E. coli NfsA: an alternative nitroreductase for prodrug activation gene therapy in combination with CB1954

Prodrug activation gene therapy is a developing approach to cancer treatment, whereby prodrug-activating enzymes are expressed in tumour cells. After administration of a non-toxic prodrug, its conversion to cytotoxic metabolites directly kills tumour cells expressing the activating enzyme, whereas the local spread of activated metabolites can kill nearby cells lacking the enzyme (bystander cell killing). One promising combination that has entered clinical trials uses the nitroreductase NfsB from Escherichia coli to activate the prodrug, CB1954, to a potent bifunctional alkylating agent. NfsA, the major E. coli nitroreductase, has greater activity with nitrofuran antibiotics, but it has not been compared in the past with NfsB for the activation of CB1954. We show superior in vitro kinetics of CB1954 activation by NfsA using the NADPH cofactor, and show that the expression of NfsA in bacterial or human cells results in a 3.5- to 8-fold greater sensitivity to CB1954, relative to NfsB. Although NfsB reduces either the 2-NO₂ or 4-NO₂ positions of CB1954 in an equimolar ratio, we show that NfsA preferentially reduces the 2-NO₂ group, which leads to a greater bystander effect with cells expressing NfsA than with NfsB. NfsA is also more effective than NfsB for cell sensitisation to nitrofurans and to a selection of alternative, dinitrobenzamide mustard (DNBM) prodrugs.Gene-directed enzyme prodrug therapy (GDEPT) is a developing strategy for cancer treatment, involving delivery to tumour cells of an exogenous gene, encoding an enzyme that can convert a non-toxic prodrug into cytotoxic products. In principle, local generation of highly reactive cytotoxins within the cancer cells allows optimal therapeutic effect, whereas systemic toxicity remains lower than with conventional chemotherapy (McNeish et al, 1997; Niculescu-Duvaz and Springer, 2005; Russell and Khatri, 2006). As specific gene delivery to all tumour cells seems unattainable using current methods, the ability of an activated prodrug to spread locally from cell to cell is considered an essential feature for a successful GDEPT system. It allows tumour cells that have escaped gene transfer to be killed as a result of prodrug activation in nearby cells that express the activating enzyme. This is known as the ‘bystander effect'' (Freeman et al, 1993).CB1954 (5-[aziridin-1-yl]-2,4-dinitrobenzamide) is a prodrug, which is converted from a weak, monofunctional alkylating agent to a potent bifunctional alkylating agent upon nitroreduction (Knox et al, 1988, 1991; Friedlos et al, 1992). Anlezark et al (1992) reported that the nitroreductase encoded by the nfsB gene of Escherichia coli could activate CB1954, leading to the initial adoption of this enzyme for use with CB1954 in GDEPT (Anlezark et al, 1992; Bridgewater et al, 1995; Grove et al, 1999; Searle et al, 2004). The expression of NfsB in cancer cells using replication-defective retrovirus or adenovirus vectors was shown to confer greatly increased sensitivity to CB1954, and anti-tumour activity was shown in vivo (McNeish et al, 1998; Djeha et al, 2000, 2001; Weedon et al, 2000). After the initial phase I clinical trials of CB1954 alone (Chung-Faye et al, 2001), and of a replication-defective adenovirus expressing NfsB (CTL102) (Palmer et al, 2004), the combination has been tested in a phase I/II trial in patients with prostate cancer. A decline in serum levels of prostate specific antigen in some patients suggests anti-tumour activity (Patel et al, 2009); however, greater efficacy is desirable.NfsB is a homodimeric flavoenzyme which can use either NADH or NADPH to reduce the tightly bound FMN; after dissociation of the NAD(P)⁺; a variety of nitroaromatic or quinone substrates can be reduced in a second reaction step (Anlezark et al, 1992; Lovering et al, 2001; Race et al, 2005). The published K_m for NfsB nitroreductase with CB1954 is 862 μM (Anlezark et al, 1992), which is considerably higher than its peak serum concentration achievable in humans (5–10 μM) (Chung-Faye et al, 2001). Thus, the activation of CB1954 by NfsB in vivo will be very inefficient, justifying the consideration of alternative enzymes (Anlezark et al, 2002). NfsB was originally identified through its role in bacterial sensitivity to nitrofuran antibiotics (Sastry and Jayaraman, 1984). Selection of E. coli for nitrofurazone resistance leads first to mutations in the major nitroreductase gene, nfsA (Whiteway et al, 1998), and only subsequently in nfsB. Alignment of the amino acid sequences shows only 28 identical and 32 conserved residues out of the 242 or 217 residues of NfsA and NfsB, respectively, and antibodies specific for NfsB do not cross-react with NfsA (unpublished results). The two enzymes share many structural features, including a 5-stranded anti-parallel β-sheet core and surrounding α-helices, with the two active sites occupying clefts at the dimer interface and presenting the re-face of the FMN isoalloxazine ring towards the substrate pocket (Kobori et al, 2001; Lovering et al, 2001). Although some residues around the active site are conserved, the active site of NfsA is more open than that of NfsB, an observation that contributed to our decision to investigate its activity with CB1954 and other prodrugs. Kinetic studies of the purified enzymes showed that NfsA has a two- to three-fold greater activity with nitrofurazone and several other nitroaromatic substrates, compared with NfsB (Zenno et al, 1996). The report that NfsA has a marked preference for the cofactor NADPH (Zenno et al, 1996) suggested that this enzyme may have been overlooked in the initial isolation of NfsB for the reduction of CB1954, which used NADH as a cofactor (Anlezark et al, 1992). One study has shown that NfsA is capable of CB1954 activation (Barak et al, 2006); however, to our knowledge, no earlier study has compared the activities of NfsA and NfsB with CB1954. In this study, we compare the ability of these two E. coli nitroreductases to sensitise cells to CB1954 and a selection of other prodrugs. We also compare their kinetics of CB1954 activation in vitro, and show that NfsA preferentially reduces the 2-NO₂ group of CB1954, resulting in an improved bystander cell killing. Overall, the results suggest that NfsA could have advantages over NfsB for use in GDEPT with CB1954 or several other nitroaromatic prodrugs.