https://orcid.org/0000-0002-2417-3331
Figures
Abstract
Background
In Bangladesh, > 50 million individuals are chronically exposed to inorganic arsenic (iAs) through drinking water, increasing risk for cancer and other iAs-related diseases. Previous studies show that individuals’ ability to metabolize and eliminate iAs, and their risk of toxicity, is influenced by genetic variation in the AS3MT and FTCD gene regions.
Methods
To identify additional loci influencing arsenic metabolism, we used data from Bangladeshi individuals to conduct genome-wide association analyses of the relative abundances of arsenic species measured in both urine (n = 6,540) and blood (n = 976). These species include iAs, monomethylated arsenic (MMA) and dimethylated arsenic (DMA) species.
Results
In analyses of urine arsenic species, we identified a novel association signal in the FMO gene cluster (1q24.3), with the lead SNP residing in FMO3 (MMA% P = 4.2x10-16). In analyses of blood arsenic species, we identified an additional signal in the FMO cluster, with the lead SNP residing in FMO4 (DMA% P = 2.3x10-22) and a novel signal at 10q25.1, with the lead SNP in GSTO1 (DMA% P = 5.3x10-13). Lead SNPs at FMO3 and GSTO1 are associated with the splicing of FMO3 and GSTO1, respectively, in multiple tissue types, but also contain missense variants. The lead SNPs at FMO4 are associated with FMO4 expression level in multiple tissue types. These newly identified SNPs did not show a clear association with risk for arsenic-induced skin lesions (P > 0.05), based on 3,448 cases and 5,207 controls.
Conclusion
We identified novel loci influencing arsenic metabolites measured in both urine and blood. FMOs are involved in the oxidation of xenobiotics but have no known direct role in arsenic metabolism, while GSTO1 has a well-established role in catalyzing the reduction of arsenic species. The novel associations we report appear specific to blood or urine, with no detectable impact on skin toxicity risk, pointing to complexities in arsenic metabolism and its genetic contributors that require further study.
Author summary
We conducted a study to better understand how humans differ in their ability to process arsenic, a harmful chemical found in drinking water that affects millions globally, particularly in Bangladesh. While we’ve already identified variants in specific genes that impact arsenic metabolism and elimination (AS3MT and FTCD), our research aimed to discover additional genetic influences. Using genetic information and arsenic measurements from both urine and blood samples of arsenic-exposed individuals, we identified new genetic regions, specifically within the FMO and GSTO gene clusters, that impact arsenic metabolism. Interestingly, our findings show that some genetic variants affect arsenic metabolites measured in urine, others in blood, and some in both. Unlike previously identified genetic factors that are linked to arsenic-induced skin lesions, these newly discovered genetic influences do not appear to have a clear association with skin toxicity. This finding points to complexities in arsenic metabolism we do not yet understand, with different genes affecting its processing in distinct ways depending on where arsenic species are measured in the body. Our work expands knowledge of how genetic variation influences arsenic metabolism and could help inform future public health strategies to protect vulnerable populations.
Citation: Tamayo LI, Tong L, Davydiuk T, Vander Griend D, Haque SE, Islam T, et al. (2025) Genetic variation in the FMO and GSTO gene clusters impacts arsenic metabolism in humans. PLoS Genet 21(9): e1011826. https://doi.org/10.1371/journal.pgen.1011826
Editor: Weichun Huang, US Environmental Protection Agency, UNITED STATES OF AMERICA
Received: July 31, 2024; Accepted: July 31, 2025; Published: September 2, 2025
Copyright: © 2025 Tamayo et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: GWAS summary statistics are provided at (https://uchicago.box.com/s/yfzeuvr9xd2n8k5yxbq666qqzhonesa1). Individual-level data are available upon reasonable request to the IPPH research group at IPPH@uchicago.edu.
Funding: This study was funded by NIEHS R35 ES028379 (to B.L.P.), NIEHS P30 ES027792 (to H.A.), NIDDK R01 DK123285 (to M.V.G.), and NIEHS R21 ES035491 (to D.V.G. and B.L.P.). The Funders have no role in study design, data collection and analysis.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Inorganic arsenic (iAs) is a human carcinogen [1], and exposure to iAs affects >200 million individuals worldwide through drinking water [2,3]. In Bangladesh, > 50 million individuals are chronically exposed to iAs through drinking water from naturally-contaminated wells that have some of the highest iAs concentrations reported in the world [3,4]. Chronic exposure to iAs above the World Health Organization (WHO) safety standard (>10 µg/L) increases the risk for multiple diseases, including cancer [5–7]. The most common signs of toxicity caused by iAs exposure are skin lesions, which appear as hyperpigmentation during early exposure and keratosis at advanced stages of exposure [8].
The liver is believed to be the primary site for the metabolism of ingested iAs, where iAs can undergo methylation catalyzed by arsenite (+3) methyltransferase (AS3MT) [9–11]. While the sequence of reactions involved in iAs biotransformation is still under debate [12], the resulting methylated arsenicals are well defined, and these include monomethylated forms of arsenic (MMAs) and dimethylated forms of arsenic (DMAs) [13]. DMA has a shorter half-life in circulation and is more readily eliminated in urine compared to MMA and iAs [14]. Consequently, DMA constitutes the majority of excreted arsenic species [15,16]. Hence, an individual’s arsenic metabolism efficiency (AME) can be defined as their capacity to methylate arsenic and generate DMA. The percentage of DMA among the total arsenic species in urine (DMA%) is commonly used as an indicator of AME [17–19].
There is inter-individual variation in AME [20–24] and this variability impacts the internal dose of arsenic and toxicity risk [4,25]. While age, sex, diet, body mass index, and smoking status are all likely to contribute to variability in AME [26], inherited genetic factors also play an important role. Prior GWAS and targeted sequencing studies of Bangladeshi individuals have identified four independent association signals for AME: three in the 10q24.32/AS3MT region (represented by rs4919690, rs11191492, and rs191177668) [20,27] and one in the FTCD gene (rs61735836) [19], a gene involved in histidine catabolism.
In addition to AS3MT and FTCD, there are likely to be other regions contributing to arsenic metabolism and toxicity risk [28]. Given the significant global health impact of arsenic exposure and the variability in AME among individuals, our objective is to identify novel genetic determinants of arsenic metabolism and toxicity phenotypes. Identifying such genetic determinants will implicate novel genes in AME, improving our understanding of biological processes underlying arsenic metabolism. Furthermore, knowledge of AME-associated variants could ultimately inform precision public health strategies focused on identifying highly susceptible groups who may benefit from targeted interventions, such as nutritional supplementation or enhanced exposure monitoring [29,30]. Here, we report the largest genome wide association study (GWAS) of urine arsenic species conducted to date and first GWAS blood arsenic species to identify inherited genetic effects on arsenic metabolism and toxicity.
Methods
Ethics statement
This research was approved by the Institutional Review Boards of the University of Chicago, Columbia University, and the Bangladesh Medical Research Council. Informed verbal consent was obtained from all participants.
Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S1 Checklist).
Study participants
DNA samples were obtained at baseline recruitment from The Health Effects of Arsenic Longitudinal Study (HEALS) [31] and the Bangladesh Vitamin E and Selenium Trial (BEST) [32]. The HEALS cohort is prospective study of health outcomes associated with arsenic exposure through drinking water of adults in located in Araihazar, Bangladesh, a rural area with substantial exposure to arsenic through naturally contaminated drinking water from local wells. HEALS began in 2000 and is comprised of >20,000 adult participants (followed over 15–20 years to ascertain health outcomes), of which 6,540 have genome-wide SNP data. Trained study physicians conducted in-person interviews, clinical evaluations, and urine collection at baseline follow-up visits. BEST is a randomized chemoprevention trial evaluating effects of vitamin E and selenium supplementation on skin cancer risk among arsenic-exposed individuals. BEST participants are residents of Araihazar, Matlab, and surrounding areas in Bangladesh. BEST has 7,000 adult participants, all with skin lesions at baseline, and 1,990 have existing genotype data. BEST uses many of the same study protocols as HEALS, including arsenic exposure assessment and biospecimen collection.
Some HEALS participants also participated in additional studies investigating correlates of AME and folate interventions to increase AME. We used data from 1,099 genotyped HEALS participants who participated in one of three such studies: the Nutritional Influences of Arsenic Toxicity study (NIAT, n = 163), the Folic Acid and Creatine Trial (FACT, n = 595), or the Folate and Oxidative Stress study (FOX, n = 341). These studies measured participants’ arsenic species in both blood and urine. Data on blood arsenic species were available for 977 of the 1099 genotyped individuals for the NIAT (n = 110), FOX (n = 273) and FACT (n = 594) studies. Blood was collected at baseline (Week 0) for FOX, at Week 0 and 12 for NIAT, and at Weeks 0, 1, 12 and 24 for FACT. Urine was collected at Week 0 for FOX, weeks 0 and 12 for NIAT, and Weeks 0, 1, 6, 12, 13, 18, and 24 for FACT.
Measurement of arsenic species
Both urinary and blood arsenic species were distinguished using a method described by Reuter et al [33]. This method entails using high-performance liquid chromatography separation of arsenobetaine, arsenocholine, AsV, AsIII, MMA, and DMA followed by detection by inductively coupled plasma-mass spectrometry with dynamic reaction cell. Because AsIII can oxidize to AsV during sample transport, storage, and preparation, we sum these two species to obtain total iAs (i.e., AsIII + AsV). The percentages of iAs, MMA and DMA in total arsenic were calculated by dividing the concentration of each species by the sum of iAs, MMA, and DMA. Data on arsenobetaine and arsenocholine, nontoxic organic arsenic from dietary sources, were not analyzed. Participants missing one of the four arsenic species were dropped from the analysis. Generation of all urine arsenic species for HEALS and HEALS ancillary studies (FOX [14], NIAT [34], FACT [35]) were conducted at Columbia University’s trace elements lab, as described previously [36] with the exception of ~1,800 HEALS urine samples which were more recently analyzed for arsenic species at the University of Alberta in the laboratory of Dr. Chris Le. Both labs use HPLC-ICP-MS to generate urine arsenic species measurements. Generation of all data on blood arsenic species used in this work has been described previously (for FOX14, NIAT32, FACT33). Bio-specimen measurements were averaged when data from multiple time points was available, as this increased power to detect known signals (AS3MT, FTCD).
Genotype data
6,665 HEALS and 1,990 BEST participants have been genotyped using either Illumina’s HumanCytoSNP-12 (299,140 SNPs), Infinium Multi-ethnic EUR/EAS/SAS arrays (1,475,140 SNPs), or the Global Screening Array (654,027 SNPs). Most participants with data from the HumanCytoSNP-12 also have complementary data from Illumina’s exome array. For each array, we removed non-rs SNPs, SNPs with a call rate of <90%, monomorphic SNPs, and samples with a call rate <90%. The Michigan Imputation Server [37] was used to genotype unmeasured SNPs using the Haplotype Reference Consortium (HRC) reference panel.31 Only high-quality imputed bi-allelic SNPs (imputation r2 > 0.3) and SNPs with minor allele frequency (MAF) >0.005 were retained (8,711,421 SNPs).
Ascertainment of skin lesion status
HEALS and BEST participants were clinically assessed for skin lesions across the entire body at each study visit [31,32]. A protocol similar to the quantitative assessment of the extent of a body surface involvement in burn patients [38] was used to quantify the size, shape and extent of skin lesion involvement. The principle is based on dividing the entire body skin surface into 11 segments (e.g., front of arm, back of arm, face) and assigning percentages to each of them based on their size relative to the whole body [39]. There are 1,458 (prevalent & incident) skin lesion cases among genotyped HEALS participants, 1990 genotyped cases from BEST, and 5,207 genotyped HEALS controls (with no history of lesions).
Statistical analysis
GWAS of arsenic species percentages (iAs%, MMA%, and DMA%) were conducted using mixed linear models as implemented in the GCTA software [40] to account for relatedness among individuals in our sample (previously described [20]). All models are adjusted for age, sex, and genotype batch.
GWAS of skin lesions status was run separately by genotyping batch. There were 4,806 participants that were genotyped using the HumanCytoSNP-12 array (2395 cases, 2411 controls), 466 genotyped using the HumanCytoSNP-12 array without exome data (92 cases, 374 controls), 2,486 genotyped using the Multi-Ethnic-Array (913 cases,1573 controls) and 1133 genotyped using the global screening array (48 cases, 849 controls). The 5 GWAS were then meta-analyzed using PLINK 1.9.
We identified SNPs that pass a standard genome-wide threshold (P-value < 5x10-8). For signals identified, we (1) ensured all QC metrics for identified SNPs were acceptable, (2) searched for secondary signals after adjusting for the top SNP, and (3) examined linkage disequilibrium for the top SNPs.
To investigate the role of identified SNPs in gene regulation we leveraged data on expression quantitative trait loci (eQTL) and splicing QTLs (sQTL) from the Genotype-Tissue Expression (GTEx) project [41]. To determine whether a common causal variant was responsible for both GWAS and eQTL/sQTL signals at a given locus we used colocalization methods. We used the coloc R package version 5.2.2. [42] a Bayesian framework to assess the probability of shared causal variants at a given locus for multiple traits using SNP-based summary statistics (from GWAS or QTL studies). Coloc estimates the posterior probabilities of different colocalization hypotheses, including one shared causal variant (H4) and two distinct causal variants (H3). We used coloc’s default priors.
Results
GWAS of arsenic species measured in urine
A GWAS of arsenic species measured in urine (DMA%, MMA%, and iAs%, distributions shown in S1 Fig) among 6,540 individuals identified previously reported associations in the AS3MT (DMA% P = 2.4x10-51) and FTCD (DMA% P = 2.2x10-48) regions (Fig 1). A novel association signal was identified in the FMO (Flavin-containing monooxygenase) gene cluster at 1q24.3, a region containing FMO1, FMO2, FMO3, and FMO4. FMOs metabolize xenobiotic chemicals through oxygenation, but currently there is no known role for FMO genes in arsenic metabolism. The minor allele (A, MAF = 6.1%) at lead SNP rs12406572, located in intron 7 of FMO3, was associated with increased urine DMA% (beta = 1.62, P = 3.9x10-8) and decreased urine MMA% (beta = -1.3, P = 3.1x10-16), but it did not show clear association with urine iAs% (beta = -0.28, P = 0.22) (Fig 2). There was evidence of a suggestive secondary association signal in this region, particularly for MMA%, represented by lead SNP rs3754494 (beta = 0.40, P = 6.5x10-7) (S2 Fig). All associations were robust to adjustment for BMI and smoking status.
Among the SNPs showing the strongest evidence of association was rs12406572 (MMA% P = 3.2x10-16), a missense variant in exon 7 of FMO3 that codes for a Glu (E) to Gly (G) amino acid substitution (CADD score of 24.4; SIFT: 0.01/deleterious, PolyPhen: 0.86/possibly damaging).
Lead uMMA% SNP rs12406572 was associated with alternative splicing of FMO3 in >10 GTEx tissue types, and the urine MMA% signal showed strong evidence of co-localization with a FMO3 sQTL in at least 7 GTEx tissue types (S1 Table), including liver (PP4 = 0.98, Fig 3), adipose (subcutaneous) (PP4 = 0.98) and lung (PP4 = 0.98). In liver, co-localization was observed for five different intron excision events (S1 Table). For example, the MMA%-decreasing allele (A) was associated with increased excision of an alternative intron (chr1:171092790:171107675) that includes exon 3, leading to loss of exon 3 in the mature mRNA. These findings suggest that the alternative splicing of FMO3 could play a role in modulating the metabolism of arsenic.
The splicing phenotype shown is excision of intron chr1:171092790:171107675.
In HEALS, the MAF of lead SNP is 6%, consistent with the MAF observed in 1 KG BEB group (6%). However, the MAF rs12406572 varies substantially across 1 KG populations (3% in SAS, 12% in AMR, 17% in EAS, 17% in EUR, and 36% in AFR), suggesting that the causal variant at this locus may make a larger contribution to variability in arsenic methylation capacity in populations outside of South Asia.
GWAS of arsenic species measured in blood
A GWAS of blood arsenic species (bDMA%, bMMA%, and biAs%) among 976 individuals identified association signals in the AS3MT and FTCD regions (Fig 4) that appear very similar to the association signals observed for urine arsenic species in those regions, in terms of both lead SNPs and direction of association (S3 Fig). In addition, we observed two novel association signals, one in the FMO gene cluster (1q24.3), spanning the FMO4 gene, and one signal spanning the GSTO1 and GSTO2 genes at 10q25.1, in close proximity (<2 Mb) to, but distinct from, the association signal at AS3MT/10q24.32 (Fig 5). All associations were robust to adjustment for BMI and smoking status.
The 1q24.3/FMO signal for blood DMA% spans the FMO4 gene and is distinct from the association signal observed for urine arsenic species that spans FMO3. The minor allele (C, MAF = 45%) at uDMA% lead SNP rs10912834 was associated with decreased blood DMA% (beta = -2.13; P = 2.3x10-22), increased blood MMA% (beta = 1.35; P = 1.2x10-11), and increased blood iAs% (beta = 0.78; P = 1.7x10-6). However, iAs% has a different lead SNP rs2011345 with minor allele C (MAF = 38%) associated with decreased blood iAs% (beta = -1.02; P = 3.4x10-9). These two SNPs have an LD r2 of 0.36, suggesting iAs% may have unique genetic determinants at this locus. The signal at FMO4 observed for blood arsenic species was not observed for arsenic species in urine, and the signal at FMO3 observed for urine metabolites was not observed for blood metabolites (Fig 6). The MAF of lead SNP rs10912834 in HEALS (45%) was consistent with the MAF observed in 1 KG BEB (46%) and SAS groups (45%). However, the MAF for rs10912834 varies somewhat across 1 KG populations (30% in AMR, 41% in EAS, 35% in EUR, and 57% in AFR). There was also evidence of a suggestive secondary association signal in this region, particularly for bDMA%, represented by lead SNP rs10798297 (beta = -1.16, P = 2.8x10-5, S4 Fig.
The FMO4 association signal for blood DMA% showed strong evidence of co-localization with a cis-eQTL for FMO4 that was present in at least 10 GTEx tissue types, including liver (PP4 = 0.84), pancreas (PP4 = 0.98) and artery-tibial (PP4 = 0.86) (Figs 7 and S2). The minor, DMA%-decreasing allele (C) was associated with increased expression of FMO4 in all of the GTEx tissues examined.
The association signal for blood arsenic species observed at 10q25.1 spanned GSTO1 and GSTO2, with GSTO1 having a well-established role in catalyzing the reduction of arsenic species (iAsV to iAsIII, MMAV to MMAIII, and DMAV to DMAIII) [43,44] The minor allele (T, MAF = 10.3%) at lead SNP rs34521730 was associated decreased blood DMA% (beta = -2.70; P = 5.3x10-13), increased blood MMA% (beta = 1.72; P = 3.5x10-7), and increased blood iAs% (beta = 0.98, P = 0.0004) (Fig 8). All associations were robust to adjustment for BMI and smoking status. SNPs in this region did not show evidence of association with arsenic species measured in urine.
Among the SNPs in the GSTO1 region showing the strongest evidence of association was rs11509438 (DMA% P = 5.3x10-13), a missense variant that codes for a GLU (E) to Lys (K) amino acid substitution. This variant has a CADD score of 0.46, a SIFT score of 0.36/tolerated, and a PolyPhen score of 0.007/benign.
Lead bDMA% SNP rs34521730 was associated with alternative splicing of GSTO1 in >40 GTEx tissue types. We tested the bDMA% signal for colocalization with the corresponding sQTL in liver and observed strong evidence of colocalization (PP4 = 0.99, Fig 9). In liver (and in many other tissue types), the minor, DMA%-decreasing allele (T) at rs34521730 was associated with increased excision of intron chr10:104259798:104262979 (intron 3) and decreased excision of intron chr10:104263077:104266084 (intron 4) (S3 Fig). (intron numbers as based on canonical GSTO1 isoform NM_004832.3/ENST00000369713.9).
In HEALS, the MAF of lead SNP rs34521730 is 10%, which is fairly consistent with the MAF observed in 1 KG BEB (15%) and SAS groups (12%). The minor allele is less common across other 1 KG populations (2% in AMR, 1% in EAS, 3% in EUR, and 2% in AFR).
We also conducted a GWAS of blood total arsenic (computed as the sum of arsenic metabolites iAsIII, iAsV, MMA, and DMA), but no clear associations were observed.
To assess the robustness of our findings, we also conducted a sensitivity analysis by adjusting our primary models for the top FMO3 (rs12406572, associated with uAsMet) and FMO4 (rs10912834, associated with bAsMet) SNPs for BMI and smoking status. The adjusted results remained consistent with the original analysis, indicating that the associations of interest are largely independent of these covariates beyond the arsenic methylation efficiency variables.
GWAS of arsenic-induced skin lesions
Using clinical data from genotyped participants from both HEALS and BEST (Bangladesh Vitamin E and Selenium Trial), we identified 3,448 participants with a diagnosis of arsenic-induced skin lesions (the most common sign of arsenic toxicity) and 5,207 participants without history of a diagnosis. We conducted a GWAS of arsenic-induced skin lesion status, and observed the strongest signal in the AS3MT region (P = 6.9x10-10; GC adjusted; P-value: 1.57x10-8, S5 Fig). For the AS3MT and FTCD SNPs that impact arsenic species in both urine and blood, the DMA%-decreasing alleles showed consistent evidence of association with increased skin lesion risk (Table 1), as we have reported previously [19] However, for the newly identified SNPs in the FMO and GSTO gene clusters, clear evidence of association with skin lesion risk was not observed. Mediation analyses were conducted for the 3 skin lesion-associated SNPs (rs4919690, rs191177668, rs6173583). We found that adjustment for DMA% substantially attenuates the association between these SNPs and skin lesion status (S4 Fig). These results support the hypothesis that genetic variability in arsenic metabolism genes impacts arsenic-related toxicity risk, with AME acting as a mediator.
Discussion
Using data from arsenic-exposed participants in Bangladesh, we performed GWAS of the relative abundances of three classes of arsenic species (iAs%, MMA%, and DMA%), measured in both urine and blood, and identified the FMO and GSTO gene clusters as novel regions in which genetic variation influences arsenic species composition in humans. While the variants we’ve identified previously (SNPs in the AS3MT and FTCD region) show association with arsenic species measured in both blood and urine, [45] the regions identified in this study show detectable association with arsenic species only in blood (GSTO1, FMO4) or only in urine (FMO3). Furthermore, the AS3MT and FTCD SNPs affecting arsenic metabolism show clear associations with risk of arsenic-induced skin lesions, while our newly identified metabolism-related variants (GSTO1, FMO3, and FMO4) do not.
Our observation that genetic effects on arsenic species can be detectable in blood, but apparently absent in urine (and vice versa), is somewhat unexpected given our prior findings that (1) AS3MT and FTCD SNPs have similar effects on metabolites in blood compared to urine and (2) arsenic species percentages (e.g., DMA%) measured in blood are positively correlated with the same metabolite percentage measured in urine [45]. These prior findings suggest that the arsenic species composition of urine reflects the composition of the blood that is being filtered by the kidney. However, our new findings point to complexities in arsenic metabolism, distribution, transport, and/or excretion that we do not yet fully understand. Such complexities could involve gene/enzyme functions that vary by cell or tissue type, important variability in the valence states of arsenic species that we are unable to measure in large numbers of human samples, and/or alternative pathways of metabolism and elimination (e.g., gut). Further research is needed to elucidate the precise mechanisms underlying these associations that differ between blood and urine arsenic species.
The association signal observed for urine species at FMO3 is unique because the association is more pronounced for MMA% compared to DMA% (and undetectable for iAs%). For all other regions identified to date (AS3MT, FTCD, FMO4, and GSTO1), the associations observed are most prominent for DMA%, suggesting a different toxicokinetic impact of the FMO3 causal variant compared to the other regions.
FMOs are known to be involved in the oxygenation of xenobiotics, but they have not previously been reported to have a role in arsenic metabolism. Oxidation by FMOs could potentially play a role in reversing the GSTO1-catalyzed reduction of arsenic species, which according to the Challenger pathway, is required prior to arsenic methylation. However, alternative pathways of arsenic metabolism have been proposed [9,10] in which trivalent arsenic species are directly methylated, without oxidation. The resulting trivalent methylated species can then undergo oxidation to form pentavalent species, but pentavalent species would then need to be reduced to trivalent species to undergo methylation (under the alternative pathway). Thus, under either the Challenger or alternative pathways, FMOs could play a role in determining the ratio of trivalent to pentavalent species, thus impacting the supply of arsenic species that can be methylated. However, additional research is needed to understand the role of FMO enzymes in the metabolism of arsenic.
SNPs in FMO genes have been reported in GWAS to be associated with blood cell traits, [46] metabolomics phenotypes [47], and hormonal phenotypes [48]. For example, our top FMO3 SNP (rs12406572) has been reported as a metabolite QTL for methylcysteine [46]. Mutations in FMO3 cause trimethylaminuria, also known as “fish odor syndrome”, a condition in which a defective FMO3 enzyme causes accumulation of trimethylamine, which results in a distinctive odor resembling rotten fish [49].
FMO2 has recently been shown to play a role in one-carbon metabolism in C.elegans, [50] an essential metabolic pathway that encompasses the folate and methionine cycles and provides one-carbon units for methylation reactions. This suggests the possibility that FMOs could influence arsenic methylation through effects on one carbon metabolism and the supply of methyl groups.
The associations identified for SNPs in the 10q25.1 region are likely to affect the function of GSTO1, a gene known to reduce pentavalent arsenic species to trivalent forms. Previous studies of candidate variants in GSTO1 [51,52] have provided only modest evidence of their involvement in risk of arsenic toxicity. Additionally, a study of variation in four GST genes among three Ecuadorian populations suggested that two non-synonymous GSTO1 variants may be under selective pressure, potentially due to environmental arsenic exposure [53]. However, the present study is the first to demonstrate a clear association between genetic variation in this region and arsenic species composition in human samples.
Previously identified SNPs in the AS3MT [54] and FTCD [19] regions show clear association of DMA%-decreasing alleles with elevated risk of arsenic-related skin lesions. This finding supports the hypothesis that genetic variants that decrease individuals’ capacity to methylate arsenic, hence reducing the elimination and increasing the internal dose of arsenic, will increase the risk of arsenic-related health conditions. However, newly identified SNPs in the FMO and GSTO gene clusters did not show clear associations with arsenic-induced skin lesion risk. This lack of association could be due to power limitations, but it is also possible that SNPs influencing the reduction (or oxidation) of arsenic species, as opposed to methylation, may influence the distribution of pentavalent versus trivalent species in human tissues, with pentavalent species being more toxic, particularly MMAIII [55]. While we cannot capture such effects in this study, they could have implications for toxicity risk, adding a layer of complexity to our underlying hypothesis that increasing DMA production should decrease toxicity risk.
For colocalization analyses, we leveraged eQTL and sQTL data from GTEx, a study of tissue donors largely of European ancestry. While the ancestry (and associated LD patterns) of GTEx are not well-matched to HEALS participants of Bangladeshi ancestry, our colocalization analyses produced strong posteriors, despite the LD mismatch.
Future directions for this research include replication of the reported associations in other populations with arsenic exposure and evaluating our metabolism-related SNPs for association with other arsenic-related health outcomes. Additional research is also needed to characterize the molecular mechanisms by which the identified variants impact gene function, their relevant cellular contexts, and their roles in pathways involved in arsenic metabolism. Importantly, identifying genetic variants associated with arsenic metabolism has the potential to inform precision public health strategies. By pinpointing genetically susceptible populations, these findings could support targeted interventions, such as nutritional supplementation or enhanced exposure monitoring, aimed at reducing arsenic-related health risks [29,30].
Supporting information
S1 Fig. Distribution of As metabolites.
A) Urine As Species (Total n = 6,540), Columbia, n = 3, 687, NIAT, n = 163, FACT, n = 594, FOX, n = 341, Alberta, n = 1800. B) Blood As Species (Total n = 977), NIAT, n = 110, FOX, n = 273, FACT, 594.
https://doi.org/10.1371/journal.pgen.1011826.s001
(TIFF)
S2 Fig. Conditional Analysis of the FMO3 signal (n = 6,540).
A) uDMA% results adjusted for lead SNP rs12406572. B) uMMA% results adjusted for lead SNP rs12406572.
https://doi.org/10.1371/journal.pgen.1011826.s002
(TIFF)
S3 Fig. Association of SNPs in the AS3MT region with A) bDMA%, bMMA%, and biAs% (n = 1,099); B) uDMA%, uMMA%, and uiAs% (n = 6,540).
https://doi.org/10.1371/journal.pgen.1011826.s003
(TIFF)
S4 Fig. Conditional Analysis of the FMO4 region (n = 1,099).
A) bDMA% adjusted for lead SNP rs10912834. B) bMMA% adjusted for lead SNP rs10912834 B) biAs% adjusted for lead SNP rs2011345.
https://doi.org/10.1371/journal.pgen.1011826.s004
(TIFF)
S4 Table. Mediation Analysis demonstrating that associations between SNPs and arsenic-induced skin lesions are attenuated by adjustment for urine DMA% (representing AME).
https://doi.org/10.1371/journal.pgen.1011826.s009
(XLSX)
S1 Checklist. Inclusivity in global research questionnaire.
https://doi.org/10.1371/journal.pgen.1011826.s010
(DOCX)
References
- 1. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans. Arsenic, metals, fibres, and dusts. IARC monographs on the evaluation of carcinogenic risks to humans. 2012;100(PT C):11.
- 2.
Ravenscroft P, Brammer H, Richards K. Arsenic pollution: a global synthesis. John Wiley & Sons. 2011.
- 3. Naujokas MF, Anderson B, Ahsan H, Aposhian HV, Graziano JH, Thompson C, et al. The broad scope of health effects from chronic arsenic exposure: update on a worldwide public health problem. Environ Health Perspect. 2013;121(3):295–302. pmid:23458756
- 4. Ahsan H, Chen Y, Kibriya MG, Slavkovich V, Parvez F, Jasmine F, et al. Arsenic metabolism, genetic susceptibility, and risk of premalignant skin lesions in Bangladesh. Cancer Epidemiol Biomarkers Prev. 2007;16(6):1270–8. pmid:17548696
- 5. Mink PJ, Alexander DD, Barraj LM, Kelsh MA, Tsuji JS. Low-level arsenic exposure in drinking water and bladder cancer: a review and meta-analysis. Regul Toxicol Pharmacol. 2008;52(3):299–310. pmid:18783726
- 6. Yuan Y, Marshall G, Ferreccio C, Steinmaus C, Liaw J, Bates M. Kidney cancer mortality: fifty-year latency patterns related to arsenic exposure. Epidemiology. 2010;:103–8.
- 7.
Smith AH, Lopipero PA, Bates MN, Steinmaus CM. Arsenic epidemiology and drinking water standards. American Association for the Advancement of Science. 2002.
- 8. Sengupta SR, Das NK, Datta PK. Pathogenesis, clinical features and pathology of chronic arsenicosis. Indian J Dermatol Venereol Leprol. 2008;74(6):559–70. pmid:19171978
- 9. Stýblo M, Venkatratnam A, Fry RC, Thomas DJ. Origins, fate, and actions of methylated trivalent metabolites of inorganic arsenic: progress and prospects. Arch Toxicol. 2021;95(5):1547–72. pmid:33768354
- 10. El-Ghiaty MA, El-Kadi AOS. The Duality of Arsenic Metabolism: Impact on Human Health. Annu Rev Pharmacol Toxicol. 2023;63:341–58. pmid:36100221
- 11. Drobná Z, Walton FS, Paul DS, Xing W, Thomas DJ, Stýblo M. Metabolism of arsenic in human liver: the role of membrane transporters. Arch Toxicol. 2010;84(1):3–16. pmid:20020104
- 12. Hirano S. Biotransformation of arsenic and toxicological implication of arsenic metabolites. Arch Toxicol. 2020;94(8):2587–601. pmid:32435915
- 13. Challenger F. Biological methylation. Chemical Reviews. 1945;36(3):315–61.
- 14. Howe CG, Niedzwiecki MM, Hall MN, Liu X, Ilievski V, Slavkovich V, et al. Folate and cobalamin modify associations between S-adenosylmethionine and methylated arsenic metabolites in arsenic-exposed Bangladeshi adults. J Nutr. 2014;144(5):690–7. pmid:24598884
- 15. Vahter M. Methylation of inorganic arsenic in different mammalian species and population groups. Sci Prog. 1999;82 ( Pt 1)(1):69–88. pmid:10445007
- 16. Loffredo CA, Aposhian HV, Cebrian ME, Yamauchi H, Silbergeld EK. Variability in human metabolism of arsenic. Environ Res. 2003;92(2):85–91. pmid:12854687
- 17. Tseng C-H. A review on environmental factors regulating arsenic methylation in humans. Toxicol Appl Pharmacol. 2009;235(3):338–50. pmid:19168087
- 18. Pierce BL, Tong L, Argos M, Gao J, Farzana J, Roy S, et al. Arsenic metabolism efficiency has a causal role in arsenic toxicity: Mendelian randomization and gene-environment interaction. Int J Epidemiol. 2013;42(6):1862–71. pmid:24536095
- 19. Pierce BL, Tong L, Dean S, Argos M, Jasmine F, Rakibuz-Zaman M, et al. A missense variant in FTCD is associated with arsenic metabolism and toxicity phenotypes in Bangladesh. PLoS Genet. 2019;15(3):e1007984. pmid:30893314
- 20. Pierce BL, Kibriya MG, Tong L, Jasmine F, Argos M, Roy S, et al. Genome-wide association study identifies chromosome 10q24.32 variants associated with arsenic metabolism and toxicity phenotypes in Bangladesh. PLoS Genet. 2012;8(2):e1002522. pmid:22383894
- 21. Agusa T, Fujihara J, Takeshita H, Iwata H. Individual variations in inorganic arsenic metabolism associated with AS3MT genetic polymorphisms. Int J Mol Sci. 2011;12(4):2351–82. pmid:21731446
- 22. Porter KE, Basu A, Hubbard AE, Bates MN, Kalman D, Rey O, et al. Association of genetic variation in cystathionine-beta-synthase and arsenic metabolism. Environ Res. 2010;110(6):580–7. pmid:20670920
- 23. Das N, Giri A, Chakraborty S, Bhattacharjee P. Association of single nucleotide polymorphism with arsenic-induced skin lesions and genetic damage in exposed population of West Bengal, India. Mutat Res Genet Toxicol Environ Mutagen. 2016;809:50–6. pmid:27692299
- 24. Balakrishnan P, Vaidya D, Franceschini N, Voruganti VS, Gribble MO, Haack K, et al. Association of Cardiometabolic Genes with Arsenic Metabolism Biomarkers in American Indian Communities: The Strong Heart Family Study (SHFS). Environ Health Perspect. 2017;125(1):15–22. pmid:27352405
- 25. Argos M, Ahsan H, Graziano JH. Arsenic and human health: epidemiologic progress and public health implications. Rev Environ Health. 2012;27(4):191–5. pmid:22962196
- 26. Shen H, Niu Q, Xu M, Rui D, Xu S, Feng G, et al. Factors Affecting Arsenic Methylation in Arsenic-Exposed Humans: A Systematic Review and Meta-Analysis. Int J Environ Res Public Health. 2016;13(2):205. pmid:26861378
- 27. Chernoff M, Tong L, Demanelis K, Vander Griend D, Ahsan H, Pierce BL. Genetic Determinants of Reduced Arsenic Metabolism Efficiency in the 10q24.32 Region Are Associated With Reduced AS3MT Expression in Multiple Human Tissue Types. Toxicol Sci. 2020;176(2):382–95. pmid:32433756
- 28. Gao J, Tong L, Argos M, Scannell Bryan M, Ahmed A, Rakibuz-Zaman M, et al. The Genetic Architecture of Arsenic Metabolism Efficiency:A SNP-Based Heritability Study of Bangladeshi Adults. Environ Health Perspect. 2015;123(10):985–92. pmid:25768001
- 29. Tamayo LI, Haque SE, Islam T, Ahmed A, Rahman M, Horayra A, et al. Returning personal genetic information on susceptibility to arsenic toxicity to research participants in Bangladesh. Environ Res. 2024;240(Pt 2):117482. pmid:37879393
- 30. Tamayo LI, Lin H, Ahmed A, Shahriar H, Hasan R, Sarwar G, et al. Research Participants’ Attitudes towards Receiving Information on Genetic Susceptibility to Arsenic Toxicity in Rural Bangladesh. Public Health Genomics. 2020;23(1–2):69–76. pmid:32069464
- 31. Ahsan H, Chen Y, Parvez F, Argos M, Hussain AI, Momotaj H, et al. Health Effects of Arsenic Longitudinal Study (HEALS): description of a multidisciplinary epidemiologic investigation. J Expo Sci Environ Epidemiol. 2006;16(2):191–205. pmid:16160703
- 32. Verret WJ, Chen Y, Ahmed A, Islam T, Parvez F, Kibriya MG, et al. A randomized, double-blind placebo-controlled trial evaluating the effects of vitamin E and selenium on arsenic-induced skin lesions in Bangladesh. J Occup Environ Med. 2005;47(10):1026–35. pmid:16217243
- 33. Reuter W, Davidowski L, Neubauer K. Speciation of Five ArsenicCompounds in Urine by HPLC/ICP-MS: PerkinElmer Life & Analytical Sciences; 2003 [cited 2021 3/22/21. ]. Available from: https://www.perkinelmer.com/PDFs/Downloads/app_speciationfivearseniccompounds.pdf
- 34. Gamble MV, Liu X, Ahsan H, Pilsner R, Ilievski V, Slavkovich V, et al. Folate, homocysteine, and arsenic metabolism in arsenic-exposed individuals in Bangladesh. Environ Health Perspect. 2005;113(12):1683–8. pmid:16330347
- 35. Peters BA, Hall MN, Liu X, Parvez F, Sanchez TR, van Geen A, et al. Folic Acid and Creatine as Therapeutic Approaches to Lower Blood Arsenic: A Randomized Controlled Trial. Environ Health Perspect. 2015;123(12):1294–301. pmid:25978852
- 36. Jansen RJ, Argos M, Tong L, Li J, Rakibuz-Zaman M, Islam MT, et al. Determinants and Consequences of Arsenic Metabolism Efficiency among 4,794 Individuals: Demographics, Lifestyle, Genetics, and Toxicity. Cancer Epidemiol Biomarkers Prev. 2016;25(2):381–90. pmid:26677206
- 37. Das S, Forer L, Schönherr S, Sidore C, Locke AE, Kwong A, et al. Next-generation genotype imputation service and methods. Nat Genet. 2016;48(10):1284–7. pmid:27571263
- 38. Burn JI, Taylor SF. Natural history of thyroid carcinoma. A study of 152 treated patients. Br Med J. 1962;2(5314):1218–23. pmid:14017111
- 39. Ahsan H, Chen Y, Parvez F, Zablotska L, Argos M, Hussain I, et al. Arsenic exposure from drinking water and risk of premalignant skin lesions in Bangladesh: baseline results from the Health Effects of Arsenic Longitudinal Study. Am J Epidemiol. 2006;163(12):1138–48. pmid:16624965
- 40. Kang HM, Sul JH, Service SK, Zaitlen NA, Kong S-Y, Freimer NB, et al. Variance component model to account for sample structure in genome-wide association studies. Nat Genet. 2010;42(4):348–54. pmid:20208533
- 41. GTEx Consortium. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020;369(6509):1318–30. pmid:32913098
- 42. Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingorani AD, Wallace C, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet. 2014;10(5):e1004383. pmid:24830394
- 43. Oakley AJ. Proposed mechanism for monomethylarsonate reductase activity of human omega-class glutathione transferase GSTO1-1. Biochem Biophys Res Commun. 2022;590:7–13. pmid:34959192
- 44. Zakharyan RA, Sampayo-Reyes A, Healy SM, Tsaprailis G, Board PG, Liebler DC, et al. Human monomethylarsonic acid (MMA(V)) reductase is a member of the glutathione-S-transferase superfamily. Chem Res Toxicol. 2001;14(8):1051–7. pmid:11511179
- 45. Tamayo LI, Kumarasinghe Y, Tong L, Balac O, Ahsan H, Gamble M, et al. Inherited genetic effects on arsenic metabolism: A comparison of effects on arsenic species measured in urine and in blood. Environ Epidemiol. 2022;6(6):e230. pmid:36530933
- 46. Shin S-Y, Fauman EB, Petersen A-K, Krumsiek J, Santos R, Huang J, et al. An atlas of genetic influences on human blood metabolites. Nat Genet. 2014;46(6):543–50. pmid:24816252
- 47. Emilsson V, Ilkov M, Lamb JR, Finkel N, Gudmundsson EF, Pitts R, et al. Co-regulatory networks of human serum proteins link genetics to disease. Science. 2018;361(6404):769–73. pmid:30072576
- 48. Ruth KS, Day FR, Tyrrell J, Thompson DJ, Wood AR, Mahajan A, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med. 2020;26(2):252–8. pmid:32042192
- 49. Treacy EP, Akerman BR, Chow LM, Youil R, Bibeau C, Lin J, et al. Mutations of the flavin-containing monooxygenase gene (FMO3) cause trimethylaminuria, a defect in detoxication. Hum Mol Genet. 1998;7(5):839–45. pmid:9536088
- 50. Choi HS, Bhat A, Howington MB, Schaller ML, Cox RL, Huang S, et al. FMO rewires metabolism to promote longevity through tryptophan and one carbon metabolism in C. elegans. Nat Commun. 2023;14(1):562. pmid:36732543
- 51. Chen J-W, Wang S-L, Wang Y-H, Sun C-W, Huang Y-L, Chen C-J, et al. Arsenic methylation, GSTO1 polymorphisms, and metabolic syndrome in an arseniasis endemic area of southwestern Taiwan. Chemosphere. 2012;88(4):432–8. pmid:22440634
- 52. Beebe-Dimmer JL, Iyer PT, Nriagu JO, Keele GR, Mehta S, Meliker JR, et al. Genetic variation in glutathione S-transferase omega-1, arsenic methyltransferase and methylene-tetrahydrofolate reductase, arsenic exposure and bladder cancer: a case-control study. Environ Health. 2012;11:43. pmid:22747749
- 53. Polimanti R, Piacentini S, De Angelis F, De Stefano GF, Fuciarelli M. Human GST loci as markers of evolutionary forces: GSTO1*E155del and GSTO1*E208K polymorphisms may be under natural selection induced by environmental arsenic. Dis Markers. 2011;31(4):231–9. pmid:22045430
- 54. Chernoff MB, Delgado D, Tong L, Chen L, Oliva M, Tamayo LI, et al. Sequencing-based fine-mapping and in silico functional characterization of the 10q24.32 arsenic metabolism efficiency locus across multiple arsenic-exposed populations. PLoS Genet. 2023;19(1):e1010588. pmid:36668670
- 55. Khairul I, Wang QQ, Jiang YH, Wang C, Naranmandura H. Metabolism, toxicity and anticancer activities of arsenic compounds. Oncotarget. 2017;8(14):23905–26. pmid:28108741