Abstract
Small-molecule kinase inhibitors often modulate signaling pathways other than the one targeted, whether by direct "off-target" effects or by indirect "pathway cross-talk" effects. The presence of either or both of these classes of complicating factors impedes the predictive understanding of kinase inhibitor consequences for cell phenotypic behaviors involved in drug efficacy responses. To address this problem, we offer an avenue toward comprehending how kinase inhibitor modulations of cell signaling networks lead to altered cell phenotypic responses by applying a quantitative, multipathway computational modeling approach. We show that integrating measurements of signals across three key kinase pathways involved in regulating migration of human mammary epithelial cells, downstream of ErbB system receptor activation by epidermal growth factor (EGF) or heregulin (HRG), significantly improves prediction of cell migration changes resulting from treatment with the small-molecule inhibitors 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one (LY294002) and 2'-amino-3'-methoxyflavone (PD98059) for both normal and HER2-overexpressing cells. These inhibitors are primarily directed toward inhibition of phosphatidylinositol 3-kinase (PI3K) and mitogen-activated protein kinase kinase (MEK) but are known to exhibit off-target effects; moreover, complex cross-talk interactions between the PI3K/Akt and MEK/extracellular signal-regulated kinase (Erk) pathways are also appreciated. We observe here that treatment with LY294002 reduces migration of HRG-stimulated cells but not EGF-stimulated cells, despite comparable levels of reduction of Akt phosphorylation under both conditions, demonstrating that the target inhibition effect is not unilaterally predictive of efficacy against cell phenotypic response. Consequent measurement of levels of Erk and p38 phosphorylation, along with those for EGF receptor phosphorylation, after LY294002 treatment revealed unintended modulation of these nontargeted pathways. However, when these measurements were incorporated into a partial least-squares regression model, the cell migration responses to treatment were successfully predicted. Similar success was found for the same multipathway model in analogously predicting PD98059 treatment effects on cell migration. We conclude that a quantitative, multipathway modeling approach can provide a significant advance toward comprehending kinase inhibitor efficacy in the face of off-target and pathway cross-talk effects.