Prediction of prognosis and chemosensitivity in breast cancer
Item Usage Stats
Breast cancer (BC) is responsible for the highest mortality and morbidity out of all the cancers in women which is primarily due to both inter and intra-tumoral molecular heterogeneity. This heterogeneity arises from stemness, epithelial to mesenchymal transition and the type of treatment given to patients. These three biological processes are highly related with each other. Traditional therapy when given to breast cancer patients generally results in the transition of epithelial cells to mesenchymal phenotype. Because treatment targets primarily non-stem cells, it can leave stem cells alive which can later result in a relapse of cancer. In this study we aimed to identify such markers that could classify breast cancer patients into stem/mesenchymal or non-stem/epithelial like phenotypes, to determine how generalized the above stated hypotheses are. We developed a gene list of 15 genes we term as CSC/non-CSC gene list (CNCL) which classifies tumors into stemness and/or EMT based phenotypes and can also classify tumor cells based on their relative sensitivity to treatment with traditional therapeutics such as paclitaxel and doxorubicin. When classified into stem/mesenchymal (CS/M) and non-stem/epithelial (NS/E) phenotypes, we showed that Lapatinib and Midostaurin have a specific growth inhibitory effects on NS/E cells, and CS/M cells, respectively. Surprisingly the CNCL showed prognostic significance only for patients who were treated with paclitaxel in neoadjuvant setting, while it could not prognosticate most other BC cohorts. We argue that this is due to the dynamic plasticity of these tumors, as studied within the third aim of this thesis. Secondly, we aimed to identify chemotherapy biomarkers for paclitaxel, cisplatin and doxorubicin to stratify patients in groups that will or will not benefit from these drugs. Using biomarkers, we selected for this purpose, we performed linear regression analysis using breast cancer cell lines to generate cytotoxicity prediction models which can predict IC50 values for these drugs, based on the expression of two genes in each model. Two models were selected for doxorubicin and cisplatin, and three models were selected for paclitaxel. All models were validated both in silico and in vitro. Thirdly, we aimed to evaluate breast cancer plasticity that occurs upon treatment or when a tumor metastasizes. We noted that some breast tumors not only switch their clinical subtype but also change their molecular subtype upon treatment or metastasis. As breast cancer patient treatment in the routine practice is routed based on breast cancer subtype, it is very important to identify the subtype switches which can be critical for changes in treatment decisions. Additionally, we also identified metastatic biomarkers using large number of cohorts. Lastly, as CNCL genes did not show any prognostic importance in terms of both overall survival and metastasis free survival, we checked if the same is true for melanoma. We used Melanin A (MLANA) and Inhibin (INHBA) genes as the markers for invasive/proliferative, stem/non-stem and mesenchymal/epithelial phenotypes. High INHBA expression, which is epithelial, proliferative and non-stem phenotype biomarker, was associated with poor survival and high MLANA expression, which is mesenchymal, invasive and stem phenotype marker, was associated with good prognosis in melanoma patients. Therefore, these findings in melanoma supported our results in breast cancer.