Levels of mercury in Nile tilapia (Oreochromis niloticus), water, and sediment in the Migori gold mining belt, Kenya, and the potential ramifications to human health

Ethical approval and consent to participate

Ethical approval was obtained from the Biosafety, Animal Care, and Use Committee of the Faculty of Veterinary Medicine, University of Nairobi (REF: FVM BAUEC/2018/148; Extended data, Figure S1¹⁶). Guidelines on humane harvesting of fish¹⁷ were applied throughout the study. Extreme caution was exercised in handling concentrated mercury reagents and acids. Good laboratory practices (GLPs) were observed (use of gloves, gas masks, overalls, and fume chamber) at all times. Before the study began, it was ensured that the fume extraction system within the lab was in good working condition by following the laboratory’s standard operating procedures.

Study design

Purposive sampling was used to collect fish, sediment, and water samples across sites within the Migori gold mining belt. The number of samples collected are summarized in Table S1 (Extended data¹⁶). The levels of total mercury in the collected samples was evaluated in the lab by following the methods of the United States Environmental Protection Agency (US-EPA) and Analytical Methods for Atomic Absorption Spectroscopy by Perkin Elmer^18–20 with slight modifications. (Modifications: 1. the target weight for all samples was 0.5 grams but for those samples whose weights were below this weight, the whole sample was weighed and processed. 2. The reaction tube was elongated by fixing a glass (Soselex) column on top to prevent bumping and spill over.)

Study area and sampling sites

The Migori gold mining belt covers five sub-counties namely Suna West, Nyatike, Rongo, Kuria West, and Kuria East within Migori County (Figure 1). Apart from mining, other economic activities undertaken in the region include livestock, maize, tobacco, and sugar cane farming. The main rivers that drain the area are the Mara, Kuja, and Migori². Rongo and Nyatike sub-counties were selected for sampling. This is because, in addition to being within the gold mining belt, inland fish farming is widely practiced in the region. Therefore, ten sites namely Minyenya, Kamagambo North, Masara, Nyabisawa, Ndiwa, Kokaka, Luanda Nyira, Kamagambo South, Siginga beach and Sori beach located within the two sub-counties (Figure 2), and bearing coordinates between 0°6’11.16’’-0°52’51.6’’S and 34°7’8.4-34°37’55.2’’E were selected as study sites.

Figure 1. Sub-counties in Migori County within the gold mining belt.

Key: 1: Nyatike, 2: Uriri, 3: Suna-East, 4: Suna-West, 5 Kuria-West, 6: Kuria-East, 7: Awendo, 8: Rongo.

Figure 2. Map of Migori County showing the sampling sites.

Sediment sampling

By use of plastic trowels, near-surface (the top 5 cm) of fish pond sediments weighing approximately 1000 g was taken from each of the selected sites²¹, packed in plastic Biological Oxygen Demand (BOD) bottles (See details under extended data¹⁶. Table S2), labeled (location, sample type and the date of collection), kept on ice and transported to the laboratory where they were stored at -20° C until the time of analysis.

Water sampling

Trace metal clean procedures, as described by Shafer, Shelton, and colleagues^22,23 were used to collect water samples. Briefly, water samples were collected in 250 ml metal-free plastic bottles and acidified to a pH <2 using 69% ultra-pure nitric acid (HNO₃) (Extended data¹⁶, Table S2) to prevent adsorption of potentially harmful elements onto the interior walls of the storage bottles as well and minimize microbial activity. The resultant mixture was then filtered through a 0.45 µm pore filter (Extended data¹⁶, Table S2) and stored in 125 ml metal-free plastic sample bottles and refrigerated at -20° C until the time of analysis. Mercury in the filtrate (0.45 μm), also referred to as “dissolved” mercury was of particular interest in this study as this is what may have had measurable biological effects on aquatic organisms²³.

Collection of fish samples and euthanasia

Five tilapia fish (regardless of their sex) were captured from each site by using gill-nets (except Minyenya where four fish were sampled). Guidelines provided by the Faculty of Veterinary Medicine, University of Nairobi were adopted (REF: FVM BAUEC/2018/148; see Extended data¹⁶, Figure S1), and all efforts were taken to ameliorate harm to the animal model (fish). Additionally, guidelines on the humane harvesting of fish¹⁷ were adopted in euthanizing the collected fish.

A two-step process involving electro-narcosis and asphyxiation was used¹⁷. Fish were initially stunned in an electric field of 2.5V/cm at 1000 Hz to make them insensible to pain¹⁷. The absence of eye-roll reflex when the fish were moved from side to side was used as a confirmation that insensibility had been achieved¹⁷. Death was then induced by asphyxiation in the air for 10 seconds and was confirmed by the lack of movement of the operculum¹⁷. Stunning fish by use of low current of electricity is considered a humane method of euthanasia as this current makes the fish insensible to pain. The length of the fish was measured by laying an extended tape measure on a flat surface and placing the fish next to the tape measure. The total length was determined by measuring from the tip of the fish’s mouth to the tip of the caudal fin. The weight of each fish was measured on an analytical balance (Sartorius: SECURA^® 324-101N). A blade (number 11) was used to make a slit incision about 3–5mm in length through the ventral abdominal wall and a 10 cm² sample of muscle tissue was taken from each fish²¹. Liver and brain tissues were also harvested by carefully dissecting them out from the fish and transferring the tissues to aluminum foil by use of gloves. The collected samples were then transferred to plastic sample bottles which were labeled, and packed in self-zipping polyethylene bags, stored in ice in an ice cool box and transferred to a -20° C freezer awaiting further analysis. The characteristics of the ecosystem in fish ponds in the study area are summarized under Table S3 (Extended data¹⁶).

Experimental procedures

Pre-treatment of equipment and sample bottles. Precautionary steps, as described by Shafer and others²³ were taken before using the equipment or sample collection bottles. Briefly, all equipment used for sample collection and storage of sediment, water, and fish samples were pre-cleaned using high-purity nitric acid and rinsed with sufficient quantities of reagent water. This was done to ensure that they were free of trace metals. After cleaning, the bottles were stored in double-bagged zip-lock polyethylene bags to ensure that no detectable metal contaminants were present in the sampling equipment.

Reagents. All chemicals and reagents used were of analytical grade (Merck, Germany; Sigma-Aldrich, France; Central Drug House, India; Fisher Scientific, UK). (Table S2, Extended data¹⁶). Double distilled, de-ionized water was used for preparing working solutions and for all analytical work. Standard stock solutions of mercury were prepared from a high purity standard stock solution with a concentration of 1000 parts per billion (ppb) and were diluted to the corresponding mercury working standard solutions (i.e.10 ppb, 20 ppb, and 30 ppb). These working solutions were freshly prepared daily by diluting an appropriate aliquot of the stock solution using 1M hydrochloric acid (HCl; Sigma-Aldrich) and diluting the resulting solution to 100 ml with reagent water.

Standard reference material for mercury in fish, i.e. the Community Bureau of Reference (BCR) – 463 (European Commission), was analyzed to ascertain the accuracy and precision of the experimental procedure. Alkaline solutions of sodium borohydride (NaBH₄) were freshly prepared daily by dissolving 1.0 g of NaBH₄ (Merck), and 0.25g of sodium hydroxide (NaOH) pellets (Merck) in 500 ml of distilled water. 3% v/v of HCl (Sigma-Aldrich) was used in the preparation of the carrier gas (Argon C-45). Stannous chloride (SnCl₂) was freshly prepared by dissolving 62.5 g in 50 ml of 6 M HCl, the solution boiled for about 5 minutes, cooled, and nitrogen bubbled through it to expel any impurities of mercury. For sample digestion, 11 M nitric acid (HNO_3; Merck), 18 M perchloric acid (HClO₄; Merck) and HCl (Sigma-Aldrich) were used. Fused alumina anti-bumping granules (Merck) were used to avoid foam formation during sample digestion.

Equipment and apparatus. All glassware used in the analysis were soaked overnight in 10 % (v/v) nitric acid (HNO₃), followed by washing with 10% (v/v) hydrochloric acid (HCl). They were then rinsed with double-distilled water and dried before use. Samples were weighed on an analytical balance (Sartorius: SECURA^® analytical balance: SECURA 324-101N), and sample digestion was carried out in a steam bath (DK Heating Digester from Velp Scientifica) in the confines of a fume hood. A Varian Model Spectr AA 220Z atomic absorption spectrometer equipped with a mercury hollow cathode lamp was used for the analysis of the total mercury content of samples. Flow injection and cold vapor generation were done via a Varian model vapor generation accessory (VGA) 77. The analytical wavelength and slit widths were 253.7 nm and 0.5 nm, respectively. The Varian model, Spectr AA 220Z software (Version: Spectr AA), was used to monitor the output.

Sample preparation and digestion. All samples, certified reference materials, standards, reagent blanks, and spiked samples were processed using slightly modified methods of the United States Environmental Protection Agency (US-EPA) and Analytical Methods for Atomic Absorption Spectroscopy by Perkin Elmer^18–20. Briefly, a top pan analytical balance (Sartorius: SECURA^®analytical balance: SECURA 324-101N) was calibrated each morning before weighing samples. Samples were removed from the freezer and allowed to thaw for about an hour. A batch of samples (approximately 20 in number) was digested simultaneously. Then, a 0.3–0.5 g aliquot of a well-homogenized sample was weighed and placed at the bottom of a glass digestion tube. For samples that were less than 0.5 g (particularly the brain and liver tissues), the weighing was done so as to obtain as much of the tissue samples as possible to enable processing. Nine milliliters of concentrated nitric acid (HNO₃), 3 ml perchloric acid (HClO₄) and 1 ml of HCl (to stabilize the pH of the matrix) were slowly added to the glass digestion tube in a fume hood. The tube was allowed to stand at room temperature (in the fume hood) until the initial reaction subsided (about 15 minutes). A spoon scoop of fused alumina anti-bumping granules (Merck) was added to the solution to prevent it from bumping and spilling over. A glass (Soselex) column was fixed on top of the glass digestion tube to prevent spilling over in the event of frothing. The tube was then placed on top of a steam bath unit (DK Heating Digester; Velp Scientifica) which was programmed to heat gradually to 150 °C over 10 minutes. This heating was maintained for 120 minutes to complete dissolution. The tube was then removed from the steam bath, and the solution allowed to cool to room temperature over 30 minutes. The solution was then carefully transferred into a 100 ml flat-bottomed volumetric flask, the tube and column rinsed thoroughly with small amounts of distilled water and the resultant contents transferred into the flat-bottomed volumetric flask. Six milliliters of saturated potassium persulfate (K₂S₂O₈) and 30 ml of potassium permanganate (KMnO₄) was added to the solution and slightly shaken to mix. The resultant was left to stand for 40 minutes. Additional portions of the KMnO₄ solution were gradually added until the resulting purple color persisted for at least 15 minutes. After thorough mixing, 6 mL of sodium chloride-hydroxylamine sulfate was added to reduce the excess permanganate (this was confirmed by the color change from purple to colorless). Reagent water was then added to the mixture up to the 100 ml mark and treated samples were then filtered through grade 541 (diameter 110 µm) filter paper. Five milliliters of stannous sulfate was then added to each of the treated samples. Thereafter, each sample bottle was immediately attached to the aeration apparatus (one at a time) of the cold vapor atomic absorption spectrophotometer.

Analytical quality control. Precautionary steps were taken to rule out any interference that may have arisen in the course of running the analysis. Briefly, method blanks were analyzed to ensure that all the materials (solvents, reagents, glassware, and other sample processing hardware) were free from artifacts or interferences which may have had the potential to compromise the integrity of the analysis. Interference from sulfide was abolished by the use of 5% w/v potassium permanganate (KMnO₄). Excess hydroxylamine sulfate reagent (about 25 mL) was used to ensure that free chlorine was absent before the mercury was reduced and swept into the cell. Also, the dead air space in the BOD bottle was purged before adding stannous sulfate.

A preliminary run without reagents (using reagent water) was also used to rule out interference by volatile organic materials. Moreover, the accuracy of the procedure was determined by analyzing three certified reference materials (CRMs) namely; Tuna fish muscle fapas CRM, BCR 463 from Community Bureau of Reference, European Commission, vegetable puree CRM (EU) and fish CRM (EU). Recovery studies were performed by adding a known amount of standard solution of mercury chloride to some samples (spiked samples), which were then taken through the digestion procedure. The concentration of mercury in the resulting solutions was then analyzed.

Analysis of mercury in collected samples. The optimum operating temperature of the CVAAS instrument was set at 18 °C, and the circulating pump was adjusted to continuously pump at the rate of 1 L/min. Maximum absorbance was noted within 30 seconds. The bypass valve was then opened, and aeration continued until absorbance returned to the minimum value. The bypass valve was then closed, the fritted tubing removed from the BOD bottle, and aeration continued. The measurement time was set at 5 seconds, with a pre-reading delay of 45 seconds in between readings. Aliquots of 1.0, 2.0 and 3.0 mL of the mercury working standard (containing 0.1 mg/L or 1000 ppb) of mercury and 1 ml of fuming (37%) hydrochloric acid (HCl) solution was transferred to a series of 100 mL volumetric flasks and made up to the mark with reagent water. The standards had 10, 20, and 30 ppb of mercury, respectively. A calibration curve was automatically generated from the instrument’s software, plotting the absorbance of the standard against the concentration of mercury in parts per billion. The absorbance of the samples and standards were determined from the recording device and corresponding mercury concentrations tabulated.

Evaluation of the degree of sediment contamination. The quantitative geochemical accumulation index (I_Geo) was used to evaluate the level of mercury contamination in fish pond sediments collected from different sampling sites. This method applies the formula proposed by Müller²⁴ to calculate the degree to which sediment is contaminated by mercury. Thus, I_Geo= log2 (C_n/1.5×B_n), where; I_Geo= the geochemical accumulation index, C_n=sediment metal concentration and B_n=geochemical background value of the metal. In this study, the global mercury background value was taken as 0.05, as described by Reimann²⁵. Accordingly, mercury pollution in collected sediments was classified into seven categories (0–6) as: class 0 (unpolluted; I_Geo≤0), class 1 (unpolluted to moderately polluted; 0≤I_Geo≤1), class 2 (moderately polluted; 1≤I_Geo≤2), class 3 (moderately to strongly polluted; 2≤I_Geo≤3), class 4 (strongly polluted; 3≤I_Geo≤4), class 5 (strongly to extremely polluted; 4≤I_Geo≤5), and class 6 (extremely contaminated; I_Geo>5)²⁴.

Risk-based consumption limits. Guidelines set by the United States Environmental Protection Agency (US-EPA)^26,27 were used to calculate the potential health risk from consumption of Nile Tilapia sampled in the region. An assumption was made that the ingestion dose was equal to the absorbed dose of Hg, as has been described previously by Chien and colleagues²⁸. Calculations on mercury consumption limits were based on the US-EPA reference dose (RfDo). The ratio between exposure and the reference dose indicated by the target hazard quotient (THQ), were calculated on the standard assumption of an integrated US-EPA risk analysis model. The methods described by Copat et al.^29,30 were used to estimate the daily intake per meal (EDI_m) and the target hazard quotient (THQ) as below;

Where EDI_m is the estimated daily intake of mercury per meal size; MS is the standard weight portion of fish (230 g) for adults according to Hosseini et al.³¹; C refers to the concentration of mercury in mg/kg wet weight according to Marrugo-Negrete and colleagues³²; BW is the body weight (assumed to be 70 kg for an adult)²⁹. According to the US-EPA, the RfDo for T-Hg is 0.1 μg/g/day²⁶.

For non-carcinogenic effects, the maximum allowable fish consumption rate in meals/week (CR_mw) according to the US-EPA²⁶ that would not be expected to cause any chronic systemic effects were calculated as below;

Where; MS is the standard weight portion of fish (230 g) for adults according to Hosseini et al.³¹, and C refers to the concentration of mercury in mg/kg wet weight according to Marrugo-Negrete and colleagues³². Considering an average adult body weight of 70 kg, the Hg USEPA Acceptable Daily Intake (ADI) can be approximated as 7 µg/day/adult (49 µg Hg/week)^26,31.

Statistical analysis

Data for mercury analysis was expressed as the mean ± standard deviation. One-Way Analysis of Variance was used to analyze the levels of mercury in fish tissues across the sites. Tukey’s HSD test was used as a post-hoc test. Pearson’s correlation coefficient was used to determine whether there were any relationships between mercury levels in water and those in fish tissues, levels of mercury in sediment and those in fish tissues, and the levels of mercury in fish tissues and the pH of the pond water. The same test was also used to determine relationships between the levels of mercury in fish tissues and pond water pH, temperature, weight, and age of the fish. The t-test was used to analyze the relationship between the level of mercury and the type of fish culture practiced. Microsoft Excel (2016) and Statistical Package for the Social Sciences (SPSS, version 20.0) were used for statistical analysis. p≤0.05 was considered significant in all cases.

Mean levels of T-Hg in fish (Nile tilapia) tissues

The mean levels of T-Hg were highest in the fish brain tissues and ranged from 0.128±0.021 µg/g wet weight (n= 5, 95% CI) at Nyabisawa in Nyatike sub-county to 3.128±1.421 µg/g wet weight (n= 4, 95% CI) at Minyenya in Rongo sub-county (Table 1 and Underlying data).

The fish muscle tissues were the second most contaminated with mean T-Hg levels ranging from 0.179±0.020 µg/g wet weight (n= 5, 95% CI) at Kamagambo South in Rongo sub-county to 0.917±0.099 µg/g wet weight (n= 5, 95% CI) in Minyenya. Table 1. The fish liver tissues were the least contaminated tissues with mean T-Hg levels ranging from 0.103±0.118 µg/g wet weight (n= 5, 95% CI) in Kamagambo South, Rongo sub-county to 0.588±0.374 µg/g wet weight (n= 5, 95% CI) in Kokaka, Rongo sub-county (Table 1).

There was a significant difference (p=0.00) in the mean levels of T-Hg in fish brain tissues sampled from the different sites (Table S4, Extended data¹⁶). Further evaluation of this variation revealed that the mean levels of T-Hg in fish brain tissues was significantly different between Nyabisawa and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036), Kamagambo South and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036), Sori beach and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036), Siginga beach and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036) and between Masara pond and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036), Sori beach and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036), and Kamagambo North and (Luanda Nyira, Kokaka, Ndiwa and Minyenya) (p=0.036) (Table S5, Extended data¹⁶).

There was a significant difference (p=0.00) in the mean levels of T-Hg in fish muscle tissues sampled from the different sites (Table S6, Extended data¹⁶). Further evaluation of this variation revealed that the mean levels of T-Hg in fish muscle tissues was significantly different between Kamagambo South and Minyenya (p=0.043), Kamagambo North and Minyenya (p=0.043), Kamagambo North and Masara pond (p=0.022), Kamagambo North and Siginga beach (p=0.022), Luanda Nyira and Masara pond, and Luanda Nyira and Siginga beach (p=0.022). (Table S7, Extended data¹⁶).

There was a significant difference (p=0.00) in the mean levels of T-Hg in fish liver tissues sampled from the different sites (Table S8, Extended data¹⁶). Post hoc analysis of this variation established that the mean levels of T-Hg in fish liver tissues was significantly different between Kamagambo South and (Sori beach, and Minyenya) (p=0.005), Masara pond and (Sori beach and Minyenya) (p=0.005), and between Ndiwa and (Sori beach and Minyenya) (p=0.005) (Table S9, Extended data¹⁶). The mean levels of T-Hg in all fish tissues (brain, liver, muscle) collected from the different study sites were found to be significantly higher than the critical value of 0.2 µg/g wet weight (n=49, 95%CI) which is the consumption limit for at-risk human populations (Table 3). However, only the fish brain tissues were found to have mean levels of T-Hg that were significantly greater than the critical value of 0.5 µg/g wet weight (n=49, 95%CI) which is the critical value for the general population (Table S10, Extended data¹⁶).

Mean levels of T-Hg in fish pond sediments collected from the study sites

The levels of T-Hg in sediments ranged from 0.208±0.000 to 1.113±0.008 µg/g wet weight (n= 3, 95%CI) with six of the eight sample sites being moderately polluted (1≤I_Geo≤2), whereas two sites (Minyenya and Kokaka; Rongo sub-county) were strongly polluted with total mercury (3≤I_Geo≤4; Table 2).

The relationship between the mean levels of T-Hg in fish pond soil sediments and the mean levels of T-Hg in fish tissues

An increase in the mean levels of T-Hg in the fish pond soil sediments coincided with an increase in the mean levels of T-Hg content in fish brain tissues (r= 0.528, sig. = 0.001, 95%CI) as well as an increase in the mean levels of T-Hg in fish muscle tissues (r= 0.524, sig. =0.001, 95%CI; Table S11, Extended data¹⁶). However, there was no significant correlation between the mean T-Hg levels in pond sediments and tilapia liver tissues (sig. =0.923, 95%CI; Table S11, Extended data¹⁶).

Mean levels of T-Hg in fish pond water collected from the different study sites

The mean levels of T-Hg in the water samples collected from the different sites ranged from 0.002±0.000 µg/ml to 0.004±0.001 µg/ml (n=3, 95% CI). (Table 3)

The water samples from all the sites were found to have mean T-Hg levels that were significantly greater than the critical value of 0.0001µg/ml (n=8, sig. = 0.000 at 95%CI) set by the food and agriculture organization for unpolluted surface waters³³. (Table S12; Extended data¹⁶). An increase in the mean levels of T-Hg in fish pond water coincided with a decrease in the mean level of T-Hg in fish brain (r= - 0.402, sig. = 0.011, 95%CI) and muscle (r= - 0.616, sig. =0.000, 95%CI) tissues. (Table S13) Extended data. There was no significant correlation between the mean levels of T-Hg in pond water and the mean levels of T-Hg in fish liver tissues (sig. =0.874, 95%CI; Table S13, Extended data¹⁶).

The relationship between water pH, water temperature, fish weight, and fish age and the level of mercury in fish tissues

An increase in the pH of fish pond water coincided with an increase in the mean levels of total mercury in fish brain tissues (r=0.48, p<0.05). (Table S14) Extended data. An increase in the temperature of the water in fish ponds coincided with an increase in the mean levels of total mercury in fish brain tissues (r=0.404, p<0.05; Table S14, Extended data¹⁶). The larger the fish, the lower the level of total mercury in fish brain tissues (r = -0.623; p<0.05; Table S14, Extended data¹⁶ and Underlying data³⁴). There was no relationship between the age of fish and the level of total mercury in the fish brain, liver, and muscle tissues. (Table S14, Extended data¹⁶).

The relationship between the type of fish culture practiced and the levels of total mercury in fish tissues

The type of fish culture practiced did not affect the mean levels of total mercury in fish brain tissues (Table S15, Extended data¹⁶). However, the type of culture practiced affected the mean levels of total mercury in fish muscle and liver tissues (Table S15, Extended data¹⁶).

Risk-based consumption limits

The health risk indices from feeding on fish sampled from the different study sites are summarized in Table 3. Based on these results, feeding on fish sampled from Sori beach, Masara, Siginga beach, and Kokaka was associated with significant exposure to mercury as the CRmw was the least (Table 4). On the other hand, the risk of exposure to mercury was least in Kamagambo South on account of the large consumption rate in meals per week (Table 4).

Underlying data

This project contains the following underlying data:

Figshare: Underlying data (1) of the study 'Levels of Mercury in Nile tilapia (Oreochromis niloticus), water, and sediment in the Migori gold mining belt, Migori, Kenya, and the potential ramifications to human health by Kola et al. https://doi.org/10.6084/m9.figshare.8869967.v3⁵⁰

This project contains the following underlying data:

Figshare: Underlying data (2) on the study 'Levels of Mercury in Nile tilapia (Oreochromis niloticus), water, and sediment in the Migori gold mining belt, Kenya and the potential ramifications to human health by Kola et al. https://doi.org/10.6084/m9.figshare.8869964.v2³⁴

This project contains the following underlying data:

Extended data

Figshare: Extended data on the study 'Levels of Mercury in Nile tilapia (Oreochromis niloticus), water, and sediment in the gold mining belt of Migori, Kenya, and the potential ramifications to human health' by Kola et al. https://doi.org/10.6084/m9.figshare.8870053.v4¹⁶

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).