Discovery and directed evolution of nitroreductase enzymes for activation of prodrugs and PET imaging compounds

Abstract

Bacterial nitroreductase enzymes, which exhibit the capacity to reduce a wide range of nitroaromatic drugs, antibiotics and environmental pollutants, have shown promise in the activation of prodrugs such as CB1954 and PR-104A. Use of these prodrugs in gene-directed enzyme prodrug therapy (GDEPT) cancer treatment would allow for targeted chemotherapy in tumour cells following specific delivery of nitroreductases to these cancerous tissues, using specialised bacterial or viral vectors. However, one key limitation in nitroreductase-based GDEPT is the current inability to rapidly and non-invasively determine vector localisation and gene delivery prior to systemic administration of prodrug. Dual-purpose nitroreductases that exhibit the ability to activate both GDEPT prodrugs and radioisotope-labelled PET imaging probes, in a manner that renders them temporarily cell-entrapped for detection using a PET scanner, would facilitate clinical development of this treatment.

Previous attempts to repurpose hypoxia-activated 2-nitroimidazole PET imaging probes for nitroreductase detection have suffered from relatively high background activation under hypoxia alone. The design of nextgeneration 5-nitroimidazole PET imaging probes, by our collaborators at the Auckland Cancer Society Research Centre (ACSRC), has resulted in much lower levels of hypoxia activation in vivo.

This thesis describes attempts to generate improved nitroreductases that can activate a bespoke 5-nitroimidazole PET-capable imaging probe, S33. A 58-membered library of nitroreductase candidates, including enzymes from many different bacterial species and oxidoreductase families, was heterologously over-expressed in E. coli screening strains. Microplate-based screening strategies were then used to identify enzymes that exhibited the most activity with S33, based on the ability of high levels of activated S33 to induce DNA damage and (at very high levels) E. coli cell death. Following this, site-targeted libraries of two different promising nitroreductase NfsA homologues were screened for S33 activity, with selected variants from each

library showing improvement in S33 activation over the parent nitroreductase. In parallel I performed error-prone PCR mutagenesis of a top NfsA variant and top NfsB variant, subjecting each to two rounds of random mutagenesis, and selecting improved variants using a specialised E. coli screening strain and fluorescence-activated cell sorting (FACS). Selected variants from the NfsB (but not NfsA) nitroreductase candidate library showed substantially improved capacity to activate S33 over the parent enzyme.

As an alternative means for developing improved nitroreductase variants, two different nitroaromatic ‘anti-prodrugs’, the anthelmintic niclosamide and the antibiotic chloramphenicol, whose cytotoxic effects on E. coli can be mitigated by the presence of an over-expressed active nitroreductase, were used to select for improved S33-activating enzymes from a site-targeted NfsA library. Variants were discovered that exhibited improved ability to active S33 as well as other nitroaromatic substrates of interest. Finally, attempts to discover novel nitroreductases from nature through the screening of cloned soil metagenomic fragments, were made utilising a novel cloning strategy to improve expression of the cloned gene fragments in E. coli. Screening and selection of nitroreductase gene ragments was conducted using niclosamide as well as nitroaromatic compounds that change colour upon activation.