Replacing animal experiments in developmental toxicity testing of phenols by combining in vitro assays with physiologically based kinetic (PBK) modelling

Many efforts have been undertaken over the past decades to develop in vitro tests for a wide range of toxicological endpoints as an alternative to animal testing. The principle application of in vitro toxicity assays still lies in the hazard assessment and the prioritisation of chemicals for further toxicity testing. The in vitro toxicity outcomes are hardly used in quantitative risk assessment of chemicals, for example to predict health-based guidance values like an acceptable or tolerable daily intake (ADI or TDI). An important reason for this limited use is that an in vitro toxicity assay cannot provide a complete in vivo dose-response curve from which a point of departure (PoD) for risk assessment, like a no observed adverse effect level (NOAEL) or a 95% lower confidence limit of the benchmark dose (BMDL) can be derived. Translation of in vitro toxicity data to the in vivo situation is of paramount importance to overcome these constraints. Physiologically based kinetic (PBK) models, which mathematically describe the absorption, distribution, metabolism and excretion (ADME) of a compound in an organism, can be used to extrapolate in vitro derived toxicity data to in vivo toxicity values. The aim of the present thesis was to demonstrate the potential of a combined in vitro PBK- based approach to translate in vitro toxicity data to in vivo toxicity values for rat and human predicting in vivo dose-response curves that allow definition of a PoD for risk assessment. This translation was carried out by applying PBK-based reverse dosimetry, using in silico and in vitro defined kinetic parameters, and by combining the in vitro PBK approach with Monte Carlo simulations to assess interindividual variability. The present thesis focused on the prediction of in vivo developmental toxicity. The ES-D3 differentiation assay of the embryonic stem cell test (EST) was selected as a representative in vitro assay for this toxicity endpoint. Phenol and a series of para-substituted phenols being p-fluorophenol, p-heptyloxyphenol, p-mercaptophenol and p-methylketophenol were chosen as model compounds. Chapter 1 introduces the methods and concepts that are used in the present thesis, namely in vitro developmental toxicity testing, PBK modelling and Monte Carlo simulations. Also the general outline of the in vitro PBK approach using reverse dosimetry is presented. Furthermore, a brief description of the ADME properties of each phenol that is included in the present thesis is given. In chapter 2 the applicability of the EST as an alternative assay to predict developmental toxicity for the selected phenols was evaluated. The embryotoxic potencies of the phenols obtained with the EST were compared with potency data reported in the literature for the WEC assay, providing a similar potency ranking. The results suggest that the EST can identify the intrinsic embryotoxic potency of the phenols. However, when comparing the potency ranking of the EST to in vivo reported potency data on developmental toxicity of the phenols, it was concluded that the EST was not able to correctly rank the phenols according these in vivo potency data. Only phenol was correctly ranked, being the least potent congener, while the relative toxic potency of p-heptyloxyphenol was overestimated in the EST, when compared to the in vivo derived relative potency value. It was hypothesised that combining in vitro toxicity data with kinetic modelling could improve the in vivo toxicity prediction. In chapter 3 we combined in vitro toxicity data from the EST with PBK-based modelling for the compound phenol. For this purpose, PBK models were developed that describe the ADME of phenol in rat and human. The PBK models were parameterised with in vitro and in silico data and data taken from the literature. Evaluation of the PBK models pointed out that the oral absorption coefficient and phenol glucuronidation are influential model parameters for the predicted phenol plasma concentration. PBK-based reverse dosimetry was applied to translate the in vitro effect concentrations obtained from the EST to in vivo effective dose levels from which an in vivo dose-response curve was generated allowing the prediction of a PoD for risk assessment. The predicted PoD thus obtained for the rat lies within the range of PoDs for phenol obtained from in vivo developmental toxicity data from the literature. The results demonstrate the potential of PBK-based reverse dosimetry to translate in vitro toxicity data to in vivo toxicity values that can be used for quantitative hazard and risk assessment purposes, like deriving safe exposure levels. In chapter 4 it was explored whether PBK-based reverse dosimetry could also successfully extrapolate in vitro toxicity data obtained in the EST to in vivo developmental toxicity values for a series of para-substituted phenols. It was of interest if this approach could overcome the differences observed between the in vitro and in vivo derived relative embryotoxic potencies for the phenols, especially for p-heptyloxyphenol. For each phenol, a PBK model for the rat was developed. In silico methods were selected and applied to predict input values for a number of PBK model parameters. In vitro assays were performed to predict metabolism of the test compounds and transport across the intestinal and the placental barrier. After completing the PBK models, reverse dosimetry was applied generating in vivo developmental toxicity data for each of the phenols from which a PoD was derived. These PoDs appeared to differ less than 3.8-fold from PoDs derived for the phenols from in vivo toxicity data available in the literature. The large difference that was observed between the in vitro and the in vivo relative embryotoxic potency for the phenols, especially for p-heptyloxyphenol, was reduced to a large extent after applying the in vitro PBK-based reverse dosimetry approach. Our modelling approach pointed towards metabolism by glucuronidation as the most important factor underlying the relative low toxic potency for p-heptyloxyphenol in vivo compared to the relatively high observed embryotoxic potency in vitro. Altogether, the results showed the feasibility of the PBK-based reverse dosimetry approach to properly extrapolate in vitro toxicity outcomes to in vivo toxicity values. In chapter 5 we demonstrated a modelling approach that integrates in vitro toxicity data, PBK modelling and Monte Carlo simulations to predict the effects of human interindividual variation in kinetics for phenol induced development toxicity. Incubations with recombinant uridine diphosphate glucuronyltransferases (UDPGTs) revealed that the enzymes UGT1A6 and UGT1A9 were mainly responsible for glucuronidation of phenol. Kinetic constants describing the glucuronidation of phenol by UGT1A6 and UGT1A9 were included in the PBK model of phenol together with information about the variation in activity of these UGTs in the population and information on the interindividual variation for the oral absorption coefficient. Monte Carlo simulations in connection with PBK modelling provided a population distribution of phenol plasma concentrations, from which chemical specific adjustment factors (CSAFs) were calculated that cover interindividual human kinetic variation. It was concluded that the default safety factor of 3.16 applied for interindividual human kinetic differences in the safety assessment of chemicals is adequate protective, given that the predicted CSAF lies within this value. Dividing the dose-response curve data obtained with in vitro PBK-based reverse dosimetry with the CSAF provided a dose-response curve that reflects the consequences of the interindividual variability in phenol kinetics for the developmental toxicity of phenol. Altogether, this chapter presented a non-animal based approach to predict interindividual human variation in sensitivity to phenol induced developmental toxicity. Chapter 6 discusses prominent aspects of the in vitro PBK-based reverse dosimetry approach. The items discussed encompass the selection of the EST as an alternative in vitro assay to predict developmental toxicity, different considerations regarding PBK modelling, the use of different kinetic reverse dosimetry approaches to perform in vitro toxicity to in vivo toxicity extrapolations (IVIVE) and the possibility to take interindividual variations into account. Finally, future perspectives related to the application of the PBK-based reverse dosimetry to predict in vivo toxicity values from in vitro toxicity data are outlined.  In conclusion, this thesis demonstrated that the approach of combining in vitro toxicity data with PBK-based reverse dosimetry has the potential to overcome a major bottleneck in the hazard and risk assessment of chemicals, since it was not only able to improve in vitro toxic potency estimates but more importantly it provided adequate quantitative toxicity estimates such as a PoD. Altogether this thesis contributes to the replacement of apical endpoints in animal studies for the safety testing of chemicals, towards an approach that is cheaper, non-animal based and more directly relevant for humans.