The influence of chemical degradation during dietary exposures to fish on biomagnification factors and bioaccumulation factors

Abstract

The chemical dietary absorption efficiency (E_D) quantifies the amount of chemical absorbed by an organism relative to the amount of chemical an organism is exposed to following ingestion. In particular, E_D can influence the extent of bioaccumulation and biomagnification for hydrophobic chemicals. A new E_D model is developed to quantify chemical process rates in the gastrointestinal tract (GIT). The new model is calibrated with critically evaluated measured E_D values (n = 250) for 80 hydrophobic persistent chemicals. The new E_D model is subsequently used to estimate chemical reaction rate constants (k_R) assumed to occur in the lumen of the GIT from experimental dietary exposure tests (n = 255) for 165 chemicals. The new k_R estimates are corroborated with k_R estimates for the same chemicals from the same data derived previously by other methods. The roles of k_R and the biotransformation rate constant (k_B) on biomagnification factors (BMFs) determined under laboratory test conditions and on BMFs and bioaccumulation factors (BAFs) in the environment are examined with the new model. In this regard, differences in lab and field BMFs are highlighted. Recommendations to address uncertainty in E_D and k_R data are provided.

---

Environmental significance

Biotransformation is a key process mitigating bioaccumulation; however, there are significant knowledge gaps in the extent of chemical biotransformation in biota and the environment and hence significant uncertainty in chemical evaluations. We have developed a new model for simulating chemical uptake efficiencies into fish following dietary exposure and calibrated and evaluated this model with evaluated experimental measurements. Further, we have derived a new database of chemical primary transformation rate constants corresponding to the dietary exposure pathway. We have incorporated the new dietary absorption efficiency model into commonly used fish bioaccumulation models to demonstrate the impact of chemical transformation rates on bioaccumulation factors (BAFs) and biomagnification factors (BMFs). Finally, we have compared lab simulations for BMFs with field simulations for BMFs highlighting key differences when considering these different BMFs for decision-making. We believe the new model, database and findings are broadly applicable and will provide valuable insights for interpreting and applying bioaccumulation data for contaminant assessment and management.

1. Introduction

Bioaccumulation data are used for chemical hazard, exposure and risk assessment.^1–4 For hydrophobic organic chemicals in the aquatic environment dietary exposures can be the dominant route of exposure to fish,⁵ particularly for persistent chemicals.^6,7 Chemicals with octanol–water partition coefficients (K_OW; unitless) greater than about 10⁵ are shown to have biomagnification potential in fish and aquatic food webs in the absence of biotransformation;^8,9 however, biotransformation rates are a key determining factor of actual bioaccumulation in fish^10–13 and food webs.^14–16 Chemical degradation processes in the lumen and epithelial tissues of the gastrointestinal tract (GIT) are recognized^17–23 and can also mitigate bioaccumulation in fish.^24,25 Degradation rate data in the GIT and models for incorporating this information have the potential to improve model predictions for chemical assessments and to provide mechanistic insights into bioaccumulation processes.

The chemical dietary absorption efficiency (E_D or AE or α; unitless) quantifies the amount of chemical absorbed by the organism relative to the amount of chemical the organism is exposed to in the GIT following ingestion.^26,27 The absorption efficiency can be determined from dietary bioaccumulation experiments such as the Organization for Economic Coordination and Development (OECD) 305 test guideline²⁷ and is a parameter in bioaccumulation and toxicokinetic models for fish.^26,28–34 Empirical models for E_D have been developed from regressions of measured E_D against K_OW for persistent organic pollutants (POPs) in various species.^35–38 These empirical models are commonly applied in bioaccumulation^34,39 and exposure models.^40,41 Empirical K_OW–E_D models for persistent chemicals do not account for possible degradation rates in the GIT that can occur for non-persistent chemicals following dietary ingestion. Databases of E_D for fish have been developed.^34,42 Models for including and estimating rates of chemical degradation in the GIT (k_R or k_GM; h⁻¹) have been developed.^24,25,43 Previous modelling work has explored the role of k_R on lab BCFs and lab BMFs for a few select chemicals,²⁴ but not environmentally relevant field BAFs and BMFs in which organisms are co-exposed to chemical in the water and the diet simultaneously. Models can be used to extrapolate laboratory data to field conditions;⁴⁴ however, comparisons of measured and modelled lab and field BMFs are limited.

In this study we develop, parameterize and apply a new model to quantify chemical processes in the GIT to estimate E_D and k_R. The new E_D model is first presented and then calibrated and evaluated with measured E_D data for POPs and is compared with other empirical models for fish. The mechanistic nature of the model is exploited to estimate k_R for chemicals that have been subject to experimental testing following similar methods as the calibration dataset. The new estimates of k_R are compared to existing k_R estimates derived from other methods for the same chemicals. The new E_D model is incorporated into bioaccumulation models to explore the role of chemical degradation in the GIT on lab and field BMFs and field BAFs.

2. Methods

2.1 Theory and terminology

Bioaccumulation assessment endpoints discussed in this study are presented in the ESI-1.† The models described below are expressed in terms of chemical fugacity. Briefly, fugacity (f; Pa) describes the escaping tendency of a chemical in a particular phase⁴⁵ or compartment such as the GIT. The fugacity capacity (Z; mol Pa⁻¹ m⁻³) quantifies the capacity of a particular phase to store a chemical under defined conditions. The relationship between chemical concentration (C; mol m⁻³) and f is a function of Z as C = fZ. Transport processes in fugacity terms are expressed by D values (mol Pa⁻¹ h⁻¹) that are analogous to conventional rate constants (k; h⁻¹). Chemical flux (N; mol h⁻¹) is then a product of f and D. Characteristic times τ (h) are 1/k or VZ/D, where V is the volume of the compartment (m³).

The E_D parameter quantifies chemical transfer from the GIT to the blood (hepatic portal vein) and the oral bioavailability parameter (F; dimensionless) quantifies chemical transfer from the GIT to systemic circulation following chemical transport and potential biotransformation in the liver. If chemical degradation rates in the GIT and liver are negligible relative to other rates in the GIT and the liver, E_D and F are equivalent values. For chemicals subject to sufficient biotransformation in the liver and/or degradation in the lumen or tissues of the GIT, F will be less than E_D. Historically most fish dietary bioaccumulation testing has been conducted with POPs such as polychlorinated biphenyls (PCBs). While E_D is a satisfactory descriptor for both absorption and oral bioavailability for persistent chemicals (i.e., chemical transformation reaction rates are negligible in the lumen of the GIT, perfused GIT tissue, liver and other compartments), F is a more appropriate description for “E_D data” reported in the literature for chemicals that are subject to degradation and transformation before systemic circulation.

Fig. 1 illustrates a series of conceptual models that can be formulated into mathematical expressions to quantify chemical transport in an organism following dietary exposure. The different models reflect different objectives and the knowledge required to reliably quantify these processes. Fig. 1A represents a model in which the organism and the GIT are considered to be a single compartment with a common fugacity f_WB to which a single reaction loss rate is applicable (e.g., N_RWB; mol h⁻¹). The E_D parameter is included in the dietary uptake rate (N_D) and the egestion rate (N_E) equations in these models. The model described in Fig. 1B separates the lumen of the GIT from the rest of the body. Different fugacities can apply to each compartment f_F and f_G and there are two reaction loss rates for each compartment. In this two-compartment model E_D is N_GF/N_D. Significant reactive loss in the GIT compartment (N_RG) reduces the flux of chemical into the fish (N_GF) and hence lowers E_D from its potential value in the absence of reaction. Fig. 1C illustrates a further refined model in which the liver is considered as a separate compartment from the rest of the body and there is one fugacity for each compartment: f_F, f_G, f_L. Multi-compartment physiologically-based pharmacokinetic models (PBPK), e.g., ref. 46–49, further separate the body into additional anatomical compartments such as the kidney, gill, muscle, adipose, etc. The model in Fig. 1C and PBPK models allow for possible first-pass effects in which the chemical is biotransformed in the liver after absorption from the GIT before systemic circulation. In the three-compartment model E_D is N_GL/N_D and F is N_LF/N_D. Significant reactive loss in the GIT compartment (N_RG) or the liver (N_RL) reduces the flux of chemical into systemic circulation.

Fig. 1 Three different conceptual models and associated chemical flux parameters (N; mol d⁻¹) for chemical uptake from water (U) and the diet (D) into an organism and elimination to the water from respiration (L) and egestion (E). Chemical loss rates from reaction (degradation) are noted by red text and red arrows. Model (A) considers the organism to consist of one “body” compartment that includes the gastrointestinal tract (GIT) and for which a single reaction loss rate (N_RWB) occurs. Model (B) separates the lumen of the GIT and the rest of the body and includes chemical transport between the two compartments and for which there are reaction loss rates (N_RG and N_RF) for each. Model (C) further separates the liver from the rest of the body and includes chemical fluxes between the three compartments and for which there are reaction loss rates (N_RG, N_RL and N_RF-L) for each.

OECD 305 test guidelines suggest whole body measurements (i.e., including the GIT and liver)²⁷ for the determination of the BMF and toxicokinetic parameters; however, the guidelines state that “specific tissues (e.g. muscle, liver) can be sampled, if desired and another option is that the GIT can be removed and analyzed separately at the end of the exposure phase and at days 1 and 3 of depuration”. Relatively few fish dietary bioaccumulation studies have separated the GIT from the fish e.g., ref. 24, 50 and 51, and the liver and GIT from the fish e.g., ref. 52–54. In cases where the GIT is removed, this includes perfused tissue, not only the lumen (cavity) of the GIT and associated microflora and secreted enzymes. Approximately 40% of fish dietary toxicokinetic data generated to-date have been derived from homogenization of the entire fish (including liver and GIT).⁴² Thus it is not possible to separate E_D from F in most existing fish bioaccumulation data and it is not possible to discriminate between N_RG and N_RL and N_RWB. Methods to calculate toxicokinetic parameters in the OECD 305 dietary test guidelines and one-compartment fish models commonly used for bioaccumulation assessment cannot differentiate between E_D and F because the models do not include a compartment for the liver and hepatic blood flow rates. The focus of this paper is on E_D data and models that can be used to parameterize and refine one-compartment fish bioaccumulation and toxicokinetic models (Fig. 1A).

2.2 Current fish E_D regression models

After ingestion, chemicals can be transported (actively and passively) from the GIT to the bloodstream through a series of aqueous and organic phases.⁵⁵ A common model used to quantify E_D is the two resistance (or two ‘film’) model,^35,56–58 and correlations for E_D values measured in fish^33,35 have been developed in the form:

1/E_D = A + BK_OW (1)

where A characterizes an organic phase resistance and B an aqueous phase resistance and octanol is assumed as a surrogate for organic phases. The general trend expressed in these correlations is that E_D decreases at higher K_OW and becomes quite small for super-hydrophobic substances of log K_OW 8 and higher. Fig. ESI-1† illustrates several E_D–K_OW empirical regression models reviewed by Barber.³⁴ Significant differences between the empirical models (Fig. ESI-1†) reflect the differences in the selection of data used in the regressions and the general variability and uncertainty in measured and predicted E_D.

2.3 New fish E_D model

The new E_D model simulates the chemical absorption process in terms of relative process rates. The model can be incorporated into one-compartment mass balance bioaccumulation and toxicokinetic models to refine predictions for chemicals subject to degradation during the complex series of events following dietary exposure. The model addresses chemical uptake in terms of kinetic (uptake and reaction) and partitioning processes and bioavailability in the GIT. The model is derived in chemical fugacity terms and also presented in terms of rate constants, corresponding characteristic times and relative efficiencies.

Conceptually, a chemical enters the GIT in a matrix comprised of aqueous and organic phases at an initial (time zero) fugacity (f₀) prior to digestion. Following ingestion the fugacity of the chemical in the food changes and the chemical has three ultimate potential fates: (1) absorption into the body, (2) fecal egestion, or (3) degradation (reaction). As a result of digestion processes in the GIT, the volume and gut contents of the ingested matrix change over time and the fugacity may increase, i.e., gastrointestinal magnification may occur. For simplicity, it is assumed that the ingested matrix is comprised of two sub-phase volumes, i.e., water (V_W; m³) and organic matter (V_O; m³). These two volumes are further assumed to represent post-digestion volumes, i.e., fecal matter following digestion and absorption of the ingested food. The total fecal volume (V_T) is then V_W + V_O and the total fecal Z-value (Z_T) is a function of the relative proportions and fugacity capacities of both phases, i.e., Z_T = (V_WZ_W + V_OZ_O)/V_T. The flow rates of aqueous and organic matter from the GIT into the organism as a result of digestion processes are represented as G_W and G_O with units of m³ h⁻¹. It is assumed that the fugacity (or dissolved concentration) in the aqueous phase provides the “driving force” for diffusion from the GIT into the organism.

First, assuming no chemical degradation in the GIT, the absorption of chemical from the GIT as a function of time can be expressed in fugacity terms by the following differential equation:

(V_TZ_T)df/dt = −D_Af (2)

where D_A is the chemical transport parameter for chemical absorption. Rearranging gives:

df/f = −(D_A/V_TZ_T)dt (3)

Integration gives:

ln f = −(D_A/V_TZ_T)t + x (4)

where x is a constant and when t = 0, f = f₀. Therefore

ln(f_E/f₀) = −D_At/V_TZ_T (5)

or

f_E/f₀ = exp(−D_At/V_TZ_T) (6)

where f_E is the final fugacity in excreted feces, and the chemical absorption efficiency E_D is 1 − f_E/f₀. Rearranging this equation gives f_E/f₀ = 1 − E_D; therefore, substituting this equality into eqn (5) gives:

ln(1 − E_D) = −D_At/V_TZ_T (7)

or

E_D = 1 − exp(−D_At/V_TZ_T) (8)

This equation shows the limiting properties that when t = 0, then E_D = 0, and at long values of t and high values of D_A/V_TZ_T, E_D approaches 1 (or 100%). In terms of rate constants the equivalent equation is:

E_D = 1 − exp(−k_At) (9)

where k_A is the absorption rate constant. The characteristic time of chemical absorption from the GIT is τ_A (h), which is calculated as V_TZ_T/D_A, or in rate constant terms 1/k_A. E_D can then also be expressed as:

E_D = 1 − exp(−τ_G/τ_A) (10)

where τ_G is the gut characteristic time (h).

If there is simultaneous chemical degradation in the GIT, an additional parameter is required to quantify this process; namely a reaction rate constant (k_R), D-value (D_R), or characteristic time (τ_R), depending on the formulation of the equations, e.g.:

f/f₀ = exp(−(k_A + k_R)t) (11)

The fraction egested is exp(−(k_A + k_R)t) and can be referred to as the egestion efficiency (E_E) and the fraction absorbed and reacted is 1 − exp(−(k_A + k_R)t), or in terms of efficiencies E_D + E_R, where E_R is the reaction efficiency. If degradation and absorption do not occur simultaneously but sequentially in whole or in part, the fractions absorbed and degraded will vary depending on the sequence but this obviously requires more detailed information on relative rates and locations within the GIT. In the absence of such information the simplest expedient is to assume simultaneous absorption and degradation in the model. The relative rates of absorption and reaction are thus proportional to k_A and k_R, e.g.:

E_D = [k_A/(k_A + k_R)][1 − exp(−(k_A + k_R)t)] (12)

We can use D values to explain D_A (and k_A and τ_A) using the two resistance model and thus the dependence on K_OW. For example, when there is negligible degradation:

τ_A = 1/k_A = V_TZ_T/D_A = (V_WZ_W + V_OZ_O)(1/G_OZ_O + 1/G_WZ_W) (13)

Dividing the numerator and denominator by Z_W and replacing Z_O/Z_W with K_OW the equation becomes:

τ_A = 1/k_A = V_TZ_T/D_A = (V_W + V_OK_OW)(1/G_OK_OW + 1/G_W) (14)

where the term (V_W + V_OK_OW) characterizes the availability of the chemical in the GIT and the term (1/G_OK_OW + 1/G_W) characterizes the kinetic aspects of the process. Measured E_D data for persistent chemicals (assuming k_R = 0) can be used to calibrate the model and parameterize values for V_W, V_O, G_O and G_W to calculate τ_A.

Following model calibration to E_D values for persistent chemicals (k_R is assumed 0) and k_As (and τ_As) are calculated for a range of K_OW, k_R for less persistent chemicals can be estimated as:

τ_R = 1/k_R = α/(1 − α/τ_G) (15)

where α is −ln(1 − E_D,X)τ_A and E_D,X is a measured “E_D value” for a non-persistent chemical derived under exposure conditions similar to those in which the calibration data were obtained (i.e., similar feeding rates, dietary lipids, fish size, water temperature). It may be that E_D,X closely approximates F; however, F is commonly estimated from measured blood concentrations (after the “first past-effect”) and it is possible the E_D,X is different because it includes GIT tissues and liver (part of “first past-effect”).

2.4 Model parameterization and calibration

The model is parameterized using estimates for τ_R, τ_G, V_O, V_W, G_O, and G_W and by calibration to measured E_D data for persistent chemicals; hence, k_R is assumed to be negligible (10⁻¹² h⁻¹); therefore, τ_R is 10¹² h. Given the highly dynamic conditions for the food ingestion and digestion process, it is difficult to obtain a single value for the gut characteristic (residence or transit) time τ_G. We estimate the gut characteristic time τ_G as V/G where V is the volume of the GIT (m³) and G is the flow rate for the gut contents (m³ h⁻¹), i.e., ingested food, digesting food, bile salts, microflora, feces, etc. This approach assumes the chemical can be absorbed throughout the GIT, whereas the true “effective residence time” for chemical absorption may be shorter. The volume fraction of the entire GIT is about 3–5% the whole fish and an initial value of 3.3% v/v was assumed for the effective volume of the GIT (sum volume fractions for pyloric ceca, upper intestine, lower intestine).⁴⁷ Historically, fish in laboratory tests are fed daily at rates typically ranging from 1–3% body weight on a dry weight basis (median = 2% body weight per day).⁴² Different assumptions can be made regarding the amount of water co-ingested with the diet and the densities of the dietary components (e.g., ref. 24). An equal volume of water is assumed ingested with the administered fish food and the densities of all ingested components are 1 kg L⁻¹ (e.g., ref. 24) corresponding to a wet weight volumetric ingestion rate of about 4% body weight per day. The initial τ_G is thus about 20 h or 0.83 d which corresponds to a k_G of 1.2 d⁻¹.

The V_O and V_W parameters can be calculated based on measured or estimated lipid and water volume fractions in the feces. In the current application we employ a version of the dietary digestion sub-model in the AQUAWEB³³ fish bioaccumulation model. The digestion model is parameterized with median values from a large collection of dietary bioaccumulation testing data (10 g fish; 15 °C)⁴² to estimate V_O and V_W from typical dietary bioaccumulation experiments. The V_O parameter was estimated to be 3.37 × 10⁻⁹ m³ from the median lipid equivalent contents reported from the feeding studies (14% ⁴²) and the fish digestion model.³³ Based on the same approach V_W was estimated to be 1.11 × 10⁻⁷ m³. Values for G_O and G_W have been derived from small fish using dietary feeding studies in the 1980s³⁵ ranging in value from about 2.42 × 10⁻¹² to 3.33 × 10⁻¹⁰ (average = 8.63 × 10⁻¹¹ m³ h⁻¹) and from about 5.83 × 10⁻⁵ to 8.25 × 10⁻³ (average = 2.12 × 10⁻³ m³ h⁻¹), respectively. The median ratio of reported G_W/G_O values is 2.44 × 10⁷.³⁵ Average estimates of 8.63 × 10⁻¹¹ and 2.12 × 10⁻³ were selected for the transport parameters G_O and G_W, respectively. The initial parameters for model calibration are summarized in Table 1.

Table 1 A summary of parameters for the new chemical dietary absorption efficiency E_D model as derived for a 10 g fish using assumed laboratory feeding conditions

Parameter	Symbol (units)	Initial	Calibrated
Reaction residence time	τ R (h)	10^12	10^12
GIT residence time	τ G (h)	20	20
Volume of octanol equivalent in the feces	V O (m^3)	3.37 × 10^−9	3.37 × 10^−9
Volume of water in the feces	V W (m^3)	1.11 × 10^−7	1.11 × 10^−7
Flow rate of organic matter from the GIT into the organism	G O (m^3 h^−1)	8.63 × 10^−11	1.19 × 10^−10
Flow rate of water from the GIT into the organism	G W (m^3 h^−1)	2.12 × 10^−3	2.12 × 10^−2

---

An evaluated database of measured E_Ds⁴² as well as another dataset of measured E_Ds²⁵ form the basis of the data used in this study. The data were critically reviewed in an effort to address experimental variability and uncertainty. The E_D data and the process for selecting relevant and reliable E_D data are summarized in ESI-2.† The new E_D model calibration dataset (Table ESI-1†) primarily consists of PCBs, which are poorly transformed, and other selected persistent chemicals that are poorly biotransformed in fish providing 250 measured E_Ds. The K_OWs for PCBs reported by Hawker and Connell⁵⁹ were used for model calibration. For other chemicals, measured K_OWs were selected preferentially to predicted values.⁶⁰ The K_OW in the calibration set range from log K_OW 4.5 to 8.2. While temperature can influence chemical partitioning, we did not include any temperature correction to K_OW.

The model was calibrated to the empirical dataset using the Solver function in Microsoft Excel® to obtain the minimal sum square error between observed and predicted E_D values. The τ_R, τ_G, V_O, V_W parameters remained fixed and the G_W parameter and the ratio between G_O and G_W were allowed to vary to allow for changes to G_O. The G_O and G_W parameters were selected for calibration because they cannot be measured directly. The calibrated empirical estimates for G_O and G_W are shown in Table 1.

2.5 Estimating k_R

Following eqn (15), the new model can be used to obtain estimates for k_R for less persistent chemicals when “E_D,Xs” are derived experimentally under exposure conditions similar to the calibration dataset. The criteria outlined for the selection of data for model calibration (ESI-2†) were also followed to obtain a database of E_D,X values (n = 255) from which preliminary estimates of k_R were calculated. Given the structure of the new E_D model, k_R is assumed to occur in the lumen of the GIT before the chemical is absorbed into the epithelium and tissues of the GIT. However, a key uncertainty with the k_R estimates is that the measured E_D,X data do not have sufficient resolution to truly determine where these reactions may be occurring because they are obtained from whole body homogenates. These preliminary rate estimates are perhaps best considered in relative terms, rather than absolute.

2.6 Simulating field BAFs and field and lab BMFs

The new E_D model was coded into bioaccumulation models to compare the new E_D model with previous E_D regression models on calculated field BAFs (i.e., C_Fish/C_Water) and lab and field BMFs (i.e. C_Fish/C_Diet). The new model was also used to examine the role of k_R and biotransformation rate constants (k_B,N; d⁻¹, normalized to 10 g body weight at 15 °C ⁶¹) on calculated BAFs and BMFs. The models were parameterized for a set of hypothetical neutral organic chemicals comprising a range of hydrophobicity (ranging in log K_OW from 0 to 12).

The AQUAWEB model³³ has been tested and applied in various ecosystems, e.g., ref. 15, 33 and 62–64 and laboratory tests, e.g.ref. 10, 42 and 65–67. The AQUAWEB model also forms the basis of the KABAM model used in US EPA pesticide evaluations.⁶⁸ The calibrated E_D model was coded into the AQUAWEB model that was then subsequently parameterized to calculate field BAFs and BMFs for a fish (0.01 kg; 5% lipid content) in an assumed environment consuming prey (5% lipid content) that is at equilibrium with the surrounding water. The same conditions were used to parameterize the model to simulate lab BMFs in which there was only exposure to the fish from the diet (no chemical exposure or uptake from the water). The simulations compare the model output with the previous and new E_D sub-models and for various assumptions of biotransformation rates in the fish and degradation rates in the GIT.

The BAF-QSAR model³² incorporated in the BCFBAF module of the United States Environmental Protection Agency's EPI Suite™ software^4,69 calculates BAFs in three general trophic levels of fish.^4,70 The E_D parameter is a variable in the BAF-QSAR model. The BAF-QSAR is calibrated to measured BAFs from the environment using assumed representative ecological parameters and requires only K_OW and k_B,N as model input parameters to simulate BAFs, thus providing environmentally relevant bioaccumulation hazard information for screening assessments. The newly calibrated E_D model was coded into an Excel spreadsheet version of the BAF-QSAR model (http://www.arnotresearch.com) that was subsequently used to calculate BAFs in upper trophic level fish and compare the results with the previous and new E_D models with different assumptions for biotransformation rates in the fish and degradation rates in the GIT.

3. Results and discussion

3.1 Comparison of E_D models with measured data

Fig. 2 compares the calibrated E_D model (black line; τ_G = 20 h) with the evaluated calibration dataset. Fig. 2 also includes three regression models: Gobas et al., 1988;³⁵ Arnot and Gobas, 2004;³³ and Lo et al., 2015.²⁴ All of the models show a plateau in E_D at about 50% for chemicals with log K_OW > ∼4 and <∼6. For chemicals with log K_OW > ∼6 there is a decrease in E_D with increasing K_OW. The decrease at higher K_OW can be explained in kinetic terms by the increased mass transfer resistance of highly hydrophobic chemicals through aqueous diffusion barriers such as the unstirred water layer in the GIT. The new model shows a decrease in E_D for chemicals with log K_OW < ∼3. This decrease cannot currently be corroborated with measured data because there are no measured E_D data for chemicals with low K_OW; however, the mechanistic explanation for this decrease is the increased mass transfer resistance of highly water soluble chemicals through the organic phases of the GIT (i.e., membranes). Similar decreases in E_D with decreasing K_OW have been observed and modelled with human data.⁷¹ The influence of lower E_D on fish at lower K_OW is insignificant because the uptake of low K_OW chemicals from food is much slower than uptake of chemicals from water. It is challenging to obtain reliable measurements of E_D in fish for lower K_OW chemicals (i.e., log K_OW < ∼3) because elimination from the gill to the water is relatively fast, e.g., gill elimination half-lives ∼≤1 h in a 10 g fish.⁴²

Fig. 2 Model predictions and 250 measurements of dietary chemical absorption efficiency (E_D) for persistent organic chemicals in fish as a function of the octanol–water partition coefficient (K_OW). Black and grey circles are considered high confidence and moderate confidence values, respectively. Three regression models are presented^24,33,35 and the new E_D model illustrating changes to E_D due to changes in the gastrointestinal tract characteristic (residence) time τ_G.

For chemicals with log K_OW > 4, the new model is similar to a recent regression model.²⁴ Both newer models show generally higher E_D at moderate and high K_OW compared to the earlier regression models.^33,35 For a persistent chemical with a log K_OW of 8 the more recent models show E_D values about 3–10 times higher than the previous E_D–K_OW regressions.^33,35 For chemicals with a log K_OW of around 9 the more recent models suggest E_D values about 5–30 times higher than the previous regression models. Many of the chemicals used to develop the earlier regression models were high K_OW chlorinated dioxins; some of which are subject to biotransformation.^61,72 If significant degradation of these chemicals occurs, the true potential E_D values for persistent high K_OW chemicals would be higher than the reported E_D values from those studies. Dioxin data were not included in the new model or in the recent E_D–K_OW regression model.^24,39,67 Reliable quality E_D data for persistent chemicals with log K_OW > 8.2 do not currently exist; hence, uncertainty in the models may be significant for chemicals with log K_OW > ∼8. Additionally, K_OW is an input parameter and there are few reliable measurements of K_OW for chemicals with log K_OW > ∼8. The mechanistic and theoretical basis of the new E_D model and its calibration to evaluated data provides some confidence in the model's application with an appreciation of the uncertainty at very high K_OW. Very hydrophobic chemicals can establish high body burdens in fish but the uptake time is prolonged thus the residence time of the chemical in the fish will be long. Even very slow rates of biotransformation and growth dilution will then become significant.

3.2 Model sensitivity and uncertainty

The sensitivity of simulated E_D as a function of changes to model parameters was examined. The gut residence time of the digesta is an uncertain variable and a function of the feeding rate. If the assumed volume for the GIT is the same and the feeding rate doubles (i.e., ∼8% bw d⁻¹) τ_G is 10 h (dashed black line in Fig. 2). If the actual effective volume of the GIT for chemical absorption is smaller and the flow rate is the same, then τ_G will also be shorter. Conversely if the assumed volume of the GIT is larger or the feeding rate is smaller, all else being equal τ_G will be larger (e.g., τ_G = 40 h, dashed black line in Fig. 2). Following changes to τ_G the model can be re-fit with changes to G_W and G_O to approximate general empirical trends.

The uncertainty in empirically derived E_D (e.g., Fig. 2) for persistent chemicals is recognized, e.g., the OECD 305 ring trials^27,73 and is difficult to reconcile without extensive examination of the raw experimental data. Fig. 2 shows that despite efforts to address experimental variability and uncertainty in selection of the calibration dataset, variability in the measured data remains large. For example, there are 34 E_Ds for hexachlorobenzene (HCB) ranging from 0.20 to 0.86 (median = 0.54). Following the OECD 305 kinetic methods for calculating E_D,²⁷ errors in the measured total elimination rate constants can result in errors in E_D. Specifically, overestimates of the total elimination rate constant (underestimation of half-life) can result in overestimates of E_D and underestimates of the actual administered dose can result in overestimates of E_D. Overestimating the initial concentration assumed to be in the organism after the dietary exposure phase (C₀) can also result in overestimates of E_D. The root mean square error between the calibrated model predictions and observations is 0.15 and there is no bias in model error as a function of K_OW (Fig. ESI-2,† plot of residuals). The high variability in E_D data is a cause for concern and in view of the importance of E_D as a parameter influencing bioaccumulation and biomagnification, there is a regulatory incentive to define the process of obtaining E_D more consistently.

3.3 Estimating k_R

The new E_D model was used to estimate k_R from a dataset of E_D,X values that were obtained following testing methods that are similar to those of the calibration dataset (ESI-2†). Fig. 3A summarizes the distribution of 255 new k_R estimates that range from about 0.015 to 84 d⁻¹ (corresponding to reaction half-lives of 0.0082 to 47 d). Table ESI-2† lists the k_R estimates. The chemicals in the new k_R database include hydrocarbons (linear and polycyclic aromatics), phthalate esters, fungicides, dioxins, furans, brominated flame retardants, and various other halogenated organics. The chemicals range in log K_OW from about 2.8 to 9.0 and molar mass of 120.2 to 959.2 g mol⁻¹. Fig. 3B compares the new k_R estimates with previous estimates of k_R (referred previously as k_GM)²⁵ with 78 data points for 75 different chemicals showing very good agreement across 3.75 orders of magnitude of k_R. Previous estimates of k_R (k_GM)²⁵ were obtained using a 2-compartment model (i.e., Fig. 1B). Conceptually the differences between the previous approach and the current approach is that in the current approach k_R is assumed to occur before absorption in the gut epithelium, whereas in the 2-compartment model it is explicitly possible for chemical transport from the perfused tissues of the fish to the GIT where it could be subject to degradation in the lumen. The greatest differences between the two methods of k_R (k_GM) estimation were for 4 chemicals with slow rates of degradation <∼0.2 d⁻¹ (Fig. 3B, bottom left corner). The general agreement using two different approaches for k_R estimation provides confidence in the estimates, including the empirical estimates selected to parameterize the new model. To improve k_R estimation, future experiments should use standardized testing methods (i.e., OECD 305) that include benchmark chemicals to address inherent variability.^24,74

Fig. 3 (A) A histogram of the new biotransformation rate (k_R) database (n = 255) and (B) corroboration of the new k_R estimates with other previous estimates²⁵ for a subset of the new database (n = 78).

The estimated k_Rs cannot be associated with specific degradation process(es) (e.g., biodegradation by microorganisms, hydrolysis, biotransformation from the secretion of enzymes, biotransformation in the epithelial tissue) or the precise location(s) of these processes (e.g., lumen, tissues, first pass effect in the liver). Because these rates are not being measured directly they are assumed and other factors could also be highly relevant. In particular, if there are key determinants other than K_OW that significantly influence chemical absorption efficiency (e.g., steric factors, or negligible solubility in water or organic phases, or active processes for absorption or secretion) the k_R estimates for these chemicals will be suspect.

3.4 Simulating field BAFs and field and lab BMFs

Fig. 4 illustrates AQUAWEB model³³ predicted field BAFs for a lower trophic level fish for a range of K_OW and a range of assumed k_B and k_R values. Fig. ESI-3† illustrates the BCFBAF model predictions for BAFs in an upper trophic level fish. One difference in model output between Fig. 4A (and ESI-3A†) and Fig. 4B (and ESI-3B†) is that rates for k_R are considered in the latter figures. The original E_D model used in the bioaccumulation models is included in Fig. 4A and ESI-3A.† The new E_D model shows higher BAFs at higher K_OW compared to the BAFs using the original E_D model, particularly when biotransformation rates are very slow. In Fig. 4Bk_R is assumed equal to k_B,N for most calculations. There are some differences with respect to bioaccumulation hazard classification with slight decreases in BAFs when k_R ∼ k_B,N. When k_B,N is about 0.5 per day, the BAFs are near or below the criterion of 5000 L kg_ww⁻¹ regardless of k_R. There are notable differences in potential bioaccumulation hazard categorization if k_R ≫ k_B,N. For example, if k_B,N and k_R are about 0.1 d⁻¹ then chemicals with log K_OW > ∼5 and <∼9 would have BAFs >5000 L kg⁻¹; however, if k_R is about 100 times faster than k_B,N (0.1 d⁻¹) then all of the chemicals would have BAFs <5000 L kg⁻¹ under these simulated conditions.

Fig. 4 Bioaccumulation factor (BAF) calculations as a function of the octanol–water partition coefficient (K_OW). (A) Assumptions for fish biotransformation rate constants (k_B,N; d⁻¹) and degradation in GIT assumed negligible, and (B) assumptions for degradation rate constants in the GIT k_R, i.e., k_R = k_B,N and k_R = 100k_B,N. BAF criteria of 1000 and 5000 L kg⁻¹ are presented. BAFs are based on total water (tw) concentrations (dissolved and sorbed).

Fig. 5 compares AQUAWEB model calculations for the BMF for the same fish simulated in Fig. 4 for a range of hypothetical chemicals under field and lab conditions. In these examples k_R is assumed equal to k_B,N for all calculations. The BMF threshold of 1 kg_lw kg_lw⁻¹ is included to illustrate the combination of some chemical properties and degradation rates that result in different bioaccumulation hazard classifications. Fig. 5A simulates a field BMF in which the diet is at equilibrium with the water and the fish is simultaneously exposed to chemical in both media and Fig. 5B simulates a laboratory BMF in which the fish is only exposed to chemical in the diet (no chemical in the water).

Fig. 5 Biomagnification factor (BMF, lipid weight ratios) calculations as a function of the octanol–water partition coefficient (K_OW) and assumptions for fish biotransformation rate constants (k_B,N; d⁻¹) and degradation rate constants in the GIT k_R. (A) Field BMFs: the diet is at equilibrium with the water, and (B) lab BMFs: fish is only exposed to chemical in the diet (no water exposure). All lab simulations include growth correction, i.e., k_G = 0, whereas the field BMFs include simulated growth rates (k_G ≅ 0.0022 d⁻¹ in this case). A BMF criterion of 1 is presented (red line).

Fig. 5A shows how the field BMF is higher at higher K_OW when the new E_D model is considered (reflecting differences between the new and old E_D models shown in Fig. 2). Furthermore the maximum field BMF (k_B,N = k_R = 0) under the assumed conditions (i.e., 10 g fish; 15 °C; k_G = 0.0022 d⁻¹) with the new E_D model is about 6.7 compared to about 5 with the earlier E_D model and the upper range of K_OW corresponding with BMFs >1 increases from about 8.5 to 10. Fig. 5A suggests that for this specific predator–prey relationship a k_B,N of about 0.05 per day results in BMFs ≤1 for all neutral hydrophobic organic chemicals. When k_R is about 0.05 d⁻¹ the impact on E_D is relatively small, i.e., in the log K_OW range of ∼4 to ∼6.5 E_D decreases from about 51% to 50%. Hence, in these cases, the decrease in biomagnification hazard classification is a result of k_B,N not k_R. The role of k_R on bioaccumulation endpoints may be more significant if k_R ≫ k_B,N (e.g., Fig. 4B).

Fig. 5B shows how the laboratory BMF is also higher at higher K_OW when the new E_D model is considered compared to previous models for E_D. These simulated lab BMFs are growth corrected as recommended in OECD guidelines,²⁷i.e., the growth rate constant in the model (k_G) is set to zero. There are no declines in maximum BMF values at higher K_OW as occurs when there is some growth even in the absence of biotransformation (Fig. 5A), see also.⁴² The maximum lab BMF (k_B,N = k_R = k_G = 0) under the assumed conditions with the new E_D model is about 10 as compared to about 7 with the earlier E_D model. The maximum lab BMF compared is higher than the maximum field BMF primarily due to growth correction, i.e., “k_G = 0” for the lab BMF.

There are notable differences between modelled field BMFs (Fig. 5A) and lab BMFs (Fig. 5B) for the “same fish”. At lower K_OW the influence of chemical uptake from the water results in field BMFs ∼1 for chemicals with log K_OW < ∼4 regardless of the assumed biotransformation rates. BMFs in the field approximate unity at lower K_OW because the predator, prey and water are near equilibrium (no fugacity gradient from fish to water). Chemical exchange kinetics between the water and the organism are relatively fast and equilibrium is approximated. However, the lab BMFs are <1 at lower K_OW because the fish is not simultaneously exposed to chemical in the water and it can efficiently eliminate chemical to its surrounding environment because of the fugacity gradient from fish to water. For poorly biotransformed chemicals, field BMFs exceed 1 (biomagnification hazard criterion) at a lower K_OW than lab BMFs because of the co-exposure to chemical in the water. Essentially the decreased fugacity gradient between the fish and the water in the environment reduces the flux of chemical from the fish to the water. This in turn shifts the BMF threshold to a lower K_OW than would be observed in the laboratory. Furthermore, the reduced chemical flux from the organism to the water also influences the degree of biotransformation required to lower concentrations in the organism at higher K_OW. For example, in the lab setting a k_B,N = k_R of 0.03 d⁻¹ is sufficient to reduce BMFs for all chemicals <1; whereas chemicals with log K_OW > ∼4.5 and <∼8.0 still show some biomagnification under environmental conditions with this rate of biotransformation. All else being equal, slightly faster rates of biotransformation may be required in the field than in the lab to lower BMFs. Predictions of E_D values corresponding to lab BMFs <1 (e.g., E_D ≤ 0.13 ²⁵) should further consider field exposure conditions for field BMFs.

Mass balance models can be used to translate toxicokinetic and bioaccumulation parameters derived from laboratory studies to field conditions with information for exposure water concentrations to simulate environmentally relevant exposure conditions. The relative ratio of the chemical concentrations in the diet and the water and the relative uptake rates constants (or fluxes, e.g., N_U and N_D in Fig. 1A) can be different at different trophic levels and these can be simulated with models and corroborated through field studies. The depicted rate constants and relationships between K_OW, BMF and BAF in Fig. 4 and 5 are a function of specific conditions including fish mass (volume), water temperature, lipid contents, trophic position and organic carbon content in the water (for BAFs) among others. Hence extrapolating the rate constants directly to other conditions requires the revision of parameters to reflect the appropriate situation.

Finally, it is worth mentioning that the BMFs and BAFs in these figures are, by definition, at steady-state. For persistent, high K_OW chemicals total elimination rates are very slow and the time to approach steady-state may take several years often exceeding the lifespan of the organism.

4. Conclusions

There are relatively few measured E_D data compared to the number of chemicals requiring evaluation. The new E_D model provides a mechanistic basis for simulating chemical absorption after dietary ingestion of chemicals in fish. The model can be used to generate hypotheses to test the influence of flow rates, digestion efficiencies, diet composition, and other related factors on E_D. Calibration to an evaluated database (n = 250) of measured E_Ds for persistent chemicals provides some confidence in the model's application for bioaccumulation and exposure assessments; however, the underlying variability in the data (and hence uncertainty in the models) requires consideration. The model simulations indicate the role of k_R on the BMF and BAF can be important for bioaccumulation hazard screening, only when k_R ≫ k_B. The present study illustrates some combinations of K_OW, k_R and k_B,N and possible outcomes for bioaccumulation assessment and there are thousands of such combinations for various trophic levels. The new database of 255 k_Rs may lead to the subsequent development and evaluation of quantitative structure–activity relationships (QSARs) to predict k_R. Obtaining additional reliable quality E_D, F and k_R data from laboratory tests for a range of chemical structures would expand the “chemical space” of the current k_R database and increase the applicability domain of subsequent QSARs.

Models that do not explicitly include a liver compartment such as AQUAWEB,³³ BCFBAF³² and others e.g., ref. 24, 30 and 74, cannot explicitly simulate first-pass effects or determine precisely where degradation is occurring throughout the dietary exposure pathway through to systemic circulation. Multi-compartment PBPK models, e.g.ref. 13 and tissue specific analyses (i.e., lumen of the GIT, perfused GIT tissues/epithelium, liver, other tissues) are required to differentiate compartment-specific rates of degradation. Future studies that combine tissue-specific in vitro biotransformation rate measurements with PBPK models may better determine details that cannot be obtained using the present methods. The application of in vitro tests and appropriate extrapolation models will provide insights into the relative rates of reaction in the GIT and liver; however, for GIT experiments it will also be important to further determine whether the reactions are occurring in the lumen, i.e., microflora and excreted enzymes, or in the epithelial tissues where biotransformation enzymes are produced and located.

There are merits and limitations of the new E_D model and the associated k_R estimates. The new E_D model is readily implemented in bioaccumulation and toxicokinetic models such as AQUAWEB and BCFBAF. The revised models require only the additional k_R parameter, if k_R data are deemed necessary to refine the model calculations and chemical assessments. This may be particularly beneficial when more complex multi-compartment models are not available or are not readily parameterized. If k_R data are not available the E_D model will provide conservative calculations (assuming no significant degradation in the GIT) appropriate for initial screening assessments. One-compartment models are highly relevant because predators consuming prey in the environment ingest entire organisms, including their GIT. Also, most field and laboratory measurements are based on whole body concentrations (including the GIT). One-compartment models are sufficient to address bioaccumulation assessments for many chemicals; however, in some cases more resolved models and data may be required to address uncertainty. A tiered framework could be developed by including different levels of model complexity (e.g., Fig. 1) that can be used based on the assessment contexts (problem formulation, uncertainty, outcomes) and data availability. Furthermore, comparisons of output from a suite of models can be used to hypothesize when more sophisticated models and more measured data are required and could thus systematically guide testing and research.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge funding support for this work provided by ILSI-Health and Environmental Sciences Institute (HESI) and Environment and Climate Change Canada. We thank HESI project monitoring team members (Mark Bonnell, Tom Parkerton, Lawrence Burkhard, Dan Merckel, Michelle Embry, Kent Woodburn and Ken Drouillard) for comments on earlier drafts.

References

1. Government of Canada, Canadian Environmental Protection Act, Canada Gazette Part III, 1999.
2. Environment Canada, Guidance Manual for the Categorization of Organic and Inorganic Substances on Canada's Domestic Substances List, Existing Substances Branch, Environment Canada, Ottawa, ON, 2003.
3. European Parliament, Regulation (EC) 1907/2006 of the European Parliament and of the council of 18 December 2006 concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH), establishing a European Chemicals Agency, amending Directive 1999/45/EC and repealing Council Regulation (EEC) 793/93 and Commission Regulation (EC) 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC, Official Journal of the European Union, 30.12.2006. L 396/1-849, 2006.
4. J. Costanza, D. G. Lynch, R. S. Boethling and J. A. Arnot, Use of the bioaccumulation factor to screen chemicals for bioaccumulation potential, Environ. Toxicol. Chem., 2012, 31, 2261–2268.
5. P. Qiao, F. A. P. C. Gobas and A. P. Farrell, Relative contributions of aqueous and dietary uptake of hydrophobic chemicals to the body burden in juvenile rainbow trout, Arch. Environ. Contam. Toxicol., 2000, 39, 369–377.
6. R. V. Thomann, J. P. Connolly and T. F. Parkerton, An equilibrium model of organic chemical accumulation in aquatic food webs with sediment interaction, Environ. Toxicol. Chem., 1992, 11, 615–629.
7. K. E. Clark, F. A. P. C. Gobas and D. Mackay, Model of organic chemical uptake and clearance from fish from food and water, Environ. Sci. Technol., 1990, 24, 1203–1213.
8. L. P. Burkhard, J. A. Arnot, M. R. Embry, K. J. Farley, R. A. Hoke, M. Kitano, H. A. Leslie, G. R. Lotufo, T. F. Parkerton, K. G. Sappington, G. T. Tomy and K. B. Woodburn, Comparing laboratory- and field-measured biota-sediment accumulation factors, Integr. Environ. Assess. Manage., 2012, 8, 32–41.
9. F. A. P. C. Gobas, W. de Wolf, L. P. Burkhard, E. Verbruggen and K. Plotzke, Revisiting bioaccumulation criteria for POPs and PBT assessments, Integr. Environ. Assess. Manage., 2009, 5, 624–637.
10. J. A. Arnot, D. Mackay and M. Bonnell, Estimating metabolic biotransformation rates in fish from laboratory data, Environ. Toxicol. Chem., 2008, 27, 341–351.
11. L. P. Burkhard, Factors influencing the design of bioaccumulation factor and biota-sediment accumulation factor field studies, Environ. Toxicol. Chem., 2003, 22, 351–360.
12. W. de Wolf, W. Seinen and J. L. M. Hermens, Biotransformation and toxicokinetics of trichloroanilines in fish in relation to their hydrophobicity, Arch. Environ. Contam. Toxicol., 1993, 25, 110–117.
13. J. W. Nichols, P. N. Fitzsimmons and L. P. Burkhard, In vitro–in vivo extrapolation of quantitative hepatic biotransformation data for fish. II. Modeled effects on chemical bioaccumulation, Environ. Toxicol. Chem., 2007, 26, 1304–1319.
14. D. Mackay, A. K. D. Celsie, J. A. Arnot and D. E. Powell, Processes influencing chemical biomagnification and trophic magnification factors in aquatic ecosystems: Implications for chemical hazard and risk assessment, Chemosphere, 2016, 154, 99–108.
15. J. Kim, F. A. P. C. Gobas, J. A. Arnot, D. E. Powell, R. M. Seston and K. B. Woodburn, Evaluating the roles of biotransformation, spatial concentration differences, organism home range, and field sampling design on trophic magnification factors, Sci. Total Environ., 2016, 551–552, 438–451.
16. D. M. Walters, T. D. Jardine, B. S. Cade, K. A. Kidd, D. C. G. Muir and P. Leipzig-Scott, Trophic magnification of organic chemicals: A global synthesis, Environ. Sci. Technol., 2016, 50, 4650–4658.
17. U. Fagerholm, Prediction of human pharmacokinetics-gut-wall metabolism, J. Pharm. Pharmacol., 2007, 59, 1335–1343.
18. K. M. Kleinow, M. O. James, Z. Tong and C. S. Venugopalan, Bioavailability and biotransformation of benzo(a)pyrene in an isolated perfused in situ catfish intestinal preparation, Environ. Health Perspect., 1998, 106, 155–166.
19. T. Sousa, R. Paterson, V. Moore, A. Carlsson, B. Abrahamsson and A. W. Basit, The gastrointestinal microbiota as a site for the biotransformation of drugs, Int. J. Pharm., 2008, 363, 1–25.
20. S. Cavret, Intestinal metabolism of PAH: in vitro demonstration and study of its impact on PAH transfer through the intestinal epithelium, Environ. Res., 2005, 98, 22–32.
21. C. M. Butt, D. C. G. Muir and S. A. Mabury, Biotransformation of the 8:2 fluorotelomer acrylate in rainbow trout. 2. In vitro incubations with liver and stomach S9 fractions, Environ. Toxicol. Chem., 2010, 29, 2736–2741.
22. P. A. Van Veld, Absorption and metabolism of dietary xenobiotics by the intestine of fish, Rev. Aquat. Sci., 1990, 2, 185–203.
23. M. O. James, C. J. Marth and L. Rowland-Faux, Slow O-demethylation of methyl triclosan to triclosan, which is rapidly glucuronidated and sulfonated in channel catfish liver and intestine, Aquat. Toxicol., 2012, 72–82.
24. J. C. Lo, D. A. Campbell, C. J. Kennedy and F. A. P. C. Gobas, Somatic and gastrointestinal in vivo biotransformation rates of hydrophobic chemicals in fish, Environ. Toxicol. Chem., 2015, 34, 2282–2294.
25. J. C. Lo, D. J. Letinski, T. F. Parkerton, D. A. Campbell and F. A. P. C. Gobas, In vivo biotransformation rates of organic chemicals in fish: relationship with bioconcentration and biomagnification factors, Environ. Sci. Technol., 2016, 50, 13299–13308.
26. W. A. Bruggeman, L. B. J. M. Matron, D. Kooiman and O. Hutzinger, Accumulation and elimination kinetics of di-, tri-, and tetra chlorobiphenyls by goldfish after dietary and aqueous exposure, Chemosphere, 1981, 10, 811–832.
27. OECD, OECD Guidelines for Testing Chemicals. Test No. 305: Bioaccumulation in Fish: Aqueous and Dietary Exposure, Organization for Economic Co-operation and Development, Paris, 2012.
28. R. V. Thomann, Bioaccumulation model of organic chemical distribution in aquatic food chains, Environ. Sci. Technol., 1989, 23, 699–707.
29. M. C. Barber, L. A. Suarez and R. R. Lassiter, Modelling bioaccumulation of organic pollutants in fish with an application to pcbs in Lake Ontario salmonids, Can. J. Fish. Aquat. Sci., 1991, 48, 318–337.
30. M. Hauck, H. W. M. Hendriks, M. A. J. Huijbregts, A. M. J. Ragas, D. van de Meent and A. J. Hendriks, Parameter uncertainty in modeling bioaccumulation factors of fish, Environ. Toxicol. Chem., 2011, 30, 403–412.
31. F. A. P. C. Gobas, A model for predicting the bioaccumulation of hydrophobic organic chemicals in aquatic food-webs: Application to Lake Ontario, Ecol. Modell., 1993, 69, 1–17.
32. J. A. Arnot and F. A. P. C. Gobas, A generic QSAR for assessing the bioaccumulation potential of organic chemicals in aquatic food webs, QSAR Comb. Sci., 2003, 22, 337–345.
33. J. A. Arnot and F. A. P. C. Gobas, A food web bioaccumulation model for organic chemicals in aquatic ecosystems, Environ. Toxicol. Chem., 2004, 23, 2343–2355.
34. M. C. Barber, Dietary uptake models used for modeling the bioaccumulation of organic contaminants in fish, Environ. Toxicol. Chem., 2008, 27, 755–777.
35. F. A. P. C. Gobas, D. C. G. Muir and D. Mackay, Dynamics of dietary bioaccumulation and faecal elimination of hydrophobic organic chemicals in fish, Chemosphere, 1988, 17, 943–962.
36. G. A. Moser and M. S. McLachlan, Modeling digestive tract absorption and desorption of lipophilic organic contaminants in humans, Environ. Sci. Technol., 2002, 36, 3318–3325.
37. K. G. Drouillard and R. J. Norstrom, Dietary absorption efficiencies and toxicokinetics of polychlorinated biphenyls in ring doves following exposure to Aroclor (R) mixtures, Environ. Toxicol. Chem., 2000, 19, 2707–2714.
38. B. C. Kelly, F. A. P. C. Gobas and M. S. McLachlan, Intestinal absorption and biomagnification of organic contaminants in fish, wildlife, and humans, Environ. Toxicol. Chem., 2004, 23, 2324–2336.
39. B. C. Kelly, M. G. Ikonomou, J. D. Blair, A. E. Morin and F. A. P. C. Gobas, Food web-specific biomagnification of persistent organic pollutants, Science, 2007, 317, 236–239.
40. J. A. Arnot and D. Mackay, Policies for chemical hazard and risk priority setting: can persistence, bioaccumulation, toxicity and quantity information be combined?, Environ. Sci. Technol., 2008, 42, 4648–4654.
41. G. Czub and M. S. McLachlan, A food chain model to predict the levels of lipophilic organic contaminants in humans, Environ. Toxicol. Chem., 2004, 23, 2356–2366.
42. J. A. Arnot and C. L. Quinn, Development and evaluation of a database of dietary bioaccumulation test data for organic chemicals in fish, Environ. Sci. Technol., 2015, 49, 4783–4796.
43. R. Xiao, J. A. Arnot and M. MacLeod, Towards an improved understanding of processes controlling absorption efficiency and biomagnification of organic chemicals by fish, Chemosphere, 2015, 138, 89–95.
44. H. Selck, K. Drouillard, K. Eisenreich, A. A. Koelmans, A. Palmqvist, A. Ruus, D. Salvito, I. Schultz, R. Stewart, A. Weisbrod, N. W. van den Brink and M. van den Heuvel-Greve, Explaining differences between bioaccumulation measurements in laboratory and field data through use of a probabilistic modeling approach, Integr. Environ. Assess. Manage., 2012, 8, 42–63.
45. D. Mackay, Multimedia Environmental Models: The Fugacity Approach - Second Edition, Lewis Publishers, Boca Raton, FL, 2001.
46. J. W. Nichols, A physiologically based toxicokinetic model for dietary uptake of hydrophobic organic compounds by fish: I. Feeding studies with 2,2′,5,5′-tetrachlorobiphenyl, Toxicol. Sci., 2004, 77, 206–218.
47. J. W. Nichols, A physiologically based toxicokinetic model for dietary uptake of hydrophobic organic compounds by fish: II. Simulation of chronic exposure scenarios, Toxicol. Sci., 2004, 77, 219–229.
48. J. W. Nichols, J. M. McKim, M. E. Andersen, M. L. Gargas, H. J. Clewell III and R. J. Erickson, A physiology based toxicokinetic model for the uptake and disposition of waterborne organic chemicals in fish, Toxicol. Appl. Pharmacol., 1990, 106, 433–447.
49. J. W. Nichols, J. M. McKim, G. J. Lien, A. D. Hoffman and S. L. Bertelsen, Physiologically based toxicokinetic modeling of three chlorinated ethanes in rainbow trout (Oncorhynchus mykiss), Toxicol. Appl. Pharmacol., 1991, 110, 374–389.
50. F. A. P. C. Gobas, J. R. McCorquodale and G. D. Haffner, Intestinal absorption and biomagnification of organochlorines, Environ. Toxicol. Chem., 1993, 12, 567–576.
51. K. Woodburn, K. Drottar, J. Domoradzki, J. Durham, D. McNett and R. Jezowski, Determination of the dietary biomagnification of octamethylcyclotetrasiloxane and decamethylcyclopentasiloxane with the rainbow trout (Oncorhynchus mykiss), Chemosphere, 2013, 93, 779–788.
52. B. J. Konwick, A. W. Garrison, J. K. Avants and A. T. Fisk, Bioaccumulation and biotransformation of chiral triazole fungicides in rainbow trout (Oncorhynchus mykiss), Aquat. Toxicol., 2006, 80, 372–381.
53. G. T. Tomy, V. P. Palace, T. Halldorson, E. Braekevelt, R. Danell, K. Wautier, B. Evans, L. Brinkworth and A. T. Fisk, Bioaccumulation, biotransformation, and biochemical effects of brominated diphenyl ethers in juvenile lake trout (Salvelinus namaycush), Environ. Sci. Technol., 2004, 38, 1496–1504.
54. C. S. Wong, F. Lau, M. Clark, S. A. Mabury and D. C. G. Muir, Rainbow trout (Oncorhynchus mykiss) can eliminate chiral organochlorine compounds enantioselectively, Environ. Sci. Technol., 2002, 36, 1257–1262.
55. W. L. Hayton, Rate-limiting barriers to intestinal drug absorption: a review, J. Pharmacokinet. Biopharm., 1980, 8, 321–334.
56. G. L. Flynn and S. H. Yalkowsky, Correlation and prediction of mass transport across membranes I: Influence of alkyl chain length on flux-determining properties of barrier and diffusant, J. Pharm. Sci., 1972, 61, 838–852.
57. G. L. Flynn, S. H. Yalkowsky and T. J. Roseman, Mass transport phenomena and models: theoretical concepts, J. Pharm. Sci., 1974, 63, 479–510.
58. A. J. Hendriks, A. van der Linde, G. Cornelissen and D. T. H. M. Sijm, The power of size. 1. Rate constants and equilibrium ratios for accumulation of organic substances related to octanol-water partition ratio and species weight, Environ. Toxicol. Chem., 2001, 20, 1399–1420.
59. D. W. Hawker and D. W. Connell, Octanol-water partition coefficients of polychlorinated biphenyl congeners, Environ. Sci. Technol., 1988, 22, 382–387.
60. U.S. EPA, Estimation Programs Interface (EPI) Suite for Microsoft® Windows, Ver. 4.1, U. S. Environmental Protection Agency, Washington, D.C., October 2011.
61. J. A. Arnot, D. Mackay, T. F. Parkerton and M. Bonnell, A database of fish biotransformation rates for organic chemicals, Environ. Toxicol. Chem., 2008, 27, 2263–2270.
62. S. B. Gewurtz, R. Laposa, N. Gandhi, G. N. Christensen, A. Evenset, D. Gregor and M. L. Diamond, A comparison of contaminant dynamics in arctic and temperate fish: A modeling approach, Chemosphere, 2006, 63, 1328–1341.
63. F. A. P. C. Gobas and J. A. Arnot, Food web bioaccumulation model for polychlorinated biphenyls in San Francisco Bay, California, USA, Environ. Toxicol. Chem., 2010, 29, 1385–1395.
64. A. M. McLeod, J. A. Arnot, K. Borga, H. Selck, D. R. Kashian, A. Krause, G. Paterson, G. D. Haffner and K. G. Drouillard, Quantifying uncertainty in the trophic magnification factor related to spatial movements of organisms in a food web, Integr. Environ. Assess. Manage., 2015, 11, 306–318.
65. D. Mackay, A. K. D. Celsie, J. M. Parnis, L. S. McCarty, J. A. Arnot and D. E. Powell, The chemical exposure toxicity space (CETS) model: Displaying exposure time, aqueous and organic concentration, activity, and onset of toxicity, Environ. Toxicol. Chem., 2017, 36, 1389–1396.
66. D. Mackay, L. S. McCarty and J. A. Arnot, Relationships between exposure and dose in aquatic toxicity tests for organic chemicals, Environ. Toxicol. Chem., 2014, 33, 2038–2046.
67. J. Stadnicka, K. Schirmer and R. Ashauer, Predicting concentrations of organic chemicals in fish by using toxicokinetic models, Environ. Sci. Technol., 2012, 46, 3273–3280.
68. K. Gerber, User's Guide and Technical Documentation KABAM Version 1.0 (K_OW (based) Aquatic Bioaccumulation Model). Report of Environmental Fate and Effects Division. Office of Pesticide Programs, US Environmental Protection Agency, Washington, DC, 2009.
69. U.S. EPA, Estimation Programs Interface (EPI) Suite for Microsoft® Windows, Ver. 4.1, U. S. Environmental Protection Agency, Washington, D.C., October 2011.
70. F. A. P. C. Gobas and J. A. Arnot, Categorization of Organic Substances on the Domestic Substances List for Bioaccumulation Potential. Report for Environment Canada, Simon Fraser University, Burnaby, BC, 2003.
71. I. A. O'Connor, M. A. J. Huijbregts, A. M. J. Ragas and A. J. Hendriks, Predicting the oral uptake efficiency of chemicals in mammals: Combining the hydrophilic and lipophilic range, Toxicol. Appl. Pharmacol., 2013, 266, 150–156.
72. J. A. Arnot, W. Meylan, J. Tunkel, P. H. Howard, D. Mackay, M. Bonnell and R. S. Boethling, A quantitative structure–activity relationship for predicting metabolic biotransformation rates for organic chemicals in fish, Environ. Toxicol. Chem., 2009, 28, 1168–1177.
73. OECD, Series on Testing and Assessment: Ecotoxicity Testing, No. 175. Part I - Validation Report of a Ring Test for the OECD 305 Dietary Exposure Bioaccumulation Fish Test and Part II - Additional Report including Comparative Analysis of Trout and Carp Results, Environment Directorate, Organization for Economic Co-operation and Development, Paris, 2012.
74. R. Xiao, M. Adolfsson-Erici, G. Åkerman, M. S. McLachlan and M. MacLeod, A benchmarking method to measure dietary absorption efficiency of chemicals by fish, Environ. Toxicol. Chem., 2013, 32, 2695–2700.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7em00539c

---