| dc.description.abstracteng | The intention of this dissertation is to gain deeper insight into mechanisms of drug action on the genomic and proteomic level in HER2-positive breast cancer cell lines with different resistance phenotypes. In the era of personalized cancer therapy such insights are essential for the development and improvement of (combined) targeted therapeutics and their application in individual therapy approaches.
On the proteomic level, cell line specific Boolean networks were reconstructed based on RPPA time course data to model signal transduction between the measured proteins under drug treatment. The drugs of interest were trastuzumab and pertuzumab which target HER2 and its dimerisation, respectively. In addition, erlotinib was analysed which targets EGFR. The time series covered a range up to 60 minutes and 30 hours, respectively. These models were used to simulate the effect of different drug combinations on the different cell lines. Therefore, the activity states of selected (phospho-)proteins in the PI3K and MAPK signalling pathways were computed. The simulation results were mostly confirmed by the actual measurement data, but still they have to be distinguished, as homoeostasis does not necessarily have to be reached after the measured maximum time points. The simulations further revealed a more diverse drug response in the short-term measurements than in the long-term measurements. This underlines the importance of early drug intervention at the top level layer of the signalling network. For the trastuzumab resistant cell line HCC1954 additional simulations were performed. These revealed specific protein interactions reinforcing the hyperactive PI3K signalling pathway in this cell line. Furthermore, the model structures were compared between the cell lines to detect potential resistance mechanisms. Indeed, cell lines with different breast cancer and resistance phenotypes seem to prefer different signalling pathways, requiring individual therapeutic strategies. In addition, the models hint at feedback loops, pathway crosstalk and hyperactive heterodimers as main resistance mechanisms.
To facilitate the analysis of RPPA data, a corresponding software was conceptually refined and methodologically enhanced. The majority of available tools for RPPA data comprises commercial or non standardized in-house solutions. Above that, these software solutions are generally limited to data preprocessing and normalization, lacking additional functions for graphical and statistical analysis. Hence, the users of the RPPA technology would benefit from improved and freely available alternatives which also allow data comparison across different RPPA platforms. The RPPanalyzer software represents such an alternative. It was extended by new functions which were further integrated in a standardized workflow. This way, users can conveniently conduct automated standard preprocessing steps. The novel methods imply variance estimation, normalization and visualization of time course data. At the same time, the modular character of the software was preserved which allows users to flexibly integrate add-on functions and to choose or adapt existing functions of the toolbox according to their specific needs.
On the genomic level, RNA-Seq data were used to detect genes whose expression differed under trastuzumab treatment or between the analysed cell lines. Both kinds of differential expression point to an essential role of these genes in the action and for the efficiency of trastuzumab. The resistance phenotypes of the cell lines covered sensitivity, intrinsic and acquired resistance. The detected and validated genes were differentially expressed between the sensitive cell line and its trastuzumab treated version as well as between the sensitive cell line and the intrinsically resistant one. The latter comparison mainly revealed genes which are already known in an oncogenic context or suspected to hinder trastuzumab efficiency. Their expression is strongly depending on steroid receptors which is in line with the results presented here. Analogously to the proteomic analysis mentioned before, also in this analysis the PI3K signalling pathway was ascribed an important role in drug resistance. Furthermore, the analyses revealed that the intrinsically resistant cell line differs more from the sensitive one than it is the case for the one with acquired resistance. The intrinsically resistant cell line was further influenced less by trastuzumab treatment than the one with acquired resistance. Additionally, mutations were detected in the untreated cell lines which also potentially impact trastuzumab action and efficiency. In the intrinsically resistant cell line more variations were detected than in the one with acquired resistance. This highlights the afore mentioned results. The intersect of mutations in the resistant cell lines very likely affects the efficiency of trastuzumab.
In conclusion, genomic and proteomic measurement data and corresponding models provided an insight into resistance mechanisms. The signalling network models unveiled protein interactions as potential resistance mechanisms and allowed simulations to infer optimal drug combinations. Based on the genomic measurements, genes and mutations were detected as potential resistance mechanisms. Additionally, within both approaches the different resistance phenotypes were compared to indicate prospective applications in personalized medicine. | de |