Cysteine 38 in p65/NF-kappaB plays a crucial role in DNA binding inhibition by sesquiterpene lactones.

Sesquiterpene lactones (SLs) have potent anti-inflammatory properties. We have shown previously that they exert this effect in part by inhibiting activation of the transcription factor NF-kappaB, a central regulator of the immune response. We have proposed a molecular mechanism for this inhibition based on computer molecular modeling data. In this model, SLs directly alkylate the p65 subunit of NF-kappaB, thereby inhibiting DNA binding. Nevertheless, an experimental evidence for the proposed mechanism was lacking. Moreover, based on experiments using the SL parthenolide, an alternative mode of action has been proposed by other authors in which SLs inhibit IkappaB-alpha degradation. Here we report the construction of p65/NF-kappaB point mutants that lack the cysteine residues alkylated by SLs in our model. In contrast to wild type p65, DNA-binding of the Cys(38) --> Ser and Cys(38,120) --> Ser mutants is no longer inhibited by SLs. In addition, we provide evidence that parthenolide uses a similar mechanism to other SLs in inhibiting NF-kappaB. Contrary to previous reports, we show that parthenolide, like other SLs, inhibits NF-kappaB most probably by alkylating p65 at Cys(38). Although a slight inhibition of IkappaB degradation was detected for all SLs, the amount of remaining IkappaB was too low to explain the observed NF-kappaB inhibition.

rylates IB on serines 32 and 36. Phosphorylation causes IB ubiquitinylation and its subsequent degradation by the 26-S proteasome. Degradation of the inhibitor allows NF-B to translocate to the nucleus, where it stimulates transcription of its target genes. 2 Genes that are regulated by NF-B include, for example, proinflammatory and inflammatory cytokines such as interleukin-1, -2, -4, and -6 or TNF-␣, as well as genes encoding immunoreceptors, cell adhesion molecules, acute phase proteins, and enzymes such as cyclooxygenase-II. Because of its central role in regulating inflammatory responses, a pharmacological inhibition of NF-B activation could be beneficial in the treatment of inflammation (4). 2 Sesquiterpene lactones (SLs) are the active constituents of many medicinal plants from the Asteraceae family. These herbal remedies are used for the treatment of inflammation in traditional medicine. It has been shown in various assays that plant extracts as well as the purified SLs possess anti-inflammatory properties. We have recently demonstrated that several SLs inhibit activation of NF-B (5)(6)(7)(8). Using helenalin as a model, we have shown that SLs inhibit neither IB degradation nor NF-B nuclear translocation. SLs interact directly with NF-B. DNA binding of NF-B is prevented by selectively alkylating cysteine sulfhydryl groups in its p65 subunit (5,6). There are strong indications that this is a general mechanism for SLs, which possess ␣,␤or ␣,␤,␥-unsaturated carbonyl structures such as ␣-methylene-␥-lactones or ␣,␤-unsubstituted cyclopentenones. These functional groups are known to react with nucleophiles, especially with cysteine sulfhydryl groups, in a Michael-type addition (9 -11). Studies on structure-activity relationships revealed that a strong NF-B-inhibitory activity correlates with the presence of two reactive centers in the SL molecule. Thus, bifunctional SLs inhibit NF-B DNA binding at a concentration 10 times lower than monofunctional molecules (8). Based on the published x-ray structure of the p65 homodimer, we have proposed a mechanism for NF-B inhibition by SLs (8,12). In our model, bifunctional SLs react with two cysteine residues in the p65 molecule (Cys 38 and Cys 120 ), thereby preventing DNA binding of the transcription factor. This model can explain both the selectivity of SLs for p65 and the observed requirement of bifunctionality for an effective inhibition (8).
In NF-B/p65, cysteine L1 (L1-Cys, position 38) participates in DNA binding by forming a hydrogen bond with the sugar/ phosphate backbone of the B-DNA motif. Several amino acids in its immediate vicinity (Arg 33 , Arg 35 , Tyr 36 , Glu 39 , and Arg 187 ) are also known to be essential for DNA recognition and binding. A second cysteine (EЈ-Cys, position 120) in the nearby EЈ region, is found at a distance of only 8 Å from the L1-Cys. We have proposed that this distance is small enough to allow the alkylation of both sulfur atoms by a single bifunctional SL. The alkylation of both amino acids would result in a cross-link between Cys 38 and Cys 120 in the p65 molecule. Between these two amino acids there is a gap that is empty except for the phenol ring of Tyr 36 ; its interaction with the DNA backbone and with two thymine residues is essential for the DNA binding. If p65 were alkylated at both cysteine residues by a SL, the Tyr 36 /DNA interaction would become impossible. The spatial arrangement of Tyr 36 is crucial for the binding activity. Hence, an alteration of the geometry in this region would dramatically affect the capability of p65 to bind to the B-DNA motif. Molecular modeling predicts that upon alkylation of both amino acids, dramatic changes in the positions of the Tyr 36 will occur. Moreover, after reacting with the protein, the SL occupies a space otherwise filled by the DNA sugar/phosphate backbone. Molecular models thus predict that alkylation would alter the p65 structure so as to make DNA binding impossible. The same model suggests that alkylation of p65 by two monofunctional SL molecules prevents its DNA binding (8).
To date, we have not been able to present direct chemical proof for our proposed mechanism. In this paper we use point mutants to demonstrate that cysteine 38 is the key amino acid in SL-mediated NF-B inhibition. Moreover, our studies show that parthenolide acts like helenalin in that it also selectively alkylates NF-B/p65 on cysteine residues. Although we also observed a slight inhibition of IB-␣ degradation with parthenolide as well as the other SLs tested, the amount of remaining IB-␣ was too low to explain the observed NF-B inhibition. These data contradict a previous report suggesting that parthenolide exerts its effect on NF-B solely by inhibiting IB-␣ degradation (13,14).
Our observation that SLs directly inhibit NF-B DNA binding reemphasizes their potential as very interesting lead compounds for the design of new anti-inflammatory drugs.
Plasmids-The mammalian expression vectors containing cDNAs for the NF-B subunits p50 and p65 have been described previously in detail (18 -20).
The mutant p65 fragments were combined using PCR-ligation; a PCR reaction using only the two fragments but no primers was performed. The resulting full-length mutated DNA was amplified using T7 and SP6 primers. All PCR reactions were conducted with recombinant Pfu DNA Polymerase (Stratagene), which has proof-reading activity, to minimize the random generation of unwanted mutations. The mutated p65 cDNA was cloned into pRC/CMV (Invitrogen) and sequenced to confirm the presence of the mutations and to rule out the appearance of unwanted mutations. Altogether, five plasmids were constructed, in which cysteines 38, 120, or both amino acids were replaced by serine (both single mutants and the double mutant), glycine (cysteine 120), or alanine (cysteine 120).
Cell Culture-Jurkat T-cells were maintained in RPMI 1640 medium, 293 cells in Dulbecco's modified Eagle's medium. Both were supplemented with 10% fetal calf serum,100 IU/ml penicillin, and 100 g/ml streptomycin (all from Life Technologies, Inc.). Cycloheximide was purchased from Sigma, and TNF-␣ was a kind gift from Dr. K. Decker, Freiburg, Germany.
Transfections-293 Cells were plated 12-16 h prior to transfection at a density of 2 ϫ 10 5 cells/60-mm dish. Cells were transfected using the SuperFect reagent (Qiagen) as described by the manufacturer.
Western Blotting-Total cell extracts (30 g) were boiled in Laemmli sample buffer and subjected to SDS-polyacrylamide gel electrophoresis. Proteins were transferred at 0.8 mA/cm 2 for 1 h onto Immobilon P membrane (Millipore) using a semi-dry blotting apparatus (Owl). Nonspecific binding sites were blocked by immersing the membrane in blocking solution (TBST: 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Tween 20 (v/v), containing 5% (w/v) nonfat dry milk (Fluka)) overnight at 4°C. After a short wash in TBST, the membrane was incubated in a 1:1000 dilution of a rabbit anti-IB-␣ antibody (Cell Signaling Technology) in TBST for 1 h at room temperature followed by 30 min of washing with TBST. Bound antibody was decorated with goat anti-rabbit horseradish peroxidase conjugate (Amersham Pharmacia Biotech; diluted 1:2000 in TBST) for 1 h at room temperature. After washing for 30 min in TBST, the immunocomplexes were detected using ECL Western blotting reagents (Amersham Pharmacia Biotech). Exposure to Kodak XAR-5 films was performed for 10 s to 2 min.
In Vitro Kinase Assay-To determine IKK activity, total Jurkat Tcell extracts were prepared using 1 ml of immunoprecipitation buffer (50 mM Hepes, pH 7.6, 250 mM NaCl, 10% glycerol, 1 mM EDTA, 0.1% Nonidet P-40) with protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin) and phosphatase inhibitors (1 mM Na 3 VO 4 , 50 mM NaF). 1000 g of protein extract was immunoprecipitated by incubating overnight at 4°C with agarose-conjugated IKK-␣ antibody (Santa Cruz Biotechnology, Inc.) and mixing on a spinning wheel. The immunoprecipitates were washed extensively with phosphate-buffered saline. The kinase assay was performed with full-length IB-␣ (amino acids 1-317, Santa Cruz Biotechnology, Inc.) protein in 20 l of kinase buffer (20 mM Tris-HCl, pH 7.6, 10 mM MgCl 2 , 0.5 mM DTT, 10 mM ATP, and 5 Ci of [␥-32 P]ATP). A portion of the immunoprecipitated IKK complex was incubated in the kinase buffer with 4 g of the substrate for 30 min at room temperature. Samples were analyzed by 10% SDS-polyacrylamide gel electrophoresis and autoradiography. The remaining portion of the immunoprecipitated samples were run on a separate 10% SDS-polyacrylamide gel electrophoresis and were Western-blotted with an IKK-␤ antibody to check for equality of loading. 38 3 Ser and Cys 38,120 3 Ser Mutants Is Inhibited at Higher SL Concentrations Than Wild Type p65-To demonstrate directly that SLs alkylate the p65 NF-B subunit, we constructed expression plasmids that encode p65 proteins in which either one cysteine (Cys 38 3 Ser, Cys 120 3 Ser) or both cysteine residues (Cys 38,120 3 Ser) were mutated to serine. By mutating cysteine to serine, a thiol functional group is replaced by an alcohol, which can no longer react with SLs. These plasmids were used to transiently transfect 293 cells. Cell extracts were prepared and assayed for NF-B DNA binding by EMSA. With the exception of the Cys 120 3 Ser mutant, all mutant p65 proteins bound to DNA To obtain a functional cysteine 120 mutant, a Cys 120 3 Gly and Cys 120 3 Ala mutant were constructed. These amino acids are incapable of reacting with the SLs. Of these mutants, only the Cys 120 3 Ala mutant showed DNA binding activity. Each p65 mutant bound DNA with different affinity, the Cys 38 3 Ser showed the strongest binding followed by the Cys 38,120 3 Ser mutant and the native protein, whereas the Cys 120 3 Ala mutant bound DNA with least affinity.

DNA Binding of NF-B/p65 Cys
Transfected cells expressing either wild type (wt) or mutant proteins were incubated with increasing concentrations of different SLs, and their DNA binding was assessed by an EMSA. The advantage of using transfected cells overexpressing p65 is that no stimulation is needed to observe a DNA binding activity. As the p65 concentration in these cells exceeds that of IB, the p65 homodimer is not inactivated and binds DNA with no need for NF-B activation. Therefore, using p65 transfected cells, the effect of direct interactions between this NF-B subunit and SLs on DNA binding can be studied.
We assessed three bifunctional SLs, 4␤,15-epoxy-miller-9Eenolide, helenalin, and parthenolide, as well as the monofunctional 11␣,13-dihydrohelenalin acetate (for structures see Fig.  1) for their effect on p65 DNA binding. We have previously shown that all four compounds inhibit DNA binding of NF-B at micromolar concentrations, the bifunctional SLs being the strongest inhibitors. For the latter, concentrations between 5 and 20 M are needed for a complete NF-B inhibition (8), whereas 200 M 11␣,13-dihydrohelenalin acetate are required for the same effect. 3 If SLs inhibit NF-B according to the mechanism proposed by us, concentrations that completely inhibit wt p65 should not affect the mutant proteins. Therefore, we first determined those concentrations that completely inhibit the DNA binding of the p65 homodimer ( Figs 2-7). It thus appears that Cys 120 , which was available for alkylation in the Cys 38 3 Ser mutant, is not critical for the effect of SLs on p65 DNA binding. Cys 38 , by contrast, appears critical for the ability of all SLs to inhibit p65 DNA binding, because its mutation rendered p65 insensitive to SL interference. Mutation of Cys 120 to alanine appears to destabilize the p65 structure, because the mutant binds DNA more weakly than wt p65 (compare Fig. 2, C, lane 2, with A, lane 2). It is thus dislodged from the DNA at lower SL concentrations than wt p65 (Figs. 2, A and B, and 3A). This conformational instability may also explain why the Cys 38,120 3 Ser double mutant is more sensitive to SL modification than the Cys 38 3 Ser single mutant at higher SL concentrations (com- pare Fig. 2D with Figs. 2B and 3B). Thus, Cys 120 appears to be targeted by SLs as well, even though Cys 38 is the major site of alkylation.
Further Evidence That the SL Parthenolide Also Directly Inhibits the p65 Subunit of NF-B-Parthenolide shows the same behavior as the bifunctional SLs helenalin and 4␤,15epoxy-miller-9E-enolide with respect to inhibition of transfected p65 proteins (Table I, Fig. 3). No difference can be observed between the inhibition of p65 mutants or the native protein by parthenolide compared with the other SLs studied. These data suggest that this SL interacts with and directly alkylates the p65 subunit of NF-B as do other SLs. However, these data contradict observations by Hehner et al. (13,14), which excluded a direct reaction with NF-B. These authors presented evidence that NF-B inhibition by parthenolide is due solely to the prevention of IB degradation. To obtain further evidence for the direct alkylation of NF-B by parthenolide, Jurkat T-cells were first stimulated with TNF-␣ for 20 min and then incubated with 20 M parthenolide for the indicated times. Subsequently, cell extracts were analyzed for NF-B DNA binding by EMSA. Fig. 4 shows that parthenolide inhibits NF-B DNA binding after the transcription factor has been activated. If parthenolide inhibited the activation of NF-B solely by preventing degradation of IB, it should not interfere with DNA binding once the transcription factor has been activated and IB is degraded. Thus, parthenolide displays the same properties reported for helenalin and germacranolides from Tithonia diversifolia (5,7).
Further proof for the direct alkylation of NF-B was obtained by incubating Jurkat T-cells with parthenolide and then stimulating with TNF-␣ for different time periods. No NF-B DNA binding could be detected by EMSA in these extracts (data not shown). The protein extracts were then treated with DOC, which can release IB from the NF-B⅐IB complex (21) (Fig. 5). If parthenolide inhibits NF-B by preventing IB degradation, de novo NF-B DNA binding should be observed after treatment with DOC. This is observed after quercetin treatment, which does not alkylate NF-B and was hence used as a control. This flavonoid inhibits NF-B activation by inhibiting IKK-␣ and -␤, thereby preventing IB degradation (22,23). Protein extracts treated with an inhibiting concentration of quercetin displayed NF-B DNA bind-   ing after DOC treatment (Fig. 5, lane 6). However, following parthenolide treatment, NF-B DNA binding could not be restored by DOC treatment (Fig. 5, lanes 7-10).
To exclude the possibility that parthenolide targets the p50 subunit of NF-B, we investigated its effect on the p50 homodimer. We transfected 293 cells with an expression vector for the p50 DNA-binding subunit of NF-B. Overexpressed p50 is not subjected to IB inactivation and will therefore constitutively bind DNA as a homodimer (24). Transfected 293 cells expressing p50 NF-B were treated with increasing concentrations of parthenolide for 2 h after which cells were analyzed by EMSA (Fig. 6). Concentrations of up to 40 M parthenolide did not impair NF-B p50 DNA binding activity (Fig. 6, lane 7). These data demonstrate that parthenolide has no effect on DNA binding of the NF-B p50 subunit. The same results were obtained with 4␤,15-epoxy-miller-9E-enolide (data not shown) and helenalin (6), suggesting that SLs in general do not target the p50 subunit.
Taken together, these three experiments clearly confirm that parthenolide directly alkylates NF-B. Furthermore, a lack of NF-B DNA binding following DOC treatment suggests that alkylation rather than any interference with its activation is the major mechanism of NF-B inhibition by SLs.
Sesquiterpene Lactones Partially Inhibit IB Degradation-Although data presented so far demonstrate direct alkylation of p65 by parthenolide, these experiments did not investigate whether this SL also inhibits IB degradation. Therefore, we repeated the experiments published by Hehner et al. (14). The protein extracts shown in Fig. 7A were analyzed in a Western blot using a polyclonal antibody against IB-␣. Fig. 7B (lanes 3,  5, 7, and 9) shows that TNF-␣ treatment for various times resulted in decreased amounts of detectable IB-␣ compared with unstimulated cells. However, this decrease occurred more slowly in the presence of parthenolide than in its absence (compare lane 2 and 3, Fig. 7B). The same experiment was repeated with helenalin and 4␤,15-epoxy-miller-9E-enolide to exclude the possibility that this effect is a specific property for parthenolide. In all cases a decrease in the amount of IB-␣ could be observed (data not shown). We have previously reported a similar retardation of IB-␣ degradation in the presence of helenalin (6).
De novo IB-␣ synthesis was observed 30 min after TNF-␣ stimulation (Fig. 7B, lane 4) (25). After 60 min the IB-␣ concentration was as high as in untreated cells (Fig. 7B, lane  6). It is therefore difficult to decide whether the IB-␣ band observed in parthenolide-treated cells (lanes 5 and 7) was due to an inhibition of its degradation or to de novo synthesis of IB-␣.
For this reason, we repeated the experiment shown in Fig. 7 but in addition incubated the cells with cycloheximide to inhibit IB-␣ resynthesis (Fig. 8) (26 -28). NF-B DNA binding was analyzed in an EMSA. Parthenolide inhibited NF-B DNA binding in the presence of cycloheximide as effectively as in untreated cells (Fig. 8A). The extracts were subsequently studied by Western blot. Fig. 8B shows that resynthesis of IB-␣ was suppressed by cycloheximide treatment (Fig. 8B, lane 8). Nonetheless, in extracts pretreated with parthenolide, IB-␣ could still be detected in small amounts (Fig. 8B, lanes 7 and 9). Thus degradation of IB-␣ is partially inhibited by parthenolide. To quantify this effect, we subjected the Western blots to densitometry. For normalization, the membrane was stripped and redecorated with an antibody against actin. The results depicted graphically in Fig. 8C clearly show that the amount of IB-␣ decreases over time, reaching a minimum after 120 min in untreated cells. IB-␣ degradation is delayed in cells exposed to parthenolide. Interestingly, although in control cells the amount of IB-␣ decreases between 60 and 120 min after treatment, in parthenolide treated cells it remains almost constant. However, of the total amount of IB-␣ present in untreated cells, only 5.6% remains intact after 120 min of TNF-␣ stimulation in the presence of parthenolide. Similar results were obtained in the presence of helenalin and 4␤,15-epoxy-miller-9E-enolide (12.8 and 6.4%, respectively). These levels of remaining IB-␣ are too low to explain the complete inhibition of NF-B DNA binding observed in the presence of SLs.
Parthenolide Does Not Effectively Inhibit the IKC-IB-␣ phosphorylation requires the activity of the IB kinase complex (IKC). Thus, one possible mechanism to explain the slight inhibitory effect of parthenolide on IB-␣ degradation is that this SL inhibits IKC activity. To test this possibility, an in vitro kinase assay was carried out. Jurkat T-cells with or without a 60-min pretreatment with different concentrations of parthenolide, were stimulated with TNF-␣ for 10 min. Total cell extracts were prepared. Subsequently, an aliquot of the protein extract was used to determine the DNA binding activity of NF-B (Fig. 9A). The IKC was immunoprecipitated using agarose-conjugated anti-IKK-␣, and the kinase activity was determined using the IB-␣ protein as substrate. The kinase activity was evaluated by incorporation of 32 P into the IB-␣ substrate. Stimulation of cells with TNF-␣ alone resulted in an increase of IB-␣ phosphorylation (Fig. 9B, lane 2). Pretreatment of the cells with parthenolide inhibited the NF-B DNA binding, but the formation of 32 P-labeled IB-␣ was only slightly inhibited at a 40 M concentration (Fig. 9B, lane 6). The use of equal amounts of IKC in each sample was confirmed by performing a Western blot against IKK-␤. These results show that the IKC is not inhibited by parthenolide at concentrations that completely inhibit NF-B DNA binding activity. Thus, NF-B inhibition by parthenolide is not because of inhibition of the IKK complex. However, the slight inhibition of IB-␣ degradation by parthenolide could be explained by the partial inhibition of IKK activity.

DISCUSSION
SLs constitute the active compounds of medicinal plants such as Arnica montana, A. chamissonis, Tanacetum parthenium, Tanacetum vulgare, or M. quinqueflora (16, 29 -32). Preparations from these plants are popular herbal remedies used for the treatment of inflammations. Several studies have shown how SLs exert an anti-inflammatory effect. They interfere with cellular processes including oxidative phosphorylation, platelet aggregation, and histamine and serotonin release, as well as neutrophil chemotaxis (33)(34)(35). In previous studies we have shown that SLs selectively inhibit DNA bind- ing of the transcription factor NF-B (5-8, 16, 36). We have demonstrated that NF-B inactivation is not caused by inhibiting IB degradation or by preventing its translocation to the nucleus (5-7). However, the molecular mechanism by which SLs cause NF-B inhibition was not clear. Here we use point mutants to show that SLs modify the active NF-B heterodimer and that cysteine 38 in the p65 subunit plays a crucial role.
The Cys 38 3 Ser mutant, in which only cysteine 120 can be alkylated by SLs, was unaffected by a concentration two to three times higher than that required to completely prevent DNA binding of the wt p65 homodimer. Only 20 M helenalin caused a slight inhibition of this mutant. These results are in agreement with our theory that inhibition of DNA binding is due to alkylation of both cysteine residues. However, the results with the Cys 120 3 Ala mutant do not fit our hypothesis and appear to contradict the proposed theory. This mutant was inhibited by all SLs studied at lower concentrations than the wild type protein. In the Cys 120 3 Ala mutant, only cysteine 38 can be alkylated. Our results would suggest that an alkylation at position 38 of p65 would be sufficient to prevent the interaction of NF-B with the DNA. However, several parameters must be considered when interpreting these results. Of the two NF-B subunits, p65 is inhibited whereas p50 is not. Although p65 contains the two cysteines mentioned at positions 38 and 120, p50 has only one equivalent cysteine, cysteine 62. In p50, the amino acid equivalent to cysteine 120 is substituted by histidine, which cannot be alkylated by SLs at physiological conditions (10). Because of the similar three-dimensional structure of both subunits, cysteine 62 in p50 has the same probability of becoming alkylated as cysteine 38 in p65. Therefore, a single alkylation of cysteine 38 in p65 cannot be solely responsible for inhibition by SLs, because if this was the case, p50 DNA binding should also be inhibited at similar SL concentrations. We have previously shown that the binding of the p50 homodimer is not influenced by treatment with helenalin (6). This was also shown here for parthenolide and 4␤,15-epoxymiller-9E-enolide. Nonetheless, the results with the Cys 120 3 Ala mutant suggest that cysteine 120 might be very important for the binding of p65 to DNA. However, alternative interpretations are also possible. This mutation could either lower the affinity of the protein for its target sequence or change its conformation, thereby destabilizing protein folding. Moreover, the Cys 120 3 Ala mutant may be expressed at a lower level than wt p65 in the transfected cells. These possibilities preclude a satisfactory interpretation of the results obtained from the Cys 120 3 Ala mutant. The differences in DNA binding strength of the p65 mutant proteins has previously been shown by Kumar et al. (37). These authors mutated the equivalent positions (Cys 35 3 Ser and Cys 117 3 Ser) on the avian virus protein v-Rel. The resulting mutants bound DNA following the same pattern observed by us. The Cys 38,120 3 Ser double mutant binds DNA more strongly than the wt p65 protein. If the mutation of cysteine 120 to serine destabilized protein DNA interactions, this effect should also be observed in the double mutant. Therefore, cysteine 120 may not play a direct role in DNA binding or recognition. It is more likely that the presence of Cys 120 is crucial for optimal folding, probably because it stabilizes DNA structure.
Even though the double mutant no longer contains reactive cysteine residues, its DNA binding is nonetheless inhibited by SLs. However, much higher concentrations are required than for the inhibition of the native protein. The fact that an inhibitory effect can be observed indicates that SLs can react with amino acids other than cysteines in the p65 molecule. This effect should be of lower significance, as it is observed only at higher concentrations.
The data presented here experimentally confirm our theory that the inhibition of NF-B DNA binding by SLs is due to alkylation of the p65 subunit and that Cys 38 is necessary for this inhibition. The role of Cys 120 is not yet unambiguously clarified. The possibility remains that inhibition of NF-B is caused by successive alkylation of Cys 38 and Cys 120 . The mechanism is valid for bifunctional SLs, as shown with the pseudoguaianolide helenalin and the germacranolides parthenolide and 4-␤,15-epoxy-miller-9E-enolide, as well as for monofunctional SLs such as 11␣,13-dihydrohelenalin acetate (Figs. 2 and 3).
However, the mechanism by which the SL parthenolide inhibits NF-B activation has remained a matter of debate (5,13,14). The data presented here (Figs. 3-6) support our hypothesis that parthenolide directly alkylates the p65 subunit of NF-B. Parthenolide thus achieves its effect by the same mechanism as all other SLs tested. Additionally, we repeated the experiment published by Hehner et al. (13) in which the authors detected an inhibition of IB degradation. We indeed detected a slight inhibition of IB-␣ degradation. However, after normalization of the protein loading using an antibody for actin and quantitation, the amount of undegraded IB-␣ remaining comprised 5.6% of the IB-␣ present in untreated cells. Similar results were also obtained with the other SLs analyzed.
The amount of IB-␣ remaining is insufficient to explain an inhibition of NF-B DNA binding. Our experiments using DOC demonstrate that alkylation is the most crucial reaction for the inhibition of NF-B DNA binding. Therefore, some previous reports in the literature, in which NF-B inhibition by parthenolide has been repeatedly reported to be caused by IB degradation, should be reexamined. (38,39). In both cases, IB-␣ resynthesis and the delay in its degradation were not taken into account. A quantitation of the undegraded IB is also missing in these studies.
We also determined the effect of parthenolide on the activity of the IB kinase complex which is composed of at least three subunits: two catalytic subunits, IKK-␣ and IKK-␤, and a regulatory protein, IKK-␥. 2 We carried out an immunoprecipitation of the IKC with an IKK-␣ antibody, which was confirmed by Western blotting with an antibody against IKK-␤ (Fig. 9C). In the subsequent in vitro kinase assay, only a slight inhibition was observed at the highest concentration (40 M) of parthenolide (Fig. 9B), whereas a concentration of 20 M was sufficient to inhibit NF-B DNA binding (Fig. 9A, lane 5). The low amount of IB-␣ remaining after parthenolide treatment could be explained by this partial inhibition of the IKC. Interestingly, both IKK-␣ and -␤ contain a cysteine residue in the activation loop of their catalytic domain. SLs could alkylate this residue, thereby inhibiting enzyme function. A similar mechanism has been reported for arsenite and cyclopentenone prostaglandinmediated NF-B inhibition (40,41).
In this study, we provide molecular evidence that SLs alkylate Cys 38 of p65/NF-B. These anti-inflammatory agents thus act directly on the transcription factor NF-B by alkylating cysteine residues in its p65 subunit. To a minor degree, inhibition of IB-␣ degradation also takes place. The fact that inhibition of NF-B activation by SLs occurs at the last step of the transduction pathway, by inhibition of its DNA binding, makes these natural products very interesting NF-B inhibitors. It was recently shown that NF-B activity can also be stimulated without IB-␣ degradation (42). Unlike NF-B inhibitors that target the IB kinase complex, SLs would also inhibit the transcription factor activated by this mechanism (43). By inhibiting NF-B activity, SLs interfere with various aspects of the inflammatory reaction, such as the production of proinflammatory cytokines and the biosynthesis of inducible nitric-oxide synthase (3,38,44,46). Therefore, SLs are very intriguing lead compounds for the development of anti-inflammatory drugs.