Search

This thesis describes the establishment of a human 3D in vitro model for the prediction of teratogenic substances in the early phase of pharmaceutical compound develop-ment. Initially, a three-dimensional cell culture model was established to depict devel-opmental processes on a transcriptomic level during the spontaneous differentiation of human induced pluripotent stem cells (hiPSC) into all three germ layers. In a first proof-of-concept study, the validity of the in vitro model was assessed using a specific panel of 96 embryonic development markers. The differential gene expres-sion was induced by eight different reference compounds with known teratogenic pro-files in various concentrations, measured with qPCR. Subsequently, a total of 45 ref-erence substances were tested and evaluated by RNA sequencing with an expanded gene panel of 1,215 markers. Various machine-learning models and statistical anal-yses evaluated the level of gene expression regulation to classify the reference sub-stances with regard to their teratogenicity and cytotoxicity. The larger selection of genes, which represent molecular signaling pathways, could not predict teratogenicity better compared to the specific selection of early developmental markers. Ultimately, the assay was validated with regard to its predictive power in terms of specificity, sen-sitivity and accuracy and it was found that it is superior to the conventional mouse model (mEST) especially in sensitivity. Another concept study with gastruloids was carried out to investigate teratogenic induced morphological and genetic changes with consideration of phenotypic effects. Data from both the described human 3D in vitro assay and the established mEST were used to compare the different models. In summary, this human in vitro model will not only serve as an improved predictive tool in the future, but will also contribute to a better characterization of molecular pro-cesses with respect to human-relevant teratogenicity and possibly reduce animal stud-ies. Due to the correspondingly higher throughput of the in vitro model, it is suitable for an extensive early selection of preclinical compounds up to industrial scale and will be used for routine application in the future. published