Characterisation and optimisation of nitroreductase-prodrug combinations for bacterial-directed enzyme-prodrug therapy

Abstract

Gene-directed enzyme-prodrug therapy (GDEPT) employs tumour-tropic vectors including viruses (VDEPT) and bacteria (BDEPT) to deliver a genetically-encoded prodrug-converting enzyme to the tumour environment, thereby sensitising the tumour to a prodrug. Bacterial nitroreductases, which are able to activate a range of anti-cancer nitroaromatic prodrugs to genotoxic metabolites, are of particular interest for GDEPT.

The bystander effect is crucial to the success of GDEPT. The bystander effect is a measure of how efficiently activated prodrug metabolites are transferred from gene-expressing cells to neighbouring tissues. This promotes more extensive tumour cell killing. The bystander effect has been quantified for multiple nitroaromatic prodrugs in mixed multilayer human cell cultures. Although this is a good model for VDEPT it cannot simulate the ability of these prodrug metabolites to exit the bacterial vectors relevant to BDEPT. Prior to this work there was an unmet need for an in vitro method of quantifying the bystander effect as it occurs in BDEPT, i.e. a bacterial model of cell-to-cell transfer of activated prodrug metabolites.

This thesis presents a method for measuring the bacterial bystander effect in vitro in a microplate based assay that was validated by flow cytometry. In this assay two Escherichia coli strains are grown in co-culture; an activator strain expressing the nitroreductase E. coli nfsA and a recipient strain containing an SOS-GFP DNA damage responsive gene construct. In this system, induction of GFP by reduced prodrug metabolites can only occur following their transfer from the activators to the recipients.

Using this method, the bacterial bystander effect of the clinically relevant prodrugs, metronidazole, CB1954, nitro-CBI-DEI, PR-104A and SN27686, was assessed. Consistent with the bystander efficiencies in human cell multilayers, reduced metronidazole exhibited little bacterial cell-to-cell transfer, whereas nitro-CBI-DEI was passed very efficiently from activator to recipient cells post-reduction. In contrast with observations in human cell multilayers, the PR-104A and SN27686 metabolites were not effectively passed between the two bacterial strains, whereas reduced CB1954 was transferred efficiently. Using nitroreductase enzymes that exhibit different biases for the 2- versus 4-nitro substituents of CB1954, I further showed that the 2-nitro reduction products exhibit substantially higher levels of bacterial cell-to-cell transfer than the 4-nitro reduction products. The outcomes of this investigation highlighted the importance of evaluating enzyme-prodrug combinations in models relevant to the intended GDEPT vector, as there can evidently be profound differences in efficacy in different settings. These findings motivated an investigation into the influence of the bystander effect on certain screening strategies used for directed evolution of nitroreductases. It was observed that the bacterial bystander effect can occur during fluorescence activated cell sorting (FACS) of a nitroreductase variant library and negatively impact the recovery of more active variants. Significantly fewer nfsA-expressing cells were recovered from FACS when using CB1954 and nitro-CBI-DEI, when the bystander effect was given time to occur, as compared to controls in which the bystander effect was given no time to occur. In contrast, at the preferred challenge concentrations the mustard prodrugs PR-104A and SN27686 did not yield significantly lower proportions of nfsA-expressing cells under bystander condition.

A subsequent investigation compared the evolutionary outcomes arising from screening a nitroreductase variant library using FACS, in which the bystander effect can occur, in parallel to a manual pre-selection method of individual clones for detoxification of structurally divergent nitroaromatic antibiotics. Overall the results of this investigation were inconclusive after just a single round of selection, but there is some evidence that the FACS strategy was more effective than niclosamide/chloramphenicol pre-selection in enriching for superior CB1954-reducing variants.

Finally, a panel of nitroreductase candidates was evaluated with the next generation prodrugs PR-104A and SN36506 for possible Clostridia-DEPT development. It was found that the Vibrio vulnificus NfsB F70A/F108Y variant displayed the highest catalytic efficiency with PR-104A reported thus far compared to any other nitroreductase, and was the only NfsB family nitroreductase to exhibit any activity with SN36506 at the purified protein level. At the time this research was performed only NfsB family nitroreductases had been successfully expressed in C. sporogenes by our collaborators, hence the V. vulnificus NfsB F70A/F108Y variant was selected as a promising lead enzyme for further development.