Vitamin E γ-Tocotrienol Inhibits Cytokine-Stimulated NF-κB Activation by Induction of Anti-Inflammatory A20 via Stress Adaptive Response Due to Modulation of Sphingolipids

Nuclear factor κB is a central transcription factor that regulates immune functions and cellular survival, and therefore plays critical roles in inflammation and cancer development (1). Under resting conditions, NF-κB p50 and p65 are bound to inhibitory IκBα that sequesters inactive NF-κB complex in the cytoplasm. During inflammation, endotoxin or cytokines such as TNF-α activate the assembly of receptor proximal signaling complexes containing receptor-interacting protein serine/threonine kinase and TNFR-associated factors (2). This receptor assembly involves ubiquitylation and phosphorylation of RIP1 and leads to recruitment of the IκB kinase (IKK) complex to the receptor in the proximity of its upstream TGF-β–activated kinase 1 (TAK1). Activated TAK1 interacts with regulatory NEMO/IKKγ and stimulates the IKKs. Subsequently, activated IKKs phosphorylate IκBα, which targets IκBα for ubiquitination and proteasomal degradation. As a result, NF-κB p50 and p65 dimers are released so that they can translocate to the nucleus, where active NF-κB binds to consensus target sequences in many promoters.

Activation of NF-κB leads to upregulation of many genes, including proinflammatory cytokines and proteins that regulate inflammation and promote proliferation and survival of many types of cells, including immune and cancer cells. To prevent excessive immune response, activation of NF-κB is tightly controlled (3). Several enzymes have been identified as negative regulators of NF-κB signaling, including cylindromatosis (CYLD), A20, and cellular zinc finger anti–NF-κB (Cezanne), a member of the A20 family (4, 5). In particular, A20 and Cezanne, both of which are NF-κB target proteins with ubiquitin-editing activity, are induced by NF-κB to prevent its prolonged and aberrant activation (5–7). A20 has been demonstrated to inhibit cytokine-triggered NF-κB activation via its ubiquitin-editing function (8–11), although a recent study indicates that the deubiquitinase activity of A20 is not required for controlling NF-κB signaling (12). Alternatively, A20 may also antagonize NF-κB activation by interaction with NEMO in a noncatalytic manner to blunt activation of TAK1 and IKKs (13).

Because of the regulatory role of NF-κB in inflammation and cancer, targeting NF-κB activation has been recognized as a potential effective strategy for preventing and treating chronic diseases. Many natural products have been shown to inhibit NF-κB in cell-based studies and animal models. For instance, γ-tocotrienol (γTE), a natural form of vitamin E rich in palm oil, has been reported to inhibit NF-κB activation in leukemia KBM-5 and other cancer cells (14) as well as macrophages (15, 16). Consistently, γTE supplementation inhibits proinflammatory cytokines in animals and human subjects (17, 18). Despite these interesting results, the molecular mechanism responsible for the anti–NF-κB effect has not been identified. In this study, we investigated inhibitory effects and mechanism of γTE on NF-κB in murine RAW 264.7 macrophages and primary bone marrow–derived macrophages (BMDMs). Our study revealed a novel anti–NF-κB mechanism in which γTE induced upregulation of NF-κB inhibitor A20 via altering sphingolipid metabolism and cellular stress.

γTE (>97% pure) was a gift from BASF (Ludwigshafen, Germany). Recombinant mouse TNF-α and IL-1β were from Sigma-Aldrich (St. Louis, MO). Primary Abs against phospho-IκBα, IκBα, and all the secondary Abs were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Abs for phospho-JNK, JNK, A20, phospho-p65, phospho–translation initiation factor 2α (eIF2α), and eIF2α were from Cell Signaling Technology (Beverly, MA). Inhibitors for MEK (U0126), p38 MAPK (SB202190), and PI3K were from Calbiochem (La Jolla, CA). Myriocin from Mycelia sterilia, C8-dihydroceramide (dhCer), MTT, and all other chemicals were from Sigma-Aldrich. Cell culture media were obtained from American Type Culture Collection (Manassas, VA).

Murine RAW 264.7 macrophages from American Type Culture Collection were routinely maintained in DMEM with 10% FBS. Confluent cells were seeded and allowed to attach overnight at 7 × 10⁵ or 5 × 10⁶ per well in a 24- or 6-well plate, respectively. γTE stock solutions were initially made in DMSO and then diluted in 10 mg/ml fatty acid–free BSA. In some studies, γTE in DMSO was directly diluted in cell culture media without BSA. Our unpublished data showed that similar results were observed with or without BSA. Confluent cells were incubated in DMEM containing 1% FBS and 0.05% DMSO (control) or γTE for indicated times and then stimulated by TNF-α or IL-1β. Cell viability was determined by MTT assays.

BMDMs from mice were prepared according to published protocols (16, 19). The protocol on animal use was approved by the Animal Care and Use Committee at Purdue University and was strictly followed. Briefly, bone marrow was obtained by flushing femur from 7- to 8-wk-old C57BL/6 black mice (Harlan Laboratories, Indianapolis, IN). Suspension cells were cultured in 10-cm dishes in the DMEM media containing 10% FBS with 100 U/ml penicillin, 100 μg/ml streptomycin, and 100 U/ml M-CSF for 5 d. Attached cells were replenished by fresh media and incubated for an additional 2 d. Cells were harvested using nonenzymatic dissociation solution (Sigma-Aldrich) and characterized using flow cytometry after staining with PE-conjugated anti-mouse F4/80/EMR1 (no. FAB5580P) by Cell Lab Quanta SC MPL flow cytometer (Beckman Coulter, Brea, CA) with excitation at 488 nm. Cells with >95% purity were seeded in 24- or 6-well plates for subsequent studies. During experiments, cells were incubated in DMEM containing 1% FBS and 0.05% DMSO (control) or γTE for indicated times and then stimulated by TNF-α.

A20^−/− and A20^+/+ mouse embryonic fibroblasts (MEFs) were prepared from the knockout and wild-type mice (7). These cells were immortalized with SV40 large T Ag and cultured in DMEM with 10% FBS, 50 μM 2-ME, 100 U/ml penicillin-streptomycin, and 2 mM l-glutamine. Cell passages less than six were used in the study.

mRNAs of A20 (TNFAIP3) and GAPDH were quantified by quantitative PCR with SYBR Green via service contract by ARQ Genetics (Bastrop, TX). Data were obtained with samples from three independent experiments.

Cells were lysed in a lysis buffer containing Tris-EDTA, 1% SDS, 1 mM DTT, 2 mM sodium vanadate, and protease inhibitor cocktails (Sigma-Aldrich). Cytosolic and nuclear proteins were extracted using the Pierce kit (Pierce, Rockford, IL). Proteins (20–50 μg) were loaded on Bio-Rad precast SDS-PAGE gels. Resolved proteins were transferred onto a polyvinylidene difluoride membrane (Millipore, Billerica, MA) and probed by Abs. Membranes were exposed to chemiluminescent reagent (PerkinElmer, Waltham, MA) and visualized on Kodak films. In all experiments, after being probed by Abs for target proteins, polyvinylidene difluoride membranes were stripped with Ab-stripping solution from EMD Millipore (Billerica, MA) and reimmunoblotted with Abs for internal controls such as β-actin. Results of Western blots were quantified by ImageJ.

To measure sphingolipids, total lipids were extracted as previously described (20). Briefly, cell pellets were resuspended in chloroform/methanol/water (10:5:1 [v/v/v]), followed by addition of internal standards containing 0.5 nmol C12:0-ceramide (Cer), C17-sphingosine, C17-sphinganine, C25:0-Cer, and C12:0-sphingomyelin from Avanti Polar Lipids (Alabaster, AL). After tip sonication, samples were incubated at 48°C overnight. The next day, 100 μl aliquots was taken out and dried under N₂ for measurement of total phospholipids (Wako Chemicals, Neuss, Germany). For the rest of the samples, 75 μl 1 M KOH in methanol was added, sonicated for 30 min, and then incubated at 37°C for 2 h and dried in a nitrogen evaporator.

The liquid chromatography–tandem mass spectrometry (LC-MS/MS) method for sphingolipid analysis was slightly modified based on a published protocol by Merrill et al. (20). Briefly, all samples were analyzed using the Agilent 6460 triple quadrupole mass spectrometer coupled with the Agilent 1200 rapid resolution HPLC (Agilent Technologies, Santa Clara, CA) with identification of each sphingolipid by multiple reaction monitoring. The MS/MS conditions used for multiple reaction monitoring detection of sphingolipids were the same as previously described (20). The LC column for sphingoid bases and ceramides was Agilent XDB-C18 (4.6 × 50 mm) with particle size of 1.8 μm, whereas the LC column for sphingomyelins was Agilent XDB-C8 (2.1 × 50 mm) with Zorbax 3.5 μm. Mobile phases A and B contain 5 mM ammonium formate and have, respectively, methanol–H₂O–formic acid (74:25:1 [v/v/v]) and methanol–formic acid (99:1 [v/v]). Agilent MassHunter software was coupled with the LC-MS/MS for data acquisition and analysis.

We used one-way ANOVA and the Student t test to analyze the data or log-transformed results. A p value <0.05 was considered significant.

Similar to previous studies (14, 16), γTE inhibited TNF-α–triggered NF-κB activation in RAW 264.7 macrophages, as indicated by decreased phosphorylation and degradation of IκBα in the cytosol and diminished phosphorylation of p65 in the nucleus compared with vehicle controls (Fig. 1A). The inhibitory potency by γTE appeared to be in a time-dependent fashion. Specifically, relatively short 2 h incubation with γTE did not lead to consistent inhibition of NF-κB (not shown) and longer preincubation resulted in stronger inhibition of phosphorylation of IκBα (compare 16 versus 8 h in Fig. 1A). Besides RAW macrophages, γTE inhibited TNF-α–stimulated phosphorylation of IκBα in BMDMs (Fig. 1B). Additionally, γTE also attenuated TNF-triggered JNK phosphorylation, another downstream signal in response to the activation of TNF receptor (Fig. 1C).

It has been well established that TNF-triggered IκBα phosphorylation is catalyzed by IKKs, which are regulated by the upstream TAK1 (2). Consistent with the key role of TAK1 in NF-κB activation, a specific inhibitor of TAK1, but not those of protein kinase Cs or PI3K (not shown), markedly inhibited TNF-α–stimulated phosphorylation of IκBα in RAW macrophages (Fig. 1D). Importantly, γTE treatment dampened TNF-α– or IL-1β–stimulated phosphorylation of TAK1 (Fig. 1E), which is in agreement with a previous study in which γTE blocked NF-κB–dependent reporter gene transcription induced by overexpression of TAK1 in A293 cells (14). These observations suggest that γTE likely targets upstream signaling that is important to the activation of TAK1 and IKKs.

Activation of TAK1 and subsequent NF-κB by TNF-α requires assembly of TNFR-proximal adaptor proteins and involves a series of ubiquitylation and deubiquitylation events (8, 9). These ubiquitylation events can be interrupted by ubiquitin-editing enzymes such as CYLD and A20, which are recognized as endogenous NF-κB inhibitor (21–23). Alternatively, A20 is shown to block activation of TAK1 and IKK via a ubiquitin-independent mechanism by direct interaction with NEMO (13). Because γTE inhibits TAK1 phosphorylation (Fig. 1E), and overexpression of CYLD or A20 blocks NF-κB activation in cells and animals (13, 21–23), we investigated whether γTE has any impact on these NF-κB negative regulators. We found that γTE treatment led to enhanced expression of A20 but had no effect on CYLD in both RAW cells and bone marrow–derived primary macrophages (Fig. 2A–C). Enhanced A20 expression was observed as early as 4 h (not shown) and sustained during longer incubation with γTE. Because A20 has recently been shown to blunt TNF-activated JNK activation (24), the induction of A20 by γTE may also explain γTE’s suppression of TNF-α–stimulated activation of JNK.

Activation of NF-κB is thought to promote cancer cell survival and contributes to drug resistance. Because γTE has been reported to suppress NF-κB activation in cancer cells (14), we next examined whether γTE induces A20 in human A549 lung, PC-3 prostate, and MCF7 breast cancer cells. We found that γTE treatment inhibited TNF-triggered phosphorylation of IκBα (Fig. 2D) and resulted in induction of A20 expression in these cells (Fig. 2E). Considering the well-established negative regulation of NF-κB by A20, γTE-mediated inhibition of NF-κB likely stems from its induction of A20 in these cells.

To further examine the role of A20 induction in γTE’s anti–NF-κB effect, we used A20^−/− MEFs. Similar to observations in macrophages, γTE treatment induced A20 expression in A20^+/+ MEFs, whereas A20 was not detectable in A20^−/− cells (Fig. 3A). Consistent with the anti–NF-κB role of A20, TNF-α–induced phosphorylation of IκBα was more pronounced in A20^−/− cells than that in wild-type cells (Fig. 3B, compare lanes 3 versus 4). The suppressive effect of γTE on phosphorylation of IκBα was significantly, although partially, reversed in A20^−/− cells compared with A20^+/+ cells (Fig. 3B, compare lanes 7 versus 8). These results, together with known functions of A20, support the idea that A20 induction plays a key role in γTE’s inhibitory effect on NF-κB. Alternatively, we observed that γTE still inhibited NF-κB in the absence of A20, suggesting that other mechanisms are involved in antagonizing NF-κB, especially under A20 knockout condition. Consistent with this hypothesis, γTE treatment increased another NF-κB inhibitor Cezanne in A20^−/− MEFs (Fig. 3C) but had no effect on CYLD (not shown). Interestingly, Cezanne was not induced by γTE in A20^+/+ MEFs (Fig. 3C).

In search of the mechanism underlying A20 induction, we found that γTE enhanced A20 mRNA, indicating that A20 was upregulated at the transcriptional level (Fig. 4A). Because A20 expression is known to be regulated by NF-κB (25), we investigated whether γTE has any effect on the basal activity of NF-κB. To this end, γTE treatment mildly increased basal IκBα phosphorylation (Fig. 4B). As previously observed in other types of cells (26, 27), γTE enhanced phosphorylation of JNK in macrophages (Fig. 4B). Additionally, γTE increased phosphorylation of eIF2α, a marker of endoplasmic reticulum (ER) stress (Fig. 4C). These data indicate that γTE treatment induced moderate stress to the cells, although no obvious changes in cell morphology or viability were observed in confluent macrophages based on microscopic examination and the MTT assays (data not shown). Because phosphorylation of eIF2α has been demonstrated to be sufficient to activate NF-κB (28) and A20 expression is regulated by NF-κB (25), we reason that γTE-induced basal activation of NF-κB and ER stress is responsible for A20 upregulation.

We have previously demonstrated that vitamin E forms including γTE induced JNK phosphorylation and ER stress in MCF7 cells and inhibited prostate cancer cell growth by modulation of de novo synthesis of the sphingolipids pathway (Fig. 5A) (27, 29, 30). Because γTE appeared to induce cellular stress in the present study, we examined whether it has any effect on sphingolipids in macrophages by a sphingolipodomic approach using LC-MS/MS (27). Consistent with a previous study (31), C16:0-, C24:0- and C24:1- dhCer or Cer were the predominant (dihydro)ceramides in macrophages (Supplemental Table I). Compared with controls, γTE treatment significantly enhanced individual and total dhCer (Fig. 5B, Supplemental Table I). In contrast, γTE treatment resulted in decreased total and specific ceramides at 8 h, but increased C24:0-Cer but had no impact on total or other ceramides at 16 h (Fig. 5B, Supplemental Table I). Prolonged incubation with γTE for 16 h but not 8 h decreased total sphingomyelin (Fig. 5, Supplemental Table I).

Modulation of sphingolipids, including increase of dhCer, has been shown to cause ER stress (32). To examine the role of sphingolipid modulation in γTE-induced effects, we pharmacologically inhibited de novo sphingolipid biosynthesis with myriocin, a specific inhibitor of serine palmitoyl-CoA transferase that catalyzes the rate-limiting reaction in de novo synthesis of sphingolipids (Fig. 5A). Myriocin blocked γTE-caused increase of dhCer (Fig. 6A) and partially reversed γTE-induced upregulation of A20 and phosphorylation of JNK (Fig. 6B). More importantly, myriocin counteracted γTE’s inhibition of NF-κB activation (Fig. 6C).

To further investigate the role of increased dhCer in induction of cell stress and A20 as well as NF-κB activation in macrophages, we incubated RAW cells with C8-dhCer. Similar to γTE, C8-dhCer treatment did not change cell morphology but led to increased phosphorylation of JNK and IκBα as well as the ER stress marker eIF2α (Fig. 7). Moreover, similar to γTE, dhCer induced A20 expression (Fig. 7A) and attenuated TNF-α–caused phosphorylation of IκBα (Fig. 7B). These results, together with myriocin’s counteraction of γTE’s anti–NF-κB, indicate that modulation of de novo synthesis of sphingolipids, especially accumulation of dhCer, plays a significant role in γTE-mediated anti–NF-κB effects (Fig. 8).

We have identified a novel anti–NF-κB mechanism in which modulation of sphingolipids leads to cellular stress and upregulation of A20, a well-established NF-κB negative regulator (Fig. 8). The importance of A20 in anti–NF-κB action was verified by the observation that NF-κB activation was enhanced in the absence of A20 and, compared with A20^+/+ cells, γTE showed diminished inhibitory effect on NF-κB in A20^−/− cells. A causal role of sphingolipid modulation in γTE’s anti–NF-κB effect and induction of A20 and stress is supported by the following lines of evidence. γTE treatment led to significant increase of intracellular dihydroceramides, which are key sphingoid bases in de novo synthesis of sphingolipids and known to cause ER stress (32, 33). Exogenous addition of C8-dhCer induced similar cellular changes to those observed in γTE-treated cells, including upregulation of A20, increase of ER stress, and suppression of TNF-α–induced IκBα phosphorylation. Moreover, myriocin, which blocks de novo synthesis of sphingolipids and thus prevents accumulation of dhCer, counteracted γTE-caused anti–NF-κB effect and induction of A20. Although γTE was previously suggested to have anti-inflammatory effects by suppressing the proteasome activity (17), there is no evidence indicating that γTE inhibits the proteasome activity in the present study. This is because in contrast to proteasome inhibitors shown to increase TNF-α–induced phosphorylation of IκBα by blocking its proteasome-mediated degradation (34), γTE inhibited TNF-α–triggered phosphorylation of IκBα and upstream regulator TAK1.

The induction of A20 explains γTE’s inhibition of TNF-α–stimulated activation of TAK1, JNK, and NF-κB and has important biological implications. Although A20 has been shown to inhibit NF-κB via its ubiquitin-editing activity, a recent study suggests that A20 may also block NF-κB in a noncatalytic manner (12) via direct association with NEMO to inhibit TAK1 and IKKs (13). Regardless of mechanistic action, A20 as a key regulator of immunopathology has been firmly established in vivo. For instance, A20^−/− mice developed severe inflammation and hypersensitivity to LPS or TNF-α and died prematurely as a result of unrestrained inflammation compared with wild-type mice (7, 35). Mice lacking A20 in dendritic cells spontaneously developed colitis (36). Besides NF-κB, A20 has also been shown to antagonize TNF-induced JNK signaling (37, 38) by targeting apoptosis signal–regulating kinase1 and mediating its degradation (24). Consistently, ectopic expression of A20 suppressed TNF-induced activation of NF-κB and JNK (13, 24). In agreement with these studies, deficiency in A20 is associated with inflammatory diseases in humans such as Crohn’s disease, arthritis, and autoimmune type I diabetes (10). Additionally, A20 has been recognized as a tumor suppressor in lymphomas, as A20 was inactivated in many lymphomas and reintroduction of A20 promoted apoptosis and growth arrest of cancer cells (4). Given that NF-κB target genes regulate cell proliferation, cytokines, and survival, the induction of A20 and consequent inhibition of NF-κB by γTE likely contributes to its antiproliferative, proapoptotic, anti-inflammatory, and immunomodulatory effects (14, 17, 18, 29).

We demonstrate a novel immunomodulatory role of dhCer and propose links among sphingolipid metabolism, ER stress, and anti–NF-κB activities. Although dihydroceramides used to be considered inert precursors of ceramides, recent evidence indicates that they are bioactive (39, 40). For instance, dihydroceramides have been shown to induce growth arrest in cancer cells (29, 41) and induce ER stress and autophagy (32, 33). In the present study, exogenous addition of C8-dhCer blocked TNF-stimulated NF-κB activation, upregulated A20, and appeared to cause ER stress in macrophages as indicated by enhanced phosphorylation of eIF2α. Interestingly, ER stress related eIF2α phosphorylation has been shown to be necessary and sufficient to activation of NF-κB (28). Consistently, both γTE and C8-dhCer moderately elevated basal levels of IκBα phosphorylation. Because A20 is a NF-κB target gene (25), the basal activation of NF-κB is likely responsible for the upregulation of A20. This is further supported by the fact that ER stress inducers such as thapsigargin, ionophore A23187, and tunicamycin have been shown to enhance A20 and blunt cytokine-induced NF-κB activation (42). It is noteworthy that besides γTE, some other bioactive phytochemicals may have similar anti-inflammatory mechanism. For instance, we found that curcumin inhibited NF-κB, modulated sphingolipids, and upregulated A20 in macrophages (Y. Wang and Q. Jiang, unpublished data).

Despite the unambiguous role of A20 in inhibition of NF-κB, γTE’s anti–NF-κB action was not exclusively dependent on A20 because γTE diminished IκBα phosphorylation even in A20^−/− cells. In this regard, we found that γTE induced another inhibitor Cezanne in the absence of A20. Cezanne is an A20 family member and is also regulated by NF-κB (5). Previously, we have shown that γTE blocked IL-13–stimulated JAK-STAT6 activation by upregulation of prostate apoptosis response-4 (PAR-4) that interacts with atypical protein kinase C to limit STAT6 signaling in lung epithelial cells (43). Interestingly, PAR-4 is known to be upregulated by ER stress (44). Therefore, we propose that upregulation of A20, Cezanne, and PAR-4 by γTE (and possibly other phytochemicals) in response to ER stress via sphingolipid modulation represents a stress-adaptive mechanism that prepares cells for antagonizing proinflammatory insults. This theory should be further verified by investigation of anti–NF-κB mechanisms by other phytochemicals in various cell types.

In addition to the present study in macrophages, γTE’s modulatory effect on sphingolipid metabolism has previously been observed in other types of cells. γTE induced accumulation of dhCer and dihydrosphingosine in prostate cancer cells where sphingolipid modulation appeared to play a key role in induction of apoptosis and autophagy (29, 30). γTE has been shown to cause ER stress and JNK phosphorylation by potentiating de novo synthesis of sphingolipids (27). Our unpublished results indicate that the mechanism underlying sphingolipid modulation by γTE appears to be rooted in its direct inhibition of dhCer desaturase (DEGS1), the enzyme converting dihydroceramides to ceramides (Y. Jang and Q. Jiang, manuscript in preparation). Alternatively, prolonged incubation of γTE likely induces other changes besides increase of dhCer. In this study, γTE treatment for 16 h resulted in decrease of sphingomyelin but increase of C24:0-Cer (Supplemental Table I). These changes may intensify inhibition of NF-κB, which should be further characterized.

We thank Amber S. Jannasch for help with LC-MS/MS analysis of sphingolipids.

The authors have no financial conflicts of interest.