A Small Molecule Ubiquitination Inhibitor Blocks NF-κB-dependent Cytokine Expression in Cells and Rats*

A small molecule inhibitor of NF-κB-dependent cytokine expression was discovered that blocked tumor necrosis factor (TNF) α-induced IκBα degradation in MM6 cells but not the degradation of β-catenin in Jurkat cells. Ro106-9920 blocked lipopolysaccharide (LPS)-dependent expression of TNFα, interleukin-1β, and interleukin-6 in fresh human peripheral blood mononuclear cells with IC50 values below 1 μm. Ro106-9920 also blocked TNFα production in a dose-dependent manner following oral administration in two acute models of inflammation (air pouch and LPS challenge). Ro106-9920 was observed to inhibit an ubiquitination activity that does not require βTRCP but associates with IκBα and will ubiquitinate IκBα S32E,S36E (IκBαee) specifically at lysine 21 or 22. Ro106-9920 was identified in a cell-free system as a time-dependent inhibitor of IκBαee ubiquitination with an IC50 value of 2.3 ± 0.09 μm. The ubiquitin E3 ligase activity is inhibited by cysteine-alkylating reagents, supported by E2UBCH7, and requires cIAP2 or a cIAP2-associated protein for activity. These activities are inconsistent with what has been reported for SCFβTRCP, the putative E3 for IκBα ubiquitination. Ro106-9920 was observed to be selective for IκBαee ubiquitination over the ubiquitin-activating enzyme (E1), E2UBCH7, nonspecific ubiquitination of cellular proteins, and 97 other molecular targets. We propose that Ro106-9920 selectively inhibits an uncharacterized but essential ubiquitination activity associated with LPS- and TNFα-induced IκBα degradation and NF-κB activation.

tors has been demonstrated with anti-TNF␣ therapies. Etanercept and Infliximab are efficacious for both rheumatoid arthritis and Crohn's disease (2)(3)(4). Transrepression of NF-B activation by glucocorticoids bound to the glucocorticoid receptor is proposed to contribute to the anti-inflammatory properties of glucocorticoids in asthma and arthritis (5,6). Strategies that directly block NF-B action have also shown efficacy in preclinical studies. Overexpression of IB␣ inhibits both inflammatory and destructive mechanisms in rheumatoid synovium but spares anti-inflammatory mediators (7). NF-B antisense has been effective in animal models of rheumatoid arthritis (8) and Crohn's disease (9). NF-B has also been shown to be important for osteoclast differentiation, leading to the suggestion that inhibitors of NF-B may hold therapeutic potential in the treatment of osteoporosis and other bone diseases (10).
Transcriptional activation by NF-B requires translocation of NF-B from the cytoplasm to the nucleus (1,(11)(12)(13). The translocation into the nucleus is controlled by IB␣. IB␣ binds to NF-B in the cytoplasm and masks the NF-B nuclear localization signal. NF-B activation is initiated by many extracellular molecules, including LPS, TNF␣, and IL-1␤, through binding to their respective cell surface receptors. The binding initiates a signaling cascade that leads to the activation of the IB kinase complex containing IKK␣ and IKK␤. Activation of the IKK complex requires phosphorylation of two serine residues located in the "activation loop" within the kinase domain of IKK␣ or IKK␤. Certain mitogen-activated protein 3-kinases (MEKK1, MEKK2, MEKK3, and NF-B inducing kinase) are capable of phosphorylating these serines in vitro and activating NF-B in transfection experiments. The activated IKK complex phosphorylates IB␣ at serines 32 and 36. The phosphorylated IB␣ is subsequently polyubiquitinated at lysine 21 or 22 and degraded by the proteasome. NF-B is then released to translocate to the nucleus and promote gene expression.
The goal of our work was to block NF-B-dependent gene expression via stabilization of the endogenous NF-B⅐IB␣ complex. Stabilization has been achieved in cells with mutations that prevent the phosphorylation at serines 32 and 36 (14 -16), in cells with mutations that prevent the ubiquitination at lysines 21 or 22 (17,18), or by inhibition of IB␣ degradation by the proteasome (19,20). Our approach was to target the ubiquitination of IB␣. Ubiquitination in general requires three enzymes: the ubiquitin-activating enzyme (E1), one of the multiple E2 ubiquitin-conjugating enzymes, and a ubiquitin E3 ligase (21)(22)(23)(24). The ubiquitination of IB␣ has been associated with a number of E2s including UBCH5b, UBCH5c, and CDC34/UBC3 (24) and an SCF E3 ligase complex in which IB␣ binds to the F-box subunit known as ␤TRCP (25)(26)(27).
The work described in this report was initiated prior to the identification of IKK as the IB␣ kinase or SCF ␤TRCP as the IB␣ ubiquitin ligase. Our approach utilized a phosphomimetic substrate of IB␣ in which serines 32 and 36 were mutated to glutamic acid (IB␣ee). We measured IB␣ee ubiquitination activity to identify an active fraction in Jurkat cell lysates. The reconstituted fraction catalyzed IB␣ee ubiquitination with the appropriate substrate selectivity and was used to identify inhibitors of IB␣ee ubiquitination. The screening paradigm identified a number of inhibitors including Ro106-9920 that demonstrated the functional activity expected for an inhibitor of IB␣ ubiquitination. With the subsequent discovery of the SCF ␤TRCP E3 ligase, we were able to compare its reported properties with those associated with IB␣ee ubiquitination. The properties of the IB␣ee E3 ubiquitin ligase in our fractions were not consistent with those of SCF ␤TRCP . We conclude that Ro106-9920 inhibits an uncharacterized E3 ligase activity essential for TNF␣-and LPS-induced IB␣ degradation and NF-B activation.

MATERIALS AND METHODS
Mutagenesis, Cloning, and Expression-The IB␣ coding sequence was cloned into the BamHI site of both the pET11 vector and the pET30 vector with an N-terminal His 6 tag (NOVAGEN, Madison, WI). The pET11 IB␣ was mutagenized following the USE mutagenesis method (Amersham Biosciences). The mutagenesis of serine 32 and 36 into glutamic acid (IB␣ee) was obtained with the mutagenic primer 5Ј-C CTC GTC TTT CAT ctc GTC CAG GCC ttc GTC GTG GCG GTC GTC CAG TAG-3Ј in the reverse, noncoding orientation. The mutated plasmid DNA was used as template in the PCR reaction to introduce flanking restriction sites. The resulting amplicon was digested with BspHI (New England Biolabs, Beverly, MA) and ligated to the NcoI sites of vector pET30a (Novagen). The subsequent mutagenesis of lysine 21 and 22 to arginine (IB␣eerr) was accomplished similarly with the mutagenic primer 5Ј-G TAG CCG CTC CcT CcT CAG CCC GTC cCG GGG GCC CTC C-3Ј also in the reverse, noncoding orientation. UBCH5a and UBCH7 were PCR-amplified from a human leukemia Jurkat cell 5Ј stretch cDNA library (CLONTECH, Palo Alto, CA) with the respective forward and reverse primers for UBCH5a (5Ј-CTA TGG ATC CCC ATG GCG CTG AAG AGG ATT C-3Ј and 5Ј-CGC GGA TCC GGT TAC ATT GCA TAT TTC TG-3Ј) and UBCH7 (5Ј-CTA TGG ATC CCC ATG GCG GCC AGC AGG AGG CTG-3Ј and 5Ј-AAG CGA CCT GTG GAC TAA CCG GAT CCG GCA-3Ј). Both E2s were cloned into the NcoI site of pET30 vector. UBCH7 was also cloned into a NcoI site of the pET11. The resulting plasmids were transformed into BL21 (DE3) competent Escherichia coli cells (Novagen) for protein expression.
Protein Purification-E. coli expressing His 6 -tagged proteins were lysed with a Gaulin homogenizer (APV Gaulin, Everett, MA) in 20 mM Tris, pH 6.8, 50 mM NaCl, 10% glycerol with 1 g/ml pepstatin, 1 g/ml leupeptin, 0.5 mg/ml Pefabloc (Centerchem, Stamford, CT), and supernatant was filtered through a 0.2-m filter. Nickel-nitrilotriacetic acid beads (Qiagen, Santa Clarita, CA) (200 ml) were washed three times with lysis buffer minus glycerol prior to adding supernatant. After adding the supernatant to the beads, they were rocked at room temperature for 3 h, washed twice with 5 ϫ 20 mM Tris pH 6.8, 50 mM NaCl, and centrifuged at 700 ϫ g for 7 min. The beads were then washed with 0.1 M imidazole at room temperature for 2 h to remove contaminants. The two washing steps were repeated three times. Purified IB␣ee was eluted with lysis buffer and 0.5 M imidazole at 4°C overnight and dialyzed against 20 mM Tris HCl, pH 6.8, 10% glycerol at 4°C. The yield for IB␣ee was 20 -25 mg/liter of culture.
The E3 was fractionated from Jurkat cells activated with 20 ng/ml TNF␣ for 5 min. The cell pellet was collected, rinsed with cold phosphate-buffered saline, and resuspended in 2 ϫ 20 mM Hepes, pH 7.5, 20 mM KCl, 2 mM DTT, 1.2% IGEPAL CA-630, 5% glycerol, and protease inhibitors (Roche Molecular Biochemicals). The cells were incubated on ice for 30 min and then broken by sonication. The S-100 supernatant was prepared, filtered, and bound in batch to DE52 resin (Whatman), and the protein was eluted with successive washes of 100, 250, and 500 mM KCl in 50 mM Hepes, pH 7.5, containing 1 mM DTT and 5% glycerol. The fractions were collected and measured for IB␣ee ubiquitination activity. The active fractions were pooled, dialyzed, and loaded onto a 21.5-mm ϫ 15-cm, 13-m TSK gel DEAE-5PW anion exchange column (TosoHaas) equilibrated in 50 mM Tris, pH 7.5, 5% glycerol, 0.05 mM DTT. The protein was eluted with a 375-ml linear KCl gradient (0 -0.5 M) at a flow rate of 5 ml/min. The fractions (7.5 ml) were collected and assayed for activity. Western transfer and immunoblotting to identify fractions containing E1 was performed with E1A/E1B C-terminal IgG (Calbiochem). The active fractions, excluding any fractions containing E1 enzyme, were pooled. The pooled fractions were precipitated with 40% ammonium sulfate. The pellet was resuspended and dialyzed against 10 mM Hepes, pH 7.5, 10% glycerol, 0.05 mM DTT and stored at Ϫ80°C. Neither E1 nor E2 was detected in this fraction as determined by thioester assays, and Western blots showed the absence of endogenous IB␣ (data not shown). Further attempts to purify the pooled material resulted in a complete loss of activity.
E3 Immunodepletion-The E3 fractions (20 g) were incubated at 4°C with purified antibody (or rabbit serum) (2 g) and 30 l of prewashed protein G-Sepharose (Amersham Biosciences). After 24 h, the reactions were spun down at 1,500 rpm in a Sorvall H6000. The cleared E3 lysate was removed. The Sepharose beads were washed with 50 l of phosphate-buffered saline and spun again. Fifty l of wash supernatant were removed and pooled with the corresponding cleared E3 lysate for use as E3 ligase in IB␣ee ubiquitination assays.
Nickel Bead Assay-I〉␣ee ubiquitination was assayed in 10 mM Hepes buffer, pH 7.5, containing 1 mM ATP, 5 mM MgCl 2 , 0.125 Ci of 5 M 125 I-ubiquitin (Amersham Biosciences), 0.2-1.0 g of E1, 0.5-3.0 g of E2 UBCH7, 0.5-2.0 g of E3, and 2 M IB␣ee in 50 l of total volume. Typically E1 and E2 were precharged with ubiquitin, ATP, and MgCl 2 for 20 min at 37°C. The E3 fraction and IB␣ee were added to the precharged mix to initiate the reaction. The reaction was stopped after 1 h by adding 50 l of 25 mM DTT to cleave thioester bonds. After 10 min the reaction was diluted with 1 ml of binding buffer and a mixture of urea and iodoacetamide to a final concentration of 5.3 M and 50 mM, respectively. Suspension of nickel-nitrilotriacetic acid beads (100 l) in 5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9, was added to the sample and shaken for 1 h at room temperature. The supernatant was removed after centrifugation, and the beads were washed (four times) with 1 ml of 1% Nonidet P-40 in 60 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9. The product was eluted with 200 l of 1 M imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9, and the amount of 125 I-ubiquitin-IB␣ee was determined in ␥ counter.
E1/E2 Counterscreen-The reactions were preincubated for 15 min at room temperature in 10 mM Hepes, pH 7.5, containing 0.8 g of E1, 5 g of His 6 -tagged E2 UBCH7, 2 mM ATP, 10 mM MgCl 2 , 1 mM DTT, 0.5 unit of pyrophosphatase, and test compound in a final volume of 50 l. The reactions were started with 0.125 Ci of 125 I-ubiquitin and 4 M ubiquitin and stopped after 20 min with 7.5 M urea (50 l). The 125 Iubiquitin-(E2-His 6 )-thioesters were bound to nickel beads, washed, and counted as above.
Nonspecific Ubiquitination Assay-Nonspecific ubiquitination was measured as the formation of ubiquitin conjugates from the S-100 fraction of Jurkat cell lysate. Jurkat S100 lysate (5 g) was combined with 0.8 g of E1, 5 g of E2 UBCH7, 2 mM ATP, 10 mM MgCl 2 , 1 mM DTT, and 0.5 unit of pyrophosphatase in a total volume of 50 l of 10 mM Hepes buffer, pH 7.5, and preincubated for 30 min at 37°C. The assay was then started by the addition of 4 M, 0.125 Ci of 125 I-ubiquitin. Following 1 h of incubation at 37°C, the reactions were stopped with 50 l of Laemmli SDS sample buffer, and the samples were electrophoresed on 10% Tris-Glycine gels (Novex) and visualized with autoradiography.
Cytokine Production in Human Peripheral Blood Mononuclear Cells (PBMNs)-Mononuclear cells were isolated from the fresh venous blood of human volunteers. The cells were washed three times with phosphate-buffered saline and resuspended at 1-1.5 ϫ 10 6 cells/ml in RPMI 1640 medium containing 100 units/ml penicillin G and 100 g/ml streptomycin sulfate. The cells were plated at 1 ml/well in 12-well plates and allowed to adhere for 90 min at 37°C. The wells were then washed twice to remove nonadherent cells, before the addition of 1.5 ml of medium (RPMI 1640 containing 100 units/ml penicillin G and 100 g/ml streptomycin sulfate and 10% fetal calf serum) with or without various concentrations of the test compounds or solvent control (Me 2 SO). After 30 min, LPS was added to a final concentration of 1 ng/ml, and the cultures were incubated for an additional 3 h. The medium was then collected and frozen at Ϫ70°C. Cytokine levels in the medium were subsequently measured using ELISA kits from PharMingen (San Diego, CA) (human TNF␣ and IL-6) and R & D Systems (Minneapolis, MN) (human IL-1␤ and IL-1ra). Compound cytotoxicity was assessed in parallel cultures, plated in 96-well dishes (100 l/well), by the addition of 100 l of a solution of 4 M propidium iodide (Molecular Probes) in phosphate-buffered saline to the cells at the end of the assay. The culture plates were read in a fluorescence plate reader; increases in the signal correlated with propidium-DNA interactions, an indication that the cytoplasmic membrane was no longer intact.
IB␣ Degradation Assay-MM6 cells (1.5 ϫ 10 6 ) were preincubated for 30 min with the inhibitor at 37°C. The reaction was started by the addition of 10 g/ml cycloheximide in Me 2 SO and 20 ng/ml TNF␣ (Promega). Following a 1-h incubation the cells were separated from the medium, resuspended in RIPA buffer (Roche Molecular Biochemicals), and broken by freezing and thawing three times at Ϫ80°C. After centrifugation the supernatants were electrophoresed on 10% Trisglycine gels, and the proteins were electrotransferred onto nitrocellulose paper. The Western blots were probed with a mouse antibody to IB␣ (C-21, Santa Cruz Biotechnology) and detected with rabbit antimouse IgG horseradish peroxidase conjugate.
Animal Studies-Hanover Wistar rats (Charles River Laboratories, Hollister, CA) were acclimated for 1 week prior to the initiation of experiments, at which time they weighed ϳ165-205 g (males) or 90 -120 g (females). Ro106-9920 was suspended in vehicle containing 0.9% NaCl, 0.5% sodium carboxylmethylcellulose (type 7L; Aqualon, Wilmington, DE), 0.4% polysorbate 80, and 0.9% benzyl alcohol. Carrageenan air pouches were created by the subcutaneous injection of air as described (29), with slight modifications. Briefly, male rats were anesthetized with a 60:40 mixture of CO 2 :O 2 , and 20 ml of sterile air was injected into the intracapsular area of the back on day Ϫ6. On day Ϫ3, the pouches were reinjected with 10 ml of sterile air to maintain the integrity of the pouch without inducing further tissue damage. Inflammation was induced on day 0 with 5 ml of 0.5% carrageenan (Type IV Lambda; Sigma) reconstituted and diluted in sterile, pyrogen-free 0.9% NaCl administered directly into the pouch. Vehicle or Ro106-9920 was orally administered at the time of the induction of inflammation. The rats were euthanized 3 h after induction of inflammation. Individual pouches were lavaged with 10 ml of solution containing 5.4 mM EDTA and 10 g/ml indomethacin in 0.9% sterile saline, 0.1% Me 2 SO. The cells in the exudate were pelleted by centrifugation. The supernatants were frozen at Ϫ80°C until analysis with ELISA kits from BIO-SOURCE International (Camarillo, CA) for rat TNF␣ or PerSeptive Biosystems, Inc. (Framingham, MA) for prostaglandin E 2 .
Female rats were challenged intraperitoneally with 50 g/kg E. coli LPS (L3129, Sigma) to induce systemic inflammation and euthanized 1.5 h later. Vehicle or Ro106-9920 was orally administered 30 min prior to LPS challenge. After euthanasia, the blood was harvested via cardiocentesis, and the serum was isolated using serum separator tubes. The serum was stored frozen at Ϫ80°C until ELISA analysis.
Miscellaneous-E1 was purified from rabbit liver according to the method of Haas and Bright (30). The phosphorylated peptide was synthesized by reported procedures (31,32).

Development and Characterization of Discovery Tools
E3 Fractionation-The first step in efforts to discover inhibitors of IB␣ ubiquitination was to develop a screening assay suitable to identify leads. IB␣ ubiquitination requires phosphorylation of serines 32 and 36 and ubiquitination at lysine 21 or 22 (17,18,33). To circumvent the need for phosphorylation, we engineered a protein in which serines 32 and 36 were mutated to glutamic acid. This protein was expressed with an N-terminal His 6 tag, purified from E. coli and used in the presence of ATP, ubiquitin, rabbit E1, and human E2UBCH7 to screen protein fractions for an activity that would catalyze the formation of IB␣ee-ubiquitin conjugates. Jurkat cells were treated with TNF␣ for 5 min to degrade endogenous IB␣ and lysed, and the S100 supernatant was fractionated over DE52 and DEAE. The activity eluted from the DEAE column in a sharp peak at ϳ0.22 M KCl (Fig. 1), immediately following E1. Western blots were run prior to pooling to ensure that E1 did not contaminate the fractions containing the E3 activity. The fractions from each peak were pooled and precipitated with 40% ammonium sulfate. Thioester assays showed that there were no detectable E1 or E2s in the final E3 fraction. Western blots showed the absence of endogenous IB␣ (data not shown). Further attempts to purify the pooled activity resulted in a complete loss of activity.
Substrate Specificity-Validation of an IB␣ ubiquitination enzyme assay requires a dependence upon phosphorylation of serines 32 and 36, ubiquitination at lysine 21 or 22, and the formation of polyubiquitin chains (17,18,33). To this end we engineered, expressed, and purified His 6 -tagged wild type IB␣ and IB␣ S32E,S36E,K21R,K22R (IB␣eerr). Neither wtIB␣ nor IB␣eerr were as efficient ubiquitination substrates for this system as IB␣ee (Table I). However, both wtIB␣ and IB␣eerr were inhibitors of IB␣ee ubiquitination with IC 50 values of 3.0 and 0.6 M, respectively.
E3 Characterization-Subsequent to the completion of this work the identity of the IB␣ E3 was discovered. The E3 was identified as a SCF type E3 with ␤TRCP as the F-box component (25)(26)(27). ␤TRCP has also been shown to be the F-box E3 component required for ␤-catenin ubiquitination (34,35). ␤TRCP was detected in the batch DE52 flow-through and the 100 mM KCl wash using a custom C-terminal peptide antibody and a N-terminal peptide antibody purchased from Santa Cruz Biotechnology. It was not detected in the 250 mM KCl wash that contained the IB␣ee ubiquitination activity. The fractions that supported IB␣ee activity were analyzed by Western blot for the components of the E3 SCF ␤TRCP complex. Cul1 was observed in the fractions with IB␣ee ubiquitination activity (Fig. 1). However, ␤TRCP was not detected with either of the two antibodies. Cul1 showed two bands consistent with reports of neddylation (36). From these data we conclude that SCF ␤TRCP is not the E3 catalyzing the ubiquitination of IB␣ee.
Efforts to characterize the source of the E3 activity by Western blots showed cIAP2 to elute in the active fractions (Fig. 1). NEMO (IKK␥) was also detected in pooled E3 fraction. Cdc34, sentrin, and nedd8 were not detected. His-tagged IB␣ee pull down experiments also showed the presence of NEMO and IKK␤ (data not shown). Immunodepletion of the E3 fraction with anti-cIAP2 decreased the activity to 31 Ϯ 4.9% (n ϭ 6) of fractions immunodepleted with rabbit serum. No decrease in activity was observed with immunodepletion using antibodies to TRAF6, TRAF2, IKK␤, IKK␣, Skp1, NEMO, and MDM2. However, the addition of recombinant N-His 6 -tagged, purified cIAP2 to the reconstituted system in the presence or absence of the E3 fraction did not result in an increase in IB␣ee ubiquitination activity (data not shown). One caveat to the observation is that we did not verify that the expressed and purified cIAP2 was functionally active. In summary, we found that the IB␣ee ubiquitination activity co-purified with Cul1, Skp1, NEMO, and cIAP2 but not with ␤TRCP and required the presence of cIAP2 or a protein associated with cIAP2.
Reaction Properties-Reconstitution of IB␣ee ubiquitination was absolutely dependent on the partially purified E3, purified recombinant human E2 UBCH7, purified rabbit E1, ubiquitin, ATP, and IB␣ee. The reaction was linear with time for over 60 min when initiated with E1 and E2 precharged with ubiquitin (Fig. 2). A lag in the initial velocity (ϳ5 min) was observed when the complete reaction was initiated with ATP. The reaction was saturable with ubiquitin, E1, E2, and IB␣ee. The apparent binding constants for ubiquitin, E1, E2 UBCH7, and IB␣ee, as determined by the saturation binding effect on IB␣ee ubiquitination, were 1. respectively, and suggests that IB␣ee, wtIB␣, and IB␣eerr all bind to the E3 with similar affinity. Therefore, the phosphomimetic glutamic acids at positions 32 and 36 in IB␣ee must increase the activity through an effect on k cat . From these results we conclude that the IB␣ association with the E3 is independent of the phosphorylation state of serines 32 and 36.
The rates of IB␣ee ubiquitination were quantitated by the amount of labeled ubiquitin associated with IB␣ee bound to nickel beads (Table I). The nickel bead bound IB␣ could also be separated by SDS gel electrophoresis and visualized by autoradiography. Fig. 3 shows the formation of high molecular mass IB␣ee-ubiquitin conjugates and the inhibition by a 14mer phosphopeptide containing the IB␣ ubiquitination recognition sequence. The high molecular mass bands Ͼ250 kDa and the smear between 30 and 250 kDa were excised from the gel and counted with scintillation spectroscopy. The percentage of inhibition by the peptide was similar to that obtained with the nickel bead assay. Fig. 3 is shown to emphasize 1) that the system catalyzes the polyubiquitination of IB␣ee, 2) that it is blocked in a dose-dependent manner by a peptide with the IB␣ degradation sequence (IC 50 ϭ 30 M), and 3) that the quantitative data obtained with the nickel bead assay is consistent with the data observed on the SDS gels.
We investigated the importance of cysteine to the functional activity by incubating the E3 fraction in the presence of Nethylmaleimide and a cysteine reactive ␣-chloroketone diglycine peptide. The peptide was synthesized to mimic the C terminus of ubiquitin. We found that 100 M of N-ethylmaleimide or the ␣-chloroketone blocked the activity when preincubated for 30 min with the E3 (Fig. 4). To ensure that the activity was not against the E1 or E2 components, DTT (5 mM) was added to quench the reaction prior to reconstitution. There was minimal loss of activity of E3-dependent activity when 5 mM DTT was preincubated with 300 M of both compounds prior to addition of E3. These data show that the E3 activity is sensitive to cysteine alkylation reagents.
E2 Specificity-We evaluated three purified E2s (UBCH7, UBCH5a, and atUBC8) for their ability to support substratespecific IB␣ee ubiquitination. All three E2s showed selectivity for ubiquitination of IB␣ee over wtIB␣ (Table II). The selectivity was determined at saturating concentrations of the E2s. The overall activity was 4-fold greater for UBCH5a and atUBC8 as compared with UBCH7. However, the E3-independ-  The rate of IB␣ee ubiquitination was quantitated using the nickel bead assay. The reaction was started with E3 following 20 min of precharging of the E1 and E2 with ATP and ubiquitin. ent ubiquitination (E1/E2 ubiquitination) was greater for UBCH5a and atUBC8 than UBCH7. UBCH7 was used in all subsequent studies because of its lower E3-independent, nonspecific ubiquitination activity, which resulted in a lower background. We also observed that UBCH5a and UBCH7 with N-terminal His 6 tags were unable to transfer ubiquitin to proteins in Jurkat cell lysates or to ubiquitinate IB␣ee, even though these E2s were able to form thioesters in the presence of E1, ATP and ubiquitin (data not shown). This is consistent with the data of Sullivan and Vierstra (37), which showed that the E2 N-terminal domain likely contains the E1-binding site. Therefore, nontagged E2 UBCH7 was used in all of the subsequent E3-dependent reactions.

Discovery Paradigm
The cell-free IB␣ee ubiquitin system was configured in a format suitable for time-resolved fluorescence spectroscopy to facilitate high throughput screening. The system catalyzed the addition of europium-labeled ubiquitin to IB␣ee bound to a 96-well plate. Compounds that inhibited that reaction were counter-screened for their ability to block the formation of His 6 -tagged UBCH7-thioesters. Compounds that inhibited E2 UBCH7 ubiquitin thioester formation were not studied further. Ro106-9920 (Fig. 5) was identified using the screening paradigm as an inhibitor of IB␣ee ubiquitination with an IC 50 value of 2.30 Ϯ 0.09 M, n ϭ 95 (Table III). It did not inhibit the formation of E1 or E2 thioesters or nonspecific ubiquitination of cellular proteins at concentrations up to 100 M. The sulfone analog of Ro106-9920 (compound 1) had similar activity (IC 50 ϭ 6.9 M), whereas the sulfide (compound 3) was inactive at 80 M. The benzimidazole and tetrazole (compound 4) analogs were inactive, as was a ring open phenylazide analog (compound 5). The 4Ј-chlorophenyl analog (compound 2) had activity similar to that of Ro106-9920 (IC 50 ϭ 2.88 M).
The key to developing an assay that was useful for discovering E3-selective compounds was to find the conditions where the E3 was rate-limiting. These conditions were achieved by increasing the concentrations of E1 and E2 UBCH7 to saturation and maximizing the ubiquitin concentration so that its consumption did not become rate-limiting. Under the conditions with which the reaction was run, E2 UBCH7 could be decreased by as much as 70% without dramatically influencing the overall rate. These conditions biased the reaction toward E3-dependent ubiquitination by increasing the window between inhibition of E1/E2-and E3-dependent ubiquitination. Another important factor was maintaining a good signal to noise ratio, which was defined as the ubiquitination in the absence versus presence of IB␣ee, or the comparison of IB␣ee ubiquitination with wtIB␣ ubiquitination. The signal to noise ratio was solely dependent upon the E3 concentration. The amount of nonspecific ubiquitination increased at the expense of specific ubiquitination when the E3 concentration or the specific activity of the E3 preparation was too low. The relative increase in nonspecific ubiquitination decreased the signal to noise ratio and also resulted in an apparent upward shift of IC 50 values of the E3-specific inhibitors. The shift in the IC 50 proved to be the most robust measurement of the integrity of system. We believe that the decrease in the signal to noise ratio resulted in an apparent shift of the IC 50 values, which in turn is due to the ability of the E2 to nonspecifically ubiquitinate other proteins present in the reaction. The increase in nonspe-  cific ubiquitination occurs when there is inadequate E3 available to direct the E2-conjugated ubiquitin to a specific substrate. Time-dependent inhibition was observed when Ro106-9920 was added at the start of the reaction (Fig. 6, top panel). It took 10 -20 min to achieve full inhibition with 10 M of Ro106-9920. A more detailed analysis showed Ro106-9920 to have an apparent irreversible component to its mechanism of action ( Fig.  6, bottom panel). The apparent irreversiblity is evident from the parallel shift in the initial velocity curves at different E3 concentrations (38). A 100-fold difference in E2 concentration (0.1-10 M) had no effect on the IC 50 (data not shown).
The functional integrity of the system was tested by the ability of the inhibitors to block IB␣ degradation and NF-Bdependent cytokine expression in stimulated lymphocytes. IB␣ degradation in MM6 cells treated with TNF␣ was blocked by Ro106-9920 at concentrations above 3 M (Fig. 7). TNF␣, IL-1␤, and IL-6 expression was also inhibited in human PBMNs by Ro106-9920 with IC 50 values of 0.7, 0.6, and 0.7 M, respectively (Table III)

TABLE III
The activity of Ro106 -9920 in cell-free, cellular, and in vivo assays The cell-free assays measured the IB␣ee ubiquitination activity (E1/E2/E3), the activity against E1/E2, and nonspecific ubiquitination of cellular protein in a cell lysate. The human PBMN cell-based assay was used to quantitate the effect of Ro106 -9920 on LPS-induced cytokine production. The ability of Ro106 -9920 to lower serum TNF␣ and prostaglandin E 2 was evaluated in two short term models of inflammation: the rat LPS challenge model and the rat air pouch. The values are presented as the median (min, max) percentage of inhibition (n ϭ 8 for the LPS study and n ϭ 5 for the air pouch). The control for the percentage of inhibition was determined as the median control value. IL-1␤, IL-6, and IL-1ra, respectively. The cytotoxicity of Ro106-9920 was evaluated in Jurkat cells and PBMNs by propidium iodide influx. No cytotoxicity was observed in PBMNs following a 3.5-h incubation at 50 M; however, cytotoxicity was observed in Jurkat cells after 2 h of incubation with Ͼ17.5 M of Ro106-9920. The analogs of Ro106-9920 that were active in the cellfree system were equally effective inhibitors of cytokine expression in the human PBMNs, whereas the inactive analogs had no effect on the cytokine profiles (Fig. 7). These data show that the inhibitors of IB␣ee ubiquitination do elicit the appropriate functional outcome with respect to blocking IB␣ degradation and the suppression of NF-B-dependent gene expression.
Ro106-9920 was evaluated for its ability to prevent ␤-catenin degradation in Jurkat cells. ␤-Catenin is phosphorylated by constitutively active GSK3␤. This targets ␤-catenin for ubiquitination by SCF ␤TRCP and degradation by the proteasome. Incubation of Jurkat cells with the GSK3␤ inhibitor SB-216763 results in an accumulation of ␤-catenin (39); however, Ro106-9920 had no effect in this assay (Fig. 8). From these results we conclude that Ro106-9920 does not inhibit SCF ␤TRCP -mediated ␤-catenin ubiquitination in Jurkat cells.
Ro106-9920 was evaluated for its ability to inhibit cytokine production in two models of acute inflammation in the rat, the carrageenan air pouch, and systemic LPS. Following an oral dose of 10 or 100 mg/kg, a dose-dependent inhibition in serum TNF␣ was observed in both models (Table III). Ro106-9920 (100 mg/kg) also caused dose-dependent suppression of prostaglandin E 2 levels in the air pouch, although this effect did not reach statistical significance (Table III). No acute toxicity was observed in these short term animal models. From these results we conclude that Ro106-9920 blocks NF-B-dependent gene expression in rats.
The selectivity of Ro106-9920 for other molecular targets was evaluated in screening assays provided by Cerep. Ro106-9920 was screened in 74 binding assays described as "Cerep high throughput profile" and 25 enzyme assays referred to as the "Cerep enzyme profile" (Table IV). Screening at a single 10 M concentration in duplicate showed Ͼ40% inhibition of only the epidermal growth factor tyrosine kinase (63%), 5-lipoxygenase (89%), and inducible nitric-oxide synthase (111%).

DISCUSSION
The strategy employed to discover molecules with anti-inflammatory properties that would inhibit the production of NF-B-dependent cytokines such as TNF␣ and IL-1␤ was to target the ubiquitination of IB␣. The first step was to develop tools and a screening paradigm to identify lead molecules. To this end we developed a cell-free assay for E3-dependent IB␣ee ubiquitination, an E2-dependent assay as a selectivity filter, and a screening paradigm that effectively identified compounds with activity in cells and animals. Ro106-9920 was discovered to inhibit IB␣ ubiquitination in a cell-free assay, to prevent IB␣ degradation in MM6 cells, to have no effect on the degradation of ␤-catenin in Jurkat cells, to block the expression of the NF-B-dependent cytokines TNF␣, IL-1␤, and IL-6 in human PBMNs, and to lower the circulating levels of TNF␣ in LPS-treated rats as well as to reduce TNF␣ in exudate harvested from the air pouches of carrageenan-challenged rats.
The cell-free reaction utilized a mutant form of IB␣ as the ubiquitination substrate, IB␣ee, and a partially purified Jurkat cell lysate containing the E3 activity. The validity of the system to reconstitute IB␣ ubiquitination was concluded by showing that 1) wtIB␣ and IB␣eerr were poor substrates but good inhibitors for the reaction, 2) the reaction formed polyubiquitin chains in an E3-dependent manner, 3) the reaction was inhibited by the IB␣ phosphopeptide, and 4) an inhibitor of the cell-free reaction, Ro106-9920, was effective in blocking IB␣ degradation in cells and NF-B-dependent cytokine expression in cells and animals.
As more information becomes available regarding the characteristics of IB␣ ubiquitination, it has become apparent that there were some inconsistencies between the biochemical properties of IB␣ee ubiquitination and those reported for SCF ␤TRCP . Perhaps most disturbing was our inability to identify the F-box component ␤TRCP in the E3 fraction. ␤TRCP was observed to elute from the anion exchange resins at a lower ionic strength than the IB␣ee ubiquitination activity. We were also unable to show an effect of Ro106-9920 on ␤-catenin degradation in Jurkat cells. Numerous reports have shown that ␤TRCP is the F-box component of the SCF complex that catalyzes the ubiquitination of ␤-catenin (34,35). Another inconsistency was that E2UBCH7 supports specific IB␣ee ubiquitination, whereas Ben-Neriah, Ciechanover, and co-workers (24,42,43) have previously shown that E2UBCH5b and E2UBCH5c supported specific IB␣ ubiquitination and that E2UBCH7 supported nonspecific ubiquitination. They demonstrated IB␣ ubiquitination in a cell-free assay using a fractionated E3 preparation isolated from HeLa cells and used in vitro translated, endogenously phosphorylated IB␣ complexed to the NF-B heterodimer as a substrate. Another inconsistency was the inhibition we observed with the thiol-modifying compounds. Strack et al. (27) reported that IB␣ ubiquitination was not blocked by pretreatment of the glutathione S-transferase-␤TRCP/Skp1 or His-Cul1/Rbx1 proteins with N-ethylmaleimide. Another inconsistency was the lack of accumulation of phospho-IB␣ in Western blot analysis of IB␣ levels in TNF␣stimulated MM6 cells treated with Ro106-9920. Kroll et al. (44) observed the accumulation of phospho-IB in TNF␣-treated HeLa cells transfected with a F-box-deleted ␤TRCP mutant. Phosphorylated IB␣ was also observed to accumulate in Jurkat cells transfected with a dominant negative ␤TRCP and stimulated with phorbol-ester and Ca ϩ2 ionophore (25). It could be that the phosphatase activity was greater in our systems or that Ro106-9920 blocks a step required for IKK activation. Altogether these data are inconsistent with what has been reported for SCF ␤TRCP . These data suggest that Ro106-9920 inhibits an ubiquitination activity in the NF-B pathway that FIG. 8. Effect of Ro106-9920 on ␤-catenin degradation. ␤-catenin degradation was measured in Jurkat cells that were incubated at 37°C with inhibitors for 24 h. The cells were washed, resuspended in lysis buffer, frozen, and thawed three times, and the supernatant was analyzed for ␤-catenin by ELISA. SB-216763 dissolved in Me 2 SO was used as the GSK3␤ inhibitor (39).
does not require ␤TRCP but associates with IB␣ and will ubiquitinate IB␣ S32E,S36E (IB␣ee) specifically at lysine 21 or 22. The IB␣ee E3 ligase activity is inhibited by cysteine alkylating reagents, supported by E2UBCH7, requires cIAP2 or a cIAP2 associated protein for activity, and may precede the phosphorylation of IB␣ by the IKK complex.
A number of proteins associated with the NF-B pathway have the properties of ubiquitin E3 ligases including TRAFs (45,46) and cIAP2 (47) or have been associated with ubiquitination including NEMO. 3 TRAF6 ubiquitination activity is sufficient to activate IKK through the TAK1 kinase (46). However, Ro106-9920 was not active against TRAF6-dependent ubiquitination. 4 cIAP2 has been reported to be critically involved in TNF␣-induced NF-B activation (48), and a cIAP2 mutant that lacked the RING domain blocked TNF␣-induced NF-B activity. We detected cIAP2 in the active E3 fraction. Immunodepletion with anti-cIAP2 reduced the IB␣ee ubiquitination activity by 70%, demonstrating cIAP2 to be essential and a promising candidate for the E3 ligase activity. However, reconstitution with recombinant cIAP2 purified from E. coli did not effect the IB␣ee ubiquitination activity. It may be that the purified recombinant cIAP2 was not functionally active or that the reaction requires additional components. Although cIAP2 is an enticing candidate for the IB␣ee E3 ligase, these data do not confirm the hypothesis.
Why was SCF ␤TRCP not identified? Why is the IB␣ee a substrate for the unidentified E3 ligase and what is the function of the unidentified ligase? The answer to why SCF ␤TRCP was not identified is most likely the use of E2UBCH7 as the E2 and IB␣ee as the substrate. Traenckner et al. (14) had previously observed in HeLa cell transfection studies that the double glutamic acid mutant did not stimulate IB␣ degradation, and as previously noted, E2 UBCH7 was reported not to support specific IB␣ ubiquitination (24,42,43). The specificity of SCF ␤TRCP -dependent IB␣ ubiquitination must be rigorously dependent upon the nature of the E2 and the substrate. As to why IB␣ee is a substrate for the unidentified E3 ligase, it is possible that IB␣ee is a ubiquitination substrate in vitro because of features in its ubiquitination domain that are similar to those of the physiological relevant substrate and/or that IB␣ is a substrate for more than one E3. As to the function of the unidentified E3, the lack of accumulation of phosphorylated IB␣ suggests that the E3 that catalyzes IB␣ee ubiquitination maybe involved in the activation of the IKK kinase. Chen et al. (28) have reported an unidentified ubiquitin activity supported by E2UBCH5 that was required for activation of IKK. As far as we are aware the source of the ubiquitination activity described by Chen has not been identified.
In summary, the work described provides 1) a small molecule ubiquitination inhibitor, 2) evidence to show that a specific E3 ubiquitination activity can be blocked by small molecules with selectivity in respect to E1 and E2, and 3) evidence for the involvement of an unidentified E3 ligase in NF-B activation.

TABLE IV
The effect of Ro106 -9920 in receptor binding and enzyme assays The results are represented as the mean percentages of inhibition of control specific binding (receptors) or activity (enzymes) (n ϭ 2) following incubation with 10 M of Ro106 -9920. The symbol Ϫ indicates less than 10% inhibition. The symbol ϩ indicates a greater than 50% increase in binding or activity.