The Antiinflammatory Sesquiterpene Lactone Parthenolide Inhibits NF-κB by Targeting the IκB Kinase Complex¹

In most cell types, the dimeric transcription factor NF-κB is trapped in the cytoplasm by binding to an inhibitory IκB³ protein (1). IκBs retain NF-κB in the cytoplasm by masking the nuclear localization sequence contained in the Rel homology domain (2, 3). Numerous stimuli including cytokines, phorbol esters, and T cell activation lead to phosphorylation, ubiquitinylation, and the subsequent degradation of IκB proteins. The DNA-binding subunits of NF-κB immigrate into the nucleus and activate the expression of numerous target genes that are important for the inflammatory and immune response, but also for other functions, such as the regulation of apoptosis (4) and cell proliferation (5, 6). The inducible phosphorylation of IκB-α is mediated by kinases contained in a high m.w. complex of relative molecular mass (M_r 700–900 kDa), the multisubunit IκB kinase complex (IKC) (7). Two of the IκB kinases contained in the IKC (IKKα and IKKβ) were independently cloned by several laboratories (8). IKKβ is attached to Nf-κB essential modulator/IKKγ, a protein composed of several coiled-coil motifs (9, 10). Both IKKs are found in association with the scaffold protein IKK complex-associated protein, which also binds to NF-κB-inducing kinase (NIK) (11). The MAP3 kinase NIK was identified as a direct IKK activator (12, 13), but there is recent evidence for further IKK-activating kinases. The Cot/Tpl-2 kinase feeds CD3/CD28-derived signals into the IKC (14), and IL-1-derived signals enter the IKC via the TAK1 kinase (15). MEKK1 was identified as an interactor and activator of both IKKs (16, 17), but dominant-negative forms of MEKK1 inhibit TNF-α-induced activation only partially (18), and NF-κB induction induced by overexpression of NIK can only partially be blocked by dominant-negative forms of both IKKs (16). These findings suggest that the incompletely defined IKC receives input from so far unidentified proteins.

The broad spectrum of immunologically relevant target genes in combination with the analysis of NF-κB knockout mice reveals NF-κB as a key mediator of the immune response (19). Therefore, the pathways leading to the activation of NF-κB are frequent targets for a variety of antiinflammatory drugs. Well-known antiinflammatory substances such as glucocorticoids or aspirin and salicylate exert, at least a part of their effects, by inhibiting NF-κB activity (20, 21). We have previously shown that sesquiterpene lactones (SLs) isolated from extracts of Mexican Indian medicinal plants act as specific inhibitors of NF-κB (22). SLs are traditionally used as antiinflammatory substances, and the SL parthenolide prevents the degradation of IκB-α and IκB-β induced by a variety of stimuli (23).

In this study, we identify the IKC as the molecular target for the NF-κB inhibitor parthenolide. This SL prevents the NIK- and MEKK1-induced activation of both IKKs, but does not interfere with the induced activation of JNK and p38.

HeLa cells were grown in DMEM supplemented with 10% FCS and 1% (v/v) penicillin/streptomycin (all from Life Technologies, Eggenstein, Germany). Transfections were performed using the Superfect reagent (Qiagen, Chatsworth, CA), according to the instructions of the manufacturer. The reporter plasmid (κB)₃-luc (24), expression vectors for IKKα and -β (25), MEKK1 and MEKK1▵ (18), NIK (12), TRAF-2 (26), and the bacterial expression vector for GST-Jun 5–89(5–89) (27) were previously described. The bacterial GST-IκB 1–54(1–54) expression vector was made by inserting the BamHI/XhoI fragment from pGEX6P1-IκB-α into pGEX4T3 (Pharmacia, Piscataway, NJ).

HeLa cells (2 × 10⁶) were washed twice with cold PBS and harvested by scraping with a rubber policeman. The pellet was resuspended in TOTEX buffer (20 mM HEPES/KOH, pH 7.9, 0.35 M NaCl, 20% (v/v) glycerol, 1% (v/v) Nonidet P-40, 1 mM MgCl₂, 0.5 mM EDTA, 0.1 mM EGTA, 1 mM PMSF) and incubated on ice for 30 min. The cell debris was pelleted upon centrifugation with 14,000 rpm at 4°C for 10 min, and equal amounts of protein contained in the supernatant were tested for DNA-binding activity essentially as described (23). Briefly, the extracts were incubated with 2 μg BSA, 2 μg poly(dI-dC), and 10,000 cpm of a ³²P-labeled oligonucleotide on ice in 5× binding buffer (100 mM HEPES/KOH, pH 7.9, 20% (w/v) Ficoll 400, 1 mM DTT, and 300 mM KCl). The supershift experiments were performed by preincubating the total cell extracts with 2 μg of the respective Abs for 15 min at 4°C. The free and protein-bound oligos were separated on a 4% polyacrylamide gel. Gel and running buffer were identical and contained 25 mM Tris, 25 mM boric acid, and 0.5 mM EDTA. The gel was dried after electrophoresis and exposed to a Kodak XAR5 film. The double-stranded oligonucleotide used for EMSAs contains a single NF-κB binding site, which is shown underlined: 5′-AGTTGAGGGGACTTTCCCAGGC-3′.

Proteins were extracted from cells in Nonidet P-40 lysis buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1 mM PMSF, 10 mM NaF, 0.5 mM sodium vanadate, leupeptine (10 μg/ml), 1% (v/v) Nonidet P-40, and 10% (v/v) glycerol) and directly analyzed by either Western blotting or immunoprecipitation. Proteins were immunoprecipitated after preclearance by adding 1–2 μg of Ab and 25 μl of protein A/G-Sepharose. After rotation for at least 4 h on a spinning wheel at 4°C, the immunoprecipitates were washed five times in lysis buffer. Immunoprecipitates were boiled in 1× SDS sample buffer and separated by SDS-PAGE before blotting to a polyvinylidene difluoride membrane (Millipore, Bedford, MA). The membrane was then incubated in a small volume of TBST containing various dilutions of the primary Abs (α-Phospho-IκB-α and α-Phospho-p38, New England Biolabs, Beverly, MA; α-I-κB-α, α-p38, α-IKKγ (FL-419), α-IKKβ (H-4), and α-IKKγ (B-3), Santa Cruz; α-IKKα and α-JNK, PharMingen, San Diego, CA; α-Flag (M2), Sigma, St. Louis, MO; α-HA (12CA5), Roche Molecular Biochemicals, Mannheim, Germany; α-Myc (9E10), Upstate Biotechnology, Lake Placid, NY). The respective proteins were detected by incubation with primary Abs at concentrations of 0.5 μg/ml. Different concentrations were only used for the α-Flag Ab (0.2 μg/ml) and the α-IKKγ (B-3) Ab (1 μg/ml). After binding of an appropriate secondary Ab coupled to HRP, proteins were visualized by enhanced chemiluminescence according to the instructions of the manufacturer (Amersham Lifescience, Arlington Heights, IL).

Harvested cells were lysed in reporter lysis buffer (25 mM Tris-phosphate, 2 mM DTT, 2 mM CDTA, 10% (v/v) glycerol, 1% (v/v) Triton X-100). Luciferase activity was determined in a luminometer (Duo Lumat LB 9507; Berthold, Nashua, NH) by injecting 50 μl of assay buffer (40 mM tricine, 2.14 mM (MgCO₃)₄Mg(OH)₂ × 5 H₂O, 5.34 mM MgSO₄, 0.2 mM EDTA, 66.6 mM DTT, 540 μM CoA, 940 μM luciferin, 1.06 mM ATP) and measuring light emission for 10 s. The results were normalized to the activity of β-galactosidase expressed by a cotransfected lacZ gene under the control of a constitutive RSV promoter.

Cells were lysed in Nonidet P-40 lysis buffer, and the IKK or JNK proteins contained in the cell lysate were immunoprecipitated. The precipitate was washed three times in lysis buffer and two times in kinase buffer (20 mM HEPES/KOH, pH 7.4, 25 mM β-glycerophosphate, 2 mM DTT, 20 mM MgCl₂). The kinase assay was performed in a final volume of 20 μl kinase buffer containing 2 μg of bacterially expressed GST-IκB-α 1–54(1–54) or GST-c-Jun 5–89(5–89) protein, 20 μM ATP, and 5 μCi [γ-³²P]ATP. After incubation for 20 min at 30°C, the reaction was stopped by the addition of 5× SDS sample buffer. After separation by SDS-PAGE, the gel was fixed, dried, and quantified using a phosphor imager.

To identify molecular targets for the antiinflammatory SL parthenolide, we tested whether the lack of IκB degradation in the presence of parthenolide can be attributed to an impaired phosphorylation of IκB. HeLa cells were stimulated with TNF-α in the absence or presence of increasing concentrations of parthenolide, and total cell extracts were analyzed in parallel for NF-κB DNA binding, IκB-α phosphorylation, and IKK activation (Fig. 1,A). The TNF-α-induced DNA binding of NF-κB, but not of a constitutively DNA-binding protein, was completely lost when parthenolide was present in concentrations exceeding 5 μM. Only the induced DNA/protein complex, but not the constitutively binding band, disappeared upon addition of a 75-fold excess of unlabeled oligonucleotide (data not shown). Further analysis of the induced NF-κB/DNA complex with supershifting Abs revealed that the TNF-α-induced NF-κB dimer contains the typical p50 and p65 DNA-binding subunits (data not shown). The parthenolide-mediated inhibition of DNA binding was paralleled by the prevention of TNF-α-induced IκB-α degradation, as detected by Western blotting. Determination of IκB-α phosphorylation by immunoblotting using an Ab recognizing phosporylated serin 32 of IκB-α revealed a loss of induced phosphorylation in the presence of inhibitory concentrations of parthenolide. Next we tested whether the prevention of IκB-α phosphorylation is due to an impaired kinase activity of the IKC. The endogenous IKC was isolated from a fraction of the HeLa cell extracts by immunoprecipitation using a monoclonal α-IKKα Ab, and its activity was analyzed by immune complex kinase assays. The TNF-α-induced IKC kinase activity was dose dependently lost in the presence of increasing concentrations of parthenolide. We then tested the impact of parthenolide on TNF-α-induced IKC activity by a complementary experimental approach. The endogenous IKC was isolated by immunoprecipitation with an α-IKKγ Ab and tested for its in vitro kinase activity. Also, this experimental setting revealed that the TNF-α-induced kinase activity of the IKC was lost in the presence of parthenolide (Fig. 1,B). The endogenous IKC is a heterodimer between IKKα and IKKβ, but gene disruption experiments revealed that mainly IKKβ is important for the TNF-α- and IL-1-induced phosphorylation and degradation of IκB-α (28, 29). To address the question, whether this NF-κB inhibitor would affect the activity of the inflammatory relevant IKKβ protein, we transfected HeLa cells with an epitope-tagged IKKβ expression vector before treatment with TNF-α in the absence or presence of parthenolide (Fig. 1 C). Likewise, the induced kinase activity of immunoprecipitated IKKβ was dose dependently impaired in the presence of increasing doses of the antiinflammatory SL. No inhibitory activity of the synthetic steroid dexamethasone on the induced activity of both IKKs was seen in a control experiment (data not shown).

The TNF-α-derived signals lead to the activation of several parallel and interconnected signaling cascades, resulting in the activation of NF-κB and the MAP kinases p38 and JNK (30). To test whether parthenolide targets the TNF-α-derived signals before the branch points leading to the activation of JNK and p38, we tested its impact on the activity of these two MAP kinases. HeLa cells were incubated with TNF-α, parthenolide, or a combination of both, and endogenous JNK activity was determined by immune complex kinase assays (Fig. 2). The TNF-α-induced JNK activation was unchanged in the presence of 20 μM parthenolide, a concentration that completely blocked NF-κB activation. Determination of p38 activation by immunoblotting using a phospho-specific Ab also revealed no effect on the induced p38 phosphorylation at Thr¹⁸⁰/Tyr¹⁸² in the presence of this SL. Also, the activation of JNK induced by overexpression of MEKK1 was not affected by parthenolide (data not shown). These results show that parthenolide targets a rather late signaling step in NF-κB activation after the separation of the TNF-α-derived signals.

Obvious direct candidates for the inhibitory activity of parthenolide are the IKKs. This possibility was tested by transiently transfecting HeLa cells either with a NF-κB-dependent reporter gene alone or in combination with expression vectors encoding IKKα and IKKβ (Fig. 3,A). The TNF-α-elicited NF-κB activation was efficiently impaired in the presence of parthenolide. In contrast, NF-κB-dependent transcription induced by overexpression of IKKα and IKKβ was not significantly affected by parthenolide. Next we investigated a possible impact of parthenolide on IKKα and IKKβ kinase activity in vitro. Following transfection, HeLa cells were stimulated with TNF-α, followed by immunoprecipitation of both IKKs. To measure a potential direct effect of parthenolide on the IKKs, the immunoprecipitated IKKs were subsequently incubated with this inhibitor in vitro (Fig. 3 B). Concentrations of parthenolide up to 50 μM remained without impact on the in vitro kinase activity of both TNF-α-activated IKKs, thus excluding these IκB-phosphorylating enzymes as direct targets for the inhibitor.

Parthenolide inhibits a TNF-α-derived signaling step after the diverging point of the signals leading to the activation of NF-κB, p38, and JNK, but does not directly block the activity of both IKKs. Previous studies have revealed an inhibitory function of SLs on NF-κB activated by a variety of stimuli, including CD3/CD28 ligation and treatment with hydrogen peroxide or phorbol esters (23), suggesting the IKC as a possible target. This assumption was experimentally tested by studying the impact of parthenolide treatment on NF-κB transcription induced by expression of various activators that initiate the activation process from different levels within the signaling cascade. NF-κB-dependent transcription was efficiently induced by expression of the TRAF-2 adaptor protein, which participates in the early events of TNF receptor signaling (Fig. 4,A). TRAF-2-elicited transcription was strongly and dose dependently impaired in the presence of parthenolide, thereby locating the SL-sensitive signaling step downstream from TRAF-2. The same experimental approach was applied for studying the effects of parthenolide on gene expression triggered by overexpression of the IKC activator MEKK1. NF-κB-dependent transcription induced by the overexpression of the catalytic domain of MEKK1 (MEKK1▵) was completely suppressed in the presence of the antiinflammatory SL (Fig. 4,B). We next assayed the effect of parthenolide on NIK-induced NF-κB activity (Fig. 4 C). Here, escalating doses of parthenolide only halved NIK-induced transcription without leading to the complete suppression of NF-κB activity that was seen upon expression of TRAF-2 or MEKK1▵.

We next asked whether the inhibition of NIK- and MEKK1▵-induced NF-κB activation is also reflected at the level of inducible NF-κB DNA-binding and IKK activation. To address this question, HeLa cells were transiently transfected with expression vectors encoding epitope-tagged NIK or MEKK1, followed by the incubation with parthenolide for 2 h, and determination of DNA-binding activity as measured by EMSAs (Fig. 5,A). Both MEKK1- and NIK-induced DNA binding of NF-κB was completely lost in the presence of this inhibitor. Similarly, IKKα-mediated phosphorylation of IκB-α induced by transient expression of MEKK1 or NIK was completely prevented by parthenolide (Fig. 5,B). A similar experimental approach also revealed a complete inhibition of MEKK1- and NIK-induced IKKβ kinase activity by preincubation with parthenolide (Fig. 5 C). This experimental setting also showed that parthenolide prevents not only the induced phosphorylation of the substrate protein IκB-α, but also of both IKKs (data not shown), thus inhibiting the coordinate activation of the IκB kinase activity downstream from the MAP3K level.

SL-containing plant extracts are frequently used in the Middle American Indian medicine for the treatment of infections and inflammation of the skin and other organs (31). Also, in the traditional Western medicine, alcoholic preparations of some SL-rich Compositae are externally applied for the treatment of rheumatic diseases and superficial inflammations. Only the 10α-methylpseudoguaianolide type of SLs displays an antiphlogistic activity, as revealed by physiological inflammation models (22). Parthenolide possesses two reactive centers in the form of an exomethylene group and an epoxide ring that are required for irreversibly inactivating its target molecule(s), probably by reacting with exposed cysteines (23). The molecular reactivity and specificity of the various SLs are determined by the molecular geometry, lipophilicity, and the environment of the target sulfhydryl groups. Therefore, it is well possible that different SLs act on distinct target structures. It was recently suggested that the SL helenalin directly alkylates the NF-κB p65 subunit (32). However, this mechanism is obviously not applying for parthenolide, which does not interfere with DNA-binding activity of activated NF-κB in vitro (23). Furthermore, a direct inactivation of a DNA-binding subunit would not be compatible with the results showing that parthenolide did not affect NF-κB-dependent transcription induced by both IKKs (compare Fig. 3 A). Therefore, alkylation of NF-κB p65 may be specific for the highly reactive SL helenalin, which causes the inactivation of multiple target enzymes and therefore displays a high cytotoxicity (33).

Parthenolide prevents IκB degradation by a variety of stimuli, suggesting that it interferes with a common step in NF-κB activation. The central role of the IKC makes it an ideal target for NF-κB-inhibiting and NF-κB-activating substances. It was proposed that the antiinflammatory agents aspirin and salicylate inhibit IKKβ in vivo and in vitro by directly binding this kinase, thereby competing with ATP (21). The HTLV Tax protein manipulates NF-κB activity by targeting the IKC on a pathway either employing MEKK1 (34) or NIK and both IKKs (35, 36, 37). Using a variety of experimental approaches, this study identifies the IKC as the molecular target for the NF-κB-inhibiting activity of parthenolide, thereby explaining the previous finding that parthenolide inhibits NF-κB activation elicited by a great variety of stimuli (23). This antiinflammatory SL efficiently inhibits the signals derived from MAP3Ks. At the level of gene expression, the MEKK1-derived signal was more efficiently inhibited than the NF-κB-activating signal submitted by NIK. Since parthenolide was only present for 8 h, this incomplete inhibition may be due to kinetic differences in transcriptional activation triggered by the NF-κB activators TRAF-2, MEKK1▵, and NIK, which would be reflected at the level of the accumulated luciferase reporter enzyme. Alternatively, it might be possible that NIK also stimulates a further, IKC-independent NF-κB-activating pathway that cannot be inhibited by parthenolide. Along this line, NF-κB activation induced by overexpression of NIK can only partially be blocked by dominant-negative forms of both IKKs (16), but on the other hand, there is presently no experimental proof for the existence of such an alternative signaling route. It is unlikely that MEKK1 itself is a direct target for parthenolide, as MEKK1-induced JNK activation remained unaffected. We speculate that parthenolide interferes with an IKC component that is necessary for the sequential transmission of signals leading to the activation of both IKKs. Since the IKC is nonfunctional in the absence of NF-κB essential modulator/IKKγ (9), this protein is a potential candidate for the inhibitory activity of parthenolide. Also, the IKK complex-associated protein protein or so far unidentified components of the IKC may be inactivated by this SL. Future studies employing radiolabeled parthenolide will help to identify the target structure(s) inactivated by this IKC inhibitor.

Taken together, this study extends our understanding on the molecular mechanisms underlying the antiphlogistic activity of SL-containing plant extracts that are used in traditional medicine, thereby providing an interesting lead compound for the development of new antiinflammatory compounds. Furthermore, the differential effects of parthenolide on the TNF-α-mediated activation of JNK, p38, and NF-κB provide a molecular tool, which allows the dissection of these different signaling pathways.

We thank Sandra Grunau for excellent technical assistance, Dr. M. Heinrich (London, U.K.) for stimulating discussions, and the following colleagues who generously provided plasmids and reagents, which made this work possible: Dr. D. Goeddel (San Francisco, CA), Dr. D. Wallach (Rehovot, Israel), Dr. T. Maniatis (Cambridge, MA), and Dr. P. Angel (Heidelberg, Germany).