Environ. Sci. Technol. 2002, 36, 3326-3334
Comparison of Five In Vitro Digestion Models To Study the Bioaccessibility of Soil Contaminants ,†
that also apply a low gastric pH, and include intestinal conditions, produce lower bioaccessibility values. The lowest bioaccessibility values are observed for a gastrointestinal method which employs a high gastric pH of 4.0.
Introduction ‡
AGNES G. OOMEN,* ALFONS HACK, MANS MINEKUS,§ EVELIJN ZEIJDNER,§ CHRISTA CORNELIS,| GREET SCHOETERS,| WILLY VERSTRAETE,⊥ TOM VAN DE WIELE,⊥ JOANNA WRAGG,# CATHY J. M. ROMPELBERG,† A D R I E¨ N N E J . A . M . S I P S , † A N D JOOP H. VAN WIJNEN∇ National Institute of Public Health and the Environment, Bilthoven, The Netherlands, Ruhr-Universita¨t Bochum, Bochum, Germany, TNO Nutrition, Zeist, The Netherlands, Vito, Mol, Belgium, LabMET, Ghent University, Ghent, Belgium, British Geological Survey, Nottingham, United Kingdom, and GG & GD, Amsterdam, The Netherlands
Soil ingestion can be a major exposure route for humans to many immobile soil contaminants. Exposure to soil contaminants can be overestimated if oral bioavailability is not taken into account. Several in vitro digestion models simulating the human gastrointestinal tract have been developed to assess mobilization of contaminants from soil during digestion, i.e., bioaccessibility. Bioaccessibility is a crucial step in controlling the oral bioavailability for soil contaminants. To what extent in vitro determination of bioaccessibility is method dependent has, until now, not been studied. This paper describes a multi-laboratory comparison and evaluation of five in vitro digestion models. Their experimental design and the results of a round robin evaluation of three soils, each contaminated with arsenic, cadmium, and lead, are presented and discussed. A wide range of bioaccessibility values were found for the three soils: for As 6-95%, 1-19%, and 10-59%; for Cd 7-92%, 5-92%, and 6-99%; and for Pb 4-91%, 1-56%, and 3-90%. Bioaccessibility in many cases is less than 50%, indicating that a reduction of bioavailability can have implications for health risk assessment. Although the experimental designs of the different digestion systems are distinct, the main differences in test results of bioaccessibility can be explained on the basis of the applied gastric pH. High values are typically observed for a simple gastric method, which measures bioaccessibility in the gastric compartment at low pHs of 1.5. Other methods * Corresponding author address: RIVM-LBM, P.O. Box 1, 3720 BA Bilthoven, The Netherlands; phone: +31 30 2742159; fax: +31 30 2744451. e-mail:
[email protected]. † National Institute of Public Health and the Environment. ‡ Ruhr-Universita ¨ t Bochum. § TNO Nutrition. | Vito. ⊥ LabMET. # British Geological Survey. ∇ GG & GD. 3326
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
In human health risk assessment, ingestion of soil can be a major route of exposure to many immobile soil contaminants (1-3). For that reason, oral bioavailability values for ingested soil contaminants are relevant to assess. Oral bioavailability of soil contaminants is defined here as the contaminant fraction that reaches systemic circulation. This definition assumes that toxicity is exerted by the parent compound and not by formed metabolites. Figure 1 describes the four major processes of oral bioavailability for soil contaminants. After soil ingestion, contaminants can be partially or totally released from the soil matrix during digestion. The fraction of the contaminant that is mobilized from soil into the digestive juice chyme is defined as the bioaccessible fraction (4, 5). This fraction is considered to represent the maximum amount of contaminant available for intestinal absorption. Bioaccessible contaminants can subsequently be absorbed, that is, transported across the intestinal wall and transferred into the blood (or lymph) stream. The compounds may then undergo the first-pass effect in which they are biotransformed and excreted by the intestinal epithelium or liver. Finally, contaminants that are not metabolized can spread through the body by the systemic circulation, and may exert systemic toxicity. Hence, oral bioavailability of soil-borne contaminants is the combined result of soil ingestion, bioaccessibility, absorption, and the first-pass effect. When test animals ingest contaminants in a soil matrix, in general less absorption occurs and fewer toxic effects are observed than when they ingest the same contaminants in a food or liquid matrix (6-8). The difference can be ascribed to a lower oral bioavailability, resulting from a lower bioaccessibility for contaminants in soil than for contaminants in food or liquid. Because risk assessments are typically based on toxicity studies in which the matrix of administration is food or liquid, the health risks posed by contaminated soil may be substantially overestimated. A relative bioavailability factor that accounts for the difference in oral bioavailability caused by the matrix of ingestion would lead to more accurate risk assessment. Previous studies suggest that bioaccessibility, and therefore oral bioavailability of soil contaminants, depends on soil type (5, 9, 10) and contaminant (2, 11, 12). This implies that for health risk assessment, site-specific results of oral bioavailability may be required. However, it is not feasible to perform in vivo studies for every soil type and every contaminated site. In vitro digestion models based on human physiology have been developed as simple, cheap, and reproducible tools to investigate bioaccessibility of soil contaminants. In the simplest approach, the in vitro stomach model, mobilization of the contaminants from soil under gastric pH conditions is simulated. Most models are static gastrointestinal models, which simulate transit through the human digestive tract by sequential exposure of the soil to simulated mouth, gastric, and small intestinal conditions. Dynamic gastrointestinal models mimic the gradual transit of ingested compounds through the digestive tract. Whereas the more complex models can simulate more aspects of human physiology, the simple models are easy to perform and allow simultaneous determination of large 10.1021/es010204v CCC: $22.00
2002 American Chemical Society Published on Web 06/28/2002
FIGURE 1. Major processes in oral bioavailability. The fraction of contaminant reaching the internal exposure is considered to be bioavailable. numbers of samples. For example, stomach models simulate the acid conditions in the stomach. However, conditions in the intestines deviate from conditions in the stomach. Because absorption mainly occurs in the intestines (13, 14), stomach models may result in bioaccessibility values that are not representative for the site of absorption. If gastric pH conditions determine bioaccessibility, then the stomach model simulates bioaccessibility correctly. In vitro digestion models are likely to become increasingly important in risk assessment of polluted soils, especially for determining the urgency of remediation. At present there is no standard method of estimating bioaccessibility. Many digestion models exist with various experimental designs. Whether these models result in similar bioaccessibility values, however, remains unknown. In this study we compared and interpreted the bioaccessibility values obtained from five different in vitro digestion models. Such a comparison provides information on the dependence of bioaccessibility values on the test system. The comparison of in vitro digestion models may also help to identify the key factors influencing bioaccessibility. This information can contribute to the decisionmaking process on methods to study the bioaccessibility of contaminants in soil. The five digestion models that were compared include one static gastric model, three static gastrointestinal models, and one dynamic gastrointestinal model. All five models were designed to estimate the fraction of a soil contaminant that is available for absorption, thereby generating a comparable set of results. We report the experimental design of the models in relation to bioaccessibility data for the three contaminants lead (Pb), arsenic (As), and cadmium (Cd), of three different soils, that were tested using all five model systems. Although As is a metalloid, it will be referred to in this paper as a metal. These metals represent common soil contaminants with contamination levels frequently near or above regulatory Intervention Values (3), indicating serious soil contamination for which the urgency for remediation should be determined. To our knowledge, this manuscript is the first to describe a multi-laboratory comparison and evaluation of bioaccessibility values determined by in vitro digestion models. These particular digestion models are being used as part of a European platform on bioavailability and bioaccessibility, the BioAvailability Research Group Europe (BARGE). The ultimate aim of BARGE is to establish a methodology for estimating more realistic relative bioavailability factors to be used in general and site-specific risk assessments.
Materials and Methods Description of in Vitro Digestion Models. Five in vitro digestion models were compared. (1) The SBET method used by the British Geological Survey (BGS, United Kingdom) is a static gastric model. (2) The German DIN model applied by the Ruhr-Universita¨t Bochum (RUB, Germany), and (3) the digestion model of RIVM (The Netherlands) are static gastrointestinal models. Also, (4) a static gastrointestinal approach is used for the SHIME procedure (LabMET/Vito, Belgium). (5) The TIM method by TNO (The Netherlands) is a dynamic gastrointestinal model. All digestion models were performed according to their respective standard procedures; no changes in the methods were made for this study. However, this meant that the amount of soil per digestion varied between different digestion models, and that different procedures were used to determine contaminant concentrations in chyme and in soil. Below we describe some important parameters of the human gastrointestinal tract. The models are then described in detail in the text and schematically in Table 1. Human Gastrointestinal Tract. Information about the residence time and the pH in the various compartments of the human gastrointestinal tract is provided here (13) to interpret the design of the in vitro digestion models. The pH in the oral cavity is about 6.5. The residence time in the oral cavity is seconds to minutes. The compound (plus matrix) is swallowed through the esophagus into the stomach. The half-emptying time for the stomach is 8 to 15 min for fasting conditions, and 0.5 to 3 h for fed conditions. The gastric pH is between 1 and 2, and between 2 and 5 for fasting and fed conditions, respectively. The small intestine consists of three sections: duodenum, jejunum, and ileum. The residence time in the duodenum is 0.5 to 0.75 h, at a pH between 4 and 5.5. The residence time in the jejunum is 1.5 to 2 h, at a pH between 5.5 and 7.0. The residence time in the ileum is 5 to 7 h, at a pH between 7.0 and 7.5. The small intestine goes over into the colon, with a residence time between 15 and 60 h, at a pH generally between 6.0 and 7.5 (15). SBET (Simple Bioaccessibility Extraction Test), British Geological Survey (BGS), United Kingdom. The SBET method simulates mobilization of contaminants in the acid conditions of the stomach. An intestinal compartment is not employed. This model is adapted from a model described by Ruby et al. (4, 16), which was developed for Pb by the Solubility/ Bioavailability Research Consortium (SBRC) in the laboratory of John Drexler (University of Colorado, Boulder, CO). The VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3327
TABLE 1. Schematic Overview of the Different in Vitro Digestion Models
specific input general
stomach
TIM method (TNO) Nutrition, Netherlands
0.6 g dry soil
10 g dry soil
10 g dry soil
static stomach
temperature mechanical treatment
37 °C end-over-end rotation, 30 ( 2 rpm no
static gastrointestinal 37 °C agitator 200 rpm
static gastrointestinal 37 °C end-over-end rotation, about 55 rpm no
static gastrointestinal 37 °C mechanical stirring at 150 rpm
dynamic gastrointestinal 37 °C peristaltic movements
cream (18 g/L) and Nutrilon plus (15 g/L) in stomach compartment no
no
saliva compartment volume of saliva pH incubation time gastric compartment volume of gastric juice pH
no
yes (whole milk powder 50 g/L), and no no
yes
yes
9.0 mL
50 mL
6.5 5 min
5.0 5 min
yes
yes
yes
yes
yes
100 mL
100 mL
13.5 mL
25 mL
250 mL
1.5
2.0
1.1
4.0
1h
2h
2h
3h
pepsin, mucin
pepsin, mucin, BSA
yes
yes
pectin, mucin, cellobiose, proteose peptone, starch yes
initial gastric pH 5.0 decreasing to pH 3.5, 2.5, 2.0 after 30, 60, 90 min, respectively gradual secretion gastric content at 0.5 mL/min (Figure 2) lipase, pepsin
volume of intestinal juice pH
100 mL
incubation time
6h
duodenal juice: 27 mL bile juice: 9 mL duodenal juice: 7.8 bile juice: 8 pH chyme mixture: 5.5 2h
0 g/L
4.5 g/L
0.9 g/L
0
porcine 1.0 mM approximately 0.15 M trypsin, pancreatine
intestinal compartment
concn of bile in chyme origin bile concn of phosphate in chyme ionic strength chyme other intestinal secretion components
9
SHIME method (LabMET/Vito), Belgium
2.0 g dry soil
gastric secretion components
3328
in vitro digestion model (RIVM), Netherlands
1.0 g dry soil
incubation time
intestine
DIN method (RUB), Germany
amount of soil added type model
food components
oral cavity
SBET method (BGS), UK
no
pH intestinal mixture 7.5
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
pancreatic fluid: 15 mL
yes (3 sections: duodenum, jejunum and ileum) 3 × 70 mL
6.5
duodenum: 6.5 jejunum: 6.8 ileum: 7.2
5h
1.5 g/L
duodenal secretion at 1 mL/min (Figure 2); total digestion time 360 min variable
bovine 2.6 mM
bovine none added
porcine variable
0.14 M
not determined
not determined
pancreatine, lipase, BSA
pancreatine
pancreatine
TABLE 1 (Continued)
specific
SBET method (BGS), UK
output centrifugation no filtration 0.45-µm cellulose acetate disk filter special pH filtrate within treatment 0.5 pH units of the starting pH
destruction
analytical method
DIN method (RUB), Germany 7000g no
in vitro digestion model (RIVM), Netherlands 3000g no
SHIME method (LabMET/Vito), Belgium 7000g no
TIM method (TNO) Nutrition, Netherlands no see special treatment hollow fiber membrane for determination bioaccessibility
supernatant determination decanted, metal in pellet stirred starting material, in 30 mL water, i.e., soil, chyme, recentrifuged, and pellet for and supernatant mass balance decanted again; decanted volumes combined 1 mL conc HF to 0.4 g soil/pellet or 1 g soil or 30-50 g microwave 3 g soil/6 mL about 0.1 g soil, 10 mL chyme in dialysate + HNO3/HBF4/ milliQ, 24 h, 0.8 mL 6 mL HNO3 H3PO4/ 4 mL HNO3 + 1 mL 65% conc HNO3 + HCl/HF (65%) + 2 mL (65%) + 12 mL HNO3, 0.4 mL conc HClO3, H2O2 (30%) microwave HCl (25%) cooked for 2.5 h 5 h at 100 °C and for 5 h at 200 °C 7 h at 190 °C; taken up in 5% HNO3 ICPAES AAS ICPMS ICPAES ICPAES (Cd, Pb) and HAAS (As)
method by Ruby et al. has been validated to the in vivo situation with swine studies (5). Digestion is performed as follows: 100 ( 0.5 mL of extraction fluid is added to 1.0 ( 0.5 g of dry soil. The extraction fluid is 0.4 M glycine adjusted to pH 1.5 with concentrated HCl. This mixture is rotated end-over-end at 37 °C at 30 ( 2 rpm for 1 h. Samples are transferred into a disposable syringe and filtered through a 0.45-µm cellulose acetate disk filter. The pH of the filtrate should be within 0.5 pH units of the starting pH, otherwise the procedure is repeated. The samples are stored in a refrigerator at 4 °C until analysis. Analysis must be performed within one week after in vitro digestion. Concentrations of As, Pb, and Cd in the filtrate are determined by ICPAES. The starting soil material is digested in HF by transferring 0.1 g of soil into a Teflon tube, to which 1 mL of concentrated HF is added. This mixture is left overnight at room temperature. Subsequently, 0.8 mL of concentrated HNO3 and 0.4 mL of concentrated HClO3 are added to the mixture. The mixture is heated for 5 h at 100 °C followed by 7 h at 190 °C. Finally, the solution is taken up in 5% HNO3 and analyzed by ICPAES using matrix matched standards. Method E DIN 19738, Ruhr-Universita¨t Bochum (RUB), Germany. The German method E DIN 19738 has its origin in the in vitro digestion models by Rotard et al. (17) and Hack and Selenka (10, 18). It is a static in vitro gastrointestinal model using synthetic digestive juices. Validation studies with mini-pigs are being performed and will be published in the future. Because it is assumed that saliva has only a negligible effect on the level of mobilization of contaminants from soil, only synthetic gastric and synthetic intestinal juices were used in the present round robin. Whole milk powder (50 g/L) may be added to the test system to simulate the influence of food on the mobilization of contaminants. The method in which milk powder is added is called “DIN”, and the method in which no milk powder is added is called “DINWM”. The digestion is started by suspending 2.0 g of dry soil in 100 mL of gastric juice for 2 h. The pH of the gastric juice is kept at 2.0. This is followed by addition of 100 mL of
intestinal juice, the pH is set to 7.5, and digestion proceeds for 6 h. The temperature is controlled by means of a water bath (37 °C). Agitation is performed at 200 rpm. The digestion mixture is centrifuged for 10 minutes at 7000g, after which the supernatant is decanted. The pellet is stirred in 30 mL of distilled water for 0.2 h, the mixture is recentrifuged, and the supernatant is decanted again. The decanted chyme solutions are combined. For determination of the total element contents of starting material, i.e., soil, a 0.4-g portion is extracted with a mixture of 6 mL of HNO3 (65%) and 2 mL of H2O2 (30%). The same mixture is used for total element determination in 10 mL of chyme. Destruction takes 5 h at a temperature of 200 °C. Analysis was performed by AAS. In Vitro Digestion Model, National Institute of Public Health and the Environment (RIVM), The Netherlands. The model by RIVM is also based on the in vitro digestion model by Rotard et al. (17). The in vitro digestion model by RIVM is a static in vitro gastrointestinal model; that is, the soil is subjected to a number of stages simulating the human digestion process. The digestion is started by addition of 9.0 mL of saliva of pH 6.5 ( 0.2 to 0.6 g of dry matter soil. This mixture is rotated for 5 min, end-over-end, at about 55 rpm at 37 °C. Then, 13.5 mL of gastric juice (pH 1.07 ( 0.07) is added, and the mixture is rotated at 37 °C. After 2 h, 27 mL of duodenal juice (pH 7.8 ( 0.2) and 9 mL of bile juice (pH 8.0 ( 0.2) are added. This mixture is rotated at 37 °C for another 2 h and subsequently centrifuged at 3000g for 5 min. The supernatant (total volume 58.5 mL) represents the chyme and has a pH of at least 5.5. For determination of the Pb, As, and Cd concentrations in chyme, the chyme is diluted 10-fold by 0.1 M HNO3, and analyzed by ICPMS. Aliquots of the soil and the whole pellet (the digested soil) are destructed by addition of 65% HNO3 and heated under pressure by microwave. These samples are centrifuged, diluted, and analyzed by ICPMS. A mass balance is obtained, as the contaminant concentration in chyme, pellet, and soil, are determined. SHIME (Simulator of Human Intestinal Microbial Ecosystems of Infants), LabMET (RUG)/VITO, Belgium. The VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3329
TABLE 2. Characteristics of Flanders, Oker 11, and Montana 2711 Soils characteristics sampling site size fractions organic carbon (%) organic matter (%) pH description
Flanders 8.0%
