Abstract
Tumor hypoxia contributes to resistance to anticancer therapies. Hypoxia-activated prodrugs (HAPs) selectively target hypoxic cells and their activity can extend to well-oxygenated areas of tumors via diffusion of active metabolites. This type of bystander effect has been suggested to be responsible for the single agent activity of the clinical-stage HAP evofosfamide (TH-302) but direct evidence is lacking. To dissect the contribution of bystander effects to TH-302 activity, we implemented a Green's function pharmacokinetic (PK) model to simulate the spatial distribution of O(2), TH-302 and its cytotoxic metabolites, bromo-isophosphoramide mustard (Br-IPM) and its dichloro derivative isophosphoramide mustard (IPM), in two digitized tumor microvascular networks. The model was parameterized from literature and experimentally, including measurement of diffusion coefficients of TH-302 and its metabolites in multicellular layer cultures. The latter studies demonstrate that Br-IPM and IPM cannot diffuse significantly from the cells in which they are generated, although evidence was obtained for diffusion of the hydroxylamine metabolite of TH-302. The spatially resolved PK model was linked to a pharmacodynamic (PD) model that describes cell killing probability at each point in the tumor microregion as a function of Br-IPM and IPM exposure. The resulting PK/PD model accurately predicted previously reported monotherapy activity of TH-302 in H460 tumors, without invoking a bystander effect, demonstrating that the notable single agent activity of TH-302 in tumors can be accounted for by significant bioreductive activation of TH-302 even in oxic regions, driven by the high plasma concentrations achievable with this well-tolerated prodrug.