Polymorphisms in Genes Encoding Potential Mercury Transporters and Urine Mercury Concentrations in Populations Exposed to Mercury Vapor from Gold Mining

Background: Elemental mercury (Hg0) is widely used in small-scale gold mining. Persons working or living in mining areas have high urinary concentrations of Hg (U-Hg). Differences in genes encoding potential Hg-transporters may affect uptake and elimination of Hg. Objective: We aimed to identify single nucleotide polymorphisms (SNPs) in Hg-transporter genes that modify U-Hg. Methods: Men and women (1,017) from Indonesia, the Philippines, Tanzania, and Zimbabwe were classified either as controls (no Hg exposure from gold mining) or as having low (living in a gold-mining area) or high exposure (working as gold miners). U-Hg was analyzed by cold-vapor atomic absorption spectrometry. Eighteen SNPs in eight Hg-transporter genes were analyzed. Results: U-Hg concentrations were higher among ABCC2/MRP2 rs1885301 A–allele carriers than among GG homozygotes in all populations, though differences were not statistically significant in most cases. MRP2 SNPs showed particularly strong associations with U-Hg in the subgroup with highest exposure (miners in Zimbabwe), whereas rs1885301 A–allele carriers had higher U-Hg than GG homozygotes [geometric mean (GM): 36.4 µg/g creatinine vs. 21.9; p = 0.027], rs2273697 GG homozygotes had higher U-Hg than A–allele carriers (GM: 37.4 vs. 16.7; p = 0.001), and rs717620 A–allele carriers had higher U-Hg than GG homozygotes (GM: 83 vs. 28; p = 0.084). The SLC7A5/LAT1 rs33916661 GG genotype was associated with higher U-Hg in all populations (statistically significant for all Tanzanians combined). SNPs in SLC22A6/OAT1 (rs4149170) and SLC22A8/OAT3 (rs4149182) were associated with U-Hg mainly in the Tanzanian study groups. Conclusions: SNPs in putative Hg-transporter genes may influence U-Hg concentrations.


Introduction
Mercury (Hg) is a ubiquitous and very reactive toxic metal. One of the most important sources of inorganic Hg exposure is gold mining, as elemental Hg (Hg 0 ) is widely used for gold extraction in small-scale gold mining (United Nations Environment Programme 2002).
Hg 0 is well absorbed in the lungs and has a high neurotoxic potential, due to its ability to pass the blood-brain barrier (Nordberg et al. 2007). Hg 0 is oxidized into mercuric Hg (Hg 2+ ) in red blood cells and other tissues (Nordberg et al. 2007). Gold miners are at risk of mercury poisoning (Baeuml et al. 2011b;Bose-O'Reilly et al. 2010a, 2010bDrasch et al. 2001, Harari et al. 2012, Lettmeier et al. 2010, and Hg 0 pollution from the gold-extraction process may also affect the health of individuals living close to mining areas (Bose-O'Reilly et al. 2010a, 2010b. It is estimated that about 10-15 million individuals work as gold miners, and approximately 30-50 million people live in small-scale gold-mining areas (Spiegel et al. 2005). Earlier studies in the Philippines (Drasch et al. 2001 (Lettmeier et al. 2010), and Ecuador (Harari et al. 2012) have shown that many individuals who work or live in gold mining areas are highly exposed to inorganic Hg from gold mining, as can be seen from high concentrations of Hg in urine (U-Hg). However, large differences in U-Hg levels among individuals with similar exposures have been reported.
5 Hg 2+ formed from Hg 0 is bound to glutathione and is excreted mainly in the urine. In contrast, there is little urinary excretion of methyl-Hg, a Hg species for which exposure is mainly by consumption of fish. Hence, U-Hg is therefore mainly a biomarker of exposure to inorganic Hg (Hg 0 and Hg 2+ ). Levels of U-Hg associated with detectable health effects in exposed workers fall between 10 and 30 µg/L (1 µg/L ≈ 1 µg/g creatinine) (Nordberg et al. 2007). U-Hg levels are lower in those who are not occupationally exposed to mercury. For example, people in the Baltic region without amalgam fillings and with normal fish consumption had U-Hg levels < 1 µg/L, while those who had dental amalgams had a few µg/L of U-Hg, however concentrations up to 50 ug/L have been reported for individuals with many fillings (Skerfving et al. 1999).
As with other metals (Nordberg et al. 2007), genetics probably influences Hg uptake, distribution, and excretion. The few genetic variants that have been associated with Hg toxicokinetics identified to date have mainly been in glutathione-related genes (Custodio et al. 2005;Harari et al. 2012;Schlawicke Engstrom et al. 2008).
The different forms of Hg is thought to be transferred across cell membranes-either into or out of cells-by transporter proteins. The capacity of specific proteins for Hg transport has not been extensively studied and no specific Hg transporters have been identified. However, Hg forms complexes with small molecules such as amino acids that can mimic essential molecules recognized by transporter proteins. Some multispecific transporter families, such as organic anion transporters (OATs), system L-amino acid transporters (LAT), and ATPbinding cassette transporters (ABC transporters) are thought to alter Hg toxicokinetics (Bridges and Zalups 2005). 6 OATs are multispecific transport proteins located on the basolateral membrane. They mediate the uptake of a variety of substrates from renal blood. In rats, OAT1 and OAT3 have been shown to be responsible for over 60% of the renal tubular uptake of Hg 2+ from renal blood into proximal tubule epithelial cells (Bridges and Zalups 2005).
LAT members transport amino acids into cells in exchange for other amino acids. LAT1 and LAT2 can transport cysteine-bound methyl-Hg, a complex that mimics certain amino acids (Bridges and Zalups 2005). LATs require interaction with SLC3A2 to function as transporters (del Amo et al. 2008), but nothing is known about their involvement in inorganic Hg transport.
ABC transporters participate in the transport of numerous toxicants across membranes, including anticancer agents and metals, and are major candidate genes for Hg transport. ABCC1, ABCC2, and ABCB1 (also known as multidrug resistance-associated proteins: MRP1, MRP2 and MDR1), are the most well-characterized ABC transporters. Upregulation of ABC transporters by Hg 2+ has been demonstrated in canine kidney cells (Aleo et al. 2005). ABCC2/MRP2 has been shown to mediate excretion of Hg 2+ from proximal tubular cells into urine in rats (Bridges et al. 2011).
The aim of the present study was to estimate associations between single nucleotide polymorphisms (SNPs) in potential Hg-transporter genes and U-Hg as an indication of possible effects of the SNPs on the toxicokinetics of inorganic Hg.

Study subjects
Participants included in the present analysis (N=1,017) were gold miners, individuals living in gold-mining areas, or control individuals living in areas with no gold mining, including men and women > 12 years of age from the Philippines (N=245) (Drasch et al. 2001 All participants were volunteers who provided written informed consent. The health examinations were performed in close cooperation with UNIDO and the United Nations Development Programme, as well as the regional health authorities and national ministries of health. All conditions were agreed upon between the governments and UNIDO, including extensive legal, formal, and ethical considerations. The conditions were approved by the governments and complied with all relevant national, state, and local regulations.

Categories and biomarkers of Hg exposure
Based on a questionnaire, the participants were divided into three exposure subgroups: (1) controls (with no particular Hg exposure from gold mining, N=139), (2) low-exposure individuals (living in Hg-contaminated mining areas, N=297), and (3) high-exposure individuals (gold miners working with Hg 0 , N=581). Participants in control subgroups were living in areas without any gold mining. Participants in the low-exposure subgroups were living in gold mining areas and were exposed to the Hg-contaminated environment.
Participants in the high-exposure subgroups were living in the same areas as the lowexposure participants and were working as miners with immediate contact with Hg, such as panning the crushed ore together with liquid Hg, or smelting the amalgams. Six of the 1,017 study participants did not have data on urinary mercury.
All U-Hg analyses were performed in the Institute of Forensic Medicine (LMU). Urine was analyzed for total Hg by cold-vapor atomic absorption spectrometry using a Perkin-Elmer 1100 B spectrometer after capturing the mercury on a gold amalgam net, with a MHS 20 attachment. Sodium-boro hydride was applied for reduction of Hg. The detection limit was 0.20 µg/L. For results below the limit of detection, the value was set to half the detection limit. U-Hg was adjusted for creatinine, analyzed by the Jaffe method (Mazzachi et al. 2000).
Strict internal and external quality control was ensured. Quality controls with reference urine samples (ClinCal® urine (41 µg Hg/L) and ClinCeck® urine level I (3.5 µg Hg/L) and II (40 µg Hg/L) for trace elements (RECIPE Chemicals + Instruments GmbH, Munich, Germany) were continuously performed to ensure the accuracy of the results. The laboratory has demonstrated good performance in external quality control tests for mercury in human specimens. The laboratory has participated in round robin tests by the German DGKL ("Deutsche Vereinte Gesellschaft für klinische Chemie und Laboratoriumsmedizin e.V.) for Hg in biological samples since 1998.

9
Genotyping DNA was extracted from whole blood using the Qiagen DNA Blood Mini kit (Qiagen, Hilden, Germany). We selected SNPs based on three different criteria: (1) functional impact according to the literature; (2) potential functional impact according to position and type of SNP (specifically, non-synonymous SNPs that may affect the protein structure/enzyme activity or 5' SNPs at putative promoter sites that can affect gene expression) (NCBI 2006); and (3) tagSNPs that capture as much of the genetic variation within a gene segment as possible due to linkage disequilibrium (LD) with other SNPs. TagSNPs were selected according to HapMap data (Thorisson et al. 2005) for YRI (Yoruba in Ibadan, Nigeria) as a proxy for the African populations, and CHB (Han Chinese in Beijing, China) and JPT (Japanese in Tokyo, Japan) for Asians. In total, 18 SNPs were analyzed in genes coding for eight different transporters (Table 1). Genotyping was done using the iPLEX® Gold assay on the MassARRAY platform (Sequenom TM ) for 15 SNPs. Additionally, three SNPs in MRP2 (rs717620, rs2273697, and rs3740066) were analyzed by Taqman allelic discrimination assay on an ABI 7900 instrument (Applied Biosystems, Foster City, CA, USA), according to the manufacturer's instructions. Genotyping failed for 2 SNPs (MRP1 rs41395947 and SLC3A2 rs2282477); for the other SNPs, Sequenom TM genotyping failed for ≤ 2% and Taqman genotyping failed for ≤ 4% of study participants. Five percent of the samples were reanalyzed for quality control purposes. There was perfect agreement between original and repeat genotyping runs for all SNPs.

Bioinformatics
Potential functional effects of SNPs with at least one statistically significant association (p <0.05) with U-Hg in at least one study group were evaluated in silico. Promoter sites and potential transcription factor binding sites were analyzed by Genomatix software suite (Genomatix Software GmbH, Munich, Germany). This was evaluated for SNPs situated in the 5´ UTR or introns.

Statistical analysis
In each country, deviations from Hardy-Weinberg equilibrium were tested using chi-square analysis and LD was evaluated using Haploview (Barrett et al. 2005).
Natural log (ln) transformed U-Hg was the dependent variable. The general linear model procedure was used, in which genotypes with three levels can be modeled as nominal variables without assuming additive effects.
First, we estimated genetic associations with U-Hg for all individuals in a country with adjustment for exposure subgroup (controls, low or high exposure) as categorical (nominal) variables, and after stratification by low or high exposure subgroup. In addition, we estimated associations adjusted for exposure subgroup according to continent (Africa, including Tanzania and Zimbabwe; or Asia, including Indonesia and the Philippines). We refer to these models as Model 1. Numbers of participants in the control groups were too small for stratumspecific analyses. Genotypes were modeled as categorical (nominal variables) coded 0, 1 or 2 according to the number of variant alleles, or were dichotomized with the common homozygous genotype. The genotype denoted last in the text in the Results section (first column in the tables) was used as reference (the common homozygote or the combination of the heterozygote and the common homozygote). However, in order to facilitate the interpretation of the results, other genotypes were sometimes used as reference, based on effect estimates.

11
Next, we estimated associations between genotypes (categorical, as above) and U-Hg by country or continent with adjustment for age (continuous) and gender, in addition to adjusting for exposure subgroup (categorical) or stratification according to low or high exposure (Model 2). In the analyses stratified for exposure group, adjustments were also made for Hg storage (categorical, categorized as never, at work, or at home). We did not include Hg storage in the analyses by country or continent because of strong co-linearity with exposure subgroup (Pearson product-moment correlation coefficient, r=0.8). Alcohol intake (never, > once a month, > once a week, or > once a day based on their response to the question "do you drink alcohol?", or categorized as "yes" or "no" for the Philippines) and BMI (continuous) were not included in models because they were correlated with gender (r around 0.5).

Background data
The study groups with the highest and lowest geometric mean (GM) creatinine-adjusted U-Hg concentrations were from Zimbabwe and Tanzania, respectively (Table 2). Correlations (r) between U-Hg and potentially influential variables in the different populations are shown in Supplemental Material, Table S1. U-Hg was significantly correlated (all p<0.001) with exposure subgroup (r=0.46-0.85), Hg storage (r=0.33-0.72), and male gender (r=0.22-0.37) in all 4 countries, but was only weakly correlated with age (r= -0.12-0.093, p>0.05). There 12 was a positive correlation between U-Hg and alcohol consumption (r=0.14-0.33, p-values of <0.050; data not available for the Philippines) and positive correlations between alcohol consumption and other factors such as gender and exposure group (r=0.3-0.5).
Genotyping was successful for 16/18 SNPs (Table 1). The minor alleles were the same in each country, for all SNPs (calculated for the total study population from each country, including the controls) except for MDR1 rs2032582 in the Philippines. There were few differences in allelic frequencies among the different exposure subgroups within each country (Fisher's exact test, not in table). In Indonesia, the allele frequency for the control group (N=19) deviated from the low exposed for one SNP (rs717620). All SNPs were in Hardy-Weinberg equilibrium in all countries, except for MRP2 rs717620 in Indonesia (Chi-square 7.1, p<0.05) and Philippines (Chi-square 23, p<0.05).
SNPs that were statistically significantly associated with U-Hg (MRP2, OAT1, OAT3 and LAT1) in at least one country/subgroup are presented in the results section. For the other SNPs evaluated, no statistically significant associations with U-Hg were found.
Unstandardised β-coefficients derived from the comparison with the reference genotype, i.e., the estimated difference in ln-transformed U-Hg for two genotypes, and corresponding pvalues and standard error (SE) are presented (Table 3 for MRP2 and Table 4 for LAT1, OAT1 and OAT3). P-value for the genotype term in the model are presented in the tables as well as when specifically stated in the text. All above are derived from Model 2 (adjusted model) if not stated otherwise.

MRP2 and U-Hg
The MRP2 rs1885301 A-allele carriers had higher observed creatinine-adjusted geometric mean U-Hg concentrations than those with the GG genotype in most populations (overall and according to high or low exposure within each country) (Supplemental material, Table S2).
Bioinformatic analyses indicated that rs1885301 is situated 5' of the promoter and that the G>A substitution results in loss of one transcription-factor binding site (Fetal Alz-50 clone 1 (FAC1) and gain of another site (FAST-1 SMAD interacting protein).
MRP2 rs717620 was rare in the African countries, where frequencies of the A-allele were only 1%, as compared to 15-20% in Asia. Individuals with one or two copies of the A-allele had higher U-Hg concentrations in all exposure subgroups in which at least four individuals carried this allele (Supplemental Material, Table S2). This was statistically significant in the Philippines (p=0.041, β=0.33, SE=0.16, AA vs. AG+GG). In the pooled analyses, associations were in the same directions: individuals with one or two copies of the A-allele had higher U-Hg concentrations (Asia: p=0.010, β=0.31, SE=0.12, AA vs. AG+GG; Africa: p=0.26, β=0.59, SE=0.53, AA vs. AG+GG). As noted above, this SNP was in Hardy-Weinberg disequilibrium in the Asian populations. In silico analysis indicated that rs717620 is situated in a promoter site, and that the G>A substitution leads to loss of two transcriptionfactor binding sites (LEF-1 and SOX4).

Page 13 of 31
14 The non-synonymous MRP2 rs2273697 was significantly associated with U-Hg among high exposed miners in Zimbabwe only, such that carriers of the A-allele (Ile) had lower U-Hg concentrations than individuals with no copies of the A-allele (p=0.001, β=-0.78, SE=0.22, AA+AG vs. GG). The A-allele was also associated with lower U-Hg among high-exposed miners in the Philippines (who had the second-highest U-Hg among all population subgroups), although this association was not statistically significant (p=0.55, β=-0.19, Overall, associations with the three MRP2 SNPs discussed above were particularly strong in the highest exposed subgroup, gold miners in Zimbabwe. In this subgroup, rs1885301 Aallele carriers had higher U-Hg than those with the GG genotype (observed GM: 36.4 vs. 21.9 µg/g creatinine, Model 1 unadjusted p=0.075, β=0.51, SE=0.23, AA+AG vs. GG), rs2273697 GG genotype had higher U-Hg concentrations than A-allele carriers (observed GM: 16.7 vs.37.4 µg/g creatinine, Model 1 unadjusted p=0.001, β=0.81, SE=0.24, AA+AG vs. GG), and rs717620 A-allele carriers had higher U-Hg concentrations than those with the GG genotype (observed GM 83.0 vs.28 µg/g creatinine, Model 1 unadjusted p =0.084, β=1.1, SE=0.63, AA+AG vs. GG).
Rs2273697 was not in linkage disequilibrium (LD) with rs1885301 or rs717620 in any population. Rs717620 was in LD with rs1885301 (the common G-alleles of both SNPs were linked to each other) in Indonesia (R 2 =0.69) and the Philippines (R 2 =0.40), but not in the African study groups (R 2 =0.02 in both countries).

LAT1 and U-Hg
For LAT1 rs33916661, individuals with the rare GG genotype showed higher observed GM U-Hg concentrations than individuals carrying GA or AA genotypes when all participants were combined by country or continent as well as in the high exposed subgroups (Supplemental Material, Table S3). However, the associations were statistically significant only in Tanzania (  Table S3). This SNP is situated in the 5' untranslated region, and in silico analysis indicated that it results in the loss of two transcription-factor binding sites: Paired box 5 Bcell-specific activator protein new (PAX5), and Avian C-type LTR CCAAT box binding factors, Zinc finger protein 217 (ZNF217).

OATs and U-Hg
Two of three SNPs evaluated in OATs (OAT1 rs4149170 and OAT3 rs4149182) were associated with lower U-Hg in the African populations (Table 4). These genotypes were rare in the Asian populations (2-4%). In most cases, individuals with the OAT1 rs4149170 AA or the OAT3 rs4149182 CC genotype had lower GM U-Hg concentrations than those with AG or GG and CG or GG genotypes, respectively (Supplemental Material, Table S3). Statistically significant associations were seen in all groups in Tanzania: individuals with AA or CC genotypes had lower U-Hg concentrations than the GG genotypes (reference). For rs4149170, p-values for the genotype term were 0.025 (low exposed), 0.024 (high exposed) and 0.023 (all individuals). When comparing AA with GG: p=0.022, β=-0.79, SE=0.33 for low exposed, p=0.039, β=-0.70, SE=0.34 for high exposed, and p=0.024, β=-0.54, SE=0.24 for all individuals. For rs4149182, p-values for the genotype term were 0.004 (low exposed), 0.059 (high exposed), and 0.017 (all individuals). When comparing CC with GG: p=0.002, β=-1.05, SE=0.32 for low exposed, p=0.046, β=-0.69, SE=0.34 for high exposed, and p=0.010, β=-0.62, SE=0.24 for all individuals. In these groups, the U-Hg concentrations were about 50% lower for individuals with AA or CC compared to GG (Supplemental Material, Table S3). Individuals with AA or CC genotypes had lower U-Hg in Zimbabwe, although these associations did not reach statistical significance. No statistically significant associations were seen for these SNPs when the African or Asian study groups were pooled.
OAT1 rs4149170 and OAT3 rs4149182 were in LD in the African populations (the common G-alleles of both SNPs were linked to each other: Tanzania, R 2 =0.82; Zimbabwe, R 2 =0.55) but this was not found in the Asian populations (Indonesia, R 2 =0.06; Philippines, R 2 =0.01).
In silico analyses indicate that rs4149182 in OAT3 is situated in a binding site for metalregulatory transcription factor 1 (MTF1), and that the C>G transversion results in loss of the MTF1-binding site.
As there were correlations between alcohol consumption and gender, and BMI and gender, we did not include them in the models above in order to avoid over-adjustment. However, we conducted a sensitivity analysis and compared model estimates adjusted for either alcohol consumption or BMI instead of gender. Results were generally consistent with those adjusted for gender (data not shown). Since gender was correlated with U-Hg concentrations we also compared associations between men and women by including an interaction term (gene x gender) in the models or by stratifying by gender in addition to the stratifications for country/continent and exposure group. However, no clear patterns could be seen (data not shown), probably due to a too low power.

Discussion
In this study, we report evidence suggesting that variations in the transporter gene MRP2 influence U-Hg in individuals exposed to elemental Hg. Associations between MRP2 rs1885301 and U-Hg were observed in study populations from two continents, and associations with two MRP2 SNPs (rs1885301 and rs2273697) appeared to be particularly strong in gold miners in Zimbabwe, the study subgroup with highest exposure.
Polymorphisms in LAT1, OAT1, and OAT3 were associated with U-Hg, but statistically significant associations were mainly observed in Tanzania. In silico analyses indicated that several of the SNPs associated with U-Hg alter putative transcription-factor binding sites, which indicate that they may influence gene expression.
This study had several strengths. The exposed populations were fairly large, ranging between 171 and 309 individuals in the different countries, although there were relatively few participants in the control populations that were not exposed to elemental mercury used in gold mining. U-Hg determinations and genotyping were done by the same methods in the same laboratories, and with good quality control. U-Hg concentrations indicated a wide range of exposures (with GM in the different exposure groups ranging from 0.12 to 29 µg/g creatinine in Zimbabwe participants, for example.) For some polymorphisms, however, the frequencies were low, which limited our ability to identify associations or consistent patterns across populations. We cannot exclude the possibility that differences in genetic background between exposure subgroups within each country might exist for other genes that influence Hg metabolism, since controls were not always living in exactly the same area, and gold miners were work migrants in many cases.
The allelic frequencies varied among the populations, as did the LD patterns (data not shown). Thus, we performed separate analyses for each country. We had a problem of multiple inference. However, the main candidate gene, well-founded in the literature, was MRP2, while associations with U-Hg for the other genes evaluated were plausible but less well supported. To increase our power to detect associations we merged the African and Asian populations in separate analyses. These results must be interpreted with caution due to differences in mining practices that may have influenced exposure levels, including differences in frequency and intensity of gold-amalgam burning, which is the main route of Hg 0 exposure (Harari et al. 2012). The Asian countries were more similar in exposure and U-Hg concentrations than the African ones, and thus provided the strongest results in the merged analyses. There were no large variations in allele frequencies between the countries within the pooled Asian and African groups.
Hg excretion among gold miners varies over time, in association with the time since the last burning of gold amalgam (Harari et al. 2012), which could obscure genetic effects, if present.
This may explain why associations with SNPs in OAT1, OAT3 and LAT1 mainly were seen in Tanzania, where the exposure was lowest, but was probably the most stable over time.

19
Alcohol intake may modify the levels of U-Hg. Alcohol consumption has been shown to increase the rate of exhalation of mercury and to reduce U-Hg levels (Clarkson and Magos 2006). However, we decided not to include it in the statistical models in order to avoid overadjustment; alcohol intake was positively associated with both exposure group and gender.
The positive correlation between alcohol consumption and U-Hg in our study populations was inconsistent with expectations, which makes us believe that the association was confounded by a positive association between alcohol consumption and exposure group/gender rather than being a true effect.
Gender differences in metal toxicokinetics and toxicodynamics have been described for some metals, including methylmercury (Vahter et al. 2007). Gender was correlated with U-Hg in our study populations, but was also associated with other factors related to exposure (alcohol consumption, BMI, work tasks -men were more likely than women to report that they worked with smelting amalgam). We did not find any gene-gender interactions for U-Hg, but we had limited power for these analyses, especially in the high-exposed groups where women constituted 20-30% of the study participants.
Transporter proteins are involved in active uptake of compounds, and as the Hg uptake in the lung is through passive absorption, we believe it is unlikely that lung absorption of Hg 0 is affected by transporter genes (Nordberg et al. 2007). Our hypothesis is that, given the same exposure, individuals with higher U-Hg would have enhanced excretion.

MRP2 is an important regulator of metal toxicity in different species, such as in rats and
Arabidopsis thaliana (Liu et al. 2001;Song et al. 2010). Hg induces upregulation of MRP2 and expression of MRP2 protein in vitro (Aleo et al. 2005). MRP2 transports Hg out of cells and MRP2 levels are high in the kidney (Wu et al. 2009). Increased MRP2 expression in the kidney may increase urinary Hg excretion. Two of the MRP2 SNPs that we evaluated (rs1885301 and rs717620) are situated 1,500 base pairs apart, both at the 5' end of the gene, which is associated with regulation of gene expression. However, there are no data for the influence of the gene expression for rs1885301. For rs717620, the data are inconclusive; no effect of the variant A-allele was found on gene or protein expression levels in liver cells or enterocytes in three studies analyzing expression (Deo et al. 2012, Moriya et al. 2002, Zhang et al 2010, whereas lower mRNA levels in kidney tissue were reported in one study (Haenisch et al. 2007). Since the in silico analysis showed a loss of two activating transcription factors for the rs717620 A-allele, this finding is in line with the lower mRNA levels found in Haenisch et al. We did not identify any reports concerning effects of MRP2 SNPs on metal transport.
Previously, LAT1 has only been shown to transport methyl-Hg in Xenopus laevis oocytes (Simmons-Willis et al. 2002). LAT1 rs33916661 is situated at the 5' end of the gene. Little is known about the functional impact of this polymorphism. Our in silico analysis suggests that the GG genotype, which was associated with higher U-Hg concentrations in most population subgroups, results in the loss of two transcription-factor binding sites and a gain of one new site.

21
OAT1 and OAT3 mediate uptake of Hg 2+ into cells, particularly in the kidney (Bridges and Zalups 2005). Rs4149182 was found to be situated in an MTF1-binding site, a so-called metal-responsive element. MTF1 induces a positive regulation of transcription in response to metal ions. The C>G transversion resulted in loss of an MTF1-binding site.
Further functional studies are, however, needed to prove causation between the SNPs identified and Hg metabolism. False-positive results may have arisen due to multiple testing of several genetic markers in different populations. Furthermore, the SNPs identified may not be functional, but may be linked to other unknown and truly functional SNPs. There is also a problem with misclassification of exposure subgroups, as the miners were placed in the same category according to their working circumstances and not based on true exposure levels.

Conclusions
Several SNPs in genes coding for transporters were associated with U-Hg concentrations in at least some study populations. Results support the hypothesis that transporters may be involved in the toxicokinetics of Hg. In particular, some MRP2 variants were associated with U-Hg in study groups from different countries.  a Adjusted for age, gender and Hg storage in the "low" and "high" groups, and for age, gender and exposure subgroup in the "All", "Asia" and "Africa" groups. Genotype, gender, exposure subgroup and Hg storage were modeled as categorical (nominal) variables, while age was modeled as a continuous variable. b Natural ln transformed. c Unstandardised β-coefficients derived from the comparison with the reference genotype, i.e., the estimated difference in ln-transformed U-Hg for two genotypes, as well as p-values and SE for these comparisons. d P-value for the genotype term in the model. e N = number of individuals. f Control subgroups are not shown separately, but are included in the "All", "Asia" and "Africa" groups. g Reference genotype. h Rs2273697 AA+AG genotypes were pooled in all populations due to a low number of individuals with AA genotype. i Denotes that no analyses were run, due to a low number of individuals (3 or less) for all genotypes but one. j Denotes that the genotype is pooled with the heterozygotes, due to a low number of individuals (3 or less).