In Silico Screening Reveals Structurally Diverse, Nanomolar Inhibitors of NQO2 That Are Functionally Active in Cells and Can Modulate NF-κB Signaling

Quinone oxidoreductase 2 (NQO2, QR2, EC.1.10.99.2) was first identified by Liao and colleagues (1) as an FAD-containing protein capable of oxidizing a variety of analogues of dihydronicotinamide. It is a homolog of NAD(P)H:quinone oxidoreductase 1 (NQO1, DT-diaphorase, EC.1.6.99.2) with which it shares significant structural similarity. The proteins are homodimeric with 2 active sites comprising residues from both monomers that are located at opposite ends of the dimer. However, major differences exist between the catalytic domains of the 2 enzymes; in particular, the binding pocket of NQO2 is more hydrophobic and slightly larger than NQO1 (2–4). In addition, whereas NQO1 can use NADPH or NADH as reducing cofactors, this does not occur with NQO2 unless the pH is reduced substantially (5). To support the enzymatic activity of NQO2, the most efficient cofactors are N-methyl, N-benzyl, or N-ribosyl nicotinamide (NRH; refs. 6, 7).

The physiologic role of NQO1 is well characterized with its major function being considered to be as a detoxification enzyme (8). A detoxification and/or chemoprotectant role for NQO2 has also been suggested from studies on quinone reduction (9, 10) and the enhanced carcinogen activity of benzo(a)pyrene observed in NQO2^−/− mice (11). However, a comparable metabolic role for NQO1 and NQO2 is confounded by the observation that genetic knockout of NQO1 increases menadione toxicity in mice, whereas in NQO2 knockout mice there is protection against menadione toxicity (12,13). Hence, other nonenzymatic properties of the proteins may contribute to the overall physiologic functions of NQO1 and NQO2. This has been the subject of some recent reviews (8, 14, 15). There is growing evidence that both NQO1 and NQO2 can protect against 20S proteasomal degradation of p53 and other onco-proteins (16). In the case of NQO1 the ability to stabilize p53 is NAD(P)H dependent (17). An inference from this is that the protein needs to be in its reduced state to efficiently interact with p53. Inhibitors of NQO1 such as dicoumarol will compete with NAD(P)H for binding to NQO1 and hence prevent reduction of the enzyme bound FAD; thus, dicoumarol promotes degradation of p53 (17, 18). The results of these pharmacologic experiments have been recapitulated in experiments with cells and tissues derived from NQO1^−/− mice, which have shown reduced levels of p53 and a variety of other proteins (19). Similar observations were made when investigations were made of p53 levels in cells and tissues from NQO2^−/− mice (13). In addition, genetic knockdown of either NQO1 or NQO2 in mouse keratinocytes revealed that both proteins could play a significant role in NF-κB signaling (20, 21).

The classical inhibitor of NQO2 is resveratrol (4), however, a variety of other structurally diverse inhibitors have been identified (3, 7, 22–25). Many of these inhibitors, including resveratrol, imatinib, and melatonin (26, 27), have other substantive pharmacologic properties, which make them less than ideal for probing the cellular function of NQO2. Thus, we recently embarked on a virtual screening strategy to identify potential candidate ligands for NQO2 (3, 22, 23). To achieve this, we took the crystal structure of NQO2 (PDB code 1QR2, resolution at 2Å; ref. 2) and used a hierarchical in silico screening approach to mine the entire National Cancer Institute (NCI) database using computational molecular docking. A range of the 250 top-ranked ligands, including various quinolines, elipticines, acridines, and furanylamidines were made available by the NCI and assessed for their effects on the enzymatic activity of NQO2. Here, we report many of these compounds to be active at nanomolar concentrations as enzyme inhibitors. In addition, we have solved the crystal structure of NQO2 containing one of the inhibitors, NSC13000 (9-aminoacridine) and used this to generate a structure–activity relationship. Compounds, representative of the different structural classes, are also shown to be functionally active as inhibitors of NQO2 enzymatic activity in cells at nontoxic concentrations. This is shown by taking advantage of the selective ability of human NQO2 to activate the drug CB1954 (5-(-aziridin-1-yl)-2,4-dinitrobenzamide) to give a potent cytotoxin (6). Thus, inhibition of the cellular toxicity of CB1954 in air can be regarded as a surrogate measure of the inhibitory potency of the different compounds in cells.

It is known that NF-κB–mediated transcription is a common feature of many tumor types (28) and blockade of NF-κB can cause cell death and tumor regression (28, 29). The cellular activity of NQO2 has been linked to NF-κB signaling (20, 21, 30). Here, we show for the first time that the cofactor for NQO2 activity, NRH, can stimulate the function of NF-κB and that this occurs in a NQO2-dependent manner. Furthermore, it is shown that the potent, structurally diverse inhibitors of NQO2 can also modulate NF-κB signaling. Many of these inhibitors have already been shown to possess anticancer activity and it is likely that this effect on NQO2/NF-κB may contribute to the overall efficacy of these agents. However, now that we have a more precise mechanistic understanding of this process, it provides a rationale for developing NQO2 inhibitors as therapeutic agents in the treatment of cancer.

Unless otherwise stated, all reagents and chemicals were obtained from Sigma-Aldrich. NQO2 inhibitors were made available by the NCI. The structures of all the compounds are given in the Supplementary Table S1. Any compounds that were recognized as being commercially available were obtained from Sigma-Aldrich.

For molecular docking, the crystallographic coordinates of the human NQO2 (PDB code 1QR2, resolution at 2Å; ref. 2) were obtained from the Brookhaven Database. Hydrogen atoms were added to the structure allowing for appropriate ionization at physiologic pH. The FAD fragment was reatom typed to avoid underestimation by ChemScore of lipophilic–aromatic interactions (31). The protonated complex was minimized within SYBYL 7.3 while holding all heavy atoms stationary. The NCI database was docked as described previously (3), and a range of the 250 top-ranked ligands were made available by the NCI and used for biochemical evaluation.

The overexpression of NQO2 was done in BL21 Codon+ cells. Cells were grown to late log phase before being induced with 1 mmol/L IPTG overnight at 20°C. NQO2 was purified to homogeneity by HisTrap HP Ni²⁺-affinity chromatography followed by size exclusion chromatography on a Superdex S200 10/300 GL column. Samples were concentrated to approximately 15 mg/mL and supplemented with 5 μmol/L FAD.

The NQO2–NSC13000 complex was obtained by cocrystallization with a 3:1 molar access of NSC13000 (PDB code–3TZB). Conditions were identified using the clear strategy II matrix screen (Molecular Dimensions) on a Mosquito nanodrop crystallization robot. Crystals suitable for diffraction experiments were obtained by sitting drop vapor diffusion in 400 nL drops containing equal volumes of protein and a solution containing 0.15 mol/L potassium thiocyanate, 0.1 mol/L sodium cacodylate pH 6.5, and 20% PEG 1.5K. Data were collected on beamline I04 at the Diamond Light Source Facility and reduced and scaled with the X-ray Detector Software suite (32).

The crystal structure of NQO2–NSC1300 was determined by molecular replacement using PHASER from PHENIX (33) and the NQO2 structure, 1QR2 (2) as a start model. The models were completed by iterative cycles of manual model building and real space refinement using the Coot program (34) and crystallographic refinement using phenix.refine (33). Structure validation was done with Molprobity. The processing and final refinement statistics are given in Supplementary Table S2.

Recombinant human NQO1 and NQO2 were taken and diluted in 50 mmol/L phosphate buffer to give an enzyme activity that would result in a change in optical absorbance of substrate of approximately 0.1 per minute. Inhibition of NQO1 and NQO2 activity was carried out as described previously (22) using 200 μmol/L NADH or NRH (synthesized in-house by converting NADH to NRH via the use of phosphodiesterase and alkaline phosphatase) as reducing cofactors, and dichlorophenolindophenol (DCPIP) as the substrate. Assays were carried out in the presence or absence of 2 μmol/L bovine serum albumin (BSA). IC₅₀ values were determined using nonlinear curve fitting as implemented in the program Excel (GraphPad) for which a 50% reduction of the initial rate was attained. Each measurement was made in triplicate and the experiments were carried out 3 times. IC₅₀ values, given in Table 1, are derived from each of these determinations.

For enzyme activity in cells, exponentially growing cultures were washed with PBS, scraped into 50 mmol/L phosphate buffer (pH 7.4) containing 5 μmol/L FAD and 250 mmol/L sucrose, sonicated twice for 5 seconds while on ice, and centrifuged at 13,000 rpm for 15 minutes at 4°C. Protein concentrations were determined using the bicinchoninic acid protein assay and enzyme activities determined by adding 10 μL of cell lysate to the reaction mixtures detailed above. The difference in the rates of reaction in the presence and absence of either dicoumarol (100 μmol/L) or resveratrol (1 mmol/L) was used to define NQO1 and NQO2 activity in the cellular lysates respectively.

MDA-MB-468 and BT474 breast cancer cells and K562 chronic myelogenous leukemia cells were obtained from the American Type Culture Collection. MDA-MB-468 and K562 cells were chosen because they express relatively high levels of NQO2 (219 ± 89 and 319 ± 92 nmoles DCPIP reduced/min/mg protein, respectively), with the activity of NQO1 being comparatively lower (8.5 ± 3.1 and 30.6 ± 13.7 nmoles DCPIP reduced/min/mg protein, respectively). BT474 cells express substantially less NQO2 compared with the other cells lines (4.2 ± 2.0 nmoles DCPIP reduced/min/mg protein). Cells were maintained in exponential phase in RPMI-1640 medium (Invitrogen) supplemented with 10% (v/v) heat-inactivated fetal calf serum (FCS; Biosera) and 2 mmol/L l-glutamine (Invitrogen).

The mouse macrophage cell line J774.2 was grown in Dulbecco's Modified Eagle's Medium supplemented with 10% FCS and 2 mmol/L l-glutamine. Macrophages (3 × 10⁶) were plated in 6-cm plates in serum-free medium for 24 hours before they were exposed to a combination of 25 ng/mL IFN-γ and 100 μg/mL LPS with or without the putative NQO2 inhibitors for 24 hours. Following treatment, the cytokines were washed out and serum was returned into the medium. The cells were reincubated for another 24 hours, at which point the culture medium was collected for subsequent measurement of nitrite levels using the Griess reaction. The amount of nitrite accumulated in the medium of cells can be used as a surrogate measure of NO production, which occurs as a consequence of induction of iNOS by the cytokine treatment (35).

MDA-MB-468 cells were seeded (7 × 10⁵) into 6-cm plates and allowed to adhere for 24 hours. Human NQO2 short interfering RNA (siRNA; GAAUGUGGCUGUAGAUGAAUU; Thermo Scientific Dharmacon) was transiently transfected (100 nmol/L) using oligofectamine transfection reagent diluted in opti-MEM media. After 5 hours of incubation at 37°C, 5% CO₂ the medium was replaced with RPMI supplemented with 10% FCS. Maximum reduction of NQO2 was seen at 48 hours posttransfection.

MDA-MB-468 cells were seeded at 2.5 × 10³ cells per well in 96-well plates and exposed to compounds in full growth media for 3, 24, or 96 hours. For the shorter exposure times, drug was removed and replaced with fresh media and allowed to grow on for a total of 96 hours. The number of surviving cells was then determined using the MTT assay (22). All toxicity experiments were repeated on at least 3 separate occasions. Data were analyzed and curves drawn using the GraphPad Prism5 software package. K562 cells were seeded at 5 × 10³ cells per well in 96-well plates and exposed to increasing concentrations of the NCI compounds for 96 hours. The number of surviving cells was then determined using the XTT colorimetric dye reduction assay (36).

To determine the potency at which the compounds could inhibit the cellular activity of NQO2, cells were treated for varying times (3, 24, or 96 hours), with either a fixed concentration of inhibitor and varying concentrations of CB1954 or a fixed concentration of CB1954 and varying concentrations of the putative inhibitors. As above, for the shorter time points, the cells were then washed with PBS and fresh media added. The plates were incubated for a total of 96 hours and the MTT (or XTT) assay was used to determine cell viability.

Two luciferase plasmids were purchased from SA Biosciences. The Cignal NF-κB reporter plasmid contains an inducible firefly luciferase construct and constitutively expressing Renilla luciferase construct. The Cignal Positive Control plasmid contains constitutively expressing firefly luciferase and Renilla luciferase constructs. The inducible firefly reporter construct is controlled by a basal promoter element (TATA box), which is driven by a 6 times transcriptional response element specific for NF-κB. The constitutive Renilla reporter construct is under the control of the CMV promoter (37).

MDA-MB-468 cells were seeded (7 × 10⁵) into 6-cm plates and allowed to adhere for 24 hours. Expression plasmids were transiently transfected (1 μg/plate) using oligofectamine transfection reagent (Invitrogen) diluted in opti-MEM (Invitrogen Gibco) according to manufacturer's instructions. After 5 hours of incubation at 37°C, the medium was replaced with RPMI plus 10% FCS and cells reseeded into a 96-well plate. The cells were treated for 1 hour with 100 μmol/L NRH and/or 20 ng/mL TNFα (Invitrogen), then given various concentrations of inhibitor [dissolved in dimethyl sulfoxide (DMSO) to a final concentration of no more than 0.5%] for 18 hours. Cells were then lysed with passive lysis buffer, 20 μL of lysate was transferred into a white-walled 96-well plate and assayed using the Dual-Luciferase Reporter Assay System (Promega) on a BMG Labtech Microplate Luminometer.

The computationally derived calculated and experimentally determined binding affinities of the top ranked compounds supplied by the NCI are given in Table 1. Also included in the table are values of IC₅₀ for their inhibition of NQO2 in the presence and absence of BSA, together with the IC₅₀ for the toxicity of the compounds toward MDA-MB-468 cells and K562 cells. The inhibitors listed in Table 1 have also been given a designation indicating their structural class [A–poly (fused) aromatic, B–ellipticine, C–acridine, D–quinoline, and E–furanylamidine]. Figure 1 gives an example of the binding mode of a compound from classes A, B, D, and E. The binding of NSC620318, a poly (fused) aromatic compound, is shown in panel A. The compound has a binding energy of −57.45 kcal/mol, an IC₅₀ of 200 nmol/L, and forms π-π stacking interactions with the isoalloxazine ring of FAD, and hydrophobic contacts with W105′, F126, F131, and F178. It also forms hydrogen bond interactions with Y155′ (2.8 Å), N161′ (2.8 Å), and G174 (2.7Å). Panel B shows the binding of the elipticine analog NSC164016. This compound is the highest scoring ligand with a binding energy of −64.06 kcal/mol and an IC₅₀ of 20 nmol/L. It forms π-π stacking interactions with FAD and hydrophobic contacts with W105′, Y132, F126, F131, and F178. Panel C shows the quinoline, NSC617933 binding in the active site of NQO2. This compound has a binding energy of −59.38 kcal/mol and an IC₅₀ of 40 nmol/L. It forms hydrophobic contacts with active site residues W105′, F126, and F178, π-π stacking interactions with FAD, and has hydrogen bond contacts with T71 (2.4 Å), the carbonyl backbone of D117 (3.0 Å) and Q122 (2.8 Å and 3.1 Å). The related compound, NSC14229 (quinacrine) is the only compound from this panel that has previously been reported as an inhibitor of NQO2 (7). The binding mode of the furanylamidine, NSC305831 is given in panel D. This compound has a binding energy of −57.01 kcal/mol and an IC₅₀ of 630 nmol/L. It forms hydrophobic interactions with W105′, F126, and F178 and hydrogen bond contacts with D117 (2.1 Å), R118 (2.8 Å), Y132 (3.0 Å), G174 (2.9 Å), and N161′ (3.1 Å). Interestingly, the related compound NSC305836 is a more potent inhibitor with an IC₅₀ of 50 nmol/L. This ligand has 2 additional methyl groups on the furanyl ring that results in additional binding with W105′ and F106′.

NSC13000 has a binding energy of −53.73 kcal/mol and an IC₅₀ of 420 nmol/L. The X-ray crystal structure of the ligand bound in the active site of NQO2 is shown in Supplementary Fig. S1. NSC13000 interacts primarily via stacking interactions with the conjugate ring system of FAD and the side chain of F178. The isoalloxazine ring of FAD provides the floor of the active site and stacks with the acridine moiety of NSC13000. Hydrophobic residues F126 and F178 form the ceiling of the binding site, whereas W105′ (primes denote residues from the other NQ02 protomer) sits at the back of the binding pocket and forms a hydrophobic interaction with the amine face of NSC13000. For comparative purposes, the docking orientation of the compound superimposed on to the crystal ligand is given in the Supplementary Fig. S2 and shows excellent concordance between the docking modes with a root mean square distance of 1.26 Å.

A plot of the calculated binding energies for all the compounds versus the experimentally measured binding energy [derived from the Cheng–Prusoff equation (ref. 38), which uses the IC₅₀ values for enzyme inhibition] is shown in Supplementary Fig. S3 and from this analysis, a correlation coefficient of R² = 0.52 was determined.

All of the compounds were also tested for their ability to inhibit NQO1. No inhibition was observed for any of the compounds (up to 100 μmol/L), with the exception of NSC628440, which we had previously reported (39) to be a relatively weak inhibitor. Compounds were also evaluated for their toxicity toward MDA-MB-468 and K562 cells, both of which express relatively high levels of NQO2 (Table 1). Values of IC₅₀ for toxicity vary over 1,000-fold in both cell lines, and there is no clear relationship between NQO2 inhibitory potency and toxicity.

In our previous work on NQO1, we took advantage of its ability to activate the anticancer drug EO9 [5-aziridinyl-3-hydroxymethyl-2-(3-hydroxyprop-1-enyl)-1-methylindole-4,7-dione, Apaziquone] to show the functional activity of the enzyme in cells. This then allowed assessment of the ability of various coumarin-based compounds to act as inhibitors in cancer cells (40, 41). A similar approach was taken here to determine the ability of the compounds identified as potent inhibitors of NQO2 to be functionally active in cells. To do this, we used the unique ability of NQO2 to activate CB1954 to give a potent cytotoxin (6). To validate the approach, MDA-MB-468 cells were used as they possess relatively high levels of NQO2. Cells were first treated with varying concentrations of CB1954 for 3 hours at 37°C. The closed circles in Fig. 2A show that concentrations of CB1954 in excess of 1 mmol/L are needed to achieve a value for IC₅₀. Previously, it was shown that the enzymatic activity of NQO2 is supported by the cosubstrate, NRH, which exists at only very low concentrations in cells (6). Therefore, varying concentrations of NRH (1 μmol/L to 1 mmol/L) were added to cells together with the CB1954. Results are given in Fig. 2A and show that the added NRH causes a substantial increase in the toxicity of CB1954. NRH itself at these concentrations shows no toxicity. As has previously been suggested (6), this strongly suggests that it is the NQO2-supported reductive activation of CB1954 that leads to toxicity. To support this contention, similar experiments were carried out with BT474 cells that express very low levels of NQO2. Data are given in Supplementary Fig. S4 and show that in the absence of NRH, the IC₅₀ value for CB1954 is in excess of 1 mmol/L, whereas in the presence of 100 μmol/L NRH there is a small increase in cellular sensitivity to give an IC₅₀ of 1 mmol/L. Finally, to confirm these observations, MDA-MB-468 cells were transfected with siRNA to NQO2. Western blotting indicated that after 48 hours, NQO2 levels reached a nadir (Fig. 2D), and this was consistent with a measured 60% reduction in enzyme activity (to 93 ± 40 nmol DCPIP reduced/min/mg of protein) at that time. Cells were then treated with varying concentrations of CB1954 plus 100 μmol/L NRH. Results shown in Fig. 2B clearly indicate that downregulation of NQO2 causes a decrease in the toxicity of CB1954.

These results provided a platform for evaluating the ability of the compounds described above to be functionally active as inhibitors of NQO2 in cells. To do this, MDA-MB-468 cells were treated with 100 μmol/L CB1954, in the presence of 100 μmol/L NRH, together with varying concentrations of the putative inhibitors for 3 hours. The maximal concentration of each compound tested depended on solubility and/or final concentration of vehicle (DMSO) or the toxicity of the inhibitor. An amount of 100 μmol/L CB1954 plus 100 μmol/L NRH reduces proliferation of MDA-MB-468 cells to about 25%; therefore, adding increasing concentrations of enzyme inhibitor should elicit a steady decrease in toxicity of CB1954. Fig. 2C shows representative dose–response curves for MBA-468 cells exposed to 100 μmol/L CB1954 and some of the NQO2 inhibitors, and it is clear that protection efficiency varies between the different compounds. Similarly, experiments were carried out with different concentrations of inhibitors with varying concentrations of CB1954. Examples are given in Supplementary Fig. S5 and show that low nanomolar concentrations of the inhibitors can alter the toxicity of CB1954. To compare the inhibitory potency of the structurally diverse compounds, the concentration needed to inhibit the toxicity of 100 μmol/L CB1954 by 50% was determined. These values of IC₅₀ for protection against CB1954 toxicity are given in Table 2 and they differ by more than 1,000-fold. NSC71795 seems to be the most potent compound for protecting against the toxicity of CB1954 in MDA-MB-468 cells, and the concentration required to reduce toxicity by 50% is 0.054 μmol/L. In contrast, resveratrol, the classical inhibitor of NQO2 is substantially less efficient with an IC₅₀ value of approximately 150 μmol/L. Also shown in Table 2 are the values for the toxicity of the putative inhibitors when given to MDA-MB-468 cells alone for 3 hours. In most instances, the ability of the compounds to inhibit the toxicity of CB1954 (i.e., inhibit the activity of NQO2) occurs at significantly lower concentrations than those that, by themselves, are toxic.

Our discovery of potent inhibitors of NQO2 came from computational screening of the NCI database. The compounds that emerged from this study represented a wide variety of structural classes including elipticines, acridines, quinolines, and furanyl amidines. Interestingly, when carrying out a Prediction of Activity Spectra for Substances (PASS) analysis of these compounds using the NCI database (42), some of the agents showed high PASS-predicted probabilities for them to have anti-inflammatory properties and/or inhibitory effects on release of TNFα. Such an analysis can only provide a pointer to activity, however, it is known that some compounds that can inhibit the activity of recombinant NQO2 can also impair gene transcription regulated by TNFα (25, 30). We, therefore, carried out experiments to determine whether compounds representative of the diverse structural classes identified above that are functionally active as inhibitors of NQO2 in cells could alter TNFα-driven NF-κB–mediated gene transcription. In addition, we evaluated whether any drug induced effects on NF-кB processes were NQO2 dependent.

Experiments were carried out in MDA-MB-468 cells transiently transfected with a dual-luciferase construct containing a CMV-driven Renilla reporter and an NF-κB responsive firefly reporter. First, we evaluated the effect of treating cells with TNFα for 18 hours on luciferase activity. Untreated, exponentially growing MDA-MB-468 cells show a small but significant level of NF-κB–mediated transcription, which was increased almost 3-fold by treatment with 20 ng/mL TNFα (Fig. 3A). Because NRH can modulate the functionality of NQO2 in cells, we also carried out experiments in which 100 μmol/L NRH was added to cells with and without TNFα. These results are also given in Fig. 3A and showed that addition of the cofactor NRH stimulates NF-κB activity to a similar level to that seen with TNFα, but combining NRH with TNFα gives no additional effect. This interesting observation may point to a redox-based mechanism linking NQO2 activity and NF-κB function and this is the subject of further investigation. To confirm that the effects were dependent on NQO2, MDA-MB-468 cells were transfected with siRNA to NQO2 and, 48 hours later, treated with or without TNFα. These results are given in Fig. 3B and showed that the reduction in NQO2 activity brought about by the transfection of siRNA causes a reduction in the basal expression of NQO2. Furthermore, the induction of NF-κB by TNFα is also significantly decreased by the reduction in NQO2 activity. The NQO2 inhibitor, NSC13000, was then evaluated for its ability to modulate NF-κB transcriptional activity in MDA-468 cells treated with or without NRH. Results are given in Fig. 3C for cells treated with 0.1 μmol/L NSC13000, and it is clear that drug treatment reduces the luciferase activity in both treatment groups. We then went on to determine whether a range of NQO2 inhibitors, representative of the different structural classes (structures given in Fig. 4B), could attenuate the increase in NF-κB activity caused by treatment with NRH and TNFα. Results given in Fig. 4A show that each of the compounds, at 0.1 μmol/L, can cause a substantial reduction in NF-κB activity. These results were confirmed by the indirect measurement of the activity of an NF-κB target gene (iNOS) in cytokine stimulated mouse macrophages treated with the different inhibitors. The concentration of nitrite in growth medium is a measure of the activity of iNOS and Fig. 4A also shows that each of the compounds given to macrophages at a concentration of 0.125 μmol/L for 24 hours results in a reduction in the level of nitrite. Examples of 6 further compounds attenuating NF-κB activity and nitrite production are given a supplementary Fig. S6.

In this work, virtual screening of the NCI database has revealed novel, nanomolar inhibitors of NQO2 and shown that these compounds are functionally active as enzyme inhibitors in cells. Using an NF-κB reporter assay, it is shown that the cofactor for NQO2 enzymatic activity, NRH, can induce NF-κB gene transcription, and this induction is similar to that level of induction seen with TNFα. The induction of NF-κB activity is reduced by the NQO2 inhibitors, and this observation is recapitulated when enzyme activity is reduced by transfection of siRNA to NQO2. Hence, we have provided some evidence to suggest a causal link between the activity of NQO2 and the regulation of NF-κB and that this can be modulated by a variety of structurally diverse inhibitors of NQO2.

There has been a recent plethora of publications on the synthesis of potential inhibitors of NQO2 (3, 7, 22–26, 43, 44). The work of the Boutin group (7, 26) has focused on the observation that NQO2 is the third melatonin binding site, MT3. From these studies, the fused poly heteroaromatic compound S29434 was identified as a nanomolar inhibitor of recombinant NQO2 with the ability to compete with melatonin for binding. In contrast, the Cushman group (24, 25) have concentrated on the potential cancer chemoprotective activity of resveratrol analogues and substituted phenazines. They measured inhibition of recombinant NQO2 (with several compounds active in the nanomolar range), as well as measuring effects on iNOS activity and NF-κB–mediated transcription (25), and they concluded that inhibition of 2 or more distinct targets will provide a foundation for the development of effective chemoprotective agents. However crucially, no molecular link was made between their observations using the different assays. Dufour and colleagues (43) and Yan and colleagues (44) have extended their work on the development of irreversible inhibitors of NQO1 (45). These agents deactivate NQO1 and NQO2 by alkylation of their active sites. However, this attractive mechanism-based approach also has limitations with regard to target selectivity (46). The work of Nolan and colleagues (3, 22, 23) and that reported here was undertaken with the specific objective of identifying novel inhibitors of NQO2 that might have therapeutic properties on the basis of the impact of NQO2 on NF-κB–mediated gene transcription that was previously suggested from genetic deletion studies (21, 30). Using a computational approach to mine the NCI Chemical data base, we discovered a remarkable variety of structural types that could potentially act as inhibitors of NQO2. The active site of NQO2 is a hydrophobic cavity, 17 Å in length and 7 Å wide, which can accommodate a range of polycyclic and polyaromatic ligands of different size and structure adopting a variety of binding modes and interactions. Among the inhibitors is NSC13000 and we have solved the structure of this compound cocrystalized with NQO2. This structure was used to aid in the construction of a structure–activity relationship between the computationally derived binding affinity of the compounds from the NCI and their experimentally measured inhibitory potency. A reasonable correlation was obtained that, considering the structural diversity of the compounds, provides confidence in the docking analysis and thus can provide leads for future drug development. This same miscellaneous set of compounds has the additional effect of modulating NF-κB activity. In addition, a recent study (47) showed that NSC13000 and NSC14229 could induce p53 function in renal carcinoma cells in a mechanism that did not involve genotoxic stress but was mediated by suppression of NF-κB. Furthermore, both compounds effectively suppressed both the basal and induced activity of NF-κB (47). If these cells contained significant levels of NQO2, then it is possible that the observed phenomenon may be mediated by this enzyme. Thus, there is good evidence to suggest a causual link between NQO2 inhibition and NF-κB function; however, whether this operates at the level of protein–protein interactions and/or maintaining local redox status (15) remains to be determined. There is no doubt that modulating NF-κB could have therapeutic antitumor activity (28, 29) and targeting NQO2 may be a useful method to achieve this. Finally, NQO2 inhibitors can have additional therapeutic effects, because it has been shown that the inhibitors, melatonin and imatinib, can protect against radiation-induced lung inflammation, a process thought to be due, in part, to TNF-dependent effects (48–50).

No potential conflicts of interest were disclosed.

The authors thank the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, NCI for the supply of compounds and John Barnes, Soo Mei Chee, and Soraya Alnabulsi (University of Manchester) for the synthesis of NRH.

This work was supported by an grants from the MRC (G0500366) and AICR (08–0152).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.