Nitrostyrene Derivatives Act as RXRα Ligands to Inhibit TNFα Activation of NF-κB

Retinoid X receptor alpha (RXRα) plays pleiotropic roles in the biologic and pathologic processes (1, 2). Dysfunctions of RXRα are implicated in a number of diseases, such as cancer. For examples, abnormal changes of RXRα expressions or modifications are associated with the development of prostate, skin, liver, and colon cancers (3–6). Like other nuclear receptors, RXRα binds to its responsive elements to regulate gene transcription in the nucleus (7). Recently, accumulating evidence demonstrates the nongenomic actions of RXRα. For example, some apoptotic stimuli induce RXRα and Nur77 translocating from the nucleus to the cytoplasm, where they execute proapoptotic effects via association of Bcl-2 and transition of Bcl-2 from an anti- to a proapoptotic molecule (8, 9). RXR was also shown to bind to the G protein Gq in a ligand-dependent manner and impair Gq-mediated Rac activation and intracellular calcium release (10).

Several small molecules have been identified as endogenous RXRα ligands, such as 9-cis-retinoic acid (9-cis-RA) and docosahexaenoic acid, and many synthetic compounds also bind to RXRα selectively and exhibit RXRα-dependent effects (11). In general, RXRα ligands are constituted of three building blocks, the hydrophobic ring, a central polyene linker, and a polar motif such as carboxyl group (12). The ionic interactions and hydrogen bonds formed between the carboxyl group of 9-cis-RA and Arg316 residue of RXRα are essential for the binding of 9-cis-RA to the ligand-binding pocket (LBP) of RXRα (13). Recently, some nonclassical ligands have been reported to bind to RXRα in different manners. K-8008, a derivative of sulindac, binds to the surface of RXRα (14). CF31, bigelovin, and magnolol do not require Arg316 for binding to the LBP of RXRα (15–17). Thus, RXRα is subjected to modulation by diverse molecules through different mechanisms.

TNFα displays a variety of physiologic activities in a cell- and tissue context–dependent manner. It stimulates the NF-κB and Akt pathways to enhance cell survival, whereas in certain contexts, it provokes apoptotic events by activating the caspase-8–mediated death pathway (18, 19). Recently, we reported that the nongenomic action of an N-terminally truncated RXRα (tRXRα) could play a role in the crosstalk with TNFα signaling (15, 20). tRXRα produced by proteolytic cleavage of full-length RXRα is highly expressed in a variety of tumor cells and tissues (21). In response to TNFα, tRXRα interacts with the p85α regulatory subunit of PI3K, followed by the activation of Akt to promote tumor cell growth. However, whether tRXRα could crosstalk with other TNFα-dependent signal pathways and whether small molecules could modulate its activities remain unknown.

Nitrostyrene derivatives have been identified as potent anticancer agents (22, 23), whereas the underlying mechanisms are still elusive. In the current study, we demonstrated that nitrostyrene derivatives (Z compounds) could inhibit the TNFα/NFκB signaling pathway by binding to tRXRα and blocking the interactions of tRXRα with TRAF2, leading to TNFα- and tRXRα-dependent apoptosis of cancer cells.

Antibodies for RXR (sc-774), PARP-1/2 (sc-7150), c-Myc (sc-40), c-Myc (sc-789), NF-κB p65 (sc-8008), α-tubulin (sc-8035), ubiquitin (sc-9133), and cyclin D1 (sc-20044) were purchased from Santa Cruz Biotechnology; Antibodies for Flag (F1804) and β-actin (A2228), and 9-cis-RA (R4643), ATRA (R2625), dexamethasone (D1756), T0901317 (T2320), and rosiglitazone (R2408) were purchased from Sigma-Aldrich; Antibody for IκBα (ab32518) was from Abcam; antibodies for cleaved caspase-8 (#9496), p-IKKα/β (#2078), TRAF2 (#4712), and p62 (#5114) were from Cell Signaling Technology; TNFα (210-TA) was from R&D Systems; RIP1 antibody (551041) was from BD Biosciences. Propyl pyrazole triol (1426) was from Tocris Bioscience; [³H]9-cis-RA was obtained from Amersham; Z compounds were dissolved in ddH₂O or dimethylsulfoxide respectively.

Cell lines were passaged for fewer than 4 months after resuscitation and were used at the fifth through tenth passage in culture for this study. MCF-7 human breast cancer and HEK293T human embryonic kidney cells were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS (Life Technologies) in a humidified atmosphere containing 5% CO₂ at 37°C. Cell transfections were carried out by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.

HEK293T cells were cotransfected with pG5-luc reporter (Promega) together with the plasmids encoding different NR-LBDs fused with the DNA-binding domain of Gal4. One day after transfection, cells were treated with DMSO, Z compounds, or ligands specific for each nuclear receptor. After 12 hours, cells were lysed by passive lysis buffer. Firefly and Renilla luciferase activities were quantitated using the Dual-Luciferase Reporter Assay System (Promega, E1960). Transfection and expression efficiency was normalized to Renilla luciferase activity.

The human RXRα-LBD (223–462) was cloned as an N-terminal histidine-tagged fusion protein in pET15b expression vector and overproduced in Escherichia coli BL21 DE3 strain. Briefly, cells were harvested and sonicated, and the extract was incubated with the His60 Ni Superflow resin. The protein–resin complexes were washed and eluted with imidazole. The eluent was collected and concentrated to 5 mg/mL for subsequent trials.

RXRα-LBD protein was incubated with different concentrations of unlabeled 9-cis-RA, Z compounds, or their derivatives in 200 mL of binding buffer [0.15 mol/L KCl, 10 mmol/L Tris-HCl (pH 7.4), 8% glycerol, and 0.5% CHAPS detergent] at 4°C for 1 hour. [³H]9-cis-RA was added to the tubes to a final concentration of 7.5 nmol/L and a final volume of 300 mL and incubated overnight at 4°C. The RXRα-LBD was captured by nickel-coated beads. Bound [³H]9-cis-RA was quantitated by liquid scintillation counting (20).

The binding kinetics between RXRα-LBD and compounds was analyzed at 25°C on a BIAcore T200 machine with CM5 chips (GE Healthcare). RXRα-LBD (20 μg/mL in 10 mmol/L sodium acetate, pH 5) was immobilized on the CM5 chip using amine coupling procedures according to the manufacturer's instructions. A serial concentration of Z compounds ranging from 1 to 10 μmol/L were used for the experiment at a flow rate of 20 μL/minute. When the data collection was finished in each cycle, the sensor surface was regenerated with Glycine-HCl (10 mmol/L, pH 2.5). Sensorgrams were fit globally with BIAcore T200 analysis using 1:1 Langmuir binding mode.

The thermodynamic properties of compounds binding to RXRα-LBD were determined using a VP-ITC titration calorimeter (MicroCal) in phosphate buffer at 25°C. The sample cell was filled with His-RXRα-LBD (50 μmol/L in 25mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 0.1% DMSO). The compounds were diluted to a concentration of 1 mmol/L in the same buffer. The injection volumes were 2 μL each with injection time 4 seconds and a 120 second delay between each injection. The heat of dilution was obtained by injecting compounds into the same buffer and subtracted from the reaction before the fitting process.

Immunostaining, immunoblotting, coimmunoprecipitation, and GST pulldown assays were performed as described (8, 9).

Size-exclusion chromatography assays were performed on an ÄktaPurifier system equipped with a HiLoad 16/600 Superdex 200 pg column (GE Healthcare). The column was pre-equilibrated with buffer [50 mmol/L sodium phosphate (pH 7.2), 150 mmol/L NaCl) and RXRα-LBD protein was run at a flow rate of 1 mL/minute.

Nude mice (BALB/c, 4–5 weeks old) were injected subcutaneously with 100 μL MCF-7 cells (2 × 10⁶). For drug treatment, mice were administered with Z-12 (30 mg/kg) diluted in Tween80 intragastrically once a day and TNFα (120 × 10⁴ U/kg) diluted in PBS plus 3 mg/mL bovine serum albumin intratumorally every 2 days alone or in combination after 6 days of transplantation. Body weight and tumor sizes were measured every 2 days. Mice were killed after drug treatment and tumors were removed for various assessments. All manipulations involving live mice were approved by the Animal Care and Use Committee of Xiamen University (Xiamen, Fujian, China).

Data were expressed as mean ± SD. Each assay was repeated in triplicate in three independent experiments. The statistical significance of the differences among the means of several groups was determined using the Student t test.

We performed mammalian one-hybrid assay to screen for RXRα modulators using our in-house chemical library, and unexpectedly found that a nitrostyrene derivative (Z-1) selectively activated the transcriptional activity of the fusion protein Gal4-DBD-RXRα-LBD but not Gal4-DBD-RARα-LBD or Gal4-DBD-ERα-LBD (Fig. 1A and B). We then designed and synthesized a series of Z-1 derivatives designated as Z-2 to Z-12 (Supplementary Table S1). Two optimized derivatives, Z-10 and Z-12, dose dependently activated RXRα transactivation, of which Z-10 and Z-12 at a concentration of 10 μmol/L reached to similar and about 50% activity of 0.1 μmol/L 9-cis-RA, respectively (Fig. 1A and C and Supplementary Fig. S1A). Interestingly, Z-11, which is different from Z-10 only in the position of the nitrovinyl group on naphthalene, was unable to activate RXRα (Fig. 1A and C), demonstrating a strict requirement of Z compound' structure for activating RXRα. In contrast to the benzene group of Z-1, the naphthalene group of Z-10 and anthracene group of Z-12 are larger aromatic groups, suggesting that larger groups induced a more favorable RXRα conformation for transactivation. The effect of Z-10 and Z-12 on RXRα transactivation was highly selective, as neither of them significantly activated the chimeric reporters of other nuclear receptors, including RARα, RARγ, ERα, GR, LXRα, and PPARγ (Fig. 1D and Supplementary Fig. S1B).

The regulation of RXRα transactivation by Z-10 and Z-12 prompted us to study their binding to RXRα protein in vitro by ligand competition assay. Similar to unlabeled 9-cis-RA, Z-10 and Z-12 dose dependently competed with [³H]-labeled 9-cis-RA for binding to RXRα-LBD, with IC₅₀ at 0.28 and 0.81 μmol/L, respectively. Interestingly, although Z-11 did not activate RXRα, it exhibited similar competition ability as Z-10 and Z-12 with an IC₅₀ at 0.33 μmol/L, implying that Z-11 bound to RXRα but induced a distinct conformation of RXRα (Fig. 2A). Our surface plasmon resonance (SPR)-based assay indicated that Z-10 and Z-12 dose dependently bound to RXRα-LBD with K_d values of 5.74 and 1.95 μmol/L, respectively (Fig. 2B). The K_d value of Z-10 binding to RXRα-LBD measured by isothermal titration calorimetry (ITC)-based assay was 2.98 μmol/L, which was in the same order of magnitude as that measured by SPR assay (Fig. 2C). Moreover, our assays based on differential scanning calorimetry (DSC) demonstrated that the T_m values of RXRα-LBD protein were shifted higher by Z-10, Z-11, and Z-12 (Supplementary Fig. S2A). Together, these data indicate that Z-10, Z-11, and Z-12 could bind to RXRα directly.

Different from classical RXRα ligands, Z serial compounds do not possess a carboxyl group but instead a nitro group. To examine the role of the nitro group, it was replaced with a carboxyl group (Supplementary Table S2). Ligand competition assay demonstrated that all the carboxyl derivatives (Z-10-1, Z-11-1, and Z-12-1) failed to displace [³H]9-cis-RA from binding to RXRα-LBD (Fig. 2A). Similar results were determined by our DSC and ITC assays (Supplementary Fig. S2A and S2B). Consistently, the carboxyl derivatives were incapable of activating Gal4-DBD-RXRα-LBD transcriptional activity (Fig. 2D). We also replaced the nitro group of Z-10 with other functional groups, including formyl, hydroxylamine, amine, and cyano (Supplementary Table S3), and none of the derivatives (Z-10-2, -3, -4 and -5) could bind to or activate RXRα (Fig. 2E and Supplementary Fig. S2B and S2C). Taken together, these results demonstrate that the nitro group is essential for Z compounds binding to RXRα and regulating RXRα transactivation.

To investigate the effects of Z compounds on RXRα homodimer transactivation, we performed reporter assay using RXRE-luciferase reporter known to bind with RXRα homodimers (24). To our surprise, both Z-10 and Z-12 failed to induce the transactivation of RXRα homodimer. Moreover, they inhibited 9-cis-RA–induced transactivation (Fig. 3A). We then examined the effects of Z compounds on RXRα homodimer formation. The purified RXRα-LBD protein exhibited two bands on the native gel, of which the upper and lower bands represented homotetramers and homodimers, respectively. Similar to 9-cis-RA, the incubation of RXRα-LBD protein with Z-10 and Z-12 induced homodimer formation, accompanying with the reduction of tetramers. However, the homodimers induced by Z-10 and Z-12 migrated slightly slower than the control and 9-cis-RA–induced dimers (Fig. 3B and Supplementary Fig. S3A). Thus, Z-10 and Z-12 could induce distinct conformations of RXRα homodimers, which was also demonstrated by our assay of size-exclusion chromatography (Fig. 3C). This provided a possible explanation for the inability of Z-10 and Z-12 to activate RXRα homodimer despite their ability to induce homodimer formation. Unlike Z-10 and Z-12, Z-11 induced higher levels of RXRα-LBD oligomerization (Fig. 3B and Supplementary Fig. S3A), consisting with its inability of activating RXRα transactivation even binding to RXRα. As expected, the carboxyl derivatives did not show any effect on the formation of RXRα-LBD tetramer and dimer (Fig. 3B and Supplementary Fig. S3A). The distinct conformations induced by Z-10 and Z-12 were also illustrated by the different patterns of the cleaved RXRα-LBD fragments produced by limited proteolysis in the presence of Z-10, Z-12, DMSO, and 9-cis-RA (Supplementary Fig. S3B).

It was imagined that the distinct RXRα conformations induced by Z compounds resulted from their unique binding mode. Cocrystallography and mutagenesis assays have revealed that Arg316 in the LBP of RXRα is essential for 9-cis-RA binding (13, 15). We therefore examined the role of Arg316 for Z-10 and Z-12 activity by reporter assay using Gal4-DBD-RXRα-LBD/R316E, a mutant with Arg316 substituted with Glu. While 9-cis-RA failed to activate the mutant, Z-10 and Z-12 strongly activated its transactivation (Fig. 3D). Consistently, Z-10 bound to RXRα-LBD/R316E protein with a K_d value of 8.30 μmol/L (Supplementary Fig. S3D). Thus, the nitro group of Z compounds might not form ionic interactions and hydrogen bonds with Arg316, suggesting a different binding manner of Z compounds, which was further characterized using other point mutants of Gal4-DBD-RXRα-LBD (Fig. 3E and Supplementary Fig. S3C). Neither 9-cis-RA nor Z compounds activated the mutants with Cys432 substituted with Gln, Trp, or Tyr. However, mutation of Cys to Ser completely incapacitated Z-10 and Z-12 but not 9-cis-RA (Fig. 3E). Thus, Cys432 was more essential for Z compounds than for 9-cis-RA, and the point mutation of C432S could be used to distinguish 9-cis-RA from Z compounds. Consistently, Z-10 failed to bind to RXRα-LBD/C432S, C432Q and C432W (Supplementary Fig. S3D), demonstrating the crucial role of Cys432 for Z compounds binding. Sequence alignment of several nuclear receptors indicated that Cys432 was a unique residue in RXRα (Supplementary Fig. S3E), which might explain the RXRα selectivity of Z compounds. Intriguingly, substitution of Cys432 with Trp or Tyr resulted in two mutants with autoactivating ability (Fig. 3E). Structural comparison indicated that the phenol group of Tyr and the indole group of Trp are similar to the benzyl group of Z-1 and the naphthalene group of Z-10, respectively, in both the molecular size and the spatial structure (Fig. 3F). Therefore, it was imagined that the two mutations could mimic the bindings of Z-1 and Z-10, which was demonstrated by our size-exclusion chromatography assay showing that the mutant RXRα-LBD/C432W could form dimer spontaneously and our native gel electrophoresis assay showing that the similar migration rate of the mutant dimer and the Z-10–induced dimer (Fig. 3G and H).

We reported previously that K-80003, a sulindac derivative, could inhibit TNFα-induced Akt activation by preventing the interaction of tRXRα and p85α (20). Unlike K-80003, Z-10, and Z-12 could not inhibit TNFα-induced interaction of tRXRα with p85α (Supplementary Fig. S4A). Instead, Z-10 dose dependently inhibited NF-κB activation by TNFα in our reporter gene assay (Supplementary Fig. S4B). Consistently, TNFα-induced expression of NF-κB target genes, including c-myc, cyclin D1, and p62, was inhibited by Z-10 and Z-12 (Fig. 4A). Further analysis revealed that Z-10 and Z-12 significantly inhibited TNFα-induced p65 nuclear translocation, IκBα degradation, and IKKα/β phosphorylation (Fig. 4B, 4C and Supplementary Fig. S4C). We next examined whether the inhibition of NF-κB activation was RXRα dependent. Compared with control siRNA, transfection of RXRα siRNA decreased the expression of both RXRα and tRXRα, which was accompanied with the impaired effect of Z compounds on TNFα-induced IκBα degradation (Supplementary Fig. S4D). The RXRα-dependent effect of Z compounds was also illustrated by our results showing that the inhibitory effect of Z-10 on TNFα-induced IκBα degradation was stronger in MCF-7 cells expressing higher levels of RXRα and tRXRα than in H460 cells with much lower RXRα and tRXRα expression (Fig. 4D). We also examined whether the effect of Z compounds could be modulated by known RXRα ligands such as 9-cis-RA, CD3254 and UVI3003, which did not show apparent effects on TNFα-induced IκBα degradation (Supplementary Fig. S4E). When Z compounds were used together with these known RXRα ligands, the inhibitory effects of both Z-10 and Z-12 on TNFα-induced IκBα degradation was reduced (Fig. 4E). In addition, the carboxyl derivatives, Z-10-1 and Z-12-1 incapable of binding to RXRα, did not show any effect on TNFα-induced IκBα degradation (Supplementary Fig. S4F). Taken together, these data demonstrate that Z-10 and Z-12 inhibit TNFα activation of the NF-κB signaling pathway in an RXRα/tRXRα–dependent manner.

To determine the molecular mechanism by which Z compounds inhibited TNFα activation of NF-κB, we first analyzed the role of RXRα and tRXRα in this signaling pathway. Overexpression of Myc-RXRα/Δ80 (representing tRXRα) but not Myc-RXRα in MCF-7 cells significantly enhanced both the basal and TNFα-stimulated NF-κB transcriptional activity and IκBα degradation (Supplementary Fig. S5A and S5B). TRAF, the IKK upstream transducer of the TNFα/NFκB signal pathway, is an essential component in this pathway (25). Ectopic expression of TRAF2 slightly activated NF-κB transcription and IκBα degradation, which was dramatically enhanced by transfection of Myc-RXRα/Δ80 but not Myc-RXRα in a dose-dependent manner (Fig. 5A and B). Thus, tRXRα, but not RXRα, is a positive regulator of TNFα activation of NF-κB. When TRAF2 expression was suppressed by siRNA-mediated knockdown, the effect of Z-12 on inhibiting TNFα-induced IκBa degradation was dramatically reduced, indicating the role of TRAF2 in the activity of Z compounds (Fig. 5C). We then determined the possibility that tRXRα interacted with TRAF2. When Flag-TRAF2 was cotransfected with either Myc-RXRα/Δ80 or Myc-RXRα, immunoprecipitation of Flag-TRAF2 resulted in coprecipitation of Myc-RXRα/Δ80 but not Myc-RXRα, which was strongly enhanced by TNFα (Fig. 5D). Similarly, when the complex was immunoprecipitated with anti-Myc antibody, Flag-TRAF2 was detected in the immunoprecipitated complex when coexpressed with Myc-RXRα/Δ80 but not Myc-RXRα (Supplementary Fig. S5C). We also found that TNFα could induce the interaction of Myc-RXRα/Δ80 with endogenous TRAF2 (Supplementary Fig. S5D). The direct interaction of RXRα/Δ80 and TRAF2 was revealed by our GST pulldown assay (Supplementary Fig. S5E). Further analysis indicated that the N-terminal region of TRAF2 was responsible for binding to RXRα/Δ80 (Supplementary Fig. S5F). When the effect of Z-10 and Z-12 was analyzed, we found that treatment of cells with either of the compounds strongly inhibited TNFα-induced interaction of tRXRα with TRAF2 (Fig. 5E). The activation of TRAF2 often leads to the ubiquitination of RIP1, which is required for the activation of NF-κB by TNFα (26, 27). We found that RXRα/Δ80 could strongly promote TNFα-induced RIP1 ubiquitination, which was largely blocked by Z-10 (Fig. 5F). Together, tRXRα may contribute to the activation of the NF-κB pathway by TNFα through binding to TRAF2 and promoting RIP1 ubiquitination, and Z compounds may suppress tRXRα-dependent activation of the NF-κB signaling pathway through their inhibition of tRXRα/TRAF2 complex formation.

We hypothesized that suppression of tRXRα activation of the NF-κB survival signaling by Z compounds may provoke the apoptotic potential of TNFα. Indeed, when cells were treated with TNFα together with either Z-10 or Z-12, a significant apoptosis of cells, indicated by PARP cleavage and nuclear fragmentation, was observed, which did not happen when cells were treated with Z-10, Z-12, or TNFα alone (Fig. 6A and Supplementary Fig. S6A). TNFα is known to induce caspase-8–dependent apoptosis (18), which was also revealed by our experiment (Fig. 6A). Treatment of cells with the general caspase inhibitor Z-VAD-FMK completely blocked the PARP cleavage induced by the combination of TNFα and Z compounds (Supplementary Fig. S6B). Thus, Z compounds are able to activate the death effect of TNFα. Consistently, TNFα combination with Z-10 exhibited much stronger inhibitory effects on MCF-7 cell colony formation than either of them alone (Fig. 6B).

In contrast to the effect of Z compounds, 9-cis-RA, CD3254 and UVI3003 showed little effect on inducing PARP cleavage in the presence of TNFα (Supplementary Fig. S6C). However, they were able to suppress the apoptotic effect of the combination of TNFα and Z compound (Fig. 6C), suggesting the involvement of RXRα in the apoptotic induction of the combination. This was first confirmed by RXRα siRNA experiments, showing that transfection of RXRα siRNA prevented PARP cleavage induced by the combination of TNFα and Z compounds (Supplementary Fig. S6D). In addition, the expression levels of RXRα and tRXRα in SW480, H460, and MCF-7 cells were directly correlated to the apoptotic status of the cells treated by TNFα/Z-10 combination (Fig. 6D). For comparison, Z-10-1 and Z-12-1 did not show any synergistic proapoptotic effects when used together with TNFα (Supplementary Fig. S6E). To further distinguish the role of RXRα and tRXRα, we stably transfected RXRα and RXRα/Δ80 in MCF-7 cells and examined the apoptosis of the resulting stable cell lines treated by the TNFα/Z compound combination. Compared with the parental MCF-7 cells, RXRα/Δ80 but not RXRα stable cell line showed increased apoptosis when treated with either TNFα/Z-10 or TNFα/Z-12 combination (Fig. 6E and Supplementary Fig. S6F). Similarly, we observed RXRα/Δ80-, but not RXRα-dependent nuclear fragments induced by the TNFα/Z-10 combination (Supplementary Fig. S6G). Together, these data demonstrate that the combination of TNFα and Z compounds induces a tRXRα-dependent apoptosis of MCF-7 cancer cells.

When Flag-TRAF2 was overexpressed in MCF-7 cells, the synergistic effects of TNFα and Z-10 on inducing PARP cleavage were largely blocked (Fig. 6F). In contrast, overexpression of dominant-negative TRAF2 (dnTRAF2) stimulated TNFα-induced PARP cleavage, which was not enhanced by Z-10 (Fig. 6G). These data suggested that the inhibition of TNFα/NFκB signaling by Z compounds led to the activation of the TNFα apoptotic pathway, providing an explanation for the synergistic proapoptotic effects of the combination. We further evaluated whether Z compounds alone or combination with TNFα inhibited MCF-7 cell growth in vivo. After subcutaneous injection of MCF-7 cells, we treated mice with vehicle, Z-12, TNFα, or TNFα/Z-12 by irrigation of Z compounds and intratumoral injection of TNFα. In MCF-7 bearing mice, TNFα/Z-12 combination induced significant suppression of tumor growth when compared with vehicle, Z-12, and TNFα (Fig. 6H). The pronounced delay in tumor growth, especially by the treatment of TNFα/Z-12 combination, was translated into tumor weight values, which was obtained at the end of the treatment (Fig. 6I). Tumor growth inhibition and extensive apoptosis of cancer cells were also observed when mice were treated with Z-10 or Z-12 alone (Fig. 6H and I and Supplementary Fig. S6H–S6J), likely due to the self-produced TNFα of tumor tissues.

By using a variety of in vitro and in vivo approaches, we showed that several nitrostyrene derivatives could bind to RXRα through a new mechanism (Fig. 2 and 3 and Supplementary Figs. S2 and S3). It has been reported that synthetic nitro compounds T0070907 and GW9662 are PPARγ antagonists (28). Endogenous nitro-fatty acids, including nitrated linoleic acid (LA-NO₂) and nitrated oleic acid (OA-NO₂), are also identified as robust PPARγ agonists, which induce PPARγ-dependent effects, including macrophage CD-36 expression, adipocyte differentiation, and glucose uptake (29, 30). Several other nuclear receptors, such as ER and HNF4α, were also found to bind to nitro compounds (31, 32). However, to our knowledge, Z compounds are the first identified nitro ligands for RXRα. Because the variety of the nitrated lipid acids and other nitrated small molecules exist in the body (33), it is much likely there are endogenous nitro ligands of RXRα, which is worthy of further investigation.

It has been reported that a replacement of nitro group with carboxyl group abolishes the effects of nitrostyrene derivatives on proapoptotic induction (34), antiplatelet activity, and inhibiting NLRP3 inflammasome activation (35, 36). Here, we also demonstrated that nitro group was essential for Z compounds to bind to RXRα (Fig. 2A and Supplementary Fig. S2), stimulate RXRα transactivation (Fig. 2D and E), regulate RXRα oligomerization (Fig. 3B and Supplementary Fig. S3A), inhibit TNFα activation of NF-κB (Supplementary Fig. S4F), and induce cancer cell apoptosis (Supplementary Fig. S6E). Therefore, nitro group is critical for the nitrostyrene derivatives to exert their wide biologic activities, which is partly due to the requirement of nitro group for nitrostyrene derivatives binding to RXRα (Fig. 2 and Supplementary Fig. S2).

Crystallographic analysis indicates that amino acid residues responsible for interacting with LA-NO₂ and rosiglitazone are different, leading to distinct conformational changes of PPARγ-LBD when complexed with LA-NO₂ and rosiglitazone (37). Similarly, Z compounds induced distinct conformations of RXRα homodimer (Fig. 3B and C and Supplementary Fig. S3A and S3B). Our mutagenesis study indicated that Cys432, which is located at the corner of the L shape LBP (13, 38), was essential for Z compounds to activate and bind to RXRα (Fig. 3E and Supplementary Fig. S3C and S3D). The distinct conformations of RXRα induced by Z compounds may rely on their unique interaction with Cys432, which was in part supported by the self-activation of RXRα mutants C432Y and C432W mimic the binding of Z compounds (Fig. 3E–H). The importance of Cys432 for specific ligand binding and activity has been illustrated by an early report. Tributyltin, an organotin compound, is able to induce an active conformation of RXRα-LBD, primarily due to the covalent bond formed between tin atom and Cys432 (39).

Crystallographic analysis demonstrates that hydrogen bonds formed between nitro group and residues Arg288 or Glu343 stabilize the complex of LA-NO₂ and PPARγ (37), while spectrometric analysis indicates that OA-NO₂ covalent binds PPARγ by Michael addition of Cys285 with nitroalkene (40). Our SPR results showed the quick dissociation of Z compounds from RXRα-LBD in the dissociation phase (Fig. 2B), implying no strong covalent binding between Z compounds and RXRα. However, we could not exclude the possibility that Z compounds bind to RXRα through a weak covalent binding via the Michael addition of Cys432 with nitroalkene, which might be microenvironment-sensitive (in RXRα LBP) and reversible referring to the nitroalkylation reactions (41).

tRXRα is able to bind to p85α to enhance TNFα-stimulated Akt activation and cancer growth (20). In this study, we showed that tRXRα could also enhance TNFα activation of the NF-κB signaling probably through its interaction with TRAF2 and induction of RIP1 ubiquitination (Fig. 5 and Supplementary Fig. S5). Interestingly, TNFα promoted the interaction of TRAF2 with tRXRα but not RXRα (Fig. 5D and F and Supplementary Fig. S5C and D), suggesting a possible mechanism by which abnormal activation of the TNFα/NFκB pathway plays a role in mediating the tumor-promoting effect of tRXRα that is specifically produced in cancer cells (15, 20). Such a tRXRα-mediated activation of the NF-κB pathway may provide a new direction for targeting tRXRα through inhibiting its interactions with TRAF2. Indeed, our data showed that Z compounds strongly inhibited TNFα-induced interaction of TRAF2 with tRXRα and RIP1 ubiquitination (Fig. 5E and F), which was associated with their inhibition of the TNFα/NFκB signaling pathway in a tRXRα-dependent manner (Fig. 4D and E and Supplementary Fig. S4D). Unlike sulindac and its derivatives (20), Z compounds did not inhibit tRXRα interaction with p85α (Supplementary Fig. S4A), reflecting the different tRXRα conformations induced by nitrostyrene derivatives and sulindac derivatives.

Suppression of the TNFα/NFκB signaling pathway may convert TNFα from a tumor promoter to a tumor suppressor (42, 43). Indeed, the potent effects of Z compounds on inhibiting the TNFα/NFκB survival pathway resulted in a synergistic effect of Z compounds and TNFα on inducing tumor cell apoptosis (Fig. 6 and Supplementary Fig. S6), likely due to the activation of the TNFα-mediated pathway of apoptosis (Fig. 6A and Supplementary Fig. S6B). The synergistic antitumor effect of the combination was tRXRα dependent (Fig. 6C–E and Supplementary Fig. S6D), consistent with the facts that Z compounds bound to RXRα and inhibited TNFα activation of NF-κB in a tRXRα-dependent manner. Thus, our results define a class of compounds that could convert TNFα signaling from survival to death in cancer cells by targeting the tRXRα-mediated TNFα/NFκB signaling pathway.

Generally, the molecular weight of RXRα ligands ranges from 300 to 500 Da determined by the effective binding and spatial size of RXRα-LBP (2, 11). The sizes of Z compounds (Z-10, 199 Da; Z-12, 249 Da) are relatively small and it was imagined that Z compounds only partially occupied RXRα-LBP, which might also explain that the binding affinity of Z compounds fall into the μmol/L but not nmol/L range. However, the μmol/L working concentrations of Z compounds for binding to RXRα, inducing RXRα conformational changes, inhibiting TNFα activation of NF-κB, and promoting cancer cell apoptosis were in the same order of magnitude, reflecting the relevance of their binding to RXRα and their tRXRα-dependent physiologic functions. The small size of Z compounds also makes it possible to optimize them. One of the optimizing approaches is to increase their molecular sizes to enhance the van der Waal interactions between Z compounds and the LBP of RXRα, and the other one is to introduce a carboxyl group in appropriate positions of Z compounds to interact with Arg316, referring to the molecular basis of the interaction between LA-NO₂ and PPARγ-LBD (37).

Taken together, our results identify the first nitro ligands of RXRα with unique RXRα-binding mode and tRXRα-dependent abilities of anti-NF-κB activation and proapoptosis of cancer cells. Our results also reveal a new mechanism by which tRXRα promotes tumor growth, providing a new strategy for inhibiting TNFα activation of the NF-κB pathway by targeting tRXRα.

No potential conflicts of interest were disclosed.

Conception and design: H. Zhou, Z. Sun, M. Huang, X. Cao, X.-K. Zhang

Development of methodology: H. Zhou, Z. Sun, M. Huang, L. Chen, F. Huang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Zeng, Z. Sun, M. Huang, W. Zhang, J. Liu, L. Chen, F. Chen, Y. Zhou, J. Lin, L. Xu, Z. Zhuang, S. Guo, G. Alitongbieke, G. Xie, B. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Zhou, Z. Sun, M. Huang, L. Chen, F. Chen, X.-K. Zhang

Writing, review, and/or revision of the manuscript: H. Zhou, Y. Su, X.-K. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Zhou, Z. Sun, J. Lin, Y. Xu

Study supervision: H. Zhou, X.-K. Zhang

This work was supported by the grants from the National Nature Science Fund of China (NSFC-31471318, NSFC-31271453, NSFC-91129302 and NSFC-81301888), the Fundamental Research Funds for the Central Universities (2013121038), Xiamen Science and Technology Project (3502Z20123015), the NIH (CA140980 and GM089927), the U.S. Army Medical Research and Material Command (W81XWH-11-1-0677), and the California Breast Cancer Research Program (20IB-0138).

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.