Intra- and Interlaboratory Evaluation of an Assay of Soil Arsenic Relative Bioavailability in Mice

Abstract

Hand-to-mouth activity in children can be an important route for ingestion of soil and dust contaminated with inorganic arsenic. Estimating the relative bioavailability of arsenic present in these media is an element in assessing the risk associated with aggregate exposure to this toxic metalloid during early life. Here, we evaluated the performance of a mouse assay for arsenic bioavailability in two laboratories using a suite of 10 soils. This approach allowed us to examine both intra-laboratory and inter-laboratory variation in assay performance. Use of a single vendor for preparation of all amended test diets and of a single laboratory for arsenic analysis of samples generated in the participating laboratories minimized contributions of these potential sources of variability in assay performance. Intra-laboratory assay data showed that food and water intake and cumulative urine and feces production remained stable over several years. The stability of these measurements accounted for the reproducibility of estimates of arsenic bioavailability obtained from repeated intra-laboratory assays using sodium arsenate or soils as the test material. Inter-laboratory comparisons found estimates of variables used to evaluate assay performance (recovery, urinary excretion factor) were similar in the two laboratories. For all soils, estimates of arsenic relative bioavailability obtained in the two laboratories were highly correlated (r²= 0.98; slope = 0.9) in a linear regression model. Overall, these findings show that this mouse assay for arsenic bioavailability provides reproducible estimates using a variety of test soils. This robust model may be adaptable for use in other laboratory settings.

Introduction

Inorganic arsenic is a potent toxicant and carcinogen.(1,2) Because it is widely distributed in the environment, it poses a significant health hazard to populations worldwide.(3) A special concern is the vulnerability of children to adverse effects caused by exposure to this metalloid. Exposure in utero or in early life to inorganic arsenic has been associated with many adverse health effects, including adverse pregnancy outcome and infant mortality, altered growth, delayed neurodevelopment, and immune system dysfunction.(4−9) There is also evidence that early life exposure to inorganic arsenic increases incidences of cardiovascular disease,(10,11) liver disease,(12) respiratory disease,(13) and cancer in adult life.(14−16) Thus, minimizing exposure to inorganic arsenic in early life likely reduces disease burden associated with this agent.

Food and drinking water are commonly the most important sources of exposure to inorganic arsenic.(17−20) However, age-dependent hand-to-mouth activity in children can create a unique pathway for ingestion of inorganic arsenic that is present in soil and dust.(21) These media can be contaminated with inorganic arsenic by mining, smelting, and manufacturing activities and by historical use of arsenic-containing pesticides.(22,23) Characterizing the contribution of hand-to-mouth activity to aggregate arsenic exposure in children requires evaluation of multiple factors including mouthing behavior, quantitation of both soil transfer from surfaces to the child’s hand and the amount of soil ingested by this route, and estimation of the bioavailability of arsenic present in the ingested material.(24,25) Estimation of intake of soil and dust by hand-to-mouth activity is a topic of research under study.(26−28) A recent study indicates that intake by this route in children under two years of age is likely less than 100 μg per day.(29)

The bioavailability of inorganic arsenic in soil and dust can be measured in bioaccessibility assays that determine the solubility of arsenic in chemically defined systems and in bioavailability assays that use intact organisms to determine the uptake of ingested arsenic across the gastrointestinal barrier. Swine and mice are commonly used as the test species in animal models for determination of the relative bioavailability (RBA) of arsenic in soils.(30−34) In general terms, these assays measure arsenic concentrations in tissues (blood, plasma, liver, and kidney) or in urine to provide data for RBA calculations. For a suite of soils that have been evaluated in both species, estimates of RBA for arsenic were comparable, although there was a trend for higher estimates in swine than in mice.(35) As a practical matter, assays that use mice are substantially less costly and less technically demanding than assays performed in swine. Given the potential for widescale use of a mouse assay for estimation of arsenic RBA, we examined the performance of a mouse assay in two laboratories. Participants were a laboratory at the U.S. Environmental Protection Agency (EPA) in Research Triangle Park, North Carolina, USA, that developed and routinely used the assay and a laboratory at the South Australia Health and Medical Research Institute (SAHMRI) in Gilles Plains, South Australia, Australia, that had experience with other mouse and swine assays for soil arsenic bioavailability. Together, results from the two laboratories allowed us to evaluate intra- and interlaboratory differences in assay performance.

Methods

Animal Origin and Acclimatization

Both laboratories used female C57BL/6 mice from local suppliers. Use of female mice facilitated group housing in metabolic cages during bioavailability assays. Female C57BL/6 mice used at the EPA were obtained at 4 weeks of age from Charles River Laboratories (Raleigh, NC). These mice were acclimatized for 12 days, while being housed in groups of three in polypropylene cages with Alpha-dri cellulose bedding (Lab Supply, Northlake, TX) with free access to Pro Lab RMH 3000 5P00 mouse diet (LabDiet, St. Louis, MO) and drinking water that contained <11 μg/L As.(36) Environmental conditions during acclimatization were a 12 h light-12 h dark photocycle and an ambient temperature of 20–22 °C. Four to 6-week-old female C57BL/6 mice used at the SAHMRI were produced at the local Bioresources facility of the SAHMRI, North Terrace. For acclimatization, mice at the SAHMRI were housed three per cage in Tecniplast cages (Sydney, New South Wales, Australia) with recycled paper bedding (Fibrecycle, Helensvale, Queensland, Australia) and free access to meat free rat and mouse diet (Specialty Feeds, Glen Forrest, Western Australia, Australia) and drinking water. Environmental conditions were a 12 h light-12 h dark photocycle and an ambient temperature of 18–24 °C.

Test Materials and Test Diet Preparation

Table1 describes test materials evaluated in mouse assays performed in both laboratories. Test materials included soils contaminated by a variety of industrial, mining, and agricultural activities and a standard reference soil with a certified concentration for arsenic. Soils used as test materials were sieved to produce a <250 μm particle size fraction for incorporation into a test diet. A diet prepared by addition of the reference arsenical (sodium arsenate heptahydrate) provided data on the absorption of a soluble arsenical. These data were used to calculate RBA estimates. On a mass basis, test materials accounted for ≤1% (w/w) of the test diet. Incorporation of ≤1% of test soil into the diet has been found to have no untoward effects on mice that consumed amended diets.(31) One vendor (Dyets, Inc., Bethlehem, PA) prepared all test diets by incorporation of test materials into a powdered AIN-93G rodent diet. One kg batches of each test diet were provided directly to the EPA and the SAHMRI. For some test materials that were assayed repeatedly at the EPA, the vendor prepared separate batches of test diets over a period of several years.

Table 1 –

Characteristics of test materials evaluated in two laboratories

Test material	Identity	Description
1	As36	Cattle dip site soil
2	Maldon	Mine impacted soil
3	MP14	Railway corridor soil
4	SH15	Railway corridor soil
5	Geo1	Gossan or highly mineralized soil
6	Mohr	Orchard soil
7	Ruston	Smelter impact soil
8	MS03	Orchard soil
9	SoFC-1	USGS SRM
10	SRM 2710a	Montana I Soil Highly Elevated Trace Element Concentrations
11	Arsenate	Sodium arsenate, heptahydrate (Sigma, St. Louis, MO)

Mouse Assay for Soil Arsenic Bioavailability

At the beginning of each assay, three adult female mice were transferred into a metabolic cage that separates urine and feces. For the EPA studies, metabolic cages were from Nalgene (Rochester, NY, USA); for the SAHMRI, metabolic cages were from Tecniplast (Via I Maggio, Buguggiate, Italy). For each laboratory, environmental conditions during assays were identical to those used during acclimatization. For these studies, four metabolic cages containing 12 mice constituted a single assay; hence, the unit of observation and analysis for these assays was the cage.

Figure 1 summarizes the design of the mouse assay. Mice were placed in metabolic cages in the morning of study day 1 with free access to diet amended with a test material and drinking water. Daily food consumption of each metabolic cage was calculated as the difference between the weight of the food hopper immediately after each morning’s filling and before replenishment the following morning. Cumulative food consumption for each cage was the sum of daily food consumption recorded on study days 2 through 9. On the mornings of study days 2 through 10, 24 h urine and feces samples were collected from each metabolic cage and pooled to prepare a cumulative urine and a cumulative feces collection. On study day 9, the test diet was replaced with a powdered unamended AIN-93G rodent diet. Substitution of the unamended diet for the last 24 h of the assay assured that absorption of arsenic from the test diet in transit through the gastrointestinal tract would be reflected in urine samples collected on study day 10. Mice were euthanized by CO₂ exposure at the conclusion of assays.

Figure 1 -.

Schematic of mouse assay for determination of bioavailability of arsenic in test materials. Schedules for collection of excreta and monitoring of food intake are indicated. Photocycle shown by alternating black and white bars.

Processing of Diet and Cumulative Excreta Samples

Diet and excreta samples from the SAHMRI were air-shipped to the EPA on ice packs and were stored at −20 °C. All samples from assays performed at either site were processed at the EPA. Multiple aliquots of each test diet were taken for determination of arsenic concentration. Each cumulative urine sample was thawed and weighed. The sample was thoroughly mixed, and multiple aliquots taken for determination of arsenic concentration. Each cumulative feces sample was thawed, weighed, and homogenized by freeze-grinding in liquid nitrogen with a 6850 Freezer/Mill (Spex CertiPrep, Metuchen, NJ). Multiple aliquots of the homogenized cumulative feces sample from each cage were taken for determination of arsenic concentration.

Determination of Arsenic Levels in Excreta and Test Diets

Arsenic concentrations of test diets, urine, and feces collected at the EPA and the SAHMRI were determined by instrumental neutron activation analysis (INAA) at the Department of Nuclear Engineering, North Carolina State University (Raleigh, NC). Analyses were calibrated by concurrent measurement of As concentration in a certified arsenic standard solution (Spex CertiPrep). INAA performance was routinely assessed by concurrent determination of arsenic concentrations in two reference materials, SRM 1566B Oyster Tissue (National Institute of Standards and Technology, Gaithersburg, MD) and CRM TORT2 Lobster hepatopancreas reference material for trace metals (National Research Council of Canada, Ottawa, Ontario, Canada). The mean arsenic mass detection limit was 0.035 μg.

Statistical Methods

All statistical analyses were performed using SAS/STAT software, version 9.4 of the SAS System for Windows SAS software. A p-value ≤0.05 was considered significant in all comparisons.

Estimation of Arsenic RBA in Mice

For each cage in an assay, cumulative arsenic intake was calculated as the product of arsenic concentration in the test diet and the mass of cumulative food intake. Cumulative urinary excretion of arsenic was calculated as the product of arsenic concentration in the cumulative urine sample and the mass of the cumulative urine sample. The urinary excretion fraction (UEF) of arsenic was calculated as the ratio of cumulative excretion of arsenic in urine (μg) to cumulative dietary intake of arsenic (μg), as shown in eq 1

$$UEF\%=100\cdot \frac{Cumulative Urinary Arsenic}{Cumulative Arsenic Intake}$$ Eq. (1)

RBA was calculated as the ratio of the UEF for arsenic in a specific soil-amended diet to the UEF for arsenic in a diet containing the reference arsenical, sodium arsenate (Equation 2):

$$RBA\%=100\cdot \frac{UEF{\%}_{soil arsenic}}{UEF{\%}_{arsenate}}$$ Eq. (2)

Each UEF in eq 2 was derived from multiple estimates of UEF for groups of three mice housed together in a single metabolic cage (the unit of measure in the assay is data from a single cage). Confidence limits on each RBA were estimated based on Fieller’s theorem for estimating confidence limits on the ratio of means.(37) Normal distributions for estimates of UEFs were assumed with adjustments for unequal variance (SAS PROC “TTEST”). This method was previously shown to yield results similar to nonparametric bootstrap methods.(32)

Recovery of Arsenic in Mouse Assays

As a measure of the performance of the mouse assay, we calculated a percentage recovery term, as shown in eq 3

$$Recovery\%=100· \frac{\left(Cumulative urinary As\right)+\left(Cumulative feces As\right)}{\left(Cumulative Arsenic Intake\right)}$$ Eq. (3)

using values defined above.

Interlaboratory Comparisons of RBA

RBA for each of 10 soils determined at the EPA or the SAHMRI were compared by regression analysis. Weighted linear regression (SAS PROC “NLIN”) was performed with uncorrelated weights assigned to each soil (W_i) calculated from eq 4

$$W_i=\frac{\text{w}\left({\text{X}}_{\text{i}}\right)\cdot \text{w}\left({\text{Y}}_{\text{i}}\right)}{{\text{w(X}}_{\text{i}}{\text{)+w(Y}}_{\text{i}}\text{)}\cdot {\text{(β}}^{\text{2}}\text{)}}$$ Eq. (4)

where w(X_i) and w(Y_i) are SAHMRI and EPA RBA weights (1/SE_i²), respectively, and β is the slope of the linear regression line fit by minimizing the weighted sum of squared residuals. (38)

Results and Discussion

Comparability of Test Diets Used in Laboratories

Figure 2 shows arsenic concentrations in test diets used at the SAHMRI and the EPA. Replicate entries for some test materials from the EPA represent test diets used in repeated assays conducted over a period of several years. Arsenic concentrations in diets used at the EPA or the SAHMRI were similar; the mean relative percent difference was 7.8% (range: −8.6, 24%). This suggests that the method of diet preparation was adequate to ensure homogenous distribution of the arsenic-containing test material in the test diet.

Figure 2 -.

Concentrations of arsenic in test diets used at SAHMRI and EPA. Parts per million (ppm) of arsenic in diets used at SAHMRI or EPA. Mean and S.D. shown.

Intralaboratory Comparisons

Calculation of estimates of arsenic RBA in the mouse assay depends on accuracy and reproducibility of measurement of cumulative food consumption (source of arsenic exposure) and cumulative urine collection (route of excretion of absorbed arsenic). Likewise, calculation of estimates of overall recovery of arsenic requires measurement of arsenic concentration in feces and cumulative feces mass. Therefore, we examined cumulative food and water consumption and cumulative urine and feces collections, using data from all contemporaneous assays of arsenic bioavailability conducted at the EPA (Supporting Information, Figure S1). Over the 5-year period when assays were performed, cumulative food and water consumption were stable (Figure S1a,b). The relative percentage difference from mean cumulative food consumption ranged from −8.0 to 9.9% (n = 28) and −25 to 29% (n = 23) for mean cumulative water consumption. In addition, cumulative urine and fecal masses were also stable over this period (Figure S1c,d). The relative percentage difference from the mean cumulative urine mass ranged from −28 to 29% (n = 28). The relative percent difference from the mean cumulative fecal mass ranged from −15 to 9.6% (n = 27). For comparison, we examined the stability of cumulative food and water consumption and cumulative urine and feces masses for studies performed over a period of 1 month at the SAHMRI. For the studies used in this analysis, the relative percent difference from the mean ranged from −11 to 7.8% (n = 11) for cumulative food consumption, −18 to 18% (n = 11) for cumulative water consumption, −27 to 21% (n = 11) for cumulative urine mass, and −22 to 18% for cumulative fecal mass.

To assess stability of estimates of arsenic bioavailability, we compared UEF estimates for the reference arsenical, sodium arsenate (iAs^V), using data from mouse assays performed at the EPA over a period of about 3 years (Supporting Information Figure S2). The relative percent difference from the mean UEF value [64% ± 10 standard deviation (SD)] ranged from −26 to 23%, and absolute difference ranged from −17 to 14%. Given the stability of UEF estimates for iAs^V, we also examined the stability of estimates of RBA for three test materials that were analyzed repeatedly at the EPA (Supporting Information Figure S3). The relative percentage difference from mean RBA for the test material 10 (N = 5) ranged from −7.7 to 5.4%, and absolute difference ranged from −3.0 to 2.1%. The relative percentage difference from mean RBA for the test material 9 (N = 3) ranged from −34 to 32%, and the absolute difference ranged from −4.2 to 3.9%. The relative percentage difference from mean for the test material 6 RBA (n = 2) ranged from −0.23 to 0.23%, and the absolute difference ranged from −0.075 to 0.075%.

Interlaboratory Comparisons

Cumulative food and water consumption and cumulative urine and feces production differed for mice used in assays at the EPA and at the SAHMRI (Table2). Notably, the EPA mice had significantly higher (P ≤ 0.0001) cumulative food cumulative consumption and cumulative urine and feces production and significantly lower cumulative drinking water consumption than did mice used at the SAHMRI. Mean body weights were also significantly higher (P = 0.010) at the EPA; however, the difference was small; 2.1 ± 5.2% (SD) at the start of the assay and 3.8 ± 4.9% (SD) at the end of the assay. Differences in the body weight of mice used in the two laboratories could account for interlaboratory differences in cumulative food and water consumption and urine and feces production. Despite differences in cumulative food consumption, arsenic dosages in the two laboratories were not significantly different (EPA:495 ± 258 μg, SAHMRI: 465 ± 280 μg, paired t (p = 0.27)). Figure 3 shows percentage recoveries for arsenic for the 11 test materials examined in both laboratories; mean recoveries were 86 ± 12 (SD) for the EPA and 85 ± 11 (SD) for the SAHMRI.

Table 2 –

Cumulative food and water consumption and excreta production in two laboratories^a

Variable	EPA	SAHMRI	P value
Cumulative food consumption (g)	75.1 5.1	70.0 7.6	0.0001
Cumulative water consumption (g)	88.7 11.0	77.5 17.8	<0.0001
Cumulative urine weight (g)	28.3 6.5	18.5 4.8	<0.0001
Cumulative feces weight (g)	10.5 1.7	6.5 0.9	<0.0001

^a.Mean (upper) and standard deviation (lower) shown. P values for t-test. n=120 for EPA and n=44 for SAHMRI.

Figure 3 -.

Inter-laboratory comparison of recoveries for arsenic in excreta for 11 test materials. Mean ± SD shown. Test materials identified in Table 1.

Figure 4 shows UEF estimates calculated from data collected at each of the participating laboratories. For the SAHMRI, there is a single determination of UEF for each test material. For the EPA, some test materials were assayed repeatedly. Results show that the assay yielded consistent results over time and nearly identical RBA estimates when implemented in the two laboratories. Figure 5 compares RBA estimates for 10 soils from the EPA and the SAHMRI in a linear regression model. The regression model had a slope of 0.90 ± 0.08 (SE) and an intercept of 0.71 ± 0.60. The regression coefficient (r²) was 0.94; thus, the model explained 94% of observed variance in RBA. The model was weighted for uncertainty (1/SE²) in the estimates from the two laboratories. Weighting was considered appropriate to account for differences in SE, which tended to be larger for the SAHMRI than the EPA. The interlaboratory difference in SE is at least partially explained by the larger sample size for some RBAs, which were repeatedly measured by the EPA (test materials 6, 9, and 10), and the larger sample size for estimates of UEF for sodium arsenate. Mean UEFs for sodium arsenate were 64 ± 11% (SD) from the EPA and 65 ± 10% (SD) from the SAHMRI. Although the weighted model is preferred for these data, weighting had a negligible effect on parameter estimates. The unweighted model had a slope of 0.91 ± 0.05 (SE), an intercept of 0.69 ± 1.3 (SE), and an r² of 0.97.

Figure 4 -.

Inter-laboratory comparison of UEF estimates for soils, reference materials, and the reference arsenical, sodium arsenate. Mean ± SD shown. Test materials identified in Table 1.

Figure 5 -.

Inter-laboratory comparison of RBA estimates for 10 soils. Mean RBA ± SE shown. The solid line is the weighted linear regression model: Intercept: 0.71 ± 0.60 (SE); slope: 0.90 ± 0.078 (SE); r²: 0.98. Dotted lines are 95% lower and upper confidence limits.

The performance of a mouse assay that provides estimates of the RBA of arsenic in soil was examined in two laboratories. Soil arsenic RBA estimates obtained with this mouse assay were similar to those obtained in swine assays.(35) In this study, the EPA and the SAHMRI performed mouse assays on 10 test materials to calculate UEF values used to estimate RBAs for arsenic dose.(32) To the best of our knowledge, this is the first interlaboratory comparison of an assay designed to estimate the RBA for an environmental contaminant. Our results show that the assay yielded consistent results over time and nearly identical RBA estimates when implemented in the two laboratories. Notably, soils that were used in this comparison of mouse assay performance were obtained from a range of sites (Table1) and would reasonably be expected to vary in chemical and physical properties. Despite these putative differences among soils, the mouse assay yielded similar estimates of As RBA in the two laboratories.

Assays performed at the EPA or the SAHMRI used diets prepared by addition of test materials to the powdered AIN-93G rodent diet. This is a purified ingredient rodent diet that meets the nutritional needs of young mice.(39) Because the AIN-93G rodent diet contains low levels of total arsenic (about 13 parts per billion) and complex organic arsenicals,(40) it was an appropriate basal diet for amendment with arsenic-containing test materials. In this study, a single vendor prepared diets used in both laboratories. This approach minimized the role of dietary composition (both in terms of arsenic concentration and nutrient composition) as a variable that potentially could affect assay performance. Similarly, the use of a single laboratory for determination of arsenic levels in all biological samples eliminated interlaboratory variability in analytical performance as a factor in RBA estimates.

Because performance of the assay depended on quantitation of arsenic intake by consumption of the arsenic-containing diet and arsenic excretion in urine, we used data from assays performed at the EPA laboratory to examine temporal variation in food and water intake and in production of urine and feces. Cumulative food and water intakes during assays were relatively stable over the period of observation, suggesting that assay conditions were sufficient to assure similar patterns of food and water intake. Similar stabilities for food and water intakes were consistent with the reports of a significant positive correlation between food and water intake in mice.(41) The relative stability of the mass of cumulative urine and feces collections during assays suggested that procedures for collection of excreta were consistently implemented.

Minimal temporal variation in estimates of the UEF for the reference arsenical sodium arsenate and in RBA estimates for three test materials indicated that assay performance remained constant over several years. Overall, the stability of these estimates probably reflects uniformity of the composition of test diets, the constancy of the laboratory procedures used for animal husbandry and sample collection, and the use of a standardized analytical method for quantitation of arsenic in diets and excreta.

Results of the interlaboratory comparison showed agreement between estimates of percentage arsenic recovery in excreta, UEF, and RBA. The r² for a linear regression model relating RBA estimates for 10 test materials assayed in the two laboratories was 0.94, and the linear slope was 0.90. This indicates nearly perfect correspondence between RBA estimates from the two laboratories. As noted above, the design of this interlaboratory comparison was primarily focused on the procedures that underlie animal husbandry and sample collection. The similarities in mean arsenic recoveries suggested that collection of data on food consumption and excreta collection was similar in the two laboratories.

Several factors can be identified as potential sources of variability in the results of studies performed in the two laboratories and as modifiers of the physiological and behavioral status of mice used in assays. First, although both laboratories used adult female C57Bl/6 mice for assays, these mice were obtained from different sources. Despite a shared inbred mouse strain designation, genetic drift within each breeding population may introduce genetic variability between the mice tested in the two laboratories.(42) Increased divergence between breeding populations could increase the prevalence of polymorphisms in the arsenic (+3 oxidation state) methyltransferase gene. These changes could alter the kinetic properties of the enzyme that catalyzes the formation of mono and dimethylated arsenical metabolites critical to arsenic excretion.(43−45) Second, the use of a powdered AIN-93G rodent diet may have physiological consequences that affect assay performance. In young male C57BL/6 mice, the physical form of the AIN-93G rodent diet affected growth and body composition.(46) Mice consumed about 15% more of the powdered form than of the pelleted form of this diet, resulting in higher body weight, increased fat mass: body weight ratio, and increased plasma insulin levels. These changes could have unrecognized effects on the processes that control the uptake, retention, and clearance of arsenic that could affect estimates of the arsenic bioavailability provided by the assay. Third, there is evidence that housing of mice in metabolic cages, as required in this assay, produces stress-related physiological effects.(47,48) However, studies of physiological effects of housing in metabolic cages have commonly been performed in singly housed mice. It is unclear whether similar physiological changes occur in female mice that are group-housed in metabolic cages and such changes would affect arsenic bioavailability.

In summary, comparable performance of the mouse assay in two laboratories suggest that it could be readily adapted in other laboratories as an efficient tool for estimation of soil arsenic RBA. The availability of a rapid, relatively low cost, and reproducible assay could provide RBA data on arsenic-contaminated soils from Brownfield and Superfund sites and urban gardens.(49−51) This information could help risk assessors and managers protect public health.