Abstract:
Background: Hypoxia is common in solid tumours and promotes tumour progression and therapy resistance. Hypoxia-activated prodrugs (HAPs) are designed to eliminate these resistant cells and to exploit hypoxia as a basis of tumour selectivity. A key concept in exploiting hypoxia with HAPs is that active drug metabolites should diffuse from hypoxic zones to overcome spatial variations in HAP activation (bystander effect). However currently there is a lack of robust methods for detection of bystander effects. Methods: We developed a novel multicellular spheroid co-culture model of drug-activator and non-activating (target) cells to quantify bystander effects of HAPs of different chemical types (SN30000, PR104A, TH-302 and SN30548). This model, coupled with monolayer and multilayer cultures and LC-MS/MS analytical methods, was used to investigate HAP cellular pharmacology. Data interpretations were critically evaluated using spatially resolved pharmacokinetic/pharmacodynamic (SR-PK/PD) models, parameterised using measured reaction-diffusion parameters. Results: Penetration of SN30000 into spheroids was compromised by its rapid metabolic consumption, and bystander effects were not sufficient to overcome this limitation. In contrast, PR104A demonstrated bystander effects in hypoxic spheroids that overcame limited diffusion of the prodrug. A discrepancy between experimental and model-simulated activity of PR104A led us to identify dichloro metabolites of the initial reduction products as major bystander mediators. The initial metabolite of TH-302, bromoisophosphoramide mustard, was also shown to be rapidly converted to the corresponding dichloro compound but neither diffused efficiently from cells, suggesting that TH-302 metabolites are unlikely to generate bystander effects. The SR-PK/PD model for TH-302 showed that its single-agent activity does not require bystander effects and can be accounted for by substantial activation outside hypoxic regions. Activation of the nitrochloromethylbenzindoline HAP SN30548 by both P450 oxidoreductase (O2-sensitive) and E.coli NfsA nitroreductase (O2-insensitive) showed the same major cytotoxic metabolite, which spheroid co-culture experiments demonstrated a highly efficient bystander effect. Conclusions: The co-culture spheroid model has potential to be used to investigate bystander effects of many therapeutic agents in 3D cultures. We demonstrate that the combined experimental and mathematical modelling approach is a powerful tool for comprehensive understanding of PK/PD in the tumour microenvironment and enables critical testing of hypotheses.