Arsenic and Environmental Health: State of the Science and Future Research Opportunities

Understanding Arsenic Speciation and Exposure Sources

A substantial amount of research has focused on exposure to arsenic via drinking water; however, more research is now being directed toward characterizing arsenic exposures from other sources. To develop a more complete understanding of arsenic exposure, more studies are needed to identify, quantify, and characterize arsenic in diet, soil, dust, and air. In addition, although much is known about some inorganic and organic forms of arsenic, data gaps in understanding exposures and toxicokinetics of other arsenic species (e.g., arsenoproteins, arsenolipids, and thiolated arsenic compounds) need to be addressed.

Understanding arsenic speciation. Arsenic exists in many different inorganic and organic forms, and in different oxidation or valence states. The valence states of arsenic compounds relevant to human health are the trivalent (As^III) and pentavalent (As^V) states. These arsenic species include arsenates (compounds containing AsO₄ ^3–), arsenites (compounds containing AsO₃ ^3–), and the monomethyl (MMA) and dimethyl (DMA) metabolites. Arsenic species in the trivalent state including arsenous acid (commonly arsenite), monomethylarsonous acid (MMA^III), and dimethylarsinous acid (DMA^III) are generally considered more toxic at lower doses than other arsenic species (ATSDR 2007; Drobna et al. 2009), although the complexity of arsenic species interconversion and the number of uncharacterized species casts uncertainty in adherance to this generalization (ATSDR 2007).

There are numerous other arsenic species, many of which we know little about. For example, fish and algae contain arsenobetaine (C₅H₁₁AsO₂) as well as arsenoproteins, arsenolipids, and arsenosugars (Feldmann and Krupp 2011; Schmeisser et al. 2006), many of which have not been characterized. Arsenobetaine is generally considered to be of low toxicity compared with some forms of inorganic arsenic (Leffers et al. 2013; Taylor et al. 2013). Some arsenosugars have been shown to be bioaccessible and metabolized in humans, and limited studies demonstrate toxicity in vitro (Feldmann and Krupp 2011; Leffers et al. 2013). Because seafood can contain up to 100 times more total arsenic than rice does, and contains a variety of poorly understood organoarsenical compounds, researchers are calling for more detailed studies of these forms of arsenic (Feldmann and Krupp 2011; Molin et al. 2015). Thiolated arsenosugars have been identified as being generated in the human gut and readily absorbed by gut epithelium (DC.Rubin et al. 2014); toxicity studies are suggestive of adverse health effects but studies are sparse (Ebert et al. 2014). In addition to toxicity differences, arsenic species can vary tremendously in terms of bioavailability, environmental fate, and transport characteristics, and remediation strategy effectiveness (Campbell and Nordstrom 2014; Gupta et al. 2012).

Future arsenic research needs to expand our understanding of the variety of arsenicals that exist in the environment, better characterize more arsenic species (e.g., currently uncharacterized arsenoproteins, arsenosugars, and arsenolipids), understand their toxicokinetics in vivo in humans and rodents, and evaluate their fate and transport in the environment. Furthermore, rather than measuring only total arsenic, researchers are moving toward more frequently measuring specific arsenic species in environmental and human samples to gain a more complete and detailed understanding of arsenic exposures and health risks. Researchers are exploring ways to overcome challenges in human sample collection, sample handling variability, and the short half-life of some arsenic species in solution (García-Salgado and Quijano 2014; NIEHS 2014).

Understanding arsenic exposure from diet. In some regions of the world including sites in the United States, arsenic exposures from drinking water are an urgent concern in the face of high concentrations of naturally occurring arsenic (Naujokas et al. 2013). Although exposure from drinking water remains a major concern, recent research reveals other sources of arsenic exposure. Identifying and characterizing these sources of arsenic exposure is very important for finding ways to minimize exposures and health risks.

One non–drinking-water source of arsenic exposure that is of increasing concern is the diet. Arsenic is present in a wide variety of foods including fish and rice (Jackson et al. 2012; Schoof et al. 1999; Tao and Bolger 1999; WHO 2011). Fish contain high amounts of organic arsenic compounds, predominantly arsenobetaine (Molin et al. 2015; Tao and Bolger 1999). In contrast, rice contains predominantly inorganic arsenic (Jackson et al. 2012). The relative contribution of the diet as a source of arsenic exposure may be substantial, particularly when drinking-water arsenic levels are low. For example, one study (Kurzius-Spencer et al. 2014) modeled dietary exposure data that was collected in three U.S. population studies: National Human Exposure Assessment Survey (NHEXAS-AZ) (Lebowitz et al. 1995), Binational Exposure Assessment Survey (BAsES, Arizona population only) (Roberge et al. 2012), and National Health and Nutrition Examination Survey (NHANES) (CDC 2015). The authors estimated that diet contributed 54–85% of total inorganic arsenic intake for individuals whose tap water contained < 10 μg/L arsenic (Kurzius-Spencer et al. 2014). A separate analysis of data from the NHEXAS-AZ and Arizona Border Survey population studies also suggests that dietary arsenic concentrations may be better predictors of urinary arsenic than drinking-water concentrations (Kurzius-Spencer et al. 2013).

It is challenging to quantify dietary exposure by measuring arsenic in foods. A duplicate diet study is one approach that uses direct measurements of duplicate samples of the foods that study participants consume over a period of time during the study (Thomas et al. 1997). These types of studies are the most accurate because they account for individual variability in food samples due to factors such as food growing conditions, preparation methods, and modifications through processing. Although duplicate diet studies have been informative for assessing dietary arsenic exposure and estimating health risks (Saipan and Ruangwises 2009), they are expensive to conduct.

An indirect measure of arsenic in foods relies on databases that contain arsenic concentration measurements for a wide variety of food types, and then using these measurements to estimate exposure based on mean values from that data. For example, arsenic exposure from rice would be estimated based on the amount consumed and the average amount of arsenic in rice reported in these databases. One such database is the ongoing U.S. FDA (Food and Drug Administration) Total Diet Study that measures about 800 contaminants and nutrients in foods present in the average U.S. diet (FDA 2015). There are a limited number of total diet studies that have measured arsenic in foods, and many of these studies measured only total arsenic concentrations (Chung et al. 2014; Schoof et al. 1999; Tao and Bolger 1999). Furthermore, arsenic concentrations in multiple samples of the same food are highly variable, and as discussed above, making generalizations of arsenic content in a specific food can be quite difficult (Lynch et al. 2014).

One study (Kurzius-Spencer et al. 2013) compared measured urinary arsenic concentrations with modeled dietary exposure estimates based on two sets of dietary data: a) the results of a duplicate diet study that measured total arsenic in duplicate diet samples, water, and urine over a 24-hr period for 252 people in the NHEXAS-AZ and Arizona Border Survey studies; and b) a total diet study using 24-hr diaries that estimated average arsenic content in food items from several published food surveys. The researchers found that the total diet study greatly underestimated dietary arsenic intake and that the duplicate diet study more accurately reflected the amount ingested to urinary biomarkers of exposure. More research is needed to unravel the complexities of dietary arsenic exposure assessment in order to better understand this exposure pathway.

Understanding arsenic exposure from dust, soil, and air. Arsenic exposure from dust, soil, and air should be better quantified and characterized, particularly near former mining sites, smelting sites, and industrial areas, including Superfund sites (Beamer et al. 2014; Menka et al. 2014; Taylor et al. 2014). A Superfund site is a location in the United States that has been contaminated with hazardous waste and identified as a priority site for cleanup by the U.S. EPA (Environmental Protection Agency) because it poses a significant risk to human health and/or the environment (U.S. EPA 2015b). The migration of arsenic from sediments and soils to groundwater sources and agricultural crops is not well understood and requires more research. For example, although a recent study in Cambodia reported that geochemical soil characteristics may be more predictive of arsenic content in rice crops than the concentrations of arsenic in water used for irrigation (Seyfferth et al. 2014), it has also been shown that high arsenic concentrations in irrigation water can increase arsenic concentrations in rice and reduce rice crop yields (Duxbury and Panaullah 2007). These types of exposure risks require better characterization, especially under special exposure scenarios such as populations who rely on rice for a large proportion of their diet and those who live near a Superfund site.

Bioavailability is another important factor to consider in allocating exposure to different sources, and ongoing research is focusing on development of cost-effective methods to measure bioavailability. For example, only a portion of the total arsenic in soil is bioavailable, or able to be absorbed, by living organisms (Juhasz et al. 2006). Arsenic bioavailability has been measured directly using expensive in vivo animal feeding studies called relative bioavailability assays (Rees et al. 2009). Recently, a less expensive in vivo assay has been developed using a mouse model (Bradham et al. 2013). Bioavailability has also been estimated using inexpensive in vitro bioaccessibility assays (IVBA) under conditions that mimic stomach and gastrointestinal environments, and several IVBA assays have demonstrated consistency in predicting bioavailability (Bradham et al. 2011; Brattin et al. 2013; Denys et al. 2012; Juhasz et al. 2015). One study performed extensive validity testing of 10 in vitro assays by comparing those results with swine in vivo assays using linear regression analysis, goodness of fit, variability in model bias and prediction error, and other parameters; validated studies had goodness-of-fit (R ²) values ranging from 0.59 to 0.71 (Juhasz et al. 2015). Although promising, these in vitro assays need to be tested further using a wider variety of sample types and larger numbers of samples. Although bioavailability testing has focused primarily on soils, more research is also warranted for bioavailability assessment of other exposure media such as dust and foods (Alava et al. 2015; Juhasz et al. 2006; Menka et al. 2014).

Modeling Aggregate Exposure

Given that arsenic is present in multiple media—food, water, soil, air, and dust—any individual is likely to have multiple routes and media of exposure. This scenario creates a substantial challenge for estimating exposure. Fate and transport, simulation, and probabilistic modeling are some approaches that can be used in conjunction with sampling measurement to estimate aggregate exposure (Dummer et al. 2015; Embry et al. 2014; Flanagan et al. 2015; Pastoor et al. 2014; U.S. EPA 2015a). These types of analyses, such as using soil sample concentrations to predict exposure and estimate health risks, are useful for risk assessment at specific sites (Gress et al. 2014). Also, some aggregate exposure modeling studies have used a multi-media, multi-pathway exposure assessment and identified house dust as an important source of exposure in mining communities (Hysong et al. 2003; O’Rourke et al. 1999).To develop a stronger foundation of data for future modeling studies, workshop participant indicated that duplicate diet studies, more sampling of food and other media, and more speciation data in all exposure media are needed to develop a stronger foundation of data for future modeling studies.

Reducing Soil and Dust Exposures

Soil and dust can be significant pathways of exposure, particularly near mining or smelting sites (Menka et al. 2014; Taylor et al. 2014). There are a number of different approaches to remediation of arsenic in soils and dust (Raj and Singh 2015; Singh et al. 2015; Wuana and Okieimen 2011), as well as daily-living exposure-reduction strategies such as hand and food washing (Defoe et al. 2014).

One example of a cost-effective and sustainable method to stabilize outdoor soils and dusts is phytostabilization. The goal of phytostabilization is to identify plants that could serve as permanent vegetative cover and, over time, may stabilize arsenic in the soil in a mineral form with low bioavailability. Stabilization may also reduce dispersal of contaminated dust. An ongoing study in Arizona is field testing several plants and optimizing growing conditions to maximize stabilization of arsenic-contaminated dust near a former smelting site (Valentín-Vargas et al. 2014). Recently, oxidized arsenic was co-localized with Actinobacteria on plant root surfaces using state-of-the-art microscopic visualization with resolution down to the 2-μm scale (Maier R, personal communication; NIEHS 2014). Actinobacteria are known to oxidize arsenic and to be resistant to metal toxicity (Banerjee et al. 2011), so their oxidizing capability combined with phytostabilization by the plants may provide powerful tools to reduce exposures from contaminated soils and dust; however, greater understanding of the relationships between these bacteria and the phytostabilizing plants is needed.

Identifying Susceptible Populations and Lifestages

There is ample evidence demonstrating that some individuals are more susceptible to arsenic than others. For example, exposure during early life is associated with increased risk of adverse effects that can persist into adulthood (Bailey et al. 2016; Smith et al. 2006; Steinmaus et al. 2014). One of the more striking examples is the nearly 50-fold increased standardized mortality ratio for bronchiectasis in a population of young adults in Chile who were exposed to high levels of arsenic from drinking water in utero and during childhood; mortality rates for this group were compared with mortality rates for the rest of the Chilean population (Smith et al. 2006). Genetic factors can also play a role in susceptibility, as demonstrated for AS3MT polymorphisms (Antonelli et al. 2014). As new biomarkers and factors of susceptibility are identified, as discussed above, researchers need to use that information to inform understanding of mechanisms of life stage and population susceptibility. Research is turning toward defining molecular mechanisms for these effects as well as biomarkers for susceptibility to disease in adulthood. Such diseases might be prevented or reduced through intervention in earlier life stages for susceptible individuals.

To better identify susceptible populations, susceptibility factors need to be investigated at the population level. For example, Engström et al. (2015) are analyzing AS3MT polymorphisms at the population level. In this study, several single nucleotide polymorphisms associated with lower urinary percent MMA occur at much higher frequency in an Argentinian Andes population living in a region where elevated arsenic levels in drinking water is common, compared with Peruvian and Colombian populations living in regions with lower arsenic level in drinking water (Schlebusch et al. 2015). These data suggest the possibility of a population adaptation to tolerate arsenic as an environmental stressor, and identify gene variants of the AS3MT gene associated with reduced risks. Characterization of these variants may inform strategies to minimize health effects from arsenic exposure. For example, protective variants might reveal proteins and pathways that have potential to prevent or reduce adverse arsenic-related health outcomes. Expanding on these types of studies, more large-scale genotyping of the AS3MT gene and other arsenic response-related genes is needed to better assess and quantitate population risks.

Clearly many factors—genetic and environmental—play a role in the response to arsenic exposure, and these factors can have various impacts in different life stages and different individuals. We are only beginning to understand how arsenic affects these processes across the life span. For example, epigenetic regulation plays an essential role in normal development whereby genes are turned on and off in sequence, and such changes are often heritable during cell division. Therefore, arsenic-associated epigenetic changes during early life may have long-term consequences. As research more clearly defines susceptibility factors, it is important to analyze those factors across populations and lifestages, and ultimately to use such information to quantitatively assess risk. The integration of molecular-level studies with in vivo animal and epidemiological studies is very important to delineate the importance of various susceptibility factors and identify new ones.

Nutrition and Health Risk Mitigation

Recent information about arsenic metabolic pathways and nutritional factors sheds light on the potential to use dietary changes to prevent or reduce arsenic-associated health effects. For example, researchers are focusing on the one-carbon metabolic pathway that is catalyzed by AS3MT and other enzymes, and converts arsenic into a variety of methylated species with varying toxicities (Hall and Gamble 2012). The goal is to find nutritional supplements or dietary changes that might prevent or mitigate arsenic toxicity. Researchers are now linking nutritional biochemistry studies with epidemiology to explore whether nutritional status and supplements may affect health outcomes stemming from arsenic exposure (Howe et al. 2014; Niedzwiecki et al. 2014).

Several epidemiology studies give credence to this possibility. In a Bangladeshi population, blood selenium levels were inversely associated with urinary arsenic concentrations (George et al. 2013). In another Bangladeshi population, blood levels of folate, which is used in the one-carbon metabolic pathway, were associated with arsenic methylation status in urine (Gamble et al. 2006; Howe et al. 2014). Other studies suggest that B₁₂, choline, homocysteine, betaine, and creatine levels may also be associated with changes in the arsenic metabolite profile in humans (Niedzwiecki et al. 2014) and in rats (Mukherjee et al. 2006). Researchers are also studying whether nutritional supplementation can prevent or ameliorate arsenic-related health outcomes. For example, the Bangladesh Vitamin E and Selenium Trial (BEST) is a population-based, double-blind, randomized controlled trial of 7,000 adults with skin lesions that is designed to test supplementation for the prevention of nonmelanoma skin cancer (Argos et al. 2013). More research is needed to better understand the balance of nutritional influences because some nutrients at certain doses may increase arsenic toxicity, as suggested for high levels of selenium (Sun et al. 2014).

Diet may negatively affect arsenic-associated effects on fat metabolism and liver function. Previous work has shown that arsenic exposure is associated with high cholesterol, liver inflammation, and liver steatosis (Sanchez-Soria et al. 2014; Shi et al. 2014). Co-exposure to a high-fat Western diet and arsenic in mice exacerbated the effects of a high-fat diet on the liver (e.g., increased size and steatosis), resulting in degeneration that was more severe and widespread than in the controls without arsenic exposure (Sanchez-Soria et al. 2014). Mice exposed to arsenic in utero were affected more strongly than mice exposed later in life. Arsenic also altered the lipid metabolic product profile in a pattern that suggests disruption in the tricarboxylic acid (TCA) cycle and increased ketogenesis (Ditzel et al. 2016). Together the data suggest that the metabolism of arsenic and fatty acids are intertwined, and can impact health outcomes.

The one-carbon and TCA metabolic pathways are parts of a complex interwoven biochemical network with other biochemical pathways, and the complexity makes it difficult to predict how specific nutrients may change arsenic metabolism. Furthermore, many nutrients may affect arsenic metabolism and toxicity. A new mathematical model of arsenic metabolism has been developed to help make such predictions, and the model performed well in comparisons with epidemiological data (Lawley et al. 2014). This model can be used to predict nutritional influences on arsenic metabolism before conducting in vivo testing.

More data are needed to define nutritional changes and supplements that may prevent or minimize health effects. Nutritional status should be considered in epidemiological studies as a possible contributing factor to the outcomes of arsenic exposure, and collecting more data on nutrients and arsenic species in biological samples could enhance our understanding of these relationships. Because of the intricate interweaving of these biochemical pathways, it is difficult to predict how supplementation will affect the entire system. Nonetheless, nutritional supplementation holds potential as a cost-effective and practical approach to reducing health impacts of arsenic exposure.

Mechanisms of Response and Susceptibility Research Needs

• Using omics technologies to identify biomarkers linking exposure to susceptibility, disease onset, and long-term disease
• Conducting molecular epidemiology studies
• Identifying susceptible populations (e.g., life stage, genetic factors)
• Finding approaches for quantitating risks for use in risk assessments
• Exploring influences of nutrition on arsenic susceptibility and health risk reduction