Soil arsenic in vivo and in vitro oral bioavailability methods for use in human health risk assessment, a review of the state of the science

Abstract

Arsenic (As) is the most frequently occurring contaminant on the Priority List of Hazardous Substances, which lists substances of greatest public health concern to people living at or near U.S. National Priorities List sites. Human exposure to arsenic via ingestion of arsenic-contaminated soils can have serious health impacts including increased cancer risk. Accurate assessment of human health risks from exposure to arsenic-contaminated soils depends on estimating its bioavailability, defined as the fraction of ingested arsenic absorbed across the gastrointestinal barrier and available for systemic distribution and metabolism. Arsenic bioavailability varies among soils and is influenced by site-specific soil physical and chemical characteristics and internal biological factors. This review describes the state-of-the science that supports our understanding of oral bioavailability of soil arsenic, the methods that are currently being explored for estimating soil arsenic relative bioavailability (RBA), and future research areas that could improve our prediction of the oral RBA of soil arsenic in humans. The following topics are addressed: (1) arsenic soil geochemistry; (2) arsenic toxicology; (3) in vivo models for estimating arsenic RBA; (4) in vitro bioaccessibility methods; and (5) conclusions and research needs.

Introduction

Human exposure to arsenic (As) via ingestion of arsenic-contaminated drinking water can have serious health impacts including increased cancer risk and other adverse health effects (ATSDR 2007; IARC 2012; U.S. EPA 2016; WHO 2012). In addition to exposure from drinking water, people are exposed to arsenic in surface soils as a result of natural geological processes as well as anthropogenic sources (ATSDR 2007). The latter include production waste, application of arsenic-containing pesticides, and wastes from mining, milling, and smelting of arsenic-containing ores. Cancer risk for ingestion of arsenic-contaminated soils is estimated by applying a cancer slope factor, derived from epidemiological studies in populations chronically exposed to arsenic in drinking water, to estimated soil arsenic ingestion (Argos et al. 2014; Chung et al. 2013; Ferreccio et al. 2013; Saint-Jacques et al. 2014; U.S. EPA 2016; Yang et al. 2005). However, because oral bioavailability of arsenic in soil is less than that from water, cancer risk, as well as risks of other adverse health effects from ingestion of arsenic-contaminated soil may be overestimated (Bradham and Wentsel 2010; NRC 2003; U.S. EPA 2007a, 2007b). A recent compilation of data from a variety of arsenic-contaminated sites demonstrated that oral bioavailability of arsenic in soil tends to be lower than that of water soluble sodium arsenate (U.S. EPA 2012). Expressed as oral relative bioavailability (RBA, or the percent ratio of bioavailability of arsenic in soil to that of dissolved sodium arsenate), the median and 95^th percentile RBA values for more than 100 soil samples were 30 and 60%, respectively (U.S. EPA 2012).

Many factors affect the oral bioavailability of arsenic in soil, including chemical form of arsenic, as well as physical and chemical characteristics of arsenic-bearing soil particles; these factors are known to vary both within and between sites. Accurate assessment of cancer risk and other adverse health effects from exposure to arsenic-contaminated soils at any given location, therefore, depends on estimating arsenic oral RBA in soil at that location. This has stimulated research into methods for predicting oral RBA of arsenic in humans from experimental models. The research described in this review includes soil geochemistry of arsenic and its determinants of bioavailability, development of a variety of animal bioassays intended to predict soil arsenic RBA in humans, and development of in vitro approaches to predicting RBA from measurements of bioaccessibility. In vitro bioaccessibility (IVBA) assays reduce the time and expense of RBA assessments, as well as the reliance upon animal studies.

At the interface between research and applications to risk assessment is an important problem that has yet to be definitely addressed: Which methods most accurately predict oral RBA of soil arsenic in humans? No animal models or IVBA assays have been evaluated directly for predicting RBA in humans, nor are such evaluations in humans feasible or likely to be undertaken in the future. Therefore, our confidence in estimating RBA in humans will continue to rely on our understanding of the science underlying the methods currently available. In particular, several problems must be addressed, which are the focus of this review: (1) Do animal models yield similar or diverging estimates of RBA when applied to the same or similar soils and arsenic minerology? (2) How much of the differences in outcomes of animal models can be attributed to interspecies differences, rather than differences in the design of the bioassays? (3) How well do IVBA assays predict the outcomes of animal bioassays? (4) What physical-chemical or physiological variables are most likely to contribute to interspecies differences in RBA relevant to extrapolation to humans?

This review describes the state-of-the science that supports our understanding of oral bioavailability of soil arsenic, the methods that are currently being explored for estimating soil arsenic RBA, and future research areas that could improve our prediction of the oral RBA of soil arsenic in humans. The following topics are addressed: (1) arsenic soil geochemistry; (2) arsenic toxicology; (3) in vivo models for estimating arsenic RBA; (4) in vitro bioaccessibility methods; and (5) conclusions and research needs.

Bioavailability-related Terms Used in this Review.

The term bioavailability has many different uses and has been assigned different meanings to accommodate specific contexts (e.g., pharmacology, ecological risk assessment, human health risk assessment). Absolute bioavailability (ABA) refers to the fraction of an ingested dose of arsenic that crosses the gastrointestinal epithelium and becomes available for internal distribution. Relative bioavailability (RBA) refers to the ratio of the absolute bioavailabilities, which in this review, is the ratio of the ABA for soil arsenic to that of a completely water soluble reference form of arsenic (typically sodium arsenate). ABA and RBA are measured in animal models; however, RBA can also be predicted from in vitro bioaccessibility assays (IVBAs). The term bioaccessibility refers to the fraction of arsenic that becomes soluble in an assay designed to predict biological absorption from the gastrointestinal tract (GIT). All of these terms are described in further detail in the pertinent sections of this review.

Overview of arsenic soil geochemistry

Arsenic (atomic number 33; atomic mass 74.9216) is the third member of Group VA of the periodic table, possessing an electronic configuration of 4s²4p³. The descending electronegativity of elements in Group VA yields non-metallic character, for which arsenic is often described as a metalloid. Ubiquitous global distribution of arsenic ranks 20^th in abundance in the earth’s crust (at 0.00005 w/w%), 14^th in seawater, and 12^th within the human body (Woolson 1975). Arsenic has only one natural isotope, ⁷⁵As; however, at least 33 radioisotopes have been synthesized. Humans have been aware of arsenic for over 3000 years, and it was initially thought to be a form of sulfur. Theophrastus (371–288 BC) called the lustrous red mineral realgar, arsenikon, meaning ‘potent’. Elemental arsenic was first separated from its sulfide form around 1250 AD by the alchemist Albertus Magnus (Boyle and Jonasson 1973).

Arsenic has four possible oxidation states: arsenide (As⁻³), elemental (As⁰), arsenite (As⁺³), and arsenate (As⁵⁺). Arsenite and arsenate valence states are the most common environmentally relevant forms in soils, sediments, and water over a wide range of pH and Eh conditions. Thermodynamic predictions of arsenic in soil pore water reveal that arsenate species (HAsO₄²⁻ > H₂AsO₄⁻ at pH 7) are more abundant in conditions that are oxidized more than pe+pH>9. Arsenite is expected to form (HAsO₂⁰ = H₃AsO₃⁰ > AsO₂⁻ = H₂AsO₃⁻ at pH 7) in relatively anoxic soil solutions with pe+pH<7 (Sadiq, 1997). Since arsenic is redox sensitive, seasonal variations affect concentration and speciation of arsenic in water and soil. For example, increased temperatures during dry periods can drive the oxidation of arsenite phases and minerals, which may generate a large pulse of arsenate release upon return to normal water conditions. In aerobic soil environments, the geochemical behavior of arsenate is similar to phosphate; however, under like conditions, arsenic tends to be more mobile than phosphorus.

Arsenic has had many commercial uses such as an additive for hardening of alloys, production of semiconductors, pigments, glass manufacturing, insecticides/pesticides, herbicides, desiccants/wood preservatives, and animal feed additives. Many of these products were directly applied to the environment, and, in some cases, end of life waste disposal also contributed to environmental contamination. Mining, milling, and smelting have also contaminated soil. Principal factors influencing the concentration of elements in soils include geogenic release and anthropogenic activities. Factors such as climate, organic and inorganic soil components, and redox potential also affect the level of arsenic in soils. Microorganisms and some abiotic forces are essential factors in the environmental fate, transformation, and toxicity of arsenic; however, anthropogenic activities have amplified arsenic environmental contamination (Smith et al. 1998).

Geological sources

There are over 245 naturally occurring minerals that contain arsenic, comprising mostly ores containing sulfide, along with copper, nickel, lead, cobalt, or other metals (Mandal and Suzuki 2002). Arsenic content of rocks depends on the rock type, with sedimentary rocks containing much higher concentrations than igneous rocks. Although discernible chemical and physical differences are evident between rock groups, the range of arsenic concentrations within a rock type may vary significantly. The mean arsenic concentrations in igneous rocks range from 1.5 to 3.0 mg As kg⁻¹, whereas the mean arsenic concentrations in sedimentary rocks range from 1.7 to 400 mg As kg⁻¹ (Mandal and Suzuki 2002). Arsenic is a common constituent of many types of mineral deposits, particularly those containing sulfides. Arsenic commonly coincides with copper, silver, gold, zinc, cadmium, mercury, uranium, tin, lead, phosphorus, antimony, bismuth, sulfur, selenium, tellurium, molybdenum, tungsten, iron, nickel, cobalt, and platinum metals in ore deposits. Under most conditions arsenic is a suitable indicator of deposits of these elements, being particularly useful in geological surveys (Boyle and Jonasson 1973). Ore sources of arsenic include iron pyrite, galena, chalcopyrite, and sphalerite, but the most common is arsenopyrite.

Background concentrations of arsenic in soil can range from 1 to 50 mg kg⁻¹, with an average around 5 mg kg⁻¹ (Lindsay 1979). Arsenic distribution in soil is often geologically influenced depending on concentrations within the parent material from which the soil is formed. Typically, sandy soils derived from granite have lower arsenic levels than alluvial and organic soils (Kabata-Pendias 1984).

The contribution of arsenic atmospheric deposition is estimated to be significant to the geochemical cycle of arsenic. Estimated global atmospheric flux is between 63,000 and 74,000 metric tons year⁻¹ (Chilvers and Peterson 1987; Nriagu and Pacyna 1988). These studies suggest that the split between geogenic and anthropogenic inputs of arsenic in the atmosphere is 60–40 or 70–30; predicting geogenic sources are more influential. The geogenic sources are believed to comprise low-temperature volatilization due to microbial activity and volcanic emissions (Chilvers and Peterson 1987).

Anthropogenic sources

The largest sources of arsenic into the environment include commercial agricultural products such as pesticides/insecticides, herbicides, desiccants/wood preservatives, feed additives, and as an impurity in phosphate fertilizer. While agricultural usage of arsenic-based pesticides has declined significantly, it was not uncommon in the mid-20^th century for the United States to consume nearly 20,000 tons of pesticides in the form of lead arsenate, calcium arsenate, copper acetoarsenite (Paris-Green), arsenic acid (H₃AsO₄), monosodium methanearsonate (MSMA), disodium methanearsonate (DSMA), and dimethylarsinic acid (DMA, cacodylic acid, Agent Blue) (Lansche 1965; Thompson 1973; USDA 1970). Lead arsenate was a popular insecticide during the first half of the 20th century because of its low toxicity to plants and great effectiveness for controlling codling moth in commercial apple orchards. Smaller but still substantial amounts were used on deciduous tree fruits other than apple, in home gardens and orchards, for mosquito control, and on lawns and golf greens (Peryea (1999). As an herbicide, sodium arsenite was widely used since the late 1800s as a non-selective weed killer (USDA 1970). Sodium and calcium versions of monomethylarsenate (MMA) and DMA are widely employed herbicides and are the active ingredient in consumer products including Weed-B-Gone Crabgrass Killer. Arsenic acid (Desiccant L-10) was extensively used as a cotton desiccant for many years to defoliate cotton to allow seeding of the next cotton crop (Rosen and Liu 2009). Arsenic has been utilized as a wood preservative since the early 1900s. Fluor-Chrome-Arsenic-Phenol (FCAP) was the first wood preservative and, later, Chromated Copper Arsenate (CCA) and Ammonical Copper Arsenate (ACA) dominated the industry (USDA 1974). As a feed additive, the most recognized organic arsenical is Roxarsone (4-hydroxy-3-nitrophenylarsonic acid), which was used as a growth enhancer in chicken production. Poultry manure containing Roxarsone was often applied to land as fertilizer, thus adding arsenic to agricultural soils.

Although often resulting in very high concentrations of arsenic and other contaminants in soils, mining-related activities tend to be quite localized in comparison to agricultural distribution issues of arsenic. However, mining districts are often more densely populated than agricultural districts. Types of arsenic ores utilized in mining activities are discussed above under Geological Sources. After ore processing and smelting activities, slag and chat materials still contained high levels of arsenic, often as mineral forms of arsenite sulfides and oxidation products. Oxidation of these sulfide minerals causes arsenic to be released in high concentrations leading to surface water, groundwater, and soil contamination, as well as food chain transfer of arsenic (Smedley and Kinniburgh 2002, 2005).

Geochemical Properties

The geochemistry of arsenic is complicated by its various oxidation states and speciation possible in soil environments. The chemical and physical properties of arsenic in soil are governed by sorption-desorption processes related to inorganic and organic soil constituent properties, soil pH and redox attributes, competing ions, and other effects such as ligands and ionic strength (Smith et al. 1998). Under typical soil conditions, arsenate and arsenite species are possible, with arsenate forms predominantly in aerobic soil environments. Both inorganic and organic compounds of arsenate and arsenite can be identified in soils (Mandal and Suzuki 2002), but inorganic species are more common. Soil arsenic resulting from weathering of parent material generally results in arsenate adsorption complexes with inorganic and organic soil materials.

However, contamination due to agricultural application or mining activities can result in a variety of arsenic forms. For example, arsenopyrite (FeAsS) oxidation can yield scorodite (FeAsO₄), iron oxides with arsenate in the oxide structure, and arsenate adsorbed to iron oxides within the Eh-pH regime of soils (Corkhill and Vaughan 2009). Soil aging is an important factor controlling arsenic mobility and bioavailability in soils (Juhasz et al. 2008; Lombi et al. 1999) resulting in reduced arsenic availability as time increases. However, most desorption studies of arsenic show a hysteric release, meaning desorption of arsenic from a surface is much slower than its adsorption kinetics (Bauer and Blodau 2006; Lin and Puls 2000; Smedley and Kinniburgh 2002; Zhang et al. 2008).

A tremendous amount of research has shown the importance of iron, aluminum, and manganese oxides and oxyhydroxides on the retention of arsenic in soils (Moreno-Jiménez et al. 2012). These interactions are both pH and redox dependent. Arsenic adsorption on metal oxides and oxyhydroxides is greatest at a pH range of 4 to 5 and declines as pH increases. Arsenate can form both inner- and outer-sphere sorption complexes on oxide and oxyhydroxide surfaces, where inner-sphere complexes are bound more strongly to the surface. However, arsenate complexes on metal oxides and oxyhydroxides can be easily released under anaerobic conditions where arsenate, iron and manganese oxides, and oxyhydroxides undergo reduction (Mello et al. 2006).

In alkaline soils (pH>8) with suitably high levels of arsenic, calcium, and/or iron, arsenates can precipitate and arsenic co-precipitation with iron sulfoxides can occur (Moreno-Jiménez et al. 2012; Zhang et al. 2008). In between the favorable retention of arsenic by iron, aluminum, and manganese oxides and oxyhydroxides at low pH and the sequestration of arsenic as precipitates at higher pH is a pH range (5–8) where clay minerals (i.e., kaolinite, illite, montmorillonite) and organic matter contribute to arsenic retention. Metal oxides are more efficient scavengers of arsenic than clay minerals, and sandy soils retain less arsenic than clayey soils (Adriano 1986; Wenzel et al. 2001). As with metal oxides and oxyhydroxides, clay minerals will adsorb and retain arsenate at a higher surface coverage than arsenite (Lin and Puls 2000). Evidence for the ability of soil organic matter to retain arsenic in soil is inconsistent as some studies show reasonable retention while others show enhanced mobility in the presence of organic ligands (Adriano 1986). These results are impacted by soil pH and the reactivity of surface functional groups. For example, fulvic and humic acid interaction with the surface of an iron oxyhydroxide can inhibit arsenate sorption (Weng et al. 2009), demonstrating complex geochemistry of arsenic and the need to understand the chemical and physical properties of soils.

Given the complexity of arsenic geochemistry, it is important to characterize arsenic-contaminated soil to understand risks and identify changes in arsenic speciation (Porter et al. 2004; Scheckel et al. 2009). A number of advanced spectroscopic analytical techniques are available to decipher the speciation of arsenic in soil environments (Lombi et al. 2011). Spectroscopic investigations coupled with basic soil characterization methods can provide valuable information to make informed decisions at impacted sites.

Elemental interactions

Previous studies have evaluated the ability of soil geochemical properties to describe arsenic bioavailability and bioaccessibility through linear or multivariate regression techniques. These predictive models have been proposed for use to provide preliminary screening estimates as well as determination of soil geochemical mechanisms that control bioavailability. In a study of 36 soils spiked with arsenic (As^v) and aged over 6 months, Yang et al. (2002) reported that three variables (log iron oxide, pH, and total inorganic carbon) to significantly influence steady-state bioaccessibility. Performance of this same model was also evaluated to predict bioaccessibility of 36 arsenic (As^III)-spiked soils, producing estimates within a root-mean square error (BMSE) of 9.5% (Yang 2005). Prediction of in vivo bioavailability in swine and Cebus monkey assays was also evaluated, with the model predicting bioavailability in swine within a RMSE of 9.5% (n = 9 soils) and 15.5% (n = 12 soils) for two sets of soils (Yang 2002, 2005). Prediction of bioavailability values (n = 5 soils) derived from the monkey assay, however, was poor, with the model consistently overpredicting bioavailability, resulting in a RMSE of 42.7% (Yang 2005). Building upon these findings, Roberts et al. (2007) evaluated arsenic mass distribution across 18 mineralogical phases in 14 soils and reported that iron sulfate was the best single linear predictor of arsenic RBA in the monkey assay, accounting for 88% of the observed variability in RBA for 8 soils that contained arsenic present in iron sulfate. The relationship, however, only accounted for 38% of observed variability when expanded to all 14 soils.

Juhasz et al. (2007a) reported that, of the five possible predictor variables evaluated (pH, total arsenic, iron, aluminum, and phosphorus), total soil arsenic together with iron concentrations were the variables that best described the variation in in vivo arsenic bioavailability in a swine assay. A regression model that included total arsenic and iron as predictor variables was able to describe 72% of the observed variability in arsenic RBA when RBA was expressed as the bioavailable concentration (μg As/g soil). Most of the explanatory power of the model derived from total arsenic concentration, for which we expect a strong correlation with the bioavailable concentration (e.g., for a constant RBA%, the bioavailable concentration will correlate with total arsenic concentration). When correlating against in vitro bioaccessibility, Juhasz et al. (2007b) reported that total soil arsenic and iron was able to describe 96% of the variability in arsenic bioaccessibility across 50 soils when bioaccessibility was expressed as the bioaccessible concentration (μg As/g soil). When soils were separated by contamination source (railway corridors, dip sites, mine sites, and gossans), model parameterization varied somewhat, with models that included total arsenic together with iron being the best predictors of arsenic bioaccessibility (μg/g) for railway corridors (herbicide impacted) and mine sites, while total arsenic and free iron were the best predictors for dip sites (pesticide impacted), and no model was achievable for gossan soils. Results from the above studies, which used the bioaccessible or bioavailable concentration in regression models, are consistent with studies that have shown significant correlations between iron content and RBA%. Bradham et al. (2011) reported that total soil iron and aluminum were significantly correlated with both arsenic RBA% using a mouse assay and bioaccessibility (%) using the SBRC method, with soil iron accounting for 48 and 32% of observed variation in arsenic bioavailability and bioaccessibility, respectively, while soil aluminum accounted for 34 and 32% of observed variation, respectively.

Studies have shown that total soil arsenic concentration and bioaccessibility values are influenced by soil particle size, albeit with inconsistent results. Ljung (2006) observed a positive correlation between sand content (higher % larger particles) and arsenic content of the finest soil fraction (<50 μm vs. 50–100 μm and <4 mm fraction). Their findings suggested that total soil arsenic will generally be lower in sandy soils due to fewer small particles, but that enrichment of arsenic to finer soil particles will be higher in sandy soils, where the number of preferred binding sites is limited and the arsenic ions will therefore be concentrated on the smallest particles. This observation is likely dependent on metal content and the number of binding sites for arsenic (or other metals). Girouard and Zagury (2009) evaluated total arsenic content and bioaccessibility in three CCA-impacted soils and found them to be differentially influenced by soil particle size (<250 μm vs. <90 μm). In the smaller size fraction (<90 μm), total arsenic concentration increased while % bioaccessibility decreased for a sandy soil; the opposite trend was observed in a loamy soil. In the third soil, there was no observed difference in total arsenic between the two size fractions, while bioaccessibility increased in the <90 μm fraction. Smith et al. (2009) evaluated arsenic distribution and bioaccessibility across 29 soils and observed that mean IVBA increased from 25 to 42% when evaluating the smaller 10 μm size fraction relative to the 250 μm size fraction, despite constant total soil arsenic concentrations.

Overview of toxicity

Inorganic arsenic and its methylated products (collectively referred to as arsenicals) produce a wide spectrum of toxic effects (cancer and/or non-cancer) in a variety of organ systems, including the cardiovascular system, eyes, gastrointestinal system, kidney, liver, nervous system, pancreas and skin (ATSDR 2007, 2016; Fowler et al. 2015; Naujokas et al. 2013; States et al. 2016). Acute exposure to inorganic arsenic (1 – 3 mg/kg) can be lethal, and recent concerns about adverse consequences of low-level chronic exposure to inorganic arsenic in drinking water include cardio-pulmonary disease, gastrointestinal disorders, diabetes, ocular effects, disturbances in the immune system, impairment of neurological function, developmental effects, and cancer, including bladder and kidney cancers following in utero exposure (ATSDR 2016; Fowler et al. 2015). The broad range of targets of arsenic toxicity is related to its mechanism of cellular toxicity, which is generic to all tissues, as well as the wide distribution of arsenic in the body following absorption. Cellular toxicity of arsenic is thought to be primarily related to its activity as a thiol reagent, which enables arsenicals to interact with a variety of thiol-dependent enzymes including pyruvate dehydrogenase, which is essential for mitochondrial respiration (ATSDR 2016; Fowler et al. 2015; Shen et al. 2013). Inhibition of mitochondrial respiration produces a cascade of effects at the cellular level, including production of reactive oxygen species (ROS), induction of stress proteins, and disruption of various cell signaling pathways (Bustaffa et al. 2014; Lee et al. 2005a, b; Madden et al. 2002; Rossi et al. 2002; Sen et al. 2005; Sherwood et al. 2011; Tsou et al. 2005; Wu et al. 1999, 2002). Arsenicals may also disrupt function of other thiol-dependent systems, including other dehydrogenases, ATPases, and zinc finger proteins (Bergquist et al. 2009; Zhou et al. 2011). Arsenicals can produce heritable toxicity or susceptibility through epigenetic mechanisms that include disruption of regulation of DNA methylation, histones, and micro-RNAs (Bailey and Fry 2014; Bhattacharjee et al. 2013; Bustaffa et al. 2014; Rossman 2003; Salnikow and Zhitkovic 2008).

Gastrointestinal absorption of arsenic in soil tends to be lower than that of water soluble arsenic (see further discussion below). Absorbed arsenic is metabolized and widely distributed in the body, including placenta and fetus (ATSDR 2007). Metabolism includes various oxidation and reduction pathways in mammalian cells and in gastrointestinal flora that participate in the interconversion of arsenicals between the +5 and +3 valence states in the formation of arsenite (As^IIIO₃) from arsenate (As^VO₄), methylarsonic acid (MMA^V), methylarsonous acid (MMA^III), and dimethylarsinic acid (DMA^V), and dimethylarsinous acid (DMA^III). Arsenicals can also form conjugates with glutathione (Hayakawa et al. 2005; Kala et al. 2000). Inorganic arsenic, MMA, and DMA are thought to participate in arsenical toxicity to varying degrees. All three arsenicals can inhibit mitochondrial respiration and stimulate ROS formation (Bergquist et al. 2009). Genetic polymorphisms in people contribute to variation in methylation and toxicity of arsenicals. These include polymorphisms for arsenite methyltransferase (Agusa et al. 2009; Engstrom et al. 2009, 2011; Porter et al. 2010; Rodrigues et al. 2012; Tellez-Plaza et al. 2013), cystathione-β-synthase (Porter et al. 2010), glutathione S-transferase π1 (Agusa et al. 2012; Antonelli et al. 2014; Marcos et al. 2006), glutathione S-transferase ω1 (Ahsan et al. 2007; Antonelli et al. 2014; Marcos et al. 2006; Porter et al. 2010; Rodrigues et al. 2012), methylenetetrahydrofolate reductase (Ahsan et al. 2007; Chung et al. 2010; Porter et al. 2010), and N-6 adenine-specific DNA methyltransferase 1 (Harari et al. 2013).

Absorbed arsenic is excreted in urine and feces, with urine being the dominant excretory pathway (Apostoli et al. 1999; Buchet et al. 1981a; Crecelius 1977; Tam et al. 1979). Urinary arsenic consists of a mixture of inorganic arsenic, MMA, and DMA (Apostoli et al. 1999; Buchet et al. 1981a). Glutathione conjugates of arsenic have also been detected in urine (Kala et al. 2004; Thomas, 2009 Studies conducted in rodents have observed excretion of arsenic-glutathione conjugates in bile following intravenous administration of inorganic arsenic (As^III and As^V) (Csanky and Gregus 2002; Kala et al. 2000). Arsenic can also be transferred to breast milk (Concha et al. 1998; Grandjean et al. 1995; Somogyi and Beck 1993). Whole body elimination half-times vary with animal species, which is influenced by methylation patterns and species differences in binding of arsenic in tissues (Drobna et al. 2009). For example, whole body elimination in rats is relatively slow (half-time 1–2 months) compared to mice (<1 day); this difference is thought to be related to retention of arsenic in red blood cells in rats (Vahter et al. 1984). Excretion of arsenic in marmoset monkeys is slower than in mice, which is thought to relate to the absence of arsenic methylation in this species (Vahter 1981; Vahter et al. 1985). The elimination half-time in humans has been estimated to be 2–3 days (Buchet et al. 1981b).

Arsenic biological interactions (ADME associated with oral exposures)

Absorption mechanisms

Evidence suggests that the oral bioavailability of inorganic arsenic in aqueous solution is high. In volunteers, about 60% of an oral dose of arsenate was excreted in urine within 5 days of ingestion (Pomroy et al. 1980; Tam et al. 1980). In monkeys and dogs, >75% of an ingested dose of inorganic arsenic was excreted in urine within a few days of administration (Charbonneau et al. 1978; Hollins et al. 1979; Tam et al. 1978). The bioavailability of arsenic in foods is highly dependent on the chemical form of arsenic present in the food. Using a swine model to evaluate the bioavailability of arsenic, Juhasz et al. (2006) found that inorganic arsenic present in rice was highly bioavailable (89%) but dimethylarsenic present in rice was less bioavailable (33%). Inorganic arsenic present in soils can be an important source of exposure for children who ingest soil. The bioavailability of arsenic in soil is dependent on the physical and chemical characteristics of the soil. In juvenile swine and mouse models that evaluated the bioavailability of arsenic in soils, relative bioavailabilities ranging from 2 to 80% have been reported (Diamond et al. 2016; U.S. EPA 2012).

The transport of inorganic arsenic across the gastrointestinal barrier is the initial step in the processes that lead to the systemic distribution, metabolism, and excretion of inorganic arsenic and its methylated metabolites. The initial uptake of inorganic arsenic across cell membranes is an example of molecular mimicry, a process by which a toxic metal or metalloid is transported by a pathway that normally mediates transport of an essential nutrient (Ballatori 2002). Arsenate in the +5 oxidation state is structurally similar to phosphate with similar shapes, ionic radii, and acid strengths (Egdal et al. 2009). For membrane transport, these similarities make it difficult to discriminate between arsenate and phosphate. The transport of arsenite, arsenic in the +3 oxidation state, depends upon its structural similarity to non-polar solutes that are transported across cell membranes by aquaglyceroporins (Mukhopadhyay et al. 2014).

Initial understanding of the relation between arsenate and phosphate transport across the gastrointestinal barrier came from studies in the ligated small intestine of the chick (Fullmer and Wasserman 1985; Wasserman and Taylor 1973), the mucosal membrane of rat small intestine (Danisi and Straub 1980), brush-border membrane vesicles (Schroder and Breves 1996), and isolated perfused rat small intestine (Gonzalez et al. 1995, 1997). Taken together, these studies showed that sodium-dependent transport of phosphate into cells of the gastrointestinal barrier was antagonized by the presence of arsenate. This antagonistic interaction suggested competition between these oxyanions for a membrane transporter. Subsequent work has identified three families of transporters that mediate the cellular uptake of phosphate (Biber et al. 2013). Transport of phosphate by these transporters takes place at the apical surface of cells and requires co-transport of sodium cations. In the GIT, the sodium-dependent transporter of phosphate (Npt2b) accounts for a large fraction of transcellular phosphate transport, although there is a paracellular pathway that accounts for uptake of phosphate across the gastrointestinal barrier (Sabbagh et al. 2011). The number and activity of Npt2b transporters is regulated by the level of phosphate in diet, vitamin D, and by growth factors. Coordination of phosphate absorption in the GIT and in kidneys works to maintain homeostasis for this essential nutrient. The uptake of phosphate by Npt2b transporters is antagonized by arsenate, consistent with its action as a potent competitive inhibitor (Villa-Bellsota and Sorribas 2008, 2010). Thus, transport of arsenate mediated by Npt2b transporters likely accounts for a substantial portion of arsenate that is transferred across the gastrointestinal barrier.

Although arsenate is probably the most common form of arsenic in oxidizing environments, under hypoxic conditions, arsenite (arsenic in the +3 oxidation state) exists as arsenous acid (As^III[OH]₃), which is an uncharged species at physiological pH (Yang et al. 2012). In this form, arsenite is a substrate for transport by aquaporins (Liu 2010). Aquaporins are a family of transmembrane channel proteins that include orthodox aquaroporins, which mediate water transport, and aquaglyceroporins, which transport non-polar solutes and metalloids (Mukhopadhyay et al. 2014). These transporters are expressed throughout the GIT (Koyama et al. 1999). Their ubiquitous expression in many tissues suggests that these channels could mediate both influx and efflux processes. Expression of aquaglyeroporin 9 in mouse hepatocytes correlates with arsenic accumulation and toxicity in these cells (Shinkai et al. 2009). Aquaglyeroporin 9-null mice, which do not express this protein, show diminished clearance of arsenic following treatment with sodium arsenite, consistent with a role of aquaporins in arsenic efflux; this defect in clearance of arsenic increases the lethality of arsenic in these mice (Carbrey et al. 2009).

In vivo models measuring of arsenic bioavailability

Monkey Urinary Excretion Fraction (UEF) Assay (Roberts et al. 2002, 2007)

The monkey UEF assay estimates soil arsenic RBA by comparing the urinary excretion of arsenic following an oral dose of soil or sodium arsenate (Roberts et al. 2002, 2007). Details of the assay protocol are presented in Table 1. Individual adult monkeys received a single oral dose of soil and sodium arsenate, with a minimum 3-week interval between dosing. Re-use of the animals is possible because of the relatively rapid elimination of absorbed arsenic in monkeys. Mean residence time (MRT) for arsenic in plasma was 1.2 hours (mean of 4 animals), and terminal elimination half-time was 0.99 hours, following an intravenous dose of sodium arsenate (Roberts et al. 2002). Animals were housed individually in metabolic cages, and urine and feces were collected for a period of 4 days following dosing. RBA was calculated as the UEF ratio for soil and sodium arsenic group (soil/arsenate). RBA was estimated for multiple soils in each of five animals, and estimates were combined into a mean and SE for the RBA for a given soil. (Roberts et al. 2007). Each animal received three dose levels of sodium arsenate (0.25, 0.50, and 1.0 mg As/kg) and a single dose level of each soil. The average UEF for the three sodium arsenate doses measured in a given animal was used to estimate the UEF for sodium arsenate for that animal. This range of doses encompassed the arsenic doses received from soil, and over this range there was no discernible effect of dose on UEF. Average recovery of the administered sodium arsenate dose in excreta was 81 ± 4% (SD), with 41 ± 10% of the dose excreted in urine and 42 ± 9% in feces. Following an intravenous dose of sodium arsenate, <1% of the administered dose was excreted in feces and 80% was excreted in urine within 4 days post-dosing (Roberts et al. 2007). RBA estimates for 18 soils obtained with the monkey assay ranged from 5 to 39% (median 19%, Brattin et al. 2013). An RBA for a soil amended in the laboratory with sodium arsenate was 93%. Standard errors on the RBA estimates were typically 17% of the mean (range 5–36%), based on estimates for 14 soils (Roberts et al. 2007). A similar assay protocol was used to estimate RBAs for 5 soils in Cebus monkeys; estimates ranged from 11 to 25% (Roberts et al. 2002).

Table 1.

Comparison of Designs of Four Arsenic Soil RBA Bioassays

Primary Reference	Brattin and Casteel et al. 2013	Juhasz et al. 2007b	Roberts et al. 2007	Bradham et al. 2011
Animal	Swine	Swine	Monkey	Mouse
Sex/body weight/age	Male, 7–12 kg, 5–6 wk	Female, 20–25 kg, adult	Male, 4–5 kg, adult	Female, 15–20 g, 4–5 wk
Diet	Low-arsenic diet	Low-arsenic pig pellets	Low-arsenic pelletized diet	AIN-93G purified powdered rodent diet
Groups	Separate test material and reference groups (5 animals/group)	Each of 3 animals receives test material and reference	Each of 5 animals receives test material and reference	Separate test material and reference groups, 12 animals/group
Fast/fed state for soil or arsenic dosing	Dosed 2 hours before meal	Dosed following overnight fast	Dosed following overnight fast	Dosed in feed
Reference material	Sodium arsenate	Sodium arsenate	Sodium arsenate	Sodium arsenate
Reference material dose delivery	Aqueous solution mixed into doughball and delivered by hand	Aqueous solution delivered by gastric gavage and intravenous infusion (for estimating ABA)	Aqueous solution delivered by gastric gavage	Mixed into feed
Soil dose delivery	Soil mixed into doughball delivered by hand feeding	Soil in aqueous slurry delivered by gastric gavage	Soil in aqueous slurry delivered by gastric gavage	Soil mixed into feed
Dosing schedule	2 times/day for 12 days administered in 2 equal portions	Single dose separated by 48-hour washout period	Single dose separated by at least 3-week washout period	Free access to amended feed for 10 days
Number of dose levels	Control (basal diet), 3 dose levels for soil and reference	Control (basal diet), single dose level for soil and reference	Control (basal diet), single dose level for soil, 3 dose levels for reference	Control (basal diet), single dose level for soil and reference
Reference material dose	300–1600 μg As/day (~30–160 mg/kg)	oral: 80–100 μg As/kg iv: 20 μg As/kg	0.25 – 1.0 mg As/kg	2 mg/kg per day
Soil dose	1–2 g soil/day	10–25 g soil	≤ 1 g soil/kg bw	1% (w/w) soil:feed ratio
Collection of biological samples for estimating RBA	Urine: 48-hour samples collected from each animal on days 6–7, 8–9 and 10–11 during dosing	Blood: prior to dosing (baseline) and then repeatedly over a 26 hour post-dosing period	Urine: cumulative sample collected over a 4 days post-dosing period	Urine: cumulative sample collected for 8 days during dosing and 1 day post-dosing
RBA calculation	RBA = UEFSoil/UEFRef UEF estimated from simultaneous regression of individual animal dose and excretion rate data.	RBA = (AUCSoil/AUCRef) * (DRRef/DRSoil) AUC calculated from serial blood samples	RBA = UEFSoil/UEFRef UEF calculated from cumulative dose and urinary excretion	RBA = UEFSoil/UEFRef UEF calculated from cumulative dose and urinary excretion.

ABA, absolute bioavailability; AUC, area under the curve for arsenic in blood; RBA, relative bioavailability; RM: Reference Material; TM: Test Material; UEF: Urinary Excretion Fraction

Juvenile Swine Assay

Swine UEF Assay (Brattin and Casteel et al. 2013)

The swine UEF assay estimates soil arsenic RBA by comparing the urinary excretion of arsenic in groups of swine that received repeated doses of arsenic in soil or sodium arsenate (Brattin and Casteel et al. 2013). Details of the assay protocol are presented in Table 1. Groups of juvenile swine (5–6 weeks of age) received daily doses of soil or sodium arsenate for a period of 12 days. Groups included three soil and three arsenate dose levels; a control group received only the basal diet. Urine was collected over 48-hour periods on days 6–7, 8–9, and 10–11 during dosing. Animals were housed individually. Urine was collected passively in a stainless steel pan placed beneath the cage and drained into plastic storage vessels. The collection trays were covered with a nylon screen to minimize contamination of the urine with feces. RBA was calculated as the UEF ratio for soil and sodium arsenic group (soil/arsenate). Group UEFs were estimated as the slope of the linear regression line relating dose rate (μg As/48 hr) and urinary excretion rate (μg As/48 hr). Data from the soil and sodium arsenate groups were fit simultaneously to weighted regression models having a common intercept. Weighting was used to adjust for heteroscedasticity of the excretion rates (increasing variance with increasing arsenic dose rate). Uncertainty in the RBA estimate (a ratio of slopes) was estimated by applying Fieller’s theorem, accounting for variance and covariance in slope estimates (Finney 1978). RBA estimates for 19 soils obtained with the swine UEF assay ranged from 19 to 60% (median 38%; Brattin et al. 2013). Based on RBAs reported for 14 soils in Brattin and Casteel (2013), the standard errors on the RBA estimates were typically 8% of the mean (range 4–17%). One soil was assayed in two different studies; the RBA estimates were 13 ± 4% SE and 18 ± 2% (U.S. EPA 2010). The average UEF for sodium arsenate, based on 14 swine studies, was 77 ± 8% (range: 62–89%, based on data reported in U.S. EPA 2010). The effect of the dosing vehicle (doughball) on arsenic bioavailability was tested by estimating UEF in groups of swine that had received sodium arsenate in a doughball or in an aqueous solution by gastric gavage. The UEF ratio (doughball/gavage) was 0.94 (U.S. EPA 2010).

Swine Area Under the Curve (AUC) Assay (Juhasz et al. 2007b; Rees et al. 2009)

The swine assay estimates soil arsenic RBA by comparing the time integrated plasma arsenic concentration (AUC) following an oral dose of soil or sodium arsenate (Juhasz et al. 2007b; Rees et al. 2009). Details of the assay protocol are presented in Table 1. Individual young adult swine received a single oral dose of soil and sodium arsenate, with a minimum 48-hour interval between dosing. Re-use of the animals is possible because of the relatively rapid elimination of absorbed arsenic in swine (Juhasz et al. 2007b). If ABA was assessed, the animal also received an intravenous injection of sodium arsenate. Serial blood samples were collected from an indwelling jugular cannula, beginning during the first hour following dosing and continuing for a period of 26 hours. Soil RBA was estimated from dose-normalized AUCs for plasma arsenic concentration following the oral dose of soil or sodium arsenate.

Estimates of RBA from three animals were combined to estimate a mean and SE for the RBA for a given soil. RBA estimates for 24 soils obtained with the swine AUC assay ranged from 7 to 80% (median 21%; Diamond et al. 2016). Standard errors on the RBA estimates were typically 15% of the mean (range 2–48%).

Although sodium arsenate was used as the reference material for estimating soil arsenic RBA, Juhasz et al. (2007b) also compared the ABA of orally administered sodium arsenate and soluble arsenite (As^III[OH]₃). ABA was estimated from dose-normalized AUCs for plasma arsenic concentrations following the oral and intravenous dose of sodium arsenate or arsenite.

ABA values for arsenate and arsenite were 92.5 ± 22.3% (SE) and 103.9 ± 25.8%, respectively.

Mouse assay

The mouse as an in vivo assay test species has low purchase and husbandry costs, ease of handling, and excellent precision gained from larger sample sizes compared with swine or primates (Bradham et al. 2011). Mice are well characterized physiologically and can be manipulated experimentally (e.g., altered dietary components, altered genotype) to determine the effects of biological variation on the gastrointestinal absorption of metals and metalloids.

The mouse UEF assay estimates soil arsenic RBA by comparing the urinary excretion of arsenic in groups of mice that receive repeated doses of arsenic in soil or sodium arsenate (Bradham et al. 2011). Details of the assay protocol are presented in Table 1. Female C57BL/6 mice, 4–6 weeks of age, were housed in metabolic cages (three mice/cage) allowing collection of urine and feces. Mice were maintained on a lowarsenic basal diet (AIN-93G) that had an arsenic concentration of <1 ppm, which contributed negligible levels of arsenic in urine. Groups of 12 mice were fed a diet amended with either sodium arsenate or the test soil, for a period of 10 days, during which urine and feces were collected daily and food consumption was also measured daily. Each group of mice was fed a single arsenic dose level. RBA of arsenic in soil was calculated as the ratio of UEF in animals that ingested arsenic in soil or sodium arsenate, where UEF was the ratio of the cumulative urinary arsenic excretion (μg per 10 days) and the cumulative dietary arsenic intake (μg per 10 days).

The mouse assay can serve as a highly cost-effective alternative or supplement to monkey and swine assays for improving arsenic risk assessments by providing site-specific assessments of RBA of arsenic in soils. Recent studies have shown strong correlations between the mouse assay and bioaccessibility assays (Bradham et al. 2011, 2015; Juhasz et al. 2014). A recent study evaluated the mouse in vivo−in vitro correlation for 40 soils with varying soil types and arsenic contamination sources. As part of this study, an independent data validation was developed for the mouse in vivo-in vitro correlation for the 0.4M glycine gastric phase bioaccessibility method (U.S. EPA Method 9200.2–86), providing critical supporting information for use in human health risk assessment (Bradham et al. 2015).

Conceptual Bases for Soil Arsenic RBA Assays

Soil arsenic RBA bioassays have relied on two metrics of absorbed arsenic dose for estimating bioavailability: area under the plasma or blood concentration-time curve (AUC; Freeman et al. 1995; Juhasz et al. 2007b) or the urinary excretion fractions (UEF; Bradham et al. 2011; Brattin and Casteel 2013; Freeman et al. 1993, 1995; Roberts et al. 2002, 2007).

Assessment of RBA from Measurement of Blood or Plasma AUC

Blood or plasma AUC provides a direct measure of the amount of arsenic that entered the systemic circulation. It is estimated from a time series of arsenic concentrations following a single dose, measured at a frequency and duration that accurately captures the absorptive and elimination phases and allows accurate calculation of the time-integrated concentration. Accurate measurement of AUC requires repeated sampling of blood at relatively high frequencies during and immediately following the absorptive phase. As a result, the AUC method is better suited for large animals, because of the requirement to maintain intravenous catheters to monitor blood arsenic concentrations over time for estimating the AUC.

ABA is estimated from the dose-normalized AUC (i.e., AUC/dose) for an oral and intravenous dose (Equation 1).

$$ABA=\frac{AU{C}_{oral}}{AU{C}_{iv}}\cdot \frac{{Dose}_{iv}}{{Dose}_{oral}}$$ Eq. (1)

Soil RBA is estimated as the ratio of the ABAs for the soil and reference, where the reference is a soluble arsenic species (e.g., As^VO₄) administered in solution or mixed into some other dosing vehicle (e.g., food). Since the dose-normalized AUC for intravenous arsenic is a constant (ABA_iv is always unity), the intravenous AUC cancels from the equation for the ABA ratio and is not needed to estimate RBA (Equation 2).

$$RBA=\frac{AB{A}_{oral,soil}}{AB{A}_{oral,ref}}=\frac{AU{C}_{oral,soil}}{AU{C}_{oral,ref}}\cdot \frac{{Dose}_{oral,ref}}{{Dose}_{oral,soil}}$$ Eq. (2)

Assessment of RBA from Measurement of UEF

Using UEF rather than plasma AUC to estimate RBA obviates the need for serial blood collections, which would be impractical for some applications (e.g., mice). However, it requires that urine be collected free of contamination with feces, which can be a challenge even when animals are housed in metabolic cages. Contamination of urine with unabsorbed arsenic in feces will result in overestimation of soil arsenic RBA if the true UEF for soil arsenic is less than that of the reference. Soil arsenic RBA has been estimated from measurements of UEF following a single dose (Freeman et al. 1993; Roberts et al. 2002, 2007) or repeated doses that result in steady-state or near steady-state conditions (Bradham et al. 2011; Brattin and Casteel 2013). When a single dose is administered, urine is collected until urinary excretion of the absorbed dose is complete (Roberts et al. 2002, 2007). When repeated doses are administered, urine is repeatedly sampled to estimate the average steady-state rate of excretion or cumulative amount excreted over the dosing period. In some RBA assays, multiple dose levels of arsenic are administered (Brattin and Casteel 2013; Freeman et al. 1993). Multiple dose levels allow examination of the relationship between urinary excretion rate and dose rate, from which an appropriate regression model can be selected for estimating UEF (Brattin and Casteel 2013).

RBA is estimated from measurements of the UEF (Equation 3):

$$UEF=AF\cdot K_u$$ Eq. (3)

where AF is the fraction of the oral dose absorbed and K_u is the fraction of the absorbed dose excreted in urine.

The ratio of arsenic UEFs following dosing with soil or reference (e.g., AsO₄) will equal the ratio of the absorption fractions (RBA), provided that K_u is a constant (K_u,soil = K_u,ref, Equation 4).

$$RBA=\frac{UE{F}_{oral,soil}}{UE{F}_{oral,ref}}=\frac{A{F}_{oral,soil}}{A{F}_{oral,ref}}\cdot \frac{K_{u,soil}}{K_{u,ref}}$$ Eq. (4)

Constancy of K_u following dosing with soil and reference has not been rigorously tested in soil RBA studies. This would require measurement of the fraction of the absorbed dose that is excreted in urine. However, several observations suggest that K_u can be expected to be similar following dosing with soil or reference. Similar K_u for soil and reference would be expected if the soil and reference dosing results in the absorption of the same arsenic species. There are several reasons to suggest that this will occur. Soluble inorganic arsenic that becomes bioaccessible from soil in the GIT is likely to be a mixture of arsenate (As^VO₄⁻, AsO₄²⁻) and arsenite (As^III[OH]_3; Bissen and Frimmel 2000; Montperrus et al. 2002). Eh/pH profiles for arsenic predict that arsenate ion and arsenite will be the dominate arsenic species in the GIT (Bissen and Frimmel 2000; Smedley and Kinniburgh 2002). Absorptive transport mechanisms for these two arsenic species have been described above. In swine, rates of elimination of arsenic from plasma are similar following a dose of arsenate or arsenite, consistent with similar excretory mechanisms (Juhasz et al. 2007b). ABAs of arsenate and arsenite in swine, estimated from dose-normalized AUC, are nearly identical; 92.5 ± 22.3% (SE) and 103.9 ± 25.8% (Juhasz et al. 2007b). In mice, UEFs of arsenate and arsenite are also nearly identical, 62 ± 5% and 66 ± 3%, respectively (Bradham et al. 2013). RBAs for soils measured by blood AUC and UEF in mice and plasma AUC in swine show good agreement (Li et al. 2016; Bradham et al. 2015). While the above observations support the validity of the use of UEF to estimate RBA, only one study has actually compared the arsenic UEF with ABA measured by blood AUC (Freeman et al. 1995). The results were not definitive. The average UEF (normalized to UEF for an intravenous dose of sodium arsenate) in monkeys that received an oral dose of sodium arsenate was 67.6 ± 2.6% and the corresponding ABA based on blood AUC was 91.3 ± 12.4%. It is unlikely that first-pass biliary secretion of arsenic absorbed from the GIT explains the intravenous-normalized UEF being lower than the ABA based on blood AUC (Roberts et al. 2002, 2007). In monkeys that received a dose of soil, the ABAs of soil arsenic measured by blood AUC and UEF were similar, 10.9 ± 5.2% and 13.8 ± 3.3%, respectively (Freeman et al. 1995). However, in monkeys that received a dose of house dust, the UEF actually exceeded the ABA (25 ± 3.2% vs. 10.9 ± 5.2%).

Additional studies that compare ABA measured by plasma AUC and UEF would be useful for evaluating whether or not they provide equivalent measures of soil arsenic RBA.

Conceptual Basis for Inter-species Differences in Arsenic RBA

The fraction of an oral dose that is absorbed (f) is the product of two processes, bioaccessibility and absorptive transport. The process of rendering soil arsenic bioaccessible may involve: (1) physical and/or chemical digestion of the soil particles to expose arsenic deposits; (2) transfer of arsenic minerals in soil to the aqueous environment of the GIT; and (3) chemical transformation of arsenic minerals to soluble arsenic species that are substrates for absorptive transport. Absorptive transport refers to all processes involved including diffusion in transferring soluble arsenic species derived from soil from the GIT to the systemic circulation. In Equation 5, bioaccessibility and absorptive transport are represented as the fraction of the oral dose that is bioaccessible (f_B) and the fraction of the bioaccessible fraction, f_B, that is absorbed by transport (f_A:B), respectively (Equation 5):

$$ABA=f_B\cdot f_{A:B}$$ Eq. (5)

The corresponding expression for RBA is the soil/reference ABA ratio (Equation 6):

$$RBA=\frac{{\left[f_B\cdot f_{A:B}\right]}_{soil}}{{\left[f_B\cdot f_{A:B}\right]}_{ref}}$$ Eq. (6)

Interspecies differences can be expected for processes involved in both bioaccessibility and absorptive transport. However, interspecies differences in absorptive transport will not result in interspecies differences in RBA if the following two conditions hold: (1) similar mechanisms participate in the absorptive transport of soil arsenic and reference; and (2) the rate of absorptive transport of bioaccessible arsenic is a linear function of bioaccessible dose (e.g., not saturated or induced at administered doses). If the second condition holds, then f_A:B will be independent of the bioaccessible dose, in which case [f_A:B]_soil will equal [f_A:B]_ref and the ratio [f_A:B]_soil/[f_A:B]_ref will equal 1; and RBA will be determined solely by the bioaccessibility ratio, [f_B]_soil/[f_B]_ref.

The first condition (similar mechanisms participate in the absorptive transport of soil arsenic and reference) would be satisfied if the same soluble arsenic species occur in the GIT following a dose of soil or reference. If the soluble inorganic arsenic species in the GIT consists of a mixture of arsenate ion and arsenite, then the first condition will be satisfied. The basis for this is the observation that arsenate and arsenite have very similar bioavailabilities when administered to mice or swine (Bradham et al. 2015; Juhasz et al. 2007b). Achieving the second condition (rate of absorptive transport of bioaccessible arsenic as a linear function of bioaccessible dose) requires consideration of possible non-linearities arising from saturable transport or induction of absorptive transport. If not linear, the experimental assessment of RBA becomes more complicated because any dose-dependence in absorption must be quantified from measurements made in the RBA bioassay. This can be achieved by including multiple dose levels in the assay so that the relationship between dose and bioavailability can be assessed (Brattin and Casteel 2013).

Some empirical evidence exists to support the idea that bioaccessibility is an important factor contributing to interspecies differences and similarities in arsenic RBA. Bioaccessibility of soil arsenic measured with in vitro extraction assays strongly correlates with RBA measured in mouse and swine assays (see discussion of bioaccessibility assays below, Bradham et al. 2015; Brattin et al. 2013; Diamond et al. 2016; Juhasz et al. 2015). A single linear regression model relating IVBA (extraction in 0.4M glycine at pH 1.5) and RBA explained 87% of the observed variance in RBA for a large dataset (83 soils) that combined RBA estimates from mouse and swine assays (Diamond et al. 2016). The specific bioassay (i.e., mouse UEF assay, swine UEF assay, or swine AUC assay) had minimal effect on the strength of the IVBA-RBA correlation, as only 6% of the variance in RBA was explained by which laboratory collected the IVBA and RBA data. The same IVBA assay failed to explain variance in soil arsenic RBA measured in a monkey UEF assay (Brattin et al. 2013). Although there are several possible explanations for this, one explanation is that conditions that determine bioaccessibility of soil arsenic in the monkey GIT are different from the determinants of bioaccessibility in mice and swine.

If the RBA assay can be designed to satisfy the above two conditions, then interspecies differences observed in RBA could be assumed to result from differences in bioaccessibility. As it is unlikely that RBA assays in humans would be performed in the future, it might be prudent to use or develop animal models in which the physiological and biochemical factors that influence bioaccessibility are most similar to humans. Although the factors that influence bioaccessibility have yet to be completely defined, pH is certainly one factor as it is an important determinant for IVBA assays predicting soil arsenic RBA (Brattin et al. 2013). In addition, for highly insoluble arsenic minerals, kinetics of dissolution may be sufficiently slow that gastric emptying times may also be important determinants of bioaccessibility. Reported values for pH of the monkey, swine, and mouse stomach fall within the range observed for humans (1.5–5.0; Desesso et al. 2012). Stomach transit times for these species also overlap (Desesso et al. 2012). This suggests that stomach pH and stomach transit times may not be discriminating factors for animal models used to predict soil arsenic RBA in humans.

Studies that have directly compared soil arsenic RBA measured on the same soils in different animal assays have been limited to a small number of soils (Bradham et al. 2013; Li et al. 2016). Swine RBAs tended to be higher than mouse RBAs in each of these studies (Table 2). These studies also found that standard errors for swine RBAs tended to be larger than for mouse RBAs. This difference is not always considered in regression analyses of these data, which calls for weighted regression. Nevertheless, a clear trend is evident for higher RBAs from swine assays than from mouse assays; although, the magnitude of the difference is relatively small (relative difference <25%) and is not universally applicable in that some soils produced very similar RBA estimates. For example, differences between RBA estimates from the mouse UEF and swine UEF assays for standard reference soils was <2.5% (Bradham et al. 2013). The trend for higher RBA estimates from the swine UEF assay compared to the mouse UEF assay may reflect real interspecies differences in bioaccessibility or absorptive transport of soil arsenic. However, a contribution of methodological differences cannot be ruled out because assay procedures differ in many ways, including dosing regimens, collections of biological samples, and data reduction (Table 1). For example, the potential contribution of fecal contamination of urine to differences in RBAs estimated from the mouse and swine UEF assays has not been experimentally investigated. Only one comparison is available that included the monkey UEF assay. When four test soils from the same site were assessed in the mouse UEF, monkey UEF, and swine UEF assays, the mean soil arsenic RBAs were not statistically different (Bradham et al. 2013). Collectively, these studies indicate that mouse, swine, and monkey RBA assays yield similar RBA estimates when applied to the same soils, although there appears to be a trend for the swine assay to predict higher RBAs than the mouse UEF assay. The mechanisms for this difference have not been determined.

Table 2.

Regression Models Relating Soil Arsenic RBAs Estimated from Mouse and Swine Models

Data Source	Li et al. 2016	Bradham et al. 2013	Li et al. 2016
Independent variable	Swine blood AUC	Swine blood AUC	Mouse blood AUC
Dependent variable	Mouse blood AUC	Mouse UEF	Mouse UEF
Number of soils	12	12	12
Regression slope	0.80	0.801	0.86
Intercept	−2.50	−0.431	7.33
R2	0.83	0.521	0.87

¹Weighted least squares regression

In vitro bioaccessibility methods

Due to time, expense, and ethical considerations associated with in vivo assessment of arsenic RBA, IVBA assays have been developed to measure the extent of arsenic solubilization in simulated gastrointestinal fluids (Bradham et al. 2011, 2015; Brattin et al. 2013; Juhasz et al. 2007a,b, 2009, 2011, 2014; Rodriguez et al. 1999; Ruby et al. 1996; Wragg et al. 2011). IVBA determination provides an indicative measurement of the amount of arsenic potentially available for absorption into the systemic circulation. These methods are attractive alternatives to in vivo assays as they are cost-effective and reduce reliance on animal studies. However, a prime assumption underlying IVBA assays is that there is a strong relationship between the fraction of arsenic solubilized in vitro and the fraction of arsenic that may cross the gastrointestinal barrier. As a consequence, if an IVBA method is to be applied as a surrogate measure for arsenic RBA, then it must be shown to reliably predict in vivo arsenic RBA for soils with varying arsenic sources and physicochemical properties (U.S. EPA 2007). Due to the complexity of the human GIT, IVBA methods have been developed to mimic key biochemical parameters known to influence the release of contaminants from soil. As detailed in Table 3, a variety of in vitro assays have been developed and utilized for the assessment of arsenic IVBA in contaminated soils. These assays vary in chemical composition of gastrointestinal solutions, the presence or absence of food additions (to represent a fed state), and operational parameters including pH, solid-to-solution ratio, and extraction times (representative of the rate of stomach emptying and small intestine transit time).

Table 3.

Commonly Utilized In Vitro Assays for the Assessment of Arsenic Bioaccessibility

Assay	Gastric Phase					Intestinal Phase				Reference
	Constituents (g l−1)	pH	S:S8 Ratio	Time (hr)		Constituents (g l−1)	pH	S:S Ratio	Time (hr)
SBRC1	30.03 g glycine	1.5	1:100	1		1.75 g bile, 0.5 g pancreatin	7.0	1:100	4	Kelley et al. (2002)
IVG2	10 g pepsin, 8.77 g NaCl	1.8	1:150	1		3.33 g bile, 0.35 g pancreatin	5.5	1:150	1	Rodriguez et al. (1999)
PBET3	1.25 g pepsin, 0.5 g sodium malate, 0.5 g sodium citrate, 420 μl lactic acid, 500 μl acetic acid	2.5	1:100	1		1.75 g bile, 0.5 g pancreatin	7.0	1:100	4	Ruby et al. (1996)
DIN4	1 g pepsin, 3 g mucin, 2.9 g NaCl, 0.7 g KCl, 0.27 g KH2PO4	2.0	1:50	2		9.0 g bile, 9.0 g pancreatin, 0.3 g trypsin, 0.3 g urea, 0.3 g KCl, 0.5 g CaCl2, 0.2 g MgCl2	7.5	1:100	6	DIN (2000)
RIVM5	Saliva (pH 6.5±0.2): 0.896 g KCl, 0.888 g NaH2PO4, 0.2 g KSCN, 0.57 g Na2PO4, 0.298 g NaCl, 0.7 g NaOH, 0.2 g urea, 0.145 g amylase, 0.05 g mucin, 0.015 g uric acid	1.2	1:37.5	2		Duodenal phase (pH7.8±0.2): 7.012 g NaCl, 3.388 g NaHCO3, 0.08 g KH2PO4, 0.564 mg KCl, 0.05 g MgCl2, 0.18 ml of 37% HCl, 0.1 g urea, 0.2 g CaCl2, 1 g bovine serum		1:98	2	Oomen et al. (2003)
	Gastric Phase: 2.752 g NaCl, 0.266 g NaH2PO4, 0.824 g KCl, 0.4 g CaCl2, 0.306 g NH4Cl, 8.3 ml of 37% HCl, 0.65 g glucose, 0.02 mg glucuronic acid, 0.085 g urea, 0.33 g glucosamine hydrochloride, 1 g bovine serum albumin, 3 g mucin, 1 g pepsin					albumin, 3 g pancreatin, 0.5 g lipase Bile phase (pH8.0±0.2): 5.259 g NaCl, 5.785 g NaHCO3, 0.376 g KCl, 0.2 ml of 37% HCl, 0.25 g urea, 0.222 CaCl2, 1.8 g bovine serum albumin, 6 g bile (chicken)
UBM6	Saliva (pH 6.5±0.5): 0.896 g KCl, 0.888 g NaH2PO4, 0.2 g KSCN, 0.57 g Na2SO4, 0.298 g NaCl, 1.8 ml of 1 M NaOH, 0.2 g urea, 0.145 g amylase, 0.05 g mucin, 0.015 g uric acid Gastric Phase: 2.752 g NaCl, 0.266 g NaH2PO4, 0.824 g KCl, 0.4 g CaCl2, 0.306 g NH4Cl, 8.3 ml of 37% HCl, 0.65 g glucose, 0.02 mg glucuronic acid, 0.085 g urea, 0.33 g glucosamine hydrochloride, 1 g bovine serum albumin, 3 g mucin, 1 g pepsin	1.2	1:37.5	1		Duodenal phase (pH7.4±0.2): 7.012 g NaCl, 5.607 g NaHCO3, 0.08 g KH2PO4, 0.564 mg KCl, 0.05 g MgCl2, 0.18 ml of 37% HCl, 0.1 g urea, 0.2 g CaCl2, 1 g bovine serum albumin, 3 g pancreatin, 0.5 g lipase Bile phase (pH8.0±0.2): 5.259 g NaCl, 5.785 g NaHCO3, 0.376 g KCl, 0.18 ml of 37% HCl, 0.25 g urea, 0.222 CaCl2, 1.8 g bovine serum albumin, 6 g bile	6.5	1:100	4	Wragg et al. (2011)
SHIME7	15 g Nutrilon, 16 g pectin, 8 g mucin, 5 g starch, 1 g cellobiose, 1 g glucose, 2 g proteose peptone	2.0 (Fasted) 4.0 (fed)	1:10 (fasted) 1:40 (fed)	3		Add 25 ml of NaHCO3 (12 g l−1), bovine bile (4 g l−1), pancreatin (0.9 g l−1).	6.5	1:15 (fasted) 1:60 (fed)	5	Van de Wiele et al. (2007)
In vitro GI model	Acidified Milli-Q-Water	1.5	1:100	2		Add 35 ml of NaHCO3 (12.5 g l−1), Oxgall (6 g l−1), pancreatin (3 g l−1)	6.5	1:150	2	Laird et al. (2011)

¹The gastric phase of the SBRC assay is also referred to as the Simplified Bioaccessibility Extraction Test (SBET), the Relative Bioaccessibility Leaching Procedure (RBALP), and USEPA 9200 and USEPA Solid Waste Method 1340. Modifications to this method have included the addition of phosphate (0.2–0.8 M), hydroxylamine (0.25 M) and sodium hypochlorite (0.25 M) (Brattin et al., 2013)

²May include the addition of a dough ball.

³See Table 2 for variations on the PBET in vitro assay.

⁴May include the addition of milk powder.

⁵A saliva phase (5 minutes) is included prior to gastric phase extraction.

⁶A saliva phase (5 minutes) is included prior to gastric phase extraction.

⁷May include the addition of intestinal microflora.

⁸Soil to solution ratio

IVBA assessment may be undertaken using a single phase extraction methodology (i.e., gastric phase) or following modification of the gastric phase to intestinal phase conditions. In addition, although not common amongst IVBA assays, a saliva phase may also be included prior to gastric phase extraction to simulate processes that may occur in the oral cavity prior to ingestion. Limited studies have investigated the influence of including a saliva phase on arsenic IVBA, although Juhasz et al. (2011) determined that its inclusion in the Unified Bioaccessibility Method (UBM) increased arsenic IVBA when saliva-gastric and gastric-only extraction was compared. Presumably, saliva phase constituents (e.g., phosphate) played a role in the enhancement of arsenic solubilization (see below).

IVBA gastric phase solutions range from simple, comprising acidified water or single/dual constituents (e.g., SRBC and IVG), to complex, where multiple organic and inorganic components are incorporated (e.g., RIVM, UBM. PBET). Pepsin (digestive protease) and glycine (smallest of the 20 amino acids) are base constituents in gastric solutions, although mucin (glycosylated protein) may also be included. When transitioning into the intestinal phase, most assays amend bile and pancreatin to gastric solutions; however, some assays include additional organic and inorganic constituents to represent simulated duodenal and bile phases (e.g., RIVM, UBM). While a systematic analysis of the influence of individual gastrointestinal solution constituents on arsenic IVBA has not been undertaken, some components have been shown to influence arsenic solubilization. The inclusion of organic acids, including malate, citrate, lactic acid, and acetic acid, has been shown to influence iron precipitation. Li et al. (2015a) determined that removal of citrate from the PBET resulted in a significant decrease in the concentration of soluble iron in the intestinal phase with a corresponding decrease in arsenic IVBA. These results suggest that the inclusion of citrate in the PBET intestinal phase extracting fluid provides a higher measure of arsenic IVBA via the inhibition of iron precipitation. The inclusion of phosphate (e.g., DIN, RIVM, UBM) has also been shown to enhance arsenic dissolution from arsenic-contaminated soils (Brattin et al. 2013), as phosphate and arsenate are chemical analogues which compete for sorption sites (Rodriguez et al. 2003). Phosphate may desorb arsenate from iron, aluminum, and manganese oxide surfaces by ligand exchange reactions. Li et al. (2015a) determined that removing phosphate from the DIN assay decreased arsenic IVBA from house dust, although the reduction in arsenic IVBA was more significant in the intestinal phase compared to the gastric phase. The inclusion of food additives to in vitro assays may also influence arsenic IVBA through phosphate-related effects. As detailed by Basta et al. (2007), arsenic IVBA in smelter-contaminated soil was higher following the inclusion of a dough dosing vehicle (used for arsenic RBA assays) when assessed using the IVG assay. The enhancement of arsenic IVBA was more pronounced following intestinal phase extraction, although the effect was variable, presumably due to differences in physicochemical properties of soil matrices.

The pH of gastric phase extractions has been shown to be particularly influential for arsenic IVBA measurement due to its effect on the dissolution of arsenic-bearing minerals and phases (e.g., iron oxides) to which arsenic may be sorbed (Meunier et al. 2010; Juhasz et al. 2009; Brattin et al. 2013). As detailed in Table 3, gastric phase pH values ranging from 1.2 to 2.5, which are similar to stomach pH conditions in fasting humans, have been utilized (Malagelada et al. 1976). However, higher pH values (e.g., 4.0) have been utilized in fed state assays as food consumption impacts gastric phase pH (i.e., increase) which may result in lower arsenic IVBA values (Ruby et al. 1996). A range of pH from 6.1 to 7.5 has been used for IVBA intestinal phase conditions. These values are similar to those observed in the small intestine (pH conditions vary from 5.5 to 6.5 in the duodenum and jejunum, respectively) and the ileum (6.5 to 7.5). For some IVBA assays, significantly lower arsenic IVBA has been observed following intestinal phase extraction compared to the gastric phase. With an increase in pH from gastric to intestinal phase conditions, dissolved iron from gastric phase dissolution becomes oversaturated, and hydrolyzed iron species precipitate as amorphous iron structures (Mercer and Tobiason 2008). Dissolved arsenic may sorb to amorphous iron by surface complexation or ligand exchange to surface hydroxyl functional groups, thereby reducing the concentration of dissolved arsenic in the intestinal phase. In addition, surface precipitation of arsenic on amorphous iron structures might also result in a decrease in dissolved arsenic at intestinal phase conditions; however, previous research has demonstrated that this removal mechanism might be minor or negligible (Fuller et al. 1993; Waychunas et al. 1993).

Some variability in solid-to-solution ratio and extraction time are present between IVBA methods. Soil-to-solution ratios ranging from 1:10 (SHIME assay) to 1:150 (IVG assay) have been utilized. Although solid-to-solution ratio may influence arsenic dissolution kinetics, Hamel et al. (1998) determined that at ratios ranging from 1:100 to 1:5000, IVBA values may not vary significantly. Gastric and intestinal extraction times range from 1 to 3 hours and 1 to 5 hours, respectively, which are similar to stomach emptying times (1–2 hours) and the time required for constituents to transition from the small to large intestines (3–5 hours). However, as detailed by Meunier et al. (2011), within the timeframes of the PBET, a steady-state bioaccessible arsenic concentration may not be achieved for some mine-impacted matrices and, as a consequence, significantly higher arsenic IVBA was observed when intestinal phase extraction was extended to 24 hours.

‘Standard’ operating parameters for SBRC, IVG, PBET, DIN, RIVM, UBM, and SHIME assays are detailed in Table 3; however, for some assays, notably the PBET (Table 4) and IVG (Table 5), some minor modifications have been implemented for the assessment of arsenic IVBA. Modifications include variations in the concentration of constituents, the inclusion or exclusion of constituents within gastric and intestinal phases, variations in gastric phase pH, and modification of intestinal phase extraction times. The justification for these modifications is unclear as is the potential influence on arsenic IVBA outcomes.

Table 4.

Variations in the PBET Utilised During a Round Robin Study to Determine the Variability in Bioaccessibility Results Using Various IVBA Methods and NIST 2710

Constituents (per liter)	PBET variants
	SOP1	V12	V23	V34	V45
Gastric pH	2.5	2.0	1.8	2.5	2.5
Pepsin	1.25 g	1.25 g	1.25 g	0.12 g	0.13 g
Sodium citrate	0.5 g	0.5 g	0.5 g	0.5 g	0.5 g
Sodium malate	0.5 g	0.5 g	0.5 g	0.5 g	0.5 g
Lactic acid	420 μl	420 μl	-	420 μl	420 μl
Acetic acid	500 μl	500 μl	500 μl	500 μl	500 μl
NaCl	-	-	0.15 M	-	-
Intestinal pH	7.0	7.0	7.0	7.0	7.0
Bile	1.75 g	1.75 g	1.75 g	0.36 g	0.36 g
Pancreatin	0.5 g	0.5 g	0.5 g	0.03 g	0.03 g
BSA	-	-	-	9.0 g	0.9 g

Date from Koch et al., 2013

¹Standard operating procedure from Ruby et al. (1996)

²PBET variant utilised by Royal Roads University, British Colombia, Canada

³PBET variant utilised by Royal Military College, Ontario, Canada

⁴PBET variant utilised by University of Guelph, Ontario, Canada

⁵PBET variant utilised by Simon Fraser University, British Colombia, Canada

Table 5.

Variations in the IVG Assay.

Constituents (per liter)	IVG Variants
	V11	V22	V33	V44	V55
Gastric pH	1.8	1.8	1.8	1.8	1.8
Extraction time (hr)	1	1	1	1	1
NaCl	8.77	8.77	8.77	5.84	5.84
Pepsin	10 g	10 g	10 g	10 g	10 g
Intestinal pH	5.5	5.5	5.5	6.1	6.1
Extraction time (hr)	1	1	1	2	2
Bile	3.33 g	4.02 g	3.5 g	3.75 g	4.02 g
Pancreatin	0.35 g	4.02 g	0.35 g	3.75 g	4.02 g

¹IVG operating procedures of Rodriguez et al. (1999)

²IVG operating procedures of Basta et al. (2007)

³IVG operating procedures of Juhasz et al. (2009), Girouard and Zagury (2009), Lu et al. (2011), Mikutta et al. (2014) and Li et al. (2015a)

⁴IVG operating procedures of Koch et al. (2013)

⁵IVG operating procedures of Whitacre et al. (2013)

Assessment of arsenic bioaccessibility in contaminated soil/dust

Figure 1 illustrates the range in arsenic IVBA values for contaminated soil and dust, determined using commonly utilized in vitro methodologies, while Table 6 provides summary statistics for individual assays and phases. Although additional arsenic IVBA data may be available from the literature, unpublished research reports, or site assessments, Figure 1 and Table 6 highlight the broad range of values (0–100%) that may be encountered depending on arsenic source, arsenic speciation, particle size distribution, and physicochemical properties of the matrix. Although significant differences in arsenic IVBA may be observed for individual soils using gastric and intestinal phase extraction, for most assays, there was no difference in mean arsenic IVBA values when gastric or intestinal phase values were compared for soils in Table 6 where both analyses were undertaken. In contrast, significantly higher arsenic IVBA values were determined for analyses using the SBRC gastric phase (mean = 24.1%) compared to intestinal phase extraction (mean = 15.9%; n = 162) as a consequence of arsenic-iron precipitation in the intestinal phase as detailed above. For a subset of soils listed in Table 6 (n = 76), arsenic IVBA assessment was undertaken using gastric and intestinal phases of the SBRC, PBET, IVG DIN, and UBM assays, allowing a direct comparison of IVBA outcomes for different methodologies. Table 7 details the relationship between IVBA results for each methodology and phase compared to data determined using the SBRC gastric phase. The SBRC gastric phase provided the most conservative (i.e., highest) measure of arsenic IVBA as demonstrated by the slope of the in vitro-in vitro relationship (0.42–0.96; Table 7 and Figure 2).

Figure 1.

Assessment of arsenic bioaccessibility in contaminated soil and dust using a variety of in vitro gastro-intestinal extraction methods (⬛: gastric phase extraction; ⬜: intestinal phase extraction).

Table 6.

Compilation of Data from the Literature Illustrating the Range in Arsenic IVBA Values for Contaminated Soil and Dust Determined Using Commonly Utilised In Vitro Methodologies

Assay	Phase	n	Arsenic Bioaccessibility (%)
			Minimum	Maximum	Mean	Median	95% CI
SBRC	Gastric	411	0.0	96.0	23.5	17.0	21.5–25.4
	Intestinal	170	0.6	80.0	16.8	9.0	14.1–19.4
PBET	Gastric	238	0.0	95.0	15.2	8.2	12.8–17.5
	Intestinal	223	0.0	73.7	17.4	11.5	15.3–19.4
IVG	Gastric	138	0.5	92.1	31.3	23.5	27.0–35.6
	Intestinal	138	0.5	90.0	26.4	22.3	22.9–30.0
DIN	Gastric	77	1.5	83.0	32.5	30.0	27.3–37.8
	Intestinal	80	3.0	96.6	32.3	29.5	27.3–37.3
UBM	Gastric	129	3.0	99.6	22.4	12.5	18.9–25.8
	Intestinal	73	2.5	88.4	28.6	22.0	23.4–33.8

See Figure 1.

Table 7.

Relationship Between arsenic IVBA Data for Soils Assessed Using a Variety of IVBA Methods

Assay	Phase	Relationship Parameters
		Slope	Y-intercept	r2
SBRC	Intestinal	0.42	5.00	0.36
PBET	Gastric	0.74	−2.56	0.84
	Intestinal	0.63	−0.15	0.78
IVG	Gastric	0.96	−4.90	0.90
	Intestinal	0.65	−0.95	0.80
DIN	Gastric	0.74	−0.42	0.81
	Intestinal	0.75	−0.49	0.80
UBM	Gastric	0.87	6.24	0.68
	Intestinal	0.86	5.02	0.64

Arsenic IVBA results for each methodology and phase are compared to data determined using the SBRC gastric phase.

¹Data (n=76) are from Juhasz et al. (2014), Juhasz et al. (2015), Li et al. (2014), Li et al. (2015b) and Li et al. (2015c)

Figure 2.

Comparison of arsenic bioaccessibility results for soils assessed using a variety of in vitro gastro-intestinal extraction methods (n = 76). Arsenic bioaccessibility results for each methodology and phase are compared to data determined using the SBRC gastric phase (Data from Juhasz et al., 2014; Juhasz et al., 2015; Li et al., 2014; Li et al., 2015b; Li et al., 2015c)

Research Needs and Conclusions

The authors have endeavored to provide a comprehensive review of the science underlying the oral bioavailability of soil arsenic and the methods that are currently being explored for estimating soil arsenic RBA. Recent research has focused on the development of models that utilize simple extraction tests of arsenic in soil to reliably predict RBA soil arsenic in animal species, and, by extension, humans. While substantial progress has been made, several important problems remain to be solved which, if explained, would further increase our confidence in risk estimates:

There is a lack of consensus on which animal model provides the most reliable prediction of soil arsenic oral RBA in humans. A definitive answer to this question is not likely to be obtained without human clinical studies; the feasibility of such studies has been demonstrated (Stanek et al. 2010). In the meantime, several problems would benefit from exploration. Concern remains that IVBA assays that perform well at predicting soil arsenic RBA in mice and swine perform poorly at predicting RBA in monkeys (Brattin et al. 2013). This suggests that there may be factors that determine soil arsenic bioaccessibility in primates which are different from those that determine bioaccessibility in other test species. Understanding these factors is important for gaining confidence in IVBA assays that are being considered for use in human health risk assessment.

Mouse and swine RBA assays tend to yield similar RBA estimates when applied to the same soils, although a trend has been observed of swine assay RBAs being higher than mouse RBA assays. An understanding of the mechanisms for this difference would increase confidence in application of these assays for human health risk assessment. In particular, it is important to discern whether or not the differences reflect interspecies or methodological differences. Measurement of RBA in additional soils would also improve estimates of how much the assays actually differ, and whether or not the magnitude is of practical significance to risk assessment.

Several IVBA assays have been described for predicting RBA. Most of these have been evaluated with a relatively small number of soils, an exception being an evaluation of the SRBC assay with 83 soils (Diamond et al. 2016). Confidence in applications of these IVBA assays would be improved by external validations. Regression models relating IVBA to RBA could be externally validated with measurements of RBA and IVBA for a representative collection of soils that have not been used to develop regression models. Such studies may also identify outlier soils that could provide insight into important physical or chemical characteristics that influence bioaccessibility.

A recent development in remediation of soils contaminated with lead has been the application of amending agents (e.g., phosphates) that decrease lead bioaccessibility and bioavailability. Methods are needed for evaluation of efficacy of amending agents and their influence on lead and arsenic, as these are common co-occurring contaminants. In vitro tests would be ideal for this purpose because of their relatively low expense. However, IVBA assays have not been rigorously tested for accuracy in predicting RBA when applied to amended soils.

It is likely that even the best IVBA assays will not accurately predict arsenic RBA for all soil arsenic mineralogical contexts. Therefore, the development of IVBA assays for predicting RBA should not preclude continued investigation of soil arsenic bioavailability in animal models and correspondence between measured RBA and IVBA. It will be important to continue to refine our understanding and confidence in the predictive accuracy and limitations of IVBA methods, and to develop methods that can identify soils for which existing IVBA assays may perform poorly.

Most of the samples with IVBA results are limited to soil, but interior house dusts, as well as interior-derived dusts, represent the proximate source of exposure to soil for young children (Lanphear et al. 1998; Succop et al. 1998).

The persistent research on bioavailability has and will continue to advance the risk assessment paradigm for metals that pose human health risks (e.g., lead and arsenic). The ultimate objectives to be realized are rapid and cost-efficient characterizations of bioavailability of metals at hazardous waste sites, and development of soil clean-up goals that more accurately reflect the risk posed by these metals.