Obtusaquinone: A Cysteine-Modifying Compound That Targets Keap1 for Degradation

Abstract

We have previously identified the natural product obtusaquinone (OBT) as a potent antineoplastic agent with promising in vivo activity in glioblastoma and breast cancer through the activation of oxidative stress; however, the molecular properties of this compound remained elusive. We used a multidisciplinary approach comprising medicinal chemistry, quantitative mass spectrometry-based proteomics, functional studies in cancer cells, and pharmacokinetic analysis, as well as mouse xenograft models to develop and validate novel OBT analogs and characterize the molecular mechanism of action of OBT. We show here that OBT binds to cysteine residues with a particular affinity to cysteine-rich Keap1, a member of the CUL3 ubiquitin ligase complex. This binding promotes an overall stress response and results in ubiquitination and proteasomal degradation of Keap1 and downstream activation of the Nrf2 pathway. Using positron emission tomography (PET) imaging with the PET-tracer 2-[¹⁸F]fluoro-2-deoxy-D-glucose (FDG), we confirm that OBT is able to penetrate the brain and functionally target brain tumors. Finally, we show that an OBT analog with improved pharmacological properties, including enhanced potency, stability, and solubility, retains the antineoplastic properties in a xenograft mouse model.

Graphical Abstract

The transcription factor Nrf2 (Nuclear factor erythroid 2 (NFE2)-related factor 2) plays a key role in maintaining cellular homeostasis in response to oxidative stress by regulating the expression of antioxidant response element (ARE) dependent genes.¹ Keap1 (Kelch-like ECH-associated protein 1) is recognized as a predominant negative regulator of Nrf2 and functions as a substrate adaptor protein for the ubiquitin ligase CRL3 (cullin 3 (CUL3)-RING ubiquitin ligase). During homeostasis, Keap1 recruits Nrf2 to CUL3 thereby promoting its ubiquitination and subsequent proteasomal degradation. Nrf2 has been recognized to exert either protumorigenic or antitumorigenic properties.² This apparent contradiction can be rationalized by differences in the cell state and the functional dependence on Nrf2 activation. Oncogenic activity is generally associated with constitutive activation of Nrf2 caused either by overexpression or somatic mutations of Nrf2 and/or other regulatory proteins.^3,4 Tumor suppressor activity, in contrast, is linked to transient activation of Nrf2, e.g., by small molecules.³ In this context, the Nrf2-Keap1 module has been validated and successfully pursued as a target for small molecules that disrupt Nrf2 binding and/or induce the dissociation of Keap1 from CUL3.^5,6 These efforts have identified several inhibitor classes, including cysteine-reactive natural products and synthetic compounds, that bind Keap1 and activate the Nrf2 pathway.⁷

We have previously identified the quinone methide obtusaquinone (OBT) as a potent antineoplastic agent with selectivity over normal cells for glioblastoma (GBM) and several other cancer types.⁸ Within the scope of this research, we have demonstrated that OBT treatment increases the production of reactive oxygen species (ROS) and induces DNA damage, leading to apoptotic cell death. However, the molecular mode of action of OBT, its medicinal chemistry, and in vivo pharmacology have not been understood, which has impeded preclinical development. We here developed novel OBT analogs with improved pharmacological properties and show that OBT is a thiol-reactive compound that reacts reversibly with cysteine residues and particularly binds to Keap1, leading to CRL3-mediated autoubiquitination and proteasomal degradation of Keap1 thereby activating the Nrf2 pathway.

RESULTS

OBT Forms Reversible Covalent Adducts with Thiols.

OBT features a 2-hydroxy-p-quinone methide, a moiety found in other natural products such as celastrol,⁹ which efficiently reacts with thiol nucleophiles including cysteine side chains to form substituted catechols (Supplemental Figure 1A). It has previously been shown that the desmethoxy analog of OBT SI1 (Supplemental Figure 1B) reacts with glutathione (GSH) in aqueous buffer to form four distinct addition products, corresponding to the diastereomers derived from direct SI2a,b and vinylogous SI3a,b addition.¹⁰ To demonstrate that OBT retains the ability to form sulfhydryl adducts, we incubated OBT with β-mercaptoethanol (BME) in ethanol and found that BME was readily added to OBT, preferentially (10:1) forming the direct addition product over the vinylogous addition product (AF20, Figure 1A; AF20a and AF20b, Supplemental Figure 1C). To investigate the reversibility of the thionucleophile addition to OBT, we incubated AF20 in the presence of 5-fold excess cysteamine. Monitoring the reaction mix by LC/MS showed the formation of the corresponding amine functionalized analogs SI4a and SI4b, demonstrating that the addition of thiols is reversible or that the substitution proceeds through an S_N2′ mechanism and that the corresponding adducts exist in a dynamic equilibrium (Supplemental Figure 1C,D).

Figure 1.

Characterization of OBT and its derivatives as reversible cysteine-modifying drugs. (A) Chemical structure of OBT and novel analogs. (B) MDA-MB231 cells were treated with different doses of OBT or analogs in 4-replicates, and cell viability was measured 4 days later and expressed as % of vehicle control. *P < 0.05 Student’s t-test versus OBT. (C) An overview of the mass spectrometry-based approach to interrogate OBT binding to cysteine residues. OBT is given to the protein and will or will not react with free thiol groups. In the second step, an alkylating reagent, iodoacetamide (IAA), is added to the mixture, which will react with free thiols if they are not blocked by OBT. In the third step a reductant, dithiothreitol (DTT), is added to the mixture. If OBT binding is reversible under these reductive conditions, free thiol groups will be available for reaction with another alkylating reagent, N-ethylmaleimide (NEM), added in the last step. The combined reaction product is digested, and peptides are subjected to mass spectrometry followed by multiplexed quantitative proteomics to determine for each cysteine if OBT did bind to it and if this binding can be reversed; results are compared to two controls: (1) sample with protein alkylated only with IAA and (2) sample alkylated only with NEM. Each cysteine-containing peptide will reveal a certain intensity pattern across the three samples allowing for the determination of site-specific binding activity of OBT. (D) Potential outcomes for the OBT-treated sample in the differential alkylation reaction process. (E) Multiplexed mass spectrometry is used to quantify peptides containing only cysteine residues generated from a consecutive trypsin and LysC digest across all samples. Quantifying IAA and NEM modified peptides across all three samples produce unique patterns of peptide intensities for each of the three options: (i) OBT does not bind, (ii) OBT binds irreversibly, and (iii) OBT binds reversibly. (F) Differential alkylation was used on bovine catalase to study OBT-binding properties. The IAA conjugate intensity patterns showed that OBT reacted with all three identified cysteines (Cys-232, −377, and −460). The NEM-conjugate patterns showed that binding was irreversible for cysteines 232 and 460 but reversible for cysteine 377 under these experimental conditions.

Next, we explored if the reactivity with thiols could be exploited for the development of novel OBT analogs with improved pharmacological properties to overcome the limited solubility of OBT in aqueous media and to buffer the abrupt oxidative stress as a result of rapid depletion of the intracellular GSH pool caused by thiol-reactive compounds.¹¹ We have previously shown that addition of GSH or N-acetyl cysteine (NAC) reduces the activity of OBT in cell culture, likely by extracellular scavenging of OBT.⁸ Based on our hypothesis that the ability to react with cysteine side chains is critical for OBT activity, we postulated that prodrugs designed to liberate OBT or analogs that retain the ability to react with cysteines would yield improved inhibitors, including compounds with enhanced solubility, while mitigating the general toxicity as a result of GSH depletion. AF20, the adduct of OBT and BME, which is more soluble and liberates OBT in PBS, potently killed patient-derived glioma stemlike cell (GSC) neurospheres, while exhibiting lower toxicity toward primary human astrocytes (HA) (Supplemental Figure 2A).⁸ As previously observed with OBT, the activity of AF20 is reversed in the presence of NAC (Supplemental Figure 2B). These results are consistent with our proposed mechanism that AF20 converts to OBT as an active compound and that NAC may function not only as a ROS antagonist but also as a direct scavenger of OBT.

To block the direct reversibility observed with AF20, we speculated that acetylation of the catechol would stabilize the thiol adduct in aqueous media but predicted that the phenolic esters would be cleaved intracellularly allowing for intracellular release of OBT via AF20 or following displacement by an S_N2′ mechanism (see the Supporting Information). Treatment of AF20 with 2 equiv and 3 equiv of acetic anhydride yielded AEN36 Bis and AEN36 Tris (Figure 1A), respectively. Both compounds demonstrated increased stability and increased potency on different cancer cells as compared to OBT (Figure 1B and Supplemental Figure 2C).

To confirm reversible cysteine modification by OBT, we established a mass spectrometry-based approach that would allow the identification of specific cysteine side chains in native proteins that covalently react with OBT and differentiate reversible from irreversible interactions. The experimental setup is outlined in Figure 1C,D, and the potential outcome is outlined in Figure 1E. The method is based on serial exposure of a protein to OBT, the cysteine-alkylating reagents iodoacetamide (IAA) to monitor derivatization with OBT, to dithiothreitol (DTT) to probe reversibility of OBT binding, and to another alkylating reagent, N-methylmaleimide (NEM), to allow a readout of probing the reversibility. In two parallel reactions, OBT is replaced by IAA and NEM. These reactions are used as standards to enable a final reaction read-out by multiplexed quantitative mass spectrometry.^12,13 We established a rule set to allow an unambiguous interpretation of the data (Supplemental Table 2) and assigned three different binding types for OBT: (1) irreversible binding, with OBT and IAA channels at an intensity at least 2-fold lower than the NEM channel and the OBT channel not being significantly (p ≤ 0.01) higher than the IAA channel intensity; (2) partially reversible binding, as irreversible binding but with an OBT channel intensity significantly higher than the IAA intensity; (3) undefined binding, the IAA channel intensity is at least as high as the NEM channel intensity. We performed this experiment using bovine catalase, a cysteine-rich antioxidant (Supplemental Figure 2D), and found that OBT binds to all detected peptides with cysteine residues (Figure 1F; Supplemental Tables 1 and 2). For Cys-377, where the cysteine is followed by a proline in the protein sequence, we observed reversible binding, while the two other monitored cysteine residues (Cys-232 and Cys-460) showed tight or irreversible binding under the tested reaction conditions (Figure 1F). Unexpectedly, we observed a high IAA signal in the NEM-conjugate, which could be attributed to deviations in the experimentally determined and predicted intensity patterns, likely due to disulfide bonds reduced and realkylated during the experiment. These results show that OBT is an effective cysteine-alkylating reagent and that the alkylation is reversible but might depend on the protein structure or amino acids adjacent to the cysteine residue, binding affinity, and/or dissociation kinetics. It is to be noted that reversibility was determined under the condition of 2-fold excess of DTT over alkylating reagents and that OBT binding may be affected differentially under other conditions.

OBT Activates the Nrf2 Pathway in Vitro and in Vivo.

To gain better insight into the molecular mechanism of OBT and its effect on the global proteome, we implemented multiplexed quantitative mass spectrometry-based proteomics using the isobaric labeling strategy with Tandem Mass Tag (TMT) reagents and the SPS-MS3 method.^12,14 Global proteomics analysis in two different patient-derived GSC specimens at 20 h after treatment with a subtoxic dose of OBT identified heme oxygenase 1 (HMOX1; HO1) as the top upregulated protein in both lines (Figure 2A,B; Supplemental Table 3). Gene ontology category analysis of the most upregulated proteins after treatment (using the DAVID bioinformatics platform) showed that the most significantly enriched category was an “oxidation-reduction process” containing the upregulated proteins HMOX1, PHS2, GLRX1, VKORL, BLVRB, and CDO1. To further explore the functional network of these proteins, we interrogated the STRING database for high-confidence direct interactors of these proteins (white circles; Figure 2C) and found a network of densely connected proteins containing five of the upregulated proteins (red) as well as Nrf2 (green), a master regulator of oxidative damage response, and Keap1 (green), a substrate adaptor protein of the E3-ligase regulating the ubiquitin-mediated degradation of Nrf2 (Figure 2C). A strong increase in HO1 mRNA expression in response to OBT was detected across different cancer cells (Figure 2D and Supplemental Figure 3A). The Nrf2 target genes NQO1 and TXNRD2 were also upregulated following OBT treatment (Supplemental Figure 3B). To further support these findings independently, we designed a functional ARE luciferase reporter which was strongly activated following treatment with subtoxic doses of OBT, even higher than the positive e control tert-butylhydroquinone (tBHQ), a potent activator of Nrf2¹⁵ (Supplemental Figure 3C,D). Importantly, OBT-induced ARE activation and loss of cell viability were completely reversed by the addition of various thiol nucleophilic antioxidants including NAC, dithiothreitol (DTT), and GSH but only partially reversed by Trolox, an antioxidant that is devoid of thiol groups (Figure 2E and Supplemental Figure 3E).

Figure 2.

OBT activates Nrf2 pathway in vitro and in vivo. (A−C) Quantitative mass spectrometry-based proteomics post-OBT treatment using 10-plexed tandem mass tags (TMT) to simultaneously map protein concentration changes of 7,904 proteins in two GSCs (GBM8 and BT07), either treated with OBT (triplicates) or untreated (duplicates) for 20 h. (A) Heat map derived from unsupervised hierarchical clustering of the data. (B) Relative protein concentration differences between OBT-treated and control cells. (C) Functional network of upregulated proteins following OBT treatment using the STRING database showing high-confidence direct interactors with HMOX1 (white circles), protein connection containing five of the proteins upregulated in the OBT-treated samples (red) as well as Nrf2 (green). (D) HO1 mRNA levels were determined by qRT-PCR (normalized to GAPDH) in MDA-MB-231 cells treated with the indicated doses of OBT for 8 h. (E) U87 cells stably expressing Gaussia luciferase under the control of ARE response elements (ARE-Gluc) and the constitutively active Vargula luciferase (Vluc under control of the SV40 promoter for normalization of cell number) were treated with OBT and different antioxidants. Sixteen h post-treatment, aliquots of the conditioned medium were assayed for Gluc and Vluc activities, and data were expressed as the ratio of Gluc/Vluc, normalized to the control (set at 1). (F) Mice carrying fat pad MDA-MB231 tumors were treated with a daily dose of either DMSO or OBT (7.5 mg/kg, intraperitoneally) for four consecutive days. At day 5, tumors were removed, and RNA was extracted and analyzed for HO1 mRNA by qRT-PCR (normalized to GAPDH). *P < 0.05; **P<0.01 Student’s t-test versus control. (G) Mapping of OBT covalent binding to cysteine residues in Keap1. OBT was found to bind covalently to all identified cysteine sites, and this modification was found to be partially reversible at two sites (Cys151 and Cys434).

Finally, we confirmed OBT-mediated Nrf2 activation in vivo using a breast cancer mouse model generated by injecting MDA-MB-231 into the fat pad of nude mice.⁸ Treatment with OBT (7.5 mg/kg for four consecutive days) resulted in 7-fold upregulation of HO1 transcripts in the tumor, consistent with our findings in cell culture (Figure 2F). Taken together, these results confirm that OBT acts as an inducer of Nrf2 both in culture and in vivo.

OBT Is a Cysteine-Modifying Drug Targeting Keap1.

Several thiol reactive Nrf2 activators have been shown to bind Keap1, suggesting that OBT could function in a similar fashion.¹⁶ To investigate if Keap1 is also an OBT target, we applied the mass spectrometry-based approach described in Figure 1 to Keap1. Similar to the experiment with bovine catalase, we found OBT to bind covalently to all identified cysteine sites, thus confirming that this compound can directly bind to Keap1. Importantly, under our experimental conditions, we found this modification to be partially reversible at Cys151 located in the BTB domain (CUL3 binding site) and Cys434 located in the Kelch domain (Nrf2 binding site) of Keap1 (Figure 2G; Supplemental Figure 4 and Supplemental Tables 4 and 5).

Next, we evaluated the effect of OBT treatment on cells following stable downregulation of Keap1. Silencing of Keap1 with shRNA (shKeap1) expectedly resulted in stabilization of Nrf2, leading to a major increase in ARE reporter activity (Figure 3A,B and Supplemental Figure 5A). There was no strong potentiation of OBT-mediated ARE activation in cells expressing shKeap1 as compared to a nontargeting shRNA (shCtrl), detected using the ARE reporter (Figure 3A,B and Supplemental Figure 5B) and mRNA expression of HO1 and NQO1 (Supplemental Figure 5C). Further, silencing of Keap1 decreased cell death following treatment with OBT (Figure 3C and Supplemental Figure 5D). These data suggest that either Keap1 expression is necessary for binding of OBT to Keap1 cysteine residues, thus stabilizing and activating Nrf2, or that a strong activation of Nrf2 protects against OBT-mediated cytotoxicity.

Figure 3.

OBT targets Keap1 and induces its degradation. (A) Western blot analysis of Keap1 protein in U87 cells stably expressing shCtrl or shKeap1. (B) ARE-Gluc activity (normalized to Vluc) 16 h post-OBT treatment in U87 glioma cells expressing a nontargeting shRNA (control; shCtrl) or shKeap1. (C) U87 cells stably expressing shCtrl or shKeap1 were treated with different doses of OBT, and cell viability was measured after 4 days. (D) U87 cells were treated with 1 μM OBT, and cell lysates were collected at the indicated time points and analyzed for Keap1 protein levels by Western blotting with β-actin as loading control. (E) Keap1 protein levels in U87 cells treated with the indicated doses of OBT for 8 h. (F) 293T cells transfected to express HA-tagged Keap1 or Keap1ΔBTB were treated with OBT (2 or 4 μM) or control. Cell lysates were collected after 6 h, and immunoblotting was performed using anti-Keap1 and anti-HA antibodies. (G) 293T cells expressing HA-keap1 were treated with OBT (2 μM) and/or MG132 (10 μM) for 6 h before immunoprecipitation with anti-HA antibody. Ubiquitination was determined using anti-ubiquitin antibody. (H) MDA-MB-231 cells (wild-type) or the same cells stably expressing shCtrl or shCUL3 were transfected to express HA-Keap1 and treated with OBT (3 μM), CHX (3 μg/mL), and/or MG132 for 6 h. Keap1 levels were detected using anti-HA. (I) Mice-bearing patient-derived GBM8 GSC tumors were treated with 3 doses of OBT (10 mg/kg within a 24 h period) or vehicle control (n = 5/group) and imaged 4 h after the last OBT administration. Representative FDG-PET-CT scans (left) and signal quantification (right) showing that OBT leads to a 50% decrease in FDG tumor uptake, determined using tumor to muscle ratio with muscle as the background. *P < 0.05; **P < 0.01 Student’s t-test. (J) Schematic representation of the proposed mode of action of OBT-mediated degradation of Keap1, resulting in subsequent nuclear translocation of Nrf2 to bind ARE and activate transcription of downstream targets.

Numerous anticancer drugs have been shown to activate Nrf2.¹⁶ We tested if reactive electrophiles and oxidants known as transient activators of Nrf2 could induce cell death when added to breast and brain cancer cells, similar to OBT. We first confirmed ARE-inducing properties of the triterpenoid CDDO-Me (2-cyano-3,12-dioxoolean-1,9-dien-28-oic acid-methyl ester),¹⁷ currently being clinically tested for the treatment of leukemia and solid tumors as well as other diseases. U87 and MDA-MB-231 cells expressing ARE-Gluc reporter treated with CDDO-Me showed increased reporter activity (Supplemental Figure 6A). Higher doses of this compound led to a marked decrease in cell viability in U87 cells (Supplemental Figure 6B). Treatment of U87 cells with additional Nrf2 activators, cinnamaldehyde,¹⁸ diethyl fumarate,¹⁹ and sulforaphane²⁰ also resulted in increased ARE activity and a moderate decrease in U87 cell viability at the doses tested (Supplemental Figure 6C,D). When combined with OBT, all three compounds showed increased cytotoxicity (Supplemental Figure 6E). These results suggest that increased electrophile or oxidant concentrations are likely to cause further cysteine modifications in the cellular proteome along with a stress response evident by an upregulation of Nrf2 signaling, thus increasing cytotoxicity.

OBT Promotes the Degradation of Keap1.

We next asked whether OBT-mediated covalent modification of Keap1 affects stability of this protein. Immunoblot analysis showed a time- and dose-dependent decrease in Keap1 protein levels as early as 4 h following treatment with OBT, which were restored to physiological levels at 24 h (Figure 3D,E). Within this time frame (<24 h), we have previously shown that treatment with OBT results in early morphological changes, elevated ROS levels, and activation of apoptosis in tumor cells.⁸ To further confirm Keap1 protein degradation, we expressed HA-tagged Keap1 or Keap1 with a deleted BTB domain (Keap1ΔBTB)²¹ which is essential for Keap1 binding to CUL3 and activation of Keap1-CUL3 E3 ligase activity. We did not observe any decrease in Keap1 protein levels in cells expressing BTB-mutated Keap1 (Figure 3F), suggesting that the E3 ligase activity is required for OBT-mediated degradation of Keap1. Further, similar to Keap1 knockdown, silencing of CUL3 with shRNA prevented ARE activation following treatment with OBT and decreased cell death (Supplemental Figure 5B,D). Ectopic expression of a dominant-negative CUL3 mutant (DN-CUL3) also protected against OBT-induced cell death, further corroborating these findings (Supplemental Figure 7A). The neddylation of CUL3 is essential for its ubiquitin ligase activity.²² To determine if CUL3 activation is essential for OBT-induced cell death, we cotreated cells with OBT and the neddylation inhibitor MLN4924 and observed protection against cell death (Supplemental Figure 7B). Overall, these findings confirm that E3 ligase activity is required for targeting of tumor cells with OBT.

Since ubiquitination of Keap1 could lead to its degradation,²³ we evaluated this process following OBT treatment by immunoblotting. Indeed, OBT treatment effectively resulted in ubiquitination of Keap1 (Figure 3G). Additionally, downregulation of CUL3 or cotreatment with the proteasome inhibitor MG132 prevented OBT-mediated degradation of Keap1 (Figure 3H). Among the reactive cysteines of Keap1, C151 was found to be necessary for Keap1-alkylating ARE inducers that promote the dissociation of the Keap1-CUL3 complex, thus stabilizing Nrf2. Accordingly, mutation of C151 impairs its alkylation by electrophiles such as sulforaphane, tBHQ, or AI-1 and impairs Nrf2 activation.²⁴⁻²⁶ However, serine substitution of Cys-151 (Keap1C151S) did not protect against OBT-mediated degradation of Keap1 (Supplemental Figure 7C). Overall, these results confirm that OBT treatment leads to degradation of Keap1 and that CUL3 is an essential regulator of this process.

OBT and Its Analog Effectively Target Tumors in Preclinical Mouse Models.

Pharmacokinetic profiling of OBT in mice showed high systemic plasma clearance with terminal plasma half-life of 24 min following intraperitoneal injection (Supplemental Figure 8A). Furthermore, we found that OBT efficiently penetrates the intact blood-brain barrier (BBB) (Supplemental Figure 8B,C). To confirm that OBT is able to penetrate the brain and functionally target brain tumors, we used positron emission tomography (PET) with the PET-tracer 2-[¹⁸F]fluoro-2-deoxy-D-glucose (FDG), which measures the rate of tumor glucose uptake in a mouse orthotopic GSC model. OBT-treated mice exhibited approximately a 50% decrease in FDG tumor uptake as measured by FDG-PET imaging (Figure 3I). Finally, we selected the analog (AEN36 Tris) with most improved pharmacological properties including enhanced potency, stability, and solubility and evaluated the in vivo antineoplastic effect in a breast cancer mammary fat pad tumor xenograft model. Treatment with AEN36 Tris (10 mg/kg daily for 22 days) induced a significant decrease in tumor volume, compared to the control group, as assessed by bioluminescence imaging (Supplemental Figure 8D). In summary, OBT penetrates the BBB and can be modified to enhance its solubility while retaining its antineoplastic properties.

DISCUSSION

Small molecules that react with cysteine side chains within Keap1²⁶ or target the kelch domain of Keap1²⁷ have been identified. We now provide direct evidence that the natural compound OBT activates the Nrf2 pathway by binding covalently to cysteine residues within the BTB-domain of Keap1 leading to its ubiquitination and subsequent proteasomal degradation. This directly impacts the ability of the CUL3-Keap1 ubiquitin ligase complex to degrade Nrf2, resulting in Nrf2 stabilization and downstream activation of ARE-mediated transcription (Figure 3J). It is highly likely that OBT also interacts with other thiol-rich proteins; however, our data supports that Keap1 is a major functional target for OBT and that the BTB-CUL3 ubiquitin ligase complex is required for OBT-mediated degradation. The cysteine-reactive compound likely engages other secondary targets in order to promote an overall stress response. In fact, downregulation of Keap1 was not sufficient to induce the same level of cytotoxicity observed after treatment with OBT or other Nrf2 activators, confirming this hypothesis.

The transcription factor Nrf2 is often viewed as a pleiotropic gene. Whether its activation or inhibition is beneficial for tumor treatment remains a paradox and seems to depend on various factors such as the cell type, tumor stage, and genetic aberrations within the tumor.²⁻⁴ Nrf2 has been suggested to act as a tumor suppressor, thus its activation can suppress carcinogenesis.^2,28−30 Nrf2 activation was shown to decrease tumor growth in established tumors,³¹ and several antineoplastic drugs enhance Nrf2 activity.¹⁶ In this study, we have demonstrated that, in addition to OBT, several other reactive electrophiles and oxidants known as transient activators of Nrf2 can induce cytotoxicity in tumor cells. On the other hand, activation of Nrf2 by cancer targeting drugs can also lead to unfavorable clinical outcomes because of Nrf2’s ability to enhance chemoresistance.²⁸ One plausible explanation of this paradox is that, unlike somatic mutations and oncogene-mediated signaling that promote a sustained Nrf2 activation accompanied by numerous adaptation mechanisms, pharmacological activation of Nrf2 is transient and does not necessarily phenocopy constitutive Nrf2 activation.^3,4 Despite these controversies, several Nrf2 activators have been developed as antineoplastic compounds in preclinical studies,³² and at least one such compound, sulforaphane, has advanced to a phase 2 clinical trial for the treatment of metastatic breast cancer (NCT02970682).

The dose−response curve of many chemopreventive agents as well as chemotherapeutic drugs is U-shaped,² resulting in opposing effects between low and high doses of the same agent. For example, synthetic oleanane triterpenoids exert chemopreventive functions at low doses but are also able to induce oxidative stress and apoptosis at higher doses.^28,33 The same model could be applied to OBT where lower doses of the compound lead to a strong Nrf2 activation, while at higher doses, OBT acts as a potent pro-oxidant that targets cancer cells leading to DNA damage and apoptosis as we have previously shown.⁸

In conclusion, we have established a mechanistic understanding of the mode of action of OBT. We show that OBT is a reversible covalent modifier of cysteine residues in Keap1, targeting it for proteosomal degradation, leading to strong activation of Nrf2. In addition, we have developed novel OBT analogs with improved in vivo activity that allow the tuning of pharmacological properties, which will facilitate the preclinical development of this compound class. Finally, given that impaired Keap1 activity and Nrf2 activation lead to increased expression of antioxidant and detoxification genes, and their role in neuroprotection,^19,34,35 we speculate that OBT and its analogs might have cross-disease applications for disorders such as diabetes, Alzheimer’s disease, and Parkinson disease;³⁶ however, this remains to be tested.

METHODS

Reagents.

OBT was purchased from Gaia Chemicals. N-Acetyl-L-cysteine, dithiothreitol, L-glutathione, tert-butylhydroquinone, N-ethylmaleimide, iodoacetamide, cinnamaldehyde, diethyl fumarate, and sulforaphane were purchased from Sigma-Aldrich. Trolox, sulfasalazine, CDDO-Me, and MG-132 were purchased from Cayman Chemical. Nontargeting shRNA control (shCtrl) and shRNA constructs targeting Keap1 (TRCN0000156676) and CUL3 (TRCN0000012778) were purchased from Sigma and packaged into lentiviral vectors using standard protocols.

Cell Culture.

MDA-MB-231 and U87 (U87-MG) cells were obtained from the American Type Culture Collection (ATCC). 293T human embryonic kidney fibroblasts were provided by Dr. Xandra Breakefield (Massachusetts General Hospital). All three cell lines were grown in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum (Gemini Bioproducts), 100U penicillin, and 0.1 mg mL⁻¹ streptomycin (Sigma). All cells were maintained at 37 °C in a humidified 5% CO₂ incubator. GSCs were obtained from tumor tissues of GBM patients following surgical resection, under approval from the corresponding Institutional Review Board. These cells have been previously characterized³⁷⁻³⁹ and were maintained in culture as neurospheres in Neurobasal medium (Gibco) supplemented with heparin (2 μg/mL; Sigma) and recombinant EGF (20 ng/mL) and bFGF-2 (10 ng/mL; Peprotech). Human astrocytes were obtained from ScienCell and cultured in Astrocytes Medium (ScienCell).

Cell Viability.

Cells were plated in 96-well plates and treated with the corresponding compounds. Cell viability was measured by adding 25 μL/well of CellTiter-Glo (Promega) followed by 10 min incubation and transfer to a white 96-well plate. Bioluminescence was quantified using a Synergy HTX multimode reader (Biotek).

Statistical Analysis.

GraphPad Prism v6.01 software (LaJolla, CA) was used for statistical analysis of all data. A p-value of less than 0.05 was considered to be statistically significant. For analysis between multiple groups, a two-tailed Student’s t-test (unpaired), ANOVA, and Tukey’s posthoc test were performed as indicated. All experiments were performed at least in 3 replicates and repeated 3 independent times.