Mechanistic Modeling of Cell Signaling Heterogeneity and Chemokine Gradients in Cancer

Heterogeneity in cancer can give rise to rare subpopulations of cells that are unlike the bulk average of the cancer cell population, such as metastatic cells which promote the formation of secondary tumors at distant sites. These subpopulations of cells can dramatically affect the progression of the disease. Recent work suggests that heterogeneity in cancer cell populations assessed by clonal analysis cannot fully account for differences in single cell behaviors. Nonclonal heterogeneity, such as variability due to the cellular microenvironment, can play a key role in promoting cancer behaviors. Signaling heterogeneity to downstream kinase effectors is one manifestation of nonclonal heterogeneity. Seemingly-identical cancer cells activate heterogeneous signaling to extracellular signal-regulated kinase (ERK) and Akt, two kinases implicated in cancer growth, survival, proliferation, and metastasis. The mechanistic drivers that promote signaling heterogeneity remain unclear and understanding them is crucial in effectively treating cancer. Advancements in single-cell experimental techniques and computational modeling can elucidate how the spatiotemporal tumor microenvironment shapes cancer cell behavior. In this thesis, we construct mechanistic computational models coupled to experimental data to understand the major drivers of nonclonal heterogeneity in cancer. First, we built a single-cell computational model to explain the heterogeneity in cell signaling responses to ERK and Akt that we observed in breast cancer cells in experiments. The model predicted that the pre-existing signaling state of cancer cells controls signaling responses through chemokine receptor CXCR4, a critical receptor in cancer initiation and metastasis, and that these pre-existing states are shaped by environmental stimuli. The model also predicted that targeted therapies currently under clinical investigation may inadvertently potentiate pro-metastatic signaling through CXCR4. Second, we expanded our computational model to test its robustness in predicting signaling through another key receptor in cancer proliferation, epidermal growth factor receptor (EGFR). Our model predicted single-cell signaling responses in two breast cancer cell lines of various mutational backgrounds to different doses of both CXCR4 and EGFR stimuli. The robustness of our model solidified our hypothesis that variation in cell signaling stems from extrinsic noise in three key pathway components: phosphatidylinositol-3-kinase (PI3K), Ras, and mammalian target of rapamycin complex 1 (mTORC1). Third, we built a spatial model of the tumor microenvironment to understand the impact of circadian rhythms and heterogeneity in the spatial tumor composition in promoting metastasis. We found that the magnitude of chemokine gradients, which can act as the molecular highway directing cancer cells where to invade and metastasize, varies throughout the course of the day with the circadian rhythm such that therapies may be more or less effective based on time of administration. Additionally, the spatial arrangement of a tumor with regard to cells secreting and scavenging these chemokines, which can vary tumor-to-tumor or within a single tumor, has a marked impact on the direction of chemokine gradients. We found that specific arrangements of cells in tumors promote chemokine gradients that can direct cancer cells to intravasate and metastasize. Overall, this thesis builds on our knowledge of heterogeneity in cancer and provides suggestions for clinical opportunities.PHDChemical EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttps://deepblue.lib.umich.edu/bitstream/2027.42/153359/1/spinospc_1.pd