Cross-Species Extrapolation of Uptake and ... - ACS Publications

Jan 21, 2016 - depuration will approximately follow first-order kinetics (i.e., for nonpolar ... in European water bodies and ubiquitously distributed...
61 downloads 3 Views 2MB Size
Download PDF
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article pubs.acs.org/est

Cross-Species Extrapolation of Uptake and Disposition of Neutral Organic Chemicals in Fish Using a Multispecies Physiologically-Based Toxicokinetic Model Framework Markus Brinkmann,*,† Christian Schlechtriem,§ Mathias Reininghaus,† Kathrin Eichbaum,† Sebastian Buchinger,∥ Georg Reifferscheid,∥ Henner Hollert,†,⊥,#,∇ and Thomas G. Preuss‡,○ †

Department of Ecosystem Analysis, Institute for Environmental Research, ABBt − Aachen Biology and Biotechnology and Department of Environmental Biology and Chemodynamics, Institute for Environmental Research, ABBt − Aachen Biology and Biotechnology, RWTH Aachen University, Aachen, 52074 Germany § Department of Ecotoxicology, Fraunhofer Institute for Molecular Ecology (IME), Schmallenberg, 57392 Germany ∥ Department G3: Biochemistry and Ecotoxicology, Federal Institute of Hydrology (BFG), Koblenz, 56068 Germany ⊥ State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, 210023 China # College of Resources and Environmental Science, Chongqing University, Chongqing, 400030 China ∇ Key Laboratory of Yangtze Water Environment, Ministry of Education, Tongji University, Shanghai, 200092 China ‡

S Supporting Information *

ABSTRACT: The potential to bioconcentrate is generally considered to be an unwanted property of a substance. Consequently, chemical legislation, including the European REACH regulations, requires the chemical industry to provide bioconcentration data for chemicals that are produced or imported at volumes exceeding 100 tons per annum or if there is a concern that a substance is persistent, bioaccumulative, and toxic. For the filling of the existing data gap for chemicals produced or imported at levels that are below this stipulated volume, without the need for additional animal experiments, physiologically-based toxicokinetic (PBTK) models can be used to predict whole-body and tissue concentrations of neutral organic chemicals in fish. PBTK models have been developed for many different fish species with promising results. In this study, we developed PBTK models for zebrafish (Danio rerio) and roach (Rutilus rutilus) and combined them with existing models for rainbow trout (Onchorhynchus mykiss), lake trout (Salvelinus namaycush), and fathead minnow (Pimephales promelas). The resulting multispecies model framework allows for cross-species extrapolation of the bioaccumulative potential of neutral organic compounds. Predictions were compared with experimental data and were accurate for most substances. Our model can be used for probabilistic risk assessment of chemical bioaccumulation, with particular emphasis on cross-species evaluations.

1. INTRODUCTION Bioconcentration is the process that leads to the accumulation of water-borne chemicals within aquatic animals through nondietary exposure routes (i.e., via the gills and skin). More specifically, it reflects the balance between the two kinetic processes of uptake and elimination. If the rate of uptake of a substance is greater than its elimination rate, it can easily reach concentrations in aquatic organisms (e.g., fish) that are orders of magnitude above that in the surrounding water.1 A high bioconcentration potential is generally considered undesirable.2 Thus, chemical legislation, including the European REACH regulation (EC no. 1907/2006),3 requires the chemical industry to provide bioconcentration data for chemicals that are produced or imported at volumes exceeding 100 tons per © 2016 American Chemical Society

annum and for potential substances of very high concern (SVHCs). The bioconcentration potential of chemicals is typically assessed using fish in accordance with the OECD technical guideline 3054 and expressed as the bioconcentration factor (BCF) (i.e., the steady-state ratio between the concentration of a chemical in fish and the corresponding aqueous concentration). At an early stage in the bioaccumulation research field, it became obvious that the bioconcentration potential of many Received: Revised: Accepted: Published: 1914

December 16, 2015 January 16, 2016 January 21, 2016 January 21, 2016 DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

an individual fish species, leaving interspecies differences largely uncharacterized. Furthermore, a maximum level of standardization was applied during development of the OECD technical guideline 305,4 which makes use of relatively narrow length and weight classes for bioconcentration experiments. Consequently, mostly because it is not necessary within the PBT assessment process, intraspecies differences between animals of different weights or physiological conditions cannot be deduced from such data. To overcome these limitations, we developed new PBTK models for zebrafish (Danio rerio) and roach (Rutilus rutilus) and combined them with existing PBTK models for rainbow trout (Oncorhynchus mykiss),27 lake trout (Salvelinus namaycush),28 and fathead minnow (Pimephales promelas). 17 Common roach is one of the most representative fish species in European water bodies and ubiquitously distributed. It typically constitutes a large proportion of the fish biomass in a water body.29−31 Compared with typically used laboratory models, only a few studies have been conducted with this species.32−35 The zebrafish, however, is probably the most widely applied and accepted model fish species in ecotoxicological research.36 The aims of the present study were (a) to develop new PBTK models for zebrafish and roach and (b) for the first time combine them with existing PBTK models into a unified multispecies PBTK model framework for probabilistic assessments of the bioconcentration potential of chemicals, (c) to test the predictive power of this framework, and (d) to evaluate its usefulness for cross-species extrapolation of the bioaccumulation potential of selected neutral organics.

neutral organic chemicals was proportional to their lipophilicity, as expressed by the n-octanol−water partitioning coefficient, log Kow.5,6 This knowledge quickly resulted in the development of quantitative structure activity relationships (QSARs) that related a substance’s bioconcentration potential to its log Kow value using simple linear or bilinear relationships.7 These models predict bioconcentration reasonably well for poorly metabolized substances, with log Kow values ranging from 1 to 6,8 and have been used to define screening-level criteria. The European Commission proposed screening-level criteria for log Kow of 4.5−5 for bioaccumulative (B) substances (BCF ≥ 2000 L kg−1 wet weight, ww) and of 5−8 for very bioaccumulative (vB) substances (BCF ≥ 5000 L kg−1 ww).9 These criteria can be used efficiently during the identification of persistent, bioaccumulative, and toxic (PBT) chemicals to decide whether they are potentially bioaccumulative and whether experimental data is required.10 An important aspect of these considerations is the effort to minimize the use of experimental animals. The OECD technical guideline 305 allows for a minimized bioconcentration test with fewer sampling points if it can be expected that uptake and depuration will approximately follow first-order kinetics (i.e., for nonpolar organic compounds with an intermediate log Kow).4 These simplifications are all favorable from an ethical and economic perspective, and they also improve the efficiency of the PBT assessment process. In the context of most regulatory frameworks, bioaccumulation is considered to be an inherent property of the substance that is independent of the actual chemical concentration in the environment.11 Nonetheless, bioaccumulation can (and in some cases, should) be viewed with a special emphasis on exposure,12 particularly because bioaccumulation represents the link between the environmental concentration of a chemical and its internal concentration in exposed wildlife. The internal concentration of a compound is a key determinant of its biological effects and thus of interspecies differences in sensitivity.13,14 In the early 1990s, this knowledge led to the development of the critical body residue (CBR) concept,14 where median effect concentrations (EC50 values) based on aqueous exposure concentrations are multiplied by the BCF to give the internal concentration at the EC50 level. A central disadvantage of this approach is that the BCF does not contain any kinetic information but is rather a description of the steady state.15 Toxicokinetic models can be used to overcome this limitation and to achieve a more holistic perspective, using the experimental data currently generated for PBT assessments to its full potential. In particular, physiologically based toxicokinetic (PBTK) models can be applied to kinetically predict the whole-body bioconcentration, as well as the tissue-specific distribution of organic chemicals, in fish.16,17 PBTK models are already available for many fish species, including dogfish sharks,18 rainbow trout,16 brook trout,19 lake trout,20 channel catfish,21 fathead minnows,17 Japanese medaka,22 tilapia,23 and recently also zebrafish.24 We have previously demonstrated that PBTK models can be used efficiently to test hypotheses about the underlying physiological processes of bioconcentration.25,26 To our knowledge, no existing PBTK models for fish have previously been combined into a single model framework (i.e., a unified model structure that can be parametrized using a speciesspecific set of parameters for each individual species). This novel approach is warranted because the current PBT assessment strategy only requires bioconcentration data from

2. MATERIALS AND METHODS 2.1. Study Design. In the present study, we experimentally determined PBTK model parameters (total lipid and water content, as well as the volume of different tissues and organs) and used physiological data from the literature to establish new PBTK models for zebrafish and roach. We also reimplemented existing PBTK models for rainbow trout, lake trout, and fathead minnow (i.e., translated them into the programming language used) and combined all five models to form a unified multispecies PBTK model framework (i.e., a unified model structure of mass-balance differential equations that can be parametrized using species-specific parameters for each individual species that will produces a species-specific set of outputs from a defined number of model inputs). This framework was then used to predict species sensitivity distributions (SSDs) for BCFs and other bioaccumulation metrics in fish; these were then compared with SSDs determined using measured values from the literature. 2.2. Fish. Roach were obtained from pond aquaculture through a local supplier (Inquadro, Aachen, Germany) and held in aerated 1000 L tanks with a flow-through of dechlorinated municipal tap water (approximately 15 °C; pH 7.8 ± 0.2; NH3 < 0.1 mg L−1). The water was continuously exchanged at a rate of 0.5−1 full exchanges per day. Light and dark phases were 12 h each. Roach were fed daily ad libitum with frozen chironomid larvae (Aquahobby, Peine, Germany) and had an average length of 104 ± 19 mm and a body ww of 18.4 ± 12.6 g. Adult wild-type zebrafish were provided by the Fraunhofer Institute for Molecular Biology and Applied Ecology (IME; Aachen, Germany) and kept in glass aquaria (26 ± 1 °C; pH 7.8). Light and dark phases were maintained at 14 and 10 h, respectively. Zebrafish were fed daily ad libitum with commercial dry flakes (TetraMin; Tetra, Melle, Germany) and live Artemia sp. nauplii 1915

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

Table 1. Relative Tissue Wet Weight, Total Lipid Levels, and Total Water Content of Tissues from Zebrafish (Danio rerio) and Roach (Rutilus rutilus)a zebrafish tissue weight (% of body wet weight) liver kidney spleen muscle brain gill viscerab carcass a b

2.09 0.21 0.11 50.55 1.60 2.26 17.51 25.68

± ± ± ± ± ± ± ±

1.33 0.14 0.08 8.86 0.73 0.47 14.21 5.98

lipid content (% of tissue wet weight) 7.79 16.61 8.52 6.53 10.37 5.03 4.37 8.30

± ± ± ± ± ± ± ±

1.18 3.78 6.04 0.47 0.63 1.44 1.22 1.33

roach water content (% of tissue wet weight) 58.35 49.31 42.74 68.98 62.51 63.13 62.41 59.16

± ± ± ± ± ± ± ±

2.14 7.32 4.96 3.01 4.57 6.04 5.53 4.53

tissue weight (% of body wet weight) 0.82 0.41 0.36 47.29 0.96 2.23 5.76 42.18

± ± ± ± ± ± ± ±

0.13 0.13 0.09 4.80 0.30 0.39 1.35 3.50

lipid content (% of tissue wet weight) 2.54 2.16 1.68 1.11 7.28 1.60 2.56 3.04

± ± ± ± ± ± ± ±

0.49 1.74 0.11 0.21 0.62 0.31 0.70 0.55

water content (% of tissue wet weight) 71.76 72.36 59.20 79.52 73.89 80.89 76.83 73.45

± ± ± ± ± ± ± ±

1.21 1.28 8.22 0.92 1.21 0.55 0.95 1.07

Tissue wet weight was determined in 15 individual fish, while lipid and water levels were determined in three pools; n = 5 animals per species. Included the stomach, pyloric ceca, intestines, viscera-associated fat, and gonads.

2.4. Model Design, Formulation, and Implementation. 2.4.1. Model Parameters and Equations. All model parameters are summarized in Supplemental Table S1. The PBTK model parameters for fathead minnow, rainbow trout, and lake trout were taken from the literature.17,20,27 Most parameters for the zebrafish and roach PBTK models (tissue volumes and lipid and water levels) were determined experimentally, as described above. The volumes deduced from the weights of the liver and kidneys were directly used in the model. The volume of the richly perfused tissue compartment was estimated by summing the experimentally determined weights of the viscera, spleen, and gill. The volume of the poorly perfused tissue compartment (mainly white muscle) was assumed to be the difference between the total body volume and the volumes of all other compartments. Compartment volumes were expressed relative to the total body volume, while all compartments were assumed to have a specific gravity of 1.0. Data on blood flow distribution, as well as allometric relationships for effective respiratory volume and cardiac output, were unavailable for the two cyprinid species. These parameter values were, therefore, adopted from the fathead minnow (also a cyprinid fish species) PBTK model. Because perfusion data of the kidney was unavailable in the fathead minnow model, this compartment was also omitted from the other two cyprinid models. The model equations are summarized in the Supporting Information. 2.4.2. Implementation. The unified multispecies PBTK model framework was implemented in the object-oriented Delphi programming language using Rapid Application Development Studio XE2 (Embarcadero, Langen, Germany). Sections of the source code for the multispecies PBTK model framework can be made available upon request by the corresponding author. 2.4.3. Estimation of Model Performance. Model performance was tested using experimentally determined values for the internal compound concentrations in rainbow trout and fathead minnow, published by Stadnicka and co-workers.17 We performed a comprehensive literature search for previously published bioconcentration data in zebrafish and roach, using the ISI Web of Knowledge and Science Direct. This search yielded a sufficient number of data sets to allow comprehensive testing of the model performance for zebrafish (31 data points for 24 different chemicals), while only four data sets were available for roach. In all studies, the water was assumed to be saturated with oxygen during exposure, and the dissolved oxygen concentration was calculated according to Weiss.38

(Silver Star Artemia, Inter Ryba GmbH, Zeven, Germany). They had an average length of 34 ± 4 mm and a body weight of 0.9 ± 0.5 g (ww). All animals were used in accordance with the animal welfare act and with the permission of the federal authorities. 2.3. Experimental Derivation of Model Parameters. A number of parameters used for the zebrafish and roach PBTK models were determined experimentally. A total of 15 fish of each species were individually euthanized using a high concentration of benzocaine (ethyl 4-aminobenzoate; SigmaAldrich). The animals were dissected to isolate the following tissues and organs: liver, kidney, spleen, muscle, brain, gills, viscera, and carcass. The tissue and organ fractions excised from each individual were weighed (0.0001 g resolution) using an ultrafine balance (Scaltec, Heiligenstadt, Germany) and stored at −20 °C until the determination of their water and total lipid contents. Total lipid and water levels were determined in three thoroughly homogenized pools generated using five animals of each species (Table 1). The tissue water content was determined gravimetrically after dying at 105 °C for 48−96 h until a constant weight was reached. The total lipid contents of the tissue samples were determined gravimetrically using the extraction method described by Smedes.37 Samples were homogenized for 2 min in 18 mL cyclohexane−propan-2-ol (5:4, vol/vol) using an UltraTurrax homogenizer (IKA, Staufen, Germany). Subsequently, 11 mL of deionized water was added, and the samples were vortex-mixed for 15 s. The phases were separated by centrifuging the mixture for 5 min at 440−460g, and the upper cyclohexane phase was transferred to a storage vial. After the addition of 10 mL of a mixture of cyclohexane−propan-2-ol (87:13, w/w) to the remaining lower aqueous phase, the phases were again mixed for 15 s and then separated by centrifugation. The upper cyclohexane phase was transferred to the storage vial holding the first extract of the same sample, and the combined extracts were concentrated under a gentle nitrogen stream. The concentrated extract was quantitatively transferred into a preweighed wide-mouthed glass flask. The solvent in the flask was evaporated to dryness for 1 h at 105 °C. After cooling to room temperature in a desiccator, the dried flasks were weighed, and the lipid contents of the tissue samples were calculated and expressed as the percent of tissue ww. The solvent volumes used for lipid extraction were adjusted to the sample size, if necessary. Not more than 1 g (ww) sample was homogenized per 3.6 mL propan-2-ol and cyclohexane (5:4, vol/vol). 1916

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology Ideally, we used quality evaluated n-octanol−water partitioning coefficients (log Kow) obtained from the Sangster Research Laboratories LOGKOW database.39 If no such data were available, we used the values provided by the US-EPA EPI Suite.40 The full data sets used to assess model performance can be found in the Tables S2−S5. As a quantitative measure of model performance, we calculated the root mean squared error (RMSE) of the residuals as well as Spearman’s rank correlation coefficient (r). 2.4.4. SSDs for Bioconcentration Factors and Rate Constants. SSDs for the bioconcentration potential of six selected organic chemicals (i.e., dieldrin, dichloro diphenyltrichloroethane (DDT), pentachlorphenol (PCP), 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), hexachlorobenzene (HCB), and 1,4-dichlorobenzene (PDCB)) were modeled using the multispecies PBTK model framework and then compared with SSDs based on published data collected from the US-EPA ECOTOX database.41 To this end, the steady-state BCF (BCFSS) (i.e., the quotient of a chemical’s internal concentration in exposed fish and its aqueous concentration at steadystate) was calculated from the model predictions. The uptake rate constant (k1) and the depuration rate constant (k2) were calculated from predicted whole-body internal concentrations (during accumulation and depuration phase) of the respective chemicals according to OECD 305 using eq 1 and eq 2 (in the same way as commonly used for calculation of k1 and k2 from experimental internal concentrations).4 BCFSS =

parameters were assumed to be equal to those used for rainbow trout. It was assumed that fish did not grow during the experiments. All SSDs were generated using the US-EPA SSD Generator v1.

3. RESULTS AND DISCUSSION The aims of the present study were to (a) develop new PBTK models for zebrafish and roach; (b) combine them with existing models to form a multispecies PBTK model framework; (c) test its predictive power; and (d) evaluate its usefulness for cross-species extrapolation to estimate the bioaccumulation potential of selected neutral organic compounds. 3.1. Parameterization and Implementation of Two New PBTK Models. To parametrize the new PBTK models for zebrafish and roach, we determined total lipid and water contents, as well as the volumes of eight different tissues and organs (Table 1). Using these data in combination with data from previously published models, we successfully parametrized novel PBTK models for zebrafish (D. rerio) and roach (R. rutilus). All model parameters used can be found in Supporting Table S1. For zebrafish, we compiled a comprehensive data set consisting of the internal (i.e., whole-body) chemical concentrations (Cint) following aqueous exposure to individual organic chemicals, along with the corresponding exposure conditions (chemical concentration, temperature, dissolved oxygen, and fish weight). The data set comprised 31 data points for 24 different chemicals with log Kow values ranging from 0.90 to 4.53 (Table S2). The zebrafish model generally showed a low deviation from the experimental data (84% of the predicted concentrations deviated less than or equal to 5-fold from measured values), with a RMSE of 0.44 log units, a Spearman’s rank correlation coefficient (r) of 0.86 and a coefficient of determination of 0.74 (Figure 1). This performance was approximately equivalent to that of another zebrafish PBTK model recently reported by Pery and co-workers (the coefficient of determination was 0.83, and 88% of the predictions were within a factor of 5 compared to the measured values).24 This model by Pery et al.24 mainly differed from the model developed within this study in the fact that the authors developed two distinct sets of parameters for male and female fish. Although roach represents one of the most abundant and ecologically relevant cyprinids in European watersheds, relevant bioconcentration data were only available for three compounds: 17α-ethinylestradiol, pentachlorophenol, and a technical nonylphenol mixture (Table S3). For these three compounds, however, the model predicted internal concentrations that were consistent with the published data, with an RMSE of 0.17 log units (data not shown). 3.2. Reimplementation of Existing PBTK Models. As part of our attempt to establish a multispecies PBTK model framework, we additionally reimplemented existing PBTK models for rainbow trout26 (O. mykiss), fathead minnow17 (P. promelas) and lake trout28 (S. namaycush), which had been previously published. The performance of the reimplemented PBTK models for rainbow trout and fathead minnow was evaluated using a data set published by Stadnicka and coworkers.17 For both species, the data set consisted of Cint values following aqueous exposure to single organic chemicals and the corresponding exposure conditions (chemical concentration, temperature, dissolved oxygen, and fish weight). The data set for rainbow trout comprised 39 data points for 24 different

Cf,SS Cw,SS

Cf (t ) = Cw(t ) ·

(eq 1)

k1 ·(1 − e−k 2·t ) k2

(eq 2)

where Cf(t) and Cw(t) were the chemical’s concentration in fish and water, respectively, at time t, and Cf,SS and Cw,SS were the concentrations in fish and water at steady state. The depuration rate constant (k2) was determined prior to k1 as the slope of a straight line fitted to ln-transformed Cf(t) plotted versus t. Input parameters for the model predictions (body ww, total lipid content, exposure temperature, dissolved oxygen concentration) were based on default values provided for the respective fish species in OECD guideline 305 (Table 2).4 Because no default values were available for lake trout, the Table 2. Default Model Inputs Used in the Multi-Species Physiologically Based Toxicokinetic Modelling Framework for Body Wet Weight (ww), Water Temperature (T), Dissolved Oxygen Concentration (Cox), and Lipid Content (lipid) ww

T

Coxa

lipid

species

kg

°C

mg L−1

fraction w.w.

roach (Rutilus rutilus) fathead minnow (Pimephales promelas) zebrafish (Danio rerio) rainbow trout (Oncorhynchus mykiss) lake trout (Salvelinus namaycush)

0.0180 0.0015

15.0 22.0

10.06 8.71

0.021 0.050

0.0009 0.0050

22.0 15.0

8.71 10.06

0.067 0.085

0.0050

15.0

10.06

0.085

a

Water was assumed to be 100% saturated with oxygen, and the dissolved oxygen concentration was calculated according to Weiss.38 1917

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

of the combined model, we predicted BCFss, BCFliver, k1 and k2 for each fish species and used them to generate the SSD for a generic chemical with log Kow = 5.5 (Figure 2). The predicted whole-body BCFss for this compound ranged from 1290 L kg−1 in roach to 5650 L kg−1 in lake trout (Figure 2), thereby spanning the classifications of not bioaccumulative (roach), bioaccumulative (fathead minnow and zebrafish), and very bioaccumulative (salmonids). These differences in bioconcentration were explained within the model and reflected species differences in total lipid content, expressed as percent body ww (Figure 3). Normalization of the predicted BCFss values to 5% total lipid, as recommended by OECD 305, resulted in BCFss,l values ranging from 3060 L kg−1 in roach to 3450 L kg−1 in zebrafish. Rainbow trout and lake trout had similar total lipid levels, and this resulted in a relatively wide 95% confidence interval of the resulting SSD for these five species (Figure 2). The BCF95ss (i.e., the BCF that corresponds to 95% of affected species (represented by the modeled species)), based on the central tendency of the SSD, for a generic compound with log Kow = 5.5, would equal 11 400 L kg−1 (with 95% confidence limits of 3500−37 100 L kg−1). This was calculated from the SSD in a similar manner to the hazard concentration affecting 5% of a population (HC5) but differed in the sense that higher BCFs represent greater bioconcentration, while lower EC50 values represent more potent effects. BCFss values based on the concentration in the liver (BCFliver) showed a different rank order, as compared with whole-body BCFss values (Figure 2). Although roach and lake trout still had the lowest and highest BCFliver values, respectively, rainbow trout had the second-lowest BCFliver, followed by fathead minnow and zebrafish. As expected, these differences in bioconcentration could again be explained by species differences in the liver lipid level (Figure 3). The resulting BCF95liver of the generic compound was 9490 L kg−1, and the confidence interval was narrower than that of BCF95ss, ranging from 5000 to 18 000 L kg−1. In light of the marked differences between whole-body BCFs and organ-specific BCFs in individual species, it might be necessary to consider such organ-specific distributions when comparing interspecies differences in sensitivity; unlike what is assumed in the classical CBR concept, EC50 and whole-body BCF values alone might not always be sufficient to understand interspecies differences in sensitivity.42 When bioconcentration is considered during investigation of the toxic effects of a chemical, tissue concentrations often need to be taken into account; this is referred to as the “tissue residue approach for toxicity assessment” (TRA). Participants of a SETAC (Society of Environmental Toxicology and Chemistry) Pellston workshop recently pointed out that the incorporation of toxicokinetics and bioavailability in toxicity assessments based on tissue residues can substantially reduce variability in intra- and interspecies comparisons of toxicity.43 Our unified multispecies PBTK model framework might thus have a potential application in “predictive TRAs”, particularly if experimental tissue residue data are unavailable. Predictions of k1 and k2 ranged from 89.8 L kg−1 d−1 (roach) to 522 L kg−1 d−1 (zebrafish) and from 0.0330 d−1 (lake trout) to 0.113 d−1 (zebrafish), respectively (Figure 2). Ideally, BCFss and kinetic BCFs (which are derived by dividing k1 by k2) should be identical. The example BCF95ss of 11 400 L kg−1 shown in Figure 2 deviates substantially from the kinetic BCF derived from the rate constants corresponding to 95% of the species within the SSD (4100 L kg−1), predominantly because

Figure 1. Relationship between measured and modeled internal concentrations (A) in zebrafish, (B) in rainbow trout, and (C) in fathead minnow. Solid lines represent the equation line, and dashed lines represent the interval of 10-fold deviation from equality. No such correlation analysis was possible for roach or lake trout because far fewer experimental values were available. Input data for the model outputs in B and C originated from Stadnicka et al.17 Experimental log Kow values were obtained from the quality-evaluated Sangster Research Laboratories LOGKOW database.

chemicals with log Kow values ranging from 1.54 to 6.53, while 68 data points for 24 chemicals ranging from log Kow 1.46 to 6.50 were available for fathead minnow (Supporting Tables S4 and S5). The PBTK models were used to predict Cint values under the given exposure conditions, which were subsequently compared with the corresponding experimentally measured Cint values. Our reimplemented models for rainbow trout and fathead minnow performed about equally well compared with the models of Stadnicka et al.,17 with RMSEs of 0.66 and 1.00 log units, respectively (Figure 1). For lake trout, no such correlation analysis was possible because far less experimental values were available. The model had, however, been used with good success in previously published work by Lien and coworkers to model the uptake and disposition of waterborne 1,1,2,2-tetrachloroethane (TCE), pentachloroethane (PCE), and hexachloroethane (HCE) for periods of 6, 12, 24, or 48 h.28 3.3. Multispecies Predictions Using the Combined PBTK Models. The individual PBTK models were combined into a multispecies model framework. To analyze the behavior 1918

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

Figure 2. Species sensitivity distributions (SSDs) for (A) the bioconcentration factor (BCF) at steady-state (BCFss), (B) the BCF in liver, (C) the uptake rate constants (k1), and (D) the depuration rate constants (k2) of a generic chemical with log Kow = 5.5 in roach, fathead minnow, zebrafish, rainbow trout, and lake trout. The SSDs, which comprise the central tendency and the prediction band, were calculated using the US-EPA SSD Generator v1. Red lines indicate the respective value that comprises 95% of the fish species represented by the model predictions and the gray area represents the 95% confidence interval of the value. Default values for the model inputs are given in Table 2.

Figure 3. Modeled whole-body and liver steady-state bioconcentration factor values are highly correlated with the total lipid content of (A) the fish body and (B) the liver. (C) The uptake rate constant predicted by the physiologically based toxicokinetic model is correlated with the model of Arnot and Gobas44 but substantially deviates from equality. (D) Regression of BCF95ss values (i.e., BCF values derived from species sensitivity distributions (SSDs) that correspond to 95% affected species (represented by the model predictions) and log Kow). Note that the log Kow that corresponds to a BCF of 2000 (the threshold for bioaccumulative substances under REACH) is 4.5 (i.e., equal to the regulation’s screening level criterion for bioaccumulative substances). Solid lines represent the linear regression lines, and dashed lines represent the 95% confidence interval.

1919

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

Figure 4. Species sensitivity distributions (SSDs) for dieldrin, DDT, PDCB, PCP, TCDD, and HCB based on the bioconcentration factor (BCF) predictions of the multispecies physiologically based toxicokinetic model framework (gray lines indicate the central tendency, as well as the 95% prediction interval) and experimental BCF values taken from the US-EPA ECOTOX database (filled circles represent measured values, and the solid black line represents the central tendency). SSDs were calculated using the US-EPA SSD Generator v1.

from the US-EPA ECOTOX database (Figure 4). In the following paragraphs, the model predictions for each individual chemical are described in relation to their physicochemical properties. 3.4.1. Dieldrin. Dieldrin is an organochlorine insecticide with reported log Kow values ranging from 5.4 to 6.2.45,46 The arithmetic mean (5.8) of these values was used in the PBTK model. Dieldrin is only slowly biotransformed in fish,47 and the Stockholm Convention on persistent organic pollutants (POPs), which took effect in 2004, classified dieldrin as a PBT substance.48 Predictions of its multispecies bioconcentration potential were very accurate (Figure 4), with all measured values (ten data points for five species) lying within the 95% confidence intervals of the model prediction. The measured and modeled BCF95ss values were 20 166 and 18 700 L kg−1, respectively (7.7% deviation). This high predictive accuracy was probably owing to the physicochemical properties of dieldrin, which favor the prediction of lipid-based partitioning.49 3.4.2. Dichlorodiphenyltrichloroethane. DDT is another organochlorine insecticide with a log Kow value of 6.91.50 It is considered a PBT substance by the Stockholm Convention, which restricted its use to vector control.51 In fish, DDT is only metabolized to a relatively small extent and is consequently

the SSDs for BCFss are subject to much greater uncertainty than those for the rate constants (Figure 2). Even though the slope of the relationship departed substantially from one, k1 correlated well with the model of Arnot and Gobas;44 this model represents this constant as a function of the chemical’s log Kow, the total fish body ww, the ambient temperature, and the dissolved oxygen concentration. Plotting the BCF95ss values for different chemicals against their log Kow revealed a linear relationship (Figure 3). Interpolation of log Kow values corresponding to the stipulated BCF thresholds for bioaccumulative (2000 L kg−1) and very bioaccumulative (5000 L kg−1) compounds resulted in log Kow 4.45 and 5.00, respectively, coinciding with the current screening level criteria of log Kow 4.5 and 5, respectively. These would cover approximately 95% of the fish species represented by the five modeled ones. 3.4. Comparison of Predicted SSDs with Measured Data. Having validated the multispecies PBTK model framework, we predicted SSDs for six chemicals (dieldrin, dichlorodiphenyltrichloroethane (DDT), 1,4-dichlorobenzene (PDCB), pentachlorophenol (PCP), 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), and hexachlorobenzene (HCB) and compared them with experimentally determined SSDs obtained 1920

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology slowly eliminated.52 Multispecies predictions of its bioconcentration potential were very accurate, with all measured values (ten data points for seven fish species) lying within the 95% confidence intervals of the model-predicted values (Figure 4). The measured and modeled BCF95ss values were 105 826 and 116 000 L kg−1, respectively (9.6% deviation). 3.4.3. 1,4-Dichlorobenzene. PDCB is applied as a pesticide, fungicide, and deodorant. Its use was restricted by the European Commission in May 2014, although it was not classified as a PBT or a very persistent and very bioaccumulative substance.53 It has a log Kow of 3.44.54 Consistent with the dieldrin and DDT studies, predictions of the multispecies model were reasonably accurate for PDCB (Figure 4). However, the experimental data from one fish species (O. mykiss) deviated substantially from the model prediction, resulting in a great difference in the measured and modeled BCF95ss values, which were 1237 and 349 L kg−1, respectively (3.5-fold deviation). Neither of these SSDs indicated an elevated bioconcentration potential because they were below the threshold of 2000 L kg−1. 3.4.4. Pentachlorophenol. PCP has been widely used as a pesticide and disinfectant. It has a log Kow of 5.1254 but is not considered a PBT substance because it is readily biotransformed in fish.55 At natural pH values it is present mostly in its dissociated form (pKa = 4.74). This fact results in an overestimation of the bioavailable fraction if only the total concentration is considered during calculation of its BCF.56 Although the measured and predicted SSD values correlated, there was a substantial difference between them; this resulted in a 3.5-fold difference between the measured and modeled BCF95ss values (Figure 4). These findings indicate that the model does not yet yield accurate predictions for ionizable or readily biotransformed compounds. In particular, the latter aspect should be addressed in dedicated future research, potentially by including data from fish in vitro assays49,57 or by using Quantitative Structure Activity Relationships (QSARs) to predict biotransformation rates in fish from chemical structure, as demonstrated by Arnot and co-workers.58 3.4.5. 2,3,7,8-Tetrachlorodibenzo-p-dioxin. TCDD, often referred to as “dioxin”, has a log Kow of 6.8059 and is classified as a PBT substance by the Stockholm Convention.48 It shows almost no biotransformation in fish.60 Although the predicted BCF values were generally accurate, two of the five measured values (Mosquitofish, Gambusia affinis; and lake trout, S. namaycush) deviated substantially from the predicted values (Figure 4). The experimental data for lake trout might be erroneous because its typical whole-body total lipid content is similar to that of rainbow trout (Table 2). Despite these anomalies, four experimental data points that were available for rainbow trout (O. mykiss), and three that were available for fathead minnow (P. promelas) supported the central tendency of the SSD in the upper part of the curve, leading to an exact prediction of the BCF95ss (Figure 4). The measured and modeled BCF95ss values were 103 597 and 98 300 L kg−1, respectively (5.4% deviation). 3.4.6. Hexachlorobenzene. HCB has been used as a fungicide and in a number of technical applications. It is classified as a PBT substance by the Stockholm Convention and has a log Kow of 5.73.9,45 It is only relatively slowly biotransformed in fish.61 Although these properties would typically favor lipid-based modeling, experimental data from three of the five fish species deviated substantially from the predictions (Figure 4), resulting in a great difference in

measured and modeled BCF95ss values; these were 83 936 and 22 100 L kg−1, respectively (3.8-fold deviation). Both of these SSDs, however, predicted a high bioconcentration potential because they exceeded the 5000 L kg−1 threshold. These apparent differences might reflect a bias in the experimental data, which was only available for fish species with a relatively low total-lipid level; no values were available for fish with high total-body lipid content (e.g., salmonids). 3.5. Potentials and Limitations of the Proposed Methodology. We conclude that our multispecies PBTK model framework can play an important role in predicting a chemical’s bioaccumulation potential across different species. The presented unified model structure can be relatively easily reparameterized to other fish species, thus increasing its predictive power. It should be emphasized that the current number of five fish species included in this study should be progressively expanded as more data becomes available to further increase the representativeness of the predictions and to reduce uncertainties. These additional species should ideally include members of other important orders of (teleost) fishes (e.g., Perciformes, Siluriformes, and Esociformes). Apart from these limitations in terms of species coverage, results of the present study indicated that the model (as most toxicokinetic models to date) does not yet yield accurate predictions for ionizable compounds. Furthermore, readily biotransformed substances require additional data for parametrization. To realize the full potential of the proposed methodology, future efforts should incorporate toxicodynamic (TD) submodels into the framework, resulting in a powerful multispecies PBTK−TD model. The explicit modeling of individual organs and tissues is particularly beneficial because it represents a first step toward enabling both researchers and regulators to predict toxicological effects semiquantitatively (e.g., acute toxicity, hepatotoxicity, aneugenic or clastogenic effects, hepatic lesions or carcinogenesis, and effects on growth or reproduction), without performing animal experiments. To this end, it might become necessary to include more organs as separate compartments (e.g., brain and gonads). From a regulatory perspective, this type of in silico method has the potential to achieve a meaningful reduction in the number of animal experiments required for effective toxicological and ecotoxicological research.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b06158. Tables showing model inputs and parameters of the PBTK models; experimental data used for modeling internal chemical concentrations in zebrafish, roach, rainbow trout, and fathead minnow using the PBTK model. Model equations. (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Phone: +49 (0) 241 80 26686; fax: +49 (0) 241 80 22182; email: [email protected]. Present Address ○

Bayer CropScience GmbH, Monheim am Rhein, Germany

Notes

The authors declare no competing financial interest. 1921

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology

■

(18) Bungay, P.; Dedrick, R.; Guarino, A. Pharmacokinetic modeling of the dogfish shark (Squalus acanthias): Distribution and urinary and biliary excretion of phenol red and its glucuronide. J. Pharmacokinet. Biopharm. 1976, 4 (5), 377−388. (19) Nichols, J. W.; Jensen, K. M.; Tietge, J. E.; Johnson, R. D. Physiologically based toxicokinetic model for maternal transfer of 2,3,7,8-tetrachlorodibenzo-p-dioxin in brook trout (Salvelinus fontinalis). Environ. Toxicol. Chem. 1998, 17 (12), 2422−2434. (20) Lien, G. J.; McKim, J. M.; Hoffman, A. D.; Jenson, C. T. A physiologically based toxicokinetic model for lake trout (Salvelinus namaycush). Aquat. Toxicol. 2001, 51 (3), 335−350. (21) Nichols, J. W.; McKim, J. M.; Lien, G. J.; Hoffman, A. D.; Bertelsen, S. L.; Gallinat, C. A. Physiologically-based toxicokinetic modeling of three waterborne chloroethanes in channel catfish. Aquat. Toxicol. 1993, 27 (1−2), 83−111. (22) Lien, G. J.; McKim, J. M. Predicting branchial and cutaneous uptake of 2,2′,5,5′-tetrachlorobiphenyl in fathead minnows (Pimephales promelas) and Japanese medaka (Oryzias latipes): Rate limiting factors. Aquat. Toxicol. 1993, 27 (1−2), 15−31. (23) Liao, C.-M.; Liang, H.-M.; Chen, B.-C.; Singh, S.; Tsai, J.-W.; Chou, Y.-H.; Lin, W.-T. Dynamical coupling of PBPK/PD and AUCbased toxicity models for arsenic in tilapia Oreochromis mossambicus from blackfoot disease area in Taiwan. Environ. Pollut. 2005, 135 (2), 221−233. (24) Péry, A. R. R.; Devillers, J.; Brochot, C.; Mombelli, E.; Palluel, O.; Piccini, B.; Brion, F.; Beaudouin, R. A Physiologically Based Toxicokinetic Model for the Zebrafish Danio rerio. Environ. Sci. Technol. 2014, 48 (1), 781−790. (25) Brinkmann, M.; Eichbaum, K.; Kammann, U.; Hudjetz, S.; Cofalla, C.; Buchinger, S.; Reifferscheid, G.; Schüttrumpf, H.; Preuss, T.; Hollert, H. Physiologically-based toxicokinetic models help identifying the key factors affecting contaminant uptake during flood events. Aquat. Toxicol. 2014, 152 (0), 38−46. (26) Brinkmann, M.; Eichbaum, K.; Buchinger, S.; Reifferscheid, G.; Bui, T.; Schäffer, A.; Hollert, H.; Preuss, T. G. Understanding receptor-mediated effects in rainbow trout: in vitro−in vivo extrapolation using physiologically based toxicokinetic models. Environ. Sci. Technol. 2014, 48 (6), 3303−3309. (27) Nichols, J. W.; McKim, J. M.; Andersen, M. E.; Gargas, M. L.; Clewell, H. J.; Erickson, R. J. A physiologically based toxicokinetic model for the uptake and disposition of waterborne organic chemicals in fish. Toxicol. Appl. Pharmacol. 1990, 106 (3), 433−447. (28) Lien, G. J.; McKim, J. M.; Hoffman, A. D.; Jenson, C. T. A physiologically based toxicokinetic model for lake trout (Salvelinus namaycush). Aquat. Toxicol. 2001, 51 (3), 335−350. (29) Baade, U.; Fredrich, F. Movement and pattern of activity of the roach in the River Spree, Germany. J. Fish Biol. 1998, 52 (6), 1165− 1174. (30) Larmuseau, M. H. D.; Freyhof, J.; Volckaert, F. A. M.; Van Houdt, J. K. J. Matrilinear phylogeography and demographical patterns of Rutilus rutilus: implications for taxonomy and conservation. J. Fish Biol. 2009, 75 (2), 332−353. (31) Ibrahim, L.; Preuss, T.; Ratte, H.; Hommen, U. A list of fish species that are potentially exposed to pesticides in edge-of-field water bodies in the European Uniona first step towards identifying vulnerable representatives for risk assessment. Environ. Sci. Pollut. Res. 2013, 20 (4), 2679−2687. (32) Machala, M.; Ulrich, R.; Neca, J.; Vykusova, B.; Kolarova, J.; Machova, J.; Svobodova, Z. Biochemical monitoring of aquatic pollution: Indicators of dioxin-like toxicity and oxidative stress in the roach (Rutilus rutilus) and chub (Leuciscus cephalus) in the Skalice river. Vet. Med.-Czech 2000, 45 (2), 55−60. (33) Wolinska, L.; Brzuzan, P.; Wozny, M.; Luczynski, M. K.; Gora, M. CYP1A expression in liver and gills of roach (Rutilus rutilus) after waterborne exposure to two phenanthrene derivatives, 1-methylphenanthrene and 4-methylphenanthrene. Environ. Toxicol. Chem. 2013, 32 (7), 1604−1610.

ACKNOWLEDGMENTS The present study was conducted in the context of the “DioRAMA: Assessment of the Dioxin-Like Activity in Sediments and Fish for Sediment Evaluation” project, which received funds from the German Federal Ministry of Transport, Building and Urban Development. The authors acknowledge the German National Academic Foundation (“Studienstiftung des deutschen Volkes”) for a personal scholarship granted to M.B. T.P. and C.S. received funding from the German Federal Environment Agency (“Further development of criteria for bioaccumulation under REACH”, FKZ 3711 63405 2).

■

REFERENCES

(1) Barron, M. G. Bioconcentration. Will water-borne organic chemicals accumulate in aquatic animals? Environ. Sci. Technol. 1990, 24 (11), 1612−1618. (2) Schafer, S.; Buchmeier, G.; Claus, E.; Duester, L.; Heininger, P.; Korner, A.; Mayer, P.; Paschke, A.; Rauert, C.; Reifferscheid, G.; Rudel, H.; Schlechtriem, C.; Schroter-Kermani, C.; Schudoma, D.; Smedes, F.; Steffen, D.; Vietoris, F. Bioaccumulation in aquatic systems: methodological approaches, monitoring and assessment. Environ. Sci. Eur. 2015, 27 (1), 5. (3) EC. Regulation No. 1907/2006 of the European Parliament and of the Council Concerning the Registration, Evaluation, Authorization and Restriction of Chemicals. 2006. (4) OECD Bioaccumulation in Fish: Aqueous and Dietary Exposure, Test No. 305; OECD Publishing: Paris, France, 2012. (5) Chiou, C. T. Partition-coefficients of organic-compounds in lipidwater systems and correlations with fish bioconcentration factors. Environ. Sci. Technol. 1985, 19 (1), 57−62. (6) Mackay, D. Correlation of bioconcentration factors. Environ. Sci. Technol. 1982, 16 (5), 274−278. (7) Bintein, S.; Devillers, J.; Karcher, W. Nonlinear Dependence of Fish Bioconcentration on n-Octanol/Water Partition Coefficient. SAR QSAR Environ. Res. 1993, 1 (1), 29−39. (8) Gobas, F. P. C.; Shiu, W.; Mackay, D. Factors Determining Partitioning of Hydrophobic Organic Chemicals in Aquatic Organisms. In QSAR in Environmental Toxicology - II; Kaiser, K. E., Ed.; Springer: Netherlands, 1987; pp 107−123. (9) ECHA Guidance on Information Requirements and Chemical Safety Assessment, Chapter R.7c: Endpoint specific guidance, v 2.0. European Chemicals Agency, Helsinki, Finland, 2014. (10) Zarfl, C.; Matthies, M. PBT borderline chemicals under REACH. Environ. Sci. Eur. 2013, 25 (1), 11. (11) Gobas, F. A. P. C.; de Wolf, W.; Burkhard, L. P.; Verbruggen, E.; Plotzke, K. Revisiting Bioaccumulation Criteria for POPs and PBT Assessments. Integr. Environ. Assess. Manage. 2009, 5 (4), 624−637. (12) Swackhamer, D. L.; Needham, L. L.; Powell, D. E.; Muir, D. C. G. Use of Measurement Data in Evaluating Exposure of Humans and Wildlife to POPs/PBTs. Integr. Environ. Assess. Manage. 2009, 5 (4), 638−661. (13) Escher, B. I.; Hermens, J. L. M. Internal Exposure: Linking Bioavailability to Effects. Environ. Sci. Technol. 2004, 38 (23), 455A− 462A. (14) McCarty, L. S.; Mackay, D. Enhancing ecotoxicological modeling and assessment. Body Residues and Modes Of Toxic Action. Environ. Sci. Technol. 1993, 27 (9), 1718−1728. (15) Mackay, D.; Fraser, A. Bioaccumulation of persistent organic chemicals: mechanisms and models. Environ. Pollut. 2000, 110 (3), 375−391. (16) Nichols, J. W.; McKim, J. M.; Lien, G. J.; Hoffman, A. D.; Bertelsen, S. L. Physiologically based toxicokinetic modeling of three waterborne chloroethanes in rainbow trout (Oncorhynchus mykiss). Toxicol. Appl. Pharmacol. 1991, 110 (3), 374−389. (17) Stadnicka, J.; Schirmer, K.; Ashauer, R. Predicting concentrations of organic chemicals in fish by using toxicokinetic models. Environ. Sci. Technol. 2012, 46 (6), 3273−3280. 1922

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923

Article

Environmental Science & Technology (34) Oikari, A.; Kukkonen, J. Acclimatization of roach, Rutilus rutilus (L.), to toxic components of kraft pulp mill effluents. Ecotoxicol. Environ. Saf. 1988, 15 (3), 282−288. (35) Smith, M. D.; Hill, E. M. Uptake and metabolism of technical nonylphenol and its brominated analogues in the roach (Rutilus rutilus). Aquat. Toxicol. 2004, 69 (4), 359−370. (36) Strähle, U.; Scholz, S.; Geisler, R.; Greiner, P.; Hollert, H.; Rastegar, S.; Schumacher, A.; Selderslaghs, I.; Weiss, C.; Witters, H.; Braunbeck, T. Zebrafish embryos as an alternative to animal experimentsa commentary on the definition of the onset of protected life stages in animal welfare regulations. Reprod. Toxicol. 2012, 33 (2), 128−132. (37) Smedes, F. Determination of total lipid using non-chlorinated solvents. Analyst 1999, 124 (11), 1711−1718. (38) Weiss, R. F. The solubility of nitrogen, oxygen and argon in water and seawater. Deep-Sea Res. Oceanogr. Abstr. 1970, 17 (4), 721− 735. (39) Sangster, J. LOGKOW database. Sangster Research Laboratories, Montreal, Quebec, Canada, 2014. (40) U.S. Environmental Protection Agency Estimation Programs Interface Suite for Microsoft Windows, v 4.11. United States Environmental Protection Agency: Washington, DC, 2015. (41) U.S. Environmental Protection Agency ECOTOX User Guide: ECOTOXicology Database System. Version 4.0. http:/www.epa.gov/ ecotox/ (accessed 19 September 2014). (42) Barron, M. G.; Anderson, M. J.; Lipton, J.; Dixon, D. G. Evaluation of critical body residue QSARS for predicting organic chemical toxicity to aquatic organisms. SAR QSAR Environ. Res. 1997, 6 (1−2), 47−62. (43) Meador, J. P.; Adams, W. J.; Escher, B. I.; McCarty, L. S.; McElroy, A. E.; Sappington, K. G. The tissue residue approach for toxicity assessment: Findings and critical reviews from a Society of Environmental Toxicology and Chemistry Pellston Workshop. Integr. Environ. Assess. Manage. 2011, 7 (1), 2−6. (44) Arnot, J. A.; Gobas, F. A. P. C. A food web bioaccumulation model for organic chemicals in aquatic ecosystems. Environ. Toxicol. Chem. 2004, 23 (10), 2343−2355. (45) De Bruijn, J.; Busser, F.; Seinen, W.; Hermens, J. Determination of octanol/water partition coefficients for hydrophobic organic chemicals with the “slow-stirring” method. Environ. Toxicol. Chem. 1989, 8 (6), 499−512. (46) Briggs, G. G. Theoretical and experimental relationships between soil adsorption, octanol-water partition coefficients, water solubilities, bioconcentration factors, and the parachor. J. Agric. Food Chem. 1981, 29 (5), 1050−1059. (47) Sudershan, P.; Khan, M. Metabolism of [14 C] dieldrin in bluegill fish. Pestic. Biochem. Physiol. 1981, 15 (2), 192−199. (48) Chemicals, U., Stockholm Convention on Persistent Organic Pollutants (POPs). 2002. (49) Nichols, J. W.; Huggett, D. B.; Arnot, J. A.; Fitzsimmons, P. N.; Cowan-Ellsberry, C. E. Toward improved models for predicting bioconcentration of well-metabolized compounds by rainbow trout using measured rates of in vitro intrinsic clearance. Environ. Toxicol. Chem. 2013, 32 (7), 1611−1622. (50) Howard, P. H. Handbook of Physical Properties of Organic Chemicals. CRC Press: Boca Raton, FL, 1996. (51) van den Berg, H. Global status of DDT and its alternatives for use in vector control to prevent disease. Environ. Health Perspect. 2009, 1656 DOI: 10.1289/ehp.0900785. (52) Gakstatter, J. H.; Weiss, C. M. The Elimination of DDT-C14, Dieldrin-C14, and Lindane-C14 from Fish following a Single Sublethal Exposure in Aquaria. Trans. Am. Fish. Soc. 1967, 96 (3), 301−307. (53) EC Commision regulation (EU) No 474/2014 of 8 May 2014 amending Annex XVII to Regulation (EC) No 1907/2006 of the European Parliament and of the Council on the Registration, Evaluation, Authorisation and Restriction of Chemicals (“REACH”) as regards 1,4-dichlorobenzene. Official Journal of the European Union 2014, L 136, 19−22.

(54) Hansch, C.; Leo, A.; Hoekman, D. Exploring QSAR: Hydrophobic, Electronic and Steric Constants; American Chemical Society: Washington, DC, 1995. (55) Cravedi, J. P.; Lafuente, A.; Baradat, M.; Hillenweck, A.; PerduDurand, E. Biotransformation of pentachlorophenol, aniline and biphenyl in isolated rainbow trout (Oncorhynchus mykiss) hepatocytes: comparison with in vivo metabolism. Xenobiotica 1999, 29 (5), 499− 509. (56) Mackay, D.; Shiu, W.-Y.; Ma, K.-C.; Lee, S. C. Handbook of Physical-Chemical Properties and Environmental Fate for Organic Chemicals. CRC Press: Boca Raton, FL, 2010. (57) Fay, K. A.; Mingoia, R. T.; Goeritz, I.; Nabb, D. L.; Hoffman, A. D.; Ferrell, B. D.; Peterson, H. M.; Nichols, J. W.; Segner, H.; Han, X. Intra- and Interlaboratory Reliability of a Cryopreserved Trout Hepatocyte Assay for the Prediction of Chemical Bioaccumulation Potential. Environ. Sci. Technol. 2014, 48 (14), 8170−8178. (58) Arnot, J. A.; Meylan, W.; Tunkel, J.; Howard, P. H.; Mackay, D.; Bonnell, M.; Boethling, R. S. A quantitative structure-activity relationship for predicting metabolic biotransformation rates for organic chemicals in fish. Environ. Toxicol. Chem. 2009, 28 (6), 1168−1177. (59) Shiu, W. Y.; Doucette, W.; Gobas, F. A. P. C.; Andren, A.; Mackay, D. Physical-chemical properties of chlorinated dibenzo-pdioxins. Environ. Sci. Technol. 1988, 22 (6), 651−658. (60) Kleeman, J. M.; Olson, J. R.; Peterson, R. E. Species Differences in 2,3,7,8-Tetrachlorodibenzo-p-dioxin Toxicity and Biotransformation in Fish. Toxicol. Sci. 1988, 10 (2), 206−213. (61) Simeonov, L. I.; Hassanien, M. A. Exposure and Risk Assessment of Chemical Pollution: Contemporary Methodology. Springer Science & Business Media: Heidelberg, Germany, 2009.

1923

DOI: 10.1021/acs.est.5b06158 Environ. Sci. Technol. 2016, 50, 1914−1923