figure 4a). This pattern is exacerbated during the dry season as a result of more eutrophic conditions, which enhance MMHg production and accumulation in the sediments [39]. During this latter season, both DOC and sulfide concentrations were positively correlated (R = 0.56, *p* < 0.05), supporting the notion that active sulfate reduction in PW favors DOC release in PWs during the reduction of Fe-oxy(hydr)oxides [40].

**Figure 4.** (**a**) Redox potential (ORP), dissolved oxygen saturation percentage (%sat), and conductivity in overlying waters (OW) above the sediment–water interface (SWI); (**b**) filtered Hg(II) (=THg − MMHg) and MMHg diffusive fluxes (Jsed) at the SWI and (**c**) partitioning of THg and MMHg between filtered and particulate fractions in surface water, during the wet (left panels) and the dry (right panels) seasons. Lines represent averages for Jsed (dotted line) and for MMHg/THg ratios (dashed line).

Interestingly, positive correlations were also found between sulfide and MMHg concentration (R = 0.73, *p* < 0.01, Figure S3) or percentages MMHg/THg (R = 0.66, *p* < 0.01), and between percentages MMHg/THg and DOC (R = 0.66, *p* < 0.01). This suggests that filtered MMHg was preferentially bound to sulfurized dissolved organic matter (DOM). In addition, the positive correlation found between sulfide and THg concentration (Figure S2) supports a limited precipitation of inorganic Hg with free sulfide or with FeS, which would have restricted the availability of Hg(II) for bio-methylation [41–43]. Hence, this rather supports the presence of Hg-SR complexes, which were reported to increase the bioavailability of Hg(II) for bio-methylation [14,15]. In contrast, sulfide was negatively correlated with chalcophile redox-sensitive elements (i.e., Fe and V, *p* < 0.01, respectively, Figures S2 and S3), supporting the notion that the sulfate reduction in anoxic PWs favors their precipitation as authigenic sulfide minerals.

The calculation of diffusive fluxes (Jsed, Figure 4b) for MMHg and HgII (=THg − MMHg) confirms both the seasonal and diel patterns observed in the PW and SW (Figures 2 and 3). Daily Jsed for HgII (16.2 <sup>±</sup> 7.3 ng m−<sup>2</sup> day−1) and MMHg (35.7 <sup>±</sup> 17.4 ng m−<sup>2</sup> day−1) were higher during the dry season compared to the wet one (13.3 <sup>±</sup> 9.1 ng m−<sup>2</sup> day−<sup>1</sup> and 5.4 <sup>±</sup> 7.8 ng m−<sup>2</sup> day−<sup>1</sup> for HgII and MMHg, respectively). These MMHg diffusive fluxes are much lower than those measured in the carbonate-rich sediments of Lake Titicaca (36 to 224 ng m−<sup>2</sup> day−1) but in the range of those reported for the organic-rich sediment of the same lake (13 to 60 ng m−<sup>2</sup> day−<sup>1</sup> [2]) and in coastal environments [3,4,6].

Strikingly, MMHg Jseds were higher than those of HgII during the dry season, whereas the opposite was found for the wet season. This greatly illustrates the major MMHg accumulation in the surface sediment PW and the upward MMHg efflux towards SW during the dry season. In addition, the positive correlation found between sulfide and MMHg Jsed (R = 0.70, *p* < 0.01, Figure S3) supports the notion that the upward shift of the redox gradient above the SWI during the dry season has likely favored the diffusion of DOC-bound MMHg or MMHg-SR complexes in the overlying water [14]. In contrast, the lower MMHg Jseds during the wet season likely result from the downward shift of the redox stratification in the epibenthic layers and/or the sediment. Indeed, higher MMHg Jsed during the night-time could be explained by a shallower redox front when the aquatic ecotope respiration dominates and favors upward Hg diffusion [2,44]. In contrast, the macrophytes and algae photosynthesis during the day may favor the water column oxygenation and the scavenging of MMHg in the Fe redox loop and onto submerged macrophytes' periphyton or benthic biofilms and algae [45,46], explaining negative Jsed during the day-time [4,47].

Hence, the stratification within the water column during the dry period likely favored upward MMHg effluxes from the sediment, which was also probably enhanced by the advection of surface PW by wind-induced shear stress in this shallow system as wind was on average 40% higher in November than in May (Figure S4).

#### *4.2. Diel and Seasonal Changes in the Hg Partitioning between Dissolved and Particulate Phases in SW*

Previous studies reported limited net Hg methylation potentials in the water column, whereas elevated ones were found in floating bio-organic substrates and surface sediment of Lake Uru Uru [18], supporting elevated MMHg production into fully anoxic environments or surface water microenvironments [12,48,49]. Although average MMHg concentrations were similar between the two seasons, the lower average percentages of filtered and unfiltered MMHg/THg with larger diel variations during the dry season compared to the wet one (i.e., 42 ± 4% and 35 ± 6%) can be explained by changes in water column biogeochemical conditions (i.e., dissolved oxygen, temperature and redox oscillations, and pelagic productivity or mobility), resulting in changes in the partitioning between the suspended solids and the solution. Alternatively, changes in the incident light radiation (see Section 4.3) between the two seasons may explain the seasonal differences.

During the wet season, average percentages of filtered (<0.22 μm) THg and MMHg in surface water were 58 ± 21% and 71 ± 23% of the UNF one, respectively, highlighting that both species were mainly present in the truly dissolved or colloidal phases (Figure 4c). Strikingly, sharp drops in the percentage of F/UNF THg and MMHg fractions (i.e., from ~80 to ~30%, Figure 4c) were found during the night, counterbalanced by the rise in both species onto particles synchronously with the drop in redox to negative values (Figure 2b). These percentages of F/UNF THg and MMHg were found to decrease with rising sulfides (R = −0.64 and −0.62, *p* < 0.05, for THg and MMHg, respectively) or DOC concentrations (R = −0.62 and −0.55, *p* < 0.05, for THg and MMHg, respectively) supporting the notion that DOC-bound MMHg or MMHg-SR complexes released from surface PW or produced in anoxic niches were sorbed onto suspended inorganic and organic particles. During the onset of reducing conditions, Mn concentrations (Figure S1) rose in the water column as a result of the dissolution of Mn oxides [50], whereas chalcophile redox-sensitive elements (i.e., Fe and V; Figure S1) decreased, likely due to their (co)precipitation as or onto authigenic sulfide minerals [51]. Sulfate concentrations (5.4 <sup>±</sup> 5.1 g L<sup>−</sup>1) were at least an order of magnitude higher than those of Fe, and thus the production of sulfides drives the Fe levels in the SW. Hence, drops in THgF and MMHgF levels during reducing conditions found in the night-time period likely resulted from their adsorption or (co)precipitation onto/with colloidal and particulate FeS [2], and onto sulfurized OM [52]. They likely enriched the particulate pool through the flocculation of dissolved OM and the aggregation of colloids when SW became reducing [53,54].

During the dry season, Hg species in the reducing SW showed the opposite trend as for the wet season. THg and MMHg were mostly bound to particles (60 ± 13% and 50 ± 25% for particulate THg and MMHg, respectively), with higher partitioning onto particles during the day (Figure 4c). Such an opposite trend likely results from the more reducing, oxygen depleted, and sulfide-rich SWs, which favored both the MMHg production, upward efflux from the PW, and its preservation in the truly dissolved or colloidal form during the night in the dry season. In particular, the drop in oxygen was more accentuated during the night in the dry compared to the wet season, possibly resulting from a higher decomposition activity, triggered in turn by a higher productivity during the day. In addition, the higher content of algae and bacteria (both living or dead) in the SW during the dry season could have enhanced Hg and MMHg entrapment during the day as these organisms are known to accumulate MMHg in lake water columns [28,45,46,55]. Indeed, higher radiation levels and SW temperatures have increased biological activities, as illustrated by higher decreases in oxygen saturation (Figure 2a), which could have led to higher active uptake of MMHg by algae during the day [56]. Again, this agrees with reported abundant bio-organic aggregates and/or suspended particles enriched in OM in the water column during the dry season in Lake Uru Uru [18]. Thus, during the nights of the dry season, more MMHg is produced, whereas more uptake and potential demethylation occurs during the day. In addition, algae activity could have facilitated Hg reduction [57], explaining an increase in DGM during the day in the dry season (Figure 2a).

#### *4.3. Solar Radiation Influence on MMHg Photodegradation and DGM Production in Surface Water*

Because solar ultraviolet (UV-B) increases by around 10 ± 20% for every 1000 m increase in elevation [58,59], UV radiation intensity at Lake Uru Uru (3686 m a.s.l) is expected to be at least 40 to 80% higher than that measured at sea level [18,60]. The large diurnal and seasonal temperature amplitude gradients between the dry season (12.2 to 20.7 ◦C) and the wet one (10.0 to 13.3 ◦C, Figure 2a) broadly reflect these gradients in solar radiation.

For both seasons, filtered and unfiltered MMHg/THg decreased with increasing temperature, with the exception of the filtered phase during the wet season (Figure 5). This trend illustrates the strong influence of UV radiation on MMHg photodegradation [61] at both the diel and seasonal scales, with enhanced degradation of MMHg during the day and the dry season [18]. One cannot exclude higher activation of demethylating bacteria at higher temperatures [62].

**Figure 5.** Plots of percentage MMHg/THg in (**a**) the filtered and (**b**) the unfiltered phases vs. surface water temperature. Crossed symbols indicate samples collected during the night-time period.

In addition, a positive correlation was found between DOC concentrations and percentages MMHg/THg (R = 0.66, *p* < 0.01) during the dry season, supporting the notion that rising DOC concentrations decreased the efficiency of MMHg photodegradation [63,64]. Although DOC concentration was lower during the wet season, the absence of a relationship between MMHg and SW temperature or DOC for the filtered fraction suggests that photodemethylation was not a primary sink for MMHg during this season. Indeed, lower insolation, higher water column, possible changes in the composition of photo-reactive DOC species, and/or different MMHg-DOC bindings [65] could have attenuated the radiation effect and reduced MMHg photodegradation efficiency. In particular, higher molecular weight dissolved OM, and a lower abundance of thiol functional groups (RS-) during the wet season compared to the dry one, could have attenuated MMHg photodegradation [66,67]. As mentioned previously, one cannot exclude the possibility of lower biological activity in the wet compared to the dry season, which would also have resulted in lower MMHg biodegradation.

As for MMHg, the overall low DGM (i.e., Hg(0)) concentrations in SW corroborate results from in situ incubations with Hg-enriched isotopes, which reported low biological and photochemical production of DGM in Lake Uru Uru [18]. Although DGM concentrations were low, the higher concentrations found during the dry season likely result from higher solar radiation, higher concentrations of suspended solids, as well as reducing bacteria/algae. Although solar radiations are the main fuel for DGM production during the day, rises in DGM found during the night support the notion of significant bacterial reductions in the dark [57].

#### **5. Conclusions**

Our results confirm previous observations made in Lake Uru Uru showing that diel changes in MMHg concentration overwhelm seasonal ones, with higher MMHg accumulation in surface water during the night-time and the dry season. We here bring new insights into the key role of the sediment compartment as a major source of MMHg for the SW, with a larger contribution in MMHg efflux towards the water column during the dry season when the lake is relatively shallow, more eutrophicated, and with the redoxcline likely moving above the SWI. Our results suggest that MMHg accumulation in the PWs fluctuates depending on diel redox oscillations around the SWI, driving both the bacterial sulfate reduction and bio-methylation. We here confirm the previous hypothesis from enriched isotope incubation that MMHg is preferentially produced in the PW and supports its release toward the SW associated with dissolved/colloidal OM or organic thiols. In the water column, both the pelagic light-dependent organisms and the flocculation/aggregation of dissolved OM with colloids drive the MMHg partitioning between dissolved and particulate phases during diel redox oscillations. Finally, although elevated MMHg diffusive fluxes demonstrate that MMHg loading from the porewater to the water column is large, both the photochemical and biological methylmercury degradation control the net MMHg accumulation in the water column.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3417/10/21/7936/s1, Figure S1: Seasonal and diel cycles for vanadium (V), manganese (Mn), calcium (Ca); potassium (K), magnesium (Mg), sodium (Na), total sulfur (S), and silicon (Si) in surface water. Figure S2: Scatter plot with linear regression for porewater concentrations of (a) manganese (Mn) vs. hydrogen sulfide (H2S) and dissolved oxygen (O2—measured in epibenthic water above the SWI with a multiparameter probe); (b) iron (Fe) and vanadium (V) vs. H2S; (c) Filtered total mercury THg vs. H2S and V. Linear regression was performed for all data (both seasons). Figure S3: Scatter plot for (upper panels) PW MMHg concentrations vs. hydrogen sulfide (H2S) and vanadium (V), and (bottom panels) MMHg diffusive fluxes (Jsed MMHg) vs. (H2S) and redox potential (measured in epibenthic water above the SWI with a multiparameter probe). Figure S4: Average precipitation and wind velocity at Oruro during the period 2010 to 2016 [68].

**Author Contributions:** Conceptualization, D.A. (David Amouroux), S.G., D.P., D.A. (Dario Achá), and S.B.; methodology, S.G., D.A. (David Amouroux), D.P., S.B., and D.A. (Dario Achá); validation, S.G., D.A. (Dario Achá), D.P., S.B., and D.A. (David Amouroux); methodology, S.G., D.A. (Dario Achá), D.P., S.B., and D.A. (David Amouroux); formal analysis, S.G., D.A. (Dario Achá), S.B., D.P., E.T., S.R.-L., C.H., P.F.-S., M.F., and S.B.; Field investigation, S.G., D.A. (David Amouroux), D.A. (Dario Achá), S.B., D.P., E.T., S.R.-L., C.H., P.F.-S., M.F., I.Q.-L., and S.B.; data curation, S.G.; writing—original draft preparation, S.G.; writing—review and editing, S.G., D.A. (Dario Achá), S.B., and D.A. (David Amouroux); supervision, S.G., D.A. (Dario Achá), S.B., and D.A. (David Amouroux); project administration, D.A. (David Amouroux) and S.G.; funding acquisition, S.G., D.A. (Dario Achá), S.B., and D.A. (David Amouroux). All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Agence National pour la Recherche, as a contribution to the LA PACHÁMAMA project (ANR CESA program, No ANR-13-CESA-0015-01, PI: D. Amouroux: david.amouroux@univ-pau.fr), COMIBOL project (INSU CNRS/IRD EC2CO Program, PI: D. Point: david.point@ird.fr), EUTITICACA project (founded by the Impuestos Directos a los Hidrocarburos IDH

administrated by the Universidad Mayor de San Andrés, PI: D. Achá: darioAchá@yahoo.ca) and TRACISOMER supported by a grant from Labex OSUG@2020 (PI: S. Guédron: stephane.guedron@ird.fr). S. Guédron and S. Bureau (ISTerre/IRD/UGA) are part of Labex OSUG@2020 (Investissements d'avenir—ANR10 LABX56).

**Acknowledgments:** We wish to thank J. Gardon, A. Terrazas, C. González, N. Clavijo, L. Salvatierra, R. Rios, J.C. Salinas, A. Castillo, M. Claure, and Don German Calizaya (Fishermen Association, MAchácamarca, Bolivia) for their help and assistance during the field campaigns.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Influence of Ecological Restoration on Mercury Mobility and Microbial Activities on Former Guyanese Mining Sites**

**Ewan Couic 1,\*, Vanessa Alphonse 2, Alexandre Livet 2, Stéphanie Giusti-Miller <sup>3</sup> and Noureddine Bousserrhine 2,\***


**Abstract:** As rehabilitation efforts in Guyana are recent, there is little information on the effect of different ecological rehabilitation protocols for Guyana's mining sites on biogeochemical cycles and mercury mobility. This study was conducted to assess the impact of different ecological restoration protocols on soil quality with the use of soil microbial indicators and by estimating the mercury mobility. We sampled soil from six rehabilitated mining sites in French Guyana with different ecological restoration procedures. We carried out measurements of enzymatic activities and an analysis of mercury environmental speciation to assess its potential toxicity according to a mobility gradient. The results obtained in this study show that the rehabilitation of mining sites has been carried out in a heterogeneous manner and soil quality is very variable, even in nearby sites. Sites that have been rehabilitated with fabaceous species have positive soil quality indicators. In addition, the results highlight a change in mercury mobility that is 82.1% correlated after co-inertia analysis with soil texture properties, which also confirms a direct effect of rehabilitation on mercury mobility. The non-restored sites had a much higher potential of mercury mobility and toxicity than the sites where ecological restoration was successful. These results highlight the positive effect of controlled rehabilitation and ecological restoration on microbiological activities and the potential toxicity of mercury.

**Keywords:** ecological restoration; mercury mobility; microbial activities; biogeochemistry; gold mining activities; French Guiana

#### **1. Introduction**

Deforestation is currently one of the sectors that emit the most greenhouse gases [1,2] and although forest areas are increasing globally, tropical forests are particularly affected by this crisis [3]. Among the causes of deforestation in the Amazon, agriculture, silviculture, cattle ranching, selective logging, coca farming, and artisanal scale gold mining (ASGM) are responsible for a large percentage of forest loss. In South America and in French Guiana in particular, the stakes of gold mining are serious for the development of the region, and are worrying for the preservation of the ecosystem. In Guyana, gold mining impacts land use and the functioning of the forest system, including the soil [4,5]. To exploit gold in alluvial terraces, it is necessary to remove all the vegetation and then to use powerful water jets (sluice) to eliminate the superficial layers of soil until reaching the gold-bearing layer [6]. In addition to the direct damage caused by deforestation, these practices lead to heavy soil losses through erosion, which will lead to increased turbidity in aquatic systems and the remobilization of toxic metallic trace elements such as mercury [7]. Due to amalgamating chemical properties for gold mining, mercury was widely and legally

**Citation:** Couic, E.; Alphonse, V.; Livet, A.; Giusti-Miller, S.; Bousserrhine, N. Influence of Ecological Restoration on Mercury Mobility and Microbial Activities on Former Guyanese Mining Sites. *Appl. Sci.* **2021**, *11*, 2231. https://doi.org/ 10.3390/app11052231

Academic Editor: Stéphane Guédron

Received: 27 January 2021 Accepted: 26 February 2021 Published: 3 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

used until 2006 and the problem of mercury in Guiana has already been described in many scientific studies [8,9]. It was shown that mercury of anthropogenic origin in Guyana was the most reactive, highly disorganized and often had suboxic physico-chemical conditions of the mining sites, which are favored its mobility and methylation [10,11].

The biogeochemistry of mercury is complex and depends on many physico-chemical parameters (i.e., temperature, redox potential, pH, metal oxides, organic matter, and mercury oxidation level) [12], but also biological parameters. Depending on these conditions, the forms of mercury in the soil will be affected and this will influence its mobility, bioavailability and toxicity to biological systems [13]. Free forms of mercury take different forms in soils such as alkyl mercury species such as methylmercury (MeHg+) or ethylmercury(II) (EtHg+) or inorganic soluble mercury species. These forms generally represent the most toxic component of mercury due to their high mobility. Metal mercury complexes, or certain amalgams, are considered to be less toxic because they are more difficult to mobilize without specific reactions, such as chemical or microbial reduction. The least toxic forms of mercury, such as mercuric sulphide or cinnabar (HgS) [14,15], are generally stable in soils, but may be weathered by simple dissolution in acidic environments [15] by significant root exudates such as those from fabaceae [16] or by microbial production of low molecular weight organic acid (LMMOAs) [17]. Determining mercury speciation at anthropized sites such as former mining sites is therefore very important for assessing the positive or negative health impact of ecological rehabilitation [15].

In order to limit the adverse effects associated with mercury mobility, ecological rehabilitation methods have been proposed with varying degrees of success in the Guyana basin. In the mining code, there is a difference between the principle of rehabilitation and restoration or revegetation [18]. Rehabilitation in the mining sense consists solely of reconstituting the original soil from the materials that have been extracted and moved, terracing the mixture, and if possible, covering it with so-called fertile horizons. Restoration will consist of applying a defined site revegetation protocol, if possible using local species in order to rebuild the original ecosystem [19,20]. While rehabilitation is mandatory, restoration is not explicitly requested in the mining code and too often remains at the discretion of mining companies. Among the methods used to evaluate the quality of ecological restoration protocols, enzymatic activity measurements and quantification of soil microbial biomass are among the most effective and widely used methods, particularly at mining sites [21,22].

While data on mercury speciation in natural soils and mining sites are fairly well documented [9,23], changes in carrier phases can alter mercury mobility and these mechanisms during ecological restoration processes are still not well known in Guyana and are essential to define a chemical quality status of the soil. In a context where the physico-chemical conditions (iron oxides, low pH and suboxic conditions) of rehabilitated sites could promote mercury mobility [11], environmental speciation measures could therefore be an interesting tool for assessing the health quality of these plots. One of the additional effects of good ecological rehabilitation is also to limit anoxic soil conditions through better regulation of the water cycle. In the context of tropical soils that are potentially rich in mercury, ecological restoration could therefore limit the conditions conducive to mercury mobility. While the effect of ecological restoration has already been shown to be positive for soil biological quality [22,24], it would also be relevant to assess the impact of different vegetation cover on mercury speciation by the use of environmental extraction procedures [25].

The main assumptions led to his work on the quality of rehabilitation, which was linked to an increase in plant cover and soil stability and can limit mercury mobility and promote soil microbial activities. We also hypothesize that the toxic fractions of mercury in the soil have a negative impact on microbial communities. In this context, the main physicochemical properties of the soil, telluric enzymatic activities indicative of the functioning of biogeochemical cycles, as well as the environmental speciation of mercury were determined in different soil samples from former mining sites rehabilitated in French Guiana, restored or not to see the effect of a vegetation gradient. The objectives of this study were: (1) assess

the biological quality of mine sites in relation to the quality of rehabilitation and the level of ecological restoration; and (2) describe the influence of heterogeneous soil rehabilitation and different ecological restoration modalities on the fate and mobility of mercury in soils. To answer these questions, six sites undergoing ecological restoration have been selected in French Guiana, in the Yaoni and Belizon regions. These sites were chosen because of their similar gold mining liabilities, as well as for their ecological restoration gradients, both in terms of quality and age.

It is difficult to anticipate the different recovery trajectories of biogeochemical cycles on a wide range of rehabilitated mining sites; however, these results could contribute to improving the assessment of the sanitary quality of rehabilitated sites in French Guiana.

#### **2. Materials and Methods**

#### *2.1. Site Description and Soil Sampling*

The work was undertaken in French Guiana, South America. The sampled soils (Figure 1) for this study were rehabilitated after gold mining extraction. After closing the alluvional mine pit, rehabilitation consisted of reconstituting the soil excavated during mine exploitation.

**Figure 1.** Illustration of some examples of different stages of ecological restoration on formers Guyanese gold mining sites and geographical location of the Yaoni and Belizon mines in French Guiana. Non-restored site (**top left**), ecological restoration with herbaceous species after 3 years ((**top right**), Bel1), ecological restoration with fabaceous species after 3 years ((**bottom left**), Bel3), ecological restoration with fabaceous species after 18 years ((**bottom right**), Yao3). The pictures were taken during a second sampling campaign in April 2016 on sites undergoing ecological restoration.

> The first three sites were in the Belizon forest (N 04◦22.200 /W 052◦19.328 ). The mining operation lasted five years and ended in January 2013. All sites in the Belizon have had the same mining history. In February 2013, an ecological restoration programme was initiated with macrocuts of *Clitoria racemosa*, a species of local fabacea. The recovery of *C. racemosa* macrocuts has not been effective on all sites in the Belizon forest. On the Bel1 site, vegetation covered between 5% and 10% of the soil surface with a majority of cyperaceae species (*Cyperus* sp. and *Carex* sp.). On the Bel2 site, vegetation covered more than 70% of the site and almost the entire site is dominated by *Lycopodiella* sp. with the presence of some cyperaceae species. On the Bel3 site, macrocuts of *C. racemosa* were recovered and the trees are nearly 5 m high.

> The following three sites were taken from the former Yaoni mine site (N 04◦30.930 /W 052◦21.275 ). Mining activity at the Yaoni sites lasted several years and ended in 1997. These three sites also have the advantage of having the same mining history. The first two Yaoni sites (Yao1 and Yao2) have not been restored by specific protocols and the low vegetation cover, mainly *Lycopodiella* sp., *Cyperus* sp. and *Carex* sp., is the result of spontaneous vegetation development. The Yao3 site has been completely restored by associations of fabaceous species, notably *Acacia mangium* and *Clitoria racemosa*.

At each site, soil samples were collected from depths between 0–10 cm with an auger in April 2016. The sampling for the five sites consisted of sampling five plots (*n* = 5) per site, with three sub-samples per plot that were pooled at the laboratory. The sampling of five replicas per site was mainly conditioned by the topography of the terrain, the accessibility of the plots and the feasibility of sampling. Soil samples were immediately sealed in sterile hermetic polyethylene bags for transportation to the Cayenne IRD laboratory. Samples were then dried at ambient temperature (25 ◦C) until they were air dried (i.e., approximately three weeks). These 90 collected samples (6 sites × 5 plots × 3 samples per plot) were then sieved at 2 mm, homogenized, and hermetically sealed at 4 ◦C until used.

#### *2.2. Soil Samples Characterization*

Granulometry was determined in the fraction less than 2 mm, and five classes of particles have been distinguished according to the NF X 31-107 standard: clay (<2 μm), fine silts (2 to 20 μm), coarse silt (20 to 50 μm), fine sands (0.050 to 0.200 mm) and coarse sands (0.200 to 2 mm).

For the total Fe and Al analyses by inductively coupled plasma optical emission spectroscopy (ICP-OES, Spectroblue, Elancourt-France), the samples were first ground to 63 μm then an acid digestion was performed in cleaned Teflon digestion tubes (SCP Science, Villebon sur Yvette-France) using a hot block (Digiprep MS, SCP Science) with two digestion HNO3 (VWR ref. 83872.330)-HCL (VWR ref. 83871.290) in a 1:3 ratio and HF (Sigma-Aldrich, Saint Louis, MO, USA, ref. 339261-100 ML).

Soil pH-H2O and pH-KCl were measured in soil suspensions according to ISO 10390. Soil total carbon (Ctot) and total nitrogen (Ntot) were determined by the Dumas method (NF ISO 13878) by chromatography with a thermal conductivity detector (NA 1500 série 2 CARLO-ERBA). Total phosphorus (Ptot) was determined after the acid digestion of soil samples, by Murphy and Riley method [26]. Water extractions were performed for total organic carbon (TOC), and concentrations were measured with a Shimadzu TOC-500 apparatus (Shimadzu, Kyoto, Japan).

#### *2.3. Soil Microbial Biomass and Enzyme Assays*

The soil microbial biomass was estimated according to Anderson and Domsch's work [27]. Using the SIR method, soil microbial biomass C (SIR-biomass) was calculated using the following equation:

$$\text{SIR-biomass C (\mu g C g}^{-1} \text{ soil)} = \text{SIR (\mu L CO}\_2 \text{ g}^{-1} \text{ soil h}^{-1}) \times 40.04 + 0.37 \tag{1}$$

The activity of acid (AcdP) and alkaline (AlkP) phosphatases, B-glucosidase (Glu) and arylsulfatase (Aryl) were assayed based on the amount of p-nitrophenol (pNP) released after cleavage of enzyme-specific substrates at the average pH of natural soil. These enzymes were measured, as described by Badiane et al. and Tabatabay and Bremner [28–30]. Dehydrogenase activity (DHA) was determined according to Klein et al. [31] using 2.3.5\_tripenyltetrazolium chloride (TTC) as substrate. Urease activity (Ur) was determined according to Kandeler and Gerber [32] using urea as substrate. The total enzymatic activity (FDA) of soil microorganisms was determined by a spectrophotometric method using fluoresceine diacetate as substrate following the method developed by Green et al. [33].

#### *2.4. Mercury Analysis*

#### 2.4.1. Total Mercury Content

Mercury concentrations (HgT) in soil samples were determined directly by thermal decomposition atomic absorption spectrometry after gold amalgamation using an automatic mercury analyzer (AMA 254) in Ultra High Sensitivity configuration (ref 002-001-001) with a detection limit of 0.001 ng and quantification limit of 0.003 ng.

2.4.2. Environmental Speciation of Mercury

• Extraction of "Highly mobile" mercury species

This extraction included three fractions (water soluble, exchangeable and bioavailable mercury fraction) of the most mobile and bioavailable forms of mercury. The sum of the three fractions (F1 + F2 + F3) in the next three paragraphs will be called E1.

Soluble mercury (F1) was estimated according to Rasmussen et al. [34] with ultrapure water. The exchangeable Hg fractions (F2) in soil samples were measured according to Silveira et al. [35] with 0.1 M CaCl2 (Sigma-Aldrich, ref. 223506-500 G). The bioavailable mercury contents (F3) in soil samples were measured after a 1:10 (*w*/*v*) soil/0.01 M EDTA (Carl Roth, Karlsruhe-Germany, ref. 8040.3) + 1 M ammonium acetate (Merck, Darmstadt-Germany, ref. 1.01115.1000) following the standard AFNOR NF X31-120. Then, solutions were centrifuged, supernatants were filtered at 0.45 mm with Teflon filters (Minisart SRP 25, Sartorius, Dourdan-France), and then 200 μL of extracted supernatants were analyzed for exchangeable mercury content with AMA-254.

• Extraction of toxic, semi-mobile and non-mobile mercury

To assess the environmental speciation of mercury in soil, we followed Han et al. [14] procedure in order to determine the mobile and toxic, semi-mobile and non-mobile fractions of mercury in the restored soils and all fractions were analyzed by the AMA-254.

Briefly, the extraction of mobile and toxic mercury, called E2, involves the use of a solution of 2% HCl + 10% ethanol to extract the mobile mercury species from the soil. The target mercury species include toxic alkyl mercury species, such as MeHg<sup>+</sup> and EtHg<sup>+</sup> species, as well as inorganic mercury species that have great mobility, such as soluble Hg2+. As this extraction targeted all forms of mobile mercury, including the three fractions detailed above, extraction E2 corresponded to the difference between the total mobile mercury extracted by this protocol and the sum of the fractions of extraction E1.

To extract the "semi-mobile" mercury, called E3, species including mainly Hg and mercury-metal amalgam, the sample remaining after the ethanol extraction was first sonicated with 5 mL of ultrapure water at 60 ◦C for 5 min. The sample was centrifuged, and the supernatant was discarded. Then, extraction was carried out with 5 mL 1:2 (*v*/*v*) HNO3: Ultrapure water. The rest of the matric material could be directly analyzed by AMA-254 for "non-mobile" mercury species, called E4, including mainly Hg2Cl2 and HgS mercury species.

#### *2.5. Statistical Analysis*

The normality of the data distributions and the equal variance between treatments were tested using the Shapiro test and Bartlett's test, respectively. To study the effect of the restoration processes, we conducted different one-way ANOVA. Tukey's HSD multiple comparison method was used to test the difference between the different restoration processes. Pearson correlation was performed between mercury speciation and the dataset of enzyme activities and soil properties.

Co-inertia analyses, a two-table ordination method, were used to analyze the impact of soil properties (physics and chemistry) on enzymes activities and mercury speciation. This involves a simultaneous projection, at the same scale, of the PCA (data not showed) conducted on soil properties and the PCA conducted on enzyme datasets and a mercury speciation dataset. Permutation tests were conducted to assess the statistical significance of the co-variation between the different datasets. The R software was used for all statistical analyses (R version 3.3.2 (31 October 2016)).

#### *2.6. Quality Assurance and Control*

To avoid contamination, all materials used in this work were acid-washed twice with HNO3 (5%) and then rinsed several times with Milli-Q water before used.

Concerning mercury assays in samples, the relative error was routinely ±5% and was always under ±10%. The detection limit (defined as three times the standard deviation (SD) of the blank) was 0.03 μg kg−1. Concentrations obtained for repeated analyses of certified reference materials never exceeded the published range of concentrations (i.e., 0.128 ± 0.017 <sup>μ</sup>g g−<sup>1</sup> and 0.091 ± 0.009 <sup>μ</sup>g g−<sup>1</sup> for BCR-277R and MESS-3, respectively).

#### **3. Results**

#### *3.1. Soil Physical and Chemical Properties of the 6 Sites*

Granulometry measurements (Table 1) showed heterogeneous textures between the rehabilitated sites. The Bel1, Bel3, Yao2 and Yao3 sites were predominantly sandy with significantly different fine sand/coarse sand ratios between the sites. The Bel2 site was mostly silty while the Yao1 site was mainly clayey. The soils rehabilitated after mining, with heterogeneous structures and textures, can be considered as anthroposols.

**Table 1.** TOC: Total Organic Carbon, Ctot, Ntot, and Ptot: Total C, N and P, Total Fe, Al: Total Iron, Aluminum content, (*n* = 5, mean ± SD). For each parameter, values followed by different letters differ significantly with *p* < 0.05 with Tukey HSD test.


Among the six rehabilitated sites, the two sites restored Bel3 and Yao3 had significantly higher levels of total CNP and TOC content than the non-restored sites. The total carbon content, nitrogen and phosphorus varied, respectively, between 3.52 and 22.88 g·kg<sup>−</sup>1; 0.14 and 1.81 g·kg−<sup>1</sup> and 0.13 and 0.75 g·kg−<sup>1</sup> between non-restored and restored sites. The C/N ratio of the Bel1 and Bel2 sites could indicate significant nitrogen deficiencies limiting ecological restoration processes. On the contrary, the old Yaoni sites have lower C/N ratios, indicative of a better cycle of organic matter degradation.

The six sites were rich in iron with minor differences, which seems consistent with Guyanese soils [36–38]. Total iron content in soil varied between 44.4 and 85.04 g·kg−<sup>1</sup> for Bel3 and Yao2, respectively. The aluminum contents were homogeneous for Bel3, Yao1, Yao2 and Yao3 sites and higher than Bel1 and Bel2 sites.

#### *3.2. Soil Microbial Biomass and Enzyme Activities*

The two sites restored with fabaceous species, Bel3 and Yao3 sites, had significantly higher soil microbial biomass levels than the non-restored sites (Table 2) with a maximum biomass of 1102 mg kg<sup>−</sup>1.

Regardless of the soil enzyme assay (Table 2) and the rehabilitated sites, soil enzyme activities showed strong significant differences between restored and non-restored sites. The total enzymatic activities (FDA) varied between 0.49 and 3.22 <sup>μ</sup>g·g−1·h−<sup>1</sup> of fluorescein for Yao2 and Yao3, respectively. For dehydrogenase activity, the maximum activity was for Yao3 site, while the minimum was for the Bel2 site with a value of 5.80 mg kg−<sup>1</sup> and 0.34 μg TPF g−<sup>1</sup> soil h−1, respectively. For the β-glucosidase activity, the Yao3 site had significantly the highest activity with 18.6 μg pNP g−<sup>1</sup> soil h−1, while the other sites had a relatively homogeneous activity. For urease activity, the highest value was measured for Yao3, then for Bel3 with 4.74 and 1.52 μg NH4 +-N g−<sup>1</sup> soil h−1, respectively. Acid phosphatase activities were higher than for the other enzymes and significantly higher for the two restored sites and Yao2 than for the other non-restored sites. For alkaline phosphatase activity, the highest value was measured for Yao2, then for Bel2 with 12.07 and 4.17 μg pNP g−<sup>1</sup> soil h−1, respectively. For Arylsulfatase activity, the Yao3 and Bel3

sites reached values of 56.0 and 8.42 μg pNP g−<sup>1</sup> soil h<sup>−</sup>1, respectively, while the other sites had very little activity.

The results, in particular the microbial biomass, illustrated the positive impact of ecological restoration on the return of microbial communities in rehabilitated soils, in direct correlation with a return of enzymatic activities responsible for the turnover of nutrients in the soil and the functioning of ecosystems. The establishment of fabaceae species seemed to be a very interesting choice as a pioneer species to be used on rehabilitated sites.

**Table 2.** Soil Microbial Biomass and soil enzymes activities involved in soil biogeochemical cycles in the six rehabilitated sites (*<sup>n</sup>* = 5, mean <sup>±</sup> SD). MBC: Soil microbial biomass carbon in mg·kg−1, DH: Dehydrogenase activity in <sup>μ</sup>g TPF (redcolored formazan) g−<sup>1</sup> soil h<sup>−</sup>1, B-Glu, AcdP, AlkP and Aryl: β-glucosidase, Acid phosphatase, Alkaline phosphatase and Arylsulfatase activities in μg pNP g−<sup>1</sup> soil h<sup>−</sup>1, Urease: Urease activity in μg NH4 +-N g−<sup>1</sup> soil h<sup>−</sup>1, FDA: Global microbial enzymes activities in μg fluorescein g−<sup>1</sup> soil h<sup>−</sup>1. For each parameter, values followed by different letters differ significantly with *p* < 0.05 with Tukey HSD test.


*3.3. Mercury Environmental Speciation*

• Total mercury

Mercury levels (Table 3) were homogeneous between the rehabilitated sites with an average of 265 ng Hg g−1. The Yao3 site had significantly higher mercury content with 338.8 ng Hg g−1. These concentrations were still low compared to the gold mined soil in French Guiana [23]; however, it remained within the normal range of mercury measurements in French Guiana [23]. The total mercury is significantly and positively correlated with microbial biomass, total organic carbon and total iron, and negatively correlated with clay, total aluminum, and pH (Table 4).

**Table 3.** Mercury environmental speciation for the six sites. E1: Extraction of highly mobile mercury species, F1: H2O soluble mercury fraction, F2: CaCl2 exchangeable mercury fraction, F3: EDTA bioavailable mercury fraction, E2: Extraction of mobile and toxic mercury species, E3: Extraction of semi mobile mercury species, E4: Non-mobile mercury species. All values are expressed in ng g−<sup>1</sup> soil. Extraction yield represents the quality of mercury extraction in the samples, the value is obtained with the following formula: (E1 + E2 + E3 + E4)/Total Hg × 100. For each parameter, values followed by different letters differ significantly with *p* < 0.05 with Tukey HSD test.



**Table 4.** Pearson correlation coefficients (*n* = 30) between enzymes activities, soil parameters and mercury mobility; bold values are significant at the *p* < 0.05.

• Total highly mobile and mobile mercury species

The quantities of highly mobile mercury E1 (Table 3), and their percentage in relation to total mercury (Figure 2b), were significantly higher for site Yao1 with a percentage of 8% for E1. Yao1 was also the richest in water soluble mercury (F1) and bioavailable mercury (F3) with 4.56 and 2.93%, respectively. For the other rehabilitated sites, the percentages of fractions F1, F2 and F3 (Figure 2a) did not show significant differences. Concerning mobile and toxic (E2) mercury, the Bel2 and Yao3 sites had the lowest grades with percentages of 6.46% and 7.68%, while Yaoni's two non-restored sites had average grades of 31.1% and 44.3% respectively. E1 and E2 are correlated negatively with microbial parameters and TOC and positively with total clay and aluminum (Table 4).

**Figure 2.** Mercury environmental speciation for the six sites as a percentage of each fraction relative to total mercury. (**a**) Focus on the most mobile fractions of mercury. F1: H2O soluble mercury, F2: CaCl2 exchangeable mercury, F3: EDTA bioavailable mercury. (**b**) Focus on the main extractions of mercury. E1: Extraction of highly mobile mercury, E2: Extraction of mobile and toxic mer cury, E3: Extraction of semi mobile mercury, and E4: Non-mobile mercury.

• Total semi-mobile mercury species

The highest percentages were detected for sites Yao1, Bel2 and Bel3 with values of 51.9%, 59.2% and 61.3%, respectively. The Bel1 site had the lowest quantity and percentage of semi-mobile mercury with 23.7%. The semi-mobile mercury fraction did not have a significant effect on microbial parameters and was significantly correlated with clay and aluminum content.

• Total non-mobile mercury species

The restored Yao3 site showed that mercury was most immobilized with 54% nonmobile mercury. The two non-restored Yaoni sites Yao1 and Yao2 had significantly lower levels of non-mobile mercury than the other sites with 4.4% and 9.9%, respectively. The two non-restored sites Bel1 and Bel2 had more mercury, 50.3% and 30.9%, respectively, than the restored site Bel3 with 19.7%. The correlation between non-mobile mercury and the dataset followed exactly the same pattern as the total mercury content.

#### *3.4. Co-Inertia between Soil Parameters and Microbial Activities Dataset*

The co-inertia analysis (Figure 3a) showed that soil parameters (six variables) had a highly significant correlation (*p* < 0.001) with the microbiological dataset (eight variables) with 41.86% of total variance explained. Total organic carbon was positively correlated to enzyme activities with the exception of acid and alkaline phosphatase. Soil pH, fine sand and clay content were negatively related to the microbiological dataset.

**Figure 3.** (**a**) Co-inertia analysis between soil physico-chemical parameters and microbiological variables. The variables linked by lines refer to the physico-chemical variables, the other variables refer to the different enzymes activities and microbial biomass in the soil sample; (**b**) discriminant analysis on co-inertia variable of soil samples. Boxes represent the projected co-ordinates of physico-chemical dataset and microbial dataset of each rehabilitated site, respectively. The length of the arrow is proportional to the divergence between the data sets. Eigen values 86.74% and 9.08% for axis 1 to 2, respectively. Randtest: simulated *p*-value: 0.0019. Explained variance: 41.81%.

Discriminant analysis of co-inertia (Figure 3b) showed that soil samples were significantly distinguished (*p* < 0.001) according to their spatial distribution in terms of the quality of ecological restoration. In the axis 1 direction, soil samples collected from restored sites (Bel3 and Yao3) were predominated on the positive axis, and were clearly distinguished from non-restored sites, all on the negative side.

According to these data, the microbial biomass and all enzyme activities tested except for phosphatases would be correlated exclusively to the organic carbon concentration. However, we observed a negative correlation of the microbiological data with the concentration of clays and fine sands. Texture and structure are among the most important factors influencing the activity of telluric enzymes [39]. Enzyme activity is generally greater in fine textured soils than in coarse textured soils [40,41] and negative correlations between enzyme activity and clays appear to show opposite results. However, while the nature of clays also impacts enzymatic activities [42,43], it has been shown that the structural stability of soil aggregates, linked to the organic matter content, could play a more important role

than soil texture alone [40,44,45]. Organic matter stabilizes aggregate structure, increases soil retention capacity and improves nutrient bioavailability [46,47], thus stimulating soil microbial flora [48].

In this study, the major factor of microbial activity was organic matter. In these rehabilitated soils, the contribution of organic matter is exclusively conditioned by the nature of the plant cover and therefore by the success of ecological restoration. The choice of the implantation of fabaceous species, fast-growing nitrogen-fixing species, seems judicious for their ability to bring a large amount of organic matter to the soil [49,50].

#### *3.5. Co-Inertia between Soil Parameters and Environmental Mercury Speciation*

The co-inertia analysis (Figure 4a) showed that soil parameters (eight variables) had a highly significant correlation (*p* < 0.001) with the environmental mercury speciation dataset (seven variables) with 82.06% of total variance explained, which could indicate relatively similar mercury retention and mobilization mechanisms between different rehabilitated sites.

**Figure 4.** (**a**) Co-inertia analysis between soil physico-chemical parameters and mercury extraction. The variables linked by lines refer to the physico-chemical variables, the other variables refer to the different extractions and fractions of mercury in the samples; (**b**) discriminant analysis on co-inertia variable of soil samples. Circles and boxes represent the projected co-ordinates of physico-chemical dataset and mercury dataset of each rehabilitated site, respectively. The length of the arrow is proportional to the divergence between the data sets. Eigen values 67.96% and 20.34% for axis 1 to 2, respectively. Randtest: simulated *p*-value: 0.0001. Explained variance: 82.06%.

Clay was strongly correlated with highly mobile mercury species (E1, F1, and F3) and negatively correlated to total and non-mobile mercury, while silt was related to exchangeable mercury and semi-mobile mercury species. Mobile mercury species were correlated with coarse sand, but not with fine sand. Soil pH was significantly and positively correlated (Table 4) with mobile mercury species (E2) and negatively correlated with total Fe content.

Discriminant analysis of co-inertia (Figure 4b) showed that soil samples were significantly distinguished (*p* < 0.001) according to their spatial distribution in terms of the quality of ecological restoration. In axis 1 direction, soil samples collected from restored sites (Bel3 and Yao3) were predominated on the positive axis and were clearly distinguished from non-restored sites. The highly mobile clay and mercury variable appeared to strongly distinguish the Yaoni sites, while the percentages of silt, sand and semi-mobile mercury appeared to more differentiate the Belizon sites along axis 2.

#### **4. Discussion**

#### *4.1. Positive Effect of Ecological Restoration on Mercury Mobility*

In this study, the results showed very distinct mercury behaviors based on sites that were not solely related to vegetation density. Indeed, the three least covered sites (Bel1, Yao1 and Yao2) presented very different speciation results with nearly 50% non-mobile mercury for Bel1 and less than 10% non-mobile mercury for Yao1 and Yao2. In addition, the results showed a significant increase in mobile and toxic mercury for sites Yao1 and Yao2 with more than 40% compared to about 25% of mobile mercury for Bel1. The presence of tree cover did not positively impact the mobility of mercury on the Bel3 site compared to adjacent plot Bel2, which showed a higher proportion of semi-extractable and nonmobile mercury and less mobile mercury. However, mercury speciation over the oldest restored area of Yao3 showed a very significant decrease in mobility and potential toxicity of mercury with 90% of mercury present in non-extractable forms.

While HCL/ethanol extraction represents more of a mobility potential linked to leaching under acidic conditions, water extraction, CaCl2 and EDTA represent more direct mobility close to natural conditions in soils. The F1, F2 and F3 extractions showed the lowest percentage of easily mobile mercury for the Yao3 site and the highest for Yao1 with 1 and 8%, respectively. While the low mobility of mercury species for the Yao3 site are close to those for the Asturias [51], Almaden [52] and Usagre [53] mines in Spain, the Cuyuni Basin [54] in Venezuela, old mining mines in Mexico [55], in the Yaoni region of French Guiana [16], and the mercury mobility and toxicity for Yao1 were significantly higher and comparable to mobility at some Wanshan mining sites in China [56]. Depending on rainfall, these sites can go into suboxic or even anoxic conditions and these results show that the pool of mercury available for rapid methylation diverges very strongly between the rehabilitated sites.

The results indicate the contribution of a dense cover of fabaceae on the Yao3 site to modify mercury speciation compared to the unrestored Yao1 and Yao2 sites. At this site, despite a higher total mercury content than at the other rehabilitated sites, most of the mercury corresponds to very stable chemical species with low mobility and high stability, such as amalgamated elemental mercury, strongly bound mercury (mineralized and metal-oxyhydroxides) and species in the form of sulphides and hydroxide oxides. For the restored site Bel3, the results are more contrasted, but are possibly due to the more recent restoration and more limited contributions in organic matter in only four years of vegetation recovery.

#### *4.2. Effect of Soil and Microbial Properties on Environmental Speciation of Mercury*

In view of the above conclusions, the nature of the vegetation cover is not only indicative of mercury mobility. The mercury speciation provide access to information on toxicity, but do not provide a clear picture of the influence of mercury-carrying phases in soils. In tropical soils, the mobility and retention of mercury are largely from the nature of carrier phases. More specifically, clay minerals, iron and aluminum (oxyhydr)oxides, and the quantity and nature of organic matter represent the most important carrier phases and vary according to the biochemical conditions of the environment [57–60].

Extractions E1, E2 as well as the highly soluble fractions F1, F2 and F3, are positively correlated with clay and total aluminum contents. In tropical soils, rich in kaolinite and aluminosilicates, the adsorption of mercury in clay minerals may be important, especially when the pH is above 5 [61]. This distribution in soil of soluble fractions of mercury in clay minerals is consistent with the literature and shows that mercury adsorbed on these soil phases appears to be readily mobilizable [61].

Concerning the relations between mercury and the organic component of the soil (organic matter and biomass), it would seem that organic matter is the main carrier phase of non-mobile and total mercury (r2 = 0.87 and 0.92), which are results consistent with the literature [62], and are already observed on restored sites in French Guiana [16]. The increase in mercury content in restored soil in relation to organic matter content could be

related to the heterogeneity of mining soils, but could also be a key factor in the processes of mercury retention in the soil and in the limitation of losses through leaching and erosion of soil particles. The enrichment of organic matter levels through ecological restoration could therefore have an additional effect by potentially limiting mercury transfers. It would also appear that these non-mobile and total mercury concentrations do not have an impact on the development of soil microbial communities with positive correlations with microbial biomass and enzymatic activities, a result already shown in some mining areas with mercury loading in soil [63]. Nevertheless, the negative correlations between E2 (toxic mercury) and microbial biomass could indicate that a predominance of toxic forms of mercury, even at relatively low concentrations, could negatively affect the development of microbial communities and their enzymatic activities. Another major point is that these potentially toxic forms of mercury do not correlate with soil organic matter. In this context, microbial mineralization of organic matter would therefore be likely to release non-toxic forms of mercury, but would have a very limited impact on the transfer of toxic mercury into the soil.

This study also shows a significant correlation between total iron and total mercury. While iron oxide concentrations in soils were not measured in this study, these results confirm the importance of iron in the biogeochemistry of mercury and its retention in soil phases. This hypothesis also seems to be confirmed by the important correlations observed between semi-mobile mercury extraction and silty soil fractions, rich in metaloxyhydroxides such as goethite or hematite. This study showed that the impact of nonhomogeneous mine site remediation processes has altered mercury speciation by acting directly on soil texture. These results also showed that the level of vegetation cover, and the level of organic matter brought to the soil, has improved mercury retention and sanitary quality.

#### *4.3. Positive Effect of Ecological Restoration on Microbial Activities*

Soil biochemical parameters such as microbial biomass or soil enzymatic activities are sensitive and relevant indicators of stressed and disturbed soils and are used to describe soil quality status [15,64]. Generally, microbial biomass varies according to soil physicochemical parameters, vegetation cover or potential contamination [15,65]. In this study, microbial biomass appeared to be very sensitive to the different modalities of ecological restoration and to the nature of the vegetation cover. The presence of a restored tree cover on the Bel3 and Yao3 plots had increased the microbial biomass by a factor of 2–4 compared to non-restored sites. The presence of *Lycopodiella* sp. on the Bel2 plot has also positively stimulated microbial biomass and shows that the increase in plant density, even herbaceous, has a significant impact on microbial biomass.

Soil enzymatic activities represent the overall degradation activity of organic soil components and are affected by many factors, such as pH, organic matter content or contamination level [66], and are mediated by microbial biomass. In our study, the restored Bel3 and Yao3 sites had the highest activities, except for alkaline phosphatase activities. However, the classification of soils according to the amount of microbial biomass is not the same as for enzymatic activities. Indeed, the Yao2 site, which has one of the lowest biomasses, has relatively important dehydrogenase, B-glucosidase, acid and alkaline phosphatase activities. This could show that these enzymatic activities are not only related to microbial biomass. These results indicated a significant and positive effect of controlled ecological restoration on certain components of biochemical activities in rehabilitated soils [67–70]. The very low level of activity on non-restored soils also highlights the stressed and anthropized nature of these former mining lands, which require adapted ecological rehabilitation protocols.

#### *4.4. Footprint of Rehabilitation on Soil Functionning and Mercury Mobility*

The various stages of rebuilding an ecosystem after human degradation, such as mining are complex, multidisciplinary and often involve ecological engineering. On mining sites in French Guiana, these protocols have to be adapted according to the degree of site contamination, soil texture, water regime and of course the nature of the plant species used for restoration [14]. *Clitoria racemosa* is a fabacea native to South America and favors sandygravelly or even sandy-clay substrates. At the Belizon mine, difficult site rehabilitation and earthworks led to a high degree of textural heterogeneity, resulting in a mosaic of vegetation and partial success in ecological restoration. The recovery of the macrocuts was not effective on the overly draining and sandy substrate of the Bel1 plots, now colonized by *Cyperaceae* sp. On the Bel2 plots, the significant presence of silt led to a rapid covering of the site by *Lycopodiella* sp., preventing any recovery of *C. racemosa* plants. The rational use of other products, such as *Acacia mangium*, adapted to gravel surfaces, and *Erythrina fusca*, adapted to silty substrates on the Bel1 and Bel2 plots, respectively, could have allowed for a more widespread success of ecological restoration. These results showed the importance of adapting ecological restoration protocols according to the nature of the substrate, as already described by Loubry [14].

On the former Yaoni mine, the very significant differences in terms of organic constituents of the soil provide interesting elements for the restoration of mining sites. The very low CNP stocks on plots Yao1 and Yao2 showed that mining activity depletes organic constituents and soil fertility. Seventeen years after the end of the rehabilitation, these sites had still not found shrub or tree cover and ecological succession seemed to be stopped. The controlled recovery of vegetation on the Yao3 site has highlighted the positive effect of the establishment of a cover of fabaceae on the renewal of the stock of nutrients and organic matter. It would appear that the organic matter brought in during ecological restoration is also a driving force behind the retention of mercury and the limitations of potential transfers to aqueous media. A preliminary study [16] showed that ecological restoration could affect the nature of the mercury bearing phases. However, it would also seem that the potential toxicity of these plots under ecological restoration could be reduced by the establishment of a plant cover.

#### **5. Conclusions**

We evaluated the effect of the rehabilitation of Guyana's mining sites in order to link soil texture with the quality of ecological restoration to the biochemical quality of the soil and the potential toxicity of mercury. The results indicated very variable and heterogeneous soil textures, even on nearby sites. Some sites rehabilitated with too much silt or sand have not allowed for the ecological restoration of a tree canopy. The results also highlight a direct impact of the success of ecological restoration on the return of microbial activities related to biogeochemical cycles. The provision of bedding using fabaceous species seems to be one of the key factors for the success of ecological restoration. The results also showed an impact of soil texture on mercury mobility and toxicity. Environmental speciation measures showed a link between mercury mobility and the nature of rehabilitation. The least restored sites have potentially higher concentrations of mobile and semi-mobile mercury, related to the quantities of clays and silts. The homogeneous texture of the restored Yaoni site, as well as the establishment of a dense tree canopy, has contributed to limiting mercury mobility by modifying its speciation towards more insoluble forms, such as cinnabar.

**Author Contributions:** Conceptualization: N.B. and E.C., The sampling campaigns for the sites presented in this study are the result of a second sampling campaign on mining sites undergoing ecological restoration. The samples were taken by E.C. and N.B., with Christian Pernaut as field guide. The methodology: E.C. and N.B.; validation, N.B.; formal analysis, E.C., V.A., A.L. and S.G.-M.; investigation, E.C.; resources, E.C.; data curation, E.C.; writing—original draft preparation, E.C.; writing—review and editing, E.C.; visualization, E.C.; supervision, N.B.; project administration, N.B.; funding acquisition, N.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** We would like to thank Christian Pernaut, the manager of the mining company Ermina (Matoury-France), who was our guide in French Guiana during the sampling campaign. We thank Philippe Matheus, the manager of the mining company C. M. B (Remire Montjoly-France). for his logistical and material support. We also thank the Cayenne and Bondy Development Research Institute for their analytical contributions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Applied Sciences* Editorial Office E-mail: applsci@mdpi.com www.mdpi.com/journal/applsci

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com