An Exceptionally Potent Inducer of Cytoprotective Enzymes

The Keap1/Nrf2/ARE pathway controls a network of cytoprotective genes that defend against the damaging effects of oxidative and electrophilic stress, and inflammation. Induction of this pathway is a highly effective strategy in combating the risk of cancer and chronic degenerative diseases, including atherosclerosis and neurodegeneration. An acetylenic tricyclic bis(cyano enone) bearing two highly electrophilic Michael acceptors is an extremely potent inducer in cells and in vivo. We demonstrate spectroscopically that both cyano enone functions of the tricyclic molecule react with cysteine residues of Keap1 and activate transcription of cytoprotective genes. Novel monocyclic cyano enones, representing fragments of rings A and C of the tricyclic compound, reveal that the contribution to inducer potency of the ring C Michael acceptor is much greater than that of ring A, and that potency is further enhanced by spatial proximity of an acetylenic function. Critically, the simultaneous presence of two cyano enone functions in rings A and C within a rigid three-ring system results in exceptionally high inducer potency. Detailed understanding of the structural elements that contribute to the reactivity with the protein sensor Keap1 and to high potency of induction is essential for the development of specific and selective lead compounds as clinically relevant chemoprotective agents.

The Keap1/Nrf2/ARE pathway regulates the ability of eukaryotic organisms to adapt and survive under various con-ditions of oxidative, electrophilic, and inflammatory stress by signaling the expression of a network of more than 100 genes, many of which code for cytoprotective ("phase 2") proteins. Under basal conditions, this pathway does not operate at its maximal capacity, but can be induced by a wide variety of different classes of small molecules (1,2). This strategy protects cells against stressful conditions, and reduces the risk of developing cancer and chronic degenerative diseases. One universal property of inducers is their capacity to modify sulfhydryl groups by alkylation, oxidoreduction, or disulfide formation. Of central importance in this respect is the reaction of inducers with specific cysteine residues of the protein sensor Keap1 (3), which thereby loses its ability to target transcription factor Nrf2 for ubiquitination and proteasomal degradation. This stabilizes Nrf2, increases its binding to the antioxidant response element (ARE), 4 and activates transcription of cytoprotective genes (4 -7).
The identification and analysis of the mechanism of action of potentially useful upregulators (inducers) of the cytoprotective response has been greatly advanced by the development of a simple, cell-based microtiter plate assay, in which the potency of compounds to increase the specific activity of NAD(P)H: quinone oxidoreductase 1 (NQO1), a prototypic Nrf2-dependent gene, can be quantified in murine hepatoma cells (8,9). This system has identified many cytoprotectors, and led to the isolation of the isothiocyanate sulforaphane [1-isothiocyanato-4R-(methylsulfinyl)butane] from broccoli (10). This discovery highlighted the importance of thiol reactivity (i.e. the isothiocyante functionality) for inducer activity. The isothiocyanate group was shown spectroscopically to react directly with specific cysteine residues of Keap1 (3,11). High potency depends on the presence of the methylsulfinyl group separated by a bridge of 4-methylene groups from the -NϭCϭS moiety. Replacement of the sulfinyl (SϭO) by a carbonyl (CϭO) group did not affect potency, but its replacement by a methylene (-CH 2 -) group drastically reduced potency (ϳ75-fold). Changes in the length of the methylene (-CH 2 ) n bridge (where n ϭ 3 or 5) resulted in modest reductions in inducer potency when compared with sulforaphane, and there were also reductions in inducer potency when the state of oxidation of the sulfinyl group was changed to sulfide (-S-) or sulfonyl (OϭSϭO). The critical relation of the distance and the topographic relation of the two reactive groups was also suggested by examining the inducer potencies of a series of isomeric norbornane analogs substituted with CH 3 CO-and -NϭCϭS groups (12). Very recently, a series of sulfoxythiocarbamate analogs of sulforaphane were synthesized that retain the structural features important for high inducer potency, i.e. a sulfinyl or a carbonyl group at a specific distance from the electrophilic functional group (13).
A critical clue to the structural features of inducers required for the chemical signaling of induction was the recognition that many inducers contained Michael acceptors i.e. olefins or acetylenes that are rendered more electrophilic (positively charged) by conjugation with electron-withdrawing groups such as carbonyl [CϭO], sulfinyl [SϭO], nitrile [C'N], or acetylene [C'C]. Moreover, the inducer potencies of these Michael acceptors are closely correlated with the degree of their electrophilicity and consequently with their rates of reaction with nucleophiles (14). It was therefore a significant and an intriguing finding that the semisynthetic triterpenoids such as CDDO (15) and TP-225 (16) (chemical structures are shown in supplemental Fig. S1) containing an activated cyano enone Michael acceptor were not only highly potent (at nanomolar concentrations) inhibitors of inflammation but also extremely potent inducers of NQO1 (17). Furthermore, these multifunctional molecules exhibit remarkable protective efficacies in a number of preclinical animal models of chronic disease, including models of carcinogenesis (18 -21), cardiovascular disease (22), and neurodegeneration (23)(24)(25); indeed, some are currently in clinical development (26).
Based on our earlier structure-activity relationship studies showing that a cyano enone in ring A and an enone in ring C are essential for the extremely high potency of CDDO (15)(16)(17), we hypothesized that the entire triterpenoid skeleton may not be necessary for inducer potency. Therefore, we designed and synthesized tricyclic compounds having the same A, B, and C rings as CDDO (27,28). Among them, the acetylenic tricyclic bis(cyano enone) TBE-31 [(Ϯ)-(4bS,8aR,10aS)-10a-ethynyl -4b,8,8-trimethyl-3,7-dioxo-3,4b,7,8,8a,9,10,10a-octahydrophenanthrene-2,6-dicarbonitrile] (Fig. 1) retains and even exceeds the potency of CDDO analogs, which are the most potent compounds in our pool of semisynthestic triterpenoids in various in vitro and in vivo assays, including induction of cytoprotective enzymes (29). Moreover, this compound inhibits proinflammatory responses in cells, and blocks the formation of aflatoxin-B 1 (AFB 1 )-DNA adducts and AFB 1 -induced tumorigenesis in vivo (29). The present investigation addresses the structural requirements that determine the exceedingly high potency of this molecule in inducing cytoprotective proteins and the spectral characteristics of its reaction product with the protein sensor for inducers Keap1.
Animals and Treatments-All animal experiments were performed in accordance with the regulations described in the UK Animals (Scientific Procedures) Act 1986. SKH-1 hairless mice (6 weeks old) were obtained from Charles River (Germany) and were bred in our animal facility. The animals were kept on a 12-h light/12-h dark cycle, 35% humidity, and given free access to water and food (RM1 diet obtained from SDS Ltd., Witham, Essex, UK).
We used 10 -12-week-old SKH-1 hairless female mice. Of note, these mice are immunocompetent, but have a defect in the hair cycle, which results in permanent hair loss during adulthood. For topical applications, the animals were treated on their backs three times at 24-h intervals with 0.03, 0.1, or 0.3 mol of TBE-31 dissolved in 80% aqueous acetone (v/v, 100 l) over ϳ5 cm 2 area. Control skin received 80% acetone. Mice were euthanized 24 h after the final dose was applied. Each treated segment of their skin was harvested, flash-frozen in liquid N 2 , and stored at Ϫ80°C until analyzed. For the feeding experiment, all animals received powdered RM1 diet for 1 week. Then, they were divided into three groups of four animals. Pure crystalline TBE-31 was mixed with the powdered diet in order to obtain 0.1 mol/3 g diet (9.2 mg of TBE-31 per kg of diet) for the low-dose, and 0.3 mol/3 g diet for the highdose treatment groups, respectively. Group 1 received the control diet, Group 2 was fed the low-dose, and Group 3, the highdose TBE-containing diet. After 11 days of feeding with these diets, the animals were euthanized and their liver, stomach, skin, and cerebral cortex were harvested, flash-frozen in liquid N 2 , and stored at Ϫ80°C until analyzed.
Biochemical Analyses-Hepa1c1c7 cells (10 4 per well) and mouse embryonic fibroblasts (MEF, 2 ϫ 10 4 per well) were grown in 96-well plates for 24 h and then exposed to serial dilutions of inducers for either 48 (Hepa1c1c7) or 24 h (MEF). Enzyme activity of NQO1 in cell lysates was determined by using menadione as a substrate as described (8,9).
To determine the enzyme activity in mouse organs, portions (ϳ50 mg) of snap-frozen tissues were pulverized under liquid nitrogen. The resulting powder was resuspended in 3 volumes (0.15 ml) of ice-cold 100 mM potassium phosphate buffer, pH 7.4, containing 100 mM KCl, 0.1 mM EDTA, and complete protease inhibitor mixture at a dose of one tablet/10 ml buffer. This material was mechanically homogenized in an ice bath. For skin samples only, mechanical homogenization was also followed by three freeze-thaw cycles. All resulting homogenates were subjected to two centrifugation steps at 4°C (15,000 ϫ g for 10 min, followed by 100,000 ϫ g for 90 min). Protein concentrations (33) and enzyme activities of NQO1 (8) and GST (34) were determined in the final 100,000 ϫ g supernatant fractions (cytosols), which were also used for Western blotting for GSTA1, GSTM1, and GSTP1.
Expression of Inducer Potencies-Potencies of induction of NQO1 were determined in a standard assay system, and were analyzed by the Median Effect Equation of Chou (35,36). In this equation, fa/fu ϭ (D/D m ) m , where fa is the fraction of a process that is affected, fu is the fraction that is unaffected, D is the concentration of a compound that is required to produce the effect fa, and D m is the concentration at which there is 50% effect. Plots of log fa/fu with respect to log D are linear and a simple computer program permits determination of goodness of fit to linearity, the linear correlation coefficient, and the slope (Hill coefficient) of the lines. Application of this procedure to the increase (induction) of NQO1 have been described (37).
Keap1 Binding-Recombinant murine Keap1 was expressed and purified as described (3) Twenty-four hours later, cells were lysed, and the firefly luciferase activity was measured after addition of Luciferase Assay Reagent (Promega) as described (39) using a luminometer (Turner Designs Model TD-20/20, Promega). Control cells were treated with acetonitrile (final concentration 0.1%, v/v).
Statistical Analysis-All values are means Ϯ 1 S.D. The differences between groups were determined by Student's t test with Dunnett's correction for multiple comparisons.

The Acetylenic Tricyclic bis(cyano enone) TBE-31 Potently
Induces Cytoprotective Enzymes in Vivo-To evaluate the inducer potency of the acetylenic tricyclic bis(cyano enone), TBE-31 ( Fig. 1) in vivo, we applied three topical doses of small quantities (0.03, 0.1, or 0.3 mol per application) of TBE-31 at 24-h intervals to the dorsal skin of hairless but immunocompetent SKH-1 mice (n ϭ 3). These applications led to a robust, highly statistically significant, and dose-dependent induction of the cytosolic dicoumarol-sensitive NQO1 activity (with menadione as a substrate) by 3.5-, 4.2-, and 4.9-fold, respectively, in the underlying skin ( Fig. 2A). Similarly, by using 1-chloro-2,4-  dinitrobenzene (CDNB) as a substrate, a small, but significant induction in glutathione transferase (GST) activity was also found, by 1.3-, 1.5-, and 1.5-fold, respectively (Fig. 2B). Oral administration of 0.1 mol (9.2 g per g of diet, low-dose) or 0.3 mol (high-dose) of TBE-31 per mouse per day for 11 days also resulted in a dose-dependent increase in NQO1 in three of four tissues (liver, stomach, skin, and cerebral cortex) examined. Importantly, no signs of toxicity were observed at either of these doses of TBE-31 as judged by similarity of body weights and behavior of the animals in all groups. Hepatic NQO1 increased by 2.4-and 3.5-fold with the low and the high oral doses of TBE-31, respectively (Fig. 3A). Remarkably, a virtually identical degree of induction was observed in skin, i.e. by 2.2and 3.5-fold, indicating that even after oral intake, significant amounts of TBE-31 reached and exerted profound biological effects on the skin. The highest basal levels of NQO1 (ϳ20-fold higher than in the other organs examined) were in stomach; induction was also most robust in this organ, by 3.8-(low-dose) and 4.5-fold (high-dose). In contrast, no changes were observed in cerebral cortex, in agreement with the exceedingly low levels of TBE-31 detected in brain of CD-1 mice after a single oral dose (29), suggesting that TBE-31 is being excluded by the blood-brain barrier. Coordinately with NQO1, the glutathione S-transferase (GST) activity was also induced in liver, skin and stomach, but not in brain (Fig. 3B). Again, the highest level of induction was in stomach, by 3.5-and 5-fold with the low and the high doses of TBE-31, respectively. Hepatic GST activity was increased by 2.4-and 3.7-fold. In close agreement with the enzyme activity data, the protein levels of GSTA1, GSTM1, and GSTP1 isoforms were markedly and dose-dependently increased in liver and stomach (Fig. 3C). Band densitometric quantification showed that the low dose of TBE-31 elevated hepatic GSTA1, GSTM1, and GSTP1 by 1.4-, 1.6-, and 2.2fold, respectively; the corresponding increases by the high dose of the compound were 1.7-, 1.9-, and 3.6-fold, respectively. In stomach, the low dose of TBE-31 induced GSTA1, GSTM1, and GSTP1 by 2.  This type of analysis showed that the tricyclic double Michael acceptor TBE-31 is an exceedingly potent inducer that elevates NQO1 at low nanomolar concentrations (D m ϭ 1.1 nM) (Fig.  4A). Thus, TBE-31 is 300-times more potent than is sulforaphane (D m ϭ 290 nM), one of the most potent naturally occurring inducers, and even one of the monocyclic analogs (MCE-1) is remarkably potent with a Dm value (34 nM) in the low nanomolar range. Notably, TBE-31 is 2-fold more potent than the pentacyclic triterpenoid CDDO (D m ϭ 2.5 nM), and only 4-fold less potent than TP-225 (D m ϭ 0.27 nM), which is the most potent inducer known to date. Nonetheless, unfavorable in vivo pharmacological properties of TP-225, the reasons for which are not presently understood, have hindered its development as a drug. The tricyclic TBE-9 (D m ϭ 20 nM), which only differs from TBE-31 by the absence of an acetylenic moiety at the juncture between rings B and C, is ϳ20-fold weaker than TBE-31 (D m ϭ 1.1 nM). The importance of the acetylenic substituent is also demonstrated by comparing the inducer potencies of the monocyclic MCE-1 (D m ϭ 34 nM) which is nearly 5-fold more potent than is MCE-2 (D m ϭ 150 nM). Thus, we infer that the acetylenic group further enhances the reactivity of the Michael acceptor functions.
Both TBE-31 and TBE-9 have two highly activated Michael acceptors: one in ring A, and another in ring C. The individual contributions of each Michael acceptor were determined by examining the inducer potencies of the monocyclic derivatives MCE-5 (representing ring A of the TBEs), and MCE-1 and MCE-2 (representing ring C). MCE-1 (D m ϭ 34 nM) is ϳ30-fold more potent than MCE-5 (D m ϭ 1.1 M), indicating that the Michael acceptor in ring C has a much greater impact on the inducer potency than does the one in ring A. Notably, the simultaneous presence of the two Michael acceptors on a single molecule has a profound effect on the inducer activity: TBE-31 is ϳ30-fold more potent than MCE-1 and ϳ1000-fold more potent than MCE-5. The much greater inducer potency of the tricyclic bis(cyano enone)s relative to their corresponding monocyclic counterparts, led us to hypothesize that the electrophilic centers in ring A and ring C could act synergistically to induce NQO1. To test this possibility, the cells were exposed to either MCE-1, MCE-2, or MCE-5 individually, or to a combination of either MCE-1 and MCE-5, or MCE-2 and MCE-5, each at equimolar ratios. Interestingly, in this assay, the inducer potencies of MCE-1 alone (D m ϭ 37 nM), or of the combination of MCE-1 and MCE-5 (D m ϭ 52 nM) were very similar (Fig. 4B). The mixture of MCE-2 and MCE-5 (D m ϭ 184 nM) had ϳ2-fold higher inducer potency than did MCE-2 (D m ϭ 320 nM) alone. These results indicate that the activity of the weaker ring A inducer MCE-5 was "overpowered" by the presence of the more potent ring C inducers. Thus, in the case of MCE-1, which is 50-fold more potent than MCE-5, essentially all the inducer activity could be attributed to the more potent monocyclic cyano enone. The additive effect that was observed when the cells were exposed to a mixture of MCE-2 and MCE-5, in comparison to MCE-2 alone, can probably be attributed to the much smaller difference in inducer potencies between MCE-2 and MCE-5 (7-fold) than the difference between MCE-1 and MCE-5 (50-fold). Most notably, these findings indicate that for high inducer potency, the two electrophilic centers must be positioned at a specific spatial orientation and distance relative to one another. Overall, these structure-activity correlations clearly establish that: (i) both Michael acceptors (in ring A and ring C) contribute to the inducer activity; (ii) the contribution to inducer potency of ring C is much greater than that of ring A; (iii) the simultaneous presence of both Michael acceptors in ring A and in ring C, as part of a contiguous three-ring structure, is essential for high inducer potency; and (iv) the addition of an acetylenic moiety at the junction between ring B and ring C leads to a marked increase in inducer potency.
Cyano Enones Induce ARE-dependent Gene Expression Induction Requires Transcription Factor Nrf2-ARE sequences are present in the promoter regions of the genes coding for NQO1 and GST. To test whether induction occurs via these sequences we examined the ability of TBE-31 and its monocyclic analogs to elevate the expression of a luciferase reporter gene under the control of 8 tandem ARE sequences by using stably transfected AREc32 cells (39). Exposure to each of the compounds resulted in a significant concentration-dependent increase in chemiluminescence (Fig. 5A). Of greatest interest, however, was that the potency order rank in elevating ARE-dependent gene expression in the AREc32 cells was similar to that for inducing NQO1 in Hepa1c1c7 cells, indicating that induction is ARE-dependent. In addition, induction is dependent on transcription factor Nrf2, because it did not occur in MEF isolated from nrf2-knock-out mice (Fig. 5B), in sharp contrast to the robust dose-dependent induction that was observed in the corresponding wild-type cells. Again, the potency order rank of the cyano enones in inducing NQO1 in the wild-type MEF cells was identical to that in Hepa1c1c7 cells.
Cyano Enones React with Cysteine Residues of Keap1, the Protein Sensor for Inducers-The presence of inducers in the cell is recognized by sensor cysteines of Keap1. When TBE-31 is added to a solution of dithiothreitol (DTT) in Tris-Cl buffer (pH 8.0), there is a red shift in the UV spectrum of TBE-31 from 242 to 268 nm, with a slight reduction in intensity (supplemental Fig. S2) (29). This spectral change is reminiscent of that observed in the reaction of the CDDO analog, TP-225 (supplemental Figs. S1 and S3) with thiols (17). The UV spectrum of the reaction product of TBE-31, but not of TP-225, with DTT, however, has a new absorption maximum at 345 nm. Thus, the difference spectrum (i.e. the spectrum of the reaction product) is characterized by two absorption maxima at 268 and 345 nm. Spectroscopic examination of the reaction of TBE-31 with purified murine recombinant Keap1 revealed a virtually identical difference spectrum (Fig. 6A) indicating that TBE-31 reacts with cysteine residues of Keap1. The stoichiometry of the reaction was investigated by successively adding small aliquots of TBE-31 to a solution of Keap1. With each aliquot of TBE-31 added, there was a proportional increase in absorbance at 345 nm until the concentration of the TBE exceeded that of Keap1 by a factor of ϳ4 -5 (Fig. 6B). It is tempting to conclude that TBE-31 reacts with four to five cysteine residues of Keap1, but in view of the undoubted complexity of reactivities of the many cysteine residues of Keap1 (25 and 27 in the murine and the human homologs, respectively), this inference may not be correct. Notably, the stoichiometry of the reaction of TP-225 (17) (in which ring A is identical to that in TBE-31, but the C rings differ) with Keap1 is 2:1, indicating that both ring A and ring C of TBE-31 participate in modification of the sensor cysteines. The number of modified cysteines of Keap1 reacting with TBE-31 was reduced by half when the protein was first incubated with an excess of sulforaphane, which binds to cysteine sulfhydryls of Keap1 (3,11), indicating competition between the two inducers, and confirming that the reaction with TBE-31 results in a Michael addition to cysteine residues of Keap1. Because the Michael addition of sulfhydryl groups to cyano enones is reversible (40), it was not possible to identify which cysteine residues of Keap1 were modified.
In determining the spectroscopic properties of the reaction product of TBE-31 with Keap1, we reasoned that the UV absorption at 268 nm reflects the formation of a Michael adduct at the C-5 position in ring A of TBE-31 with the sulfhydryl group of a cysteine, indeed a similar absorption is observed in the reaction product of Keap1 with the CDDO analog TP-225. Given that the absorption at 345 nm is unique for the Michael adduct of Keap1 with the tricyclic compound bearing a Michael acceptor in ring C, we concluded that it must be a consequence of the formation of a Michael adduct at the C-1 position in ring C. This conclusion was unequivocally proven by spectroscopic examination of the reaction of Keap1 with the MCEs. Thus, the difference spectrum at the longer (Ͼ300 nm) wavelengths ( max 332 nm) derived from the Michael adduct of Keap1 with MCE-1 is very similar to that of Keap1 with TBE-31 (compare Fig. 6, A and C). The difference spectrum at the shorter (Ͻ300 nm) wavelengths ( max 268 nm) derived from the Michael adduct of Keap1 with MCE-5 closely resembles that of the product of Keap1 with either TP-225 (17) or with TBE-31 (compare Fig. 6, A and D). Comparison of the spectral intensities of the reaction products of Keap1 with the monocyclic cyano enones indicates that the reactivity of MCE-1 with Keap1 is higher than that of MCE-5, in close agreement with the difference in their inducer potencies in the NQO1 assay.
Because the Michael adduct with a sulfhydryl group in ring C exists predominantly in an enol form, we considered the possibility that the absorption maximum at 345 nm is due to ionization of the enol. Because Keap1 cannot tolerate drastic changes in pH, we tested this possibility using a model reaction, that is, by use of DTT and TBE-31, the product of which, as shown above, has spectral characteristics identical to the reaction product of Keap1 and TBE-31. First, the product was formed at pH 8.0 by mixing equimolar amounts of DTT and TBE-31. Then, small aliquots of HCl were titrated into this solution resulting in progressive decrease in pH: from 8.0 to 6.0 to 4.0, and finally to pH 2.0. The absorbance at 345 nm decreased as the pH fell, and was almost fully suppressed at pH 2.0 (Fig. 7A). Conversely, addition of base greatly enhanced the absorbance at 345 nm, consistent with formation of an enolate. Moreover, the suppression of absorbance at pH 2.0 was fully reversed by raising the pH to 9.0. This pH titration experiment led us to conclude that the absorbance at 345 nm is due to the ionization of the enol resulting from the formation of the Michael adduct in ring C of TBE-31. Further support for this conclusion was obtained by the synthesis and examination of the UV spectra of the reduced monocyclic analogs dihydro-MCE-1 and dihydro-MCE-5. Thus, the UV spectrum of dihydro-MCE-1 (Fig. 7B) has two characteristic absorption maxima, one at 224 nm, and the other, at 332 nm. Notably, the absorption at 332 nm is enhanced by addition of NaOH, consistent with enolate formation. In contrast, dihydro-MCE-5 (Fig. 7C) only absorbs at the shorter wavelengths with a maximum at 265 nm. Thus, the sulfhydryl addition reaction of TBE-31 to Keap1 can be envisioned as shown in Scheme 1.

DISCUSSION
The Keap1/Nrf2/ARE cytoprotective pathway is a target for the prevention and treatment of cancer, cardiovascular and neurodegenerative diseases. By activating this pathway, small-molecule inducers trigger a broad cytoprotective response. Its versatility and duration (in the order of several days) provides a very efficient and long-lasting protection against a wide variety of damaging agents, including exogenous carcinogens and their metabolites, as well as endogenous oxygen-and nitrogen-based oxidants. Conversely, deficiencies in Nrf2-dependent cytoprotective genes, and especially in transcription factor Nrf2 itself, are associated with increased disease risk and accelerated pathogenesis. Inducers of the Keap1/Nrf2/ARE pathway belong to 10 distinct chemical classes, and include Michael acceptors, oxidizable diphenols, isothiocyanates, conjugated polyenes, vicinal dimercaptans, trivalent arsenicals, and heavy metals (5). The enone functionality was early on recognized as a feature of inducers and this recognition led to the identification of Michael acceptors (14) as one major inducer class. Both naturally occurring and synthetic enones were found to up-reg- ulate NQO1 in the Hepa1c1c7 cell-based bioassay long before the discovery of the Keap1/Nrf2/ARE pathway. Examples include the phenylpropenoid cinnamic acid esters, chalcones and curcuminoids, the avicins, and the cyclopentenone prostaglandins. Inducers that contain two enone functionalities, which we designate as "double Michael acceptors," are invariably much more potent than their corresponding "single Michael acceptor" counterparts. Thus, bis(2-hydroxybenzylidene)acetone (supplementa1 Fig. S1) is Ͼ30-fold more potent than trans-4-(2-hydroxyphenyl)-3buten-2-one; and curcumin is Ͼ10-fold more potent than ferulic acid methyl ester (41,42). As found in this study, exceedingly high inducer potency of a double Michael acceptor can be achieved by a combination of: (i) a doubly activated electrophilic carbon, i.e. by its simultaneous conjugation to both nitrile [C'N] and carbonyl [CϭO] moieties, and (ii) by a restricted spatial arrangement of the two electrophilic centers such that they are part of a contiguous and rigid system within the same molecule. This fixed geometric requirement strongly suggests that both electrophilic carbons react with two cysteines within Keap1. Indeed, the first attempts to identify "the most reactive sensor cysteine" of purified recombinant murine Keap1, by using dexamethasone mesylate, an irreversible sulfhydryl reagent, as a probe, led to the consistent isolation not of a single, but of multiple covalent adducts (3).
Whereas it is well recognized that cyano enones are highly susceptible to nucleophilic attack by sulfhydryl agents to form covalent Michael adducts, perhaps less appreciated is that such adduct formation is usually easily reversible. NMR variable temperature studies previously revealed that the pentacyclic triterpenoid CDDO, which has an identical cyano enone ring A to the monocyclic MCE-5 and the tricyclic TBE-31 used in this study, readily forms covalent adducts with DTT, but that these adducts quickly undergo elimination as the temperature is raised (40). Similar NMR studies demonstrated that formation of the Michael adduct of MCE-1 with DTT is also reversible, even though the reactivity of MCE-1 is very high (30). The chemical reversibility of these reactions has very significant biological implications: (i) it enhances inducer bioavailability; (ii) it allows reversible cysteine modifications of the protein sensor Keap1, which does not need to be permanently inactivated (and possibly subsequently destroyed), but could be easily regenerated without requiring de novo protein synthesis; (iii) it leads to a pulse of activation, rather than constitutive up-regulation of the pathway; and (iv) it may explain, at least in part, some of the reasons why TBE-31 is such a potent inducer in vivo, able to react with Keap1 at nanomolar concentrations despite the presence of millimolar concentrations of glutathione. An analogy could be made with the isothiocyanate class of inducers. The isothiocyanates are taken up by the cell predominantly, if not entirely, through glutathione conjugation reactions, and furthermore, cellular glutathione S-transferases promote the uptake and accumulation of isothiocyanates by enhancing the conjugation reaction (43).
In conclusion, the design, synthesis, and use of novel monocyclic cyano enones allowed us to establish the structural features that are responsible for the activity of an extremely potent tricyclic acetylenic bis(cyano enone) in inducing cytoprotective enzymes via the Keap1/Nrf2/ARE pathway in cells and in vivo. Furthermore, these monocyclic derivatives revealed the contributions of each of the two Michael acceptors present on the tricyclic compound, as FIGURE 7. Formation of the reaction product of ring-C cyano enones with DTT is accompanied by enolization. A, 50 M TBE-31 was added to a solution of 50 M DTT in 20 mM Tris-HCl/0.005% Tween 20 (pH 8.0) at 25°C and the absorption spectrum was recorded against a DTT blank. The pH of the solution was lowered by addition of small aliquots of HCl. Once the pH reached pH 2, NaOH was added to raise the pH to 9. Absorption spectra of 100 M dihydro-MCE-1 (B) and 100 M dihydro-MCE-5 (C) before (black lines) and after (gray lines) addition of 10 mM NaOH in phosphate-buffered saline (pH 7.4), containing 1% EtOH (v/v), at 25°C. SCHEME 1. Proposal for the reaction of TBE-31 with Keap1.
well as the acetylenic moiety, to the reactivity with the protein sensor for inducers Keap1 and to the spectral characteristics of the reaction product.