Pseudomonas aeruginosa NfsB and nitro-CBI-DEI – a promising enzyme/prodrug combination for gene directed enzyme prodrug therapy

In gene-directed enzyme prodrug therapy (GDEPT) tumour cells are sensitised, via selective transgene delivery and/or expression, to a systemically administered prodrug. A key aspect of GDEPT is the bystander effect, the ability of activated prodrugs to transport either passively or actively out of the cell of origin and into neighbouring non-transfected cells, which provides an elegant solution to the unavoidable issue of low cell transfection rates [1]. Studies employing prodrugs with high bystander effects have demonstrated that significant tumour reduction can occur when less than 0.1% of the tumour population expresses the activating enzyme [2].

Bacterial type I nitroreductase enzymes, which catalyse the simultaneous two-electron bioreductive activation of nitroaromatic prodrug substrates, hold great potential for GDEPT. To date, the majority of nitroreductase GDEPT studies have focused on the prodrug CB1954 [5-(aziridin-1-yl)-2,4-dinitrobenzamide] (Figure 1B, structure inset), which exhibits only a modest bystander effect upon activation [3]. However, the intrinsically oxygen insensitive nature of the two-electron reduction mechanism enables nitroaromatic prodrugs that were originally designed to target tumour hypoxia (i.e. by exploiting the oxygen-sensitive one-electron reduction mechanism of endogenous human reductases), to potentially be re-purposed for nitroreductase GDEPT [1]. The main class of hypoxia-activated prodrugs to have been considered in this context is the dinitrobenzamide mustards (DNBMs) e.g. [4‐6]; not only are these substantially more cytotoxic and generally better tolerated than CB1954, they also typically exert a greater bystander effect [3, 4, 7].

Also promising for GDEPT are the nitro-chloromethylbenzindolines (nitro-CBIs), originally designed to be hypoxia-activated prodrugs [9] of amino analogues of the cyclopropylindoline anti-tumour antibiotics, exemplified by CC-1065 and duocarmycin SA [10, 11]. It has been shown that the Escherichia coli nitroreductase NfsB (NfsB_Ec) can reduce nitro-CBIs in an oxygen-independent fashion, generating highly cytotoxic metabolites that alkylate the N3 of adenine in the minor groove of DNA [12]. However, it was inferred that the lead nitro-CBI prodrug in that study, nitro-CBI-5-[(dimethylamino)ethoxy]indole (nitro-CBI-DEI; Figure 1A, structure inset) is a poor substrate for NfsB_Ec relative to CB1954 or the DNBMs [12]. In this work we sought to identify more active nitroreductases for metabolism of nitro-CBI-DEI, reasoning that superior enzymes will be required to extend the therapeutic index of nitro-CBI prodrugs in GDEPT.

To identify novel nitroreductase enzymes we previously constructed an over-expression library of eleven candidate oxidoreductases from E. coli[13]. Here, preliminary screens of this library indicated that in addition to NfsB_Ec, E. coli NfsA (NfsA_Ec) was able to activate nitro-CBI-DEI to a measurable extent (not shown). We next tested an orthologous collection of oxidoreductase enzymes from Pseudomonas aeruginosa (Table 1), which we had independently assembled to investigate their hypothesised role in reducing antioxidant quinones in this bacterium (LK Green, AC la Flamme and DF Ackerley, unpublished work). When NfsA_Ec, NfsB_Ec and the P. aeruginosa enzymes were individually over-expressed in the E. coli strain SOS-R2, which contains a lacZ reporter gene under control of the SOS responsive promoter sfiA[6], NfsA_Ec, NfsB_Ec, and NfsB_Pa all generated a powerful induction of the SOS (DNA damage repair) response following challenge with 5 μM nitro-CBI-DEI (Figure 1A). In contrast, only NfsA_Ec and NfsB_Ec induced a substantial SOS response following challenge with 20 μM CB1954 (Figure 1B).

To further distinguish the ability of NfsA_Ec, NfsB_Ec, and NfsB_Pa to activate each prodrug we employed a bacteria-delivered enzyme cytotoxicity assay, as previously described [8]. E. coli strains individually over-expressing each nitroreductase or containing an empty plasmid control were incubated in co-culture with a monolayer of untransfected human colon carcinoma (HCT-116) cells across a range of concentrations of CB1954 (25–400 μM) or nitro-CBI-DEI (0.06-15 μM). Consistent with the SOS assays, the two E. coli enzymes were far more effective than NfsB_Pa in sensitizing HCT-116 cells to CB1954 (Figure 2A); whereas NfsB_Pa exhibited a significantly lower IC₅₀ (the concentration of prodrug required to inhibit growth of HCT-116 to 50% of unchallenged control levels) with nitro-CBI-DEI than either NfsA_Ec or NfsB_Ec (p ≤ 0.001; T-test) (Figure 2B).

We next created HCT-116 cell lines stably transfected with NfsA_Ec, NfsB_Ec or NfsB_Pa, using the Gateway™ compatible expression plasmid F527-V5, which expresses inserted genes from a constitutive human elongation factor-1 alpha promoter (all details as per [13]). Functional nitroreductase activity was confirmed qualitatively for all three stably transfected cell lines, but not the parental HCT-116 cells, using the bacterial nitroreductase specific fluorogenic probe FSL81 as previously described [6] (Figure 3). The sensitivity of each cell line towards nitro-CBI-DEI was then evaluated using an in vitro proliferation assay as described [3]. Briefly, replicate monolayers (n = 3) of NfsA_Ec, NfsB_Ec, NfsB_Pa or non-transfected HCT-116 cells were exposed to a range of concentrations of nitro-CBI-DEI (dilution series from 15.2 pM to 0.3 μM) for 18 h. Cells were washed free of drug and incubated for a further 4 days, after which wells were stained with sulforhodamine B to detect protein as a measure of cell proliferation. Consistent with the results of the bacteria-delivered cytotoxicity assays, the stably transfected NfsB_Pa cell line (IC₅₀ = 2.0 ± 0.4 nM) was significantly more sensitive to nitro-CBI-DEI than either the NfsA_Ec (IC₅₀ = 24 ± 5 nM), NfsB_Ec (IC₅₀ = 9.0 ± 1.2 nM) or parental HCT-116 (IC₅₀ = 47 ± 9.2 nM) cell lines (p ≤ 0.003; T-test). The 5.2-fold differential between WT and NfsB_Ec cells is consistent with the 4.6-fold differential reported previously for these cell lines [12] but is considerably smaller than the 23.5-fold differential observed here between WT and NfsB_Pa cells.

The role of NfsB_Ec in the activation of nitro-CBI-DEI has previously been explored in a three dimensional cell culture model, in which expression of NfsB_Ec sensitised HCT-116 cells to nitro-CBI-DEI by a factor of 12-fold, and there was evidence of a substantial bystander effect for the activated metabolite(s) [12]. Having shown NfsB_Pa to be superior to NfsB_Ec at activating nitro-CBI-DEI in cell monolayers, we next sought to quantify the ability of a minority of NfsB_Pa transfected ‘activator’ cells to kill untransfected HCT-116 ‘target’ cells in three dimensional (3D) co-cultures [3], via the bystander effect of activated nitro-CBI-DEI metabolites (Figure 4). Sensitivity to nitro-CBI-DEI was calculated using a post-exposure clonogenic endpoint [3], where the C₁₀ value represents the concentration of nitro-CBI-DEI that yields one log of cell kill. In cultures of 100% target cells, the C₁₀ value was 52.3 μM (Figure 4). In contrast, in co-cultures of target cells and a minority (5.8% ± 3.2%) of NfsB_Pa-transfected activator cells, the activator cells were 540-fold more sensitive to nitro-CBI-DEI, with a C₁₀ value of 0.097 μM (Figure 4). The target cells in this 3D co-culture were nearly as sensitive as the activators, their C₁₀ value of 0.19 μM indicating an almost perfectly uniform transfer of toxicity (Figure 4). The overall bystander effect efficiency (BEE) was calculated to be 89%; this strikingly efficient transfer of cytotoxicity compares to previously measured BEEs of 13% for CB1954, and 48% for the lipophilic DNBM prodrug SN27686 (using NfsB_Ec expressing activator cells) [4]. The BEE is a measure of the extent to which activator cells cause the dose response curve for co-cultured target cells to shift toward the activator dose response curve; and is calculated according to the formula (LogC₁₀T-LogC₁₀T_C)/(LogC₁₀T-LogC₁₀A_C), where C₁₀T is the C₁₀ value for a pure culture of target cells, C₁₀T_C is the C₁₀ value for target cells in co-culture with activators, and C₁₀A_C is the C₁₀ value for those activator cells [4]. It is important to acknowledge that the quoted BEEs for CB1954 and SN27686 were measured using only a 1% activator cell population [4]. However, in another study that used a 50% activator cell population [7], the shift in dose response curve for the co-cultured target cells was still less substantial with any of the five prodrugs examined (CB1954, RSU-1069, triapazamine, or the DNBMs SN23862 or SN23816) than with nitro-CBI-DEI and the 5.8% activator cell population in our study.

The disparity between killing of NfsB_Pa transfected HCT-116 cells due to nitro-CBI-DEI activation in the in vitro IC₅₀ proliferation assay and the 3D mixed cell cultures highlights the importance of evaluating potential GDEPT enzyme-prodrug partnerships in a 3D bystander model, as low cell density may dramatically underestimate cytotoxic potential. We surmise that the cytotoxic potential of nitro-CBI-DEI may be dramatically underestimated in low cell density proliferation assays due to washout and loss in the media, as suggested for other high-bystander prodrugs [3].

The unprecedented level of bystander killing, coupled with the substantial superiority of NfsB_Pa over NfsB_Ec in sensitising 3D cultures of nitroreductase-expressing HCT-116 cells to nitro-CBI-DEI, suggests that the NfsB_Pa/nitro-CBI-DEI combination is worthy of further evaluation in preclinical GDEPT models. Should it prove necessary to further enhance enzyme activity via directed evolution or targeted mutagenesis studies, our demonstration that nitro-CBI-DEI induces the E. coli SOS response upon activation indicates that it should be possible to recover improved NfsB_Pa variants by SOS screening, as previously used to generate superior CB1954 activating variants of the Vibrio fischeri nitroreductase FRaseI [8].