Towards next generation risk assessment of chemicals: Development and application of physiologically based kinetic models in farm animals

Leonie S. Lautz 2020

Towards next generation risk assessment of chemicals: Development and application of physiologically based kinetic models in farm animals

Proefschrift

ter verkrijging van de graad van doctor aan de Radboud Universiteit Nijmegen op gezag van de rector magnificus prof. dr. J.H.J.M. van Krieken, volgens besluit van het college van decanen

in het openbaar te verdedigen op

dinsdag 2 juni 2020,

om 16.30 uur precies

door

Leonie Sophie Lautz geboren op 5 maart 1990 te Bad Oeynhausen, Duitsland

2 Promotoren Prof. dr. A.M.J. Ragas (Open Universiteit) Prof. dr. ir. A.J. Hendriks

Copromotor Dr. J.L.C.M. Dorne (European Food Safety Authority, Parma, Italïe)

Manuscriptcomissie Prof. dr. A.M. Breure Prof. dr. F.G.M. Russel Prof. dr. R. Gehring (Universiteit Utrecht)

This research described in this thesis was funded by the European Food Safety Authority (EFSA) [Contract number: EFSA/SCER/2014/06].

3

This is how you do it: you sit down at the keyboard and write one word after another until it is done. It’s that easy, and that hard. (Neil Gaiman)

4 Table of Contents

Chapter 1 ...... 6

General introduction ......

Chapter 2 ...... 13

Physiologically based kinetic models for farm animals: critical review of published models and future perspectives for their use in chemical risk assessment ......

Chapter 3 ...... 32

Generic physiologically based kinetic modelling for farm animals: Part I. Data collection of physiological parameters in swine, cattle and sheep ......

Chapter 4 ...... 42

Generic physiologically-based kinetic modelling for farm animals: Part II. Predicting tissue concentrations of chemicals in swine, cattle, and sheep ......

Chapter 5 ...... 55

An open source physiologically-based kinetic model for the chicken (Gallus gallus domesticus): Calibration and validation for the prediction of residues in tissues and eggs ......

Chapter 6 ...... 78

General discussion and synthesis ......

Bibliography ...... 87

Appendices ...... 110

Summary / Samenvatting ...... 117

Acknowledgments ...... 122

Curriculum Vitae ...... 124

5

CHAPTER 1

GENERAL INTRODUCTION

6 1. GENERAL INTRODUCTION

1.1. Principles of animal health risk assessment of chemicals

Chemicals from anthropogenic and natural origins enter animal feed and human food as undesirable contaminant or components of a diet, such as persistent organic pollutants, phthalates and veterinary medicinal products (Dorne and Fink-Gremmels, 2013). These substances can have potential adverse effects on health and productivity of farm and companion animals (Berny et al., 2010; Guitart et al., 2010). Furthermore, the presence of residues of unauthorized substances, veterinary medicinal products or chemical contaminants in food may pose a risk factor for human health, due to potential accumulation and transfer within the food chain (EFSA, 2018b; Verstraete, 2013). The control of risks possibly associated with chemical contamination in food and feed is a priority of the European Union (Silano and Silano, 2017).

Compared to human risk assessment, toxicological risk assessment in farm and companion animals is still a young discipline. In the food and feed safety area, animal health risk assessment of chemicals follows the classical steps of the risk assessment framework namely, hazard assessment, exposure assessment and risk characterisation (Dorne and Fink-Gremmels, 2013). Scientific advisory boards have not developed specific frameworks for risk assessment in animal health (farm and companion animals), but in practice, mostly the principles of human risk assessment are applied (EFSA, 2014a; EFSA Scientific Committee, 2019). Available methods and tools in human risk assessment were developed mostly for the pharmaceutical industry, where chemicals are developed for specific targets with certain physicochemical properties and with well-understood kinetics and metabolism (Hartung and Hoffmann, 2009; Sager et al., 2015). Very different from this, the risk assessment of most undesirable chemicals requires identifying unknown modes of action, usually in the absence of kinetic data, for many diverse chemicals with rather limited resources (Hartung and Hoffmann, 2009). Next to this, animal health risk assessment also has to account for differences related to species-specific and interspecies differences in kinetics, dose- dependent toxicity and methodologies to estimate exposure to characterise the risk in variety of animal species.

In that respect, a major data gap and research need in this area applies to understanding species differences in kinetics and dynamic (pharmacodynamic or toxicodynamic) processes for single chemicals and mixtures of chemicals and the implementation of such data in current risk assessment practices (EFSA Scientific Committee, 2019). Harmonisation in methodology and terminology between human, animal and environmental risk assessment enhances the understanding of risk assessment issues associated with the

7 food chain and enables efficient use of resources and consistency among assessments (EFSA, 2014b; EFSA Scientific Committee, 2019). Currently, the use of kinetic data in animal health risk assessment is limited due to limited data availability, and when it occurs, it is often on an ad-hoc basis (Alexander et al., 2012; Clapham et al., 2019; Dorne and Fink-Gremmels, 2013). The differences in absorption, distribution, metabolism and excretion (ADME processes) between animal species and the limited data availability on those differences have consequences for the outcome of risk assessment and introduce uncertainty in risk characterisation. In order to protect animal health, it is essential to understand the transfer of chemicals residues in farm animals and to derive safe levels of exposure in feed. Improvement of uncertainty factors and the application of in silico models incorporating kinetics, e.g. one compartment models and physiologically based kinetic (PBK) models, are examples of refinements that may reduce uncertainties in current risk assessment practices, particularly in the extrapolation of toxicity data between various species (Dorne and Fink-Gremmels, 2013). To move towards the integration of new approaches in chemical risk assessment in the areas of human health, animal health and environment, several projects were launched by the European Food Safety Authority (EFSA). These projects involve the integration of kinetic tools (EFSA-Q-2014-00918; EFSA-Q-2015-00640) the modelling of population dynamics of aquatic and terrestrial organisms (EFSA-Q-2015-00554), and the modelling of human variability in kinetic and dynamic processes (EFSA-Q-2015-00641).

1.2. Allometric scaling

Body mass is an important parameter, since it largely determines the morphology, physiology, and biology of organisms (Hendriks, 2007; Lindstedt and Schaeffer, 2002). Changes in body mass can result in changes of proportions of the body fluids and organs (Boxenbaum, 1980). In the context of risk assessment, some important kinetic parameters, including the clearance, volume of distribution and elimination half-life, can also be predicted based on the body mass (Huang and Riviere, 2014). Allometric scaling can be used to correlate variables between organisms of different size and is based on a power-law function (Equation 1):

푌 = 푎 ∗ 푀푏 (1)

Where Y is the parameter of interest that has to be correlated with body mass (M). The constant a is proportionality factor for the particular parameter, whereas b is the allometric exponent, which varies in magnitude and is specific for each parameter according to how it scales with mass. If log transformed the equation turns into a linear function (Equation 2):

8 log 푌 = log 푎 + 푏 ∗ log 푀 (2)

An important advantage of allometric scaling is that it is a widely applied and understood technique. Especially for zoo animals and wildlife, allometric scaling has a definite role since data for these animals are scarce, particularly for large animals such as camels, large cats and elephants. Interspecies allometric scaling is a powerful tool to extrapolate the use of approved drugs to non-approved species (Hunter, 2010). However, allometric scaling is not applicable for all chemicals. Specific chemical properties, like protein binding, hepatic metabolism and renal reabsorption can limit the accuracy of allometric scaling (Riviere et al., 1997). Nevertheless, allometric scaling is a simple and fast approach to predict species- specific parameters for which no data are available.

1.3. Kinetic modelling approaches

Kinetics is the study of the time course of chemical concentrations in biological fluids, tissues and excreta as well as the construction of mathematical models to interpret the data (Wagner, 1981). Kinetics of chemicals in various animals makes it possible to have a closer look at the distribution of a chemical in an animal. Classical compartment modelling and physiologically based kinetic (PBK) modelling are commonly used for the simulation of kinetic parameters and concentration-time curves. The following paragraphs will briefly introduce these two techniques.

Classical compartment models

Simple kinetic models are used to analyse the exposure of a chemical, mostly drugs, via a specific route. A classical kinetic model has a central compartment representing the animal’s plasma, which is linked to one or two peripheral compartments via rate constants, such as absorption and clearance (Cl), describing the uptake and excretion of a chemical (Jones and Rowland-Yeo, 2013). The model is obtained by fitting mathematical equations to (usually) in vivo data resulting from animal experiments; typically, plasma concentrations of the chemical measured at various time points. The values derived from simple one or two compartment model analyses, specifically apply to the experimental conditions of the study. As such, they do not provide insight into generic kinetic behaviour of the chemical assessed, outside the circumstances in which the data were obtained. Since no (physiological) volumes are defined for the included compartments, tissue-specific concentration curves cannot be obtained from classical compartment models. In the context of this thesis, the classic compartment analysis will be of limited value, since the focus is on the variation in internal dose within various animal species.

9 Physiologically based kinetic models

PBK models integrate mathematical relationships that link key physiological and anatomical parameters (e.g. blood flow, organ volumes) and biochemical parameters (e.g. partition coefficients, protein binding), to dynamically predict absorption, distribution, metabolism and excretion processes of a chemical using differential mass balance equations (WHO, 2010). Typically, PBK models consist of a variable number of compartments representing different organs and tissues (e.g. liver, kidney, gut, lung, and blood) and provide means to simulate concentration-time profiles of chemicals and corresponding metabolite(s) in these compartments (Bois et al., 2010; Chiu et al., 2007) (Figure 1). Compared to classical compartment models, the included compartments in the PBK model are defined with physiological relevant tissue volumes and blood flows. In PBK models, the ADME of a chemical in time is simulated in a closed, physiologically realistic system, while classical modelling approach consists of fitting empirically obtained chemical concentrations to estimate kinetic parameters. The advantages of PBK modelling include the ability to predict target tissue dosimetry of the modelled compound and the robust extrapolation power across species (e.g. animal to human), exposure routes, doses and duration. PBK modelling has been applied in various areas, such as risk assessment of environmental chemicals, drug tissue residue estimations, safety assessment of animal-derived food products, and drug development (Lin et al., 2016a; Reddy et al., 2005; Rowland et al., 2011).

Figure 1. A) Scheme of a general PBK model. Chemical uptake routes can be oral, intravenous, inhalation, intramuscular, or subcutaneous. Metabolism is assumed to occur in the liver only (blue arrow). Elimination of the chemical can take place via urinary excretion (kidney), egestion (gut lumen) or exhalation (lung). Other compartments, such as muscle, adipose tissue, and bone, can be included as storage organs. B) Example of PBK model output (time-dependent chemical concentration in blood).

10 “Physiologically based pharmacokinetic” (PBPK) model is the widely used term introduced by the pharmaceutical field to simulate the kinetics of drugs. However, this term is not entirely correct in the context of general chemical risk assessment. In the European Union the term “PBTK”, with TK standing for “toxicokinetics”, is preferred. This term is not entirely appropriate either, since the dose differentiates a poison and a remedy (Clewell et al., 2008). A more general term, such as physiologically based biokinetic (PBBK), might be seen as more appropriate. Regardless the terminology used, PBK, PBPK, PBTK, can all be considered synonyms (Paini et al., 2019), and so throughout this thesis the more general term PBK model is used consistently.

Human PBK models for specific compounds including pharmaceuticals, cosmetics, pesticides and contaminants as well as generic models are readily available (Madden et al., 2019). The same accounts for fish PBK models (Grech et al., 2017). In addition, a range of New Approach Methods (NAMs) have been developed to reduce animal testing such as quantitative in vitro in vivo extrapolation (QIVIVE) models which can be included in PBK models as well as other in silico tools such as quantitative structure activity (QSAR) relationship models (Paini et al., 2019). In the animal health and veterinary medicine area, several PBK models have been developed in recent years, mostly for the prediction of tissue residues and withdrawal intervals in farm animals (Lautz et al., 2019; Lin et al., 2016a). For farm animals, availability of PBK models is much more limited and mostly restricted to specific compounds. However, the development and application of separate PBK models for each unique compound-species combination is inefficient and practically unfeasible. This leaves a gap regarding generic PBK models for farm animals incorporating historical kinetic information as well as implementing NAMs (e.g. QIVIVE and QSARs), applicable to a large number of compounds and species, within the risk assessment framework.

1.4. Aim and outline of the thesis

The main aim of the thesis was to develop and apply generic PBK models for the prediction of chemical kinetics and tissue concentrations in farm animals and provide recommendations to refine and adapt guidance documents in animal health risk assessment. Such models particularly aim to improve methods for exposure assessment based on evaluation of the internal dose.

In Chapter 2 an overview of existing PBK models is given for chicken, cattle, swine, sheep, goat, horse, deer/reindeer, turkey, duck, and goose and their use in animal risk assessment of chemicals. Furthermore, data gaps in available information are identified and recommendations are proposed for the future development and application of PBK models in farm animals. Compared with the described farm animal

11 PBK models that were available for specific chemicals only, the focus of the developed PBK models is on both pharmaceuticals and undesirable chemicals as part of the animal’s diet in farm animal species. In Chapter 3, data collection of physiological and anatomical parameters and their associated variability for swine, cattle and sheep through an extensive literature search and meta-analysis is described. In Chapter 4, the development and performance of the generic PBK model for swine, cattle, and sheep is illustrated for melamine and oxytetracycline. Chapter 5 is about the development and performance of the generic PBK model for chicken and the transfer of chemicals into the egg. Finally, in Chapter 6 the perspectives of implementation of uncertainty factors and generic PBK model in animal health risk assessment of chemical are discussed as well as future perspectives for further research particularly to improve the PBK models in this area.

12

CHAPTER 2

PHYSIOLOGICALLY BASED KINETIC MODELS FOR FARM ANIMALS: CRITICAL REVIEW OF PUBLISHED MODELS AND FUTURE PERSPECTIVES FOR THEIR USE IN CHEMICAL RISK ASSESSMENT

L.S. Lautz, R. Oldenkamp, J.L.C.M. Dorne, A.M.J. Ragas

Published in Toxicology in Vitro, 60:61-70

13 Abstract

Physiologically based kinetic (PBK) models in the 10 most common species of farm animals were identified through an extensive literature search. This resulted in 39 PBK models, mostly for pharmaceuticals. The models were critically assessed using the WHO criteria for model evaluation, i.e. 1) purpose, 2) structure and mathematical representation, 3) computer implementation, 4) parameterisation, 5) performance, and 6) documentation. Overall, most models were calibrated and validated with published data (92% and 67% respectively) but only a fraction of model codes were published along with the manuscript (28%) and local sensitivity analysis was performed without considering global sensitivity analysis. Hence, the reliability of these PBK models is hard to assess and their potential for use in chemical risk assessment is limited. In a risk assessment context, future PBK models for farm animals should include a more generic and flexible model structure, use input parameters independent on calibration and include assessment tools to assess model performance. Development and application of PBK models for farm animal species would furthermore benefit from the setup of structured databases providing data on physiological and chemical-specific parameters as well as enzyme expression and activities to support the development of species-specific QIVIVE models.

14 1. INTRODUCTION

Chemical risk assessment in food and feed safety aims to set safe levels of regulated compounds and contaminants to protect farm animals after exposure through feed, and to protect humans against carry over and residues in animal products (e.g. meat, milk, eggs). Currently, hazard and exposure metrics used in risk assessment for farm animal species and humans are typically based on external exposure. Basing such metrics on internal exposure provides opportunities to better account for interindividual and interspecies differences in physiology, toxicokinetics and toxicity (Dorne and Fink-Gremmels, 2013).

PBK models integrate mathematical relationships that link key physiological and anatomical parameters (e.g. blood flow, organ volumes) and biochemical parameters (e.g. partition coefficients, protein binding), to dynamically predict absorption, distribution, metabolism and excretion (ADME) processes of a chemical using differential mass balance equations (WHO, 2010). Typically, PBK models consist of a variable number of compartments representing different organs and tissues (e.g. liver, kidney, gut, lung, and blood) and provide means to simulate concentration-time profiles of chemicals and corresponding metabolite(s) in these compartments (Bois et al., 2010; Chiu et al., 2007) (Figure 1). For this reason, PBK models are extensively applied in the pharmaceutical industry in drug discovery and development (EMA, 2018; Jones and Rowland-Yeo, 2013). PBK models vary in complexity, ranging from full PBK models where all of the distribution organs and tissues are represented as separate perfused compartments to more simplified PBK models in which tissues with similar kinetics are lumped (Bois et al., 2010). While the simultaneous modelling of ADME processes provides several advantages, complex PBK models have the disadvantage of requiring many physiological and chemical-specific data (Sager et al., 2015).

In literature, the term “physiologically based pharmacokinetic” (PBPK) model is widely used. In the context of general chemical risk assessment this term is not entirely correct (Clewell et al., 2008; Paini et al., 2019). A more general term, such as PBK, covering the field of pharmacokinetics (PK) as well toxicokinetics (TK), might be more appropriate. However, PBPK, PBTK, and PBK can be considered synonyms of each other. Throughout this document, we will use the more general term PBK model.

15

Figure 1. Scheme of a generic PBK model. Chemical uptake routes can be oral, intravenous, inhalation, intramuscular, or subcutaneous. In this example, metabolism is assumed to occur in the liver only (blue arrow). Elimination of the chemical takes place via urinary excretion (kidney), egestion (gut lumen) and exhalation (lung). Muscle, adipose tissue, and bone are included as storage organs. The “Rest of body” compartment account for anatomically missing organs. Other organs such as skin and spleen can be added as additional compartments.

A recent survey across academia, industry and regulatory agencies such as the European Food Safety Authority (EFSA), the European Medicine Agency (EMA), the US Environmental Protection Agency (US EPA), the US Food and Drug Administration (US FDA), and the Organisation for Economic Co-operation and Development (OECD) highlighted the importance of PBK models for regulatory risk assessment (EFSA, 2014b; c; Paini et al., 2017). While the use of PBK models in risk assessment is encouraged, there are also challenges, such as the need for well-trained persons capable of model development and to ensure model quality (Barton et al., 2007; Chiu et al., 2007; Tan et al., 2018). The increasing use of PBK modelling in regulatory human health risk assessment (HRA) has led to several guidance documents facilitating good modelling practice and standardising their review, acceptance and implementation (EMA, 2018; EPA, 2006; FDA, 2018; WHO, 2010).

In the European Union, the chemical risk assessment for animal health, including farm as well as companion animals, follows the same principles as human health risk assessment (Alexander et al., 2012).

16 Main differences towards human health risk assessment are related to species-specific and interspecies differences in kinetics, dose-dependent toxicity and methodologies to estimate exposure to characterise the chemical risk in each species (Dorne and Fink-Gremmels, 2013). Furthermore, animal health risk assessment accounts for animal health due to chemical exposure, as well as possible impacts on human health due to transfer of chemicals into the food-chain (Alexander et al., 2012). In the animal health and veterinary medicine area, several PBK models have been developed in recent years, mostly for the prediction of tissue residues and withdrawal intervals in farm animals (Henri et al., 2017; Li et al., 2017; Zeng et al., 2017). While PBK models are relevant for regulatory agencies that monitor chemical concentrations in animal target tissues and animal products, e.g. meat, milk and eggs (Craigmill, 2003), as well as for the veterinary pharmaceutical industry, a specific guidance document on their use in animal health risk assessment is currently lacking.

A previous review on PBPK models for farm animal species focused on their application in veterinary medicine (Lin et al., 2016a). However, several models specific for farm animals were not included and the review did not assess the application potential of the PBK models for risk assessment purposes. Therefore, the aim of the current paper is to critically review existing PBK models for farm animals in order to develop recommendations for their use in animal and human health risk assessment. Existing models were identified through an extensive literature review and these models were characterised and assessed using criteria set by WHO for practice in human PBK modelling (WHO, 2010). Recommendations are proposed for the future development and application of PBK models in farm animals in chemical risk assessment.

2. MATERIALS AND METHODS

2.1. Species selection

There are more than 40 domestic animal species, whereof 13 species contribute to most of the world’s food production or are of veterinary relevance (Toutain et al., 2010). In Europe, cattle, sheep, swine, chicken and salmonids are classified as major food-producing animals. In contrast, other animals such as goat, horse, rabbit, deer/reindeer, turkey, duck, goose, and non-salmonid fin fish species are classified as minor food-producing animals (EMEA/CVMP, 2003). Based on this list of food-producing animals in Europe, 10 species were selected for inclusion in this review, namely chicken, cattle, swine, sheep, goat, horse, deer/reindeer, turkey, duck, and goose (Figure 2). Fish species were excluded, since a review on PBK models in fish is available elsewhere (Grech et al., 2017). Since the focus of this review is on farm

17 animals, test species used in human risk assessment, such as rabbit, mouse, rats and companion animals (cat, dog) were excluded.

Figure 2. Major food producing animals: cattle (milk, meat), sheep (meat, milk), swine (meat), and chicken (meat, eggs). Minor food-producing animals: goat (milk, meat), deer (meat), horse (meat), avian species (meat, eggs).

2.2. Search strategy

An extensive literature search (up to February 2019) was performed in PubMed and Scopus to identify available PBK models for farm animals. Table 1 summarises the keywords applied. Peer reviewed publications in English were selected, with a focus on PBK modelling of chemicals in the 10 species selected. Individual publications were then assessed based on title and abstract, followed by further cross- referencing of publications from the reference lists. In a second step, a grey literature search was conducted by consulting websites of EFSA, US EPA and the Dutch National Institute for Public Health and the Environment. Publications with a primary focus on preclinical studies for human health were excluded.

Table 1. Keywords identified for extensive literature search Category Keywords General keywords OR OR OR OR OR OR OR OR OR More complex tools and methods and full biologically-based models

18 2.3. Model description and assessment

The selected PBK models for farm animals were characterised and assessed using the model features listed in guidance documents for the evaluation and application of PBK models in HRA (EPA, 2006; WHO, 2010). We grouped these features in six categories: (1) purpose, (2) structure and mathematical representation, (3) computer implementation, (4) parameterisation, (5) performance, and (6) documentation (Table 2).

Table 2. Model features used to characterize and assess PBK models Category Model feature Purpose - What is the model purpose? - What is the included species? - What are the included chemicals? - What age(s), life stage(s), sex, exposure window(s) is included? - What exposure route(s), and dose metric(s) is/are modelled? - What target organs or whole body? Structure and mathematical representation - Is a graphical representation of the model available? - What is the complexity in terms of number of compartments? - Are steady-state or differential calculations used? - Are the mass balance (ADME-) equations given? Computer implementation - What software (package) is used for the implementation? Parameterisation - Are physiological parameter values derived from experiments, literature, or predicted? - Are physicochemical parameter values derived from experiments, literature, or predicted? - Are biochemical parameter values derived from experiments, literature, or predicted? - Has the model been calibrated with a dataset? - Are the calibration data adequately reported? Model evaluation of predictive ability - Has the model been validated against external data (i.e., not those used for calibration)? - Are the validation data adequately reported? - Is a variability analysis performed? - Is a sensitivity analysis performed? - Is a visual inspection of the adequacy of the model predictions possible (e.g., via concentration-time plots?) Model documentation - Has the model been peer-reviewed? - Are model codes and syntax available? - Is the model publicly available?

3. MODEL DESCRIPTION

3.1. Purpose

Thirty-nine PBK models were identified through extensive literature search, covering nine of the farm animal species included (Table 3). Specifically, PBK models were available for cattle (n=10), swine (n=17), chicken (n=6), sheep (n=1), goat (n=1), horse (n=1), and turkey (n=1) while two models included multiple

19 poultry species, i.e., chicken, goose, turkey and duck (Cortright et al., 2009; MacLachlan, 2010). For deer and reindeer, no PBK models were found. Most PBK models were developed for either juvenile or adult life stage, with the exception of five models that included growth rates to simulate lifetime exposure for cattle, swine and poultry (Henri et al., 2017; Hoogenboom et al., 2007; MacLachlan, 2010; van Eijkeren et al., 1998; Zeng et al., 2019) (Table 3). The vast majority of models were chemical-specific and mostly developed for antibiotics and contaminants (e.g., melamine, PFOS) with the exception of four generic models developed for lipophilic chemicals (Freijer et al., 1999; MacLachlan, 2009; 2010; van Eijkeren et al., 1998). The models were developed for specific scientific purposes, such as concentrations predictions of a specific chemical in certain target tissues (e.g. liver, muscle).

3.2. Structure and mathematical description

All available publications provided a graphical representation of the model structure, and 28 out of 37 articles provided mass balance equations. The conceptual structures of the models ranged from two compartments to ten compartments, showing a lack of harmonisation across the published models. Model structures were mostly based on the model purpose rather than on physiological differences between species (Li et al., 2017; Lin et al., 2016b). PBK models with a focus on distribution of the metabolite had at least as many compartments for the metabolite as for the parent compound (Huang et al., 2015a; Yang et al., 2014a). Furthermore, differences between model structures resulted from a reduction in model complexity through “lumping” tissues with similar blood flows (Buur et al., 2009), and from species-specific differences such as the inclusion of a milk compartment for dairy cattle (Leavens et al., 2014; Li et al., 2018) and an egg compartment for chicken (Hekman and Schefferlie, 2011; MacLachlan, 2010; Schefferlie and Hekman, 2016).

In cattle, the simplest model was structured into two compartments (carcass and serum) (van Asselt et al., 2013) and more complex models split liver, adipose tissue, and blood into separate compartments lumping the remaining tissue as richly and poorly perfused (Derks et al., 1994; Freijer et al., 1999; Hoogenboom et al., 2010; van Eijkeren et al., 1998). In addition to the compartments cited above, Leavens et al. (2014) published a comprehensive PBK model with separate compartments for blood, liver, adipose tissue, kidney, muscle, lung, and richly and poorly perfused tissue. Three models included all compartments listed above, while all remaining tissues were aggregated in a “rest of the body” lumped compartment, including richly and poorly perfused tissues (Achenbach, 2000; Li et al., 2017; Lin et al., 2016b; MacLachlan, 2009). Another model included the gastrointestinal tract as a single compartment (Lin et al., 2016b), and four models included milk/mammary tissue compartments to take into account

20 secretion from plasma into milk for dairy lactating cattle (Derks et al., 1994; Leavens et al., 2014; Li et al., 2018; MacLachlan, 2009). Finally, two models accounted for metabolite formation through the inclusion of metabolism rates (Leavens et al., 2014; Lin et al., 2016b).

For swine, 17 PBK models were available, with one model structured into two compartments (carcass, fat) (Hoogenboom et al., 2007) and four models consisting of three compartments (liver, blood, rest of the body) or as a lumped distinction between richly and poorly perfused tissues (Buur et al., 2009; Yang et al., 2014a; Yang et al., 2015b; Zeng et al., 2017). Other models included liver, blood, and kidney (Buur et al., 2008), and a compartment for the rest of the body, or compartments for muscle and adipose tissue (Buur et al., 2005; Buur et al., 2006; Huang et al., 2015a; Qian et al., 2017; Yang et al., 2012; Zhu et al., 2017). Four models included the lung as an additional compartment (Chen and Seng, 2012; Li et al., 2017; Lin et al., 2016b; Yuan et al., 2011), whereas Qian et al. (2017) divided the lung in three sub-compartments. The anatomical most detailed models contained 8-10 compartments (Chen and Seng, 2012; Sjögren et al., 2012; Yuan et al., 2011). In terms of exposure route, nine PBK models integrated oral dosing modules consisting of either stomach and intestine or intestine only (Buur et al., 2008; Buur et al., 2006; Huang et al., 2015a; Lin et al., 2016b; Yang et al., 2014a; Yang et al., 2012; Yang et al., 2015b; Zeng et al., 2017; Zhu et al., 2017). In contrast, Yang et al. (2012) and Qian et al. (2017) included a separate compartment for intramuscular injection. With regards to metabolism, nine models integrated metabolite formation through implementing metabolism rates. The number of compartments for the major metabolite was similar to the parent compound ranging from three to seven compartments. In two models, the metabolite was modelled using a more complex PBK model than that for the parent compound (Yang et al., 2014a; Yang et al., 2015b).

For chicken, most PBK models had similar compartment numbers, albeit with some variations. Exception to this, are two models with two compartments (blood, fat/egg) (Hekman and Schefferlie, 2011; van Eijkeren et al., 2006). In four models liver, blood, (leg) muscle and kidney was included (Cortright et al., 2009; Yang et al., 2015a; Yang et al., 2014b; Zeng et al., 2019), a lung compartment in two models (Yang et al., 2015a; Yang et al., 2014b), and a compartment for adipose tissue in four models (Cortright et al., 2009; Henri et al., 2017; MacLachlan, 2010; Zeng et al., 2019). Oral dosing modules consisting mainly of intestine, and sometimes including crop, were only implemented in four models (Henri et al., 2017; Yang et al., 2015a; Yang et al., 2014b; Zeng et al., 2019). MacLachlan (2010) incorporated egg formation in a PBK model to account for partitioning from plasma into eggs, while Hekman and Schefferlie (2011) split the egg into yolk and white compartment. Metabolite formation was included in a model describing the

21 metabolism of monensin via Michaelis-Menten kinetics (Henri et al., 2017). Metabolite formation for T-2 toxins was described by a metabolism rate (Zeng et al., 2019). For other avian species (turkey, duck, goose), two PBK models were identified (Cortright et al., 2009; MacLachlan, 2010) with identical structures to those described for chicken. One model was specifically developed for turkey, consisting of four compartments, i.e. blood, liver, gastrointestinal tract, and a compartment to lump the rest of the body (Pollet et al., 1985).

For sheep, goat, and horse, only one model per species was available. The PBK model for sheep consisted of six compartments, i.e., adipose tissue, muscle, kidney, liver, blood and a compartment to lump the rest of the body. An intramuscular injection site was included as a specific absorption compartment (Craigmill, 2003). In contrast, the goat PBK model contained eight compartments, i.e., lung, adipose tissue, muscle, liver, kidney, blood, richly perfused tissue, and poorly perfused tissue (Leavens et al., 2012). The PBK model for horse included four compartments, i.e., liver, blood, gastro intestinal circulatory system, and a compartment lumping the rest of the body (Abbiati et al., 2017). This model also included metabolite formation and elimination of tramadol through hepatic clearance.

3.3. Computer implementation

Most PBK models were written in the Advanced Continuous Simulation Language (ACSL, 76%), a computer language designed for modelling and evaluating the performance of continuous systems described by time-dependent, nonlinear differential equations (Mitchell and Gauthier Associates, 1981). 26% used other software (Berkley Madonna, Excel). For 18% of models, information on computer implementation was lacking.

3.4. Parameterisation

Parameters incorporated in PBK models can be divided into physiological parameter (species-specific, e.g. body weight), chemical parameters (chemical-specific, e.g. molecular weight), and biochemical parameters (species- and chemical-specific, e.g. blood-tissue partitioning, protein binding, clearance, absorption). Physiological parameters are dependent on the organisms modelled, e.g. small- or large-sized animals, and are typically obtained from literature or the experiment that is being simulated. Data on interindividual variation in physiological parameters are needed to parameterise a typical individual for deterministic analyses (e.g., worst-case) or to perform probabilistic analyses describing the variation in the population.

22 For most models, physiological and biochemical parameter values were either derived for the species of interest from the literature or using in vivo experiments conducted and described in the same paper (Appendix S1). From the 29 PBK models available, 90% determined some of the biochemical parameters, e.g. clearance and absorption, by means of calibration, i.e. fitting model predictions to experimental data. van Eijkeren et al. (1998) provided a quantitative structure activity relationship (QSAR) to derive blood- tissue partition coefficient, and Buur et al. (2008) and Yuan et al. (2011) extrapolated biochemical parameter values form rats to pigs.

3.5. Performance

From the 39 PBK models available, 67% were validated against an external dataset independent from the calibration dataset. 31% of the PBK model applications included interindividual variability in biochemical parameter values (blood-tissue partition coefficients, clearance, absorption rate), while some also included interindividual variability in physiological parameters (Henri et al., 2017; Li et al., 2017; Zeng et al., 2017; Zeng et al., 2019). Local sensitivity analysis was performed for 54% of the models to assess the impact of variations in a single input parameter on the model output. A global sensitivity analysis investigating the interactions and the impact of simultaneous variations in input parameters is lacking for all models.

3.6. Documentation

Most models were published in peer-reviewed literature (90%), while others were available in reports only. Model codes were published for 28% of the models.

23 Table 3. Published PBK models for various farm animals Reference Species Life Chemicals Exposure Number of Mass Software Model Calibration3 Validation3 Variability Sensitivity Peer- stage route(s) compart balance code analysis analysis reviewed ments1 equations available performed performed Derks et al. Cattle Mature 2,3,7,8-TCDD oral 6 No N/A2 No Literature Literature No No No (1994) van Eijkeren et Cattle Lifetime Lipophilic oral 5 No ACSL Yes N/A2 N/A2 No No No al. (1998) contaminates Freijer et al. Cattle Mature 2,3,7,8-TCDD, oral 5 Yes ResAnA No Literature Literature No No No (1999) Lindane, PCB- 169 Achenbach Cattle N/A2 Oxytetracycline im 8 No Microsoft No Experiment Experiment No Yes No (2000) (steers); Excel sc (calves) MacLachlan Cattle, Mature Lipophilic oral 7 Yes lcc C No N/A2 N/A2 No No Yes (2009) Swine, xenobiotics compiler Sheep Hoogenboom Cattle Mature Dioxin oral 5 No N/A2 No Literature N/A2 No No Yes et al. (2010) van Asselt et Cattle Mature PFOS oral 2 Yes ACSL No Literature Literature No No Yes al. (2013) Leavens et al. Cattle Mature Flunixin iv; im 9 Yes acsIX No Literature N/A2 No Yes Yes (2014) Lin et al. Cattle, Mature, Ceftiofur, iv, sc, im, 8; 7 No acslX Yes Literature Literature No Yes Yes (2016b) Swine Juvenile Enrofloxacin, oral Flunixin, Sulfamethazine Li et al. (2017)4 Cattle, Mature, Penicillin G im, sc 7 Yes Berkeley Yes Literature Literature Yes Yes Yes Swine Juvenile Madonna Leavens et al. Goat Juvenile, Tulathromycin sc 8 Yes acslX No Experiment N/A2 No Yes Yes (2012) market age Abbiati et al. Horse Mature Tramadol iv 4 Yes N/A2 No Literature N/A2 No No Yes (2017) van Eijkeren et Chicken Mature Dioxin, PCBs oral 2 Yes N/A2 No Literature Literature No No Yes al. (2006) Cortright et al. Chicken, Mature Midazolam iv 6 Yes ACSLXtreme Yes Experiment Experiment No Yes Yes (2009) Turkey MacLachlan Chicken, Lifetime Lipophilic oral 6 Yes lcc C No N/A2 N/A2 No No Yes (2010) Duck, pesticides compiler Goose, Turkey

24 Hekman and Chicken Mature sulphanilamide, oral 3 Yes Multisim No Literature N/A2 No No Yes Schefferlie pyrimethamine, (2011)5 sulphaquinoxali ne Yang et al. Chicken Juvenile Marbofloxacin oral 7 Yes ACSLXtreme No Literature Literature Yes Yes Yes (2014b) Yang et al. Chicken Mature Danofloxacin oral 7 Yes ACSLXtreme No Literature Literature Yes No Yes (2015a) Henri et al. Chicken Lifetime Monensin oral 6 Yes ACSLXtreme Yes Experiment Experiment Yes Yes Yes (2017) Zeng et al. Chicken Lifetime T-2 oral, iv 6; 6 No Berkeley Yes Experiment Literature Yes Yes Yes (2019) Madonna Pollet et al. Turkey N/A2 Chlortetracyclin oral 4 Yes N/A2 No Experiment N/A2 No No Yes (1985) e Buur et al. Swine Market Sulfamethazine iv 6; 3 Yes ACSLXtreme No Literature Literature No Yes Yes (2005) age Buur et al. Swine Market Sulfamethazine iv, oral 8; 3 Yes ACSLXtreme No Literature Experiment; Yes Yes Yes (2006)6 age Literature Hoogenboom Swine Lifetime Dioxins oral 2 No N/A2 No Literature N/A2 No No Yes et al. (2007) Buur et al. Swine Market Melamine iv, oral 6 Yes ACSLXtreme No Literature Literature No Yes Yes (2008) age Buur et al. Swine Market Sulfamethazine iv 4; 3 Yes ACSLXtreme No Literature Literature No No Yes (2009) age , flunixin Yuan et al. Swine N/A2 Valnemulin oral 8 No ACSLXtreme No Experiment N/A2 No No Yes (2011) Chen and Seng Swine N/A2 Soman Iv 10 No N/A2 No Experiment Experiment No No Yes (2012) Sjögren et al. Swine N/A2 Repaglinide iv 9 Yes Berkeley No Experiment Experiment No Yes Yes (2012)7 Madonna Yang et al. Swine Market Doxycycline im, iv 7 Yes ACSLXtreme No Literature Literature Yes No Yes (2012) age Yang et al. Swine Juvenile Olaquindox oral 5; 6 Yes ACSLXtreme No Experiment; Literature Yes Yes Yes (2014a) Literature Yang et al. Swine Juvenile Cyadox oral 4; 6 Yes ACSLX No Literature Literature No Yes Yes (2015b) Huang et al. Swine Juvenile Cyadox oral 7; 7 Yes ACSLX No Literature Experiment Yes Yes Yes (2015a) Zhu et al. Swine N/A2 Quinocetone, oral 8; 7 Yes acslX No Experiment Experiment Yes Yes Yes (2017) Mequindox Zeng et al. Swine Juvenile Mequindox oral, im 5; 6 Yes Berkeley Yes Experiment Experiment Yes Yes Yes (2017) Madonna

25 Qian et al. Swine N/A Florfenicol im 7 Yes ACSLXtreme Yes Experiment Literature No No Yes (2017) Craigmill Sheep Mature Oxytetracycline im 7 Yes ACSL Yes Experiment Experiment No Yes Yes (2003) (from calibration) im: intramuscular, sc: subcutaneous; iv: intravenous; 1 first value indicates the number of the compartments parent compound, the second value when available indicates the number of compartments metabolite 2 N/A: not available; 3 Experiment means that data is based on experiments described in the article itself. Literature means that experimental data is based on other articles. 4 extended with udder compartment (Li et al., 2018); 5 extended by detailed egg compartment (Schefferlie and Hekman, 2016); 6 validated by Mason et al. (2008); 7 extended with fexofenadine (Sjogren et al., 2014)

26 4. MODEL ASSESSMENT

In the previous section, an overview of the 39 PBK models for farm animals identified in from the extensive literature was presented (Table 3). In total, four PBK models were fully described and characterised according to the WHO guidance document (Henri et al., 2017; Li et al., 2017; Zeng et al., 2017), whereas all other models were missing information. The aim of the present section is to assess their potential usefulness for application in chemical risk assessment based on the model features presented in Tables 2 and 3.

4.1. Scope and purpose of the model

None of the reviewed farm animal models were developed and applied to a risk assessment context for either animal or human health purposes. Instead, most models were developed for a specific scientific purpose, such as the predicting the concentrations of a specific chemical in certain target tissues (e.g. liver or muscle). Consequently, these models were unique, each tailored for a specific purpose, often focusing on one particular substance in one particular species. These models can be useful for substance- and species-specific risk assessments, but risk assessors in practice have to deal with a diversity of substances and species. Within this context, the development and application of separate PBK models for each unique substance-species combination is inefficient and practically unfeasible (Punt et al., 2011). Risk assessors need more generic PBK models that can simulate ADME processes for multiple substances, and ideally also for multiple farm animal species. The four models developed for lipophilic substances (Freijer et al., 1999; MacLachlan, 2009; 2010; van Eijkeren et al., 1998) constitute a first step in the right direction since these models can simulate multiple substances in one species. Grech et al. (2019) went one step further by publishing an open source generic PBK model for four fish species (rainbow trout, fathead minnow, zebra fish, and European stickleback) together with an anatomical and physiological database and nine case studies to validate the models.

4.2. Model structure and mathematical description of ADME

Although the PBK models reviewed have many commonalities, each model has its own specific structure, often reflecting the uniqueness of the substance-species combination under study. The simplest model has two compartments (van Asselt et al., 2013), whereas the most extensive model has ten (Chen and Seng, 2012). As noted in the previous section, risk assessors require generic PBK models which can be applied to multiple substances and species. This triggers the question how these generic models should be structured. A generic PBK model should be sufficiently detailed to capture the diversity of processes relevant to a multitude of chemicals and the diversity of organs that may potentially be of interest to risk assessors. Liver, intestine and kidney should be included as compartments because of their metabolic and excretory functions. Muscle, adipose tissue, bone, milk

27 and/or eggs should be included because of their role as storage organs, which is important for analysing tissue residues in food and feed safety. In addition, organs critical to the oral route or inhalation should also be included, i.e. gastrointestinal tract and lung respectively (EPA, 2006).

In terms of structure, some compartments are only relevant to specific species because of specific traits including crop and egg compartment for birds and multiple stomachs for ruminants. In addition, compartments can also be substance-specific including the subdivision of the lung compartment into bronchioles, alveolae and mucosa to simulate the so-called washin–washout principle for polar chemicals after inhalation (Johanson, 1991). Overall, these species- and substance-specific adaptations provide a challenge for generic PBK models and an option would be to develop a flexible modelling structure that can be adjusted based on the species and substances of interest. The basic structure would then include compartments common to all species, e.g. lungs, stomach, intestines, liver, kidney, bone and muscle, while specific compartments could be implemented as optional, i.e. switched on or off to accommodate species- or substance-specific requirements.

4.3. Computer implementation

Application of PBK models in a risk assessment context for either regulatory products of contaminants requires transparency and reproducibility. These criteria can partially be met through implementation of the model in an open source environment. However, none of the reviewed models were implemented as such. The most widely used computer language ACSL is of commercial nature and is no longer available. Another type of software is Berkeley-Madonna, which is non-commercial proprietary shareware but is not open source, since a license is necessary for full access. A more promising trend is the development of PBK models in R (Wambaugh et al., 2015). The R language and environment was originally developed for statistical computing and graphics, but is increasingly being used for dynamic modelling. This language is freely available under the terms of the GNU General Public License in source code form (Free Software Foundation, 2007). It compiles and runs on a variety of platforms, including UNIX, Windows and MacOS. Recently, a methodology was published for converting existing PBK models that were initially developed in acslX, Berkeley Madonna to R (Lin et al., 2017).

4.4. Parametrisation

PBK models applied in chemical risk assessment must be parameterised transparently since the selection of input values have an important impact on the results. The selected physiological parameters should reflect the population of interest. This is highly relevant for farm animals, since the variation in their physiology can be much larger compared with humans. For example, the bodyweight

28 of pigs can vary between 100 kg and 300 kg depending on the breed. It is therefore important that risk assessment includes a detailed and motivated description of the population of interest, and considers how variation in physiology will be dealt with. For human physiology, a reference dataset has been compiled to parameterise PBK models (Brown et al., 1997). Such datasets are largely lacking for farm animal species. Ad-hoc parameterisation is inefficient and can potentially lead to inconsistencies between specific risk assessments. Hence, it is recommended to develop peer reviewed reference datasets reporting physiologically parameters for farm animal species.

In a risk assessment context, calibration of parameters such as clearance and absorption will rarely be feasible because experimental data on the species of interest are generally lacking, particularly for new substances. In such cases, extrapolation across chemicals and species may be an option, e.g. using QSAR-based approaches and allometric scaling (Peyret and Krishnan, 2011; Sharma and McNeill, 2009). Although these methods can provide valuable input in data-poor situations, uncertainty in prediction of kinetic parameters may be high due to potential interspecies differences in metabolic and transporter activities. An alternative is to use in vitro data, e.g. for the prediction of metabolism. Unfortunately, specific in vitro systems for farm animal species are rarely available and the quantitative extrapolation of in vitro results to in vivo conditions (QIVIVE) remains a challenge. Therefore, data collection on abundance and expression of CYP enzymes and transporters in a variety of farm animal species can support QIVIVE and the parameterisation of ADME processes in PBK models. Insight in the quantitative differences in metabolism for key farm animal species will allow to include relative expression and activity of phase I (cytochrome P-450 isoforms, esterases, alcohol dehydrogenase etc.), phase II (UDP-glucuronyltransferases, glutathione-s-transferases, glycine conjugation, methyl-transferases etc.) and transporters (P-glycoprotein, organic anion transporter proteins etc.) (Dorne and Fink-Gremmels, 2013; Fink-Gremmels, 2008; Giantin et al., 2008; Gusson et al., 2006; Martinez et al., 2018a). This will help to address three major challenges in animal risk assessment: 1. Moving away from default test species (rat, rabbit, mice, dog) by enabling species- specific assessments for farm animals and other animal species (e.g., companion animals such as cats and dogs), 2. Reducing and replacing animal testing, and (3) future development of quantitative in vitro in vivo extrapolation models.

4.5. Model performance

Before PBK models can be applied in a risk assessment context, it is essential to assess their performance (EFSA, 2014c). Model predictions should be sufficiently accurate to support risk assessment or regulatory decision to be taken. The most obvious way to assess model performance is by comparing predicted internal concentrations with measured ones under the same exposure

29 scenario. Unfortunately, measured data are rarely available for the substance and scenario under assessment since this would make model application redundant. Model performance should therefore be assessed based on other criteria, such as the performance for other substances and scenarios, biological plausibility, and a review of the model code. Variability, sensitivity and uncertainty analysis can further support the assessment of model performance. Uncertainty analysis should not only focus on the propagation of input uncertainties, but also on uncertainties underlying model equations and assumptions, e.g. the homogeneous distribution of substances within compartments and the estimation of absorption or partitioning coefficients based on Kow (Ragas et al., 1999).

In the context of the review, it is difficult to assess the performance of the reviewed models since: 1) only local sensitivity analysis was performed, 2) model validation was not performed systematically, and 3) codes for computer implementation were rarely published (Leavens et al., 2014; MacLachlan, 2009; van Eijkeren et al., 1998). Local sensitivity analysis does not allow to investigate interactions and simultaneous variations in input parameters and may give misleading results (Hsieh et al., 2018). In terms of good modelling practice, global sensitivity analysis has been demonstrated to be much more robust than local sensitivity analysis, providing means to assess the variation of all parameters simultaneously, their interactions and quantifying the relative contribution of each parameter to the overall sensitivity of the model (Hsieh et al., 2018; McNally et al., 2011; McNally et al., 2018).

4.6. Model documentation

Only a small part of the reviewed PBK models were appropriately documented, i.e. published in peer- reviewed literature and publicly available (model and underlying code). Appropriate documentation contributes to transparency and reproducibility; two essential criteria for application of PBK models in a regulatory context. It is a prerequisite for well-informed use of the models (McLanahan et al., 2012).

5. CONCLUSIONS AND RECOMMENDATIONS

The available PBK models for farm animal species have generally been developed for specific purposes, often focusing on one particular substance in a specific species. This approach of developing unique models for each substance and each species is less suitable for a risk assessment context because a risk assessor has to deal with multiple regulated substances and a diversity of species. Most models also do not meet the criteria for application in risk assessment, such as (1) implementation in an open source environment, (2) parameterisation in the absence of calibration data, (3) assessment of model performance in the absence of validation data (e.g. uncertainty analysis), and (4) appropriate

30 model documentation. Future PBK models for farm animals to be applied in a risk assessment context should therefore: 1) Have a generic and flexible model structure covering all relevant target organs; 2) Be published in peer-reviewed literature, implemented in an open source environment and publicly available (model & code); 3) Use input data that are not dependent on calibration, preferably empirical data, or otherwise data estimated using techniques such as allometric scaling, QSAR approaches and QIVIVE, provided the limitations of these techniques are accounted for; 4) Include tools to assess model performance, e.g. variability and uncertainty analyses. Development and application of PBK models for farm animal species would furthermore benefit from the setup of structured databases providing data on physiological parameters (e.g., body weight, organ weights, and organ blood flow), ADME and chemical-specific parameters as well as data on enzyme expression and activities in farm animal species to support the development of species- specific QIVIVE models.

Funding information This work was supported by the European Food Safety Authority (EFSA) [Contract number: EFSA/SCER/2014/06].

31

CHAPTER 3

GENERIC PHYSIOLOGICALLY BASED KINETIC MODELLING FOR FARM ANIMALS: PART I. DATA COLLECTION OF PHYSIOLOGICAL PARAMETERS IN SWINE, CATTLE AND SHEEP

L.S. Lautz, J.L.C.M. Dorne, R. Oldenkamp, A.J. Hendriks, A.M.J. Ragas

Published in Toxicology Letters, 319:95-101

32 Abstract

Physiologically based kinetic (PBK) models for farm animals are of growing interest in food and feed safety with key applications for regulated compounds including quantification of tissue concentrations, kinetic parameters and the setting of safe exposure levels on an internal dose basis. The development and application of these models requires data for physiological, anatomical and chemical specific parameters. Here, we present the results of a structured data collection of anatomical and physiological parameters in three key farm animal species (swine, cattle and sheep). We performed an extensive literature search and meta-analyses to quantify intra-species variability and associated uncertainty of the parameters. Parameters were collected for organ weights and blood flows in all available breeds from 110 scientific publications, of which 29, 48 and 33 for cattle, sheep, and swine, respectively. Organ weights were available in literature for all three species. Blood flow parameter values were available for all organs in sheep but were scarcer in swine and cattle. Furthermore, the parameter values showed a large intra-species variation. Overall, the parameter values and associated variability provide reference values which can be used as input for generic PBK models in these species.

33 1. INTRODUCTION

Physiologically based kinetic (PBK) models are increasingly being used by scientific advisory bodies and regulatory authorities to estimate internal concentrations of chemicals or their metabolites in a various body fluids and tissues (Paini et al., 2017). These models can be applied to support exposure assessment to perform route-to-route extrapolation, to evaluate biomonitoring data, to characterise variability between individuals, and to quantify variability and uncertainty in physiological and kinetic parameters (Beaudouin et al., 2010; Bois et al., 2010; EFSA, 2014b; Fierens et al., 2016; Huizer et al., 2014; McNally et al., 2012). In the pharmacology field, PBK models are also applied in drug discovery and development (EMA, 2018).

PBK models are built by integrating mathematical relationships describing the kinetics of a chemical in different physiological compartments of an organism processes. A number of parameters are required to build a PBK model and these can be divided into physiological/anatomical descriptors (e.g., blood flow, organ volumes), and biochemical descriptors (e.g., partition coefficients, protein binding) (EFSA, 2014b; WHO, 2010). Biochemical descriptors are chemical-specific, whereas the physiological and anatomical descriptors are species-specific. It is foreseen that systematic data collection of physiological parameters in relevant species, and the creation of open source databases will support the development and application of PBK models in risk assessment (Madden et al., 2019). In addition, quantification of variability and uncertainty associated with these parameters (e.g. Marchov Chain Monte Carlo) and global sensitivity analysis using probabilistic approaches, are becoming increasingly important to investigate key variables that impact on model outputs and to support validation of the models (Bois et al., 2010; Li et al., 2019a).

Over twenty years ago, Brown et al. (1997) published a thorough comparative data collection of physiological and anatomical parameters for four standard laboratory test species and humans which provided a basis to develop generic PBK models for those species. Especially for human, various databases have been described (Madden et al., 2019). For fish species, an anatomical and physiological database for four species (rainbow trout, fathead minnow, zebra fish and European stickleback) has recently been published (Grech et al., 2019). For sheep and swine, only limited data have been published, e.g. mean physiological data for one sheep breed (Merino) and small pigs (25kg), both often used for preclinical drug studies (Upton, 2008). Thorough data for anatomical and physiological parameters in a range of farm species and breeds are currently lacking, impeding the development of generic open source PBK models and this has been highlighted in recent review papers (Lautz et al., 2019; Li et al., 2019a; Lin et al., 2016a).

34 This paper aims to perform a systematic data collection on physiological and anatomical parameters for swine, cattle and sheep as key farm animal species by means of extensive literature searches and meta-analyses. A part II of this manuscript describes the development of the generic PBK models in swine, cattle and sheep, their validation using global sensitivity analysis as well as case studies to predict blood and tissue concentrations for compounds that are renally excreted (Lautz et al., 2020b).

2. MATERIAL AND METHODS

2.1. Extensive Literature Search

Extensive literature searches were conducted in PubMed and Google Scholar until December 2018 to identify all relevant peer-reviewed publications reporting physiological parameters, such as organ volumes and blood flow rates, using relevant keywords described in Table 1. Each reference identified in the extensive literature searches and its associated bibliography were screened for relevance based on title and abstract. In a next step, the full text of each individual study was scrutinised and evaluated critically for their relevance. Parameter values for body weight, organ weight, and values for blood flow distribution were extracted when available for healthy, mature animals. Animals were classified as healthy when no information on altered health conditions were given. When parameters for healthy animals were not available, healthy, juvenile animals (pigs >25 kg, sheep and cattle: ruminants only) were included. Data for animals fed on a conventional (grass, corn-based) diet were included. In contrast, data for animals fed an altered diet such as protein rich or protein poor diet, or with added supplements were excluded to collect all parameters under physiological conditions. Data for animals with non-typical traits (e.g. obese animals), kept under adverse environmental conditions, or exposed to drugs or toxic agents, were excluded. Mean values for organ weight, represented as a fraction of body weight, were only derived if studies explicitly provided body weight data or data presented as percent body weight. If organ blood flow was reported in ml/100g/min or similar units, the mean organ weight was used to calculate the percent of cardiac output.

Table 1. Keywords applied in the extensive literature searches for the data collection of physiological and anatomical parameters in swine, cattle and sheep Type Keywords Species OR OR OR Biological and physiological OR OR OR OR OR OR OR OR OR OR OR

2.2. Meta-analyses

From the data collection, meta-analyses were performed for each physiological parameter and species obtained. All studies assumed that the physiological and biological parameters followed a normal

35 distribution (arithmetic mean and standard deviation). Organ weight and blood flow fractions were calculated based on the reported arithmetic mean (AM) and standard deviation (SD) (Equations 1 & 2):

퐀퐌퐎,퐢 퐀퐌퐟,퐢 = (Equation 1) 퐀퐌퐁

where AMf,i is the arithmetic mean body weight or blood flow fraction for organ i; AMO,i is the arithmetic mean weight or blood flow for organ i; and AMB is the arithmetic mean body weight or cardiac output;

ퟐ ퟐ 퐒퐃퐎,퐢 퐒퐃퐁 퐒퐃퐎,퐢∙퐒퐃퐁 퐒퐃퐟,퐢 = 퐀퐌퐟,퐢 ∙ √( ) + ( ) − ퟐ ∙ ∙ 훒[퐎𝑖, 퐁] (Equation 2) 퐀퐌퐎,퐢 퐀퐌퐁 퐀퐌퐎,퐢∙퐀퐌퐁

where SDf,i is the standard deviation of the body weight or blood flow fraction for organ i; SDO,i is the standard deviation of body weight or blood flow for organ i; SDB is the standard deviation of the body weight or cardiac output; ρ[Oi, B] is the Pearson’s correlation coefficient between the organ weight or blood flow and body weight or cardiac output, respectively. When ρ[Oi, B] was not given, no correlation was assumed (ρ[Oi, B]=0).

When multiple studies were available for the assessed parameter, the studies were combined using the method described by Ragas and Huijbregts (1998), in which AM and SD were combined based on the sample size of the respective studies (Equation 3 & 4):

퐱=퐧 ∑퐱=ퟏ 퐍퐱∙퐀퐌퐱 퐀퐌퐜퐨퐦퐛퐢퐧퐞퐝 = 퐱=퐧 (Equation 3) ∑퐱=ퟏ 퐍퐱

퐱=퐧 ퟐ ퟐ ퟐ ퟐ ∑퐱=ퟏ[(퐍퐱−ퟏ)∙퐒퐃퐱+퐍퐱∙퐀퐌퐱 −ퟐ∙퐍퐱∙퐀퐌퐱∙퐀퐌퐜퐨퐦퐛퐢퐧퐞퐝+퐍퐱∙퐀퐌퐜퐨퐦퐛퐢퐧퐞퐝] 퐒퐃퐜퐨퐦퐛퐢퐧퐞퐝 = 퐱=퐧 (Equation 4) (∑퐱=ퟏ 퐍퐱)−ퟏ

where AMcombined represents the combined mean; AMx is the mean of study x; SDx is the standard deviation of individual study x; Nx is the sample size of study x; and n is the number of studies to be combined.

The original dataset reporting all anatomical and physiological parameters included in the meta- analyses is available in excel format on EFSA knowledge junction under the DOI: 10.5281/zenodo.3433224 with a Creative Common Attribute 4.0 license.

Since dairy cattle and beef cattle are selected for on different traits (i.e. milk yield versus meat yield), a two-tailed Student t-test (p<0.001) was performed to test for significant intra-species differences between dairy cattle and beef cattle.

36 3. RESULTS

3.1. Anatomical and physiological parameters

Anatomical and physiological parameters were retrieved from 110 scientific publications, i.e. 29 for cattle, 48 for sheep, and 33 for swine. For each species, data was collected for organ weights (adipose tissue, blood, brain, bone, heart, intestine, kidney, liver, lung, muscle, and mammary tissue) and blood flows. In order to get representative physiological values covering the whole species, various breeds were included. For cattle the data on the following breeds were included: Aberdeen Angus, Angus x Gelbvieh, Angus x Hereford, Angus x Simmental, Ayrshire, Braunvieh, Charolais, Friesian x Ayrshire, Gelbvieh, Guernsey, Hereford, Holstein, Holstein x Friesian, Holstein x Gyr, Jersey, Jersey x Limousin, Limousin, Pinzgauer, Red Poll, Salers, Simmental, and Swedish Red and white. For sheep most data §was available for Merino sheep, however also other breeds were described, such as Clun Forest, Dorper, East Friesian, Friesian, INRA 401, Lacaune, Leicester x Swaledale, Omani, Rambouillet, Romney Marsh, Santa Ines, Sarda, Swifter, Targhee, and Western Whiteface. For swine, physiological values for the following breeds were available: Duroc, Duroc x Large White x Yorkshire, Landrace and Landrace crossbreeds, Meishan, Pietrain, Purebred, Wild Boar, Yorkshire, and York x Hampshire. Meta-analyses were conducted to characterise the intra-species variability of the parameters and their associated uncertainty.

Results of the meta-analyses for the most common physiological and anatomical parameter values expressed as percentage of body weight are provided in Table 2. A distinction between dairy cattle and beef cattle was made the weight fraction of muscle, due to significant differences. The highest intra-species variability was observed in the fraction of brain weight for swine, the fraction of adipose tissue for sheep, and the fraction of lung tissue in cattle. For other parameters, the coefficient of variation was below 30%.

Results of the meta-analyses of blood flow parameter values as percentage of cardiac output are presented in Table 3. A significant difference was observed between dairy cattle and beef cattle for kidney and liver blood flows. Due to these differences, blood flows for dairy cattle and beef cattle are reported separately. From the three species included, sheep is the only one, for which blood flow to each organ was reported in literature. Large intra-species variability for blood flow was observed, especially for the percentage blood flow through the hepatic artery. In sheep and swine, blood flow fractions through the muscles also shows high intra-species variability.

37 Table 2. Mean organ weights of cattle, sheep, and swine as percentage of body weight Body Adipose Mammary Species Blood Brain Carcass Heart Intestine Kidney Liver Lung Muscle References weight tissue gland N 38976 3663 70 1116 3705 1566 71 1557 1592 1541 54 23 BW (%) 526.1 18.4 3.8 0.1 12.8 0.4 1.7 0.2 1.3 0.8 34.0a; 47.0b 2.2 Cattle 1-21 CV (%) 14 26 12 14 16 15 27 14 16 33 21b, 16c 34 ρ(BW,OW) - 0.94 0.96 0.74 NA 0.96 0.41 0.92 0.93 0.91 NA NA N 21231 86 102 133 62 309 84 242 293 207 62 75 BW (%) 58.4 19.2 4.7 0.2 9.8 0.4 5.4 0.3 1.5 1.1 35.3 1.7 Sheep 22-53 CV (%) 21 39 21 26 16 21 19 24 22 28 24 49 ρ(BW,OW) - 0.9 NA 0.73 NA 0.39 NA NA NA NA NA NA N 18221 15860 216 46 497 2562 206 458 2523 150 15880 NA BW (%) 108.0 17.6 3.0 0.1 9.3 0.4 3.2 0.3 1.7 0.8 45.0 NA Swine 54-76 CV (%) 15 27 18 68 21 23 24 21 19 27 6 NA ρ(BW,OW) - NA NA 0.72 0.95 0.95 NA 0.92 0.91 0.748 0.98 NA N: sample size (number of individuals); BW%: percentage body weight; CV: coefficient of variation; rho: ρ(BW,OW); Bodyweight in kg; NA: not available; a muscle fraction of dairy cattle; b muscle fraction for beef cattle Cattle 1-21: (Ballarin et al., 2016; Bellmann et al., 2004; Crile and Quiring, 1940; Ellis et al., 2016; Holt et al., 1968; Holtenius and Björnhag, 1989; Hunter, 2010; Jenkins and Ferrell, 1997; Jurie et al., 2007; Long et al., 2010; Long et al., 2012; Martinez et al., 2006; McDowell et al., 1987; Morris et al., 2010; Nephawe et al., 2004; Pfuhl et al., 2007; Ren et al., 2002; Rotta et al., 2015; Swett, 1937; von Soosten et al., 2012; Wood et al., 2013) Sheep 22-53: (Barnes et al., 1983; Bennett, 1973; Boxenbaum, 1980; Brown and Swan, 2014; Burrin et al., 1990; Carlson et al., 2009; Charismidou et al., 2000; Clarke et al., 2001; Delavaud et al., 2007; Ebinger, 1974; Gardner et al., 2005; Grace, 1983; Graham et al., 1982; Hales and Fawcett, 1993; Holt et al., 1968; Holtenius and Björnhag, 1989; Jenkinson et al., 1995; Juca et al., 2016; Kamalzadeh et al., 1998; Louey et al., 2005; Mahgoub and Lodge, 1998; McCutcheon et al., 1993; Metcalfe et al., 1962; Moss et al., 2005; Norberg et al., 2005; Pethick and Lindsay, 1982; Reed et al., 2007; Sinclair et al., 2010; Swanson et al., 2008; Tilahun et al., 2014; Wallace et al., 2002) Swine 54-77: (Andersson-Eklund et al., 1998; Fraga et al., 2009; García-Valverde et al., 2008; He et al., 2015; Holt et al., 1968; Hunter, 2010; Kerr et al., 1995; Kruska and Rohrs, 1974; Martinez et al., 2006; Martinsen et al., 2015; Minervini et al., 2016; Moughan et al., 1990; Müller et al., 2000; Nieto et al., 2012; Njoku et al., 2015; Orcutt et al., 1990; Pekas, 1983; Quinious and Noblet, 1995; Razmaite et al., 2009; Ruusunen et al., 2007; Thein et al., 2003; Tranquilli et al., 1982; Wiseman et al., 2007; Yang and Lin, 2010)

38 Table 3. Average blood flows of cattle, sheep, and swine as percentage of cardiac output Cardiac Adipose Mammary Species Brain Carcass Heart Kidney Liver HA Liver PV Lung Muscle References output tissue gland N 89 1 NA NA NA 15 163 163 NA 1 14 Dairy Cattle CO (%) 59.7 6.8 NA NA NA 1.6 11.8 54.2 NA 1.9 16.5 CV (%) 21 24 NA NA NA 19 62 33 NA 30 25 1-10 N 89 1 NA NA NA NA 140 140 NA 1 NA Beef Cattle CO (%) 59.7 6.8 NA NA NA NA 3.1 21.5 NA 26.3 NA CV (%) 21 24 NA NA NA NA 75 30 NA 31 NA N 77 15 41 5 40 131 56 80 8 21 40 Sheep CO (%) 6.4 2.3 1.9 6.6 4.5 14.3 2.6 38.5 3.0 33.2 7.4 11-30 CV (%) 26 46 39 42 42 51 123 39 54 104 46 N 228 11 20 NA 17 17 37 37 11 11 NA Swine CO (%) 10.9 11.0 23.0 NA 4.2 9.8 4.7 17.8 2.1 29.2 NA 31-41 CV (%) 34 66 56 NA 43 38 77 32 63 63 NA N: sample size (number of individuals); CO%: percentage cardiac output; CV: coefficient of variation; Cardiac output in L/min; Liver HA: liver hepatic artery; Liver PV: Liver portal vein; NA: not available; no correlation assumed between cardiac output and blood flow Cattle 1-10: (Bell et al., 1976; Boonsanit et al., 2012; Doyle et al., 1960; Ellis et al., 2016; Fisher and Dalton, 1961; Hallowell et al., 2007; Lomax and Baird, 1983; McDowell et al., 1987; Purdie et al., 2008; Will et al., 1962) Sheep 11-30: (Barnes et al., 1983; Bergman et al., 1971; Di Giantomasso et al., 2003; Dodic et al., 2001; Evans et al., 1998; Fegler and Hill, 1958; Fleet and Mepham, 1985; Freetly and Ferrell, 1997; Hales, 1973; 1976; Hales and Fawcett, 1993; Liu et al., 2015; Norberg et al., 2005; Pethick and Lindsay, 1982; Runciman et al., 1984; Runciman et al., 1986; Talke et al., 2000; Thompson, 1980; Ullman et al., 2001; von Engelhardt and Hales, 1977) Swine 31-41: (Carretero et al., 2010; Kuipers et al., 1999; Kurita et al., 2002; Kurita et al., 2013; Lundeen et al., 1983; Mosing et al., 2015; Shih et al., 2013; Stonestreet et al., 1998; Thein et al., 2003; Tranquilli et al., 1982; van Essen et al., 2018)

39 4. DISCUSSION AND CONCLUSION

Here, extensive literature searches and structured data collection on physiological and anatomical parameters in three key farm animal species (swine, cattle and sheep) are described. Subsequently, meta- analyses were performed to quantify intra- and inter-species variation in those parameters. Such information can for example be used to perform a Monte Carlo simulation to analyse the impact of interindividual variability and uncertainty on PBK predictions (Huizer et al., 2012).

Organ weights for all three species were available in literature. The larger variability observed in the brain weight fractions in swine might be caused by large differences in body weight influencing the brain-to- body ratio (Minervini et al., 2016). The variability observed in adipose tissue fractions in sheep might be caused by differences between male and female, but also various breeds (Bennett, 1973; Clarke et al., 2001; Delavaud et al., 2007). It should be noted that the reported organ weight fractions cannot be used directly as input for PBK models since PBK model compartments are defined by volume rather than mass. The density of most organs is ranging from 1.02–1.06, meaning mass-to-volume conversion is not necessary (Brown et al., 1997). Blood flow parameter values were available for all organs in sheep but were very limited in cattle. For swine, blood flow parameter values were available for juvenile pigs (body weight range 25-38kg); however, none were available for mature pigs. Furthermore, the parameter values showed a large variation, which may reflect true intra-species differences, but it is more likely that the observed variation is caused by interlaboratory variability using different experimental techniques for blood flow measurements (Brown et al., 1997). With regard to blood flow, experimental studies using radiolabelled microsphere techniques, originally reported by Rudolph and Heymann (1967), have become the preferred method of choice for measuring the distribution of blood flow among and within organs in animals (Hoffman et al., 1977). In addition, data for studies using other techniques for blood flow measurements were also included, such as those using a thermodilution or indicator-dilution method (Bergman et al., 1971; Fleet and Mepham, 1985; Freetly and Ferrell, 1997; Ullman et al., 2001). Consequently, blood flow parameters characterising the vascular system in these three species are likely to have considerable uncertainty.

The meta-analyses provide estimates for organ weights and tissue blood flows expressed in percentage of body weight or cardiac outputs. The means of the included parameters do not add up to 100% of body weight or cardiac output. Variability and uncertainty are associated with these estimates particularly because of inter-study variability and data gaps in the availability of physiological parameters for each species. Allometric scaling between species can be used to estimate mean physiological parameters to fill 40

such data gaps as inputs for PBK modelling for individuals using a mean scaling exponent of 0.75. For intra- species differences, the exponent applied shows higher variation (0.3-1.8) and quantifying intra-species variability and uncertainty for the whole population is more challenging because of data gaps (Feldman and McMahon, 1983; Glazier, 2005; Lindstedt and Schaeffer, 2002). In order to fill in such data gaps for physiological parameters (e.g. blood flow and organ weights), authors have proposed to include a 30% default coefficient of variation associated with mean parameter values in the model to provide quantify intra-species variability and uncertainty for the whole population (Clewell and Clewell, 2008). Finally, global sensitivity analysis for the PBK model is recommended to assess the influence of each parameter on the final model output (Hsieh et al., 2018; McNally et al., 2011).

Overall, the anatomical and physiological parameters reported in the present study provide reference values for the development of generic PBK models and other allometric scaling models in cattle, swine and sheep. It is foreseen that such data collection can be performed for other species of farm animals (e.g. chicken, goat, duck, salmon) and vertebrate species of ecological relevance (e.g. quail and other birds, test fish species) in adults and other life-stages (Li et al., 2019a; Li et al., 2019b). In addition, such anatomical and physiological data should be available as open source databases to allow the scientific community to use and update the datasets when more data become available. In order to facilitate the further use of physiological databases in the field of food and feed safety and risk assessment, the anatomical and physiological databases for the three species have been harmonised with those previously described for fish (Grech et al., 2019) and are available for download on EFSA’s knowledge junction together with the codes for the generic PBK models described in the associated manuscript (Lautz et al., 2020b).

Funding information

This work was supported by the European Food Safety Authority (EFSA) [Contract number: EFSA/SCER/2014/06].

41

CHAPTER 4

GENERIC PHYSIOLOGICALLY-BASED KINETIC MODELLING FOR FARM ANIMALS: PART II. PREDICTING TISSUE CONCENTRATIONS OF CHEMICALS IN SWINE, CATTLE, AND SHEEP

L.S. Lautz, S. Hoeks, R. Oldenkamp, A.J. Hendriks, J.L.C.M. Dorne, A.M.J. Ragas

Published in Toxicology Letters, 318:50-56

42

Abstract

The development of three generic multi-compartment physiologically based kinetic (PBK) models is described for farm animal species, i.e. cattle, sheep, and swine. The PBK models allow one to quantitatively link external dose and internal dose for risk assessment of chemicals relevant to food and feed safety. Model performance is illustrated by predicting tissue concentrations of melamine and oxytetracycline and validated through comparison with measured data. Overall, model predictions were reliable with 71% of predictions within a 3-fold of the measured data for all three species and only 6% of predictions were outside a 10-fold of the measured data. Predictions within a 3-fold change were best for cattle, followed by sheep, and swine (82%, 76%, and 63%). Global sensitivity analysis was performed to identify the most sensitive parameters in the PBK model. The sensitivity analysis showed that body weight and cardiac output were the most sensitive parameters. Since interspecies differences in metabolism impact on the fate of a wide range of chemicals, a key step forward is the introduction of species-specific information on transporters and metabolism including expression and activities.

43

1. INTRODUCTION

Animal health risk assessment of chemicals aims to protect a range of farm and companion animals from the harmful effects of chemicals present in the feed chain (Alexander et al., 2012; Toutain et al., 2010). It accounts also for impacts on human health due to transfer of chemicals into the food-chain (Alexander et al., 2012). For a given chemical, inter-species differences in toxicokinetics (TK) and toxicodynamics (TD) can be investigated through the quantification of inter-species variability in physiology, diet, absorption, metabolism, distribution and excretion (ADME), target receptors and toxicological sensitivity in different life-stages (Dorne and Fink-Gremmels, 2013).

Recently, the European Food Safety Authority (EFSA) recommended to further support quantitative risk assessment through a better understanding of such inter-species differences in TK and TD processes and the development of generic biologically-based models (EFSA, 2014b). Such biologically-based models include tools for allometric scaling, physiologically based kinetic (PBK) and physiologically based TK-TD (PBTK-TD) models (Huang et al., 2015b; Riviere et al., 2016). Generic PBK models integrate physiological and anatomical features subdivided into body compartments (i.e. organ volumes), connected through blood flow, chemical specific parameters (e.g. physico-chemical properties) and the compound’s biochemistry (e.g. Vmax and Km, metabolic clearance) (Brochot et al., 2007; Reddy et al., 2005). Model parameters correspond to physiological and biochemical entities specific to the body and compounds, such as affinities of the compounds for the tissues, or metabolic clearances. PBK models can be refined as chemical-specific, class-specific or generic for a given species or range of species and can be applied to a wide range of data poor compounds bearing in mind the underlying uncertainties associated with data gaps and poor characterisation of kinetics (Beaudouin et al., 2010; Clewell et al., 2004; Cohen Hubal et al., 2019; Edginton et al., 2006).

In the area of animal health, the use of kinetic information and PBK models is currently mostly limited to two main applications. The first one is on therapeutical dosing of veterinary drugs in a given species, the second one is on determining residue levels and transfer of regulated compounds (feed additives, pesticides, veterinary drugs) or contaminants (persistent organic pollutants, toxins) in animal products (e.g. meat, milk, eggs). Such carry over and residue levels can then be used as occurrence inputs and combined with human consumption patterns of animal products for human exposure assessment (Dorne and Fink-Gremmels, 2013; Huang et al., 2015b; Riviere et al., 2016). A recent review assessed the available models in farm animal species and proposed the future development of generic PBK models in risk

44

assessment (Lautz et al., 2019). In addition, generic models have been recently published for four fish species (fathead minnow, zebrafish, European stickleback, rainbow trout) (Grech et al., 2019).

This manuscript describes the development and application of generic multi-compartment models in three farm animal species, namely swine, cattle, and sheep. Anatomical and physiological parameters were collected via meta-analysis and are presented in a related manuscript (Lautz et al., 2020a). Model performance is illustrated in two case studies, i.e. one on melamine and one on oxytetracycline and a global sensitivity analysis was performed to identify the most sensitive model parameters.

2. MATERIALS AND METHODS

2.1. PBK model structure

The PBK model structure is generic for all three farm animal species and is presented in Figure 1. Each model is structured with eleven compartments: arterial blood, venous blood, gastrointestinal tract (GIT), liver, heart, brain, adipose tissue, muscle, bone, lung, and kidney. Sheep and cattle have a twelfth compartment i.e. the milk compartment. All organs and tissues were modelled as homogenous compartments with a blood-flow limited distribution.

The model covers two exposure routes, i.e. oral exposure and intravenous injection. The GIT consists of two sub-compartments, the gut lumen and the gut tissue, enabling the inclusion of biliary excretion as a separate process. Absorption from the gut lumen into the gut tissue is modelled as a first order process and distribution is modelled throughout the body via the systemic circulation. Chemical elimination is included through different routes, i.e. renal excretion, hepatic metabolism, exhalation, egestion and accumulation in milk. The differential equations describing the mass balance in each compartment and parameters abbreviations are provided in the supporting information (SI, part 1).

45

Figure 1. Schematic description of the PBK model developed for cattle, sheep, and swine. Uptake and excretion sites are presented in the purple and green, respectively.

2.2. Anatomical and physiological parameter values

An extensive literature search was performed to identify experimental data on anatomical and physiological parameters (e.g. organ volumes, relative blood flow). Meta-analyses were performed to define distributions reflecting intra-species variability in these parameters (Lautz et al., 2020a). Data availability for blood flow parameters and their associated variability for the three species was limited. Data gaps were filled using allometric scaling (Lindstedt and Schaeffer, 2002) and a default variability of 30% was allocated as suggested by Clewell and Clewell (2008). Tissue composition parameters were expressed as fractions of neutral lipids, polar lipids, proteins and water. These fractions were reported for humans and assumed applicable to swine, cattle, and sheep (Schmitt, 2008).

2.3. Case studies

The PBK models were validated by comparing model predictions of tissue concentrations with experimental data measured in tissues of cattle, swine, and sheep for two compounds that are eliminated through renal excretion: melamine as an environmental contaminant and oxytetracycline as an antibiotic. An extensive search of the published literature in PubMed and Scopus using search terms 'species',

46

'substance' and 'tissue' yielded experimental data for these two compounds (see SI, part 3.1). Papers were included when the experiment was conducted in healthy animals, a clear descriptions of the exposure was available, and measured tissue concentration was reported. Papers not fulfilling these criteria were excluded. The tissue:blood partition coefficients as well as the blood:air partition coefficient were calculated by available QSARs (Hendriks et al., 2005). The QSAR allowed the calculation of the chemical affinity for all tissues based on the octanol/water partition coefficient and tissue composition through considering the tissues constituents’ lipids (both neutral and polar), proteins and water. For both compounds, characteristics, kinetic data and data for model evaluation are presented in Table 1. Model input parameters were predicted by allometric scaling or were based on values from peer-reviewed publications that were independent of the values used for model evaluation (see SI, part 3.2).

Table 1. Compound-specific parameters and data used for PBK model evaluation Parameter values and references Parameter Unit Melamine Oxytetracycline Class - Contaminant Drug Molecular weight g/mol 126.12 460.44 Log Kow - -1.37 -0.9 mg/L 3230 313 Vapour pressure Pa 4.37E-08 1.29E-22 b Absorption rate -1 a 0.0058 min 0.0059 c constant (kabs) 0.032 Clearance (renal) L/min/kg 0.001d 0.002d (Ahmad et al., 1990; Black and Gentry, 1984; Immelman and Dreyer, 1986; Kumar and Malik, (Battaglia et al., 2010; 1998; Kumar and Malik, 2003; Mevius et al., 1986; References for model Cruywagen et al., 2011; Sun et Moreno et al., 1998; Nielsen and Gyrd-Hansen, evaluation al., 2011; Tkachenko et al., 1996; Nouws et al., 1985; Phibro Animal Health, 2015) 2004; Pijpers et al., 1990; Pijpers et al., 1991; Sun et al., 2002; Toutain and Raynaud, 1983; Yar et al., 2000) a Poapolathep et al. (2015); b absorption rate constant for cattle and sheep (Schifferli et al., 1982); c absorption rate constant for swine (Pijpers et al., 1991), d based on allometric scaling and dependent on the body weight of each species, equations are given in the SI (SI part 3.2).

2.4. Global sensitivity analysis

Global sensitivity analysis of the model was performed for each species using the variance-based Sobol method, based on variance decomposition quantifying the relative contribution of each parameter to the total variance of the model output (Saltelli et al., 2008; Sobol’ et al., 2007). The global sensitivity analysis orders the model inputs by importance so that the main contributors to the variation in the output (e.g. blood concentration) can be identified. Sensitivity was assessed for oxytetracycline concentrations in the whole animal, blood, and kidney at three different time points, i.e. absorption phase, fast elimination phase, and slow elimination phase. Parameter values and exposure scenarios are described in detail in SI

47

(part 2.1). Uniform distributions were used for all physiological parameters to compare model performance between species.

2.5. Software

The PBK models were implemented in the R software (version 3.3.3) (R Core Development Team, 2014). The function “soboljansen” in the “sensitivity” package was used to carry out the sensitivity analysis (Pujol et al., 2017). The R codes for the PBK models in the three farm animal species are presented in the SI (part 4) and are also available on the EFSA knowledge junction under the DOI [https://doi.org/ 10.5281/zenodo.3432796].

3. RESULTS

3.1. Model evaluation and validation

For melamine and oxytetracycline, 19 publications were identified with experimental data and Figures 2 and 3 illustrate the comparisons between their measured concentrations in different tissues and organs and the PBK model predictions. Tables S5, S6, and S7 summarise the fold-changes (FC) between the measured data and PBK model predictions for melamine and oxytetracycline (SI part 3.3). Globally, 71% of the model predictions were within a 3-FC of the measured data for all species and only 6% of the PBK predictions were outside a 10-FC.

On a compound-specific basis, 40% and 79% of the predictions of the PBK models were within a 3-FC for melamine and oxytetracycline respectively and only 5-6% of the predictions were over- or underpredicted by more than a 10-fold. The predicted curves for blood concentrations were very close to the measured blood concentration-time data (Figure 4). For other organs, single time point measurements were mostly reported from the literature. And at the tissue or organ level, 58-84% of the predictions in blood, kidney, and muscle were within 3-FC, whereas 67% of the predicted concentrations in liver were between 3-FC to 10-FC.

From an inter-species perspective, the model predictions showed relatively low variability with 82%, 76%, and 63% of predictions within a 3-FC for cattle, sheep, and swine respectively. Model predictions exceeded 10-FC in only 3% of cases in sheep and cattle and 9% in swine.

48

Figure 2. Comparison between concentrations measured in various organs of cattle (CA), sheep (SH), and swine (SW) and PBK model predictions for melamine. Dotted lines represent the 3-fold and 10-fold changes. Organs, species, and references of experimental datasets are indicated in legend: colours and shapes represent organs and studies, respectively.

Figure 3. Comparison between concentrations measured in various organs of cattle (CA), sheep (SH), and swine (SW) and PBK model predictions for oxytetracycline. Dotted lines represent the 3-fold and 10-fold changes. Organs, species, and references of experimental datasets (full references given in Table 1) are indicated in legend: colours and shapes represent organs and studies, respectively.

49

Figure 4. Comparison of model prediction (solid line and 95th confidence interval) and observed data (red dots)(Sun et al., 2002) are shown for concentrations of oxytetracycline in blood from sheep.

3.2. Global sensitivity analysis

Global sensitivity analysis of the PBK models for cattle, sheep, and swine with oxytetracycline shows that body weight (BW), cardiac output (CO) and renal blood flow (fCO_kidney) were the main contributors to the overall variance of the model output (Figure 5). However, the relative contribution of each of those varied between cattle, sheep and swine. During the absorption phase, the model output for the blood concentration was impacted by the intestinal blood flow (fCO_intestine) as well as distribution of the chemical towards other organs such as muscle and adipose tissue (fBW_adipose, fCO_muscle). In the elimination phase, renal blood flow was the most sensitive parameter and had a strong influence on model outputs. The results of the global sensitivity analysis for model outputs with regards to whole animal and kidney concentrations were similar to that for blood concentrations (see SI part 2.2).

50

Figure 5. Sensitivity analysis of the cattle (upper panels), sheep (middle panels) and swine (lower panels) PBK model applied to oxytetracycline. Sobol’s sensitivity indices were estimated for the blood concentration in the three species at three time points (from left to right): 1.25, 9.5, and 24 hours for cattle, 0.5, 6.5, and 21 hours for sheep; 0.5, 5, and 21 hours for swine. Estimation of the Sobol' total sensitivity indices (TI) are presented in dark green and estimation of the Sobol' first-order indices (FOI) in light green. The nine most influencing parameters according to the total sensitivity indices are shown.

4. DISCUSSION

Generic PBK models have been developed for three farm animal species namely, swine, cattle and sheep. The performance of the models was illustrated and validated for melamine and oxytetracycline eliminated via renal excretion while comparing measured data and predicted outputs in these species. Overall, the generic models provided reliable predictions within 3-fold of the measured data.

51

For melamine, a data gap for melamine was identified with regards to absorption rate in the digestive tract of ruminants (Cruywagen et al., 2009; Cruywagen et al., 2011). However, differences between exposure and excreted concentrations in ruminants suggest that the absorption of melamine is higher than 75%. In monogastric animals, such as swine, the absorption of melamine is nearly 100% and unchanged melamine is detected in urine only (Cruywagen et al., 2011; Dorne et al., 2013a). Since absorption rates for melamine in the included species were not available in literature, melamine absorption rates were extrapolated from chicken, leading to uncertainty in the PBK model inputs for this parameter with still reliable predictions in cattle and sheep (Poapolathep et al., 2015). Overall, literature data on melamine in various tissues of the included species were very limited, so the quality of the included papers is of high relevance for the reliability of the model performance. Milk concentrations were overpredicted by the model in most cases when only about 2% of ingested melamine has been reported to be excreted in milk (Cruywagen et al., 2009). Nevertheless, melamine patterns in milk were dose dependent and variation was based on the milk yield which differed between studies (Battaglia et al., 2010; Cruywagen et al., 2009; Sun et al., 2011). This discrepancy might be due to the modelled blood flow for the mammary gland, which is higher compared to the renal blood flow. In swine, the organ concentrations in muscle and liver were overpredicted and showed large variability at high exposure concentration (1000 mg/kg). The rationale for such overpredictions are threefold. First, measured melamine tissue concentrations in swine showed large variability within the animals tested as well as between the measured concentrations of laboratories, leading to discrepancies between model prediction and observed data (Tkachenko et al., 2015). Second, experimental concentrations in swine were very high compared with those for sheep and cattle (Cruywagen et al., 2011; Sun et al., 2011; Tkachenko et al., 2015). However, the calculated clearance using allometric scaling was similar to measured clearance at high dose and steady-state conditions (0.072 vs 0.079 L/h/kg) (Wang et al., 2010) even though renal clearance may be altered at high concentrations (Juhlin et al., 2008). Third, due to limited data availability of experimental data for tissue: blood partition coefficients, tissue partitioning was estimated with a QSAR for polar chemicals (Hendriks et al., 2005) which is based on tissue constituents (e.g. polar lipids, non-polar lipids, water, protein) and octanol/water partition coefficient. This limitation however, did not affect the prediction of organ concentrations for sheep and cattle which were reliable.

For oxytetracycline, a veterinary antibiotic administered orally and intravenously, absorption is only partial in the swine’ intestine (Pijpers et al., 1991) and was not characterised in adult ruminants, i.e. cattle

52

and sheep, and may vary compared to monogastric species. For other substances, such differences in absorption between monogastric and ruminants already have been observed (Ratz et al., 1995). Oxytetracycline undergoes no metabolism and is excreted in urine unchanged (Nouws et al., 1985). Overall, measured blood concentrations were often available in literature, whereas tissue concentrations were scarce for more than one time point. For oxytetracycline, model predictions were reliable compared to observed data. However, Nielsen and Gyrd-Hansen (1996) reported very low blood concentrations after oral exposure, leading to overestimation of the model predictions, compared to similar studies in swine (Pijpers et al., 1990; Pijpers et al., 1991). Differences in the measured blood concentrations between the studies cannot be explained by small changes in the administered dose. Furthermore, kidney concentration of oxytetracycline in swine were underestimated by the model. These differences might by due to inclusion of the urinary formation in the kidney in the measured data, while the model predicts kidney tissue only (Black and Gentry, 1984). Another explanation may be the underestimation of tissue partitioning of oxytetracycline in the kidney. In the literature, a cattle and a sheep PBK model for oxytetracycline were developed to predict concentrations in different organs, after intramuscular administration and fitted model parameters based on experimental data (Achenbach, 2000; Craigmill, 2003). Based on this approach, the predicted concentration were within a 3-FC in 95% of cases (Achenbach, 2000; Craigmill, 2003). With regards to concentrations, our PBK model provided predictions with slightly lower accurate predicted values for cattle (82% predictions within a 3-FC) while for sheep, 76% of the predictions were within a 3-FC.

5. CONCLUSION AND RECOMMENDATIONS

Generic PBK models with species-specific physiological parameters were developed for three farm animal species. The models share the same structure for all three species with the exception of the milk compartment, which has been added for cattle and sheep as an extra elimination route of high relevance to food and feed safety. The model performance was illustrated for melamine and oxytetracycline which are renally excreted. In this context, further model validation is needed through PBK modelling of residues in edible tissues (e.g. muscle, kidney, liver) and animal-based products (i.e. milk) from cattle, swine and sheep for a range of compounds However, a range of data gaps and recommendations for future work can be highlighted for these three species:

1. Peer-reviewed literature providing tissue residues and milk residues for a range of regulated compounds (pesticides, feed additives) and anthropogenic or naturally-occurring contaminants (e.g. persistent organic pollutants, mycotoxins) are still relatively scarce. Extensive literature searches and data 53

collection from pre-market dossiers should be performed to develop kinetic databases to further explore the predictability of the models for a larger group of compounds.

2. Current generic model do not incorporate developmental aspects. Further anatomical and physiological parameters should be collected to test the impact of growth and age on the PBK prediction and sensitivity.

3. Chemicals may undergo transport and phase I and/or Phase II metabolism and databases on species- specific information on transporters, enzyme expression and activities are currently not available. Enzyme expression and activities for these species may therefore be first collected from the literature or generated experimentally, when data gaps have been identified. Experimentally, a range of approaches can provide important datasets with regards to expression levels and activities including 1. OMICs technologies (i.e. transcriptomics, proteomics, metabolomics) providing qualitative and quantitative information on expression of enzymes and transporters. 1. Enzyme activities in liver microsomes and other organs (kidney, gut etc.) providing kinetic parameters for specific probe substrates (phase I, phase metabolism and transporters), regulated compounds and contaminants. In addition, integration of the in vitro metabolism, kinetic data and generic enzymes activities can also provide the basis to further develop quantitative in vitro to in vivo extrapolation models, which can be implemented in PBK models for animal risk assessment and ultimately limit in vivo testing in farm animal species.

Finally, these data gaps and recommendations can be broaden to a wider range of mammalian and avian animal species of public health and animal health importance. It is foreseen that these research efforts will lead to open source databases and PBK tools for multiple species and developmental stages and allow the implementation of a ‘One health’ approach in chemical risk assessment while integrating species differences in chemical kinetics and toxicity.

Acknowledgments

This work was supported by the European Food Safety Authority (EFSA) [Contract number: EFSA/SCER/2014/06].

Appendix A. Supplementary data Supplementary data to this article can be found online. The R code is also included in the Appendix A of this thesis.

54

CHAPTER 5

AN OPEN SOURCE PHYSIOLOGICALLY-BASED KINETIC MODEL FOR THE CHICKEN (GALLUS GALLUS DOMESTICUS): CALIBRATION AND VALIDATION FOR THE PREDICTION OF RESIDUES IN TISSUES AND EGGS

L.S. Lautz, C. Nebbia, S. Hoeks, R. Oldenkamp, A.J. Hendriks, A.M.J. Ragas, J.L.C.M. Dorne

Environmental International [136]: 105488

55

Abstract

Xenobiotics from anthropogenic and natural origin enter animal feed and human food as regulated compounds, environmental contaminants or as part of components of the diet. After dietary exposure, a chemical is absorbed and distributed systematically to a range of organs and tissues, metabolised, and excreted. Physiologically based kinetic (PBK) models have been developed to estimate internal concentrations from external doses. In this study, a generic multi-compartment PBK model was developed for chicken. The PBK model was implemented for seven compounds (with log Kow range -1.37-6.2) to quantitatively link external dose and internal dose for risk assessment of chemicals. Global sensitivity analysis was performed for a hydrophilic and a lipophilic compound to identify the most sensitive parameters in the PBK model. Model predictions were compared to measured data according to dataset- specific exposure scenarios. Globally, 71% of the model predictions were within a 3-fold change of the measured data for chicken and only 7% of the PBK predictions were outside a 10-fold change. While most model input parameters still rely on in vivo experiments, in vitro data were also used as model input to predict internal concentration of the coccidiostat monensin. Future developments of generic PBK models in chicken and other species of relevance to animal health risk assessment are discussed.

56

1. INTRODUCTION

Xenobiotics from anthropogenic and natural origins enter animal feed and human food as regulated compounds, dietary components or environmental contaminants. Risk assessment associated with exposure to such xenobiotics in food and feed is a priority of the European Union (Silano and Silano, 2017) because of the impact they may have on (1) the feed chain potentially affecting productivity and health of farm and companion animals, and (2) the food chain potentially affecting human health via transfer of residues in farm animal products (i.e. meat, milk, eggs) (EFSA, 2018b; Verstraete, 2013). In the European Union and worldwide, regulated chemicals of high relevance to the food and feed safety area include substances added to raw commodities as feed additives, pesticides/biocides or veterinary medicinal products (e.g. antibiotics, coccidiostats, histomonostats) as well as nutrients including amino acids or oligo-elements (e.g. zinc, copper). Environmental contaminants as undesirable substances include anthropogenic substances such as persistent organic pollutants (i.e., dioxins, polychlorinated biphenyls, brominated flame retardants, perfluoroalkyls, phthalates) and natural toxins such as mycotoxins, plant alkaloids and marine biotoxins to cite but a few (Dorne and Fink-Gremmels, 2013).

Chemicals present in the diet are absorbed (A), distributed (D) over a range of organs and tissues, metabolised (M) to a range of more polar metabolites or bioactivated to a toxic moiety, and finally excreted (E) (ADME). Understanding such ADME properties in a quantitative fashion can provide means to determine internal concentrations, predict target organ concentrations and adverse effects in farm animals and humans. Physiologically based kinetic (PBK) models integrate anatomical and physiological characteristics of an organism into algorithms and provide practical means to quantify internal dose metrics and ultimately refine human and animal health risk assessment in food and feed (Andersen et al., 2006; Bois and Brochot, 2016; Cortright et al., 2009). Moving towards internal dose metrics using PBK models has major two major advantages over the use of external dose metrics 1. quantify target organ concentrations in humans and animals on a species-specific basis to set safe levels of exposure for chemicals or therapeutic doses for drugs, 2. determine residue levels and transfer of chemicals in animal products (e.g. meat, milk, eggs) for animal health or as occurrence inputs combined with human consumption patterns for human exposure assessment (Lautz et al., 2019). Human PBK models for specific compounds including pharmaceuticals, cosmetics, pesticides and contaminants as well as generic models are readily available. In addition, a range of New Approach Methods (NAMS) have been developed to reduce animal testing and these include quantitative in vitro in vivo extrapolation (QIVIVE) models which

57

can be incorporate PBK models as well as other in silico tools such as quantitative structure activity (QSAR) relationship models (Paini et al., 2019).

For farm animals, availability of PBK models is much more limited compared to human PBK models and is mostly restricted to specific compounds. A recent literature search and review identified 39 available PBK models for farm animals which were mostly focused on veterinary drugs (Lautz et al., 2019). In addition, the review critically assessed whether the model evaluation was performed using the WHO criteria namely purpose, structure, mathematical representation, computer implementation, parameter estimation and analysis, model performance and documentation. For most compound-specific PBK models available in farm animals, model evaluation was found to be rarely performed using the WHO criteria (WHO, 2010). Development and application of PBK models for farm animals would furthermore benefit from publication of open source model codes and databases, use of global sensitivity analysis and data collection on enzyme expression and activities to support the development of species-specific QIVIVE models (Lautz et al., 2019). Recently, PBK models have been developed as generic open source tools in the R free software environment for four fish species (zebrafish, trout, fathead minnow, and European stickleback) and three farm animal species (cattle, swine and sheep) and validated for a range of chemicals (Grech et al., 2019; Lautz et al., 2020b).

With regards to avian species, compound-specific models have been developed for a limited number of compounds such as marbofloxacin, danofloxacin, midazolam, and monensin, residues of lipophilic pesticides and T-2 Toxin. All these available compound-specific models mostly focused on body distribution and target tissue concentrations of the parent compound without considering generic models (Cortright et al., 2009; Henri et al., 2017; MacLachlan, 2010; Yang et al., 2015a; Yang et al., 2014b; Zeng et al., 2019). However, developing PBK models for every single chemical separately requires a large amount of information regarding their parametrisation including physiological, anatomical, biochemical entities (e.g. organ volume, cardiac output, biotransformation enzymes, and drug transporter expression) and kinetic parameters. Since such kinetic parameters are often sparse for a large number of chemicals, the development of generic PBK models provides means to predict kinetic properties (persistence, clearance, half-life, etc) of known chemicals and emerging chemical hazards for which limited information and data are available (EFSA, 2014d). The present paper describes the development of an open source PBK model for the chicken (Gallus gallus domesticus) and its implementation in the R freeware environment. The open source chicken PBK model aims to predict kinetic properties of orally administered chemicals with a particular focus on tissue and egg concentrations. The model has been evaluated using

58

all WHO criteria including structure, anatomical and physiological parameters, mathematical representation, results from global sensitivity analysis, calibration and validation of its performance for a range of regulated compounds and contaminants. A summary table is also provided to transparently report the model evaluation against each WHO criteria. Finally, conclusions and recommendations for future work are formulated to refine PBK models in chicken and other species of relevance to animal health risk assessment.

2. MATERIALS AND METHODS

2.1. Generic model structure, mathematical representation and computer implementation

The generic chicken PBK model structure consists of eleven compartments, as previously described by Lautz et al. (2020b) for cattle, swine and sheep, with the addition of a twelfth compartment for eggs. All organs and tissues are modelled as well-mixed compartments with a blood-flow limited distribution considering the gastrointestinal tract (GIT) as two compartments: gut lumen and tissue collecting venous blood from the portal vein into the liver. Absorption from the gut lumen into the gut tissue is modelled as a first order process and distribution is modelled throughout the body by systematic circulation. Elimination processes can be included in the model through implementation of data reporting hepatic metabolism or renal excretion, as well as transfer to eggs. Structure and Mathematical representation of the model are provided in Figure 1 and Table 2 of the result section respectively (3.1). Mathematical equations were the same as described previously (Lautz et al., 2020b), except for the egg compartment. Computer implementation of all differential equations was performed in the R software (version 3.3.3) to provide model codes and syntax (R Core Development Team, 2014). The model implementation is available as an open source model code in the R freeware on EFSA knowledge junction under the DOI [10.5281/zenodo.3603114] with a Creative Commons Attribution 4.0 license.

2.2. Parameter estimation and analysis

Data collection of anatomical and physiological parameters

An extensive literature search was performed in PubMed and Google Scholar to identify experimental data providing quantitative anatomical and physiological parameters for the parameterisation of the generic PBK model (e.g. relative organ volumes, relative blood flows). A list of relevant keywords for the extensive literature search is provided below in Table 1.

59

Table 1. Keywords for the extensive literature search for the data collection of anatomical and physiological parameters in chicken Type Keywords Species OR Anatomical and OR OR OR OR OR OR OR OR OR OR OR OR

Each individual reference was screened for anatomical and physiological data which were then computed in an Excel database. Each physiological parameter was assumed to follow a normal distribution with a given arithmetic mean and standard deviation. Meta-analyses for the parameters were performed as described previously (Lautz et al., 2020a) to estimate summary statistics such as arithmetic means and interspecies variability expressed as the coefficient of variation (CV) in male and female adult chicken. The complete dataset with references is available as structured, open source excel databases with a Creative Commons Attribution 4.0 license on EFSA knowledge junction [DOI: 10.5281/zenodo.3603114]. Data gaps were identified for blood flow parameters in different organs and were filled using allometric scaling allocating a default variability of 30% (Clewell and Clewell, 2008; Lindstedt and Schaeffer, 2002). Tissue composition parameters were expressed as fractions of neutral lipids, polar lipids, proteins and water using available values in humans (Schmitt, 2008) and assuming similarity in chicken.

Case studies

Chemicals were selected from the literature using specific criteria: 1. relevance to food and feed safety, e.g. veterinary drugs, feed additives and environmental contaminants, 2. covering a broad range of physico-chemical properties (solubility, vapour pressure, Log P, Kow, pKa) including polar and lipophilic structures, 3. availability of in vivo kinetic data in chicken (constant of absorption (kabs), hepatic and renal clearances (Clhepatic, Clrenal), concentrations in whole body, blood, individual organs, tissues, and eggs) as well as exposure scenarios (single and multiple oral dosing). For each chemical, physico-chemical parameters such as tissue:blood partition coefficients and blood:air partition coefficients were calculated using an available QSAR (Hendriks et al., 2005). This allowed us to determine chemical affinity for all tissues based on the octanol/water partition coefficient and tissue composition by considering the tissues constituents’ lipids (both neutral and polar), proteins and water (Hendriks et al., 2005). Biochemical parameters were collected for each chemical from the literature when available.

60

2.3. Model evaluation

Global sensitivity analysis

Global sensitivity analysis aims to identify the main parameters of the PBK model that contribute to the variation in the model outcome while ordering the inputs by relative importance and was performed using the variance-based Sobol method. The Sobol method is based on variance decomposition allowing to quantify the relative variance contribution of each parameter to the unconditional variance of the model output expressed as tissue, blood or whole body concentration and can handle nonlinear and non- monotonic functions (Saltelli et al., 2008; Sobol’ et al., 2007). Model output variances were computed using the Monte Carlo method with two independent input sample n × p matrices (the “sample” matrix M1 and the “resample” matrix M2 as individual rows), where n is the sample size and p the number of parameters. M1 and M2 represents a possible parameter combination for the model and the first order sensitivity index (Si) reflects the relative contribution of one parameter to the total model variance. Finally, total Sobol sensitivity indices (STi) reflect the relative contribution of the parameter and its inter- relations with the other individual parameters (up to the pth order). In order to generalise the applicability of the global sensitivity analysis to a broad range of polar and lipophilic chemicals, two sensitivity analyses were run for a polar and a lipophilic compound (melamine, deltamethrin) for three time points: 1) uptake phase (0.75 hour), 2) initial elimination phase (5.5 hours), and 3) delayed elimination phase (19.5 hours) after a single oral dose. For each sensitivity analysis, since CVs differ between physiological parameters, uniform distributions were assigned for each parameter as U (min = 0.9 × median; max = 1.1 × median) and the median value was fixed by the calibration step. The function “soboljansen” in the “sensitivity” package was used to carry out the global sensitivity analysis (Pujol et al., 2017).

Model calibration and validation

For each compound, model calibration was performed using partition coefficients estimated with an available QSAR model (Hendriks et al., 2005). The QSAR model allows the calculation of the chemical affinity for all tissues based on the Kow and tissue composition as given below:

푎,푛푙 푎,푝푙 푎,푝푟 푃퐶푡𝑖푠푠푢푒 = 푓푛푙,푡 ∗ 푏푛푙,푡 ∗ 퐾표푤 + 푓푝푙,푡 ∗ 푏푝푙,푡 ∗ 퐾표푤 + 푓푝푟,푡 ∗ 푏푝푟,푡 ∗ 퐾표푤 + 푓퐻2푂,푡

where PCtissue is the tissue:water partition coefficient, Kow is the octanol-water partition coefficient, fnl,t, fpl,t, fpr,t, and fH2O,t are the fractions of neutral lipids, polar lipids, protein and water in tissue, bnl,t, bpl,t, and bpr,t are the intercepts for neutral lipids, polar lipids, and proteins, anl, apl, and apr are the affinity exponents for neutral lipids, polar lipids and proteins. 61

In vivo kinetic data included constant of absorption (kabs), hepatic and renal clearances, concentrations in whole body, blood, individual organs, tissues, and eggs as well as exposure scenarios (single and multiple oral dosing). One set of experimental data was used to estimate absorption rate constant or elimination rates by visual fitting. This dataset was subsequently excluded from model validation, which was performed on the remaining experimental studies. Model validation of the generic chicken PBK model was assessed through comparison of fold changes (FC) between model predictions of organ, tissue and egg concentrations and experimental data from the literature.

3. RESULTS AND DISCUSSION

3.1. Generic model structure, mathematical representation and computer implementation

The purpose of this generic PBK model for the adult chicken is to predict quantities such as area under the curve (AUC) or residues in edible tissues (muscle, liver, kidney) and eggs after oral acute or chronic exposure. In practice, these predictions can be applied to derive either reference points or points of departure on an internal dose basis or carry over and residues as an input for basis for human exposure assessment. The structure of the generic PBK model is illustrated in Figure 1 and Table 2 provides its mathematical representation for mass balance in each anatomical and physiological compartment with associated abbreviations for physiological and chemical specific parameters.

62

Figure 1. Structure of the generic PBK model for chicken. Uptake, excretion and metabolism sites are illustrated in purple, green, and blue, respectively.

Table 2. Mathematical representation of the differential equations describing mass balance in each anatomical and physiological compartment for the generic PBK model in chicken Tissue compartment Equation

Gut lumen 푑푀푙푢푚푒푛 = 푄 ∗ (퐶 − 퐶 ) 푑푡 푓표표푑 푓표표푑 푓푎푒푐푒푠

Gut tissue 푑푀푔푢푡 퐶푔푢푡 = 푄푔푢푡 ∗ (퐶푎푟푡 − ) + 푘푎 ∗ 푀푙푢푚푒푛 푑푡 푃푔푢푡

Liver 푑푀푙푖푣푒푟 퐶푙푖푣푒푟 푉푚푎푥∗퐶푙푖푣푒푟 퐶푔푢푡 퐶푙푖푣푒푟 = 푄푙𝑖푣푒푟 ∗ (퐶푎푟푡 − ) − ∗ 푉푙𝑖푣푒푟 + 푄푔푢푡 ∗ ( ) − 푄푔푢푡 ∗ ( ) 푑푡 푃푙푖푣푒푟 퐾푚+퐶푙푖푣푒푟 푃푔푢푡 푃푙푖푣푒푟 Metabolism

푑푀 퐶 퐶 퐶 푙푖푣푒푟 = 푄 ∗ (퐶 − 푙푖푣푒푟) − (퐶 ∗ 퐶푙 ) + 푄 ∗ ( 푔푢푡) − 푄 ∗ ( 푙푖푣푒푟) 푑푡 푙𝑖푣푒푟 푎푟푡 푃 푙𝑖푣푒푟 ℎ푒푝푎푡𝑖푐 푔푢푡 푃 푔푢푡 푃 Clearance 푙푖푣푒푟 푔푢푡 푙푖푣푒푟

Heart, Brain, Bone, 푑푀푡𝑖푠푠푢푒 퐶푡 Adipose tissue, Muscle, = 푄 ∗ (퐶 − ) 푑푡 푡 푎푟푡 푃 Lung 푡

Kidney 푑푀푘𝑖푑푛푒푦 퐶푘𝑖푑푛푒푦 = 푄푘𝑖푑푛푒푦 ∗ (퐶푎푟푡 − ) − (퐶푘𝑖푑푛푒푦 ∗ 퐶푙푟푒푛푎푙) 푑푡 푃푘𝑖푑푛푒푦

63

Venous blood 푛표푡 푙𝑖푣푒푟+푔푢푡 푑푀푣푒푛 퐶푡 퐶푙𝑖푣푒푟 = ∑ 푄푡 ∗ ( ) + (푄푔푢푡 + 푄푙𝑖푣푒푟) ∗ ( ) − 푄푡표푡 ∗ 퐶푣푒푛 푑푡 푃푡 푃푙𝑖푣푒푟 푇

Arterial blood 푑푀푎푟푡 푄푡표푡 = 푄푡표푡 ∗ 퐶푣푒푛 ∗ ( ) − 푄푡표푡 ∗ 퐶푎푟푡 푑푡 푄푡표푡 + 푄푒푥ℎ푎푙푒 ∗ 푃푎𝑖푟

Egg 푑푀푟푒푝푟표푑 퐶푟푒푝푟표푑 = 푄푟푒푝푟표푑 ∗ (퐶푎푟푡 − ) − 푄푒푔푔 ∗ 퐶푟푒푝푟표푑 푑푡 푃푒푔푔

Vi: Tissue volume (L); Qi: Tissue blood flow (L/min); Qtot: Cardiac output (L/min); Qegg: Egg production (L/min); Qexhale: Exhalation (L/min); Ci: concentration in tissue (mg/kg); Mi: mass in tissue (mmol); Pi: tissue:blood partition coefficient; ka: absorption rate -1 constant (h );Clhepatic: hepatic clearance (L/min); Clrenal: renal clearance (L/min); Vmax: maximal metabolic velocity (mg/min/L liver); Km: Michaelis-Menten constant (mg/L); Mathematical equations were the same as described previously (Lautz et al., 2020b), except for the egg compartment.

3.2. Parameter estimation and analysis

Data collection of anatomical and physiological parameters

Individual anatomical and physiological data from 102 scientific publications in adult male and female chicken (Gallus gallus domesticus) were collected from the literature through an extensive literature search (see Section 2.2) for a wide range of chicken male and female chicken breed, cross-breed and un- named breeds (50% of the dataset). The chicken breeds included were as follows: Arbor Acres, Anak, Anak 2000, Araucana, Archer Arbor, Avian 43, Brahma, Brown Hyssex, Cobb, Cobb 315, Cornish cross, Creeper, Crossbreed, Hy-line Brown chicken, Japanese Bantam, Naked neck, New Hampshire, Novo Brown, Ovambo, Ross, Ross 308, Ross PM3, Shamo, Silky, Starbro, Venda, White Crested Polish, White Leghorn, White Leghorn hybrid. Meta-analyses on data for individual tissue weights and blood flows were performed and normalised as a percentage of body weight and percentage of cardiac output respectively to characterise means and CVs for each compartment of the PBK model in males, females and mixed chicken population as aggregated values (Table 3).

Overall, physiological parameters did not show significant differences (means and CVs) between male and female chicken, with the exception of variability in intestinal weight which was higher in males (Table 3). Blood flow values could only be described for the whole chicken population, since these literature values were very limited available and mostly reported as combined value. For the aggregated meta-analysis, large intra-species variation between studies were observed for brain weight, spleen weight and brain blood flow (>50%). Blood flows for the carcass, lung and muscle were estimated using allometric scaling. A default value for the CV of 30% was allocated to the population variability (Clewell and Clewell, 2008). Intestinal weight showed large intra-species variation across studies (CV= 57%) which mostly reflected differences between chicken breeds while influencing tissue-to-body ratio. Blood flow parameters were

64

only available from few studies and these were measured with different experimental design, however variability was below 20 and 33% with the exception of intestine (39%) and liver blood flow (43%).

Table 3. Mean relative organ weights, tissue weights and blood flows and associated coefficient of variations in males and female chicken breeds (Gallus gallus domesticus) Mixed chicken population Blood flowsa Female (Ref. 1-44) Male (Ref. 1-44) (Ref. 1-44) (Ref. 45-60) Anatomical BW BW CV BW ρ(BW, CO N CV (%) N N CV N CV Parameters (%) (%) (%) (%) OW) (%)d Adipose tissue 100 12.1 17 100 10.4 19 228 10.7 24 0a 1 1.5 30b Blood 69 7.5 22 59 6.8 17 201 7.1 18 0.47 NA NA NA Brain 59 0.2 49 NA NA NA 207 0.3 39 0a 13 0.4 20 Carcass NA NA NA 27 22.9 NA 45 20.3 36 0a 0c 12.4 30b Heart 561 0.6 21 572 0.7 19 2624 0.6 20 0.48 35 5.5 33 Intestine 12 5.8 16 80 3.1 51 193 3.9 57 0.36 13 17.7 39 Kidney 83 0.7 30 68 0.7 21 182 0.8 33 0a 48 11.4 29 Liver 669 2.6 17 614 2.6 19 2833 2.4 17 0.6 116 6.6 43 Lung 562 0.8 27 538 0.8 21 1167 0.8 27 0.17 0c 3.0 30b Muscle 67 39.8 12 67 40.9 11 152 40.8 13 0a 1 19.8 30b Rep. tissue 25 2.8 4 NA NA NA 25 2.8 4 0a 72 14.3 26 Body weight(kg) and CV (2.11, 10); Cardiac output and CV (0.34, 30); Abbreviations: N: sample size (number of individuals); NA: not applicable; BW (%): Organ or tissue normalised as a percentage of body weight, CO (%): blood flow normalised as percentage of cardiac output, CV : Coefficient of variation; a no correlation between body weight and organ weights or cardiac output and blood flow assumed; b ± 30% variability; c fCO based on allometric scaling, d no correlation assumed between cardiac output and blood flow; rho: ρ(BW,OW); Organ and tissue weights (1-44): (Adil et al., 2010; Agnvall et al., 2017; Alikwe et al., 2014; Becker et al., 1979; Bochno et al., 1999; Bond and Gilbert, 1958; Bowes and Julian, 1988; Chikumba and Chimonyo, 2014; Cortright et al., 2009; Dairo et al., 2010; Diarra et al., 2014; Dominguez-Romero et al., 2016; Fernandez et al., 1994; Frahm and Rehkamper, 1998; Hanif et al., 2008; Harris and Koike, 1977; Hassan et al., 2010; Koike et al., 1983; Kosarachukwu et al., 2010; Landers et al., 2008; Lee et al., 2015; Mabelebele et al., 2015; Manafi et al., 2015; Mavromichalis et al., 2000; Medway and Kare, 1959; Mirsalimi and Julian, 1993; Mirsalimi et al., 1993; Moura et al., 2016; Park and Kim, 2015; Ryu et al., 2016; Shahzad et al., 2012; Sieo et al., 2005; Stoev et al., 2004; Szabo et al., 2014; Tickle et al., 2014; Togun et al., 2006; Venturini et al., 2014; Viscor et al., 1985; Wels et al., 1967; Williams and Rodbard, 1960; Wolfenson et al., 1978; Wolfenson et al., 1981; Yahav et al., 1997; Yokhana et al., 2016); Cardiac output and blood flows (45-60): (Boelkins et al., 1973; Cortright et al., 2009; Merrill et al., 1981; Moynihan and Edwards, 1975; Nightingale, 1976; Sapirstein and Hartman, 1959; Stebel and Wideman, 2008; Sturkie and Abati, 1975; Sturkie and Vogel, 1959; Vogel and Sturkie, 1963; Whittow et al., 1964; Wideman, 1999; Wideman et al., 1998; Wideman and Tackett, 2000; Wolfenson et al., 1978; Wolfenson et al., 1981).

Case studies

Seven compounds (melamine, florfenicol, monensin, salinomycin, fipronil, deltamethrin, and sanguinarine) were identified from the literature as relevant to food and feed safety, covering a range of molecular weights and physico-chemical properties and with available in vivo kinetic data in chicken (kinetic parameters in body fluids, tissues, and eggs). Key features are given in Table 4 including chemical name, classification and use, molecular weight, physico-chemical properties (log Kow, pKa) and structure. Melamine is an environmental contaminant that can be present in animal feed and is eliminated by renal

65

excretion (Dorne et al., 2013a). Deltamethrin is a broad-spectrum pyrethroid insecticide which is mainly oxidised into 3-phenoxybenzoic acid in chickens by CYP enzymes. Such metabolism is consistent with oxidative metabolism in humans (Abass et al., 2012; Akhtar et al., 1994; Huyuk and Eraslan, 2017). Fipronil is a hydrophobic chiral insecticide (phenylpyrazole) commonly used in agriculture which is metabolised by a range of cytochrome P-450 isoforms (CYP) in different species. In chicken, fipronil sulfone is the main detected metabolite mediated via CYP3A and CYP2C oxidation (JMPR, 2001; Wang et al., 2016). Florfenicol is a veterinary medicine (i.e., broad-spectrum antibiotic) and monensin and salinomycin are ionophoric (poly-ethers) coccidiostatics used in poultry (Anadon et al., 2008). Monensin and florfenicol are extensively metabolised by CYP3A in chicken and salinomycin is nearly completely metabolised by individual CYP isoforms which have not been characterised to date in chicken (Dorne et al., 2013b; EFSA FEEDAP Panel, 2017; Nebbia, 2001; Wang et al., 2018). Sanguinarine is an alkaloid applied as a feed additive which is widely distributed in the plant families Papaveraceae, Fumariaceae, and Rutaceae. Sanguinarine metabolism involves oxidation into dihydro-metabolites which are generated by CYP1A oxidation in humans and rats (Deroussent et al., 2010; Hu et al., 2018; Xie et al., 2015).

66

Table 4. Chemical characteristics and classification of the compounds selected for the case studiesa Chemical name MW Log Solubility Vapour Classification pKa Structure (CAS number) (g/mol) Kow (mg/L) Pressure (Pa)

Deltamethrin Pesticide 505.2 6.20 0.002 120E-08 NA (52918-63-5)

Fipronil Pesticide 437.1 4.00 2.1 1.29E-22 NA (120068-37-3)

Florfenicol - Drug 358.2 1320 3.70E-07 9.03b (73231-34-2) 0.12b

Melamine Contaminant 126.1 -1.37 3230 4.37E-08 5.0 (108-78-1)

Monensin Drug 670.9 5.43 0.1 6.93E-21 6.6 (17090-79-8)

Salinomycin Drug 773.0 5.12b 622.3 8.30E-23 6.4b (55721-31-8)

Sanguinarine Botanical 332.3 -0.9c 0.001 1.33E-22 NA (2447-54-3) a Data was retrieved from PubChem unless stated otherwise; b Zhao and Ball (2009); b (EFSA, 2018b); c estimated; NA: not available, MW: molecular weight

3.3. Model evaluation

Global sensitivity analysis

Melamine and deltamethrin were selected for the global sensitivity analysis as polar and lipophilic compounds. Global sensitivity analysis using variance-based Sobol method was performed for each PBK 67

input parameter for concentrations in the whole animal, blood, liver and kidney. Parameter values and exposure scenarios are described in Table 5. Results of the global sensitivity analysis for both compounds are presented in Figure 2 for three time points (0.75, 5.5, and 19.5 hours); parameters are ordered by relative importance. Overall, body weight (BW) and the cardiac output (CO) were parameters of the PBK model that contributed the most to the variation in the model outcome (Figure 2). Since the global sensitivity analysis was run for the oral route of exposure, intestinal blood flow (fCO_intestine) represents an important source of variance, particularly during the absorption phase. For melamine, renal blood flow (fCO_kidney) was the most sensitive parameter in the elimination phase since its elimination is driven by renal clearance. In contrast, for deltamethrin, the neutral fraction of the tissue (nl), the adipose tissues relative volume (fBW_adipose), blood flow to the adipose tissue (fCO_adipose) and the lipid content of the tissues contributed the most to the overall variance and predictions of internal concentrations. These results are consistent with deltamethrin’s lipophilicity (MacLachlan, 2010). Other organ blood flows did not contribute significantly to the overall variance of concentration predictions. Overall, results of the sensitivity analysis for model outputs in the whole animal and kidney concentrations were similar compared to those for melamine blood concentrations, but were more variable for deltamethrin.

Table 5. Exposure scenario characteristics for melamine and deltamethrin analysis. Chemical Reference Characteristics Study design Organ Melamine Bai et al. (2010) Species: Chicken Dose: 8.6 mg/kg Plasma Age: 55 weeks Route: food Sex: Female Multiple doses Weight: NA Duration: 34 days N: 12 Deltamethrin Huyuk and Species: Chicken Dose: 0.75 mg/kg Plasma Eraslan (2017) Age: 5 weeks Route: oral Sex: Male Single dose Weight: 1.75 N: 10

68

Figure 2. Sensitivity analysis of the chicken PBK model applied to melamine (upper panel) and deltamethrin (lower panel). Sobol’ sensitivity analysis indices were estimated for the blood concentrations at three time points: 0.75, 5.5, and 19.5 hours (from left to right). Sobol’s total indices (TI) are presented in dark green and Sobol’s first-order indices (FOI) in light green. Parameters were ordered according to the TI. The eleven most influencing parameters according to the total sensitivity indices are shown. BW: body weight, CO: cardiac output, fCO_tissue: fraction cardiac output of a specific tissue; fBW_tissue: fraction body weight of a specific tissue; tissue constituents neutral lipids (nl), polar lipds (pl), proteins (pr) and water (w).

Model calibration and validation

Model calibration has been conducted on a compound-specific basis depending on data availability. For melamine, pharmacokinetic parameters were estimated for renal clearance using allometric scaling, and the model was calibrated with an independent dataset (Lautz et al., 2020b; Poapolathep et al., 2015). Monensin intrinsic clearance was determined based on in vitro parameters (Vmax (646 pmol/mg/min) and Km (28.6 µM) (Henri et al., 2008)) and these were scaled to total liver weight using liver weight and milligrams of microsomal protein per gram liver (value of 9.31) (Henri et al., 2017). Hepatic clearances of salinomycin, florfenicol, deltamethrin and fipronil were estimated by visual fitting (Henri et al., 2012; Huyuk and Eraslan, 2017; MacLachlan, 2008; Shen et al., 2003). Finally, an absorption rate constant for sanguinarine was not available in literature and was therefore estimated by visual fitting (Xie et al., 2015).

69

Compound-specific kinetic parameters including absorption rate constant (kabs), hepatic clearances

(Clhepatic) and renal clearance (Clrenal) are reported in Table 6.

Table 6. Chemical specific kinetic parameters collected for the case studies Unit Melamine Monensin Salinomycin Fipronil Florfenicol Deltamethrin Sanguinarine Parameter sodium a b c g d e g kabs 1/min 0.0059 0.064 0.061 0.01 0.022 0.064 0.01 h g g g g f Clhepatic L/min/kg 0 0.006 0.006 0.00013 0.003 0.001 0.76 i Clrenal L/min/kg 0.0029 0 0 0 0 0 0 Data were retrieved from: a) Poapolathep et al. (2015); b) Henri et al. (2017); c) Atef et al. (1993); d) Shen et al. (2003); e) Huyuk and Eraslan (2017); f) Hu et al. (2018); g estimated; h in vitro in vivo extrapolation; i allometric scaling

Model validation for each compound was performed using the comparison between in vivo kinetic data reported from peer-reviewed experimental studies (n=13) providing concentrations in blood, tissues including mostly adipose, muscle, liver and kidney and eggs and their PBK prediction counterparts. Availability of residue changes in blood and tissues over time were limited to salinomycin and florfenicol in various tissues (Figure 3), showing model performance for the prediction of blood, adipose tissue, kidney and muscle after single and multiple oral doses. Overall, for salinomycin and florfenicol 63% and 51% of the predictions were within a 3-FC. However, slight overestimations were noted for the prediction of internal concentrations (absorption phase) and concentrations in adipose tissue for salinomycin and blood and muscle concentrations (absorption phase) for florfenicol. Model discrepancies might be due to variability in feeding patterns or overestimation of tissue:blood partition coefficients. However, based on the available Kow for salinomycin sodium, higher concentrations in adipose tissue would be expected.

70

Figure 3. Validation of the generic chicken PBK model for salinomycin (A-D) and florfenicol (E-H). Comparison of model predictions (solid lines) and observed data (red dots) are shown for concentrations of both salinomycin and florfenicol in blood, adipose tissue, kidney, and muscle from chicken exposed to salinomycin via a single oral dose (A-D; 20 mg/kg), and to florfenicol via multiple oral doses (E-H; 30 mg/kg).

Model validation is provided for the seven compounds in Figure 4 and Figure 5 comparing experimental and PBK predicted concentrations in blood, tissues and the egg compartment. Globally, 71% of the model predictions were within a 3-fold change of the measured data for chicken and only 7% of the PBK predictions were outside a 10-fold change. Accuracy of the model predictions between compounds was

71

variable but similar between water soluble and lipophilic chemicals. For water soluble compounds (melamine, florfenicol, and sanguinarine), over 70% of the predictions were within 3-fold of the experimental data with only 9% of predictions over- or underestimating by more than a 10-fold. For lipophilic compounds (fipronil, salinomycin, monensin and deltamethrin), 72% of the predictions were within a 3-fold of the experimental data. Only 5% of concentrations were over- or underestimated by more than a 10-fold factor particularly for florfenicol concentration predictions in the brain and the bone compartments (Afifi and Abo el-Sooud, 1997). Florfenicol has low bioavailability, high clearance rates, and is subject to efflux transport (P-glycoprotein (P-gp) and ABCG2) limiting enteric absorption and blood brain barrier transfer and ultimately influencing internal concentrations (Afifi and Abo el-Sooud, 1997; Liu et al., 2018; Wang et al., 2018). Deviations between predicted concentrations and experimental data for florfenicol may be explained by the fact that the current PBK model in chicken does not include quantitative information to model efflux transporters since such data are scarce in the literature (Nebbia, 2001; Schrickx and Fink-Gremmels, 2008). Most information available on transporters in chicken include expression and probe substrates in liver and intestine of Multiple drug resistance 1-gene (MDR1/ABCB1 encoding P-gp), multidrug resistance-associated protein 2 (MRP2/ABCC2) and Breast Cancer Resistant Protein (ABCG2) encoding for membrane proteins from the ABC superfamily (Antonissen et al., 2017; Guo et al., 2013; Guo et al., 2016; Guo et al., 2014; Liu et al., 2018; Osselaere et al., 2013b). Currently, selective transport can only be taken into account in the model when partition coefficients are measured as they are poorly predicted by QSAR models and including active transport across blood-tissue barriers may improve predictions in a mechanistic way for relevant chemicals.

Overall, literature data on tissues concentrations of the included chemicals were limited, so the quality of the included papers is of high relevance for the reliability of the model performance. Therefore, study design and availability of ADME parameters for model calibration may impact model performance and validation. First, chemical uptake has been modelled in chicken after feed ingestion and intestinal absorption as the major relevant exposure route for environmental contaminants and feed additives. While oral absorption directly impacts on internal concentrations, it is required as a model input variable and was well described in the literature for most of the compounds (Atef et al., 1993; Henri et al., 2017; Huyuk and Eraslan, 2017; Poapolathep et al., 2015; Shen et al., 2003). Exceptions included fipronil for which oral absorption patterns in chicken were not available and sanguinarine for which absorption was very low and absorption rate were not reported due to low plasma concentrations (Hu et al., 2018). For melamine and salinomycin, the study design involved ad libitum feeding patterns through contaminated

72

feed, leading to large variability of oral intake in chicken (Bai et al., 2010; Dong et al., 2010; Henri et al., 2009; Henri et al., 2012). Such study design may introduce variability in time dependent patterns of chemical intake impacting internal concentrations. In contrast, PBK model predictions for salinomycin using data from intra-crop administration as a controlled route for chemical intake resulted in more accurate predictions (Atef et al., 1993; Henri et al., 2012).

Figure 4. Comparison between quantities measured in chicken and PBK model predictions for four chemicals in various organs. Dotted lines represent 3-fold and 10-fold changes. Data in FC<3, 3 10 for the chemicals: melamine (80%, 18%, 2%), florfenicol (51%, 30%, 19%), monensin (75%, 25%, 0%), salinomycin (63%, 28%, 9%). Organs and references of experimental dataset obtained are indicated in legends: colours and shapes represent organs and studies, respectively.

73

Figure 5. Comparison between quantities measured in chicken and PBK model predictions for three chemicals in various organs. Dotted lines represent 3-fold and 10-fold changes (FC). Data in FC<3, 3 10 for the chemicals: fipronil (88%, 12%, 0%), deltamethrin (83%, 17%, 0%), sanguinarine (100%, 0%, 0%). Organs and references of experimental dataset obtained are indicated in legends: colours and shapes represent organs and studies, respectively.

Table 7 provides the overall performance of the model for the prediction of tissues and egg concentrations expressed in fold changes compared with the experimental data. Overall, accuracy of the model predictions across compounds was best for egg concentrations with 97% within a 3-fold of the experimental data, followed by blood and kidney (82% and 76%) and other organs (57%-67%). Eggs form an important storage compartment for lipophilic contaminants and veterinary drug residues. In addition, the 1999 dioxin and the 2017 fipronil crises provide examples that egg transfers of lipophilic compounds are also important to consider for human exposure assessment (EFSA, 2018c; 2019; Goetting et al., 2011; Pajurek et al., 2019). Current limitations in PBK models include partitioning of lipophilic chemicals between blood and tissues which is dependent on blood lipid fraction and in chicken, such fraction varies

74

with diet and physiological state such as laying and may introduce variability in egg residue predictions (Máchal, 2000; Peebles et al., 2004; Pinchasov et al., 1994). In the egg compartment, residue profiles can be different between yolk and albumen as part of the egg formation process. Inclusion of both yolk and albumen compartments in the PBK model would support better predictions of residue profiles in this process of egg formation (Hekman and Schefferlie, 2011).

Table 7. Validation of the generic chicken PBK model: Overall performance for the prediction of organs, tissues, blood and egg concentrations Organ FC < 3 3 < FC < 10 FC > 10 Adipose tissue 57% 33% 10% Blood 82% 18% 0% Heart 67% 33% 0% Kidney 76% 12% 12% Liver 61% 35% 4% Muscle 58% 34% 8% Egg 97% 3% 0%

3.4. Open source chicken PBK model: summary of model evaluation

Evaluation of the generic PBK model in chicken (Gallus gallus domesticus) using the WHO criteria is summarised in Table 8 namely scope and purpose of the model, model structure and mathematical implementation, parameter estimation and analysis, model calibration and validation and model documentation. In the future, this table can provide practical means to increase transparency in model evaluation for non-PBK specialists and increase the confidence in using such generic models for risk assessment and regulatory purposes.

Table 8. Evaluation of the generic PBK chicken model according to WHO criteria Category Characteristic Scope and purpose of the model - Model purpose: generic PBK model - Species: Chicken - Age, life stage(s), sex, exposure window(s): adult, males and females, single and multiple doses - Exposure route(s), and dose metric(s): Oral - Target organs and tissues: whole body, blood, organs and tissues, eggs Model structure and mathematical - Graphical representation of the model available description - 12 compartments including eggs - Steady-state and differential calculations - Mass balance equations given Computer implementation - Model implemented in R - Model codes and syntax available Parameter estimation and analysis - Anatomical and physiological parameter values from the literature or predicted - Physicochemical and biochemical parameter values from literature or predicted

75

Model calibration and validation - Global sensitivity analysis performed - Model calibrated with measured data from 7 compounds - Calibration data adequately reported - Model validation against independent data - Validation data reported (Figures 4 and 5) - Variability analysis of the model predictions: predicted versus experimental data expressed as fold changes (Table 7) Model documentation - Peer-reviewed model - Publicly available model

4. CONCLUSION AND FUTURE RECOMMENDATIONS

This manuscript describes the development of a generic open source PBK model in chicken (Gallus gallus domesticus) through integration of meta-analysed physiological parameters into an R-based algorithm. The generic model can be applied to predict blood, tissue and egg concentrations in chicken for risk assessment and can also provide input data for analysis of carry over and residues in human exposure assessment. Model evaluation has been performed using WHO criteria i.e. model purpose, mathematical representation, computer implementation, model calibration and validation. Global sensitivity analysis has been illustrated for melamine (hydrophilic) and deltamethrin (lipophilic) to identify the major physiological variables contributing to the overall variance of the model outputs. Methods such as e-FAST, Sobol Plots or Lowry plots can be recommended as future systematic tools to determine parameters of the PBK model which have little influence on model outputs so that they can be fixed to improve computational speed (Hsieh et al., 2018; McNally et al., 2011; McNally et al., 2018).

Model calibration and validation was performed for seven compounds of relevance to food and feed safety with predictions of blood, tissue and egg concentrations and these are in good agreement with published data. While most model input parameters still rely on physiological data and in vivo chemical- specific data, future opportunities to predict internal concentrations of chemicals using in vitro kinetic data (Vmax, Km, intrinsic clearance) as model inputs and QIVIVE models are increasingly highlighted in the literature and have been explored here for the predictions of monensin residues in chicken (Lautz et al., 2019). However, such in vitro kinetic data are still anecdotic in the literature for chicken and other avian test species (i.e. turkey, quail etc.). Hence, the developments of in vitro test systems in avian species is recommended to support the generation of such datasets while reducing in vivo testing. In addition, data collection on the relative expression and activity of phase I (e.g. cytochrome P-450 isoforms (CYP) etc.), phase II (UDP-glucuronyltransferase set etc.) and transporters (P-glycoprotein, organic anion transporter proteins (OATPs) etc.) using combination of genomic data, in vivo and in vitro assays will further improve

76

QIVIVE based PBK models (Dorne and Fink-Gremmels, 2013; Fink-Gremmels, 2008; Gusson et al., 2006; Martinez et al., 2018a). In this context, CYP expression and activities in chicken and other bird species are becoming increasingly available and could constitute a starting point for such data collection. The current generic model can be modified, and applied to other avian species of relevance to food and feed safety such as ducks, turkey, goose and quails. This would require data mining and collection from databases and the available literature of physiological parameters, enzyme activities, calibration, validation using available kinetic for relevant chemicals.

Comparative in vitro studies revealed that, despite a very low CYP content, broiler chick liver microsomes are more efficient than those from horses, cattle, pigs, farmed rabbits and rats in CYP-mediated oxidation of aromatic model substrates (e.g. 7-ethoxycoumarin, benzo(a)pyrene, aminopyrine, aniline) as well as the coccidiostatic monensin (Nebbia et al., 2001; Nebbia et al., 2003). More recently, the relative expression of hepatic CYP isoforms (CYP 1A4, 1A5, 1B1, 1C1, 2C23a, 2C23b, CYP2C45, CYP2D49, CYP3A37 and CYP3A80) in chicken has been characterised using available genome databases for Gallus gallus domesticus to identify the most important ones (CYP2C45 > CYP1A5 > CYP2C23a > CYP3A37) while the remaining CYPs were barely detectable. Interestingly, no gender-related differences between male and female chickens were observed with regards to relative expression of hepatic CYP isoforms (Watanabe et al., 2013). CYP intestinal expression and activities in chicken have been detected for CYP1A, CYP2C and CYP3A isoforms and characterised for their involvement in pre-systemic metabolism of chemicals. This is of particular relevance for feed additives, pesticides and contaminants administered by the oral route (Kulcsar et al., 2017; Osselaere et al., 2013a; Osselaere et al., 2013b).

It is foreseen that integration of open sources databases reporting physiological and chemical parameters as well as model codes into open source workflows such as the US-EPA computational dashboard and the EFSA TKPlate will provide increase the confidence in these models for risk assessment and regulatory purposes (Baas et al., 2018; Dorne et al., 2018; Pearce et al., 2017; Williams et al., 2017).

Funding information

This work was supported by the European Food Safety Authority (EFSA) [Contract number: EFSA/SCER/2014/06]. Dr. Dorne works as member of staff at the European Food Safety Authority and no conflicts of interest were identified.

77

CHAPTER 6

GENERAL DISCUSSION AND SYNTHESIS

78

“Begin at the beginning,” the King said, very gravely, “and go on till you come to the end: then stop”

The King (Alice in Wonderland)

1. WHERE DID WE COME FROM?

Current US strategies in toxicity-testing have their basis in the year 1937, when the USA only had a couple of laws preventing the sale of unsafe food or drugs (Gad and Chengelis, 2001). In that specific year, a drug labelled ‘Elixir of Sulfanilamide’ entered the US market leading to at least 73 deaths due to diethylene glycol poisoning, in the period from the recognition of the drug’s toxicity until removal from the market by the US Food and Drug administration. This incident facilitated the introduction of the Federal Food, Drug and Cosmetic Act in the year 1938, which required evidence of drug safety before marketing (Krewski et al., 2010). On a European level, safety science started in 1957 with the creation of the European commission Joint Research Center (JRC). However, the first EU executive agency for product safety, i.e. the European Medicines Agency (EMA), was established in 1995. Since then, various agencies were established and considerable efforts and progress were made by the international scientific community to develop methodologies and laws to protect humans, animal species, and the environment against potential risks associated with the exposure to toxic chemicals (Busquet and Hartung, 2017; Krewski et al., 2010; Weinroth et al., 2018).

With the release of the report ‘Toxicity testing in the 21st century: A vision and Strategy’ by the U.S. National Academy of Sciences in 2007, a major change in the way toxicity testing is conducted was initiated. The document describes the investigation of more mechanistic, animal sparing, techniques for toxicity testing (Krewski et al., 2010). A lot of initiatives and research programmes, such as the ExpoCast, ToXCast, Human Toxome, and EU-ToxRisk projects were initiated, in order to integrate advances of in vitro and in silico toxicology, read-across methods and adverse outcome pathways. Furthermore, these projects put specific emphasis on their acceptance and implementation in regulatory risk assessment (Bouhifd et al., 2015; Cronin et al., 2019; Daneshian et al., 2016; Richard et al., 2016; Wambaugh et al., 2015; Wetmore et al., 2015). At its heart there is the need for more robust, reproducible, translatable, cheaper, rapid and ethically acceptable assessment of chemicals. Bringing safe products to the market without animal testing requires new approaches (Grimstein et al., 2019; Mahony, 2019; Paini et al., 2019; Patlewicz et al., 2018; Punt et al., 2017; Thomas et al., 2019). Next Generation Risk Assessment’ (NGRA), defined as an exposure- led, hypothesis-driven risk assessment approach that integrates in silico, in chemico and in vitro approaches, provides such an opportunity (Dent et al., 2018; Desprez et al., 2018). All these tools generate

79

and integrate data from alternative methods under the 3R (replacement, reduction, refinement) principle and improve the mechanistic basis of ADME parameters, which can ultimately be useful to provide PBK input parameters.

Recent developments in chemical risk assessment focus on the harmonisation of various methodologies in human health, animal health and environment. Although the fields have similar backgrounds, they have largely developed independently. Harmonisation of methods enables an efficient use of available resources and facilitates consistency among the assessments of chemicals and stimulates the ‘one health’ approach encompassing public health, animal health and the protection of the environment. Over the last few years, efforts in the US and Europe have been made to develop and harmonise methodologies and approaches to assess the chemical risks between humans, animal species and environment. An example of specific relevance to this thesis is the identification by EFSA of the need to integrate exposure data with kinetic processes, including the development of new in silico tools, such as PBK models, QSARs, and read- across methods as a key priority (EFSA, 2014b).

The main aim of this thesis was to review available farm animal PBK models, collect physiological data, develop and validate PBK models for the prediction of tissue concentrations and chemical kinetics in farm animals. Existing PBK models and their use in animal health risk assessment of chemicals have been reviewed for important farm animal species (Chapter 2). Based on the identified data gaps and recommendations, physiological data were collected on farm animal species including various breeds (Chapter 3 & 5). In the last step, generic PBK models for farm animals, such as cattle, sheep, swine, and chicken were developed and validated for a range of substances to predict internal concentrations in tissues as well as in eggs and milk, with the purpose to move towards the use of internal dose as a basis for the derivation of safe levels in animal health risk assessment (Chapter 4 & 5).

2. WHERE ARE WE NOW?

“I can’t go back to yesterday because I was a different person then”

Alice (Alice in Wonderland)

In Chapter 2, existing PBK models and their use in animal health risk assessment of chemicals have been reviewed for important farm animal species including chicken, cattle, swine, sheep, goat, horse, deer/reindeer, turkey, duck, and goose. The available PBK models for farm animal species have generally been developed for specific purposes, often focusing on a specific substance in a given species. Since

80

generic PBK models for chemical risk assessment were not available for farm animals, my focus has been first to collect physiological data and develop and validate such models for chemicals in food and feed safety. Traditionally, PBK models have been used for various regulatory applications, such as inter-species and intra-species extrapolation. More recently, to reduce the reliance of chemical risk assessments on whole animal toxicity data, increasing emphasis is being placed on the use of in vitro and in silico approaches to link key molecular mechanisms with adverse outcomes (Madden et al., 2019; Punt et al., 2017). Thus, there is a need to establish consensus on approaches for developing and validating PBK models in situations where no new in vivo kinetic data are generated, and which must therefore rely on historical in vivo data, as well as data generated by in vitro and/or in silico methods (Paini et al., 2019).

One prerequisite for the development of PBK models is the availability of accurate physiological parameters as mentioned in Chapter 2. PBK models require many physiological parameters and not all of them may be available experimentally for a particular animal species, strain or breed. In Chapters 3 and 5 extensive literature searches and meta-analyses were described in order to collect information on physiological parameters for cattle, swine, sheep, and chicken, including breed specific information. The data collection performed, included physiological parameter mean values and coefficients of variation for mature farm animals, which can be used as reference values for the development of generic PBK models and allometric scaling models in cattle, swine, sheep and chicken. In the past, data for physiological parameters compiled for laboratory animals and humans were restricted to mean values for the main laboratory animal breed and average adult human (Brown et al., 1997). Recently, an updated version of the PopGen database on human physiological parameters was released, providing the ability to generate realistic anatomically correct human populations for various age groups, e.g. adults, infants, toddlers, (young) teenagers, as well as other user-specified populations (McNally et al., 2015). Compared to the database by Brown (1997), the databases (Chapter 3 and 5) include breed specific information on physiological parameters as well as parameter variability for population estimations for the ‘adult’ life- stage of the included farm animals. However, to reach the detailed level of the PopGen database, further data collection should focus on additional life-stages to create realistic subpopulations for various farm animals. Furthermore, additional data collection can be performed for other species of farm animals (e.g. chicken, goat, duck, salmon) and vertebrate species of ecological relevance (e.g. quail and other birds, test fish species) in adults and other life-stages (Li et al., 2019a; Li et al., 2019b). In order to facilitate the further use of physiological databases in the field of food and feed safety and risk assessment, the anatomical and physiological databases for the four species (cattle, sheep, swine, chicken) have been

81

harmonised with those previously described for fish (Grech et al., 2019) and are available as open source databases for download on EFSA’s knowledge junction. These literature searches and databases should stimulate further PBK development in this field (Lautz et al., 2019; Lin et al., 2016a). If a parameter is not available for the species of interest, it can often be estimated from another species by using allometric scaling (Boxenbaum, 1980; West et al., 1997). Similarly, interspecies scaling has been proposed as a means of extrapolating tissue concentrations across species (Mahmood and Balian, 1996). Care must be taken when applying these algebraic scaling laws in cases where they are based entirely on weight-based scaling factors, since they do not account for underlying kinetic processes, such as metabolism and active transport (Hall et al., 2012; Huang et al., 2015b).

In the next step, generic PBK models were developed for cattle, swine, and sheep focused on chemicals which were only renally excreted (Chapter 4), while the generic chicken model included metabolism (Chapter 5). These are the first generic PBK models developed for farm animals including NAMs, such as QSARs and QIVIVE, to support regulatory decisions on the use of chemicals. Based on the inclusion of NAMs and historical kinetic and tissue data of animals, these models fit in the new strategy of next generation PBK (Paini et al., 2019). The models were developed following the WHO guidance document for PBK model development and the inclusion of the recommendations made previously (Chapter 2) (Lautz et al., 2019; Lin et al., 2016a; WHO, 2010). Similar models were developed for fish and humans in order to further harmonise risk assessment of chemicals in the various fields (Grech et al., 2019; Pearce et al., 2017). However, in the field of animal health as well as environmental risk assessment, most of the included ADME data was based on historical in vivo data and/or (bio-)monitoring data. In some cases (e.g. melamine, oxytetracycline, monensin case studies), models were also parameterised and validated with in silico and in vitro methods, accounting for data poor situations, and fitting in the new strategy of next generation PBK (Paini et al., 2019). The developed models share the same structure regarding the included tissues (gut tissue, liver, heart, brain, adipose tissue, muscle, bone, kidney, and lung). However, the models include two species-specific elements, such as the milk compartment for sheep and cattle, and an egg compartment in the chicken model, since both are relevant due to potential accumulation and transfer within the food chain. The PBK models include the results of a meta-analysis of the literature to provide new parameterisation of physiological parameters as well as related variability. However, not all parameters will have the same impact on the overall model prediction. The relative importance of the parameters in driving the prediction was assessed by a sensitivity analysis (SA). The many SA methods that exist for the analysis of complex deterministic models can be grouped into two categories: Local sensitivity

82

analysis distinguishes the contribution of each parameter, using the one-at-a-time sampling method (Zhang et al., 2015). Local SA techniques are rapid and simple to implement but can give misleading results when there are large differences in the plausible ranges of model parameters or if there are substantial interactions among multiple parameters (Hsieh et al., 2018; McNally et al., 2011; McNally et al., 2018; Saltelli et al., 2008). In a global sensitivity analysis, parameters are varied simultaneously, allowing for the estimation of total variability resulting from both the relative contribution of all parameters as well as their interactions (Zhang et al., 2015). Since concentration-time profiles of chemicals in biological systems are commonly studied with PBK models, it is important to account for the fact that the sensitivity of model parameters may vary with time, so a global sensitivity analysis is preferred. Global sensitivity analyses were performed using the variance-based Sobol method (Saltelli et al., 2008; Sobol’ et al., 2007). Based on the outcome of the sensitivity analyses of the PBK models (Chapter 4 & 5), some general conclusions can be formulated: 1) Sensitivity of the PBK models for lipophilic as well as polar chemicals show that the model output was strongly impacted by body weight and cardiac output. A similar pattern was observed previously (Grech et al., 2019); 2) For renal excreted chemicals, kidney blood flow is a sensitive parameter, which drives differences in internal concentrations; 3) For lipophilic chemicals, the adipose tissue’s relative volume and corresponding blood flow were sensitive parameters. As the case studies for the sensitivity analysis consisted of an exposure via feed, the intestinal absorption rate had a strong influence on the outputs monitored, which was also observed in the sensitivity analysis of fish PBK models (Grech et al., 2019).

Given the trend to reduce animal testing in chemical risk assessment, and the need to derive safe exposure levels from internal points of departure (based on in silico models or in vitro toxicity tests), the challenge for the PBK modelling community is to base the model parameterisation increasingly or entirely on ADME properties derived from non-animal studies, with limited or no availability of in vivo kinetic data (Leonard and Tan, 2019; Madden et al., 2019; Punt et al., 2017). An overview of in silico and in vitro methods for generating human ADME parameters is available (Bessems et al., 2014), while such methods are scarcer for animal species (Huang et al., 2015b; Lee et al., 2017). Especially in vitro methods are suggested to describe metabolic clearance and give insight in underlying processes. For humans, various approaches have been suggested to attain accurate in vivo prediction of clearance from in vitro data. Understanding the particular in vitro system used and any bias or inherent underprediction is crucial for reliable extrapolation of clearance measurements (Da-Silva et al., 2018; Hallifax and Houston, 2019; Poulin and

83

Haddad, 2013; 2018). However, for animal species other than laboratory animal species, in vitro approaches are not yet established.

The main task is to address the credibility of these new tools for their use in a regulatory context (Madden et al., 2019; Paini et al., 2019; Sachana, 2019). Integration of open source databases reporting physiological and chemical parameters as well as model codes into open source workflows such as the US-EPA computational dashboard and the EFSA TKPlate will increase the confidence in these models for risk assessment and regulatory purposes (Baas et al., 2018; Dorne et al., 2018; Laroche et al., 2018; Pearce et al., 2017; Williams et al., 2017). At the same time, a new OECD guidance document was written to address the credibility of these models for their intended purposes to promote the acceptance of next generation PBK models and use in a regulatory context. The generic PBK models as well as datasets (Chapter 3-5) are included as case studies in the new OECD guidance document. This harmonised guidance will inform model developers, reviewers, and risk assessors on the current state of science regarding: 1. characterisation of PBK models using non-animal ADME data (e.g. model elements and construction, underlying principles and assumptions, model parameters and their estimation); 2. assessment of the model performance (e.g. model verification, sensitivity analysis, uncertainty assessment, applicability domain, limitations); 3. reporting and documentation of model characterisation, validation, and intended applications (Sachana, 2019).

3. WHERE ARE WE GOING?

"If you don't know where you want to go, then it doesn't matter which path you take"

The Cheshire Cat (Alice in Wonderland)

Species-specific PBK models can be applied in a generic way for various chemicals as demonstrated in chapter 4 and 5. However, data gaps and recommendations for future work related to PBK model input parameters and model validation can be highlighted. In general, kinetic data and models can help to refine the hazard metrics and risk metrics into internal doses. Metabolism plays a major role in the elimination of most compounds and the lack of chemical-specific metabolism data together with the lack of data on expression and activities for phase I and phase II enzymes in farm animals makes the quantification of inter-species differences in metabolism a challenge. There is an increase in determining expression profiles and functions of the enzymes to gain more understanding of species differences (Girolami et al., 2016; Vimercati et al., 2019; Visser et al., 2019). Considerable inter-species differences exist in the nature,

84

amount and activity of the enzymes and transporters or excretion pathways, influencing ADME processes (Gergs et al., 2015; Toutain et al., 2010). Next to inter-species differences, also intra-species differences caused by polymorphisms, can significantly impact the effect of a chemical. Although, it is conceivable that various types of polymorphisms exist in animal species, only very few studies have investigating such polymorphisms (Martinez et al., 2018a; Martinez et al., 2018b; Vimercati et al., 2019). OMICs technologies (i.e. transcriptomics, proteomics, metabolomics) providing qualitative and quantitative information on expression of enzymes and transporters in various animal species can help to gain insight into differences in ADME processes (EFSA, 2018a). Overall, in vitro kinetic data are still anecdotic in the literature for various animal species. Hence, the developments of in vitro test systems in farm animal species and others (e.g. companion animals) are recommended to support the generation of such datasets while reducing in vivo testing (EFSA Scientific Committee, 2019). The advantage of in vitro assay systems using primary cells from farm animals lies in the similarity between the cells and the original organ and the results may be seen as essentially comparable to in vivo results. However, cell systems from farm animal tissues are rarely available and are less stable (Ogorevc et al., 2016; Yin et al., 2019). In this context, data on CYP expression and activities, as well as genomic data, in various species is becoming more important and essential to further support the development of species-specific QIVIVE models. All these methods can be integrated in PBK modelling to evaluate extrapolation assumptions (Honda et al., 2019). These should be addressed in the future through data collection from the literature and bioinformatic analysis of genome sequences, proteomic expression and the design of in vitro assays in farm animal hepatocytes and intestinal cells to depict such metabolism. These approaches will allow the refinement of PBK models in farm animals and QSAR models allowing for inter-species differences in metabolism and toxicity, respectively. Regarding model validation, peer-reviewed literature describing tissue residues, milk and egg residues for a range of regulated compounds (pesticides, feed additives) and anthropogenic or naturally-occurring contaminants (e.g. persistent organic pollutants, mycotoxins) are still scarce. Extensive literature searches and data collection from pre-market dossiers should be performed to develop kinetic databases to further explore the predictive capacity of the models for a larger group of compounds. Furthermore, animals can be chronically exposed to chemicals, which may include various life-stages, but the current generic models do not incorporate developmental aspects yet. In the human and environmental area, such PBK models are available (El-Masri et al., 2016; Grech et al., 2019). Further physiological parameters and growths patterns should be collected to test the impact of growth and age on the PBK prediction and sensitivity. Finally, these data gaps and recommendations can be broadened to a wider range of animal species of public health and animal health importance, such as goat, horses, salmon, duck, quail, other bird species, 85

as well as test fish species. It is foreseen that these research efforts will lead to open source databases and PBK tools for multiple species and developmental stages and allow the implementation of a ‘one health’ approach in chemical risk assessment while integrating species differences in chemical kinetics and toxicity and moving further towards next generation risk assessment of chemicals.

86

BIBLIOGRAPHY

Abass, K., Turpeinen, M., Rautio, A., Hakkola, J., Pelkonen, A., 2012. Metabolism of pesticides by human cytochrome P450 enzymes in vitro – a survey. In: Perveen, F. (Ed.), Insecticides – Advances in Integrated Pest Managemen. InTech, Austria. Abbiati, R.A., Cagnardi, P., Ravasio, G., Villa, R., Manca, D., 2017. A physiologically based model for tramadol pharmacokinetics in horses. J Theor Biol, 429: 46-51. Achenbach, T.E., 2000. Physiological and classical pharmacokinetic models of oxytetarcycline in cattle, Simon Fraser University, Burnaby, BC, Canada, 1-129 pp. Adil, S., Banday, T., Bhat, G.A., Mir, M.S., Rehman, M., 2010. Effect of dietary supplementation of organic acids on performance, intestinal histomorphology, and serum biochemistry of broiler chicken. Vet Med Int, 2010: 479485. Afifi, N.A., Abo el-Sooud, K.A., 1997. Tissue concentrations and pharmacokinetics of florfenicol in broiler chickens. Br Poult Sci, 38: 425-8. Agnvall, B., Belteky, J., Jensen, P., 2017. Brain size is reduced by selection for tameness in Red Junglefowl- correlated effects in vital organs. Sci Rep, 7: 3306. Ahmad, M., Nawaz, M., Zia-Ur-Rahman, 1990. Disposition kinetics and dosage of oxytetracycline in sheep and lamb. Veterinarski Arhiv 60: 173-179. Akhtar, M.H., Mahadevan, S., Paquet, A., 1994. Comparative metabolism of deltamethrin and 3- phenoxybenzoic acid in chickens. J Environ Sci Health B, 29: 369-94. Alexander, J., Benford, D., Boobis, A., Eskola., M., Fink-Gremmels, J., Fürst, P., Heppner, C., Schlatter, J., van Leuuwen, R., 2012. Risk assessment of contaminants in food and feed. EFSA Journal, 10: 1004-1015. Alikwe, P., Ohimain, E., Dairo, F., 2014. Performance, carcass quality and organ characteristics of broiler finishers fed rumen epithelial scrappings meal (RESM) as replacement for fish meal American Journal of Life Sciences, 2: 29-34. Anadon, A., Martinez, M.A., Martinez, M., Rios, A., Caballero, V., Ares, I., Martinez-Larranaga, M.R., 2008. Plasma and tissue depletion of florfenicol and florfenicol-amine in chickens. J Agric Food Chem, 56: 11049-56. Andersen, M.E., Lutz, R.W., Liao, K.H., Lutz, W.K., 2006. Dose-incidence modeling: consequences of linking quantal measures of response to depletion of critical tissue targets. Toxicol Sci, 89: 331-7. Andersson-Eklund, L., Marklund, L., Lundstrom, K., Haley, C.S., Andersson, K., Hansson, I., Moller, M., Andersson, L., 1998. Mapping quantitative trait loci for carcass and meat quality traits in a wild boar x Large White intercross. J Anim Sci, 76: 694-700. Antonissen, G., Devreese, M., De Baere, S., Martel, A., Van Immerseel, F., Croubels, S., 2017. Impact of Fusarium mycotoxins on hepatic and intestinal mRNA expression of cytochrome P450 enzymes and drug transporters, and on the pharmacokinetics of oral enrofloxacin in broiler chickens. Food Chem Toxicol, 101: 75-83. Atef, M., Ramadan, A., Youssef, S.A., Abo el-Sooud, K., 1993. Kinetic disposition, systemic bioavailability and tissue distribution of salinomycin in chickens. Res Vet Sci, 54: 179-83. Baas, J., Augustine, S., Marques, G.M., Dorne, J.-L., 2018. Dynamic energy budget models in ecological risk assessment: From principles to applications. Science of The Total Environment, 628-629: 249-260. Bai, X., Bai, F., Zhang, K., Lv, X., Qin, Y., Li, Y., Bai, S., Lin, S., 2010. Tissue deposition and residue depletion in laying hens exposed to melamine-contaminated diets. J Agric Food Chem, 58: 5414- 20.

87

Ballarin, C., Povinelli, M., Granato, A., Panin, M., Corain, L., Peruffo, A., Cozzi, B., 2016. The Brain of the Domestic Bos taurus: Weight, Encephalization and Cerebellar Quotients, and Comparison with Other Domestic and Wild Cetartiodactyla. PLoS One, 11: e0154580. Barnes, R.J., Comline, R.S., Dobson, A., 1983. Changes in the blood flow to the digestive organs of sheep induced by feeding. Q J Exp Physiol, 68: 77-88. Barton, H.A. et al., 2007. Characterizing Uncertainty and Variability in Physiologically Based Pharmacokinetic Models: State of the Science and Needs for Research and Implementation. Toxicological Sciences, 99: 395-402. Battaglia, M., Cruywagen, C.W., Bertuzzi, T., Gallo, A., Moschini, M., Piva, G., Masoero, F., 2010. Transfer of melamine from feed to milk and from milk to cheese and whey in lactating dairy cows fed single oral doses. Journal of Dairy Science, 93: 5338-5347. Beaudouin, R., Micallef, S., Brochot, C., 2010. A stochastic whole-body physiologically based pharmacokinetic model to assess the impact of inter-individual variability on tissue dosimetry over the human lifespan. Regul Toxicol Pharmacol, 57: 103-16. Becker, W., Spencer, J., Mirosh, L., Verstrate, J., 1979. Prediction of Fat and Fat Free Live Weight in Broiler Chickens Using Backskin Fat, Abdominal Fat, and Live Body Weight1. Poultry Science, 58: 835- 842. Bell, A.W., Hilditch, T.E., Horton, P.W., Thompson, G.E., 1976. The distribution of blood flow between individual muscles and non-muscular tissues in the hind limb of the young ox (Bos taurus): values at thermoneurality and during exposuer to cold. J Physiol, 257: 229-43. Bellmann, O., Wegner, J., Rehfeldt, C., Teuscher, F., Schneider, F., Voigt, J., Derno, M., Sauerwein, H., Weingärtner, J., Ender, K., 2004. Beef versus dairy cattle: a comparison of metabolically relevant hormones, enzymes, and metabolites. Livestock Production Science, 89: 41-54. Bennett, J., 1973. Regional body surface area in sheep J. Agric. Sci. Comb., 81: 429-432. Bergman, E.N., Havel, R.J., Wolfe, B.M., Bohmer, T., 1971. Quantitative studies of the metabolism of chylomicron triglycerides and cholesterol by liver and extrahepatic tissues of sheep and dogs. J Clin Invest, 50: 1831-9. Berny, P., Caloni, F., Croubels, S., Sachana, M., Vandenbroucke, V., Davanzo, F., Guitart, R., 2010. Animal poisoning in Europe. Part 2: Companion animals. Vet J, 183: 255-9. Bessems, J.G. et al., 2014. PBTK modelling platforms and parameter estimation tools to enable animal- free risk assessment: Recommendations from a joint EPAA – EURL ECVAM ADME workshop. Regulatory Toxicology and Pharmacology, 68: 119-139. Black, W.D., Gentry, R.D., 1984. The distribution of oxytetracycline in the tissues of Swine following a single oral dose. The Canadian veterinary journal = La revue veterinaire canadienne, 25: 158- 161. Bochno, R., Rymkiewicz, J., Janiszewska, M., 1999. Regression equations for the estimation of the meat and fat content of broiler carcasses. Journal of Animal and Feed Sciences, 8: 73-80. Boelkins, J.N., Mueller, W.J., Hall, K.L., 1973. Cardiac output distribution in the laying hen during shell formation. Comparative Biochemistry and Physiology Part A: Physiology, 46: 735-743. Bois, F.Y., Brochot, C., 2016. Modeling Pharmacokinetics. In: Benfenati, E. (Ed.), In Silico Methods for Predicting Drug Toxicity. Springer New York, New York, NY, pp. 37-62. Bois, F.Y., Jamei, M., Clewell, H.J., 2010. PBPK modelling of inter-individual variability in the pharmacokinetics of environmental chemicals. Toxicology, 278: 256-267. Bond, C., Gilbert, P., 1958. Comparative Study of Blood Volume in Representative Aquatic and Nonaquatic Birds. American Journal of Physiology-Legacy Content, 194: 519-521. Boonsanit, D., Chanpongsang, S., Chaiyabutr, N., 2012. Effects of supplemental recombinant bovine somatotropin and mist-fan cooling on the renal tubular handling of sodium in different stages of lactation in crossbred Holstein cattle. Research in Veterinary Science, 93: 417-426. 88

Bouhifd, M. et al., 2015. The human toxome project. Altex, 32: 112-24. Bowes, V.A., Julian, R.J., 1988. Organ weights of normal broiler chickens and those dying of sudden death syndrome. Can Vet J, 29: 153-6. Boxenbaum, H., 1980. Interspecies variation in liver weight, hepatic blood flow, and antipyrine intrinsic clearance: Extrapolation of data to benzodiazepines and phenytoin. Biopharmaceutics, 8: 165- 176. Brochot, C., Smith, T.J., Bois, F.Y., 2007. Development of a physiologically based toxicokinetic model for butadiene and four major metabolites in humans: Global sensitivity analysis for experimental design issues. Chemico-Biological Interactions, 167: 168-183. Brown, D.J., Swan, A.A., 2014. Genetic analysis of adult body weight and condition scores in Merino sheep. In: 10th World Congress on Genetics Applied to Livestock Production Science. Brown, R.P., Delp, M.D., Lindstedt, S.L., Rhomberg, L.R., Beliles, R.P., 1997. Physiological Parameter Values for Physiologically Based Pharmacokinetic Models. Toxicology and Industrial Health, 13: 407-484. Burrin, D.G., Ferrell, C.L., Britton, R.A., Bauer, M., 1990. Level of nutrition and visceral organ size and metabolic activity in sheep. Br J Nutr, 64: 439-48. Busquet, F., Hartung, T., 2017. The need for strategic development of safety sciences. Altex, 34: 3-21. Buur, J., Baynes, R.E., Craigmill, A.L., Riviere, J.E., 2005. Development of a physiologic-based pharmacokinetic model for estimating sulfamethazine concentrations in swine and application to prediction of violative residues in edible tissues. Am J Vet Res, 66: 1686-93. Buur, J., Baynes, R.E., Riviere, J.E., 2008. Estimating meat withdrawal times in pigs exposed to melamine contaminated feed using a physiologically based pharmacokinetic model. Regulatory Toxicology and Pharmacology: 324-331. Buur, J., Baynes, R.E., Smith, G., Riviere, J.E., 2006. Use of Probabilistic Modeling within a Physiologically Based Pharmacokinetic Model To Predict Sulfamethazine Residue Withdrawal Times in Edible Tissues in Swine. Antimicrobial Agents and Chemotherapy, 50: 2344-2351. Buur, J., Baynes, R.E., Smith, G., Riviere, J.E., 2009. A physiologically based pharmacokinetic model linking plasma protein binding interactions with drug disposition. Research in Veterinary science: 293-301. Carlson, D.B., Reed, J.J., Borowicz, P.P., Taylor, J.B., Reynolds, L.P., Neville, T.L., Redmer, D.A., Vonnahme, K.A., Caton, J.S., 2009. Effects of dietary selenium supply and timing of nutrient restriction during gestation on maternal growth and body composition of pregnant adolescent ewes. J Anim Sci, 87: 669-80. Carretero, M., Fontanals, J., Agustí, M., Arguis, M., Martínez-Ocón, J., Ruiz, A., Rios, J., 2010. Monitoring in resuscitation: Comparison of cardiac output measurement between pulmonary artery catheter and NICO. Resuscitation, 81: 404-409. Charismidou, M., Bizelis, J., Rogdakis, E., 2000. Metabolic changes during the perinatal period in dairy sheep in relation to level of nutrition and breed. I. Late pregnancy. Animal Physiology and Animal Nutrition, 84: 61-72. Chen, K., Seng, K.Y., 2012. Calibration and validation of a physiologically based model for soman intoxication in the rat, marmoset, guinea pig and pig. J Appl Toxicol, 32: 673-86. Chikumba, N., Chimonyo, M., 2014. Effects of water restriction on the growth performance, carcass characteristics and organ weights of naked neck and ovambo chickens of southern Africa. Asian- Australasian journal of animal sciences, 27: 974-980. Chiu, W.A., Barton, H.A., DeWoskin, R.S., Schlosser, P., Thompson, C.M., Sonawane, B., Lipscomb, J.C., Krishnan, K., 2007. Evaluation of physiologically based pharmacokinetic models for use in risk assessment. Journal of Applied Toxicology, 27: 218-237.

89

Clapham, M.O., Martin, K.L., Davis, J.L., Baynes, R.E., Lin, Z., Vickroy, T.W., Riviere, J.E., Tell, L.A., 2019. Extralabel drug use in wildlife and game animals. J Am Vet Med Assoc, 255: 555-568. Clarke, I.J., Tilbrook, A.J., Turner, A.I., Doughton, B.W., Goding, J.W., 2001. Sex, fat and the tilt of the earth: effects of sex and season on the feeding response to centrally administered leptin in sheep. Endocrinology, 142: 2725-8. Clewell, H.J., Andersen, M.E., Blaauboer, B.J., 2008. On the incorporation of chemical-specific information in risk assessment. Toxicology Letters, 180: 100-109. Clewell, H.J., Gentry, P.R., Covington, T.R., Sarangapani, R., Teeguarden, J.G., 2004. Evaluation of the potential impact of age- and gender-specific pharmacokinetic differences on tissue dosimetry. Toxicol Sci, 79: 381-93. Clewell, R.A., Clewell, H.J., 2008. Development and specification of physiologically based pharmacokinetic models for use in risk assessment. Regulatory Toxicology and Pharmacology, 50: 129-143. Cohen Hubal, E.A., Wetmore, B.A., Wambaugh, J.F., El-Masri, H., Sobus, J.R., Bahadori, T., 2019. Advancing internal exposure and physiologically-based toxicokinetic modeling for 21st-century risk assessments. Journal of Exposure Science & Environmental Epidemiology, 29: 11-20. Cortright, K.A., Wetzlich, S.E., Craigmill, A.L., 2009. A PBPK model for midazolam in four avian species. Journal of Veterinary Pharmacology and Therapeutics, 32: 552-565. Craigmill, A.L., 2003. A physiologically based pharmacokinetic model for oxytetracycline residues in sheep. Journal of Veterinary Pharmacology and Therapeutics, 26: 55-63. Crile, G., Quiring, D., 1940. A record of the body weight and certain organ and gland weights of 3690 animals. Ohio Journal of Science, 40: 219-260. Cronin, M.T.D., Madden, J.C., Yang, C., Worth, A.P., 2019. Unlocking the potential of in silico chemical safety assessment - A report on a cross-sector symposium on current opportunities and future challenges. Comput Toxicol, 10: 38-43. Cruywagen, C.W., Stander, M.A., Adonis, M., Calitz, T., 2009. Hot topic: Pathway confirmed for the transmission of melamine from feed to cow's milk. Journal of Dairy Science, 92: 2046-2050. Cruywagen, C.W., van de Vyver, W.F.J., Stander, M.A., 2011. Quantification of melamine absorption, distribution to tissues, and excretion by sheep. Journal of Animal Science, 89: 2164-2169. Da-Silva, F., Boulenc, X., Vermet, H., Compigne, P., Gerbal-Chaloin, S., Daujat-Chavanieu, M., Klieber, S., Poulin, P., 2018. Improving Prediction of Metabolic Clearance Using Quantitative Extrapolation of Results Obtained From Human Hepatic Micropatterned Cocultures Model and by Considering the Impact of Albumin Binding. J Pharm Sci, 107: 1957-1972. Dairo, F., Adesehinwa, A., Oluwasola, T., Oluyemi, J., 2010. High and low dietary energy and protein levels for broiler chickens. African Journal of Agricultural Research, 5: 2030-2038. Daneshian, M., Kamp, H., Hengstler, J., Leist, M., van de Water, B., 2016. Highlight report: Launch of a large integrated European in vitro toxicology project: EU-ToxRisk. Arch Toxicol, 90: 1021-4. Delavaud, C., Bocquier, F., Baumont, R., Chaillou, E., Ban-Tokuda, T., Chilliard, Y., 2007. Body fat content and feeding level interact strongly in the short- and medium-term regulation of plasma leptin during underfeeding and re-feeding in adult sheep. Br J Nutr, 98: 106-15. Dent, M. et al., 2018. Principles underpinning the use of new methodologies in the risk assessment of cosmetic ingredients. Computational Toxicology, 7: 20-26. Derks, H.J.G.M., Berende, P.L.M., Olling, M., Everts, H., Liem, A.K.D., de Jong, A.P.J.M., 1994. Pharmacokinetic modeling of polychlorinated dibenzo-p-dioxins (PCDDs) and furans (PCDFs) in cows. Chemosphere, 28: 711-715. Deroussent, A., Re, M., Hoellinger, H., Cresteil, T., 2010. Metabolism of sanguinarine in human and in rat: characterization of oxidative metabolites produced by human CYP1A1 and CYP1A2 and rat

90

liver microsomes using liquid chromatography-tandem mass spectrometry. J Pharm Biomed Anal, 52: 391-7. Desprez, B. et al., 2018. A strategy for systemic toxicity assessment based on non-animal approaches: The Cosmetics Europe Long Range Science Strategy programme. Toxicology in Vitro, 50: 137- 146. Di Giantomasso, D., May, C.N., Bellomo, R., 2003. Vital Organ Blood Flow During Hyperdynamic Sepsis. Chest, 124: 1053-1059. Diarra, S., Sandakabatu, S., Perera, D., Tabuaciri, P., Mohammed, U., 2014. Growth Performance, Carcass Measurements and Organs Weight of Broiler Chickens Fed Cassava Copra Meal-based or Commercial Finisher Diets in Samoa. Asian Journal of Poultry Science, 8: 16-22. Dodic, M., Samuel, C., Moritz, K., Wintour, E.M., Morgan, J., Grigg, L., Wong, J., 2001. Impaired cardiac functional reserve and left ventricular hypertrophy in adult sheep after prenatal dexamethasone exposure. Circ Res, 89: 623-9. Dominguez-Romero, E., Cariou, R., Omer, E., Marchand, P., Dervilly-Pinel, G., Le Bizec, B., Travel, A., Jondreville, C., 2016. Tissue Distribution and Transfer to Eggs of Ingested alpha- Hexabromocyclododecane (alpha-HBCDD) in Laying Hens (Gallus domesticus). Journal of Agricultural and Food Chemistry, 64: 2112-2119. Dong, X.F., Liu, S.Y., Tong, J.M., Zhang, Q., 2010. Carry-over of melamine from feed to eggs and body tissues of laying hens. Food Addit Contam Part A Chem Anal Control Expo Risk Assess, 27: 1372- 9. Dorne, J.L., Doerge, D.R., Vandenbroeck, M., Fink-Gremmels, J., Mennes, W., Knutsen, H.K., Vernazza, F., Castle, L., Edler, L., Benford, D., 2013a. Recent advances in the risk assessment of melamine and cyanuric acid in animal feed. Toxicol Appl Pharmacol, 270: 218-29. Dorne, J.L.C.M. et al., 2018. Reconnecting exposure, toxicokinetics and toxicity in food safety: OpenFoodTox and TKplate for human health, animal health and ecological risk assessment. Toxicology Letters, 295: S29. Dorne, J.L.C.M., Fernández-Cruz, M.L., Bertelsen, U., Renshaw, D.W., Peltonen, K., Anadon, A., Feil, A., Sanders, P., Wester, P., Fink-Gremmels, J., 2013b. Risk assessment of coccidostatics during feed cross-contamination: Animal and human health aspects. Toxicology and Applied Pharmacology, 270: 196-208. Dorne, J.L.C.M., Fink-Gremmels, J., 2013. Human and animal health risk assessments of chemicals in the food chain: Comparative aspects and future perspectives. Toxicology and Applied Pharmacology, 270: 187-195. Doyle, J.T., Patterson, J.L., Jr., Warren, J.V., Detweiler, D.K., 1960. Observations on the circulation of domestic cattle. Circ Res, 8: 4-15. Ebinger, P., 1974. A cytoarchitectonic volumetric comparison of brains in wild and domestic sheep. Z Anat Entwicklungsgesch, 144: 267-302. Edginton, A.N., Schmitt, W., Willmann, S., 2006. Development and evaluation of a generic physiologically based pharmacokinetic model for children. Clin Pharmacokinet, 45: 1013-34. EFSA, 2014a. Guidance on Expert Knowledge Elicitation in Food and Feed Safety Risk Assessment. EFSA Journal, 12: 3734. EFSA, 2014b. Modern methodologies and tools for human hazard assessment of chemicals. EFSA Journal, 12: 3638-3725. EFSA, 2014c. Scientific opinion on good modelling practice in the context of mechanistic effect models for risk assessment of plant protection products. EFSA Journal, 12: 3589-3681. EFSA, 2014d. A systematic procedure for the identification of emerging chemical risks in the food and feed chain. EFSA Supporting publication: 547-586.

91

EFSA, 2018a. EFSA Scientific Colloquium 24 – 'omics in risk assessment: state of the art and next steps. EFSA Supporting Publications, 15: 1512E. EFSA, 2018b. Report for 2016 on the results from the monitoring of veterinary medicinal product residues and other substances in live animals and animal products. EFSA Supporting publication, 15: 75pp. EFSA, 2018c. Scientific report on the occurrence of residues of fipronil and other acaricides in chicken eggs and poultry muscle/fat. EFSA Journal, 16. EFSA, 2019. Report for 2017 on the results from the monitoring of veterinary medicinal product residues and other substances in live animals and animal products. EFSA Supporting Publications, 16: 1578E. EFSA FEEDAP Panel, 2017. Scientific opinion on the safety and efficacy of Sacox® microGranulate (salinomycin sodium) for chickens for fattening and chickens reared for laying. EFSA Journal, 15: 4670pp. EFSA Scientific Committee, 2019. Guidance on harmonised methodologies for human health, animal health and ecological risk assessment of combined exposure to multiple chemicals. EFSA Journal, 17: e05634. El-Masri, H., Kleinstreuer, N., Hines, R.N., Adams, L., Tal, T., Isaacs, K., Wetmore, B.A., Tan, Y.-M., 2016. Integration of Life-Stage Physiologically Based Pharmacokinetic Models with Adverse Outcome Pathways and Environmental Exposure Models to Screen for Environmental Hazards. Toxicological sciences : an official journal of the Society of Toxicology, 152: 230-243. Ellis, J.L., Reynolds, C.K., Crompton, L.A., Hanigan, M.D., Bannink, A., France, J., Dijkstra, J., 2016. Prediction of portal and hepatic blood flow from intake level data in cattle. J Dairy Sci, 99: 9238- 9253. EMA, 2018. Guideline on the qualification and reporting of physiologically based pharmacokinetic (PBPK) modelling and simulation. EMEA/CVMP, 2003. Position paper regarding availibility of products for minor uses and minor species (MUMS). EMEA/CVMP/477/03/Final. EPA, 2006. Approaches for the application of physiologically based pharmacokinetic (PBPK) models and supporting data in risk assessment., National Center for Environmental Assessment, Washington, DC. Evans, W., Phernetton, T.M., Magness, R.R., 1998. 17beta-estradiol effect on critical cardiac output with reduction of cardiac output in oophorectomized sheep. Am J Physiol, 275: H57-64. FDA, 2018. Physiologically Based Pharmacokinetic Analyses — Format and Content Guidance for Industry”. Fegler, G., Hill, K.J., 1958. Measurement of blood flow and heat production in the splanchnic region of the anaesthetized sheep. Q J Exp Physiol Cogn Med Sci, 43: 189-96. Feldman, H.A., McMahon, T.A., 1983. The 3/4 mass exponent for energy metabolism is not a statistical artifact. Respir Physiol, 52: 149-63. Fernandez, A., Verde, M.T., Gascon, M., Ramos, J., Gomez, J., Luco, D.F., Chavez, G., 1994. Variations of clinical biochemical parameters of laying hens and broiler chickens fed aflatoxin-containing feed. Avian Pathol, 23: 37-47. Fierens, T., Van Holderbeke, M., Standaert, A., Cornelis, C., Brochot, C., Ciffroy, P., Johansson, E., Bierkens, J., 2016. Multimedia & PBPK modelling with MERLIN-Expo versus biomonitoring for assessing Pb exposure of pre-school children in a residential setting. Sci Total Environ, 568: 785- 793. Fink-Gremmels, J., 2008. Implications of hepatic cytochrome P450-related biotransformation processes in veterinary sciences. European Journal of Pharmacology, 585: 502-509.

92

Fisher, E., Dalton, R., 1961. Determination of the cardiac output of cattle and horses by the injection method. Br Vet J, 118: 143-151. Fleet, I.R., Mepham, T.B., 1985. Mammary uptake of amino acids and glucose throughout lactation in Friesland sheep. J Dairy Res, 52: 229-37. Fraga, A.L., Thomaz, M.C., Kronka, R.N., Budiño, F.E.L., Huaynate, R.A.R., Scandolera, A.J., Ruiz, U.S., Nadai, A., 2009. Qualitative feed restriction for heavy swines: effect on digestibility and weight of organs of digestive tract, and environmental impact of feces. Arquivo Brasileiro de Medicina Veterinária e Zootecnia, 61: 1353-1363. Frahm, H.D., Rehkamper, G., 1998. Allometric comparison of the brain and brain structures in the white crested polish chicken with uncrested domestic chicken breeds. Brain Behav Evol, 52: 292-307. Free Software Foundation, 2007. GNU GENERAL PUBLIC LICENSE. Version 3, 29 June 2007. Boston, USA. Freetly, H.C., Ferrell, C.L., 1997. Oxygen consumption by and blood flow across the portal-drained viscera and liver of pregnant ewes1. Journal of Animal Science, 75: 1950-1955. Freijer, J., Eijkeren, J.v., Sips, A., 1999. Model for estimating initial burden and daily absorption of lipophilic contaminants in cattle. 643810005, National Institute of Public Health and the Environment, Bilthoven The Netherlands. Gad, S., Chengelis, C., 2001. Human pharmaceutical products. In: Gad, S. (Ed.), Regulatory toxicology. Taylor & Francis, London, pp. 9-69. García-Valverde, R., Barea, R., Lara, L., Nieto, R., Aguilera, J.F., 2008. The effects of feeding level upon protein and fat deposition in Iberian heavy pigs. Livestock Science, 114: 263-273. Gardner, D.S., Tingey, K., Van Bon, B.W., Ozanne, S.E., Wilson, V., Dandrea, J., Keisler, D.H., Stephenson, T., Symonds, M.E., 2005. Programming of glucose-insulin metabolism in adult sheep after maternal undernutrition. Am J Physiol Regul Integr Comp Physiol, 289: R947-54. Gergs, A., Kulkarni, D., Preuss, T.G., 2015. Body size-dependent toxicokinetics and toxicodynamics could explain intra- and interspecies variability in sensitivity. Environ Pollut, 206: 449-55. Giantin, M., Carletti, M., Capolongo, F., Pegolo, S., Lopparelli, R.M., Gusson, F., Nebbia, C., Cantiello, M., Martin, P., Pineau, T., Dacasto, M., 2008. Effect of breed upon cytochromes P450 and phase II enzyme expression in cattle liver. Drug Metab Dispos, 36: 885-93. Girolami, F., Spalenza, V., Benedetto, A., Manzini, L., Badino, P., Abete, M.C., Nebbia, C., 2016. Comparative liver accumulation of dioxin-like compounds in sheep and cattle: Possible role of AhR-mediated xenobiotic metabolizing enzymes. Science of The Total Environment, 571: 1222- 1229. Glazier, D.S., 2005. Beyond the '3/4-power law': variation in the intra- and interspecific scaling of metabolic rate in animals. Biol Rev Camb Philos Soc, 80: 611-62. Goetting, V., Lee, K.A., Tell, L.A., 2011. Pharmacokinetics of veterinary drugs in laying hens and residues in eggs: a review of the literature. J Vet Pharmacol Ther, 34: 521-56. Grace, N., 1983. Amounts and distribution of mineral elements associated with fleece-free empty body weight gains in the grazing sheep. New Zealand Jounral of Agricultural Reseach, 26: 59-70. Graham, A., Nicol, A., Christopherson, R., 1982. Rumen motility repsonses to adrenaline and noradrenaline and organ weights of warm- and cold-acclimated sheep. Canadian Journal of Animal Science, 62: 777-786. Grech, A., Brochot, C., Dorne, J.-L., Quignot, N., Bois, F.Y., Beaudouin, R., 2017. Toxicokinetic models and related tools in environmental risk assessment of chemicals. Science of The Total Environment, 578: 1-15. Grech, A., Tebby, C., Brochot, C., Bois, F.Y., Bado-Nilles, A., Dorne, J.-L., Quignot, N., Beaudouin, R., 2019. Generic physiologically-based toxicokinetic modelling for fish: Integration of environmental factors and species variability. Science of The Total Environment, 651: 516-531.

93

Grimstein, M., Yang, Y., Zhang, X., Grillo, J., Huang, S.M., Zineh, I., Wang, Y., 2019. Physiologically Based Pharmacokinetic Modeling in Regulatory Science: An Update From the U.S. Food and Drug Administration's Office of Clinical Pharmacology. J Pharm Sci, 108: 21-25. Guitart, R., Croubels, S., Caloni, F., Sachana, M., Davanzo, F., Vandenbroucke, V., Berny, P., 2010. Animal poisoning in Europe. Part 1: Farm livestock and poultry. Vet J, 183: 249-54. Guo, M., Bughio, S., Sun, Y., Zhang, Y., Dong, L., Dai, X., Wang, L., 2013. Age-related P-glycoprotein expression in the intestine and affecting the pharmacokinetics of orally administered enrofloxacin in broilers. PloS one, 8: e74150-e74150. Guo, M., Dai, X., Hu, D., Zhang, Y., Sun, Y., Ren, W., Wang, L., 2016. Potential pharmacokinetic effect of rifampicin on enrofloxacin in broilers: Roles of P-glycoprotein and BCRP induction by rifampicin. Poult Sci, 95: 2129-35. Guo, M., Sun, Y., Zhang, Y., Bughio, S., Dai, X., Ren, W., Wang, L., 2014. E. coli infection modulates the pharmacokinetics of oral enrofloxacin by targeting P-glycoprotein in small intestine and CYP450 3A in liver and kidney of broilers. PLoS One, 9: e87781. Gusson, F., Carletti, M., Albo, A.G., Dacasto, M., Nebbia, C., 2006. Comparison of hydrolytic and conjugative biotransformation pathways in horse, cattle, pig, broiler chick, rabbit and rat liver subcellullar fractions. Vet Res Commun, 30: 271-83. Hales, J.R., 1973. Radioactive microsphere measurement of cardiac output and regional tissue blood flow in the sheep. Pfluegers Archiv, 344: 119-132. Hales, J.R., 1976. Effects of acute cold exposure on the distribution of cardiac output in the sheep. Pflügers Archiv, 366: 153-157. Hales, J.R., Fawcett, A.A., 1993. Wool production and blood supply to skin and other tissues in sheep. J Anim Sci, 71: 492-8. Hall, C., Lueshen, E., Mosat, A., Linninger, A.A., 2012. Interspecies scaling in pharmacokinetics: a novel whole-body physiologically based modeling framework to discover drug biodistribution mechanisms in vivo. J Pharm Sci, 101: 1221-41. Hallifax, D., Houston, J., 2019. Use of Segregated Hepatocyte Scaling Factors and Cross-Species Relationships to Resolve Clearance Dependence in the Prediction of Human Hepatic Clearance. Drug Metab Dispos, 47: 320-327. Hallowell, G.D., Potter, T.J., Bowen, I.M., 2007. Methods and normal values for echocardiography in adult dairy cattle. Journal of Veterinary Cardiology, 9: 91-98. Hanif, N.Q., Muhammad, G., Siddique, M., Khanum, A., Ahmed, T., Gadahai, J.A., Kaukab, G., 2008. Clinico-pathomorphological, serum biochemical and histological studies in broilers fed ochratoxin A and a toxin deactivator (Mycofix Plus). Br Poult Sci, 49: 632-42. Harris, K.M., Koike, T.I., 1977. The effects of dietary sodium restriction on fluid and electrolyte metabolism in the chicken (Gallus domesticus). Comparative Biochemistry and Physiology Part A: Physiology, 58: 311-317. Hartung, T., Hoffmann, S., 2009. Food for thought ... on in silico methods in toxicology. Altex, 26: 155-66. Hassan, Z., Khan, M., Khan, A., Javed, I., 2010. Pathological Responses of White Leghorn Breeder Hens Kept on Ochratoxin A Contaminated Feed. Pak Vet J, 30: 118-123. He, Y., Li, X., Zhang, F., Su, Y., Hou, L., Chen, H., Zhang, Z., Huang, L., 2015. Multi-breed genome-wide association study reveals novel loci associated with the weight of internal organs. Genet Sel Evol, 47: 87. Hekman, P., Schefferlie, G.J., 2011. Kinetic modelling and residue depletion of drugs in eggs. Br Poult Sci, 52: 376-80. Hendriks, A.J., 2007. The power of size: A meta-analysis reveals consistency of allometric regressions. Ecological Modelling, 205: 196-208.

94

Hendriks, A.J., Traas, T.P., Huijbregts, M.A., 2005. Critical body residues linked to octanol-water partitioning, organism composition, and LC50 QSARs: meta-analysis and model. Environ Sci Technol, 39: 3226-36. Henri, J., Burel, C., Sanders, P., Laurentie, M., 2009. Bioavailability, distribution and depletion of monensin in chickens. J Vet Pharmacol Ther, 32: 451-6. Henri, J., Carrez, R., Meda, B., Laurentie, M., Sanders, P., 2017. A physiologically based pharmacokinetic model for chickens exposed to feed supplemented with monensin during their lifetime. J Vet Pharmacol Ther. Henri, J., Manceau, J., Sanders, P., Laurentie, M., 2008. Cytochrome P450-dependent metabolism of monensin in hepatic microsomes from chickens and turkeys. J Vet Pharmacol Ther, 31: 584-6. Henri, J., Maurice, R., Postollec, G., Dubreil-Cheneau, E., Roudaut, B., Laurentie, M., Sanders, P., 2012. Comparison of the oral bioavailability and tissue disposition of monensin and salinomycin in chickens and turkeys. J Vet Pharmacol Ther, 35: 73-81. Hoffman, J.I., Heymann, M.A., Rudolph, A.M., Payne, B.D., 1977. Uses and abuses of the radioactive microsphere method of measuring regional blood flow. Bibl Anat: 20-3. Holt, J.P., Rhode, E.A., Kines, H., 1968. Ventricular volumes and body weight in mammals. Am J Physiol, 215: 704-15. Holtenius, K., Björnhag, G., 1989. The significance of water absorption and fibre digestion in the omasum of sheep, goats and cattle. Comparative Biochemistry and Physiology Part A: Physiology, 94: 105-109. Honda, G.S., Pearce, R.G., Pham, L.L., Setzer, R.W., Wetmore, B.A., Sipes, N.S., Gilbert, J., Franz, B., Thomas, R.S., Wambaugh, J.F., 2019. Using the concordance of in vitro and in vivo data to evaluate extrapolation assumptions. PLoS One, 14: e0217564. Hoogenboom, L.A., Van Eijkeren, J.C., Zeilmaker, M.J., Mengelers, M.J., Herbes, R., Immerzeel, J., Traag, W.A., 2007. A novel source for dioxins present in recycled fat from gelatin production. Chemosphere, 68: 814-23. Hoogenboom, R., Zeilmaker, M., Eijkeren, J.v., Kan, K., Mengelers, M., Luykx, D., Traag, W., 2010. Kaolinic clay derived PCDD/Fs in the feed chain from a sorting process for potatoes. Chemosphere, 78: 99-105. Hsieh, N.H., Reisfeld, B., Bois, F.Y., Chiu, W.A., 2018. Applying a Global Sensitivity Analysis Workflow to Improve the Computational Efficiencies in Physiologically-Based Pharmacokinetic Modeling. Front Pharmacol, 9: 588. Hu, N.X., Chen, M., Liu, Y.S., Shi, Q., Yang, B., Zhang, H.C., Cheng, P., Tang, Q., Liu, Z.Y., Zeng, J.G., 2018. Pharmacokinetics of sanguinarine, chelerythrine, and their metabolites in broiler chickens following oral and intravenous administration. J Vet Pharmacol Ther. Huang, L., Lin, Z., Zhou, X., Zhu, M., Gehring, R., Riviere, J.E., Yuan, Z., 2015a. Estimation of residue depletion of cyadox and its marker residue in edible tissues of pigs using physiologically based pharmacokinetic modelling. Food Additives & Contaminants: Part A, 32: 2002-2017. Huang, Q., Gehring, R., Tell, L.A., Li, M., Riviere, J.E., 2015b. Interspecies allometric meta-analysis of the comparative pharmacokinetics of 85 drugs across veterinary and laboratory animal species. J Vet Pharmacol Ther, 38: 214-26. Huang, Q., Riviere, J.E., 2014. The application of allometric scaling principles to predict pharmacokinetic parameters across species. Expert Opin Drug Metab Toxicol, 10: 1241-53. Huizer, D., Huijbregts, M.A., van Rooij, J.G., Ragas, A.M., 2014. Testing the coherence between occupational exposure limits for inhalation and their biological limit values with a generalized PBPK-model: the case of 2-propanol and . Regul Toxicol Pharmacol, 69: 408-15.

95

Huizer, D., Oldenkamp, R., Ragas, A.M., van Rooij, J.G., Huijbregts, M.A., 2012. Separating uncertainty and physiological variability in human PBPK modelling: The example of 2-propanol and its metabolite acetone. Toxicol Lett, 214: 154-65. Hunter, R.P., 2010. Interspecies allometric scaling. In: Cunningham, F., Elliott, J., Lees, P. (Eds.), Handbook of experimental pharmacology. Springer, Berlin, Heidelberg, pp. 139-57. Huyuk, R., Eraslan, G., 2017. Toxicokinetics of the broad-spectrum pyrethroid insecticide deltamethrin in broiler chickens. Br Poult Sci, 58: 95-99. Immelman, A., Dreyer, G., 1986. Oxytetracycline concentrations in plasma and semen of rams. J S Afr Vet Assoc, 57: 103-4. Jenkins, T.G., Ferrell, C.L., 1997. Changes in proportions of empty body depots and constituents for nine breeds of cattle under various feed availabilities. J Anim Sci, 75: 95-104. Jenkinson, C.M.C., Peterson, S.W., Mackenzie, D.D.S., McDonald, M.F., McCutcheon, S.N., 1995. Seasonal effects on birth weight in sheep are associated with changes in placental development. New Zealand Journal of Agricultural Research, 38: 337-345. JMPR, 2001. Pesticide residues in food-2001 evaluation-Part I, FAO Plant Production and Protection Paper 171, FAO and WHO 2002. Johanson, G., 1991. Modelling of respiratory exchange of polar solvents. Ann Occup Hyg, 35: 323-39. Jones, H., Rowland-Yeo, K., 2013. Basic concepts in physiologically based pharmacokinetic modeling in drug discovery and development. CPT Pharmacometrics Syst Pharmacol, 2: e63. Juca, A.F., Faveri, J.C., Melo Filho, G.M., Ribeiro Filho, A.L., Azevedo, H.C., Muniz, E.N., Pedrosa, V.B., Pinto, L.F., 2016. Effects of birth type and family on the variation of carcass and meat traits in Santa Ines sheep. Trop Anim Health Prod, 48: 435-43. Juhlin, T., Jonsson, B.A., Hoglund, P., 2008. Renal effects of aspirin are clearly dose-dependent and are of clinical importance from a dose of 160 mg. Eur J Heart Fail, 10: 892-8. Jurie, C., Picard, B., Hocquette, J.F., Dransfield, E., Micol, D., Listrat, A., 2007. Muscle and meat quality characteristics of Holstein and Salers cull cows. Meat Sci, 77: 459-66. Kamalzadeh, A., Koops, W.J., van Bruchem, J., Tamminga, S., Zwart, D., 1998. Feed quality restriction and compensatory growth in growing sheep: development of body organs. Small Ruminant Research, 29: 71-82. Kerr, B.J., McKeith, F.K., Easter, R.A., 1995. Effect on performance and carcass characteristics of nursery to finisher pigs fed reduced crude protein, amino acid-supplemented diets. J Anim Sci, 73: 433- 40. Koike, T.I., Pryor, L.R., Neldon, H.L., 1983. Plasma volume and electrolytes during progressive water deprivation in chickens (Gallus domesticus). Comparative Biochemistry and Physiology Part A: Physiology, 74: 83-87. Kosarachukwu, C., Iheshiulor, O., Omede, A., Ogbuewu, P., 2010. Effect of strain on growth, carcass characteristics and meat quality of broilers for 12 weeks. New York Science Journal, 3: 112-116. Krewski, D. et al., 2010. Toxicity testing in the 21st century: a vision and a strategy. J Toxicol Environ Health B Crit Rev, 13: 51-138. Kruska, D., Rohrs, M., 1974. Comparative--quantitative investigations on brains of feral pigs from the Galapagos Islands and of European domestic pigs. Z Anat Entwicklungsgesch, 144: 61-73. Kuipers, J.A., Boer, F., Olofsen, E., Olieman, W., Vletter, A.A., Burm, A.G., Bovill, J.G., 1999. Recirculatory and compartmental pharmacokinetic modeling of alfentanil in pigs: the influence of cardiac output. Anesthesiology, 90: 1146-57. Kulcsar, A., Matis, G., Molnar, A., Petrilla, J., Wagner, L., Febel, H., Husveth, F., Dublecz, K., Neogrady, Z., 2017. Nutritional modulation of intestinal drug-metabolizing cytochrome P450 by butyrate of different origin in chicken. Res Vet Sci, 113: 25-32.

96

Kumar, R., Malik, J.K., 1998. Some Pharmacokinetic Parameters and Dosage Regimens for a Long-Acting Formulation of Oxytetracycline in 6- to 8-Month-Old Male Calves. Veterinary Research Communications, 22: 533-544. Kumar, R., Malik, J.K., 2003. Influence of endotoxin on the disposition kinetics and dosage regimens of oxytetracycline in calves. J Vet Pharmacol Ther, 26: 159-64. Kurita, T., Morita, K., Kazama, T., Sato, S., 2002. Influence of cardiac output on plasma propofol concentrations during constant infusion in swine. Anesthesiology, 96: 1498-503. Kurita, T., Uraoka, M., Jiang, Q., Suzuki, M., Morishima, Y., Morita, K., Sato, S., 2013. Influence of cardiac output on the pseudo-steady state remifentanil and propofol concentrations in swine. Acta Anaesthesiol Scand, 57: 754-60. Landers, K.L., Moore, R.W., Herrera, P., Landers, D.A., Howard, Z.R., McReynolds, J.L., Bryd, J.A., Kubena, L.F., Nisbet, D.J., Ricke, S.C., 2008. Organ weight and serum triglyceride responses of older (80 week) commercial laying hens fed an alfalfa meal molt diet. Bioresour Technol, 99: 6692-6. Laroche, C. et al., 2018. Finding synergies for 3Rs - Toxicokinetics and read-across: Report from an EPAA partners' Forum. Regul Toxicol Pharmacol, 99: 5-21. Lautz, L.S., Dorne, J.L.C.M., Oldenkamp, R., Hendriks, A.J., Ragas, A.M.J., 2020a. Generic physiologically based kinetic modelling for farm animals: Part I. Data collection of physiological parameters in swine, cattle and sheep. Toxicology Letters, 319: 95-101. Lautz, L.S., Hoeks, S., Oldenkamp, R., Hendriks, A.J., Dorne, J.L.C.M., Ragas, A.M.J., 2020b. Generic physiologically-based kinetic modelling for farm animals: Part II. Predicting tissue concentrations of chemicals in swine, cattle, and sheep. Toxicology Letters, 318: 50-56. Lautz, L.S., Oldenkamp, R., Dorne, J.L., Ragas, A.M.J., 2019. Physiologically based kinetic models for farm animals: Critical review of published models and future perspectives for their use in chemical risk assessment. Toxicology in Vitro, 60: 61-70. Leavens, T.L., Tell, L., Kissell, L.W., Smith, G., Smith, D., Wagner, S.A., Shelver, W., Wu, H., Baynes, R.E., Riviere, J.E., 2014. Development of a physiologically based pharmacokinetic model for flunixin in cattle (Bos taurus). Food Additives and Contaminants Part a-Chemistry Analysis Control Exposure & Risk Assessment, 31: 1506-1521. Leavens, T.L., Tell, L.A., Clothier, K.A., Griffith, R.W., Baynes, R.E., Riviere, J.E., 2012. Development of a physiologically based pharmacokinetic model to predict tulathromycin distribution in goats. J Vet Pharmacol Ther, 35: 121-31. Lee, K., Kim, J., Oh, S., Kang, C.W., CW, An, B.K., 2015. Effects of dietary sanguinarine on growth performance, relative organ weight, cecal microflora, serum cholesterol level and meat quality in brioler chickens. The Journal of Poultry Science, 52: 15-22. Lee, Y.S., Lo, J.C., Otton, S.V., Moore, M.M., Kennedy, C.J., Gobas, F., 2017. In vitro to in vivo extrapolation of biotransformation rates for assessing bioaccumulation of hydrophobic organic chemicals in mammals. Environ Toxicol Chem, 36: 1934-1946. Leonard, J.A., Tan, Y.-M., 2019. Tiered approaches for screening and prioritizing chemicals through integration of pharmacokinetics and exposure information with in vitro dose-response data. Computational Toxicology, 12: 100101. Li, M., Cheng, Y.-H., Chittenden, J.T., Baynes, R.E., Tell, L.A., Davis, J.L., Vickroy, T.W., Riviere, J.E., Lin, Z., 2019a. Integration of Food Animal Residue Avoidance Databank (FARAD) empirical methods for drug withdrawal interval determination with a mechanistic population-based interactive physiologically based pharmacokinetic (iPBPK) modeling platform: example for flunixin meglumine administration. Archives of Toxicology. Li, M., Gehring, R., Riviere, J.E., Lin, Z., 2017. Development and application of a population physiologically based pharmacokinetic model for penicillin G in swine and cattle for food safety assessment. Food and Chemical Toxicology, 107: 74-87. 97

Li, M., Gehring, R., Riviere, J.E., Lin, Z., 2018. Probabilistic Physiologically Based Pharmacokinetic Model for Penicillin G in Milk From Dairy Cows Following Intramammary or Intramuscular Administrations. Toxicol Sci, 164: 85-100. Li, M., Mainquist-Whigham, C., Karriker, L.A., Wulf, L.W., Zeng, D., Gehring, R., Riviere, J.E., Coetzee, J.F., Lin, Z., 2019b. An integrated experimental and physiologically based pharmacokinetic modeling study of penicillin G in heavy sows. J Vet Pharmacol Ther. Lin, Z., Gehring, R., Mochel, J.P., Lave, T., Riviere, J.E., 2016a. Mathematical modeling and simulation in animal health - Part II: principles, methods, applications, and value of physiologically based pharmacokinetic modeling in veterinary medicine and food safety assessment. J Vet Pharmacol Ther. Lin, Z., Jaberi-Douraki, M., He, C., Jin, S., Yang, R.S.H., Fisher, J.W., Riviere, J.E., 2017. Performance Assessment and Translation of Physiologically Based Pharmacokinetic Models From acslX to Berkeley Madonna, MATLAB, and R Language: Oxytetracycline and Gold Nanoparticles As Case Examples. Toxicol Sci, 158: 23-35. Lin, Z., Vahl, C.I., Riviere, J.E., 2016b. Human Food Safety Implications of Variation in Food Animal Drug Metabolism. Scientific Reports, 6: 27907. Lindstedt, S.L., Schaeffer, P.J., 2002. Use of allometry in predicting anatomical and physiological parameters of mammals. Laboratory Animals, 36: 1-19. Liu, N.T., Kramer, G.C., Khan, M.N., Kinsky, M.P., Salinas, J., 2015. Is heart-rate complexity a surrogate measure of cardiac output before, during, and after hemorrhage in a conscious sheep model of multiple hemorrhages and resuscitation? J Trauma Acute Care Surg, 79: S93-100. Liu, Y., Guo, L., Zloh, M., Zhang, Y., Huang, J., Wang, L., 2018. Relevance of Breast Cancer Resistance Protein to Pharmacokinetics of Florfenicol in Chickens: A Perspective from In Vivo and In Vitro Studies. Int J Mol Sci, 19. Lomax, M.A., Baird, G.D., 1983. Blood flow and nutrient exchange across the liver and gut of the dairy cow. Effects of lactation and fasting. Br J Nutr, 49: 481-96. Long, N.M., Prado-Cooper, M.J., Krehbiel, C.R., DeSilva, U., Wettemann, R.P., 2010. Effects of nutrient restriction of bovine dams during early gestation on postnatal growth, carcass and organ characteristics, and gene expression in adipose tissue and muscle. J Anim Sci, 88: 3251-61. Long, N.M., Tousley, C.B., Underwood, K.R., Paisley, S.I., Means, W.J., Hess, B.W., Du, M., Ford, S.P., 2012. Effects of early- to mid-gestational undernutrition with or without protein supplementation on offspring growth, carcass characteristics, and adipocyte size in beef cattle. J Anim Sci, 90: 197-206. Louey, S., Cock, M.L., Harding, R., 2005. Long term consequences of low birthweight on postnatal growth, adiposity and brain weight at maturity in sheep. J Reprod Dev, 51: 59-68. Lundeen, G., Manohar, M., Parks, C., 1983. Systemic distribution of blood flow in swine while awake and during 1.0 and 1.5 MAC isoflurane anesthesia with or without 50% nitrous oxide. Anesth Analg, 62: 499-512. Mabelebele, M., Alabi, O., Ngambi, J., Norris, D., Ginidza, M., 2015. Comparison of gastrointestinal tracts and pH values of digestive organs of Ross 208 broiler and indigenous venda chickens fed the same diet. Asian Journal of Animal and Veterinary Advances, 9: 71-76. Máchal, L., 2000. Concentration of plasma cholesterol and total lipids in hens of various laying lines during the laying period. Journal of Animal and Feed Sciences, 9: 687-696. MacLachlan, D.J., 2008. Transfer of fat-soluble pesticides from contaminated feed to poultry tissues and eggs. Br Poult Sci, 49: 290-8. MacLachlan, D.J., 2009. Influence of physiological status on residues of lipophilic xenobiotics in livestock. Food Addit Contam Part A Chem Anal Control Expo Risk Assess, 26: 692-712.

98

MacLachlan, D.J., 2010. Physiologically based pharmacokinetic (PBPK) model for residues of lipophilic pesticides in poultry. Food Addit Contam Part A Chem Anal Control Expo Risk Assess, 27: 302-14. Madden, J.C., Pawar, G., Cronin, M.T.D., Webb, S., Tan, Y.-M., Paini, A., 2019. In silico resources to assist in the development and evaluation of physiologically-based kinetic models. Computational Toxicology, 11: 33-49. Mahgoub, O., Lodge, G., 1998. A comparative study on growth, body composition and carcass tissue distribution in Oamni sheep and goats. The Journal of Agricultural Science, 131: 329-339. Mahmood, I., Balian, J.D., 1996. Interspecies scaling: predicting clearance of drugs in humans. Three different approaches. Xenobiotica, 26: 887-95. Mahony, C., 2019. Building confidence in non-animal methods: Practical examples of collaboration between regulators, researchers and industry. Computational Toxicology, 10: 78-80. Manafi, M., Pirany, N., Noor Ali, M., Hedayati, M., Khalaji, S., Yari, M., 2015. Experimental pathology of T-2 toxicosis and mycoplasma infection on performance and hepatic functions of broiler chickens. Poult Sci, 94: 1483-92. Martinez, M., Mahmood, I., Hunter, R.P., 2006. Interspecies allometric scaling: prediction of clearance in large animal species: part II: mathematical considerations. J Vet Pharmacol Ther, 29: 425-32. Martinez, M.N., Court, M.H., Fink-Gremmels, J., Mealey, K.L., 2018a. Population variability in animal health: Influence on dose-exposure-response relationships: Part I: Drug metabolism and transporter systems. J Vet Pharmacol Ther, 41: E57-e67. Martinez, M.N., Gehring, R., Mochel, J.P., Pade, D., Pelligand, L., 2018b. Population variability in animal health: Influence on dose-exposure-response relationships: Part II: Modelling and simulation. J Vet Pharmacol Ther, 41: E68-e76. Martinsen, K.H., Odegard, J., Olsen, D., Meuwissen, T.H., 2015. Genetic variation in efficiency to deposit fat and lean meat in Norwegian Landrace and Duroc pigs. J Anim Sci, 93: 3794-800. Mason, S.E., Baynes, R.E., Buur, J.L., Riviere, J.E., Almond, G.W., 2008. Sulfamethazine water medication pharmacokinetics and contamination in a commercial pig production unit. J Food Prot, 71: 584- 9. Mavromichalis, I., Emmert, J.L., Aoyagi, S., Baker, D.H., 2000. Chemical Composition of Whole Body, Tissues, and Organs of Young Chickens (Gallus domesticus). Journal of Food Composition and Analysis, 13: 799-807. McCutcheon, S., NBlair, H., Purchas, R., 1993. Body composition and organ weights in fleeceweight- selected and control Romney rams. New Zealand Jounral of Agricultural Reseach, 36: 445-449. McDowell, G.H., Gooden, J.M., Leenanuruksa, D., Jois, M., English, A.W., 1987. Effects of exogenous growth hormone on milk production and nutrient uptake by muscle and mammary tissues of dairy cows in mid-lactation. Aust J Biol Sci, 40: 295-306. McLanahan, E.D., El-Masri, H.A., Sweeney, L.M., Kopylev, L.Y., Clewell, H.J., Wambaugh, J.F., Schlosser, P.M., 2012. Physiologically based pharmacokinetic model use in risk assessment--Why being published is not enough. Toxicol Sci, 126: 5-15. McNally, K., Cotton, R., Cocker, J., Jones, K., Bartels, M., Rick, D., Price, P., Loizou, G., 2012. Reconstruction of Exposure to m-Xylene from Human Biomonitoring Data Using PBPK Modelling, Bayesian Inference, and Markov Chain Monte Carlo Simulation. J Toxicol, 2012: 760281. McNally, K., Cotton, R., Hogg, A., Loizou, G., 2015. Reprint of PopGen: A virtual human population generator. Toxicology, 332: 77-93. McNally, K., Cotton, R., Loizou, G.D., 2011. A Workflow for Global Sensitivity Analysis of PBPK Models. Frontiers in pharmacology, 2: 31-31. McNally, K., Hogg, A., Loizou, G., 2018. A Computational Workflow for Probabilistic Quantitative in Vitro to in Vivo Extrapolation. Frontiers in Pharmacology, 9.

99

Medway, W., Kare, M., 1959. Blood and Plasma Volume, hematocrit, blood specific gravity and serum protein electrophoresis of the chicken. Poult Sci, 38: 624-631. Merrill, G.F., Russo, R.E., Halper, J.M., 1981. Cardiac output distribution before and after endotoxin challenge in the rooster. Am J Physiol, 241: R67-71. Metcalfe, J., Meschia, G., Hellegers, A., Prystowsky, H., Huckabee, W., Barron, D.H., 1962. Observations on the growth rates and organ weights of fetal sheep at altitude and sea level. Q J Exp Physiol Cogn Med Sci, 47: 305-13. Mevius, D.J., Nouws, J.F., Breukink, H.J., Vree, T.B., Driessens, F., Verkaik, R., 1986. Comparative pharmacokinetics, bioavailability and renal clearance of five parenteral oxytetracycline-20% formulations in dairy cows. Vet Q, 8: 285-94. Minervini, S., Accogli, G., Pirone, A., Graïc, J.-M., Cozzi, B., Desantis, S., 2016. Brain Mass and Encephalization Quotients in the Domestic Industrial Pig (Sus scrofa). PLoS ONE, 11: e0157378. Mirsalimi, S.M., Julian, R.J., 1993. Effect of excess sodium bicarbonate on the blood volume and erythrocyte deformability of broiler chickens. Avian Pathol, 22: 495-507. Mirsalimi, S.M., O'Brien, P.J., Julian, R.J., 1993. Blood volume increase in salt-induced pulmonary hypertension, heart failure and ascites in broiler and White Leghorn chickens. Canadian Journal of Veterinary Research, 57: 110-113. Mitchell and Gauthier Associates, 1981. ACSL: Advanced continuous simulation language - User guide/reference manual. P.O. Box 685, Concord, Mass. Moreno, L., Serrano, J.M., Guimera, M.E., Carceles, C.M., 1998. Pharmacokinetics of oxytetracycline after intramuscular administration with lidocaine in sheep, comparison with a conventional formulation. Res Vet Sci, 65: 209-13. Morris, C.A., Bottema, C.D., Cullen, N.G., Hickey, S.M., Esmailizadeh, A.K., Siebert, B.D., Pitchford, W.S., 2010. Quantitative trait loci for organ weights and adipose fat composition in Jersey and Limousin back-cross cattle finished on pasture or feedlot. Anim Genet, 41: 589-96. Mosing, M., Kutter, A.P., Iff, S., Raszplewicz, J., Mauch, J., Bohm, S.H., Tusman, G., 2015. The effects of cardiac output and pulmonary arterial hypertension on volumetric capnography derived- variables during normoxia and hypoxia. J Clin Monit Comput, 29: 187-96. Moss, T.J.M., Doherty, D.A., Nitsos, I., Sloboda, D.M., Harding, R., Newnham, J.P., 2005. Effects into adulthood of single or repeated antenatal corticosteroids in sheep. American Journal of Obstetrics and Gynecology, 192: 146-152. Moughan, P.J., Smith, W.C., Stevens, E.V.J., 1990. Allometric growth of chemical body components and several organs in the pig (20–90 kg liveweight). New Zealand Journal of Agricultural Research, 33: 77-84. Moura, A.S., Ledur, M.C., Boschiero, C., Nones, K., Pinto, L.F., Jaenisch, F.R., Burt, D.W., Coutinho, L.L., 2016. Quantitative trait loci with sex-specific effects for internal organs weights and hematocrit value in a broiler-layer cross. J Appl Genet, 57: 215-24. Moynihan, J.B., Edwards, N.A., 1975. Blood flow in the reproductive tract of the domestic hen. Comparative Biochemistry and Physiology Part A: Physiology, 51: 745-748. Müller, E., Moser, G., Bartenschilager, H., Geldermann, H., 2000. Trait values of growth, carcass and meat quality in Wild Boar, Meishan and Pietrain pigs as well as their crossbred generations. Journal of Animal Breeding and Genetics, 117: 189-202. Nebbia, C., 2001. Biotransformation enzymes as determinants of xenobiotic toxicity in domestic animals. Vet J, 161: 238-52. Nebbia, C., Ceppa, L., Dacasto, M., Nachtmann, C., Carletti, M., 2001. Oxidative monensin metabolism and cytochrome P450 3A content and functions in liver microsomes from horses, pigs, broiler chicks, cattle and rats. Journal of Veterinary Pharmacology and Therapeutics, 24: 399-403.

100

Nebbia, C., Dacasto, M., Rossetto Giaccherino, A., Giuliano Albo, A., Carletti, M., 2003. Comparative expression of liver cytochrome P450-dependent monooxygenases in the horse and in other agricultural and laboratory species. Vet J, 165: 53-64. Nephawe, K.A., Cundiff, L.V., Dikeman, M.E., Crouse, J.D., Van Vleck, L.D., 2004. Genetic relationships between sex-specific traits in beef cattle: mature weight, weight adjusted for body condition score, height and body condition score of cows, and carcass traits of their steer relatives. J Anim Sci, 82: 647-53. Nielsen, P., Gyrd-Hansen, N., 1996. Bioavailability of oxytetracycline, tetracycline and chlortetracycline after oral administration to fed and fasted pigs. Journal of Veterinary Pharmacology and Therapeutics, 19: 305-311. Nieto, R., Lara, L., Barea, R., Garcia-Valverde, R., Aguinaga, M.A., Conde-Aguilera, J.A., Aguilera, J.F., 2012. Response analysis of the Iberian pig growing from birth to 150 kg body weight to changes in protein and energy supply. J Anim Sci, 90: 3809-20. Nightingale, T.E., 1976. Acute isovolemic anemia in anesthetized chickens. Am J Physiol, 231: 1451-6. Njoku, C.P., Adeyemi, O.A., Sogunle, O.M., Aina, A.B.J., 2015. Growth performance, carcass yield and organ weight of growing pigs fed different levels of feed. Slovak Journal of Animal Science, 48: 16-22. Norberg, A., Brauer, K.I., Prough, D.S., Gabrielsson, J., Hahn, R.G., Uchida, T., Traber, D.L., Svensen, C.H., 2005. Volume turnover kinetics of fluid shifts after hemorrhage, fluid infusion, and the combination of hemorrhage and fluid infusion in sheep. Anesthesiology, 102: 985-94. Nouws, J.F., Vree, T.B., Termond, E., Lohuis, J., van Lith, P., Binkhorst, G.J., Breukink, H.J., 1985. Pharmacokinetics and renal clearance of oxytetracycline after intravenous and intramuscular administration to dairy cows. Vet Q, 7: 296-305. Ogorevc, J., Orehek, S., Dovč, P., 2016. Cellular reprogramming in farm animals: an overview of iPSC generation in the mammalian farm animal species. Journal of Animal Science and Biotechnology, 7: 10. Orcutt, M.W., Forrest, J.C., Judge, M.D., Schinckel, A.P., Kuei, C.H., 1990. Practical means for estimating pork carcass composition. J Anim Sci, 68: 3987-97. Osselaere, A., De Bock, L., Eeckhaut, V., De Backer, P., Van Bocxlaer, J., Boussery, K., Croubels, S., 2013a. Hepatic and intestinal CYP3A expression and activity in broilers. J Vet Pharmacol Ther, 36: 588- 93. Osselaere, A., Li, S.J., De Bock, L., Devreese, M., Goossens, J., Vandenbroucke, V., Van Bocxlaer, J., Boussery, K., Pasmans, F., Martel, A., De Backer, P., Croubels, S., 2013b. Toxic effects of dietary exposure to T-2 toxin on intestinal and hepatic biotransformation enzymes and drug transporter systems in broiler chickens. Food Chem Toxicol, 55: 150-5. Paini, A. et al., 2019. Next generation physiologically based kinetic (NG-PBK) models in support of regulatory decision making. Computational Toxicology, 9: 61-72. Paini, A., Leonard, J.A., Kliment, T., Tan, Y.M., Worth, A., 2017. Investigating the state of physiologically based kinetic modelling practices and challenges associated with gaining regulatory acceptance of model applications. Regul Toxicol Pharmacol, 90: 104-115. Pajurek, M., Pietron, W., Maszewski, S., Mikolajczyk, S., Piskorska-Pliszczynska, J., 2019. Poultry eggs as a source of PCDD/Fs, PCBs, PBDEs and PBDD/Fs. Chemosphere, 223: 651-658. Park, J.H., Kim, I.H., 2015. The effects of the supplementation of Bacillus subtilis RX7 and B2A strains on the performance, blood profiles, intestinal Salmonella concentration, noxious gas emission, organ weight and breast meat quality of broiler challenged with Salmonella typhimurium. J Anim Physiol Anim Nutr (Berl), 99: 326-34.

101

Patlewicz, G., Cronin, M.T.D., Helman, G., Lambert, J.C., Lizarraga, L.E., Shah, I., 2018. Navigating through the minefield of read-across frameworks: A commentary perspective. Computational Toxicology, 6: 39-54. Pearce, R.G., Setzer, R.W., Strope, C.L., Wambaugh, J.F., Sipes, N.S., 2017. httk: R Package for High- Throughput Toxicokinetics. J Stat Softw, 79: 1-26. Peebles, E.D., Burnham, M.R., Walzem, R.L., Branton, S.L., Gerard, P.D., 2004. Effects of fasting on serum lipids and lipoprotein profiles in the egg-laying hen (Gallus domesticus). Comp Biochem Physiol A Mol Integr Physiol, 138: 305-11. Pekas, J.C., 1983. Effect of gastric-feeding on feed consumption, growth, organ size, and body composition of swine. Appetite, 4: 87-95. Pethick, D.W., Lindsay, D.B., 1982. Acetate metabolism in lactating sheep. Br J Nutr, 48: 319-28. Peyret, T., Krishnan, K., 2011. QSARs for PBPK modelling of environmental contaminants. SAR QSAR Environ Res, 22: 129-69. Pfuhl, R., Bellmann, O., Kühn, C., Teuscher, F., Ender, K., Wegner, J., 2007. Beef versus dairy cattle: a comparison of feed conversion, carcass composition, and meat quality. Arch Anim Breed, 50. Phibro Animal Health, 2004. Freedom of information summary, Supplemental new animal drug application, NADA 008-804. NADA 008-804. Pijpers, A., Schoevers, E.J., van Gogh, H., van Leengoed, L.A., Visser, I.J., van Miert, A.S., Verheijden, J.H., 1990. The pharmacokinetics of oxytetracycline following intravenous administration in healthy and diseased pigs. Journal of Veterinary Pharmacology and Therapeutics, 13: 320-326. Pijpers, A., Schoevers, E.J., van Gogh, H., van Leengoed, L.A., Visser, I.J., van Miert, A.S., Verheijden, J.H., 1991. The influence of disease on feed and water consumption and on pharmacokinetics of orally administered oxytetracycline in pigs. J Anim Sci, 69: 2947-54. Pinchasov, Y., Elmaliah, S., Bezdin, S., 1994. Plasma apolipoprotein VLDL-II and egg production in laying hens: establishment of an ELISA method. Reprod Nutr Dev, 34: 361-9. Poapolathep, S., Klangkaew, N., Arreesrisom, P., Isariyodom, S., Sugita-Konishi, Y., Kumagai, S., Poapolathep, A., 2015. Toxicokinetics and absolute oral bioavailability of melamine in broiler chickens. J Vet Pharmacol Ther, 38: 101-4. Pollet, R.A., Glatz, C.E., Dyer, D.C., 1985. The pharmacokinetics of chlortetracycline orally administered to turkeys: Influence of citric acid andPasteurella multocida infection. Journal of Pharmacokinetics and Biopharmaceutics, 13: 243-264. Poulin, P., Haddad, S., 2013. Toward a new paradigm for the efficient in vitro-in vivo extrapolation of metabolic clearance in humans from hepatocyte data. J Pharm Sci, 102: 3239-51. Poulin, P., Haddad, S., 2018. Extrapolation of the Hepatic Clearance of Drugs in the Absence of Albumin In Vitro to That in the Presence of Albumin In Vivo: Comparative Assessement of 2 Extrapolation Models Based on the Albumin-Mediated Hepatic Uptake Theory and Limitations and Mechanistic Insights. J Pharm Sci, 107: 1791-1797. Pujol, G. et al., 2017. Sensitivity: global sensitiv-ity analysis of model outputs. R package version 1.14.0. Punt, A., Peijnenburg, A., Hoogenboom, R., Bouwmeester, H., 2017. Non-animal approaches for toxicokinetics in risk evaluations of food chemicals. Altex, 34: 501-514. Punt, A., Schiffelers, M.J., Jean Horbach, G., van de Sandt, J.J., Groothuis, G.M., Rietjens, I.M., Blaauboer, B.J., 2011. Evaluation of research activities and research needs to increase the impact and applicability of alternative testing strategies in risk assessment practice. Regul Toxicol Pharmacol, 61: 105-14. Purdie, N.G., Trout, D.R., Poppi, D.P., Cant, J.P., 2008. Milk Synthetic Response of the Bovine Mammary Gland to an Increase in the Local Concentration of Amino Acids and Acetate. Journal of Dairy Science, 91: 218-228.

102

Qian, M.R., Wang, Q.Y., Yang, H., Sun, G.Z., Ke, X.B., Huang, L.L., Gao, J.D., Yang, J.J., Yang, B., 2017. Diffusion-limited PBPK model for predicting pulmonary pharmacokinetics of florfenicol in pig. J Vet Pharmacol Ther, 40: e30-e38. Quinious, N., Noblet, J., 1995. Prediction of tissular body composition from protein and lipid deposition in growing pigs. J Anim Sci, 73: 1567-75. R Core Development Team, 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. Ragas, A.M., Huijbregts, M.A., 1998. Evaluating the coherence between environmental quality objectives and the acceptable or tolerable daily intake. Regul Toxicol Pharmacol, 27: 251-64. Ragas, A.M.J., Etienne, R.S., Willemsen, F.H., Van De Meent, D., 1999. Assessing model uncertainty for environmental decision making: A case study of the coherence of independently derived environmental quality objectives for air and water. Environmental Toxicology and Chemistry, 18: 1856-1867. Ratz, V., Maas, R., Semjén, G., van Miert, A.S.P.J.A.M., Witkamp, R.F., 1995. Oral bioavailability of sulphonamides in ruminants: A comparison between sulphamethoxazole, sulphatroxazole, and sulphamerazine, using the dwarf goat as animal model. Veterinary Quarterly, 17: 82-87. Razmaite, V., Kerziene, S., Jatkauskiene, V., 2009. Body and carcass measurements and organ weights of Lithuanian indigenous pigs and their wild boar hybrids. Animal Science Papers and Reports, 27: 331-342. Reddy, M.B., Clewell, H.J., Andersen, M.E., 2005. Physiology based pharmacokinetic Modeling: Science and Applications. John Wiley & Sons, Inc., Hoboken, New Jersey. Reed, J.J., Ward, M.A., Vonnahme, K.A., Neville, T.L., Julius, S.L., Borowicz, P.P., Taylor, J.B., Redmer, D.A., Grazul-Bilska, A.T., Reynolds, L.P., Caton, J.S., 2007. Effects of selenium supply and dietary restriction on maternal and fetal body weight, visceral organ mass and cellularity estimates, and jejunal vascularity in pregnant ewe lambs. J Anim Sci, 85: 2721-33. Ren, M.Q., Wegner, J., Bellmann, O., Brockmann, G.A., Schneider, F., Teuscher, F., Ender, K., 2002. Comparing mRNA levels of genes encoding leptin, leptin receptor, and lipoprotein lipase between dairy and beef cattle. Domest Anim Endocrinol, 23: 371-81. Richard, A.M. et al., 2016. ToxCast Chemical Landscape: Paving the Road to 21st Century Toxicology. Chem Res Toxicol, 29: 1225-51. Riviere, J.E., Gabrielsson, J., Fink, M., Mochel, J., 2016. Mathematical modeling and simulation in animal health. Part I: Moving beyond pharmacokinetics. J Vet Pharmacol Ther, 39: 213-23. Riviere, J.E., Martin-Jimenez, T., Sundlof, S.F., Craigmill, A.L., 1997. Interspecies allometric analysis of the comparative pharmacokinetics of 44 drugs across veterinary and laboratory animal species. J Vet Pharmacol Ther, 20: 453-63. Rotta, P.P., Filho, S.C., Gionbelli, T.R., Costa, E.S.L.F., Engle, T.E., Marcondes, M.I., Campos, M.M., Menezes, A.C., Lobo, A.A., 2015. Effects of day of gestation and feeding regimen in Holstein x Gyr cows: II. Maternal and fetal visceral organ mass. J Dairy Sci, 98: 3211-23. Rowland, M., Peck, C., Tucker, G., 2011. Physiologically-based pharmacokinetics in drug development and regulatory science. Annu Rev Pharmacol Toxicol, 51: 45-73. Rudolph, A.M., Heymann, M.A., 1967. The circulation of the fetus in utero. Methods for studying distribution of blood flow, cardiac output and organ blood flow. Circ Res, 21: 163-84. Runciman, W.B., Ilsley, A.H., Mather, L.E., Carapetis, R., Rao, M.M., 1984. A sheep preparation for studying interactions between blood flow and drug disposition. I: Physiological profile. Br J Anaesth, 56: 1015-28. Runciman, W.B., Mather, L.E., Ilsley, A.H., Carapetis, R.J., Upton, R.N., 1986. A sheep preparation for studying interactions between blood flow and drug disposition VI: effects of general or

103

subarachnoid anastehsia on blood flow and chlormethiazole disposition. BJA: British Journal of Anaesthesia, 58: 1308-1316. Ruusunen, M., Partanen, K., Pösö, R., Puolanne, E., 2007. The effect of dietary protein supply on carcass composition, size of organs, muscle properties and meat quality of pigs. Livestock Science, 107: 170-181. Ryu, S.T., Park, B.S., Bang, H.T., Kang, H.K., Hwangbo, J., 2016. Effects of anti-heat diet and inverse lighting on growth performance, immune organ, microorganism and short chain fatty acids of broiler chickens under heat stress. J Environ Biol, 37: 185-92. Sachana, M., 2019. An international effort to promote the regulatory use of PBK models based on non- animal data. Computational Toxicology, 11: 23-24. Sager, J.E., Yu, J., Ragueneau-Majlessi, I., Isoherranen, N., 2015. Physiologically Based Pharmacokinetic (PBPK) Modeling and Simulation Approaches: A Systematic Review of Published Models, Applications, and Model Verification. Drug Metab Dispos, 43: 1823-37. Saltelli, A., Chan, K., Scott, E.M., 2008. Sensitivity analysis, New York. Sapirstein, L., Hartman, F., 1959. Cardiac output and its distribution in the chicken. American Journal of Physiology-Legacy Content, 196: 751-752. Schefferlie, G.J., Hekman, P., 2016. Prediction of the residue levels of drugs in eggs, using physicochemical properties and their influence on passive diffusion processes. J Vet Pharmacol Ther, 39: 381-7. Schifferli, D., Galeazzi, R.L., Nicolet, J., Wanner, M., 1982. Pharmacokinetics of oxytetracycline and therapeutic implications in veal calves. J Vet Pharmacol Ther, 5: 247-57. Schmitt, W., 2008. General approach for the calculation of tissue to plasma partition coefficients. Toxicol In Vitro, 22: 457-67. Schrickx, J.A., Fink-Gremmels, J., 2008. Implications of ABC transporters on the disposition of typical veterinary medicinal products. Eur J Pharmacol, 585: 510-9. Shahzad, M.N., Javed, M.T., Shabir, S., Irfan, M., Hussain, R., 2012. Effects of feeding urea and copper sulphate in different combinations on live body weight, carcass weight, percent weight to body weight of different organs and histopathological tissue changes in broilers. Experimental and Toxicologic Pathology, 64: 141-147. Sharma, V., McNeill, J.H., 2009. To scale or not to scale: the principles of dose extrapolation. Br J Pharmacol, 157: 907-21. Shen, J., Hu, D., Wu, X., Coats, J.R., 2003. Bioavailability and pharmacokinetics of florfenicol in broiler chickens. J Vet Pharmacol Ther, 26: 337-41. Shih, P.-C., Hung, Y.-C., Chen, Y.-L., Tsai, H.-J., Chen, C.-Y., Huang, C.-J., 2013. Agreement of cardiac output measurement between pulse contour analysis and thermodilution in various body positions: a porcine study. Journal of Surgical Research, 181: 315-322. Sieo, C.C., Abdullah, N., Tan, W.S., Ho, Y.W., 2005. Influence of beta-glucanase-producing Lactobacillus strains on intestinal characteristics and feed passage rate of broiler chickens. Poult Sci, 84: 734- 41. Silano, M., Silano, V., 2017. Food and feed chemical contaminants in the European Union: Regulatory, scientific, and technical issues concerning chemical contaminants occurrence, risk assessment, and risk management in the European Union. Crit Rev Food Sci Nutr, 57: 2162-2217. Sinclair, L.A., Weerasinghe, W.M., Wilkinson, R.G., de Veth, M.J., Bauman, D.E., 2010. A supplement containing trans-10, cis-12 conjugated linoleic acid reduces milk fat yield but does not alter organ weight or body fat deposition in lactating ewes. J Nutr, 140: 1949-55. Sjögren, E., Bredberg, U., Lennernäs, H., 2012. The Pharmacokinetics and Hepatic Disposition of Repaglinide in Pigs: Mechanistic Modeling of Metabolism and Transport. Molecular Pharmaceutics, 9: 823-841. 104

Sjogren, E., Hedeland, M., Bondesson, U., Lennernas, H., 2014. Effects of verapamil on the pharmacokinetics and hepatobiliary disposition of fexofenadine in pigs. Eur J Pharm Sci, 57: 214- 23. Sobol’, I.M., Tarantola, S., Gatelli, D., Kucherenko, S.S., Mauntz, W., 2007. Estimating the approximation error when fixing unessential factors in global sensitivity analysis. Reliability Engineering & System Safety, 92: 957-960. Stebel, S., Wideman, R.F., 2008. Pulmonary hemodynamic responses to intravenous prostaglandin E2 in broiler chickens. Poult Sci, 87: 138-45. Stoev, S.D., Stefanov, M., Denev, S., Radic, B., Domijan, A.M., Peraica, M., 2004. Experimental mycotoxicosis in chickens induced by ochratoxin A and penicillic acid and intervention with natural plant extracts. Vet Res Commun, 28: 727-46. Stonestreet, B.S., Ocampo, S.S., Oh, W., 1998. Reductions in cardiac output in hypoxic young pigs: systemic and regional perfusion and oxygen metabolism. Journal of Applied Physiology, 85: 874- 882. Sturkie, P.D., Abati, A., 1975. Blood flow in mesenteric, hepatic portal and renal portal veins of chickens. Pflugers Arch, 359: 127-35. Sturkie, P.D., Vogel, J.A., 1959. Cardiac output, central blood volume and peripheral resistance in chickens. Am J Physiol, 197: 1165-6. Sun, P., Wang, J.Q., Shen, J.S., Wei, H.Y., 2011. Residues of melamine and cyanuric acid in milk and tissues of dairy cows fed different doses of melamine. Journal of Dairy Science, 94: 3575-3582. Sun, Y., Peng, Y., Aksornkoae, N., Johnson, J.R., Boring, J.G., Scruggs, D., Cooper, R.C., Laizure, S.C., Shukla, A.J., 2002. Controlled release of oxytetracycline in sheep. Journal of Controlled Release, 85: 125-134. Swanson, T.J., Hammer, C.J., Luther, J.S., Carlson, D.B., Taylor, J.B., Redmer, D.A., Neville, T.L., Reed, J.J., Reynolds, L.P., Caton, J.S., Vonnahme, K.A., 2008. Effects of gestational plane of nutrition and selenium supplementation on mammary development and colostrum quality in pregnant ewe lambs. J Anim Sci, 86: 2415-23. Swett, W., 1937. Comparative conformation, anatomy, and udder characteristics of cows of certain beef and dairy breeders. Journal of Agricultural Research, 55: 239-288. Szabo, J., Andrasofszky, E., Tuboly, T., Bersenyi, A., Weisz, A., Hetenyi, N., Hullar, I., 2014. Effect of arginine or glutamine supplementation on production, organ weights, interferon gamma, interleukin 6 and antibody titre of broilers. Acta Vet Hung, 62: 348-61. Talke, P.O., Traber, D.L., Richardson, C.A., Harper, D.D., Traber, L.D., 2000. The effect of alpha(2) agonist- induced sedation and its reversal with an alpha(2) antagonist on organ blood flow in sheep. Anesth Analg, 90: 1060-6. Tan, Y.M., Worley, R.R., Leonard, J.A., Fisher, J.W., 2018. Challenges Associated With Applying Physiologically Based Pharmacokinetic Modeling for Public Health Decision-Making. Toxicol Sci, 162: 341-348. Thein, E., Becker, M., Anetzberger, H., Hammer, C., Messmer, K., 2003. Direct assessment and distribution of regional portal blood flow in the pig by means of fluorescent microspheres. J Appl Physiol (1985), 95: 1808-16. Thomas, R.S. et al., 2019. The Next Generation Blueprint of Computational Toxicology at the U.S. Environmental Protection Agency. Toxicological Sciences, 169: 317-332. Thompson, G.E., 1980. The distribution of blood flow in the udder of the sheep and changes brought about by cold exposure and lactation. J Physiol, 302: 379-86. Tickle, P.G., Paxton, H., Rankin, J.W., Hutchinson, J.R., Codd, J.R., 2014. Anatomical and biomechanical traits of broiler chickens across ontogeny. Part I. Anatomy of the musculoskeletal respiratory apparatus and changes in organ size. PeerJ, 2: e432. 105

Tilahun, M., Kefelegn, K., Abebe, G., Goetsch, A.L., 2014. Feed intake, digestibility, weight gain, and slaughter characteristics influenced by genetic percentage of Boer in goats and Dorper in sheep in the central highlands of Ethiopia. Trop Anim Health Prod, 46: 593-602. Tkachenko, A., Clark, J., Knutson, N., Wallace, B., Bomba, M., Yacopucci, M., Rhodes, B., Nemser, S.M., Guag, J., Reimschuessel, R., 2015. Investigation of melamine and cyanuric acid deposition in pig tissues using LC-MS/MS methods. Food and Chemical Toxicology, 80: 310-318. Togun, V., Farinu, G., Olabanji, R., 2006. Feeding Graded Levels of Wild Sunflower (Tithonia diversifolia Hemsl. A. Gray) Meal in Replacement of Maize at Pre-pubertal Age, Negatively Impacts on Growth and Morphormetric Characteristics of the Genitalia of Anak 2000 Broiler Cocks at Their Pubertal Age. World Applied Sciences Journal, 2: 115-121. Toutain, P.L., Ferran, A., Bousquet-Melou, A., 2010. Species differences in pharmacokinetics and pharmacodynamics, Comparative and Veterinary Pharmacology. Handbook of Experimental Pharmacology. Springer, Berlin, Heidelberg, pp. 19-48. Toutain, P.L., Raynaud, J.P., 1983. Pharmacokinetics of oxytetracycline in young cattle: comparison of conventional vs long-acting formulations. Am J Vet Res, 44: 1203-9. Tranquilli, W.J., Manohar, M., Parks, C.M., Thurmon, J.C., Theodorakis, M.C., Benson, G.J., 1982. Systemic and regional blood flow distribution in unanesthetized swine and swine anesthetized with halothane + nitrous oxide, halothane, or enflurane. Anesthesiology, 56: 369-79. Ullman, J., Eriksson, S., Rundgren, M., 2001. Effects of losartan, prazosin and a vasopressin V1-receptor antagonist on renal and femoral blood flow in conscious sheep. Acta Physiol Scand, 171: 99-104. Upton, R., 2008. Organ weights and blood flows of sheep and pig for physiological pharmacokinetic modelling. Journal of Pharmacological and Toxicological Methods, 58: 198-205. van Asselt, E.D., Kowalczyk, J., van Eijkeren, J.C.H., Zeilmaker, M.J., Ehlers, S., Fürst, P., Lahrssen- Wiederholt, M., van der Fels-Klerx, H.J., 2013. Transfer of perfluorooctane sulfonic acid (PFOS) from contaminated feed to dairy milk. Food Chemistry, 141: 1489-1495. van Eijkeren, J.C.H., Jager, D.T., Sips, A.J.A.M., 1998. Generic PBPK-modelling of lipophilic contaminants in the cow. 679102042, National Institute for Public Health and the Environment, Bilthoven, The Netherlands. van Eijkeren, J.C.H., Zeilmaker, M.J., Kan, C.A., Traag, W.A., Hoogenboom, L.A.P., 2006. A toxicokinetic model for the carry-over of dioxins and PCBs from feed and soil to eggs. Food Additives & Contaminants, 23: 509-517. van Essen, G.J., te Lintel Hekkert, M., Sorop, O., Heinonen, I., van der Velden, J., Merkus, D., Duncker, D.J., 2018. Cardiovascular Function of Modern Pigs Does not Comply with Allometric Scaling Laws. Scientific Reports, 8: 792. Venturini, G.C., Cruz, V.A., Rosa, J.O., Baldi, F., El Faro, L., Ledur, M.C., Peixoto, J.O., Munari, D.P., 2014. Genetic and phenotypic parameters of carcass and organ traits of broiler chickens. Genet Mol Res, 13: 10294-300. Verstraete, F., 2013. Risk management of undesirable substances in feed following updated risk assessments. Toxicol Appl Pharmacol, 270: 230-47. Vimercati, S., Elli, S., Jagannathan, V., Pandey, A.V., Peduto, N., Leeb, T., Mevissen, M., 2019. In silico and in vitro analysis of genetic variants of the equine CYP3A94, CYP3A95 and CYP3A97 isoenzymes. Toxicol In Vitro, 60: 116-124. Viscor, G., Marques, M.S., Palomeque, J., 1985. Cardiovascular and organ weight adaptations as related to flight activity in birds. Comp Biochem Physiol A Comp Physiol, 82: 597-9. Visser, M., Weber, K.L., Lyons, L.A., Rincon, G., Boothe, D.M., Merritt, D.A., 2019. Identification and quantification of domestic feline cytochrome P450 transcriptome across multiple tissues. Journal of Veterinary Pharmacology and Therapeutics, 42: 7-15.

106

Vogel, J.A., Sturkie, P.D., 1963. Cardiovascular responses of the chicken to seasonal and induced temperature changes. Science, 140: 1404-6. von Engelhardt, W., Hales, J.R., 1977. Partition of capillary blood flow in rumen, reticulum, and omasum of sheep. Am J Physiol, 232: E53-6. von Soosten, D., Meyer, U., Piechotta, M., Flachowsky, G., Danicke, S., 2012. Effect of conjugated linoleic acid supplementation on body composition, body fat mobilization, protein accretion, and energy utilization in early lactation dairy cows. J Dairy Sci, 95: 1222-39. Wagner, J.G., 1981. History of pharmacokinetics. Pharmacology & Therapeutics, 12: 537-562. Wallace, J.M., Bourke, D.A., Aitken, R.P., Leitch, N., Hay, W.W., Jr., 2002. Blood flows and nutrient uptakes in growth-restricted pregnancies induced by overnourishing adolescent sheep. Am J Physiol Regul Integr Comp Physiol, 282: R1027-36. Wambaugh, J.F., Wetmore, B.A., Pearce, R., Strope, C., Goldsmith, R., Sluka, J.P., Sedykh, A., Tropsha, A., Bosgra, S., Shah, I., Judson, R., Thomas, R.S., Setzer, R.W., 2015. Toxicokinetic Triage for Environmental Chemicals. Toxicological Sciences, 147: 55-67. Wang, G.Y., Zheng, H.H., Zhang, K.Y., Yang, F., Kong, T., Zhou, B., Jiang, S.X., 2018. The roles of cytochrome P450 and P-glycoprotein in the pharmacokinetics of florfenicol in chickens. Iran J Vet Res, 19: 9-14. Wang, X., Martinez, M.A., Wu, Q., Ares, I., Martinez-Larranaga, M.R., Anadon, A., Yuan, Z., 2016. Fipronil insecticide toxicology: oxidative stress and metabolism. Crit Rev Toxicol, 46: 876-899. Wang, Z., Ma, X., Zhang, L., Yang, W., Gong, L., He, P., Li, Z., 2010. Screening and determination of melamine residues in tissue and body fluid samples. Anal Chim Acta, 662: 69-75. Watanabe, K.P., Kawai, Y.K., Ikenaka, Y., Kawata, M., Ikushiro, S., Sakaki, T., Ishizuka, M., 2013. Avian cytochrome P450 (CYP) 1-3 family genes: isoforms, evolutionary relationships, and mRNA expression in chicken liver. PLoS One, 8: e75689. Weinroth, M.D., Belk, A.D., Belk, K.E., 2018. History, development, and current status of food safety systems worldwide. Animal Frontiers, 8: 9-15. Wels, A., Schnappauf, H., Horn, V., 1967. Blutvolumenbestimmung bei Hühnern mit Cr 51 und T-1824. Blackwell Verlag GmbH. West, G.B., Brown, J.H., Enquist, B.J., 1997. A general model for the origin of allometric scaling laws in biology. Science, 276: 122-6. Wetmore, B.A., Wambaugh, J.F., Allen, B., Ferguson, S.S., Sochaski, M.A., Setzer, R.W., Houck, K.A., Strope, C.L., Cantwell, K., Judson, R.S., LeCluyse, E., Clewell, H.J., Thomas, R.S., Andersen, M.E., 2015. Incorporating High-Throughput Exposure Predictions With Dosimetry-Adjusted In Vitro Bioactivity to Inform Chemical Toxicity Testing. Toxicological Sciences, 148: 121-136. Whittow, G.C., Sturkie, P.D., Stein, G., Jr., 1964. Cardiovascular changes associated with thermal polypnea in the chicken. Am J Physiol, 207: 1349-53. WHO, 2010. Characterization and Application of Physiologically Based Pharmacokinetic Models in Risk Assessment, World Health Organization (WHO), Geneva, Switzerland. Wideman, R.F., Jr., 1999. Cardiac output in four-, five-, and six-week-old broilers, and hemodynamic responses to intravenous injections of epinephrine. Poult Sci, 78: 392-403. Wideman, R.F., Jr., Forman, M.F., Hughes, J.D., Jr., Kirby, Y.K., Marson, N., Anthony, N.B., 1998. Flow- dependent pulmonary vasodilation during acute unilateral pulmonary artery occlusion in Jungle Fowl. Poult Sci, 77: 615-26. Wideman, R.F., Jr., Tackett, C.D., 2000. Cardio-pulmonary function in broilers reared at warm or cool temperatures: effect of acute inhalation of 100% oxygen. Poult Sci, 79: 257-64. Will, D.H., Alexander, A.F., Reeves, J.T., Grover, R.F., 1962. High altitude-induced pulmonary hypertension in normal cattle. Circ Res, 10: 172-7.

107

Williams, A.J., Grulke, C.M., Edwards, J., McEachran, A.D., Mansouri, K., Baker, N.C., Patlewicz, G., Shah, I., Wambaugh, J.F., Judson, R.S., Richard, A.M., 2017. The CompTox Chemistry Dashboard: a community data resource for environmental chemistry. J Cheminform, 9: 61. Williams, F.L., Rodbard, S., 1960. Increased circulating plasma volume following phenoxybenzamine (dibenzyline). Am J Physiol, 198: 169-72. Wiseman, T.G., Mahan, D.C., Peters, J.C., Fastinger, N.D., Ching, S., Kim, Y.Y., 2007. Tissue weights and body composition of two genetic lines of barrows and gilts from twenty to one hundred twenty- five kilograms of body weight. J Anim Sci, 85: 1825-35. Wolfenson, D., Berman, A., Frei, Y.F., Snapir, N., 1978. Measurement of blood flow distribution by radioactive microspheres in the laying hen (gallus domesticus). Comparative Biochemistry and Physiology Part A: Physiology, 61: 549-554. Wolfenson, D., Frei, Y.F., Snapir, N., Berman, A., 1981. Heat stress effects on capillary blood flow and its redistribution in the laying hen. Pflugers Arch, 390: 86-93. Wood, K.M., Awda, B.J., Fitzsimmons, C., Miller, S.P., McBride, B.W., Swanson, K.C., 2013. Influence of pregnancy in mid-to-late gestation on circulating metabolites, visceral organ mass, and abundance of proteins relating to energy metabolism in mature beef cows. J Anim Sci, 91: 5775- 84. Xie, H., Yang, J., Feng, S., Cheng, P., Zeng, J., Xiong, X., 2015. Simultaneous quantitative determination of sanguinarine, chelerythrine, dihydrosanguinarine and dihydrochelerythrine in chicken by HPLC- MS/MS method and its applications to drug residue and pharmacokinetic study. J Chromatogr B Analyt Technol Biomed Life Sci, 985: 124-30. Yahav, S., Straschnow, A., Plavnik, I., Hurwitz, S., 1997. Blood system response of chickens to changes in environmental temperature. Poult Sci, 76: 627-33. Yang, B., Huang, L.L., Fang, K., Wang, Y.L., Peng, D.P., Liu, Z.L., Yuang, Z.H., 2014a. A physiologically based pharmacokinetic model for the prediction of the depletion of methyl-3-quinoxaline-2-carboxylic acid, the marker residue of olaquindox, in the edible tissues of pigs. Journal of Veterinary Pharmacology and Therapeutics, 37: 66-82. Yang, F., Liu, H.W., Li, M., Ding, H.Z., Huang, X.H., Zeng, Z.L., 2012. Use of a Monte Carlo analysis within a physiologically based pharmacokinetic model to predict doxycycline residue withdrawal time in edible tissues in swine. Food Additives & Contaminants, 29: 73-84. Yang, F., Sun, N., Liu, Y.M., Zeng, Z.L., 2015a. Estimating danofloxacin withdrawal time in broiler chickens based on physiologically based pharmacokinetics modeling. Journal of Veterinary Pharmacology and Therapeutics, 38: 174-182. Yang, F., Yang, Y.R., Wang, L., Huang, X.H., Qiao, G., Zeng, Z.L., 2014b. Estimating marbofloxacin withdrawal time in broiler chickens using a population physiologically based pharmacokinetics model. Journal of Veterinary Pharmacology and Therapeutics, 37: 579-588. Yang, T.S., Lin, J.H., 2010. Variation of heart size and its correlation with growth performance and vascular space in domestic pigs. Animal Science, 64: 523-528. Yang, X., Zhou, Y.F., Yu, Y., Zhao, D.H., Shi, W., Fang, B.H., Liu, Y.H., 2015b. A physiologically based pharmacokinetic model for quinoxaline-2-carboxylic acid in rats, extrapolation to pigs. Journal of Veterinary Pharmacology and Therapeutics, 38: 55-64. Yar, M., Ahmad, M., Bukhari, N.I., 2000. Pharmacokinetics of oxytetracycline in sheep after various intravenous doses. Turk J vet Anim Sci, 24: 135-138. Yin, Y.-B., Guo, S.-G., Wan, D., Wu, X., Yin, Y.-L., 2019. Enteroids: Promising in Vitro Models for Studies of Intestinal Physiology and Nutrition in Farm Animals. Journal of Agricultural and Food Chemistry, 67: 2421-2428.

108

Yokhana, J.S., Parkinson, G., Frankel, T.L., 2016. Effect of insoluble fiber supplementation applied at different ages on digestive organ weight and digestive enzymes of layer-strain poultry. Poult Sci, 95: 550-9. Yuan, L.G., Luo, X.Y., Zhu, L.X., Wang, R., Liu, Y.H., 2011. A physiologically based pharmacokinetic model for valnemulin in rats and extrapolation to pigs. J Vet Pharmacol Ther, 34: 224-31. Zeng, D., Lin, Z., Fang, B., Li, M., Gehring, R., Riviere, J.E., Zeng, Z., 2017. Pharmacokinetics of Mequindox and Its Marker Residue 1,4-Bisdesoxymequindox in Swine Following Multiple Oral Gavage and Intramuscular Administration: An Experimental Study Coupled with Population Physiologically Based Pharmacokinetic Modeling. Journal of Agricultural and Food Chemistry, 65: 5768-5777. Zeng, D., Lin, Z., Zeng, Z., Fang, B., Li, M., Cheng, Y.-H., Sun, Y., 2019. Assessing Global Human Exposure to T-2 Toxin via Poultry Meat Consumption Using a Lifetime Physiologically Based Pharmacokinetic Model. Journal of Agricultural and Food Chemistry, 67: 1563-1571. Zhang, X.-Y., Trame, M., Lesko, L., Schmidt, S., 2015. Sobol Sensitivity Analysis: A Tool to Guide the Development and Evaluation of Systems Pharmacology Models. CPT: Pharmacometrics & Systems Pharmacology, 4: 69-79. Zhao, L., Ball, C.H., 2009. Determination of Chloramphenicol, Florfenicol, and Thiamphenicol in Honey Using Agilent SampliQ OPT Solid-Phase Extraction Cartridges and Liquid Chromatography- Tandem Mass Spectrometry. Agilent Technologies: 1-6. Zhu, X., Huang, L., Xu, Y., Xie, S., Pan, Y., Chen, D., Liu, Z., Yuan, Z., 2017. Physiologically based pharmacokinetic model for quinocetone in pigs and extrapolation to mequindox. Food Addit Contam Part A Chem Anal Control Expo Risk Assess, 34: 192-210.

109

APPENDICES

110

APPENDIX A

PBK model was developed using R software (version 3.3.3). PBK model input ### Model PBK (cattle, sheep, swine) generic ### # 01/2019 # L.S. Lautz, S. Hoeks, R. Oldenkamp #======# # probabilistic # # multi-compartment model for # # farm animal species # #======#

#source functions source('singlechemical_dmdt.R')

#Loading physiology data fBW <- read.csv("data_fBW_animals.csv",stringsAsFactors = FALSE) #organ fractions, incl distribution parameters fCO <- read.csv("data_fCO_animals.csv",stringsAsFactors = FALSE) #blood flow fractions, incl distribution parameters rates <- read.csv("data_rates_animals.csv",stringsAsFactors = FALSE) #physiological rates fPC <- read.csv("data_pc_tissue.csv",stringsAsFactors = FALSE) #tissue composition

#Loading TK data TK <- read.csv("data_TK.csv",stringsAsFactors=FALSE) #toxicokinetic parameters

#Loading physicochemical data chem <- read.csv("data_chem.csv", stringsAsFactors = FALSE) #physicochemical parameters

# Inputs ---- #Defining the exposure scenario species <- "species" #cattle_d/cattle_b/sheep/swine chemical <- "chemical" #add chemical name

regime <- "bolus" #exposure regime (bolus) route <- "route" #exposure route (oral/iv) E_dose <- 0 #exposure dose (mg/kgbw) E_start <- 0 #start of exposure phase (h) E_end <- 24 #end of exposure phase (h) E_int <- 24 #interval between doses (h)

#Simulation parameters A_type <- "single" #type of probabilistic analysis ("single" or "VA") chem_fix <- TRUE #fixing the chemical and TK parameters or not (TRUE/FALSE) n_sim <- 1000 #number of iterations n_out <- 9 #number of compartments to output (blood, total body) t_start <- 0 #start of simulation (h) t_end <- 24 #end of simulation (h) t_A <- c(seq(0.025,0.225,by=0.025),seq(0.25,24,by=0.25)) #time points (h), only relevant when chem_fix=TRUE

#======#Setup single animal simulation-----

if(A_type=="single"){ Names_fix <- c("BW", paste("fBW",colnames(fBW[2+3*c(1:19)]),sep="_"), "CO", paste("fCO",colnames(fCO[2+3*c(1:19)]),sep="_"), colnames(rates[2+3*c(0:4)]), colnames(fPC[2:57]), "nl","pl","pr","w", colnames(TK[3:8]), colnames(chem[2:5])) NP_fix <- length(Names_fix) #number of fixed parameters

fix_in <- cbind(fBW[fBW$species==species,"BW"], fBW[fBW$species==species,colnames(fBW)%in%substr(Names_fix[grep('fBW',Names_fix)],start=5,stop=nchar(Names_fix[grep('fBW',Names_fi x)]))], fCO[fCO$species==species,"CO"], fCO[fCO$species==species,colnames(fCO)%in%substr(Names_fix[grep('fCO',Names_fix)],start=5,stop=nchar(Names_fix[grep('fCO',Names_fi x)]))], rates[rates$species==species,colnames(rates)%in%Names_fix], fPC[fPC$species==species,colnames(fPC)%in%Names_fix], data.frame(nl=1,pl=1,pr=1,w=1), TK[TK$species==species&TK$chemical==chemical,colnames(TK)%in%Names_fix], chem[chem$chemical==chemical,colnames(chem)%in%Names_fix]) colnames(fix_in) <- Names_fix

par_in <- fix_in }

#======#Setup probabilistic simulations----

if(A_type=="VA") { #variability analysis with fixed chemical and toxicokinetic parameters Names_var <- c("BW", paste("fBW",colnames(fBW[2+3*c(1:19)]),sep="_"), 111

"CO", paste("fCO",colnames(fCO[2+3*c(1:19)]),sep="_"), colnames(rates[2+3*c(0:4)]))

#initial list with mean + sd values of all varying parameters Means <- cbind(fBW[fBW$species==species,2+3*c(0:19)], fCO[fCO$species==species,2+3*c(0:19)], rates[rates$species==species,2+3*c(0:4)]) SDs <- cbind(fBW[fBW$species==species,3+3*c(0:19)], fCO[fCO$species==species,3+3*c(0:19)], rates[rates$species==species,3+3*c(0:4)]) Distributions <- cbind(fBW[fBW$species==species,4+3*c(0:19)], fCO[fCO$species==species,4+3*c(0:19)], rates[rates$species==species,4+3*c(0:4)]) colnames(Means) = colnames(SDs) = colnames(Distributions) = Names_var Names_var_not <- colnames(SDs[which(is.na(SDs))])

Means <- Means[!(colnames(Means)%in%Names_var_not)] #exclude all parameters without SD SDs <- SDs[!(colnames(SDs)%in%Names_var_not)] #exclude all parameters without SD Distributions <- Distributions[!(colnames(Distributions)%in%Names_var_not)] #exclude all parameters without SD

Names_var <- colnames(SDs) NP_var <- length(Names_var) #number of varying parameters

#list with names of all fixed parameters (including 'varying parameters' that are 0 or NA) Names_fix <- c(colnames(fPC[2:57]), "nl","pl","pr","w", colnames(TK[3:8]), colnames(chem[2:5]), Names_var_not) NP_fix <- length(Names_fix) #number of fixed parameters

fix_in <- cbind(fPC[fPC$species==species,colnames(fPC)%in%Names_fix], data.frame(nl=1,pl=1,pr=1,w=1), TK[TK$species==species&TK$chemical==chemical,colnames(TK)%in%Names_fix], chem[chem$chemical==chemical,colnames(chem)%in%Names_fix], fBW[fBW$species==species,colnames(fBW)%in%substr(Names_fix[grep('fBW',Names_fix)],start=5,stop=nchar(Names_fix[grep('fBW',Names_fi x)]))], fCO[fCO$species==species,colnames(fCO)%in%substr(Names_fix[grep('fCO',Names_fix)],start=5,stop=nchar(Names_fix[grep('fCO',Names_fi x)]))], rates[rates$species==species,colnames(rates)%in%Names_fix]) colnames(fix_in) <- Names_fix

#create data frames with random samples X1 <- matrix(NA, nrow = n_sim, ncol = NP_var) colnames(X1) <- Names_var X1 <- as.data.frame(X1) SARes <- X1

for(i in 1:NP_var){ if (Distributions[i] == "N") { SARes[,i] <- rnorm(n_sim, mean = Means[,i], sd = SDs[,i]) } else if (Distributions[i] == "B") { alpha <- ((1 - Means[,i]) / SDs[,i]^2 - 1 / Means[,i]) * Means[,i] ^ 2 beta <- alpha * (1 / Means[,i] - 1) SARes[,i] <- rbeta(n_sim, shape1 = alpha, shape2 = beta) } else if (Distributions[i] == "LN") { CV <- SDs[,i]/Means[,i] mlog <- log(Means[,i]/sqrt(1+CV^2)) #mean of log values slog <- sqrt(log(1+CV^2)) #sd of log values SARes [,i] <- rlnorm(n_sim, meanlog = mlog, sdlog = slog) } }

var_in <- SARes par_in <- cbind(var_in,fix_in) write.csv(par_in,"par_in.csv",row.names = FALSE)

}

#======#Model application----- #setup table model results SimRes <- matrix(NA, nrow = nrow(par_in), ncol = n_out*length(t_A)) colnames(SimRes) <- c(paste("blood_t",t_A,sep=""), paste("fat_t",t_A,sep=""),paste("liver_t", t_A, sep=""), paste("kidney_t",t_A,sep=""), paste("muscle_t",t_A,sep=""),paste("heart_t", t_A, sep=""), paste("brain_t",t_A,sep=""), paste("carcass_t",t_A,sep=""),paste("lung_t", t_A, sep=""))

#Input model function for (j in 1:nrow(par_in)) { print(paste0("Running loop: ", j)) print(paste("Current time: ", Sys.time()))

SimRes[j,] <- multi_tool( par_in = par_in[j,], #input data frame species = species, regime = regime, route = route, E_dose = E_dose, E_start = E_start, E_end = E_end, E_int = E_int, n_out = n_out, t_start = t_start, t_end = t_end, t_A = t_A, chem_fix = chem_fix) }

write.csv(SimRes,"results.csv",row.names = FALSE)

112

PBK model function with integrated QSAR for tissue:blood partition coefficients ### Model PBK (cattle, sheep, swine) generic ### # 01/2019 # R. Oldenkamp, S. Hoeks, L.S. Lautz #======# # probabilistic # # multi-compartment model function # # for farm animal species # #======# dMdt <- function(i, M, C, h, physnames, Qin, Qout, CO, PC, OW, E_bolus, E_cont, E_iv, kgastric, Qbolus, Qingest, Qiv, Qexhale, kabs, Vmax, Km, Cl, fbile, fbact, Qmilk) { #input via food dMfood <- E_bolus[i] * Qbolus + E_cont[i] * Qingest #input via iv dMiv <- E_iv[i] * Qiv #absorption over intestinal wall dMabs <- M[physnames=="lumen"] * kabs #delivery via arterial blood dMart <- Qin * C[physnames=="art"] dMart[physnames=="art"] <- -sum(dMart) #delivery to venous blood and passage through portal vein dMven <- Qout * (C / PC) dMart[physnames=="liver"] <- dMart[physnames=="liver"] - dMven[physnames=="intestine"] dMven[physnames=="liver"] <- dMven[physnames=="liver"] + Qout[physnames=="intestine"] * (C[physnames=="liver"] / PC[physnames=="liver"]) dMven[physnames=="ven"] <- -sum(dMven[physnames!="intestine"]) #metabolism an enterohepatic circulation and retransformation dMmet <- ifelse(!is.na(Vmax)&!is.na(Km),(Vmax*(C/PC))/(Km+(C/PC))*OW,0) dMmet[physnames=="liver"] <- ifelse(!is.na(Cl[physnames=="liver"]),Cl[physnames=="liver"]*C[physnames=="liver"]) dMmet[physnames=="lumen"] <- -dMmet[physnames=="liver"] * fbile * fbact dMmet[physnames=="metab"] <- -dMmet[physnames=="liver"] - dMmet[physnames=="lumen"] #transport over lung and exhalation dMexh <- -CO * C dMexh[physnames!="ven"] <- 0 dMexh[physnames=="art"] <- -dMexh[physnames=="ven"] * (CO / (CO + Qexhale * PC[physnames=="air"])) dMexh[physnames=="air"] <- -dMexh[physnames=="ven"] * ((Qexhale * PC[physnames=="air"])/(CO + Qexhale * PC[physnames=="air"])) #excretion to urine, milk, feces dMexc <- M[physnames=="lumen"] * kgastric dMexc[physnames=="kidney"] <- C[physnames=="kidney"] * Cl[physnames=="kidney"] dMexc[physnames=="mamgland"] <- C[physnames=="mamgland"] * Qmilk[physnames=="mamgland"] dMexc[physnames=="urine"] <- -dMexc[physnames=="kidney"] dMexc[physnames=="milk"] <- -dMexc[physnames=="mamgland"]

dMdt <- dMfood + dMiv + (dMabs + dMart + dMven + dMmet + dMexh + dMexc) * (h / 60)

M <- M + dMdt C <- ifelse(OW==0,0,M/OW)

return(list(M=M,C=C)) }

#Function to run model multi_tool <- function(par_in, species, regime, route, E_dose, E_start, E_end, E_int, n_out, t_start, t_end, t_A, chem_fix) {

#Physicochemical properties ---- MW <- par_in$MW Kow <- par_in$Kow #Kow S <- (0.001*par_in$S) / MW #solubility Temp <- 298 #Temperature 298 K = 25 degC R <- 8.314 #Gas constant (J/mol/K) Pv <- par_in$Pv #vapor pressure

#General physiology ---- BW <- par_in$BW #bodyweight (kg) CO <- par_in$CO #cardiac output (L/min)

Qexhale <- (0.499 * (BW^0.81)) #Ventilation mammalian (L/min)

#one matrix per relevant phys parameter (in right order) comp <- colnames(par_in[grep('fBW',colnames(par_in))]) comp <- comp[order(comp)] physnames <- c(substr(comp,start=5,stop=nchar(comp)),"ven","art","lumen","milk","urine","air","feces","metab","feed","iv")

#Organ weights fBW <- c(t(par_in[comp]),rep(0,11)) fBW[physnames=="ven"] <- 2/3 * fBW[physnames=="blood"] fBW[physnames=="art"] <- 1/3 * fBW[physnames=="blood"] fBW <- fBW[physnames!="blood"] OW <- BW*fBW/sum(fBW) #organ volums (L), based on normalized weight fractions and overall assumed density of 1 L/kg

#Blood flows Q comp <- gsub('fBW','fCO',comp) fCO <- c(t(par_in[comp]),rep(0,11)) fCO <- (fCO[physnames!="blood"]) Qin <- fCO*CO/sum(fCO) #blood flows (L/min), normalized to CO Qout <- -Qin

#Tissue-blood partitioning comp <- c(colnames(par_in[grep('_nl',colnames(par_in))]),colnames(par_in[grep('_pl',colnames(par_in))]),colnames(par_in[grep('_pr',colname s(par_in))]),colnames(par_in[grep('_w',colnames(par_in))])) 113

comp <- comp[order(comp)]

PCcomp <- comp[c(-grep('exp_',comp),-grep('int_',comp))] #all tissue composition -names fPC_old <- c(t(par_in[PCcomp])) #'old fractions' fPC_new <- fPC_old*c(par_in$nl,par_in$pl,par_in$pr,par_in$w) #new fractions

#normalisation to mean sum sumold <- rep(tapply(fPC_old,rep(seq(length(fPC_old)/4),each=4),sum),each=4) sumnew <- rep(tapply(fPC_new,rep(seq(length(fPC_old)/4),each=4),sum),each=4) fPC_new <- fPC_new*(sumold/sumnew)

PCexps <- comp[grep('exp_',comp)] fPC_exp <- c(t(par_in[PCexps])) #all QSAR exponents PCints <- comp[grep('int_',comp)] fPC_int <- c(t(par_in[PCints])) #all QSAR intercepts

tissnames <- c(substr(PCcomp[grep('_nl',PCcomp)],start=1,stop=nchar(PCcomp[grep('_nl',PCcomp)])-3)) PC1 <- fPC_int*fPC_new*Kow^fPC_exp PC1 <- tapply(fPC_int*fPC_new*Kow^fPC_exp,rep(seq(length(PC1)/4),each=4),sum) #partitioning coefficients tissue-water PC <- rep(PC1[tissnames=="blood"],length(physnames)) PC[match(tissnames,physnames)] <- PC1 PC[physnames=="air"] <- (Pv/S)/(R*Temp) PC[is.na(PC)] <- PC[physnames=="blood"] PC <- PC/PC[physnames=="blood"] PC <- PC[physnames!="blood"]

#Rates and flows (not chemical-dependent) physnames <- physnames[physnames!="blood"] Qingest <- rep(0,length(physnames)) Qiv <- rep(0,length(physnames)) Qbolus <- rep(0,length(physnames)) kgastric <- rep(0,length(physnames)) Qmilk <- rep(0,length(physnames)) fbile <- par_in$fbile #fraction of metabolites reentering lumen with bile (-)

Qingest[physnames=="lumen"] <- par_in$Qingest / (24*60) #ingestion rate kgfeed/min Qingest[physnames=="feed"] <- -par_in$Qingest / (24*60) Qiv[physnames=="ven"] <- 1 Qiv[physnames=="iv"] <- -1 Qbolus[physnames=="lumen"] <- 1 Qbolus[physnames=="feed"] <- -1 kgastric[physnames=="feces"] <- par_in$kgastric kgastric[physnames=="lumen"] <- -par_in$kgastric #gastric emptying rate constant (1/min) Qmilk[physnames=="mamgland"&Qin!=0] <- -par_in$Qmilk #milk production rate (L/min)

#TK parameters kabs <- rep(0,length(physnames)) Vmax <- rep(NA,length(physnames)) Km <- rep(NA,length(physnames)) Cl <- rep(0,length(physnames))

fbact <- par_in$fbact kabs[physnames=="intestine"] <- par_in$kabs kabs[physnames=="lumen"] <- -par_in$kabs Vmax[physnames=="liver"] <- -par_in$Vmax_tot Km[physnames=="liver"] <- par_in$Km_tot Cl[physnames=="liver"] <- -par_in$Cl_hepatic * BW Cl[physnames=="kidney"] <- -par_in$Cl_renal * BW

#Time vector for exposure ---- h <- 3 #stepsize in seconds t <- seq(3600*t_start,3600*t_end,by=h) #vector with timepoints (seconds) E_iv <- rep(0,times=length(t)) E_bolus <- rep(0,times=length(t)) E_cont <- rep(0,times=length(t))

if (route=="oral" & regime == "bolus") { E_bolus[t>=E_start*3600 & t<3600*E_end & t%%(3600*E_int)==0] <- (E_dose/MW)*BW } else if (route=="oral" & regime == "continuous") { E_cont[t>=E_start*3600 & t<3600*E_end] <- ((E_dose/MW)/par_in$Qingest)*BW #mmol/kgfeed } else { E_iv[t==E_start*3600] <- (E_dose/MW)*BW }

M <- rep(0,length(physnames)) #mass per compartment (mmol) C <- rep(0,length(physnames)) #concentration per compartment (mmol/L)

#Create output sheet ---- if (chem_fix) { results <- matrix(NA,nrow=1,ncol=n_out*length(t_A)) colnames(results) <- c(paste("blood_t",t_A,sep=""), paste("fat_t",t_A,sep=""),paste("liver_t", t_A, sep=""), paste("kidney_t",t_A,sep=""), paste("muscle_t",t_A,sep=""),paste("heart_t", t_A, sep=""), paste("brain_t",t_A,sep=""), paste("carcass_t",t_A,sep=""),paste("lung_t", t_A, sep="")) } else { results <- matrix(0,nrow=1,ncol=3) colnames(results) <- c("Cmax","tmax","AUC_24h") }

#Model simulation ---- for (i in 1:length(t)) { output <- dMdt(i=i, M=M, C=C, h=h, physnames=physnames, Qin=Qin, Qout=Qout, CO=CO, PC=PC, OW=OW, E_bolus=E_bolus, E_cont=E_cont, E_iv=E_iv, kgastric=kgastric, Qbolus=Qbolus, Qingest=Qingest, Qiv=Qiv, Qexhale=Qexhale, kabs=kabs, Vmax=Vmax, Km=Km, Cl=Cl, fbile=fbile, fbact=fbact, Qmilk=Qmilk) 114

M <- output$M C <- output$C

#Write to results for t_A if (chem_fix & any(t[i]==round(t_A*3600))) { results[,paste("blood_t",t[i]/3600,sep="")] <- C[physnames=="ven"] * MW results[,paste("fat_t",t[i]/3600,sep="")] <- C[physnames=="adipose"]* MW results[,paste("liver_t",t[i]/3600,sep="")] <- C[physnames=="liver"] * MW results[,paste("kidney_t",t[i]/3600,sep="")] <- C[physnames=="kidney"] * MW results[,paste("muscle_t",t[i]/3600,sep="")] <- C[physnames=="muscle"] * MW results[,paste("heart_t",t[i]/3600,sep="")] <- C[physnames=="heart"] * MW results[,paste("brain_t",t[i]/3600,sep="")] <- C[physnames=="brain"] * MW results[,paste("carcass_t",t[i]/3600,sep="")] <- C[physnames=="carcass"] * MW results[,paste("lung_t",t[i]/3600,sep="")] <- C[physnames=="lung"] * MW

} else if (!chem_fix) { results[,"tmax"] <- ifelse(C[physnames=="ven"]>results[,"Cmax"],t[i]/3600,results[,"tmax"]) results[,"Cmax"] <- ifelse(C[physnames=="ven"]>results[,"Cmax"],C[physnames=="ven"],results[,"Cmax"]) results[,"AUC_24h"] <- results[,"AUC_24h"] + C[physnames=="ven"] } }

return(results) }

115

APPENDIX B

This thesis research has been carried out under the RDM policy of the Institute for Water and Wetland Research, version [5 September 2019] accessed on https://www.ru.nl/iwwr/organisation/data- management/.

LS Lautz (2019). Physiological parameters for three farm animal species (cattle, sheep, and swine) as the basis for the development of generic physiologically based kinetic models. DOI: 10.5281/zenodo.3433224. [Chapter 3]

LS Lautz, S Hoeks, R Oldenkamp (2019). Development of generic physiologically based kinetic models for three farm animal species, cattle, sheep, and swine. DOI: 10.5281/zenodo.3432796. [Chapter 4]

LS Lautz, R Oldenkamp, J Hendriks, A Ragas, JL Dorne. 2020. Open source physiological data and physiological-based kinetic model code for the chicken (Gallus gallus domesticus). DOI: 10.5281/zenodo.3603114. [Chapter 5]

116

SUMMARY

Refinement of risk assessment procedures and including new approaches on risk assessment of chemicals, such as improvement of uncertainty factors to determine safe level of exposure as well as in silico models incorporating kinetics, like one compartment models and physiologically based kinetic (PBK) models, are methods to extrapolate between various animal species to reduce uncertainties in current practices. To move towards the integration of new approaches in chemical risk assessment in the areas of human health, animal health and environment, several projects were launched by the European Food Safety Authority (EFSA). These projects involve the integration of kinetic tools, the modelling of population dynamics of aquatic and terrestrial organisms, and the modelling of human variability in kinetic and dynamic processes. The main aim of the thesis was to develop and apply generic PBK models for the prediction of chemical kinetics and tissue concentrations in farm animals and provide recommendations to refine and adapt guidance documents in animal health risk assessment. Such models particularly aim to improve methods for exposure assessment based on evaluation of the internal dose. In Chapter 2, PBK models in the 10 most common species of farm animals were identified through an extensive literature search. This resulted in 39 PBK models, which were critically assessed using the WHO criteria for model evaluation. The available PBK models for farm animal species have generally been developed for specific purposes, often focusing on a specific substance in a given species. Hence, the reliability of these PBK models is hard to assess and their potential for use in chemical risk assessment is limited. In a risk assessment context, future PBK models for farm animals should include a more generic and flexible model structure, use input parameters independent on calibration and include assessment tools to assess model performance. Development and application of PBK models for farm animal species would furthermore benefit from the setup of structured databases providing data on physiological and chemical-specific parameters. Therefore, in Chapter 3 and 5, the results of a structured data collection of anatomical and physiological parameters in four farm animal species (swine, cattle, sheep and chicken) are presented. We performed an extensive literature search and meta-analyses to quantify intra-species variability and associated uncertainty of the parameters. Parameters were collected for organ weights and blood flows in all available breeds from 170 scientific publications for cattle, sheep, swine, and chicken, respectively. The parameter values showed a large intra-species variation. Overall, the parameter values and associated variability provide reference values which can be used as input for generic PBK models in these species. In the next step, generic PBK models were developed for cattle, swine, and sheep focused on chemicals which were only renally excreted (Chapter 4), while the generic chicken model included metabolism

117

(Chapter 5). The PBK model performance for cattle, sheep, and swine is illustrated by predicting tissue concentrations of melamine and oxytetracycline and validated through comparison with measured data. Overall, model predictions were reliable with 71% of predictions within a 3-fold of the measured data for all three species and only 6% of predictions were outside a 10-fold of the measured data. Predictions within a 3-fold change were best for cattle, followed by sheep, and swine (82%, 76%, and 63%). The chicken PBK model was implemented for seven compounds (with log Kow range -1.37-6.2) to quantitatively link external dose and internal dose for risk assessment of chemicals. Model predictions were compared to measured data according to dataset specific exposure scenarios. Globally, 71% of the model predictions were within a 3-fold change of the measured data for chicken and only 7% of the PBK predictions were outside a 10-fold change. The global sensitivity analysis showed that body weight and cardiac output were the most sensitive parameters for the four developed PBK models. Species-specific PBK models can be applied in a generic way for various chemicals as demonstrated in chapter 4 and 5. However, data gaps and recommendations for future work related to PBK model input parameters and model validation can be highlighted (Chapter 6). Since interspecies differences in metabolism impact on the fate of a wide range of chemicals, a key step forward is the introduction of species-specific information on transporters and metabolism including expression and activities. Given the trend to reduce animal testing in chemical risk assessment, and the need to derive safe exposure levels from internal points of departure (based on in silico models or in vitro toxicity tests), the challenge for the PBK modelling community is to base the model parameterisation increasingly or entirely on ADME properties derived from non-animal studies, with limited or no availability of in vivo kinetic data. This should be addressed in the future through data collection from the literature and bioinformatic analysis of genome sequences, proteomic expression and the design of in vitro assays in farm animal hepatocytes and intestinal cells to depict such metabolism. These approaches will allow the refinement of PBK models in farm animals and QSAR models allowing for inter-species differences in metabolism and toxicity, moving further towards next generation risk assessment of chemicals.

118

SAMENVATTING

Het verfijnen van risicobeoordeling procedures en het toevoegen van nieuwe benaderingen aan de risicobeoordeling van chemicaliën, zoals het verbeteren van onzekerheidsfactoren om veilige blootstellingsnormen te kunnen bepalen en in silico modellen die kinetiek meenemen, zoals één compartiment modellen en fysiologisch gebaseerde farmacokinetische (PBK: physiologically based kinetic) modellen, zijn methodes om te extrapoleren tussen verschillende diersoorten om onzekerheden te verminderen. Om de integratie van nieuwe benaderingen in chemische risicobeoordeling in humane gezondheid, diergezondheid en milieu te stimuleren, zijn verschillende projecten uitgeschreven door de Europese voedselveiligheid autoriteit (EFSA: European Food Safety Authority). Deze projecten betrekken de integratie van kinetische modellen, het modelleren van populatiedynamiek van water- en landdieren en het modelleren van humane variabiliteit in kinetiek- en dynamiek processen. Het hoofddoel van deze thesis was om generieke PBK modellen te ontwikkelen en toe te passen, om chemische kinetiek en weefselconcentraties te kunnen voorspellen in boerderijdieren en om aanbevelingen te geven om rapporten in diergezondheidsrisicobeoordeling te verfijnen en aan te passen. Zulke modellen hebben vooral als doel om methodes voor blootstellingsbeoordeling te verbeteren, gebaseerd op de evaluatie van de interne dosis. In hoofdstuk 2 werden PBK modellen in de 10 meest voorkomende boerderijdieren verzameld met behulp van een uitgebreid literatuuronderzoek. Dit heeft geresulteerd in 39 PBK modellen, die kritisch werden beoordeeld met behulp van de wereldgezondheidsorganisatie (WHO: World Health Organization) criteria voor modelevaluatie. De beschikbare PBK modellen voor boerderijdieren zijn in het algemeen ontwikkeld met een specifiek doel: ze focussen meestal op een specifieke stof in een bepaald dier. Daardoor is de betrouwbaarheid van deze PBK modellen moeilijk te beoordelen en is het potentiële gebruik van deze modellen in chemische risicobeoordeling gelimiteerd. Binnen de context van risicobeoordeling zullen toekomstige PBK modellen voor boerderijdieren een meer generieke en flexibele structuur moeten hebben, input parameters moeten gebruiken die onafhankelijk zijn van kalibratie en beoordelingscriteria moeten bevatten om de prestaties van het model te kunnen beoordelen. De ontwikkeling en toepassing van PBK modellen voor boerderijdieren zou ook baat kunnen hebben bij het opzetten van gestructureerde databases, die data leveren over fysiologische en stof-specifieke parameters. De resultaten van een gestructureerde dataverzameling over de anatomische en fysiologische parameters in vier boerderijdieren (varkens, koeien, schapen en kippen) worden daarom weergegeven in hoofdstukken 3 en 5. We hebben een uitgebreid literatuuronderzoek en meta-analyses uitgevoerd om verschillen tussen dieren en de bijbehorende onzekerheid van de parameters te

119

kwantificeren. Parameters voor orgaangewicht en bloedstroom in alle beschikbare rassen voor koeien, schapen, varkens en kippen werden verzameld uit 170 wetenschappelijke publicaties. De parameter waardes lieten grote verschillen zien binnen diersoorten. Over het algemeen leveren de parameter waardes en bijbehorende variabiliteit-referentiewaardes, die gebruikt kunnen worden als input voor generieke PBK modellen in deze diersoorten. In de volgende stap werden generieke PBK modellen ontwikkeld voor koeien, varkens en schapen, met een focus op stoffen die alleen via de nieren worden uitgescheiden (hoofdstuk 4) en een model voor kippen dat metabolisme includeert (hoofdstuk 5). Het PBK model voor koeien, schapen en varkens voorspelt weefselconcentraties van melamine en oxytetracycline en is gevalideerd door te vergelijken met gemeten data. Over het algemeen waren de model voorspellingen betrouwbaar: 71% van de voorspellingen lagen binnen drievoud van de gemeten data voor alle drie de soorten en maar 6% van de voorspellingen lagen buiten tienvoud van de gemeten data. Voorspellingen binnen de 3 keer verandering waren het beste voor koeien, gevolgd door schapen en varkens (82%, 76% en 63%). Het kip PBK model is geïmplementeerd voor zeven stoffen (met een log Kow range -1.37-6.2) om kwantitatief externe dosis en interne dosis te linken voor de risicobeoordeling van chemicaliën. Model voorspellingen zijn vergeleken met gemeten data volgens specifieke blootstellingsscenario’s. Globaal gezien lagen voor kippen 71% van de voorspellingen binnen drievoud van de gemeten data en maar 7% van de voorspellingen lagen buiten tienvoud van de gemeten data. De globale gevoeligheidsanalyse liet zien dat lichaamsgewicht en hartminuutvolume de meest gevoelige parameters waren voor de vier ontwikkelde PBK modellen. Soort-specifieke PBK modellen kunnen op een generieke wijze toegepast worden voor verschillende stoffen, zoals aangetoond in hoofdstuk 4 en 5. Echter, ontbrekende data en aanbevelingen voor toekomstig werk gerelateerd aan PBK model input parameters en model validatie kunnen worden aangekaart (hoofdstuk 6). Omdat verschillen tussen diersoorten in metabolisme invloed hebben op het lot van veel verschillende stoffen, is het introduceren van soort-specifieke informatie over transporteiwitten en metabolisme, inclusief expressie en activiteit, een grote stap voorwaarts. Gezien de trend richting het verminderen van dierproeven in chemische risicobeoordeling en de noodzaak om veilige blootstellingswaarden af te leiden van interne vertrekpunten (gebaseerd op in silico modellen of in vitro toxiciteit testen), is het de uitdaging voor de PBK- modeleergemeenschap om de model parametrisatie steeds meer, of geheel, te baseren op ADME eigenschappen niet afgeleid van dierproeven, met gelimiteerde of geen beschikbaarheid van in vivo kinetische data. Dit moet in de toekomst aangepakt worden door data uit de literatuur te verzamelen en bioinformatische analyses van genoom sequenties, proteoom expressie en het ontwerp van in vitro assays in boerderijdier-hepatocyten en -darmcellen te doen, om dit metabolisme te ontrafelen. Deze 120

benaderingen zullen zorgen voor de verfijning van PBK modellen in boerderijdieren en voor QSAR modellen, die zullen zorgen voor het karakteriseren van verschillen tussen soorten in metabolisme en toxiciteit, op weg naar de volgende generatie risicobeoordeling van stoffen.

121

ACKNOWLEDGMENTS

Von Bad Oeynhausen meinem Heimatort, Thanks also to my supervisors, Zog ich bald nach Nimwegen fort. Who are very critical analysers. Und an dieser Universität, Fand ich dann die Stabilität, Jan, Exzellent ist dieses Haus, Bedanken wil ik mij heel graag, Es bildet höchste Geister aus. Dat geldt niet enkel voor vandaag, Want het leven is steeds goed, Biology was my discipline, Door al jouw steun, kracht, en moed. I mastered without caffeine. Bedankt voor al je wijze woorden, I studied medicine in-between Wat was het zonder die geworden. But my interest was more in melamine. Toxicology was my master choice Ad, Which I finished with a lot of joice. Ik ging je af en toe vervelen, Want je moest meestal als laatst oordelen, After my study I continued with the animal Over al mijn papers hier, For my supervisors best practical. Hoe moesten ze anders goed op papier. Testing nearly every chemical, Bedankt voor alles wat je deed, With this choice I was very flexible. Daarom is mijn boek ook zo breed.

Pesticides and anti-inflammatory, Jean-Lou, One from every category, You have gone over and beyond, Feed additives and antibiotics, You had always a good respond! In rare cases even narcotics. You were always the best to chatter, Day and night, that does never matter! From sediment and water, You had faith in me a lot, Into blood from the trotter, Even you knew that I did not. All the species I was looking at, The project included even the cat! Silke, Richard, Arne, Tom, Heinke, Ingrid, The chemical risk was calculated Elisabeth, Ulrike, und andere: And the results in papers stated. Fliegen heißt nur aufwärts fallen, Doch ich stehe hier vor euch allen, However, doing the PhD alone is not nice, Der Grad der Entscheidung war oft schmal, A lot of people add the extra spice. Denn gar nichts kommt zum zweiten Mal. When deadlines seem to pile, Summa summarum steht unterm Strich, When there is no reason to smile, Da liegt das Buch und es gibt wieder mich. Colleagues, friends and a culture whack, Still keep you on the right track. In Memoriam: Hans und Manfred Ihr beiden wäret stolz auf mich, s ’Avonds samen naar cultuur café, Leider ist so vieles endlich, Was altijd een heel goed idee! In allen Lebenslagen kanntet ihr euch aus, Samen gezellig op thee en koffie eruit, Und jetzt seid ihr wieder einen Schritt voraus. Dat was ook een goed besluit.

122

After taking so much time, I want to finish with one last rhyme: The decisions at the Radboud University In the end led to my PhD. This was the final academical test, So I want to finish with: ”Hora est”.

123

CURRICULUM VITAE

About the Author

Leonie Sophie Lautz was born on the 5th of March 1990 in Bad Oeynhausen, Germany. She did her undergraduate degree in Biology in 2012 and continued with a master education in Human and Environmental Risk Assessment. She obtained her master degree in 2015. Leonie was offered a job as researcher for eight months at the department of Environmental Science, which resulted in one first author publication. After this, she continued as a researcher working on animal and environmental risk assessment at the same department. In 2017 she got involved in a PhD program at Radboud University and research under the supervision of Prof. dr. A.M.J. Ragas and Prof. dr. A.J. Hendriks. Her work has resulted in numerous reports for the European Food Safety Authority and several first author publications. From March 2019, she continued her career as a researcher at the French Agency for Food, Environmental and Occupational Health & Safety in Paris, France.

List of publications

K Darney, JLCM Dorne, E Testai, LS Lautz, 2020. Human variability in carboxylesterases and implementation in physiologically based kinetic models. Toxicology Letters, submitted.

LS Lautz, MZ Jeddi, C Nebbia, JLCM Dorne, 2020. No truce between cats and rats: Comparative assessment of pharmacokinetic differences, implications for chemical risk assessment and future direction. Archives of Toxicology, submitted.

N Quignot, WA Wiecek, JLCM Dorne, LS Lautz, B Amzal, 2020. Inter-phenotypic differences in CYP2C9 and CYP2C19 metabolism: A Bayesian meta-analysis and meta-regression of human population variability in kinetics and application to chemical risk assessment. Archives of Toxicology, submitted.

EEJ Kasteel, K Darney, NI Kramer, JLCM Dorne, LS Lautz, 2020. Human variability in isoform-specific UDP- glucuronosyltransferases metabolism: from markers of acute and chronic exposure to polymorphisms and uncertainty factors. Archives of Toxicology, accepted.

N Thunnissen, LS Lautz, T van Schaik, AJ Hendriks, 2020. Ecological risks of imidacloprid to aquatic species in the Netherlands using measurements from monitoring data and species sensitivity distributions. Chemosphere [254]:126604.

K Darney, L Turco, FM Buratti, E Di Consiglio, S Vichi, AC Roudot, C Bechaux, E Testai, JLCM Dorne, LS Lautz, 2020. Human variability in influx and efflux transporters in relation to uncertainty factors for chemical risk assessment. Food and Chemical Toxicology [140]:111305.

K Darney, EEJ Kasteel, FM Buratti, L Turco, S Vichi, C Béchaux, AC Roudot, NI Kramer, E Testai, JLCM Dorne, E Di Consiglio, LS Lautz, 2020. Bayesian meta-analysis of inter-phenotypic differences in human serum paraoxonase-1 activity for chemical risk assessment. Environmental International [138]: 105609.

124

LS Lautz, C Nebbia, S Hoeks, R. Oldenkamp, AJ Hendriks, AMJ Ragas, JLCM Dorne, 2020. An open source physiologically based kinetic model for the chicken (Gallus gallus domesticus): Calibration and validation for the prediction of residues in tissues and eggs. Environmental International [136]: 105488.

LS Lautz, JLCM Dorne, R Oldenkamp, AJ Hendriks, AMJ Ragas, 2020. Generic physiologically based kinetic modelling for farm animals: Part I. Data collection of physiological parameters in swine, cattle and sheep. Toxicology Letters [319]:95-101.

LS Lautz, S Hoeks, R Oldenkamp, AJ Hendriks, JLCM Dorne, AMJ Ragas, 2020. Generic physiologically based kinetic modelling for farm animals: Part II. Predicting tissue concentrations of chemicals in swine, cattle, and sheep. Toxicology Letters [318]:50-56.

LS Lautz, R Oldenkamp, JLCM Dorne, AMJ Ragas, 2019. Physiologically based kinetic models for farm animals: Critical review of published models and future perspectives for their use in chemical risk assessment. Toxicology in Vitro [60]:61-70.

JLCM Dorne, B Amzal, N Quignot, W Wiecek, C Bechaux, K Darney, A Grech, C Brochot, R Beaudouin, F Bois, A Ragas, L Lautz, R Oldenkamp, N Kramer, E Kasteel, E Testai, S Vichi, E Di Consiglio, L Turco, F Buratti, C Tebby, J Cortinas-Abrahantes, A Paini, JC Madden, T Robinson, 2018. Developing TK databases and tools to support food safety assessment. Toxicology Letters [295S1]: S5-S6.

JLCM Dorne, B Amzal, N Quignot, W Wiecek, A Grech, C Brochot, R Beaudouin, F Bois, A Ragas, L Lautz, R Oldenkamp, C Bechaux, K Darney, N Kramer, E Kasteel, E Testai, et al., 2018. Reconnecting exposure, toxicokinetics and toxicity in food safety: OpenFoodTox and TKplate for human health, animal health and ecological risk assessment. Toxicology Letters [295S1]: S29.

LS Lautz, J Struijs, TM Nolte, D van de Meent, T Breure, E van der Grinten, R van Zelm, 2017. Evaluation of SimpleTreat 4.0 for pharmaceuticals and personal care products. Chemosphere [168]:870–876.

Reports

OECD 2020, Draft: OECD guidance document in the characterisation, validation, and reporting of PBK models. Series on Testing & Assessment, OECD Publishing, Paris.

E Testai, C Bechaux, FM Buratti, K Darney, E Di Consiglio, EEJ Kasteel, NI Kramer, LS Lautz, L Turco, S Vichi, 2020. Deliverable 6: Modelling human variability in toxicokinetic and toxicodynamic processes using Bayesian meta-analysis, physiologically based modelling and in vitro systems. EFSA supporting publication 2020.

E Testai, C Bechaux, FM Buratti, K Darney, E Di Consiglio, EEJ Kasteel, NI Kramer, LS Lautz, L Turco, S Vichi, 2019. Deliverable 5: Modelling human variability in toxicokinetic and toxicodynamic processes using Bayesian meta-analysis, physiologically based modelling and in vitro systems. EFSA supporting publication 2019.

125

N Quignot, W Wiecek, B Amzal, LS Lautz, R Oldenkamp, AMJ Ragas, A Grech, R Beaudouin, C Brochot, 2018. Integrating toxicokinetics in chemical risk assessment: application to human, animal and environmental risk assessment. EFSA supporting publication 2018.

LS Lautz, R Oldenkamp, AMJ Ragas, 2017. Data collection of TK variables to calibrate and develop “generic” TK tools for animal risk assessment of multiple chemicals. EFSA supporting publication 2017.

LS Lautz, R Oldenkamp, AMJ Ragas, 2017. Data collection of biological, physiological and toxicological variables calibrate and develop “generic” TK tools for animal health risk assessment of single chemicals. EFSA supporting publication 2017.

R Oldenkamp, LS Lautz, AMJ Ragas, 2016. Availability of tools and models for the integration of TK in chemical risk assessment for animal health. EFSA supporting publication 2016.

Award

SOT Exposure Specialty Section Best Abstract Award 2020. Title: Alice in Numberland: Mathematics of variability and uncertainty – Can kinetics help us understand variability in toxicity?

Blablablaba;balnalnla

126

127

Institute for Water and Wetland Research (IWWR), Radboud University