Targeting Truncated Retinoid X Receptor-α by CF31 Induces TNF-α–Dependent Apoptosis

Retinoid X receptor (RXR)-α, a unique member of the nuclear receptor superfamily, regulates diverse biologic processes, including growth, differentiation, apoptosis, and immune response, and ligands for RXR-α show promise as therapeutic agents for many diseases, such as cancer (1–4). Alterations in the expression and function of RXR-α are implicated in the development of a number of diseases and cancer. Targeted disruption of RXR-α gene leads to prostatic preneoplastic lesions (5) and skin abnormalities (6). Dysregulation of RXR-α function by phosphorylation is associated with the development of human liver cancer (7) and colon cancer (8). Levels of RXR-α protein are often reduced in cancer cells and tumor tissues (9–14), suggesting that diminished RXR-α expression is associated with the development of certain malignancies. Numerous studies have shown that proteolytic cleavage of RXR-α protein is primarily responsible for its diminished levels in tumor cells and represents an important mechanism controlling the function and activities of RXR-α (13–18). Recent studies have shown that RXR-α binding to PML/RAR-α is essential for the development of acute promeylocytic leukemia (19, 20), further showing the oncogenic potential of this protein when it acts inappropriately.

Although significant progress has been made, how RXR-α regulates cancer cell growth and how its ligands suppress tumorigenesis is still poorly understood. Like other nuclear receptors, RXR-α interacts with DNA to regulate the transcription of target genes in a ligand-dependent manner (1–4). Aside from its role in DNA binding and transactivation, accumulating evidence indicates that RXR-α also has extranuclear nongenomic actions (15, 21–24). RXR-α resides in the cytoplasm at certain stages during development (25, 26). It migrates from the nucleus to the cytoplasm in response to differentiation (23), apoptosis (21), and inflammation (22, 24). Cytoplasmic localization of RXR-α facilitates nuclear export of its heterodimerization partner Nur77, leading to Bcl-2 conversion and apoptosis (21, 27, 28).

We showed recently that proteolytic cleavage of RXR-α results in production of truncated RXR-α protein, tRXR-α, which acquires new function that is different from the full-length RXR-α (29). Unlike the full-length RXR-α that resides in the nucleus, tRXR-α is cytoplasmic and interacts with the p85α subunit of phosphoinositide 3-kinase (PI3K) in a TNFα–dependent manner, leading to activation of the PI3K/AKT pathway, a major survival pathway important for uncontrolled growth of tumor and as drug resistance (29). Activation of the PI3K/AKT pathway by tRXR-α contributes significantly to anchorage-independent growth of cancer cells in vitro and tumor growth in animals (29), offering an opportunity to suppress cancer cell growth by targeting the tRXR-α–mediated PI3K/AKT pathway.

Epidemiologic studies have shown that dietary phytochemicals provide beneficial effects for cancer prevention, and among them, xanthones are of great interest as cancer chemopreventive agents because of their potent antioxidative and anticancer activity (30–33). We previously purified a series of xanthones from a medicinal plant, Cratoxylum formosum ssp. pruniflorum, which belongs to the Clusiaceae family and is widely distributed in several Southeast Asian countries (34, 35). Here, we report our identification of one of the xanthones, CF31, which suppresses tRXR-α–dependent AKT activation through its unique RXR-α binding. CF31, when combined with TNF-α inhibits TNF-α–induced tRXR-α–p85α interaction and AKT activation. Moreover, the CF31 and TNF-α combination results in synergistic induction of TNF-α–dependent caspase-8 activation and apoptosis in cancer cells. Thus, CF31 represents a new modulator of the tRXR-α survival pathway and a potent converter of TNF-α signaling from survival to death.

Compounds CF31, also called cochinchinone B or 1,3,6,7-tetrahydroxy-2-(3-methyl-2-butenyl)-5-(3,7-dimethyl-2,6-octadienyl)xanthone (36), CF13 (1,3,5-trihydroxy-4-geranylxanthone), and CF26 (pruniflorone Q) were isolated from the stems of Cratoxylum formosum ssp. pruniflorum as described (35).

The pGAL4-RXR-α-LBD plasmid was obtained by inserting RXR-α ligand-binding domain (LBD) sequence (amino acids 198–462) in-frame with the Gal4 DBD coding sequence in the pBIND vector (Promega).

Antibodies for phospho-AKT (Ser473, D9E) and cleaved caspase-8 (p43/p41; Cell Signaling); Bax (6A7; Sigma-Aldrich); p85α (Millipore); AKT1 (C-20), actin, c-Myc (9E10), RXR-α (ΔN197), RXR-α (D20), and PARP (H-250; Santa Cruz Biotechnology) were used. Caspase-3 and -8 activity assay kits were from Biovision.

GFP-RXR-α/Δ80 was stably transfected into SW480, HepG2, andMCF-7 cells to obtain SW480/RXR-α/Δ80, HepG2/RXR-α/Δ80, and MCF-7/RXR-α/Δ80, respectively.

RXR-α LBD was incubated with [³H]-9-cis-RA in the presence or absence of unlabeled 9-cis-retinoic acid (9-cis-RA) or CF31. Bound [³H]-9-cis-RA in RXR-α. LBD protein was determined in a scintillation counter (29).

CV-1 green monkey kidney cells were grown in Dulbecco's Modified Eagle's Medium supplemented with 10% FBS. For chloramphenicol acetyltransferase (CAT) reporter assays, cells were seeded at 5 × 10⁴ cells per well in 24-well plates and transfected with 50 ng TREpal-tk-CAT reporter plasmid (37), 20 ng β-galactosidase expression vector (pCH 110; Amersham), 20 ng RXR-α expression vectors using Lipofectamine 2000 (Invitrogen). Cells were then treated with CF31 at different dose for 20 hours. CAT activity was normalized with β-galactosidase activity. For luciferase reporter assay, 24 hours after transfected with pGL5 luciferase reporter vector (40 ng/well) and pGAL4-RXR-α-LBD expression vector (40 ng/well), cells were incubated with varied concentrations of compounds for 12 hours. Luciferase activities were measured using the Dual-Luciferase Assay System Kit (Promega).

AutoDock version 4.0 was used for the docking screening. The Lamarckian genetic algorithm was selected for ligand conformational searching. Docking parameters were as follows: grid box of 40 Å × 40 Å × 40 Å, xyz-coordinates of gridcenter: 49.192 63.991 −7.523, population size of 150, random starting position and conformation, translation step ranges of 2 Å, rotation step ranges of 50°, elitism of 1, mutation rate of 0.02, cross-over rate of 0.8, local search rate of 0.06, and 2.5 million energy evaluations. Molecular dynamics simulations were adopted to sample possible configurations of each RXR-α LBD complex using the AMBER (version 7.0) program (38). To set up each molecular dynamics simulation, the electrostatic potentials of the ligand was computed by using the Gaussian 98 package at the HF/6-31G* level. Atom-centered partial charges were derived by using the Restrained Electrostatic potential method implemented in the AMBER package.

HepG2 cells mounted on glass slides were permeabilized with PBS containing 0.1% Triton X-100 and 0.1 mol/L glycine for 15 minutes, and blocked with 1% bovine serum albumin in PBS for 30 minutes at room temperature. Cells were then incubated with primary antibody [anti-Bax (6A7; 1:500)] at 37°C for 1 hour and detected by anti-rabbit immunoglobulin G (IgG) conjugated with Cy3 (1:400) at room temperature for 30 minutes. Cells were costained with 4′6′-diamidino-2-phenylindole (DAPI) to visualize nuclei. The images were taken using LSM-510 confocal laser scanning microscope system (Carl Zeiss).

Equal amounts of the lysates were electrophoresed on 8% SDS-PAGE gel and transferred onto polyvinylidene difluoride membranes. The membranes were blocked with 5% nonfat milk in Tris-Buffered Saline and Tween 20 [50 mmol/L Tris–HCl (pH 7.4), 150 mmol/L NaCl, and 0.1% Tween 20] for 1 hour, incubated with various primary antibodies for 2 hours and detected with either anti-rabbit (1:5,000) or anti-mouse (1:5,000) secondary antibodies for 1 hour, all of which were undertaken under room temperature. The final immunoreactive products were detected by using enhanced chemiluminescence (ECL) system.

Caspase activities were determined by a colorimetric assay based on the ability of caspase-3 and -8 to change acetyl–Asp–Glu–Val–Asp p-nitroanilide (Ac-DEVD-pNA) and acetyl–Ile–Glu–Thr–Asp p-nitroanilide Ac-IETD-pNA) into a yellow formazan product [p-nitroaniline (pNA)], respectively. After treatment with CF31 and/or TNF-α, lysates were prepared and assayed according to the manufacturer's protocol (Biovision Research Products).

RXR-α siRNA siGENOME SMARTpool (M-003443-02) and siRNA Nonspecific Control IX (D-001206-09-05) were from DHARMACON, and they were transfected into cells by RNAiMAX reagent (Invitrogen).

Cells extracts were cleared by incubation with the Protein A/G plus Agarose beads and then incubated with appropriate antibody and 30 μL of Protein A/G plus Agarose beads overnight at 4°C. Immunoreactive products were detected by chemiluminescence with an ECL system (Amersham).

HeLa cells were seeded in 6-well plate (500 cells/well) for 5 days, treated with CF31 in 0.5% serum medium for 3 days, and fixed with 4% paraformaldehyde. Colonies were stained with 0.1% crystal violet.

Data were analyzed using an analysis of variance or Student test and were presented as the mean ± SEM of triplicate.

To identify new agents for regulating the tRXRα survival pathway, we screened a natural product library prepared from Chinese Herbal Medicine for compounds capable of regulating RXR-α transcriptional function. For this purpose, transient transfection assays using a CAT reporter containing TREpal capable of binding to RXR-α homodimer (37) were used. Compounds were evaluated for their effect on TREpal activity in the absence or presence of 9-cis-RA, the natural RXR ligand known to induce the formation of RXR-α homodimer and its activation of TREpal (37). Among compounds that were identified to modulate RXR-α activity, CF31 (Fig. 1A), which was prepared from Cratoxylum formosum ssp. Pruniflorum (35), potently inhibited 9-cis-RA–induced TREpal-reporter activity in a dose-dependent manner (Fig. 1B). CF31 at 10 μmol/L showed a similar inhibitory effect when compared with BI1003 (1 μmol/L), a known RXR-α antagonist (39). To further study the inhibitory effect of CF31, we cloned the LBD of RXR-α as a Gal4 fusion, and used the resulting Gal4-RXR-α/LBD chimera and Gal4 reporter system to evaluate the inhibitory effect of CF31. Gal4-RXR-α/LBD strongly activated the Gal4 reporter in the presence of 9-cis-RA, which was inhibited by 1 μmol/L UVI3003, another RXR-α antagonist (40; Fig. 1C). Similar to UVI3003, treatment of cells with CF31 resulted in inhibition of 9-cis-RA–induced reporter activity in a CF31 dose-dependent manner. Because UVI3003 was derived from a potent and selective RXR-α agonist CD3254 (40), the effect of CF31 on CD3254-induced RXR-α activity was also evaluated. CD3254 strongly induced Gal4-RXR-α/LBD activity, which was comparable with that induced by 9-cis-RA. Interestingly, CD3254-induced reporter activity was similarly inhibited by UVI3003 and CF31 (Fig. 1C). We next determined whether CF31 could bind to RXR-α. Thus, a ligand competition assay with [³H]9-cis-RA binding to RXR-α LBD was used. While unlabeled 9-cis-RA was able to efficiently displace [³H]9-cis-RA from binding to RXR-α LBD with an IC₅₀ of 7.6 nmol/L, CF31 displaced [³H]9-cis-RA with an IC₅₀ of 9.6 μmol/L (Fig. 1D). Together, our data show that CF31 acts as a RXR-α antagonist by binding to the RXR-α LBD.

Molecular conformation analyses showed there was a large degree of structural overlapping between CF31 and LG100754 (Fig. 2A), an antagonist of RAR-α/RXR-α heterodimer (41). This suggested that CF31 might bind to the antagonist form of the RXR-α LBD. However, CF31, unlike many natural and synthetic RXR-α ligands, lacks a carboxylate moiety known to form salt bridges with Arg316 in the L-shaped RXRα ligand-binding pocket (LBP; refs. 41–43). To understand how CF31 binds RXR-α, a docking study using the antagonist structure of the RXR-α LBD in complex with LG100754 (41) was conducted. CF31 docked well with a good score to the RXRα antagonist conformation. Superposing the docked configuration of CF31 and the crystal structure of LG100754 showed that their scaffold orientations were highly conserved (Fig. 2A). An interesting difference, however, is that while the carboxylate group of LG100754 made a strong salt bridge with Arg316 in helix H5, CF31 could not form such a bridge (Fig. 2B).

To study the requirement of Arg316 in CF31 binding, Arg316 was replaced with Glu, and the resulting mutant RXR-α/R316E was evaluated for its activation by 9-cis-RA and CD3254. As shown in Fig. 2C, mutation of Arg316 completely abolished the effect of 9-cis-RA on activating the Gal4 reporter. In contrast, CD3254 could still activate the RXR-α/R316E mutant. Thus, while Arg316 is capable of establishing an ionic interaction with the carboxylate group of 9-cis-RA (43), it fails to form salt bridges with CD3254 (40). Consistently, our computational analysis showed that 9-cis-RA formed 2 shorter hydrogen bonds (2.098 Å and 2.423 Å) with Arg316 than those formed with CD3254 (2.901 Å and 2.553 Å; Fig. 2D). Therefore, Arg316 plays a more important role in the binding of 9-cis-RA than that of CD3254. We then studied whether CF31 could inhibit CD3254-induced RXRα/R316E activity. When RXR-α/R316E was cotransfected with the Gal4 reporter, CD3254-induced reporter activity was potently inhibited by CF31 in a dose-dependent manner (Fig. 2C), showing that Arg316 was not required for CF31 binding.

To identify which amino acid residues in the RXR-α LBP were key to CF31 binding, we used the Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PB/SA) method (44) to evaluate binding free energy of CF31 in complex with the RXRα LBP. For comparison, 9-cis-RA was also studied. Our examination of the interaction spectrum of CF31 with amino acid residues in the RXR-α LBP with that of 9-cis-RA revealed that Arg316 was indeed mainly responsible for the binding of 9-cis-RA but not CF31 (Fig. 2E; Supplementary Table S1). However, CF31 made considerable contacts with several amino acid residues in the RXR-α LBP, including Val265 and Ile268 in H3, Ile310 and Phe313 in H5, V342, I345, F346 in H7, and Leu436 in H11 (Fig. 2E). Thus, the nonpolar van der Waals interactions of CF31 with the hydrophobic residues in the RXRα LBP may serve to stabilize the binding of CF31, revealing its distinct mode of RXRα binding.

In an attempt to gain more understanding of the antagonist effect of CF31, we evaluated 2 additional xanthones isolated from Cratoxylum formosum ssp. pruniflorum, CF13 and CF26 (Fig. 3A), for their antagonist effect (Fig. 3B). While CF13 had no inhibitory effect of 9-cis-RA–induced RXR-α transactivation, the inhibitory effect of CF26 was even stronger than that of CF31 (Fig. 3B). Modeling studies suggested that a potential steric hindrance between side chain of Leu451 in H12 and substituent at position 3 of ring A of these compounds might mediate their antagonist effect (Fig. 3C). The size of the substituent correlated well with their antagonist effect, with 3-methyl-2-butenyl in CF26 being the most effective.

In the first step to study the effect of CF31 on tRXRα–dependent AKT survival pathway, we determined whether CF31 could regulate AKT activation. Figure 4A shows that CF31 dose-dependently inhibited the activation of AKT in HeLa cells, revealed by its inhibition of the expression of phospho-AKT but not total AKT using Western blotting. Time-course analysis showed that the inhibition of AKT activation by CF31 was observed when cells were treated by 10 μmol/L CF31 for as short as 8 hours (Fig. 4B). The inhibition of AKT activation by CF31 also closely correlated with apoptosis induction as shown by enhanced cleavage of PARP protein (Figs. 4A and B).

The ability of CF31 to inhibit AKT activation prompted us to examine the effect of CF31 on the survival of HeLa cells by colonogenic survival assays. Treatment of cells with CF31 for 3 days completely inhibited colony formation of HeLa cells, which was as effective as K-80003, a sulindac analog that inhibits AKT activation through its binding to tRXR-α (29). In contrast, sulindac showed little effect under the same conditions (Fig. 4C). We also determined whether CF31 inhibition of AKT activation could enhance the apoptotic response of cancer cells to chemotherapeutics. To this end, we examined the death effect of the topoisomerase-I inhibitor camptothecine (CPT), which has shown significant antitumor activity across a broad spectrum of human tumors (45) in the presence or absence of CF31 in HepG2 liver cancer cells. Immunoblotting showed that PARP cleavage induced by camptothecine was significantly enhanced when cells were cotreated with CF31, whereas CF31 alone had undetectable effect on PARP cleavage under the conditions used (Fig. 4D). When cells were transfected with RXR-α siRNA, the enhancing effect of CF31 was impaired. Thus, CF31 may sensitize cancer cells to the apoptotic effect of certain anticancer drugs through its inhibition of AKT activation.

TNF-α, a multifunctional cytokine, can activate AKT in a tRXR-α–dependent manner in certain cancer cells (29). In HeLa cells, TNF-α treatment enhanced AKT activation, whereas CF31 and K-80003 showed comparable effect on inhibiting AKT activation. Both compounds were much more effective than sulindac when used at 10 μmol/L (Fig. 5A). We then examined whether CF31 could inhibit TNF-α activation of AKT. Treatment of HeLa cells with TNF-α for 30 minutes strongly activated AKT, which was inhibited by either CF31 or K-80003 in a dose-dependent manner (Fig. 5B). Both compounds at 20 μmol/L completely inhibited the effect of TNF-α on activating AKT. CF31 could also inhibit basal and TNF-α–induced AKT activation in other cancer cell lines, including A549 lung cancer and HepG2 liver cancer cells (Fig. 5C). All these cancer cell lines expressed high levels of tRXR-α, as revealed by immunoblotting using ΔN197 anti-RXR-α antibody recognizing the LBD of RXR-α (29; Fig. 5C).

To study the role of tRXR-α in CF31 action, we first assessed the effect of RXR-α siRNA transfection. Transfection of RXR-α siRNAs in A549 cells, which reduced levels of both the full-length RXR-α and tRXR-α (29), abolished the inhibitory effect of CF31 on AKT activation either in the absence or presence of TNF-α (Fig. 5D). To complement this study, we determined whether overexpression of tRXR-α could enhance the inhibitory effect of CF31. For this purpose, we stably expressed GFP-RXR-α/Δ80 (29), a GFP-tagged RXR-α mutant lacking 80 N-terminal amino acid residues, in SW480 colon cancer cells that expressed little tRXR-α protein (Fig. 5E), SW480 cells exhibited low basal AKT activation, and their AKT activation was insensitive to regulation by TNFα and (Fig. 5E). However, stable expression of GFP-RXRα/Δ80 enhanced not only their basal AKT activation but also their response to TNF-α induction of AKT activation (Fig. 5E). Moreover, the sensitivity of SW480 cells to CF31 was restored, as it could effectively inhibit TNF-α–induced AKT activation in SW480/RXR-α/Δ80 cells. Thus, tRXR-α expression is involved in AKT activation by TNF-α and is essential for its inhibition by CF31.

Previously, we reported that tRXR-α binding to p85α was essential for the activation of the PI3K/AKT pathway (29). Thus, we determined whether CF31 inhibition of tRXR-α–dependent AKT activation could be attributed to its interference with tRXR-α interaction with p85α. Consistent with our previous observation, TNF-α induced interaction of tRXR-α with p85α, as revealed by coimmunoprecipitation using the ΔN197 anti-RXR-α antibody (Fig. 5F). However, when cells were cotreated with CF31, TNF-α–induced tRXR-α–p85α interaction was largely abolished. Thus, CF31 may modulate AKT activation through its ability to inhibit the interaction of tRXR-α with p85α.

To study whether CF31 inhibition of AKT activation could result in apoptosis, we first examined the effect of CF31 on activation of proapoptotic Bcl-2 family member Bax in HepG2 cells in the presence of TNF-α by immunostaining using a Bax conformation-sensitive antibody Bax/6A7 that recognizes active Bax protein (46). As shown in Fig. 6A, immunostaining with Bax/6A7 antibody showed that activated Bax protein was undetectable in control cells. However, when cells were treated with CF31, a strong immunostaining was observed, suggesting the activation of Bax by CF31. The apoptotic effect of CF31 in the presence of TNF-α was also illustrated by its induction of PARP cleavage in HepG2 and HeLa cells (Fig. 6B). Induction of PARP cleavage by TNF-α/CF31 combination treatment was associated with reduction of AKT activation, suggesting that the inhibition of AKT activity plays a role in their induction of apoptosis. In support of the role of AKT inhibition by the combination treatment, transfection of the constitutive-active AKT (CA-AKT) expression vector prevented its induction of PARP cleavage (Fig. 6C).

We also determined the role of RXR-α in CF31-induced apoptosis by transfecting control and RXR-α siRNA in HeLa cells. Comparison of the effect of CF31 in cells transfected with control siRNA or RXR-α siRNA showed that transfection of RXR-α siRNA impaired the ability of CF31 to induce PARP cleavage and to inhibit AKT activation (Fig. 6D). To address the role of tRXR-α, we examined the effect of CF31 on PARP cleavage in cells stably expressing RXR-α/Δ80. In A549 cells, CF31 showed little effect on inducing PARP cleavage (Fig. 6E). However, it significantly induced PARP cleavage in cells stably expressing RXR-α/Δ80 tagged with Myc epitope, Myc-RXR-α/Δ80. Similar results were obtained in MCF-7 breast cancer cells stably expressing GFP-RXRα/Δ80 (Fig. 6F). Thus, the expression of tRXR-α plays a role in the apoptosis induction by CF31.

TNF-α is capable of inducing opposing biologic activities, such as cell survival and death, and its killing effect in cancer cells are often antagonized by its survival function that is mainly mediated by activation of the NF-κB and PI3K/AKT pathways (47, 48). Our observation that both inhibition of AKT activation and induction of apoptosis by CF31 occurred in the presence of TNF-α prompted us to study whether CF31 could convert TNF-α from a survival to a killing molecule. Thus, caspase-3 activation was examined in HeLa cells treated with CF31 or TNF-α alone or their combination. Treatment of cells with either CF31 or TNF-α alone had little effect on caspase-3 activation, whereas the combination treatment resulted in strong caspase-3 activation (Fig. 7A). The combination treatment also strongly induced the activation of caspase-8, the downstream mediator of the TNF-α–dependent apoptotic pathway (47, 48). Caspase-8 activation by CF31/TNF-α combination was also revealed by immunoblotting showing strong induction of cleaved caspase-8 products, p43/p41, an indication of caspase-8 activation (Fig. 7B). The induction of caspase-8 activation and PARP cleavage by the CF31/TNF-α combination was partially impaired by transfection of RXR-α siRNA (Fig. 7B), showing a role of RXR-α.

Induction of PARP cleavage by CF31/TNF-α combination treatment was closely correlated with caspase-8 activation (Fig. 7B), suggesting an activation of TNF-α–dependent apoptotic pathway by CF31. To further address the role of caspase-8 activation in apoptosis induction by CF31, we examined the effect of caspase-8 siRNA transfection. Transfection of HeLa cells with caspase-8 siRNA reduced the level of p43/p41 and PARP cleavage (Fig. 7C). In addition, PARP cleavage induced by the CF31/TNF-α combination treatment was largely inhibited in cells treated with the caspase-8 inhibitor Z-IETD-fm (Fig. 7D). These results showed that apoptosis induction by the CF31/TNF-α combination treatment was largely mediated by their activation of TNF-α–mediated extrinsic apoptotic pathway, suggesting the role of CF31 in shifting the TNFα signaling from the survival to death.

Traditional medicinal plants and dietary factors provide a fertile ground for modern drug development, with some compounds, such as paclitaxel, etoposide, camptothecin, and vincristine, being successfully used as anticancer drugs (49). Xanthones, a class of 3-membered heterocyclic ring compounds mainly found as secondary metabolites in higher plants and microorganisms, exert very diverse biologic profiles, including antihypertensive, antioxidative, antithrombotic, and anticancer activity, depending on their diverse structures (30–33). A large number of naturally occurring and synthetic xanthones such as psorospermin (50), dimethylxanthesone-4-acetic acid (DMXAA; ref. 51), and α-mangostin (52) have shown potent anticancer activities. However, the biologic targets of xanthone compounds remain elusive (30–33). Cratoxylum is a small genus distributed in Southeast Asia with some of its species used medicinally, and xanthones are the most characteristic biologically active components of this genus (31, 35, 52, 53). We have previously identified several bioactive xanthones from the stems of Cratoxylum formosum ssp. pruniflorum (34, 35). We report here that CF31, one of the xanthones that we isolated, acts as a potent negative regulator of the tRXR-α–mediated PI3K/AKT survival signaling pathway by binding to RXR-α, thus identifying CF31 as a new lead for a class of anticancer agents targeting this newly identified cancer survival pathway.

CF31 binds to RXR-α in a unique mode and acts as a RXR-α antagonist. The LBP of RXR is highly restrictive to flexible and elongated ligands. The published crystal structures of RXR-α bound to natural or synthetic ligands show that a carboxylate group in these ligands forms salt bridges with basic residue Arg316 at the end of the L-shaped RXR-α LBP to establish anchoring ionic interaction for stabilization (41–43). However, CF31 lacks such a carboxylate moiety (Fig. 1A) and is therefore incapable of interacting with Arg316. This was supported by our mutagenesis study, which showed that Arg316 was not required for its antagonist effect (Fig. 2C). Although Arg316 was not required for CF31 binding, our MM-PB/SA analysis suggested that the binding of CF31 was stabilized by its extensive van der Waals interactions with several hydrophobic residues in the RXR-α (Fig. 2E), thus revealing a distinct binding mode for CF31. It is noteworthy that mutation of Arg316, which completely impaired the transactivation function of 9-cis-RA, did not show much effect on CD3254 (Fig. 2C), implying that CD3254 binding to RXR-α does not require its ionic interaction with Arg316. Very recently, crystal structure studies showed that begelovin, a RXR-α agonist lacking the carboxylate moiety, bound to RXR-α (54) in a mode similar to CF31. Thus, the LBP of RXR-α is more flexible than expected to mediate diverse activities of compounds with different structural features.

CF31 effectively inhibited constitutive and inducible AKT activation and cell survival in several cancer cell lines (Figs. 4 and 5). It was much more effective than sulindac, an nonsteroidal anti-inflammatory drug (NSAID) that was previously reported to inhibit tRXR-α–mediated AKT activation (29), and was comparable with K-80003, an improved sulindac analog (29), on inhibiting AKT activation (Fig. 5A and data not shown). The inhibitory effect of CF31 on AKT activation occurred at concentrations under that CF31 could bind to RXR-α, suggesting that it achieved its inhibitory effect on AKT activation by RXRα binding. In support of this conclusion, we showed that knocking down RXR-α expression by RXR-α siRNA impaired its inhibitory effects on basal and TNF-α–induced AKT activation (Fig. 5D), whereas overexpression of tRXR-α resulted in an enhancement (Fig. 5E).

Another unique property of CF31 is its ability to convert TNF-α from a survival molecule to a killer of cancer cells. TNF-α is a multifunctional cytokine that plays roles in diverse cellular events, such as cell survival and death (47, 48, 55, 56). The apoptotic effect of TNF-α is mediated by caspase-8–dependent apoptotic pathway, whereas its survival function involves activation of PI3K/AKT and NF-κB pathways. Because the death effect of TNF-α is antagonized or suppressed by its abnormally elevated survival function in cancer cells, TNF-α often acts as a survival instead of killer in the cells (47, 48, 55, 56). Because TNF-α is produced by malignant or host cells in the tumor microenvironment but not in normal cells, there has been tremendous interest in developing strategies to change a tumor-promoting microenvironment to a tumor-inhibiting state by shifting TNF-α signaling from survival to death (47, 48, 57–59). Our previous discovery that tRXR-α mediates AKT activation by TNF-α provides an opportunity to convert TNF-α's function from survival to death in cancer cells by targeting tRXR-α (29). Our current study provides several lines of evidence that CF31 could act as such a converter through its ability to suppress tRXRα–mediated AKT activation by TNF-α. First, we found that CF31 could potently induce PARP cleavage in cancer cells when used together with TNF-α (Fig. 6). Moreover, combination of CF31 and TNF-α led to a synergistic effect on PARP cleavage (Fig. 6C) and caspase activation (Fig. 7A), showing that the apoptotic effect of CF31 requires the TNF-α signaling pathway. Second, induction of PARP cleavage by the combination treatment in cancer cells was associated with their inhibition of AKT activation (Fig. 6B), which was also tRXR-α dependent (Fig. 6D–F). In contrast, transfection of CA-AKT abolished the inducing effect of CF31/TNF-α combination on PARP cleavage (Fig. 6C). Thus, inhibition of AKT activation was essential for their induction of apoptosis. Third, induction of caspase-3 activation and PARP cleavage by the CF31/TNF-α combination was mediated by activation of caspase-8 (Fig. 7), a critical mediator of TNF-α–induced intrinsic apoptotic pathway. Thus, CF31 can serve to relieve the antiapoptotic function of AKT, leading to activation of TNF-α–dependent apoptosis. Together, our results identify CF31 as a new converter of TNF-α survival signaling in cancer cells. With its ability to shift TNF-α signaling from survival to death by its unique RXR-α binding, CF31 represents a promising lead for a class of RXR-α modulators that selectively induce apoptosis of cancer cells. Because of the proven natural product drug discovery track record, identification of a natural product that targets an RXR-mediated cell survival pathway that regulates PI3K/AKT may offer a new therapeutic strategy to kill cancer cells.

No potential conflicts of interest were disclosed.

Conception and design: G.-H. Wang, F.-Q. Jiang, J.-Z. Zeng, Y. Su, X.-S. Yao, X.-K. Zhang

Development of methodology: G.-H. Wang, F.-Q. Jiang, Z.-P. Zeng, F. Chen, Y. Dai, J. Liu, J. Liu, H. Zhou, X.-K. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.-H. Wang, F.-Q. Jiang, Z.-P. Zeng, Y. Dai

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.-H. Wang, F.-Q. Jiang, Z.-P. Zeng, F. Chen, Y. Dai, J.-B. Chen, H.-F. Chen, Y. Su, X.-K. Zhang

Writing, review, and/or revision of the manuscript: Z.-P. Zeng, Y. Su, X.-S. Yao, X.-K. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z.-P. Zeng, J.-B. Chen, Y. Su, X.-S. Yao, X.-K. Zhang

Study supervision: X.-K. Zhang

Preparation of natural products and pure compounds for activity test: Y.-H. Duan

This work was supported by Grants from the U.S. Army Medical Research and Material Command (W81XWH-11-1-0677), the NIH (CA140980 and GM089927), the 985 Project from Xiamen University, the National Natural Science Foundation of China (NSFC-91129302, NSFC-30873146 and NSFC-81001664), and the Fundamental Research Funds for the Central Universities (2010121100, 2011121058, and 2010111081).

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.