The multidrug resistance P-glycoprotein (P-gp) is characterised by the ability to bind and/or transport an astonishing array of drugs. This poly-specificity is imparted by at least four pharmacologically distinct binding sites within the transmembrane domain. Whether or not these sites are spatially distinct has remained unclear. Biochemical and structural investigations have implicated a central cavity as the likely location for the binding sites. In the present investigation, a number of contact residues that are involved in drug binding were identified through biochemical assays using purified, reconstituted P-gp. Drugs were selected to represent each of the four pharmacologically distinct sites. Contact residues important in rhodamine123 binding were identified in the central cavity of P-gp. However, contact residues for the binding of vinblastine, paclitaxel and nicardipine were located at the lipid-protein interface rather than the central cavity. A key residue (F978) within the central cavity is believed to be involved in coupling drug binding to nucleotide hydrolysis. Data observed in this investigation suggest the presence of spatially distinct drug binding sites connecting through to a single translocation pore in the central cavity.

PR-104 is a phosphate ester pre-prodrug that is converted in vivo to its cognate alcohol, PR-104A, a latent alkylator which forms potent cytotoxins upon bioreduction. Hypoxia selectivity results from one-electron nitro reduction of PR-104A, in which cytochrome P450 oxidoreductase (POR) plays an important role. However, PR-104A also undergoes ‘off-target’ two-electron reduction by human aldo-keto reductase 1C3 (AKR1C3), resulting in activation in oxygenated tissues. AKR1C3 expression in human myeloid progenitor cells probably accounts for the dose-limiting myelotoxicity of PR-104 documented in clinical trials, resulting in human PR-104A plasma exposure levels 3.4- to 9.6-fold lower than can be achieved in murine models. Structure-based design to eliminate AKR1C3 activation thus represents a strategy for restoring the therapeutic window of this class of agent in humans. Here, we identified SN29176, a PR-104A analogue resistant to human AKR1C3 activation. SN29176 retains hypoxia selectivity in vitro with aerobic/hypoxic IC50 ratios of 9 to 145, remains a substrate for POR and triggers γH2AX induction and cell cycle arrest in a comparable manner to PR-104A. SN35141, the soluble phosphate pre-prodrug of SN29176, exhibited superior hypoxic tumour log cell kill (>4.0) to PR-104 (2.5–3.7) in vivo at doses predicted to be achievable in humans. Orthologues of human AKR1C3 from mouse, rat and dog were incapable of reducing PR-104A, thus identifying an underlying cause for the discrepancy in PR-104 tolerance in pre-clinical models versus humans. In contrast, the macaque AKR1C3 gene orthologue was able to metabolise PR-104A, indicating that this species may be suitable for evaluating the toxicokinetics of PR-104 analogues for clinical development. We confirmed that SN29176 was not a substrate for AKR1C3 orthologues across all four pre-clinical species, demonstrating that this prodrug analogue class is suitable for further development. Based on these findings, a prodrug candidate was subsequently identified for clinical trials.

No abstract