| Improving the efficacy of targeted cancer treatments requires quantitative knowledge of how potential therapeutic targets at the molecular level drive tumorigenesis at the physiological level. In this dissertation, we employ quantitative, multiscale models at the biochemical, cell population, and physiological levels to understand how deregulation of the ErbB signaling network, for which several targeted cancer therapeutics have been developed and used successfully, drives tumorigenesis.;At the biochemical level, we develop mechanistic, ordinary differential equation (ODE) models that relate stimulation of all four ErbB receptors with two different ligands, epidermal growth factor (EGF) and heregulin (HRG), to activation of the key ERK and Akt pathways and immediate early expression of the transcription factor c-Fos in MCF-7 breast cancer cells. These models constitute a major advance over previous models that are limited to only a single receptor and ligand and do not consider gene expression changes. Model simulation results corroborated by experimental data show that (i) EGF-induced ERK activation is more sensitive than HRG-induced ERK activation to the potential targeted pharmaceutical U0126, (ii) ErbB2 overexpression, which occurs in ∼25% of breast cancers and for which the targeted pharmaceutical Herceptin was developed, sustains EGF-induced ERK activation, and (iii) sustained (HRG) vs. transient (EGF) ERK signaling induces all (HRG) or nothing (EGF) c-Fos expression responses that lead to differentiation (HRG) vs. proliferation (EGF) of MCF-7 cells.;Although these mechanistic modeling efforts were successful in that relevant predictions were experimentally validated, more than half the model parameters were not accurately known, inherently limiting the predictive power of the models. To address this problem, we develop a procedure for designing experiments such that unknown parameter values can be identified to specified precisions from experimental data. Our approach, as opposed to traditional methods, is specifically developed for the typical high-dimensional signal transduction model that routinely contains over 100 ODEs and several hundred parameters. Applying our procedure to a previously published ErbB1 signaling model shows that although approximately half the unknown parameter values can be identified from a small number of data points (∼100), identifying all the unknown parameter values requires a multitude of input perturbations followed by low frequency, system wide measurements and high frequency, specific measurements. These results suggest that combining mass spectrometry (low frequency, system wide measurements) with live cell microscopy (high frequency, specific measurements) is better suited than the conventionally-used immunoblotting for identifying signal transduction model parameter values.;Since cell fate decisions involve switch-like selection from a finite number of discrete options, at the cell population level we investigate the potential for switch-like ErbB signaling. Flow cytometry measurements of EGF-induced ERK activation in individual HEK293 cells reveal bimodal ERK activation response distributions, implying digital, switch-like signaling, rather than analog signaling. We derive a probabilistic model to characterize the ERK cell population response distributions, and upon fitting this model to the response distributions we find that (i) HEK293 cells exist in discrete "ERK on" or "ERK off" populations, (ii) the fraction of cells in the "ERK on" population is a function of time and EGF dose, (iii) the ERK response appears digital at long times (30 min) but analog at short times (2 and 5 min), and (iv) cell-to-cell variability in ERK abundance is significant and can corrupt otherwise clear "on" and "off" signals. As population average ERK activation measurements show strictly analog responses, these results challenge the current paradigm of assuming population average behavior can be extrapolated to single cells.;At the physiological level, we propose a theoretical framework for incorporating the ErbB and other single cell signaling models into the larger picture of cancer in vivo, a necessary step in translating quantitative ErbB signaling network knowledge into improved cancer treatments. At the heart of this framework is a process engineering-inspired "cancer flowsheet" that allows a modular, scalable mathematical description of how a fundamental set of deregulated physioglogical processes interact with one another and depend on single cell signal transduction to drive tumorigenesis. This framework facilitates integration of the widely diverse and bewilderingly complex body of cancer biology knowledge into a single comprehensive product that can provide clinically testable hypotheses for novel, improved, and optimal treatment strategies, and has the potential to change radically the way cancer is treated. |
