Predicting Arsenic Relative Bioavailability Using ... - ACS Publications

Article pubs.acs.org/est

Predicting Arsenic Relative Bioavailability Using Multiple in Vitro Assays: Validation of in Vivo−in Vitro Correlations Albert L. Juhasz,*,† Paul Herde,‡ Carina Herde,‡ John Boland,§ and Euan Smith† †

Centre for Environmental Risk Assessment and Remediation, University of South Australia, Mawson Lakes, South Australia 5095, Australia ‡ South Australian Health and Medical Research Institute, Gilles Plains, South Australia 5086, Australia § Centre for Industrial and Applied Mathematics, University of South Australia, Mawson Lakes, South Australia 5095, Australia

Downloaded by UNIV OF WINNIPEG on August 29, 2015 | http://pubs.acs.org Publication Date (Web): August 28, 2015 | doi: 10.1021/acs.est.5b02508

S Supporting Information *

ABSTRACT: In this study, previously established arsenic (As) in vivo−in vitro correlations (IVIVC) were assessed for their validity using an independent data set comprising As relative bioavailability (RBA) and bioaccessibility values for 13 herbicide- and mine-impacted soils. The validation process established the correlation between As RBA (swine model) and bioaccessibility (five in vitro assays), determined whether correlations differed significantly from previous relationships and assessed model bias and error. The capacity of in vitro assays to predict As RBA was demonstrated by the strength of IVIVC; goodness of fit ranged from 0.53 (DIN-I) to 0.74 (UBM-I). When compared to previous IVIVC (Juhasz et al. Environ. Sci. Technol. 2009, 43, 9487; Juhasz et al. J. Hazard. Mater. 2011, 197, 161), there was no significant difference (P < 0.01) in the slope and y-intercept for IVG-G, UBM-G, and UBM-I indicating the consistency of these assays for predicting As RBA. However, variability in model bias and prediction error was observed with significantly lower (P < 0.01) error determined for IVG-G suggesting that As RBA predictions using IVG-G may be more robust compared to UBM-G and UBM-I. In contrast, differences in the slope and/or y-intercept were observed for SBRC-I, IVG-I, PBET-G, PBET-I, DIN-G, and DIN-I suggesting that these methodologies may not be suitable for predicting As RBA.

■

al.,18 to have confidence in the use of bioaccessibility assays for predicting As RBA for human health exposure refinement, validation of the relationship between in vivo and in vitro measures is required. In a previous study, Juhasz et al.12 validated the predictive capabilities of the SBRC gastric phase assay (SBRC-G) for estimating As RBA in contaminated soil. Validation was achieved by assessing As RBA (swine assay encompassing area under the blood As concentration time curve) and As bioaccessibility (SBRC-G) in 13 contaminated soils, developing an in vivo−in vitro correlation (IVIVC) and comparing this correlation to one that had previously been established.11 Comparison of IVIVC determined that there was no significant difference in the slope and y-intercept while cross-validation analysis established that residuals and prediction errors were low, illustrating the capacity of SBRC-G to accurately predict As RBA.12 As detailed previously, a number of studies have demonstrated the potential of other in vitro assays for predicting As RBA in contaminated soil.11,14−17 As a consequence, this study

INTRODUCTION Incidental ingestion of contaminated soil and dust is a major nondietary exposure pathway for arsenic (As). However, human exposure is influenced by the amount of As in soil or dust which is actually absorbed into the systemic circulation (i.e., the bioavailable fraction).1 Arsenic bioavailability is influenced by physicochemical properties of the contaminant and the matrix in addition to physiological characteristics of the individual.1,2 Quantifying exposure for human health risk assessment requires robust methodologies for measuring or predicting As relative bioavailability (RBA). Although animal models, including mice,3,4 swine,5,6 and monkeys,7−9 may be utilized for the measurement of As RBA in contaminated soil, these assays are prohibitively expensive, require extended time frames, and are ethically challenging. To overcome these challenges, a wealth of research has investigated the use of in vitro assays to act as surrogate measures of As RBA. A number of researchers have reported the correlation between As RBA and As bioaccessibility using in vitro assays including the solubility bioaccessibility research consortium assay (SBRC),3,6,10−13 in vitro gastrointestinal extraction method (IVG),11,14,15 physiologically based extraction test (PBET),11 Deutsches Institut für Normunge.V. (DIN),11 and the unified bioaccessibility research group of Europe method (UBM).16,17 However, as detailed by Juhasz et © XXXX American Chemical Society

Received: May 25, 2015 Revised: July 27, 2015 Accepted: August 21, 2015

A

DOI: 10.1021/acs.est.5b02508 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

Table 1. Selected Properties of Soils ( 0.05) between values derived using the UBM and SBRC gastric phases for samples S1, S2, S3, S5, and S8. Lower gastric phase As bioaccessibility values were obtained using the IVG (mining-impacted soils) and PBET (pesticide-impacted soils) assays; however, there was no significant difference (P > 0.05) between As bioaccessibility values (for either As source) when assessed using the PBET or DIN gastric phases (Figure 1a). The variability in As bioaccessibility was not unexpected as the assays varied in gastric phase pH (1.5−2.5) which has been shown to influence the dissolution of As and associated mineral phases.10,11,14,20,24−27 The increase in As bioaccessibility with decreasing gastric phase pH has previously been observed.11,18 Although both SBRC and UBM gastric phase extractions were conducted at pH 1.5, higher As bioaccessibility values were obtained using the UBM due to the influence of gastric phase constituents (see SI Table S1) on the enhancement of As solubilization.26 Higher gastric phase As bioaccessibility values

have previously been reported when applying the UBM in vitro assay compared to the SBRC.16 When assays were modified to include the intestinal phase, As bioaccessibility was also variable depending on the assay utilized. In 8 of the 13 soils analyzed, the UBM intestinal phase provided the highest As bioaccessibility values of the 5 assays (Figure 1b). For the SBRC assay, As bioaccessibility decreased by 1.2−3.2-fold (P < 0.05) when gastric and intestinal phase values were compared. For IVG and UBM assays, As bioaccessibility was either equivalent or decreased following modification of gastric phase conditions to include the intestinal phase. DIN intestinal phase As bioaccessibility values were also lower compared to the gastric phase value for miningimpacted soils; however, higher values were obtained for pesticide-impacted soils. In contrast, utilizing the PBET intestinal phase, As bioaccessibility was either congruent (7 of 13 soils) or higher (6 of 13 soils) than gastric phase values (Figure 1a,b). The decrease in As bioaccessibility following intestinal phase extraction has previously been observed11,20,26 and may result from the precipitation of Fe, as a result of the increase in pH, and the subsequent sorption of As via surface complexation or ligand exchange.28,29 For the SBRC assay, the concentration of Fe was reduced by up to 23-fold following transition from the gastric to the intestinal phase (data not shown). In contrast, an increase in As bioaccessibility was observed for some assays. This may occur due to the enhanced solubilization of Fe as a result of the presence of organic acids in the extracting solution. At the intestinal phase pH, these organic acids will provide more dissociated, deprotonated functional groups that can bind Fe thereby enhancing C


Article

Environmental Science & Technology Table 2. Comparison of Linear Regression Models from This Study and Juhasz et al.11,16 for Predicting As Relative Bioavailability Using SBRC, IVG, PBET, UBM, and DIN Assays Incorporating Gastric and Intestinal Phases linear regression model (this study) assay d

SBRC-G SBRC-I IVG-G IVG-I PBET-G PBET-I UBM-G UBM-I DIN-G DIN-I

linear regression model (Juhasz et al.11,16)

slope

y-interc

R2

slope

y-inter

R2

0.69 1.02 0.89 0.91 0.60 0.66 0.54 0.58 0.64 0.88

5.24 6.85 5.14 6.06 10.20 8.42 4.53 5.91 9.03 5.00

0.75 0.71 0.69 0.62 0.59 0.56 0.70 0.74 0.63 0.53

0.99 1.67 0.85 1.11 1.16 1.77 0.99 1.08 1.78 1.45

1.89 5.25 14.33 13.88 10.06 5.56 0.80 −3.73 5.64 9.29

0.92 0.83 0.71 0.71 0.80 0.83 0.52 0.59 0.68 0.66

comparison of modelsa slope P P P P P P P P P P

= = = = < < = = < =

0.06 0.05 0.88 0.53 0.05 0.01 0.13 0.08 0.01 0.18

y-inter

grouped modelb slope

y-inter

0.14 0.05 0.09 0.05

0.83 1.29 0.87 1.02

3.90

P = 0.06 P = 0.22

0.65 0.71

5.87 5.11

P < 0.05

1.12

P P P P

= < =
0.05) in the slopes or y-intercepts of IVIVC from this study and those from Juhasz et al.11,16 (Table 2), similar to findings for SBRC-G.12 This indicates the consistency in the IVIVCs (between studies) and satisfies a validation criterion detailed in Juhasz et al.18 Grouped linear regression models were developed for the respective in vitro assays with slopes and y-intercepts ranging from 0.65 to 0.87 and from 3.90 to 9.68, respectively. When models developed using SBRC-I, IVG-I, or DIN-I were compared to previous IVIVC, there was no significant difference between the slopes of the relationships although a significant difference (P < 0.05) in the y-intercepts was observed. In contrast, the slope of IVIVC developed in this study using PBET-G, PBET-I, and DIN-G differed significantly (P < 0.05) from those developed by Juhasz et al.11,16 (Table 2). As detailed in Juhasz et al.,12 a number of cross-validation methodologies may be utilized to confirm the predictive capabilities of in vitro assays for predicting As RBA. In this study, repeat random subsampling cross-validation was utilized as it provides a more robust approach compared to “2-fold” and “leave one out” methodologies for the assessment of prediction accuracy.12,23 Data sets exhibiting no significant difference in the slope and y-intercept for linear regression models developed from this study and Juhasz et al.11,16 (Table 2) were assessed using this approach. As detailed in Materials and Methods, paired in vivo−in vitro data (IVG-G, UBM-G, and UBM-I) from this study and Juhasz et al.11,16 were combined resulting in data for 25 soils. These data sets were randomly divided into training (two-thirds of the data) and holdout (onethird of the data) sets which were used to construct and test

and 1.02 (SBRC-I); however, there was no significant difference when the slopes were compared (P = 0.16). Similarly, for IVG, PBET, UBM, and DIN assays, there was no significant difference (P > 0.05) in the slope and y-intercept of the IVIVCs for gastric and intestinal phases. Between-assay comparisons yielded similar results to within-assay comparisons for most methodologies. There was no significant difference (P > 0.05) in the slopes and y-intercepts of IVIVCs except when SBRC-I and UBM-G correlations were compared (P < 0.05). This indicates that the UBM-G is more efficient at extracting As from soil matrices than SBRC-I. The IVIVCs detailed in Table 2 are similar to previously published correlations. Previously, Juhasz et al.12 reported the similarity in IVIVC (SBRC-G) to those derived by Brattin et al.6 and Bradham et al.3 Additionally, for IVIVCs derived by Rodriguez et al.14 and Basta et al.15 using swine and IVG assays, the slopes of gastric (0.91, R2 = 0.85 and 1.09, R2 = 0.85, respectively) and intestinal (1.02, R2 = 0.82 and 0.99, R2 = 0.92, respectively) phase relationships were similar to those reported here (0.89 and 0.91, respectively; Table 2). However, the slopes of IVIVCs determined for both gastric (0.54) and intestinal (0.58) phases of the UBM assay were significantly different (P < 0.001 and P < 0.01, respectively) from those reported by Denys et al.17 (1.03, R2 = 0.99 and 1.10, R2 = 0.99, respectively). These differences may have arisen from operational differences in the determination of As RBA (urinary excretion versus AUC analysis) in addition to differences in gastric phase extractions (pH 1.5 versus 1.2) and its influence on As dissolution from soils with varying physicochemical properties. Validating the Relationship between As Relative Bioavailability and As Bioaccessibility. In order to validate the As RBA predictive performance of in vitro methodologies, data generated from this study were used as an independent data set to test the strength of previously established IVIVC.11,16 As detailed in Materials and Methods, the initial E


Article


Table 3. Repeat Random Subsampling Cross-Validation Analysis To Determine the Accuracy and Precision of As Relative Bioavailability Predictions Using IVG-G, UBM-G, and UBM-I training set slope assay IVG UBM

SBRCi

phase g

gastric gastrich gastricg gastrich intestinalg intestinalh gastricg gastrich

n

d

17 16 17 16 17 16 17 16

a

MBE e

mean

z

0.76 0.76 1.03 1.14 1.06 1.04 1.03 1.02

0.07 0.08 0.14 0.11 0.14 0.09 0.02 0.03

holdout set

b

RMSE

c

MBE f

b

RMSEc

mean

std dev

mean

std dev

n

mean

std dev

mean

std dev

−0.12 −1.08 4.58 2.45 3.43 2.46 0.11 −0.76

1.78 2.83 1.87 1.76 1.63 1.16 1.06 0.74

10.22 9.95 17.40 14.72 15.09 14.40 7.94 6.43

1.51 1.40 3.14 3.15 1.59 1.60 2.23 0.68

8 8 8 8 8 8 8 8

0.17 −1.41 6.77 −3.83 1.21 3.48 1.30 −1.89

6.79 6.31 9.39 8.15 9.66 6.43 3.16 2.46

13.68 11.70 20.34 21.31 17.32 17.03 11.24 6.92

3.75 3.79 6.27 6.08 5.50 3.04 3.71 1.70


a

Slope of the in vivo−in vitro linear regression models (with zero y-intercept). bMean bias error. cRoot mean square error. dNumber of paired As relative bioavailability−bioaccessibility data points used for in vivo−in vitro linear regression model construction. e95% confidence interval for the mean of the slope of best fit for the training set. fNumber of paired As relative bioavailability−bioaccessibility data points used to test the strength of the in vivo−in vitro relationship. gThe cross-validation method was repeated 10 times utilizing randomly selected training (∼2/3 of the data; n = 17) and holdout (∼1/3 of the data; n = 8) sets from this study and Juhasz et al.11,16 Values reported (slope, MBE, and RMSE) represent the mean and standard deviation of 10 analyses. hThe cross-validation method was repeated 10 times utilizing randomly selected training (∼2/3 of the data; n = 16) and holdout (∼1/3 of the data; n = 8) sets from this study and Juhasz et al.11,16 where S3 was omitted. Values reported (slope, MBE, and RMSE) represent the mean and standard deviation of 10 analyses. iThe accuracy and precision of SBRC-G for predicting As relative bioavailability (Juhasz et al.12) is included for comparison.

significant difference (P > 0.05) in residuals and prediction errors when SBRC-G and IVG-G were utilized for As RBA estimates, values were significantly higher (P < 0.05) when linear regression models utilized UBM-G or UBM-I. This suggests that the relative error for predicted As RBA values would be higher when utilizing linear regression models incorporating UBM-G and UBM-I in vitro data compared to SBRC-G and IVG-G. Withholding sample S3 from repeat random subsampling cross-validation analysis significantly (P < 0.01) reduced holdout set prediction errors (6.92 ± 1.70) for SBRC-G;12 however, exclusion of this sample had no significant influence on training set residuals and holdout set prediction errors for IVG-G, UBM-G, and UBM-I. Comparison of As Relative Bioavailability Predictive Performance. In order to assess the variability in As RBA predictive performance for different in vitro assays, As bioaccessibility data from Juhasz et al.13 was utilized as input parameters into validated IVIVC from this study (Tables 2 and 3). Previously, these mine-impacted soils (1−9) were assessed for As bioaccessibility using a variety of in vitro assays including SBRC-G, IVG-G, UBM-G, and UBM-I (SI Table S5). As detailed in Figure 4, when As bioaccessibility and IVIVC were utilized to predict As RBA, some variability was observed between models. When the range in predicted As RBA for individual soils was determined using zero y-intercept models from Table 3 (i.e., 95% confidence intervals for the mean of the slope of best fit for repeat random subsampling training sets), values determined using SBRC-G were the most conservative and presented the highest precision (Figure 4a). Predicted As RBA values using IVG-G were the least conservative for 6 of the 9 soils while predicted values were congruent for 7 of 9 soils when determined using gastric and intestinal phases of the UBM. However, when As RBA was predicted using grouped models incorporating y-intercepts (Table 2), predictions based on SBRC-G and IVG-G models resulted in more conservative estimates compared to UBM-G and UBM-I (Figure 4b). Although differences in predicted As RBA values were evident, when varying in vitro assays were utilized, data presented in Figure 4b does not account for, or incorporate, model

IVIVC, respectively. Training and holdout sets were randomly repeated 10 times with mean values determined for the slope of the line of best fit, model bias, and model error. For each IVIVC, because the y-intercept was not significantly different from zero (P > 0.05), all analyses were performed with a zero yintercept. In addition, as sample S3 was identified as an outlier (as detailed in Juhasz et al.12) and appeared to bias the results (Figures 2 and 3), repeat random cross-validation was repeated with this sample withheld to determine its influence on As RBA prediction accuracy. For IVIVC determined using IVG-G, low mean residuals (10.22 ± 1.51) and holdout set prediction errors (13.68 ± 3.75) were calculated indicating the accuracy of this in vitro assay for predicting As RBA (Table 3). The mean MBE in the holdout set (0.17) indicated that predicted values were similar to the true or measured As RBA values. In contrast, mean residuals for the training set (17.40 ± 3.14) and prediction errors (20.34 ± 6.27) were significantly higher (P < 0.01) for UBM-G compared to IVG-G. Similar training and holdout set RMSEs were calculated for UBM-G and UBM-I although differences in holdout set MBEs (6.77 versus 1.21) suggest that UBM-I predicted values are closer to the true or measured As RBA values compared to those predicted using UBM-G. Although Denys et al.17 determined that there was no significant difference between As IVIVC when UBM-G or UBM-I was utilized, conceivably UBM-I may provide a more robust approach when As-contaminated soils, other than those impacted from mining and smelting activities, are assessed (i.e., those with discrete As minerals compared to sorbed As species). When data from sample S3 (outlier) was withheld from repeat random subsampling cross-validation analysis, mean residuals, and holdout set prediction errors varied for IVG-G, UBM-G, and UBM-I; however, values were not significantly different (P > 0.05) compared to its inclusion. Previously, Juhasz et al.12 validated the relationship between As RBA and As bioaccessibility determined using the SBRC-G assay. Low residuals (7.94 ± 2.23) and prediction errors (11.24 ± 3.71) were determined using the 10-fold repeat random subsampling cross-validation approach. Although there was no F


Article



Figure 4. Predicted As relative bioavailability for mine-impacted soils 1−9 from Juhasz et al.13 using As bioaccessibility data derived from SBRC-G (black filled bars), IVG-G (unfilled bars), UBM-G (gray filled bars), and UBM-I (dotted filled bars) and validated linear regression models. The range in predicted As relative bioavailability for individual soils was determined using zero y-intercept models incorporating 95% confidence intervals for the mean of the slope of best fit for repeat random subsampling training sets, as detailed in Table 3 (A). Data with low variability in predicted As relative bioavailability has been circled for clarity. Predicted As relative bioavailability was also determined using grouped models incorporating y-intercepts detailed in Table 2 (B).

Figure 5. Predicted As relative bioavailability for mine-impacted soils 2 (A) and 8 (B) from Juhasz et al.13 using linear regression models detailed in Table 2 and incorporating prediction intervals (1 standard deviation of the mean). The range in As relative bioavailability was predicted using the corresponding model for duplicate in vitro analyses undertaken using SBRC-G, IVG-G, UBM-G, and UBM-I assays. Overlap in predicted values (gray bars) indicates that even though there is an apparent difference in predicted As relative bioavailability (Figure 4B), the within-sample error in the data used to construct the models is sufficient to explain the differences between predicted values.

prediction errors. In order to assess the influence of model error on predicted As RBA values, prediction intervals (1 standard deviation of the mean) were constructed and examined to determine whether there was “significant” overlap. Overlap would indicate that even though there is an apparent difference in predicted As RBA as shown in Figure 4, the within-sample error in data used to construct the models is sufficient to explain the differences between predicted values. In 8 of 9 soils, there was sufficient overlap (>25% of the interquartile range) of the prediction intervals when SBRC-G, IVG-G, UBM-G, and UBM-I and the corresponding models were utilized to predict As RBA; a representative graph is presented in Figure 5a. As highlighted in Figure 5a, a common As RBA region of between 61 and 70% was calculated when As RBA was predicted in soil 2 using duplicate in vitro analyses undertaken with the respective bioaccessibility assays. Similar overlap in As RBA prediction was observed for other soils with the exception of soil 8 (Figure 5b), although the reason for the lack of overlap is unclear. For 8 of the 9 soils, the within-sample error of the data used to define prediction models (Table 3) explained the apparent discrepancy in As RBA estimations (Figure 4) between the four validated IVIVC. This implies that either model may be utilized to predict As RBA in contaminated soil; however, prediction precision will vary.

While it is difficult to define an acceptable standard for prediction error when validating IVIVC, the smaller the better. Conceivably, prediction error for the four validated IVIVCs may be reduced by increasing the size of the data set (i.e., assessment of additional As-contaminated soils), reducing data variability (i.e., additional in vivo replicates), and increasing the range of values being predicted (i.e., incorporating additional soils with As RBA values > 50%).

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b02508. Details of As bioaccessibility methodologies (Table S1), As-contaminated soil from Juhasz et al.11,16 (Table S2), data for IVIVC and validation analysis (Tables S3 and S4), and soil parameters for comparative As RBA predictive performance (Table S5) (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Tel.: +618 8302 5045; fax: +618 8302 3057; e-mail: Albert. [email protected]. G


Article

Environmental Science & Technology Notes

(15) Basta, N. T.; Foster, J. N.; Dayton, E. A.; Rodriguez, R. R.; Casteel, S. W. The effect of dosing vehicle on arsenic bioaccessibility in smelter-contaminated soil. J. Environ. Sci. Health, Part A: Toxic/ Hazard. Subst. Environ. Eng. 2007, 42, 1275−1281. (16) Juhasz, A. L.; Weber, J.; Smith, E. Influence of saliva, gastric and intestinal phases on the prediction of As relative bioavailability using the unified Bioaccessibility Research Group of Europe method (UBM). J. Hazard. Mater. 2011, 197, 161−168. (17) Denys, S.; Caboche, J.; Tack, K.; Rychen, G.; Wragg, J.; Cave, M.; Jondreville, C.; Feidt, C. In vivo validation of the unified BARGE method to assess the bioaccessibility of arsenic, antimony, cadmium, and lead in soils. Environ. Sci. Technol. 2012, 46, 6252−6260. (18) Juhasz, A. L.; Basta, N. T.; Smith, E. What is required for the validation of in vitro assays for predicting contaminant relative bioavailability? Considerations and criteria. Environ. Pollut. 2013, 180, 372−375. (19) Kelley, M. E.; Brauning, S. E.; Schoof, R. A.; Ruby, M. V. Assessing Oral Bioavailability of Metals in Soil; Battelle Press: Columbus, OH, USA, 2002. (20) Ruby, M. V.; Davis, A.; Schoof, R.; Eberle, S.; Sellstone, C. M. Estimation of lead and arsenic bioavailability using a physiologically based extraction test. Environ. Sci. Technol. 1996, 30, 422−430. (21) Soil quality−Absorption availability of organic and inorganic pollutants from contaminated soil material, Standard DIN E19738; Deutsches Institut fur Normung e.V: Berlin, 2000. (22) Wragg, J.; Cave, M.; Basta, N.; Brandon, E.; Casteel, S.; Denys, S.; Gron, C.; Oomen, A.; Reimer, K.; Tack, K.; Van de Wiele, T. An inter-laboratory trial of the unified BARGE bioaccessibility method for arsenic, cadmium and lead in soil. Sci. Total Environ. 2011, 409, 4016− 4030. (23) Kohavi, R. A study of cross-validation and bootstrap for accuracy estimation and model selection. IJCAI’95 Proceedings of the 14th International Joint Conference on Artificial Intelligence, Vol. 2; Morgan Kaufman: San Francisco, CA, USA, 1995; pp 1137−1143. (24) Basta, N. T.; Rodriguez, R. R.; Casteel, S. W., Bioavailability and risk of arsenic exposure by the soil ingestion pathway. In Environmental Chemistry of Arsenic; Frankenberger, W. T., Jr., Ed.; Marcel Dekker, New York, 2001. (25) Juhasz, A. L.; Smith, E.; Weber, J.; Rees, M.; Rofe, A.; Kuchel, T.; Sansom, L.; Naidu, R. In vitro assessment of arsenic bioaccessibility in contaminated (anthropogenic and geogenic) soils. Chemosphere 2007, 69, 69−78. (26) Smith, E.; Scheckel, K. G.; Miller, B. W.; Weber, J.; Juhasz, A. L. Influence of in vitro assay pH and extractant composition on As bioaccessibility in contaminated soils. Sci. Total Environ. 2014, 473− 474, 171−177. (27) Oomen, A. G.; Hack, A.; Minekus, M.; Zeijdner, E.; Cornelis, C.; Schoeters, G.; Verstraete, W.; Van de Wiele, T.; Wragg, J.; Rompelberg, C. J. M.; Sips, A. J. A. M.; Van Wijnen, J. H. Comparison of five in vitro digestion models to study the bioaccessibility of soil contaminants. Environ. Sci. Technol. 2002, 36, 3326−3334. (28) Fuller, C. C.; Davis, J. A.; Waychunas, G. A. Surface chemistry of ferrihydrite 0.2. Kinetics of arsenate adsorption and coprecipitation. Geochim. Cosmochim. Acta 1993, 57, 2271−2282. (29) Waychunas, G. A.; Rea, B. A.; Fuller, C. C.; Davis, J. A. Surface chemistry of ferrihydrite 0.1. EXAFs studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta 1993, 57, 2251−2269. (30) Nierop, K. G. J. J.; Jansen, B.; Verstraten, J. M. Dissolved organic matter, aluminium and iron interactions: Precipitation induced by metal/carbon ratio, pH and competition. Sci. Total Environ. 2002, 300, 201−211. (31) Bowell, R. J. Sorption of arsenic by iron oxides and oxyhydroxides in soils. Appl. Geochem. 1994, 9, 279−286. (32) Grafe, M.; Eick, M. J.; Grossl, P. R.; Saunders, A. M. Adsorption of arsenate and arsenite on ferrihydrite in the presence and absence of dissolved organic carbon. J. Environ. Qual. 2002, 31, 1115−1123.

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowledge the support of the Centre for Environmental Risk Assessment and Remediation and Centre for Industrial and Applied Mathematics, University of South Australia, and the South Australian Health and Medical Research Institute for this research.


■

REFERENCES

(1) Basta, N. T.; Juhasz, A. L. Using in vivo bioavailability and/or in vitro gastrointestinal bioaccessibility testing to adjust human exposure to arsenic from soil ingestion. Rev. Mineral. Geochem. 2014, 79, 451− 472. (2) Ruby, M. V.; Schoof, R.; Brattin, W.; Goldade, M.; Post, G.; Harnois, M.; et al. Advances in evaluating the oral bioavailability of inorganics in soil for use in human health risk assessment. Environ. Sci. Technol. 1999, 33, 3697−3705. (3) Bradham, K. D.; Scheckel, K. G.; Nelson, C. M.; Seales, P. E.; Lee, G. E.; Hughes, M. F.; Miller, B. W.; Yeow, A.; Gilmore, T.; Serda, S. M.; Harper, S.; Thomas, D. J. Relative bioavailability and bioaccessibility and speciation of arsenic in contaminated soils. Environ. Health Perspect. 2011, 119, 1629−1634. (4) Bradham, K. D.; Diamond, G. L.; Scheckel, K. G.; Hughes, M. F.; Casteel, S. W.; Miller, B. W.; Klotzbach, J. M.; Thayer, W. C.; Thomas, D. J. Mouse assay for determination of arsenic bioavailability in soil. J. J. Toxicol. Environ. Health, Part A 2013, 76, 815−826. (5) Rees, M.; Sansom, L.; Rofe, A.; Juhasz, A. L.; Smith, E.; Weber, J.; Naidu, R.; Kuchel, T. Principles and application of an in vivo swine assay for the determination of arsenic bioavailability in contaminated matrices. Environ. Geochem. Health 2009, 31, 167−177. (6) Brattin, W.; Drexler, J.; Lowney, Y.; Griffin, S.; Diamond, G.; Woodbury, L. An in vitro method for estimation of arsenic relative bioavailability in soil. J. Toxicol. Environ. Health, Part A 2013, 76, 458− 478. (7) Freeman, G. B.; Schoof, R. A.; Ruby, M. V.; Davis, A. O.; Dill, J. A.; Liao, S. C.; Lapin, C. A.; Bergstrom, P. D. Bioavailability of arsenic in soil and house dust impacted by smelter activities following oral administration in cynomolgus monkeys. Toxicol. Sci. 1995, 28, 215− 222. (8) Roberts, S. M.; Weimar, W. R.; Vinson, J. R. T.; Munson, J. W.; Bergeron, R. J. Measurement of arsenic bioavailability in soil using a primate model. Toxicol. Sci. 2002, 67, 303−310. (9) Roberts, S. M.; Munson, J. W.; Lowney, Y. W.; Ruby, M. V. Relative oral bioavailability of arsenic from contaminated soils measured in the cynomolgus monkey. Toxicol. Sci. 2007, 95, 281−288. (10) Juhasz, A. L.; Smith, E.; Weber, J.; Rees, M.; Rofe, A.; Kuchel, T.; Sansom, L.; Naidu, R. Comparison of in vivo and in vitro methodologies for the assessment of arsenic bioavailability in contaminated soils. Chemosphere 2007, 69, 961−966. (11) Juhasz, A. L.; Weber, J.; Smith, E.; Naidu, R.; Rees, M.; Rofe, A.; Kuchel, T.; Sansom, L. Assessment of four commonly employed in vitro arsenic bioaccessibility assays for predicting in vivo relative arsenic bioavailability in contaminated soils. Environ. Sci. Technol. 2009, 43, 9487−9494. (12) Juhasz, A. L.; Herde, P.; Herde, K.; Boland, J.; Smith, E. Validation of the predictive capabilities of the SBRC-G in vitro assay for estimating arsenic relative bioavailability in contaminated soils. Environ. Sci. Technol. 2014, 48, 12962−12969. (13) Juhasz, A. L.; Smith, E.; Nelson, C.; Thomas, D.; Bradham, K. Variability associated with As in vivo-in vitro correlations when using different bioaccessibility methodologies. Environ. Sci. Technol. 2014, 48, 11646−11653. (14) Rodriguez, R. R.; Basta, N. T.; Casteel, S. W.; Pace, L. W. An in vitro gastrointestinal method to estimate bioavailable arsenic in contaminated soils and solid media. Environ. Sci. Technol. 1999, 33, 642−649. H


Article



(33) Grafe, M.; Grossl, P. R.; Eick, M. J. Adsorption of arsenate(V) and arsenite(III) on goethite in the presence and absence of dissolved organic carbon. Soil Sci. Soc. Am. J. 2001, 65, 1680−1687. (34) Zhu, J.; Pigna, M.; Cozzolino, V.; Caporale, A. C.; Violante, A. Sorption of arsenite and arsenate on ferrihydrite: Effect of organic and inorganic ligands. J. Hazard. Mater. 2011, 189, 564−571.

I


Predicting Arsenic Relative Bioavailability Using ... - ACS Publications

Recommend Documents