Potential hazard and risks of substances, pesticides and pharmaceuticals are traditionally addressed with animal experiments, for which numerous standardized, recognized and harmonized methods are available. In accordance with the paradigm of the 21st century concerning toxicological risk assessment and the 3R principle (“Reduction, Replacement, Refinement”), i.e. replacing animal experiments, in vitro and in silico based methods are increasingly being used to assess concentration-effect relationships. To derive risk assessments from results of in vitro methods, relevant human and animal doses should be extrapolated from concentrations in cell-based assays (in vitro-in vivo extrapolation, IVIVE). Usually, the effective concentration in vitro is based on a pre-defined amount of substance per unit cell culture volume, the so-called nominal concentration. However, factors such as binding to proteins in the cell culture medium, to adsorption to the cell culture vessel, and processes such as volatilization and enzymatic degradation contribute to the reduction of the concentration of a test substance in an in vitro test system. This hampers the extrapolation to relevant in vivo doses. The actual concentration of test substances in in vitro systems was examined in this work. For this purpose, analytical methods were developed, validated and used to determine actual in vitro concentrations of twelve test substances. Finally, the obtained effective in vitro concentrations were used to assess the endocrine disrupting potential of substances based on in vitro study results. The work on determining actual in vitro concentrations was divided into three parts: (1) Binding to proteins, which depends on the amount of protein in the medium, has a major influence on the unbound concentration of a substance in an in vitro test system. Likewise, substances bin to proteins, such as albumin, in blood plasma in vivo. Since the free fraction (fu) of a substance might be responsible for an effect, the fraction bound to proteins of the test substances in human plasma was determined. Applied methods for the investigation of protein binding are the rapid equilibrium dialysis (RED), ultrafiltration (UF) and ultracentrifugation (UC), which were checked for repeatability, accuracy and robustness in this work. In comparison to published literature data, more reliable fu ten test substances were obtained using RED with recovery values of 70 – 130 %. Physico-chemical properties of the substances, e.g. the octanol-water partition coefficient, are significant when selecting the appropriate separation method. The methods UF and UC are also recommended for polar substances (fu >70 % and logPow < 2) since nonspecific binding to devices are not significant, while the protein binding of lipophilic substances (logPow of 3.6 - 6.84) should be rather determined using RED based on comparable derived fu with literature data and suitable recovery values (68.2 – 118.1 %). Recoveries below 50 % and higher fu in comparison to published reference data were derived for moderately lipophilic to lipophilic substances when using UC of UF.
(2) Methods for the quantification of test substances in the culture medium and cell lysates and as well as different sample preparations for the cell lysates were developed and the impact of sampling preparation on the measured cellular concentrations evaluated. A diffusion-based model for predicting the concentration in the compartments of the in vitro test system was used and the experimental data utilized to evaluate the model. Based on the results obtained from the protein binding studies, analytical concentrations of the test substances at 6, 24 and 48 h of incubation were corrected for the experimentally determined free fraction. The experiment was carried out with the Balb/c 3T3 cell line assuming that the cellular uptake of substances is based on diffusion due to the low transporter expression in this cell line. With regards to the mass balances and the measured data, minor effects on the concentration of the test substances with little or no protein binding in the culture medium were determined. On the other hand, substances with a defined mechanism of action and particular cellular targets or high octanol-water partition coefficient (TAM) showed deviations from the nominal concentration or analytically measurable concentration in the medium and thus resulted in a shift of the equilibrium towards the cell compartment. Cellular concentrations are up to 2 – 274 times higher than the nominal concentration and thus confirmed the discrepancies between medium and cell compartment. Computational and experimental data on the total and free concentration of the test substances in the culture medium agreed for eleven substances for which accordance below a factor of 2 were observed. In contrast, up to 4-fold higher concentrations were predicted for the Balb/c 3T3 cells. Discrepancies could be explained by the lack of involvement of target compartments such as cell organelles where substance specific mechanisms may be triggered (e.g. lysosomes as target organelle for tamoxifen), metabolic activation, transporter-mediated uptake and the degree of ionization of the molecules in the model. The model can be expanded by including these parameters in future work and needs to be validated again. The simple model can be used as a first screening method to assess the behavior of a substance in diffusion-based test systems.
(3) Finally the approaches from previous work, the investigation on plasma protein binding and in vitro dosimetry, were combined with a reverse dosimetry approach and applied to in vitro effect concentrations in YES/YAS and steroidogenesis assay to determine oral doses in rats with regards to endocrine effects. For this purpose, concentrations of seven test substances (APAP, BPA, CAF, FEN, FLU, GEN, KET) were quantified in culture media and yeasts and human H295R adenocarcinoma cells using validated analytical methods. These serve as the POD for a QIVIVE based on physiologically based toxicokinetic model (PBTK model) to calculate an external, oral in vitro dose. This extrapolated in vivo does was and to compare the data obtained with published results on in vivo doses based on nominal in vitro concentrations. Increasing concentrations in cells has been found for for 4/7 test substances (BPA, FEN, FLU, GEN) which are known to have an affinity to the estrogenic/androgenic receptor or interfere with steroidogenesis respectively. An equilibrium between the medium and cell compartments was observed in the negative controls (APAP and CAF) which are known to not induce effects in the YES-/YAS- and steroidogenesis assay. Using cellular concentration, oral doses for 6/7 compounds were correctly calculated within a factor of 10, providing results with higher correlation to in vivo data than the estimated LOEL based on total and medium concentrations. However, the study highlights the analytical difficulties that arise when determining cellular concentrations over multiple time points (3, 6 and 12 h). Accordingly, no statements could be made on the actual substance uptake over time in the YES-/YAS-assay (APAP, BPA, CAF, FLU, GEN). Nominal as well as total concentrations of substances can serve as POD if there is no affinity to cellular targets, low protein binding (less than 10%) and no metabolic activation by the cell lines are present. Future studies may include other endocrine disruptors and promote the use of reverse dosimetry in relation to other toxicological endpoints.
