Quantcast
Last updated on April 20, 2014 at 8:28 EDT

Use of In Vitro Absorption, Distribution, Metabolism, and Excretion (ADME) Data in Bioaccumulation Assessments for Fish

December 20, 2007

By Nichols, John Erhardt, Susan; Dyer, Scott; James, Margaret; Moore, Margo; Plotzke, Kathleen; Segner, Helmut; Schultz, Irvin; Thomas, Karluss; Vasiluk, Luba; Weisbrod, Annie

ABSTRACT A scientific workshop was held in 2006 to discuss the use of in vitro Absorption, Distribution, Metabolism, and Excretion (ADME) data in chemical bioaccumulation assessments for fish. Computer-based (in silico) modeling tools are widely used to estimate chemical bioaccumulation. These in silico methods have inherent limitations that result in inaccurate estimates for many compounds. Based on a review of the science, workshop participants concluded that two factors, absorption and metabolism, represent the greatest sources of uncertainty in current bioaccumulation models. Both factors can be investigated experimentally using in vitro test systems. A variety of abiotic and biotic systems have been used to predict chemical accumulation by invertebrates, and dietary absorption of drugs and xenobiotics by mammals. Research is needed to determine whether these or similar methods can be used to better predict chemical absorption across the gills and gut of fish. Scientists studying mammals have developed a stepwise approach to extrapolate in vitro hepatic metabolism data to the whole animal. A series of demonstration projects was proposed to investigate the utility of these in vitro-in vivo extrapolation procedures in bioaccumulation assessments for fish and delineate the applicability domain of different in vitro test systems. Anticipating research progress on these topics, participants developed a “decision tree” to show how in vitro information for individual compounds could be used in a tiered approach to improve bioaccumulation assessments for fish and inform the possible need for whole-animal testing.

Key Words: fish, bioaccumulation, bioconcentration, metabolism, biotransformation, absorption.

LIST OF ABBREVIATIONS

ADME Absorption, Distribution, Metabolism, Excretion

BAF Bioaccumulation Factor

BCF Bioconcentration Factor

CEPA Canadian Environmental Protection Act

CDSL Canadian Domestic Substances List

D^sub 7.4^ n-octanol/phosphate buffer distribution coefficient at a pH of 7.4

GIT Gastrointestinal Tract

ILSI-HESI International Life Sciences Institute-Health and Environmental Sciences Institute

K^sub ow^ n-octanol/water partition coefficient

OECD Organization for Economic Cooperation and Development

PAMPA Parallel Artificial Membrane Permeability Assay

PBiT Persistent, Bioaccumulative, and inherently Toxic substance

POP Persistent Organic Pollutant

QSAR Quantitative Structure-Activity Relationship

REACH Registration, Evaluation, and Authorization of Chemicals Program (European Union)

UNEP United Nations Environment Program

USEPA U.S. Environmental Protection Agency

INTRODUCTION

The accumulation of xenobiotics in fish and other aquatic biota is an issue of long-standing concern to industry, government regulators, the academic community, and the general public. Extensive research has been conducted to understand the chemical and biological processes that promote bioaccumulation, and detailed information is available for a small number of compounds, several of which are now banned from production and use. Increasingly, however, there is a need to perform less intensive assessments for a much larger number of compounds. Regulatory programs in Europe and North America are being revised to support the 2004 enactment of the United Nations Stockholm Convention (also known as the Persistent Organic Pollutants (POPs) Protocol), which governs the assessment, use, trade, release, and replacement of all persistent (P), bioaccumulative (B), and inherendy toxic substances (iT), or PBiTs (UNEP 2006). For example, the Canadian Environmental Protection Act (CEPA) required the Ministers of Environment and Health to categorize the hazard of approximately 23,000 chemicals on a Domestic Substances List (CDSL) and, as necessary, conduct screening level assessments to determine whether they are “… toxic or capable of becoming toxic to the environment or human health” (Government of Canada 1999; the word “toxic” is defined in Part 5, Section 64 of the Act). Legislation in Europe (Registration, Evaluation, and Authorization of Chemicals program; REACH) could result in similar reviews of tens of thousands of compounds (Rogers 2003).

In most cases, these reviews are conducted in the absence of measured bioaccumulation data. Moreover, because of ethical concerns, many government agencies and animal welfare organizations are advocating large reductions in vertebrate testing, including testing with fish. These considerations suggest a need for alternative methods to assess the potential for chemicals to accumulate in fish. One method that is receiving considerable attention involves the use of in vitro test systems, alone or in combination with mathematical models.

This report describes the results of a workshop held March 3-4, 2006, in San Thego, California, USA. Workshop participants were asked to review the state-of-the-science regarding the incorporation of in vitro Absorption, Distribution, Metabolism, and Excretion (ADME) information into bioaccumulation assessments for fish, and identify research needed to expand the utility and applications of this approach. An important outcome of this workshop was a proposal to conduct research on hepatic biotransformation in fish, with the goal of relating in vitro metabolic rate, in vivo metabolic rate, and measured levels of accumulation for a set of strategically selected compounds. Participants also discussed how in vitro data could be used in a tiered approach for bioaccumulation assessments. Based on this discussion, a “decision tree” was proposed to identify information required at each tier in the assessment process and provide guidance on the need for whole animal testing.

The scope of the workshop was limited to consideration of in vitro methods that could be used to predict chemical accumulation in fish under standardized laboratory conditions. Participants recognized that a large number of factors may complicate efforts to predict accumulation in a natural setting, including the distribution of chemicals among environmental compartments, food web structure and function, and seasonal movements of animals. The extrapolation of bioaccumulation predictions from the lab to the field, and among different environmental settings, was identified as an important topic for future scientific workshops.

Defining Bioconcentration and Bioaccumulation

As applied to fish, the term bioconcentration refers to chemical accumulation that occurs in a waterborne exposure due to uptake across the gills and skin, whereas the term bioaccumulation refers to chemical accumulation resulting from all possible routes of exposure, including dietary uptake. Bioconcentration is generally measured in controlled laboratory exposures, whereas bioaccumulation is typically characterized by measuring chemical concentrations in field-collected animals. The extent of bioconcentration may be expressed by calculating a bioconcentration factor (BCF; L/kg), which is the total chemical concentration in the animal (mg/kg) divided by that in water (mg/L). A bioaccumulation factor (BAF) with the same units can be developed to describe the extent of bioaccumulation. Alternatively, bioaccumulation may be referenced to the chemical concentration in sediment, resulting in a biota- sediment accumulation factor (BSAF). Unless otherwise indicated, BCFs, BAFs, and BSAFs represent the extent of accumulation that would be expected in a long-term exposure (i.e., under steady-state conditions). These ratios are often normalized to the lipid content offish and the freely dissolved chemical concentration in water (or, in the case of the BSAF, the lipid content offish and the organic carbon content of sediment). The goal of this normalization is to account for differences in factors that control the uptake and accumulation of hydrophobic organic chemicals such as binding to organic material in water or sediment, and partitioning to tissue lipid. Bioaccumulation is the condition that results from a natural exposure, particularly for hydrophobic substances; however, BAFs and BSAFs are difficult to measure experimentally. As a result, regulators often use measured or modeled BCFs to estimate the potential for a compound to bioaccumulate, and legislated criteria for “bioaccumulation” are generally expressed as BCF values (Arnot and Gobas 2006).

The major processes that determine the extent of chemical accumulation in fish are illustrated in Figure 1. Uptake processes include chemical absorption across the gills, skin, and gut. Loss processes include chemical efflux across the gills, skin, and gut; urinary and biliary elimination; and biotransformation. Growth affects the measured concentration of a chemical by increasing the tissue mass into which it is diluted. Additional processes are responsible for the internal distribution of a chemical. Among these are blood flow rates to individual tissues and tissue-specific differences in lipid content.

Current Methods to Estimate Bioaccumulation in Fish Three techniques are currendy employed to assess the potential for a chemical to accumulate in fish: controlled exposures offish to test chemicals in the water or thet, measurement of chemical residues in field-collected animals, and computational modeling.

Measured in vivo accumulation data for fish are relatively scarce. For example, Arnot and Gobas (2006) reported that measured BCFs and BAFs are available for

Bioaccumulation assessments for many compounds are conducted using computer-based modeling tools. Quantitative Structure Activity Relationships (QSPRs) are widely used to predict the BCF for a compound based on its n-octanol/water partition coefficient, K^sub ow^ (Veith et al. 1979). Compartmental models have been used to predict BCFs from measured or estimated uptake and elimination rate constants (Branson et al. 1975). Additional models have been developed to predict bioaccumulation in aquatic food webs occurring within a natural environment (Thomann et al. 1992; Gobas 1993). These approaches are described briefly in another ILSI-HESI workshop publication (Weisbrod et al. 2007). Generally, models of chemical accumulation in fish have been developed and evaluated using data for chemicals that passively diffuse across biological membranes, partition non-specifically to tissue lipid, and undergo little or no metabolism. These models may provide inaccurate estimates of accumulation for compounds that exhibit more complex behaviors.

RESEARCH TO SUPPORT THE INCORPORATION OF IN VITRO DATA INTO BIOACCUMULATION MODELS

Workshop participants concluded that two factors, absorption and metabolism, represent the greatest sources of uncertainty in current models of chemical accumulation by fish. Although empirical relationships exist to predict branchial (McKim et al. 1985) and dietary uptake (Gobas et al. 1988) of chemicals as functions of K^sub ow^, there is considerable scatter in the data used to develop these relationships, and limited knowledge of uptake mechanisms other than simple diffusion.

Metabolism may substantially reduce the extent to which a compound accumulates in fish, and was recognized as the single most important source of modeling uncertainty. This conclusion was based on known variation in metabolism rates among species and chemicals, as well as the sensitivity of models to changes in the value of this parameter (Clark et al. 1990; Dimitrov et al. 2005). In vitro methods have been developed to predict absorption and metabolism in a variety of biological systems. The following sections describe these methods, emphasizing their current or potential application to bioaccumulation assessments for fish.

Absorption

Chemical absorption by fish may be viewed as a process that depends on both chemical speciation and membrane permeatic. Most compounds cross biological membranes by passive diffusion. In general, ionized and bound chemical species exhibit low rates of diffusion. Neutral, freely dissolved forms may exhibit high rates of diffusion, depending on other physicochemical properties (see later). Membrane transport proteins may facilitate the absorption of some compounds, resulting in uptake rates much greater than those expected from simple diffusion. Alternatively, efflux transporters may promote the elimination of a compound, reducing the extent to which it accumulates. Both types of transporters are active in the gastrointestinal tract (GIT) offish. Biotransformation at the site of absorption tends to limit the rate of uptake by consuming chemical that diffuses into the tissue.

Physicochemical properties that may limit membrane diffusion

Physicochemical properties of a compound may determine the extent of passive diffusion across biological membranes. For example, Lipinski et al. (1997) found that four physicochemical parameter ranges were associated with 90% of all orally administered drugs. Drugs that were considered bioavailable had a molecular weight =500, a log K^sub ow^ value =5, =5 hydrogen (H)-bond donors, and =10 H-bond acceptors. Wenlock et al. (2003) reviewed data for several hundred additional oral drugs and found that 90% had a molecular weight

Previous work has shown that organochlorines with a molecular weight >600 or molecular volume >3 nm^sup 3^ are poorly absorbed across the GIT of rainbow trout (Niimi and Oliver 1988). Other researchers have suggested that molecular shape and flexibility are important determinants of chemical uptake and accumulation by fish (Opperhuizen et al. 1985; Dimitrov et al. 2002). After reviewing these and other stuthes, de Wolf et al. (2007) recommended several guidelines based on molecular weight, lengtii, cross-sectional diameter, and lipophilicity to screen chemicals for their potential to bioconcentrate in fish. The value of this approach is that it may be possible to identify compounds which, based on their physicochemical properties, can be eliminated from further consideration in bioaccumulation assessment efforts. The authors cautioned, however, that there were no clear cut-off values for chemical absorption by fish.

General considerations for assay selection

The standard test for chemical absorption is to measure the extent of uptake by experimental animals in a controlled laboratory exposure. Such tests are impractical, however, for screening a large number of compounds. Alternatively, it may be possible to conduct screening-level assessments of absorption using one or more in vitro assays.

A variety of in vitro test systems have been used to predict in vivo absorption. These are listed in Table 1, along with their advantages and limitations, and the type of data that they generate. For a particular test system and compound, bioavailability (defined here as the amount of compound present as a neutral, diffusing form) or membrane transport may constitute the principal rate limitation on membrane flux. The selection of an appropriate in vitro test system should therefore reflect knowledge of the predominant route of exposure, chemical speciation, and predicted modes of uptake.

Because chemical speciation has a large impact on absorption, experimental factors that influence speciation should be controlled and measured. These include pH, alkalinity, hardness, temperature, dissolved organic carbon, and particulate organic carbon. Membrane bilayer structure also can influence the rate of passive diffusion (Frezard and Garnier-Suillerot 1998). In fish and other poikilotherms, membrane bilayer composition may change with season due to changes in water temperature. These changes have been shown to influence the absorption of lipophilic chemicals (Ehkus et al. 2005).

Degradation of a compound in the external medium (water, gut contents) as well as metabolism within tissues of the gill or gut may modify uptake (Van Veld et al. 1988; Barron et al. 1989). In some cases it may be difficult to determine whether the rate of absorption is controlled by the rate of diffusion across the membrane or the rate of loss of parent chemical. Hence, appropriate controls must be included to account for the potential contribution of metabolism.

In vitro systems to predict xenobiotic absorption

Various partitioning systems (e.g., n-octanol and water) have been used to characterize the relative hydrophobicity of a compound, which is equivalent in these systems to lipophilicity. Because biological membranes consist mostly of lipid, measures of hydrophobicity often correlate with the rate of membrane diffusion. In general, diffusion rates are greatest for moderately hydrophobic chemicals and decline substantially for chemicals that exhibit extreme hydrophilic or hydrophobic character.

Using both liquid and solid phase partitioning devices, researchers have characterized the bioavailability of hydrophobic compounds in water and sediment, and predicted chemical accumulation in invertebrates (Sabaliunas and Sodergren 1996; Mayer et al. 2000; Wilcockson and Gobas 2001; Cornelissen et al. 2001). In this application, the liquid or solid phase may be viewed as a surrogate for tissue lipid. In principle, biomimetic partitioning devices could be used to predict chemical accumulation in fish under a particular set of environmental conditions, but this usage remains largely unexplored.

Artificial lipid bilayers have been used extensively to predict the membrane transport of drugs (Parallel Artificial Membrane Permeability Assay (PAMPA); Ransy et al. 1998). For many drugs, values obtained using PAMPA exhibit good agreement with those obtained using cell-based systems such as Caco-2 cells (Bermejo et al. 2004). Kwon et al. (2006) used PAMPA to estimate passive uptake and elimination rates for 23 simple aromatic compounds in small fish. In general, the experimentally determined values exhibited good correspondence with literature values; however, there were some exceptions for which the in vitro system did not accurately predict measured in vivo rates. The principal drawback of physical systems such as artificial membranes and partitioning devices is that they do not account for complex membrane processes such as active transport and metabolism. Chemical uptake from the diet represents a special case because conditions within the GIT (low pH, gut microflora, and the presence of digestive enzymes) may radically alter speciation and bioavailability. “Physiologically based” in vitro extraction systems that duplicate these conditions have been developed to predict dietary uptake of chemicals by mammals (Hack and Selenka 1996; Van de Wiele et al. 2004; Ruby 2004). No such systems have been developed for fish.

Caco-2 cells, alone or in combination with physiologically based extraction methods, have been used to predict the dietary uptake of drugs (Hidalgo 2001; van Breemen and Li 2005) and environmental contaminants (Oomen et al. 2001; Buesen et al. 2003; Vasiluk et al. 2005; Minhas et al. 2006) by mammals. Caco-2 cells exhibit morphological characteristics of small intestinal cells and express most of the enzymes and transporters that exist in vivo. In vitro assays are performed using a cell monolayer that can be scaled to predict in vivo permeability. It is not known whether Caco-2 cells or a cell type derived from fish tissue could be used to predict chemical uptake across the gills or gut of fish. Enterocytes from fish have been used to examine ion transport and uptake of fatty acids (Schoenmakers et al. 1993; Perez et al. 1999), whereas cultured gill epithelial cells have been used to study ion regulation and branchial excretion of ammonia (Wood et al. 2002). To date, however, neither of these cell-based systems has been used to study the absorption of environmental contaminants.

Research is needed to develop in vitro assays that can be used to better predict chemical absorption by fish. Partitioning-based approaches provide useful information related to bioavailability and passive diffusion but are unlikely to deliver data needed to predict the accumulation of chemicals that undergo metabolism in the gills or gut. The successful use of cell-based systems to predict dietary uptake of chemicals in mammals provides an important example for future research with fish. This work should be conducted using piscine cell lines because taxonomic differences in membrane transporters, enzyme complement, and fatty acid composition could influence the absorption process.

Metabolism

For many xenobiotics, metabolism information is required to accurately predict in vivo bioaccumulation. The major questions that must be addressed are whether a compound is metabolized at all, and if so at what rate. At an early stage of the assessment process, a yes/no answer concerning the likelihood for metabolism may be sufficient. Methods to derive this information include QSAR in silico models (Borodina et al. 2003) and knowledge-based expert systems such as METEOR (Balmat et al. 2005), OASIS (Mekenyan et al. 2006), and META (Klopman et al. 1994; Talafous et al. 1994).

If the information required for a particular assessment goes beyond a yes/no decision, then quantitative estimates of biotransformation may be needed. Methods used in studies with mammals to extrapolate in vitro hepatic clearance data to the whole animal are well described and widely accepted (Rane et al. 1977; Wilkinson 1987; Houston 1994; Houston and Carlile 1997; Iwatsubo et al. 1997). It has been proposed that these or similar methods could be used to extrapolate in vitro hepatic metabolism data for fish, and that clearance constants calculated from this information could serve as inputs to bioconcentration and bioaccumulation models (Nichols et al. 2006). Each of the major in vitro hepatic metabolism assays has strengths and limitations. The selection of a “preferred” assay system relates in part to the questions that are being asked, ease of preparation, and metabolic pathways of concern. The characteristics of each in vitro hepatic metabolism assay as well as their advantages and limitations are summarized in Table 2.

Purified enzymes and recombinant enzyme systems

Purified enzymes and recombinant enzyme systems provide information for a specific pathway of interest (Dierickx 1985; Buhler and Wang-Buhler 1998; Leaver and George 1998; Tong and James 2000; Sugahara et al. 2003; Chung et al. 2004). From the perspective of predicting effects of metabolism on bioaccumulation, this focus on a single pathway may be disadvantageous if other (uncharacterized) pathways contribute substantially to elimination. Additional disadvantages include the technical challenges of preparing these systems and difficulty relating biotransformation data to the whole animal in the absence of information concerning tissue enzyme content.

Subcellular preparations

Most of the in vitro metabolism data for fish have been collected using subcellular S9 or microsomal fractions (Fitzsimmons et al. 2007). Subcellular preparations are amenable to cryopreservation and can be frozen for up to 1 year at -80[degrees]C while maintaining initial levels of enzymatic activity (Hodson et al. 1991). The appropriateness of a particular preparation depends on whether it contains the enzymes required to metabolize the chemical of interest. Co-factors required for Phase I and II metabolic activity must be supplied as appropriate to a given system and reaction pathway.

Isolated and cultured primary liver cells

Isolated hepatocytes are obtained by enzymatic treatment of liver tissue, followed by mechanical dissociation of cells (Mommsen et al. 1994; Segner 1998a). In comparison to subcellular fractions, intact hepatocytes provide greater biological realism because membrane transport mechanisms and intracellular compartments are maintained. In some cases, diffusion across the hepatocyte membrane may limit the rate at which a chemical is metabolized (Lu et al. 2006), resulting in a lower level of activity than that observed in subcellular fractions. For other compounds, transport proteins located within the membrane (e.g., p-GP, OATp) may contribute to high rates of hepatic clearance (Sturm and Segner 2005). In stuthes with mammals, hepatic clearance rates calculated from freshly isolated hepatocytes tend to correlate with rates obtained using liver microsomes, but often provide more accurate predictions of in vivo clearance (Houston 1994; Houston and Carlile 1997; Jones and Houston 2004).

Cryopreserved hepatocytes can be purchased from commercial sources for several mammalian species, but are not yet available for fish. The use of primary hepatocytes therefore requires access to live animals. A significant problem when using primary hepatocytes from fish is the variability among preparations with respect to cell viability and metabolic capability. This variability can be accounted for, however, by using benchmark compounds with known metabolic characteristics to characterize each cell preparation (Segner 1998a).

Immortalized cell lines from fish liver tissue

Immortalized fish cell lines derived from liver tissue have been used extensively to study cytotoxicity and induction of metabolic enzymes. The primary advantages of these systems are their availability and the fact that they do not require the use of animals. The principal disadvantage is that metabolic pathways and rates may not reflect the biotransformation capabilities of tissues from which these cell lines originated (Nehls and Segner 2001; Castano et al. 2003). In general, metabolism rates in fish cell lines are lower than those in primary liver cells (Segner 1998b; Dyer et al. 2003; Dyer and Bernhard 2004).

Precision-cut liver slices

A liver slice represents the complete cellular environment for hepatic biotransformation. Procedures used to obtain precision-cut liver slices from fish have been described (Kane and Thohan 1996; Gilroy et al. 1996; Singh et al. 1996), but the use of slices in metabolism studies with fish has been highly limited. A potential disadvantage of liver slices is that diffusion limitations on chemical flux can create internal concentration gradients, reducing the rate of metabolism and complicating in vitro-in vivo extrapolation efforts (Worboys et al. 1996a,b).

Assay conditions

A number of factors may complicate attempts to use in vitro metabolism data in bioaccumulation assessments for fish. Included among these are factors that relate to the collection of in vitro data, as well as the application of this data to diverse chemicals and species living within a complex natural environment. Here we single out two factors for special consideration: in vitro bioavailability and temperature.

Chemicals that tend to accumulate in fish often exhibit physicochemical characteristics (e.g., high log K^sub ow^ values) that make them difficult to work with. Chemical adsorption to the reaction vessel and/or binding to macromolecules within the assay system may substantially reduce the free concentration of a chemical in solution. The interpretation of in vitro test data may be improved by expressing the results in relation to the freely dissolved concentration of the test compound. Established methods for measuring in vitro binding include equilibrium dialysis (Obach 1997), ultracentrifugation (Tang et al. 2002), and solid phase microextraction (Heringa et al. 2004). Empirical models have been developed to estimate free concentrations in vitro (Austin et al. 2002; Gulden et al. 2005), but these must be re-evaluated when using biological material from a new source.

The maximum in vitro activity (V^sub max^) of metabolic enzymes from fish acclimated to a particular temperature generally doubles with a 10[degrees]C increase in temperature (Fitzsimmons et al. 2007). Changes in acclimation temperature may be accompanied, however, by ideal (or near ideal) temperature compensation. The result of this compensation is that activities of metabolic enzymes measured at the temperature to which individuals are acclimated tend to be similar. To facilitate extrapolation of in vitro biotransformation data to the whole animal, it is recommended that all metabolism studies be performed at or near the acclimation temperature of the animal. In vitro-in vivo extrapolation of hepatic metabolism

Scientists using mammals have developed a stepwise approach to predict in vivo hepatic clearance from measurements of in vitro hepatic metabolism (Rane et al. 1977; Wilkinson 1987; Houston 1994; Houston and Carlile 1997; Iwatsubo et al. 1997). Nichols et al. (2006) reviewed this procedure and suggested ways that it could be adapted to studies with fish. Briefly the steps in this procedure are: (1) calculate in vitro intrinsic clearance (CL^sub in vitro, int^; [mu]L/min/mg microsomal protein) from the ratio V^sub max^/ K^sub m^, or the rate of decline of parent chemical concentration in a substrate depletion experiment; (2) employ extrapolation factors to calculate in vivo intrinsic clearance normalized to fish body weight (CL^sub in vivo,int^; L/d/kg); and (3) convert CL^sub in vivo,int^ to in vivo hepatic clearance (CL^sub in vivo,int^; L/d/ kg) using a liver model that interrelates biotransformation, liver blood flow rate, and the effect of binding to proteins and other macromolecules.

The simplest liver model is termed the venous equilibration model, which can be expressed as (Rowland et al. 1973; Wilkinson and Shand 1975):

where Q^sub liver^ (L/d/kg) is the liver blood flow rate and f^sub u^ (unitless) is a term that corrects for the binding of a chemical in blood and in the in vitro system used to assess metabolism.

Bioaccumulation models for fish are generally referenced to the whole-body concentration of a chemical. The incorporation of CL^sub h^ into this type of model requires that CL^sub h^ be divided by an apparent volume of distribution (V^sub d,blood^; L/kg), referenced to the total chemical concentration in blood. The result of these calculations is a whole-animal clearance rate constant with units of inverse time (k^sub met^; 1/d).

An examination of this procedure suggests the need for several types of information specific to the fish species tested. First, one or more extrapolation factors are required to convert CL^sub in vitro,int^ to CL^sub in vivo,int^, depending on the type of in vitro system used to characterize enzyme activity. Calculation of CL^sub h^ also requires knowledge of liver blood flow. Extrapolation factors and liver blood flow estimates have been published for trout (Nichols et al. 2006). These terms are not yet available for most other fish species. Empirical relationships based in part on measured binding of chemicals in trout plasma may be used to estimate f^sub u^ (Nichols et al. 2006); however, the role that binding plays in limiting the metabolism of hydrophobic environmental contaminants is poorly known (see later).

DEMONSTRATION PROJECT PROPOSAL FOR METABOLISM EXTRAPOLATIONS

A key finding of this workshop was the need to evaluate the relationship between in vitro hepatic metabolism rate, in vivo metabolic rate, and measured bioconcentration/bioaccumulation. The goal of this effort is to determine the practical limits of in vitro- in vivo extrapolation procedures across a range of chemicals and metabolic pathways. To this end, workshop participants developed specific recommendations for a proposed set of demonstration projects.

In Vitro-In Vivo Experimental Design

Procedures used to extrapolate metabolism information should be evaluated using a linked in vitro-in vivo experimental design. In each case, it is necessary to use the same chemical and fish species for both in vitro and in vivo experiments. Because metabolism within a species can vary with gender, season, and life-stage, these factors should be controlled. Most of the existing BCF data has been obtained using rainbow trout and carp (Arnot and Gobas 2006). These species provide sufficient liver tissue for the preparation of subcellular fractions and isolated hepatocytes, and are therefore recommended for initial testing efforts.

Important experimental issues include the selection of an in vitro assay, selection of a higher order system (in vivo or in situ) to evaluate the accuracy of extrapolated in vitro values, and the need to collect appropriate supporting data. A subcellular fraction is recommended for most applications due to ease of preparation. An S9 or microsomal system is appropriate if the rate of hepatic clearance is limited by Phase I metabolic activity. An S9 fraction is preferred if clearance is limited by Phase II activity. Isolated hepatocytes would constitute the system of choice if chemical diffusion across the hepatocyte membrane limited the rate metabolic clearance.

In Situ and In Vivo Experimental Approaches

Two general approaches exist to evaluate the accuracy of extrapolated in vitro metabolism values. The first is to measure hepatic clearance directly using an isolated perfused liver or one of several in vivo experimental approaches. The second is to estimate in vivo whole-animal clearance based on either the disappearance of parent chemical or appearance of metabolic products.

Isolated perfused liver

Liver perfusion studies integrate the two major factors that determine hepatic clearance in viva enzyme activity and perfusion limitations on the rate of chemical delivery to the site of metabolism (Forlin and Andersson 1981; Andersson et al. 1983). The main advantage of this technique is that measured concentrations of parent chemical in the perfusion buffer provide a direct estimate of hepatic clearance. These studies are technically demanding, however, and are best suited to collecting detailed information for a small number of animals and compounds. The need to use relatively large fish also limits this technique to laboratories with adequate fish holding facilities.

In vivo measurement of hepatic clearance

Researchers working with mammals have developed methods to estimate hepatic clearance by kinetic analysis of plasma data from bolus i.v. injection, continuous i.v. infusion (Iwatsubo et al. 1997), or intraportal infusion (Carlile et al. 1997) stuthes. The applicability of these methods to stuthes with fish is largely unexplored.

In vivo measurement of whole-body clearance

The protocol most commonly used to estimate in vivo metabolism rates in fish involves dosing animals with a chemical for a period of time and then allowing them to depurate in clean water. The rate of metabolism is calculated as the difference between the observed rate of elimination of parent chemical and the rate of elimination expected due to non-metabolic padiways (Sijm and Opperhuizen 1988; de Wolf et al. 1993; Fisk et al. 1998). Accurate determinations of metabolism rate are difficult to make if the total rate of elimination is very slow, or the rate of metabolism is slow relative to that of non-metabolic elimination pathways. Exposures can be conducted by exposing fish to chemicals in water, although the dietary route may be better suited for compounds that possess low water solubility (Fisk et al. 1998; Wong et al. 2002).

Whole-animal metabolism rates also can be estimated by analysis of measured whole-body chemical concentrations resulting from defined laboratory or field exposures (Arnot and Gobas 2003; Dimitrov et al. 2005). Using an appropriate mathematical model, the apparent rate of metabolism is determined by adjusting a metabolism rate constant until model simulations predict the observed level of accumulation. This approach requires high-quality BCF (or BAF) data, and is highly dependent on the accuracy of other model parameters (in particular, the rate of elimination that would be expected in the absence of metabolism). The main advantage of this approach is that metabolism rates may be estimated using existing test data.

In a small number of studies, investigators have estimated whole- animal rates of metabolism by measuring the appearance of metabolic products in fish and/or the exposure water (Karara and Hayton 1984; Barron et al. 1990; Bradbury et al. 1993). This approach provides a direct measure of metabolism but is technically demanding. The utility of metabolite data also depends on whether all major metabolic pathways are accounted for. Otherwise, an incomplete accounting of metabolic products could result in underestimation of the true rate of parent chemical metabolism.

Supporting Data

Supporting data are required to maximize the utility of in vitro data, in metabolism extrapolations for fish. As indicated earlier, the model used to translate CL^sub in vitro,int^ to an estimate of CL^sub h^ (Eq. [1]) contains a term (f^sub u^) that accounts for binding of chemicals in blood and in the in vitro system used to assess metabolism. The role that binding plays in limiting hepatic metabolism of hydrophobic compounds (log K^sub ow^ > 4) is poorly understood. Although theoretical approaches suggest that these compounds are unlikely to be metabolized due to low bioavailability in plasma, experimental data and measured chemical concentrations in field-collected animals suggest that this is not always the case. As a class, the PAHs represent a good example of this apparent discrepancy. Although they possess relatively high log K^sub ow^ values, PAHs are readily metabolized by fish (Varanasi et al. 1989). Research is needed to determine the effects of binding on in vitro and in vivo metabolism of hydrophobic chemicals by fish.

Extrapolation parameters required to translate CL^sub in vitro,int^ into an estimate of CL^sub h^ vary with the in vitro system used to characterize this activity, but may include liver size, liver blood flow rate, S9 protein content (mg/g liver, total), microsomal protein content (mg/g liver, total), and hepatocellularity (106 hepatocytes/g liver). Methods used to estimate these parameters were reviewed by Nichols et al. (2006), along with existing data for rainbow trout. Additional work is required to characterize these parameters for other species.

Strategic Selection of Chemicals Issues that pertain to the selection of appropriate test chemicals are summarized in Table 3. Chemicals representing four classes (A-D) can be segregated based on predicted or measured BCF (or BAF) values, and predicted degree of metabolic clearance. Within the two chemical classes (A and C) that undergo significant metabolism, there is a need to test compounds that are substrates for each of the major Phase I and II metabolic pathways. Because the endpoint of concern is chemical accumulation in fish, demonstration projects should be performed using compounds for which measured BCFs (or BAFs) already exist. Additional priority should be given to chemicals for which established analytical methods are available.

Modeling studies suggest that even high rates of metabolism are unlikely to reduce the accumulation of chemicals with log K^sub ow^ values less than 3 (Nichols et al. 2007). In contrast, relatively low rates of metabolism may have a large influence on the accumulation of very hydrophobic compounds (log K^sub ow^ > 6). However, these chemicals are often difficult to work with. Adsorption to the reaction vessel and/or binding to macromolecules can substantially reduce the amount of chemical available to interact with an in vitro test system. Hydrophobic chemicals also exhibit low aqueous solubility, complicating efforts to conduct in vivo waterborne exposures. Based on these considerations, it is recommended that initial validation studies be carried out on compounds with log K^sub ow^ values ranging from 3 to about 6.

Methods Standardization and Cross-Laboratory Comparisons

In vitro assay methods must be standardized if the data they generate are going to be used for environmental and public health priority-setting and decision-making. This includes standardization of the assay protocols themselves, as well as procedures used to obtain biological material. The acceptance of a “standardized” method may be facilitated by demonstrating its reproducibility in a set of cross-laboratory comparisons using the same chemicals and source material.

Potential Errors Due to Extrahepatic Metabolism

The demonstration project proposal described earlier is focused on hepatic metabolism because the liver is thought to be the primary organ of xenobiotic metabolism in fish. Workshop participants recognized, however, that extrahepatic metabolism may substantially impact the extent to which chemicals accumulate. This is particularly true when presystemic metabolism in tissues of the gills or gut limits the rate of chemical uptake at these exchange surfaces (Van Veld et al. 1988; Barron et al. 1989). Under these circumstances, the extent of accumulation predicted using in vitro hepatic metabolism data could substantially overestimate observed values.

A need exists to better characterize the circumstances (species, chemical, route of exposure) under which extrahepatic metabolism contributes to reduced uptake and increased whole-body clearance. In vitro subcellular systems have been adapted to investigate both branchial (Barron et al. 1999) and intestinal (James et al. 1997; Kleinow et al. 1998) metabolism. Metabolism by tissues of the intestinal tract also has been studied using an in situ perfused intestinal preparation (Van Veld 1988; James et al. 1996; Kleinow et al. 1998).

DECISION TREE FOR REGULATORY APPLICATION OF IN VITRO DATA

Requirement for a Decision Tree

To ensure its credibility and defensibility, the process for deciding which chemicals should undergo animal testing and which could be evaluated using models and/or in vitro data should be standardized according to a consensus of the scientific community. From a global perspective, this standardization would encourage the use of experimental results obtained from different international stakeholders and minimize duplication of testing. In this section we describe how in vitro data could be used within the context of a “decision tree” to better assess chemical bioaccumulation in fish and prioritize testing needs.

The decision tree consists of five tiers. Early tiers rely on QSARs, computer-based fate and transport models, and a review of physicochemical properties to evaluate the potential for a compound to accumulate in fish. As the process moves through the tiers, data from in vitro systems are used to refine modeled bioconcentration predictions. Whole animal testing becomes necessary if, at the end of this process, predicted bioconcentration exceeds a pre- determined criterion value. A summary of the tiers is presented in Figures 2 and 3, and each is explained briefly in what follows.

Tier 1-Initial Screen for Bioaccumulation

The first step of this process is to conduct a search for BCFs or BAFs for the parent compound and/or structural analogs that have been measured using acceptable test methods (Weisbrod et al. 2007). If these data are lacking or thought to be of limited value due to methodological concerns, a simple regression model (QSAR) can be used to estimate the BCF for a chemical based on its measured or estimated log K^sub ow^ value (Veith et al. 1979). If, based on this approach, the chemical is determined to have a measured or predicted BCF greater than 500 (the Globally Harmonized System value recommended by UNEP [2006]), it would be passed on for further evaluation.

Tier 2-Environmental Modeling

The BCF models used in Tier 1 have been developed using data for compounds with log K^sub ow^ values of 7 or lower and generally predict bioconcentration as a function of the total concentration of chemical in water. These models do not account for processes such as volatilization, sediment binding, and binding to dissolved organic carbon that may reduce the freely dissolved concentration of the chemical in water. The second step in this evaluation process, therefore, is to determine how the chemical of interest is likely to partition within the environment and whether it would be available for uptake by fish.

Fugacity-based models (Level III, Mackay and Paterson 1991) provide a useful means of predicting the distribution of a compound within the environment and may be formulated to include abiotic or biotic degradation rates. Additional empirical models can be used to predict the free concentration of chemical in water as a function of K^sub ow^, DOC, and POC (Arnot and Gobas 2004). The results from these models can serve in turn as inputs to food web bioaccumulation models to obtain initial predictions of chemical concentrations in biota occupying the lowest levels of an aquatic food web (Thomann et al. 1992; Gobas 1993). If a chemical is determined to be available for uptake by fish, either as freely dissolved chemical in the water column or as a contaminant in lower trophic level organisms, it would be passed to Tier 3 of the assessment process.

Tier 3-Potential for Absorption

In vitro systems may offer a means of predicting gill and gut uptake for specific compounds of interest. Here we propose a three- pronged approach using physicochemical information in combination with in vitro data to estimate the potential for a chemical to be absorbed by fish.

Physicochemical parameters

The success of “Lipinski’s rule of five” for predicting the oral bioavailability of potential drug candidates demonstrates the value of physicochemical properties (Lipinski et al. 1997). At this stage in the assessment process, physicochemical parameters for the compound of interest would be compared to guidelines for uptake and accumulation in fish (de Wolf et al. 2007). A chemical that possessed physicochemical parameters greatly exceeding one or more of these guidelines would be viewed as having a low probability for bioaccumulation.

Bioaccumulation prediction based on simple phase partitioning

Biomimetic sampling devices such as those listed in Table 1 have been used to conduct site-specific assessments of chemical bioavailability under natural conditions. These or similar systems could be used to predict the extent of chemical accumulation in fish that would be expected from simple phase partitioning (water and tissue lipid). Unlike the BCF prediction from Tier 1 of this process, this experimentally determined partitioning estimate would account for factors (e.g., pH, DOC) that alter the freely dissolved concentration of the chemical in water.

Experimental assessment of membrane transport

Although biomimetic systems reflect the physicochemical process of phase partitioning, they lack properties such as membrane transport and metabolism that may be important determinants of chemical uptake. The Caco-2 cell line has been used extensively to predict the intestinal absorption of drugs and environmental contaminants in mammals. The applicability of this or a similar approach to fish is largely unexplored. If a cell-based system were available to predict chemical absorption for the suspected route (s) of exposure, data from that system would be used at this time.

Tier 4-Metabolism Assessment

If it appeared that a compound would be classified as “bioaccumulative” based on physicochemical characteristics, environmental modeling, and its potential for absorption, then it would be necessary to characterize its potential for biotransformation using a stepwise approach (Figure 3). The first step in this process is to use a QSAR (Borodina et al. 2003) and/or a knowledge-based expert system (Klopman et al. 1994; Talafous et al. 1994; Balmat et al. 2005; Mekenyan et al. 2006) to predict likely metabolic pathways and products. This information would be used to select an appropriate in vitro test system, as the applicability domain of these systems becomes better known. If the results of this analysis suggest that the compound is unlikely to be metabolized, it may be appropriate to skip to Tier 5 in the process.

Assuming that there is some potential for the compound to undergo metabolism, the first experimental step would be to screen for metabolic activity using a subcellular liver fraction. Due to its ease of preparation and the presence of both of Phase I and II activities, the S9 fraction is well suited for this purpose. Selection of a test species may be based on regulatory requirements or geographic region. In the absence of specific guidance, it is recommended that the experiment be conducted using rainbow trout or common carp. The assay would be performed using a substrate- depletion approach, intended to quickly determine if fish can metabolize the compound at a relatively high rate as compared to benchmark materials with known in vivo rates of metabolism. If the rate of metabolism is not sufficient to employ a substrate depletion approach, it may be advisable to run a traditional enzyme kinetics study utilizing multiple substrate concentrations. The results of a fish hepatocyte assay also could be used at this time as a check on the results of the S9 assay, particularly if there was reason to believe that the two assays would provide different results due to membrane diffusion limitations on metabolic rate, or because the compound is a potential substrate for membrane transport proteins. The results from these assays would be used to calculate in vitro intrinsic hepatic clearance (CL^sub in vitro,int^). CL^sub in vitro,int^ would then be extrapolated to estimate in vivo hepatic clearance and incorporated into an established kinetic model to re- estimate the BCF. If the revised BCF is close to the criterion value, consideration should be given to repeating the analysis using metabolism information from other fish species. Tier 5-In Vivo Determination of BCF or Risk Assessment

If, after completing Tiers 1-4, the predicted BCF is still greater than the established criterion, then one may assume that the chemical will bioaccumulate to an unacceptable level or conduct whole-animal exposures to test this possibility. The purpose of this test would be to provide a high-quality estimate of bioconcentration and, by extension, provide a reliable prediction of bioaccumulation within a natural exposure setting. Whole-animal tests of this type are generally performed using an accepted protocol such as the OECD 305 test (OECD 1996).

SUMMARY AND RECOMMENDATIONS

This report summarizes the results of a workshop on the use of in vitro ADME data in bioaccumulation assessments for fish. Workshop participants agreed that absorption and metabolism represent the two greatest sources of uncertainty in current bioaccumulation assessments, and that both can be investigated using established in vitro mediods. To a considerable degree, the proposal to use in vitro data in bioaccumulation assessments builds on the successful experience of the pharmaceutical industry, which routinely uses in vitro data during the drug development process. Although the physicochemical characteristics of chemicals that accumulate in fish often differ from those of drugs (in general, possessing much higher log K^sub ow^ values), the problem faced in both enterprises is the same: rapid identification of compounds that clearly do or do not meet established criteria for acceptance (e.g., the BCF criterion or, in the case potential drug candidates, criteria for oral bioavailability and elimination half-life). This allows testing resources to be focused on a much smaller number of chemicals for which high-confidence predictions cannot be obtained by simpler means.

Although in vitro methods exist to investigate the processes of absorption and metabolism, very few studies have been performed to compare in vitro data with in vivo outcomes for fish. The available database is too small to support conclusions on the quality of in vitro-in vivo metabolism extrapolations in piscine systems, which in vitro system is most suitable for in vivo prediction, or which in vitro system is applicable for which purpose. Established in vitro methods exist to estimate dietary absorption of drugs by mammals (e.g., Caco-2 cells, PAMPA). Whether these models can be adapted to predict dietary uptake of xenobiotics by fish remains to be investigated. There is an urgent need to conduct research to fill these data gaps; this will require new sources of research funding.

Because bioaccumulation assessments must be performed for a large number of compounds, the methods used to perform these assessments must be amenable to high-throughput. One issue in this regard is the possible existence of a tradeoff between the accuracy of an in vitro assay (defined as whether it predicts the same process in vivo) and its ease of use. One system may be easy to use, but yields inaccurate predictions for some compounds. A second may be more technically demanding, but provides data of greater accuracy. Under these circumstances, the first system may be more appropriate for initial screening and prioritization of large chemical inventories, whereas the second is better suited to collecting definitive information for a smaller number of compounds.

An issue of critical concern for the future use of in vitro data in bioaccumulation assessments is that of methods standardization. Much of the variability noted in comparisons of in vitro metabolism data among species may be related to methodological differences among studies. Standard methods for each of the major in vitro assays should be developed with the stated intent of supporting in vitro-in vivo extrapolations. We recommend that a future workshop be held to address this issue.

A “decision tree” approach was used to describe how in vitro data might be used in future bioaccumulation assessments. Some reluctance on the part of scientists and regulators to accept in vitro data in bioaccumulation assessments can be anticipated for the simple reason that it has not been done before. We recognize that existing in vitro methods may be insufficient to generate highly accurate predictions of in vivo bioaccumulation for all chemicals. It may be possible, however, to use this information in a “binning” approach (e.g., low, medium, or high metabolism) to place individual compounds into categories that would lead to more scientifically defensible assessments and reduce the need for animal testing. This would represent a major improvement over current assessment methods.

ACKNOWLEDGMENTS

This workshop was co-sponsored by the International Life Sciences Institute-Health and Environmental Sciences Institute (ILSI-HESI) and the Society of Environmental Toxicology and Chemistry (SETAC), and was organized by an ILSI-HESI Emerging Issues Bioaccumulation Workgroup. Financial support for the workshop was provided by ILSI- HESI. We thank the individuals who participated in the workshop, most of whom did so at their own expense. We also thank Dr. Keith Sappington and Dr. Larry Burkhard for constructive reviews of this report. The information in this document has been funded in part by the U.S. Environmental Protection Agency.

REFERENCES

Andersson T, Forlin L, and Hansson T. 1983. Biotransformation of 7-ethoxycoumarin in isolated perfused rainbow trout liver. Drug Metab Disposit 11:494-8

Arnot JA and Gobas FAPC. 2003. A generic QSAR for assessing the bioaccumulation potential of organic chemicals in aquatic food webs. QSAR Comb Sci 22:337-45

Arnot JA and Gobas FAPC. 2004. A food web bioaccumulation model for organic chemicals in aquatic ecosystems. Environ Toxicol Chem 23:2343-55

Arnot JA and Gobas FAPC. 2006. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in aquatic-organisms. Environ Rev 14:257-97

Austin RP, Barton P, Cockroft SL, et al. 2002. The influence of nonspecific microsomal binding on apparent intrinsic clearance, and its prediction from physicochemical properties. Drug Metab Disposit 30:1497-503

Balmat A-L, Judson P, Long A, et al. 2005. Predicting drug metabolism-An evaluation of the expert system METEOR. Chem Biodiversity 2:872-85

Barron MG, Schultz IR, and Hayton WL. 1989. Presystemic branchial metabolism limits di-2-ethylhexyl phdialate accumulation in fish. Toxicol Appl Pharmacol 98:49-57

Barron MG, Mayes MA, Murphy PG, et al. 1990. Pharmacokinetics and metabolism of tichlopyr butoxyethyl ester in coho salmon. Aquat Toxicol 16:19-32

Barron MG, Charron KA, Stott WT, et al. 1999. Tissue carboxylesterase activity of rainbow trout. Environ Toxicol Chem 18:2506-11

Bermejo M, Avdeef A, Ruiz A, et al. 2004. PAMPA-a drug absorption in vitro model 7. Comparing rat in situ, Caco-2, and PAMPA permeability of fluoroquinolones. Eur J Pharm Sci 21:429-41

Borodina Y, Sadym A, Flimonov D, et al. 2003. Predicting biotransformation potential from molecular structure. J Chem Inf Comput Sci 43:1636-46

Bradbury SP, Dady JM, Fitzsimmons PN, et al. 1993. Toxicokinetics and metabolism of aniline and 4-chloroaniline in medaka (Oryzias latipes). Toxicol Appl Pharmacol 118:205-14

Branson DR, Blau GE, Alexander HC, et al. 1975. Bioconcentration of 2,2′,4,4′-tetrachlorobiphenyl in rainbow trout as measured in an accelerated test. Trans Am Fish Soc 104:785-92

Buesen R, Mock M, Nau H, et al. 2003. Human intestinal Caco-2 cells display active transport of benzo(a)pyrene metabolites. Chem Biol Interac 142:201-21

Buhler DR and Wang-Buhler J-L. 1998. Rainbow trout cytochrome P450s: Purification, molecular aspects, metabolic activity, induction, and role in environmental monitoring. Comp Biochem Physiol 121C:107-37

Carlile DJ, Zomorodi K, and Houston JB. 1997. Scaling factors to relate drug metabolic clearance in hepatic microsomes, isolated hepatocytes, and the intact liver. Drug Metab Disposit 25:903-11

Castano A, Bols NC, Braunbeck T, et al 2003. The use offish cells in ecotoxicology. The report and recommendations of ECVAM workshop 47. Altern Lab Anim 31:317-51

Chung WG, Sen A, Wang-Buhler J-L, et al. 2004. cDNA-directed expression of a functional zebrafish CPY1A in yeast. Aquat Toxicol 70:111-21

Clark KE, Gobas FAPC, and Mackay D. 1990. Model of organic chemical uptake and clearance by fish from food and water. Environ Sci Technol 24:1203-13 Cornelissen G, Rigterink H, ten Hulscher DE, et al. 2001. A simple Tenax extraction method to determine the availability of sediment-sorbed organic compounds. Environ Toxicol Chem 20:706-11

de Wolf W, Seinen W, and Hermens JLM. 1993. Biotransformation and toxicokinetics of trichloroanilines in fish in relation to their hydrophobicity. Arch Environ Contam Toxicol 25:110-7

de Wolf W, Comber M, Douben P, et al. 2007. Animal use replacement, reduction and refinement: Development of an integrated testing strategy for bioconcentration of chemicals in fish. Integr Environ Assess Manag 3:3-17

Dierickx PJ. 1985. Purification and partial characterization of the glutathione S-transferases in carp liver, and their interaction with 2,4-dichlorophenoxyacetic acid and 1,4-benzoquinone. Biochem Int 11:755-63

Dimitrov SD, Dimitrova NC, Walker JD, et al. 2002. Predicting bioconcentration factors of highly hydrophobic chemicals. Effects of molecular size. Pure Appl Chem 74:1823-30

Dimitrov S, Dimitrova N, Parkerton T, et al. 2005. Baseline model for identifying the bioaccumulation potential of chemicals. SAR QSAR Environ Res 16:531-54

Dyer SD and Bemhard MJ. 2004. Follow-up to feasibility study on in vitro biotransformation systems: Determination of uptake, loss and bioconcentration of two surfactants. ERASM final report. Brussels, Belgium. Available at http://www.erasm.org/study.htm

Dyer SD, Bernhard MJ, and Versteeg DJ. 2003. Identification of an in vitro method for estimating the bioconcentration of surfactants in fish. ERASM final report. Brussels, Belgium. Available at http:// www.erasm.org/study.htm

Elskus AA, Collier TK, and Monosson E. 2005. Interactions between lipids and persistent organic pollutants (POPs) in fish. In: Mommsen TP and Moon TW (eds), The Biochemistry and Molecular Biology of Fishes, vol. 6, Environmental Toxicology, pp 119-52. Elsevier, St. Louis, MO, USA

Fisk AT, Norstrom RJ, Cymbalisty CD, et al. 1998. Dietary accumulation and depuration of hydrophobic organochlorines: Bioaccumulation parameters and their relationship with the octanol/ water partition coefficient. Environ Toxicol Chem 17:951-61

Fitzsimmons PN, Lien GJ, and Nichols JW. 2007. A compilation of in vitro rate and affinity values for xenobiotic biotransformation in fish, measured under physiological conditions. Comp Biochem Physiol 145C:485-506

Forlin Land Andersson T. 1981. Effects of Clophen A50 on the metabolism of paranitroanisole in an in vitro perfused rainbow trout liver. Comp Biochem Physiol 68G239-42

Frezard F and Garnier-Suillerot A. 1998. Permeability of lipid bilayer to anthracycline derivatives. Role of the bilayer composition and of the temperature. Biochim Biophys Acta 1389:13-22

Gilroy DJ, Miranda CL, Siddens LK, et al. 1996. Dieldrin pretreatment alters ^sup 14^C-dieldren and [^sup 3^H] 7,12- dimediylbenz(a)anthracene uptake in rainbow trout liver slices. Fundam. Appl Toxicol 30:187-93

Gobas FAPC. 1993. A model for predicting the bioaccumulation of hydrophobic organic chemicals in aquatic food-webs: Application to Lake Ontario. Ecol Mod 69:1-17

Gobas FAPC, Muir DCG, and Mackay D. 1988. Dynamics of dietary bioaccumulation and faecal elimination of hydrophobic organic chemicals in fish. Chemosphere 17:943-62

Government of Canada. 1999. Canadian Environmental Protection Act, 1999. Statutes of Canada. Chapter 33. Act assented to September 14, 1999. Department of Justice, Ottawa, ON, Canada

Gulden M, Dierickx P, and Seibert H. 2005. Validation of a prediction model for estimating serum concentrations of chemicals which are equivalent to toxic concentrations in vitro. Toxicol In Vitro 20:1114-24

Hack A and Selenka F. 1996. Mobilization of PAH and PCB from contaminated soil using a digestive tract model. Toxicol Lett 88:199- 210

Heringa MB, Schreurs RHMM, Busser F, et al. 2004. Toward more useful in vitro toxicity data with measured free concentrations. Environ Sci Technol 38:6263-70

Hidalgo IJ. 2001. Assessing the absorption of new pharmaceuticals. Curr Topics Med Chem 1:385-401

Hodson PV, Kloepper-Sams KR, Munkittrick KR, et al. 1991. Protocols for measuring mixed function oxygenases of fish liver. Can Tech Rep Fish Aquat Sci No. 1829

Houston JB. 1994. Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol 47:1469- 79

Houston JB and Carlile DJ. 1997. Prediction of hepatic clearance from microsomes, hepatocytes, and liver slices. Drug Metab Rev 29:891-922

Iwatsubo T, Hirota N, Ooie T, et al 1997. Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data. Pharmacol Ther 73:147-71

James MO, Kleinow KM, Tong Z, et al. 1996. Bioavailability and biotransformation of [^sup 3^H]-benzo(a)pyrene mteabolites in in situ intestinal preparations of uninduced and BNF-induced channel catfish. Mar Environ Res 42:309-15

James MO, Altman AH, Morris K, et al. 1997. Dietary modulation of phase 1 and phase 2 activities with benzo(a)pyrene and related compounds in intestine but not liver of the channel catfish, Ictalurus punctatus. Drug Metab Disp 25:346-54

Jones HM and Houston JB. 2004. Substrate depletion approach for determining in vitro metabolic clearance: Time dependencies in hepatocyte and microsomal incubations. Drug Metab Disposit 32:973- 82

Kane AS and Thohan S. 1996. Use of fish hepatic tissue slices to evaluate species-specific metabolic capacity and alterations due to environmental exposure. In: Ostrander G (ed), Methods in Aquatic Toxicology, pp 371-91. Lewis, Boca Raton, FL, USA

Kansy M, Senner F, and Gubernator K. 1998. Physicochemical high throughput screening: Parallel artificial membrane permeation assay in the description of passive absorption processes. J Med Chem 41:1007-10

Karara AK and Hayton WL. 1984. Pharmacokinetic model for the uptake and disposition of di-2-ethylhexyl phthalate in sheepshead minnow Cyprinodon variegates. Aq