4. Discussion
An ideal GDEPT prodrug is substantially less cytotoxic than its corresponding active metabolite(s) and is a unique substrate for the chosen prodrug activating enzyme under physiological conditions. PR-104 is a clinical stage nitroaromatic prodrug that is demonstrably more active than the prototypical NTR prodrug CB1954 in gene therapy models [24]. Nevertheless, several mechanisms of PR-104A activation including hypoxia-selective metabolism [10], aerobic activation by AKR1C3 [15] and oxygeninsensitive non-CYP450 hepatic enzymes [36], indicate this prodrug lacks selectivity for NTRs in the systemic setting. In order to generate an NTR-specific prodrug for gene therapy applications, the chemical structure of PR-104A was modified to minimize activation by endogenous human enzymes. Rational design led to the synthesis of the prodrug SN34507, an analog of PR-104A lacking the ancillary ortho nitro moiety predicted to reduce recognition by the AKR1C3 active site. Additionally, the potential for ortho nitro reduction leading to intramolecular cyclization of the activate metabolite to form non-cytotoxic by-products (as seen for PR-104A [10]) is negated. Loss of AKR1C3-dependent cytotoxicity was confirmed in 2D anti-proliferative IC50 assays, 3D multicellular layer assays and tumor bearing murine models (Fig. 3). Suggestion of residual activity in low cell density assays did not result in biologically meaningful activity in high cell density MCL and tumor models, with no evidence of AKR1C3 driven antitumor activity.