The NF-κB family of inducible transcription factors regulates the expression of numerous genes, which are central to developmental and immune processes, cell survival, proliferation, and differentiation (Baeuerle and Henkel, 1994). However, dysregulated NF-κB activity leads to the onset of several human pathologies, including cancer and inflammatory diseases such as rheumatoid arthritis, asthma, and inflammatory bowel disease (Karin and Greten, 2005). The central dogma on the activation of NF-κB is that these proteins generally exist as dimers in the cytoplasm of resting cells and are bound to the typical inhibitory IκB proteins, such as IκBα and IκBβ (Chen et al., 1995; DiDonato et al., 1995). In response to a variety of agonists, IκBs are rapidly phosphorylated, polyubiquitinated, and degraded through the S26 proteasome pathway, thus releasing NF-κB, which translocates into the nucleus to initiate gene transcription (DiDonato et al., 1996). However, accumulating evidence shows that the atypical IκB members, such as Bcl-3 and IκBζ, are localized in the nucleus; therefore, the activation and functions of NF-κB are complex and cell context-dependent (Hayden and Ghosh, 2008). In support of this notion, although the development of secondary lymphoid tissues and the maturation of B cells are mainly dependent on the RelB/p100 heterodimers, most of the inflammatory processes are associated with the activation of the p65/p50 heterodimers (Bonizzi and Karin, 2004). Post-translational modifications, including phosphorylation and acetylation, have been shown to modulate NF-κB transcriptional activity (Ghosh and Karin, 2002). For example, p65 is phosphorylated on Ser536 by IKK-1 and IKK-2 in vitro and in a variety of activated cells (Sakurai et al., 1999; Jiang et al., 2003; Kishore et al., 2003; Yang et al., 2003; Mattioli et al., 2004).

IKK-1/IKK-α and IKK-2/IKK-β are the catalytic subunits of the IKK complex in which NF-κB essential modulator (NEMO)/IKKγ functions as a regulatory subunit (Karin and Lin, 2002). The findings that mice with gene deletion of NEMO, IKK-2, or p65 display a similar phenotype strongly support the notion that these components are indispensable and belong to the same signaling pathway (Beg et al., 1995; Li et al., 1999a; Rudolph et al., 2000). Different models of activation of the IKK complex have been depicted, one proposing that K63-linked polyubiquitination of NEMO is required for the recruitment of TAK-1 complex and subsequent phosphorylation of IKK-1/IKK-2 (Neumann and Naumann, 2007). Although IKK-1 has been shown to activate the canonical NF-κB pathway, overwhelming evidence indicates that IKK-2 is required for the activation of this pathway by inflammatory stimuli. Early evidence supporting this claim originated from studies using IKK-2 and IKK-1 knockout models and transgenic expression of IKK-1 and IKK-2 dominant negative constructs (Hu et al., 1999; Li et al., 1999b; Tas et al., 2006). The proof that IKK-2 activity accounts for most of the activity of the IKK complex in response to inflammatory stimuli ultimately made IKK-2 the target of choice for the drug discovery programs. In fact, several small-molecule inhibitors of IKK-2 have been identified with some of them being currently in clinical development (Burke et al., 2003; Wen et al., 2006). The efficacy of these inhibitors has been demonstrated clearly in animal models of arthritis, asthma, and cancer (McIntyre et al., 2003; Podolin et al., 2005; Ziegelbauer et al., 2005; Schopf et al., 2006). However, to our knowledge, none of the available inhibitors of IKK-2 have been shown to bind tightly to this enzyme, and preclinical safety assessments of IKK-2 inhibitors based on a comprehensive understanding of pharmacokinetic/pharmacodynamic relationships have yet to be reported. Therefore, novel IKK-2 inhibitors with unique binding properties and proven efficacy and safety in animal models of chronic inflammation are still of great interest as additional tools to unravel the mechanisms of IKK-2 regulation of NF-κB signaling and to further validate IKK-2 as a therapeutic target. Here, we describe a potent and highly selective IKK-2 inhibitor, PHA-408, which binds tightly to IKK-2 and inhibits the NF-κB pathway in vitro and in vivo in an arthritis disease model at doses within the safety margin. Furthermore, we demonstrate that PHA-408 is an ideal tool to evaluate IKK-2 activity and efficacy endpoints in human blood samples, which is unprecedented.

Materials and Methods

Reagents. Nonidet P-40, bovine serum albumin (BSA), ATP, ADP, lipopolysaccharide (LPS), protein A agarose, anti-FLAG M2-agarose, Ficoll histopaque-1077, phosphatase inhibitor cocktail, dithiothreitol, and dexamethasone were obtained from Sigma-Aldrich (St. Louis, MO). Antibodies to p-HSP-27, p-p65, and total p65 and ELISA for human and rat TNF-α, PGE₂, and p-p65 were developed by Pfizer Inc. (St. Louis, MO). Antibodies specific for NEMO, IKK-2, IκBα, p-IκBα, p65, p38 MAPK, extracellular signal-regulated kinase-2, and HSP-27 were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). p-p38 MAPK, p-extracellular signal-regulated kinase, p-JNK, and p-IKK-1/2 were from Cell Signaling Technology Inc. (Danvers, MA), and secondary antibodies were supplied by Jackson ImmunoResearch Laboratories Inc. (West Grove, PA). DMEM, DPBS, gentamicin, penicillin-streptomycin, nitrocellulose membranes, Tris-glycine acrylamide gels, IL-6 and IL-8 ELISA kits, and Alamar Blue were obtained from Invitrogen (Carlsbad, CA). Biotinylated peptides, 5-carboxyfluorescein (5-FAM) peptides, and Z-Leu-Leu-Leu-CHO were from American Peptide Co., Inc. (Sunnyvale, CA). Great EscAPeTM SEAP detection kit (Clontech, Mountain View, CA), Biotin Capture Plates (Promega, Madison, WI), and rhIL-1β (R&D Systems, Minneapolis, MN) were all purchased as indicated. ECL plus kits and ³³P- and ³⁵S-labeled ATP were obtained from GE Healthcare (Chalfont St. Giles, UK). All other reagents used were of the highest grade commercially available.

Inhibitors. PHA-408 (H.-C. Huang, unpublished data describing the synthesis procedure) and p38 MAPK inhibitor SC-806 (Graneto et al., 2007) were synthesized at Pfizer Inc. For in vitro assays, inhibitors were dissolved in Me₂SO and stored at -20°C in 50 mM aliquots.

Kinase Assays with Recombinant IKK. The kinase activities of the rhIKK isoforms and K_m determinations of rhIKK-2 have been described previously (Huynh et al., 2002; Kishore et al., 2002). In brief, various concentrations of [³³P]ATP (2500 Ci/mmol) or biotinylated IκB peptide substrate were used in the assay at a fixed (3× K_m) concentration of the second substrate and 100 ng of rhIKK-2 or other IKK isoforms in a final volume of 50 μl of kinase buffer (25 mM HEPES, pH 7.6, 2 mM MgCl₂, 2 mM MnCl₂, 10 mM NaF, 5 mM DTT, and 1 mM phenylmethylsulfonyl fluoride). After incubation for 30 min at 25°C, the reaction was stopped by the addition of 150 μl of AG1XB resin in 900 mM sodium formate buffer, pH 3.0 (the resin is in slurry of 1 volume of resin/2 volumes of sodium formate buffer). The resin was allowed to settle, and 50 μl of supernatant was transferred to a top count plate followed by the addition of 150 μl of Microscint 40, mixed well, and counted. Once K_m graphs were fitted using the GraFit program, kinetic constants were calculated from V_max values.

For the kinase reversibility assay, rhIKK-2 was incubated with ADP, PHA-408, or 2% Me₂SO vehicle for 1 h. After incubation, a set of samples were passed through micro-biospin columns (Bio-Rad, Hercules, CA) according to the manufacturer's instructions and further diluted 10-fold. Samples were assayed for IKK-2 activity as described.

The dissociation rate constant of PHA-408 from the phosphorylated form of IKK-2 was determined by following the recovery of IKK-2 kinase activity over time. For these studies, the kinase activity of rhIKK-2 was determined by following the phosphorylation of 5-FAM-GRHDSGLDSMK (American Peptide Co., Inc.) using the Caliper Lab Chip 3000. PHA-408 was preincubated (200 nM) with 100 nM p-IKK-2 for 2 h at room temperature in kinase buffer (25 mM HEPES, 10 mM NaF, 0.1% BSA, 0.0005% Triton X-100, 5 mM MnCl₂, 5 mM MgCl₂, 1 mM dithiothreitol, and 2% Me₂SO, pH 7.5). The preincubation mixture was diluted 100-fold into an assay mixture containing 100 μM MgATP (100× K_m) and 2.0 μM 5-FAM-GRHDSGLDSMK. The production of the phosphorylated 5-FAM-GRHDSGLDSMK was followed for 6 h. The dissociation rate constants were determined using GraFit 4.0 (Leatherbarrow, 1992) and eq. 1, where [P] is the product concentration at any time (t), v₀ and v_s are the initial and final steady-state rates, respectively, and k_-1 is the apparent first order rate constant for the establishment of the final steady-state velocity. Under the conditions of this assay, the value of v₀ and the effective inhibitor concentration were considered to be approximately zero. Thus, the rate of activity regenerated will provide the dissociation rate constant, k_-1. The final steady-state rate, v_s, was determined from a control incubated without inhibitor:

Kinase Selectivity Assays. A panel of 30 tyrosine and serine/threonine kinase assays were run in a 384-well streptavidin capture format. Recombinant kinases were incubated in the presence of specific-biotinylated peptides and a mix of both unlabeled and ³³P-labeled ATP (K_m of ATP). The reaction was stopped via EDTA chelation, and then the reaction mix transferred to a flash plate for readout. More extensive targets were evaluated using assays provided by Invitrogen and Millipore (Billerica, MA) using 10 μM PHA-408.

Endogenous IKK Complex Assay. Kinase activity was determined from immunoprecipitated IKK complexes from IL-1β-treated rheumatoid arthritis-derived synovial fibroblasts (RASFs) and LPS-treated peripheral blood mononuclear cell (PBMC) whole-cell lysates and from liver and paw extracts from rats treated with LPS and streptococcal cell wall (SCW), respectively. Livers or paws were powdered under dry ice, and 5 to 20 mg of the powder was lysed with 1 ml of whole-cell lysis buffer supplemented with fresh protease and phosphatase inhibitors on ice with intermittent vortexing. Samples underwent a high-speed spin, and the supernatants were removed and precleared with protein A agarose beads for 30 min before use. IKK complex was captured using an NEMO antibody (3-10 μg) conjugated to protein A agarose beads and was pelleted by centrifugation and washed three times with 1 ml of cold lysis buffer followed by two washes in kinase buffer. One hundred micrograms of the immunoprecipitated IKK complex was analyzed for kinase activity in a reaction containing 10 μM biotinylated IκB peptide and 1 μM [³³P]ATP as described previously (Huynh et al., 2002; Kishore et al., 2002). After incubation at room temperature for 40 min, 25 μl of the reaction mixture was withdrawn and added to a SAM 96 biotin capture plate. After successive wash steps, the plate was allowed to air-dry, and 25 μl of scintillation fluid was added to each well. Incorporation of [³³P]ATP was measured using a Top-Count NXT (PerkinElmer Life and Analytical Sciences, Waltham, MA).

Rheumatoid Arthritis-Derived Synovial Fibroblast Cultures. Adherent RASFs were isolated via enzymatic digestions from primary synovial tissues isolated after knee synovectomy and were cultured in DMEM high glucose, containing 15% defined bovine serum (HyClone Laboratories, Logan, UT) and 50 μg/ml gentamicin. For cytokine and PGE₂ release, NF-κB-linked reporter activity, and toxicity determination, 1.5 × 10⁴ cells/well were plated in a 96-well plate and allowed to attach overnight. The growth media were replaced with fresh DMEM containing 1% serum, and the cells were pretreated with increasing concentrations of PHA-408 (0.2% Me₂SO final) for 1 h before an 18-h stimulation with 1 ng/ml IL-1β. Cytokines and PGE₂ secreted into the culture media were measured by ELISA. Cytotoxicity was assessed using an Alamar Blue assay. The AdNF-κB SEAP construct and assay was performed as described previously (Kishore et al., 2003). In brief, RASF cells were transduced with the κB-linked SEAP adenoviral construct, and SEAP activity was measured using a Great EscAPe SEAP detection kit. For Western analysis, EMSA analysis, and in vitro kinase assays, 5 × 10⁵ cells were seeded in six-well plates and allowed to attach overnight. The growth medium was replaced with fresh DMEM containing 1% serum before PHA-408 pretreatment (10 μM or the indicated concentrations) and IL-1β stimulation.

Peripheral Blood Mononuclear Cell Cultures. Human whole blood (HWB) was collected from healthy donors in sodium heparinized tubes (BD Biosciences, Franklin Lakes, NJ), and PBMCs were isolated by Ficoll separation. Cells were washed in DPBS, resuspended in DMEM containing 5% endotoxin-free fetal bovine serum and 10 U/ml penicillin-streptomycin, and plated at 2.5 × 10⁵ cells/well in 96-well tissue culture plates. Cells were pretreated with increasing concentrations of PHA-408 (0.001-3 μM) or the p38 MAPK inhibitor, SC-806 (1 μM), for 1 h before the 18-h stimulation with 100 ng/ml LPS. Cytokines secreted into the culture media were measured by ELISA. Cells (5 × 10⁶) were plated in six-well dishes for inhibitor pretreatment and LPS stimulation for Western analysis and in vitro kinase assays. For PHA-408 washout or duration of action studies, individual plates were utilized, and inhibitor-treated cells were washed twice with media 0 and 4 h before LPS stimulation for 20 min before measuring IKK-2 activity.

Human Whole-Blood Cultures. Blood from healthy volunteers was aliquoted into 3-ml samples and received the following: PHA-408 (0.1-30 μM) or Me₂SO vehicle control and 10 μM proteasome inhibitor. After 1-h pretreatment, 10 μg/ml LPS was added to all tubes and incubated for 20 min at 37°C except for unstimulated controls. PBMCs for lysates were prepared from the whole blood by layering 1 ml of blood onto 900 μl of Histopaque 1077 in 2-ml tubes. Tubes were centrifuged at 16,000g for 90 s at room temperature. PBMCs at the interface were drawn off and rapidly washed in ice-cold DPBS. Pelleted cells were flash-frozen and stored at -80°C until further use. For TNF-α production, 180 μl of whole blood was aliquotted in 96-well plates, pretreated with PHA-408 (0.1-30 μM) or Me₂SO, and stimulated with LPS at 37°C for 4 h. Serum for TNF-α determinations was separated from red blood cells by centrifugation at 3000g for 10 min.

Western and EMSA Analyses. Cell lysates from RASFs or PBMCs were prepared for Western analysis at the indicated times by incubating previously washed and pelleted cells on ice for 30 min in cell lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM NaVO₃, complete protease inhibitor cocktail, phosphatase inhibitor cocktail, 1 mM DTT, and 0.5% Nonidet P-40). PBMC lysates were transferred to a Qiashredder (QIAGEN, Valencia, CA), washed with an additional 100 μl of lysis buffer, and centrifuged at 16,000g for 4 min to ensure complete lysis. Liver lysates were prepared as described above in the endogenous IKK complex assay. Western analysis methods were described previously (Kishore et al., 2003). In brief, equal amounts of proteins (10-30 μg) were separated by SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes. Blots were then incubated overnight with 1:1000 primary antibodies washed three times and incubated with either horseradish peroxidase- or fluorescence (IR700, IR800)-labeled secondary antibodies for 1 h. Enhanced chemiluminescence (ECL plus; GE Healthcare) or fluorescence (LI-COR, Lincoln, NE) was used for detection.

Electrophoretic Mobility Shift Assay Analysis. The preparation of nuclear lysates and EMSA analysis were performed as described previously (Kishore et al., 2003). In brief, NF-κB probe was generated in a 10-μl reaction containing 20 ng of NF-κB double-stranded oligonucleotide, 2 μl of [³²P]ATP (3000 Ci/mmol), 1 μl of T4 polynucleotide kinase (10 units/μl), and nuclease-free H₂O for 10 min at 37°C. Unincorporated [³²P]ATP was removed with a G-50 spin column. For the binding assay, 5 μg of nuclear extracts was preincubated in 10 mM HEPES, pH 7.7, 10% glycerol, 50 mM NaCl, 0.5 mM MgCl₂, 1 mM DTT, and 2 μg of poly(dI-dC) for 30 min on ice. Then, 1 μl (50,000 cpm) of ³²P oligonucleotide was added, and the incubation proceeded for an additional 30 min. Samples were separated on a 4 to 20% gradient acrylamide gel in 1× Tris borate-EDTA. The gel was dried, and NF-κB binding was visualized by autoradiography.

Phospho-p65 ELISA Assay. Ninety-six-well Maxisorp Immuno plates (Nalge Nunc International, Rochester, NY) were coated with rabbit anti-p-p65 (2 μg/ml), washed three times, blocked with BSA, and washed an additional three times before use. PBMC lysates (10 μg), prepared from whole blood (Fig. 4B) and diluted in 0.05% Tween 20 buffer containing protease and phosphatase inhibitors, were added to the plates and allowed to shake at 4°C for 2 h. Plates were washed three times, and mouse total anti-p65 antibody was added for 1.5 h. After washing, horseradish peroxidase-labeled goat anti-mouse was added to both total and p-p65 plates for 1 h. TMB 1 substrate (BioFX Laboratories, Inc., Owings Mills, MD) was used to develop the ELISA. Total protein levels in lysates were extrapolated from standard curves using recombinant total p65 or p-p65 full-length proteins.

Rat Model of Acute Inflammation. Vehicle (0.5% methylcellulose, 0.025% Tween 20), PHA-408 (0.5-50 mg/kg), dexamethasone (1 mg/kg), or SC-806 (p38 MAPK inhibitor, 20 mg/kg) was administered by oral gavage to adult male Lewis rats (5 rats/group) that had been deprived of food overnight. At the indicated times after inhibitor treatment, 1 mg/kg LPS (Escherichia coli) in saline was administered intravenously. Ninety minutes after LPS administration, the animals were bled, and serum TNF-α levels were analyzed by a rat-specific TNF-α ELISA. Livers were also harvested and flash-frozen for Western analysis and kinase assays.

Rat Streptococcal Cell Wall Model of Arthritis. Arthritis was induced in 125- to 140-g female Lewis rats (Harlan, Indianapolis, IN) by a single intraperitoneal administration of peptidoglycan-polysaccharide complexes isolated from group A SCW (15 μg/g body weight). The SCW preparation was purchased from Lee Laboratories (Grayson, GA). The disease course is biphasic, in which an acute inflammatory arthritis develops within days 1 to 3 (non-T-cell-dependent phase) followed by a chronic erosive arthritis (T cell-dependent phase) developing on days 14 to 28 (Kuiper et al., 1998). Only animals developing the acute phase (8 rats/group) were treated or not with PHA-408 from days 10 to 21 after SCW injection. Paw volume was measured on day 21 using a water displacement plethysmometer.

For micro-computed tomography (CT) imaging, hind paws obtained from the SCW study (day 21) arthritic rats were scanned using the high-resolution micro-CT (vivaCT; SCANCO USA, Inc., Southeastern, PA). Four normal paws and four paws from the vehicle group and 10 mg/kg/day treatment group were scanned. Micro-CT images were acquired during a 15-min scan at 70 kVP, 85 μA, and standard resolution, and 636 slices of 21 μm thickness each were measured, and the resulting scan length of 13.4 mm covers the entire foot of the rat. Three-dimensional images of the ankle were generated using sigma, 0.7; support, 1; and threshold, 225. All animal studies were reviewed and approved by the Pfizer Institutional Animal Care and Use Committee. The Pfizer animal care and use program is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, International.

Liquid Chromatography/Tandem Mass Spectrometry. Plasma samples were analyzed for PHA-408 by liquid chromatography mass spectrometry (liquid chromatography/tandem mass spectrometry). The mass spectrometer was a Sciex API3000 (Applied Biosystems, Foster City, CA) equipped with a TurboIonSpray source operated in positive ion mode. The liquid chromatography was a Shimadzu LC system (Shimadzu, Kyoto, Japan) with a 50 × 2.1-mm Alltima C18 column (Alltech Associates, Deerfield, IL). A 10-μl aliquot of sample was injected onto the column for analysis. Solvent A was 95% water/5% acetonitrile with 0.1% formic acid, and solvent B was 100% acetonitrile with 0.1% formic acid. PHA-408 was eluted using a gradient started with 10% solvent B for 1 min, then increased to 100% solvent B in 3 min and maintained for 1 min, followed by returning to 10% solvent B and held for 1 min. Total run length was 5.2 min at flow of 0.4 ml/min. PHA-408 was quantified by the multiple reaction monitoring method (Q1/Q3, 560.246/514.113) versus an internal standard (an in-house compound) with declustering potential, 51 V; collision energy, 51 V; and collision exit potential, 16 V.

Rat Safety Study. A 2-week safety study was carried out using female Lewis rats (∼200g). PHA-408 (prepared as a suspension in 0.5% methylcellulose/0.1% Tween 80) was administered by oral gavage at doses of 0, 15, 30, 45, and 60 mg/kg/day (n = 6 animals/group). A t.i.d. dosing regimen was employed to minimize peak to trough ratios. Animals were dosed at approximately 8-h intervals. Endpoints assessed included clinical and gross observations, hematology, clinical chemistry, and histopathology. Plasma was collected on day 14 at 1, 2, 4, and 8 h after dose for determination of PHA-408 pharmacokinetic parameters.

Results

PHA-408 Is a Potent and Highly Selective Inhibitor of IKK-2. PHA-408 (Fig. 1A) was initially identified from a tricyclic pyrazole chemical series with potent activity against rhIKK-2. Upon further enzymatic characterization, we determined that PHA-408 inhibited rhIKK-2 homodimer and rhIKK-2/IKK-1 heterodimer with comparable IC₅₀ values in the 10 to 40 nM range (Fig. 1B). PHA-408 weakly inhibited rhIKK-1 with an IC₅₀ of 14 μM (350-fold selectivity) and did not have any inhibitory activity against its closely related members, TBK1 and IKKi (Fig. 1C). Furthermore, when evaluated against a panel of 30 tyrosine and serine/threonine kinases, PHA-408 only inhibited PIM1, with 15-fold selectivity compared with rhIKK-2 (Fig. 1C). PHA-408 was also screened at 10 μM against a broader panel of enzymes, receptors, and channels and showed at least >100-fold selectivity (data not shown). These data indicate that PHA-408 is a potent and highly selective inhibitor of IKK-2.

PHA-408 Is ATP Competitive and a Tight Binding Inhibitor of IKK-2. PHA-408 showed competitive inhibition with respect to the ATP site (Fig. 2A), with K_i = 6 nM, and noncompetitive inhibition with respect to the IκB site (data not shown). This suggests that PHA-408 was targeted to the ATP-binding site of IKK-2. Using an ATP binding assay that we described previously (Kishore et al., 2003), we found that PHA-408 competitively inhibited the binding of [γ-³⁵S]ATP to rhIKK-2, with an IC₅₀ similar to that obtained from the kinase studies above (data not shown).

To characterize further the binding properties of PHA-408 to IKK-2, 1 μM ADP or 1 μM PHA-408 were preincubated with rhIKK-2 at concentrations that inhibited >95% enzyme activity. The resulting samples were passed through micro-biospin columns to remove unbound PHA-408, and the flow through was assayed for kinase activity. PHA-408-treated rhIKK-2 remained inhibited by 90%, whereas the ADP-treated rhIKK-2 regained activity similar to buffer-treated enzyme (Fig. 2B). Upon an additional 1:10 dilution of the filtrate, the PHA-408-treated enzyme was still significantly inhibited (Fig. 2B).

Because it was unclear whether PHA-408 was a slow off or an irreversible inhibitor, the dissociation rate constants were evaluated by preincubating rhIKK-2 with PHA-408 for 1 h. The reaction was then diluted extensively (100-fold) into a reaction mixture containing a high concentration of ATP (100 × K_m), and the formation of phosphorylated peptide was evaluated as a function of time. Under these conditions, the change in the rate of product formation reflects the dissociation of the enzyme-inhibitor complex. As shown in Fig. 2C, based on the results of three independent experiments, the activity of PHA-408-treated IKK-2 recovered very slowly, with off rate (k_-1) of approximately 0.36 ± 0.0023 h^-1 and the half-life of the PHA-408-IKK-2 complex of approximately 2 h. Although the binding of PHA-408 was reversible, the data showed that the recovery of enzyme activity was slow. These data indicate collectively that PHA-408 is an ATP-competitive inhibitor that binds tightly to IKK-2.

PHA-408 Also Binds Tightly to Endogenous IKK-2 in Cells. As part of the efforts to establish optimal cellular assay conditions, we showed that upon activation of PBMCs by LPS, IKK activity from immunoprecipitated lysates was rapidly increased, reaching peak activity at 15 min and declining to baseline by 90 min (Fig. 3A). In agreement with these data, IκBα levels in the same samples decreased within 20 min post-LPS stimulation, an observation that is also consistent with increased phosphorylation of IκBα and p65 at Ser536 (Fig. 3B). As expected, the levels of total p65 were not affected by LPS treatment. Based on the data from Fig. 2C and because PHA-408 is selective for IKK-2 over IKK-1 (350-fold selectivity), we reasoned that PHA-408 should coimmunoprecipitate with IKK-2 if in fact it binds tightly to this enzyme. To test this idea, PBMCs were preincubated with varying concentrations of PHA-408 and stimulated with LPS for 20 min (optimal time of IKK activation) followed by several wash steps to remove free PHA-408 before cell lysis. IKK was immunoprecipitated from lysates using anti-NEMO antibody and IKK kinase activity measured (Fig. 3C). A duplicate set of samples was used to analyze the state of phosphorylation of IκBα and p65 by Western blot (Fig. 3D), and the intensity of the bands corresponding to each protein was quantified by fluorescence densitometry (Fig. 3C). The data show that PHA-408 inhibited IKK-2 activity and the phosphorylation of IκBα and p65 on Ser536, with a similar potency in the 50 nM range. Note that PHA-408 inhibited IKK-2 with equal potency in living cells (Fig. 3C) and cell-free systems (Fig. 1, B and C). As expected, the levels of total p65 were not affected by PHA-408.

To evaluate the duration of action and determine whether PHA-408 binds to inactive and activated IKK-2, PBMCs were pretreated with PHA-408 for 1 h, then either washed or not washed with media before 20-min LPS stimulation. Kinase activity was evaluated from immunoprecipitated IKK complexes. The data show that washing had no effect on PHA-408 potency in inhibiting IKK-2 activity (Fig. 3E), whereas the efficacy of a reversible control IKK-2 inhibitor was completely abolished after this washing protocol (data not shown). Furthermore, the data suggest that PHA-408 can bind tightly to either the unstimulated kinase complex (Fig. 3E) or activated IKK-2 as shown in Fig. 2. The duration of action of PHA-408 in cells extended over 4 h and was consistent with biochemical data.

Using synovial fibroblasts isolated from rheumatoid arthritis patients (RASFs), we further confirmed that PHA-408 inhibited IL-1β-stimulated IKK-2 activity when added to cells before IL-1β stimulation and immunoprecipitation or when added directly to the immunoprecipitated kinase with a similar potency (Supplemental Fig. S1A). Furthermore PHA-408 inhibited p65 phosphorylation and DNA binding activity (Supplemental Fig. S1, B and C). Thus, because of the tight binding action of PHA-408, for the first time, we were able to correlate directly IKK-2 activity with p65 phosphorylation on Ser536 and IκBα phosphorylation and degradation.

PHA-408 Selectively Inhibits IL-1β- and LPS-Induced Expression of Various Inflammatory Mediators. The effects of PHA-408 on IL-1β- and LPS-induced production of inflammatory mediators were evaluated in PBMCs, HWB, and RASFs. In addition, the effects of PHA-408 on IL-1β-stimulated NF-κB reporter gene (SEAP) in RASFs were analyzed. PHA-408 demonstrated a concentration-dependent inhibition of inflammatory stimuli-induced production of inflammatory mediators in PBMCs (Fig. 4A), HWB (Fig. 4B), and RASFs (Fig. 4C). There was no cell toxicity at concentrations of PHA-408, which completely inhibited gene expression (Fig. 4C). Although the IC₅₀ values for PGE₂ and SEAP in RASFs were comparable with those of IL-6 and IL-8 in PBMCs in the 0.2 to 0.6 μM range, the IC₅₀ values for TNF-α inhibition in PBMCs and HWB (after correction for free fraction; PHA-408 binds 99% to proteins) were in the 0.02 to 0.03 μM range. The IC₅₀ values for PHA-408 inhibition of TNF-α were closely correlated with the IC₅₀ values on IKK-2. Likewise, the similar potencies of PHA-408-mediated inhibition of NF-κB-dependent reporter activity and those of PGE₂, IL-6, and IL-8 suggest that they are also inhibited through an IKK-2 mechanism of action. This differential inhibition of gene expression by PHA-408 probably reflects the complex nature of the interactions between NF-κB and the regulatory elements of its targets in the context of chromatin structures. Similar observations were made with other reported IKK-2 inhibitors (Burke et al., 2003; Podolin et al., 2005; Ziegelbauer et al., 2005; Wen et al., 2006). Here, not only was the inhibition of IKK-2 activity by PHA-408 correlated to p65 phosphorylation but also to the inhibition of the production of inflammatory mediators measured in a complex matrix such as HWB (Fig. 4B).

Although we demonstrated that PHA-408 had an excellent selectivity profile against numerous targets in cell-free systems; however, these findings did not rule out potential PHA-408 off-target effects in living cells. To address this point, experiments were carried out to show that treatment of RASFs with 10 μM PHA-408 prevented IL-1β-induced IκBα phosphorylation in RASFs while sparing the JNK and p38 MAPK pathways as measured by phosphorylation of p-JNK and HSP-27, respectively (Fig. 4D). PHA-408 had no effect on IKK-2 phosphorylation (data not shown), suggesting that it inhibited IKK-2 activity, but not its upstream activation. Although partial IκB degradation was observed in PHA-408-treated cells 45 min after IL-1β-stimulation, IκBα and p65 phosphorylation and subsequent gene products were completely blocked, suggesting that IκBα degradation is only one of the mechanisms by which IKK-2 regulates the expression of NFκB targets in RASF cells. These data reinforce the notion that PHA-408 is a highly selective inhibitor of IKK-2 in cells.

PHA-408 Inhibits LPS-Induced TNF-α Production in Vivo in the Rat Acute Model of Inflammation. To evaluate the efficacy of PHA-408 to suppress inflammation in vivo, PHA-408 was given orally and was assessed for its ability to inhibit TNF-α production in the rat acute model of LPS-induced cytokine production. PHA-408 dosed at 50 mg/kg, which yielded plasma concentrations in the range 4 μM, maximally inhibited LPS-induced serum TNF-α production when administered either 1 or 4 h before LPS challenge (Fig. 5A). The magnitude of PHA-408 inhibition of TNF-α production was similar to that of the benchmark inhibitor, dexamethasone, and the potent and selective p38 MAPK inhibitor, SC-806 (Graneto et al., 2007). Liver lysates from LPS and/or PHA-408-treated rats were generated to assess IκBα and other MAPK pathways by Western analysis. As in in vitro cell systems, PHA-408 reduced IκBα degradation but had no inhibitory activity against the phosphorylation of p38 MAPK (Supplemental Fig. S2A), indicating that the selectivity of PHA-408 was preserved at efficacious doses in vivo. PHA-408-inhibitory effects on LPS-induced TNF-α production and IKK-2 activity were dose-dependent (Fig. 5B) and were consistent with the degradation of IκBα (Supplemental Fig. S2B). Confirming the data shown above, the inhibition of IKK-2 activity immunoprecipitated from rat liver lysates correlated well with TNF-α inhibition (Fig. 5B). Furthermore, the translatability of these studies was demonstrated by the comparable potency of PHA-408 inhibition of TNF-α production in HWB and rat LPS assay, with IC₅₀ values of 2.7 and 2 μM, respectively (Supplemental Fig. S2C). Likewise, PHA-408 IC₅₀ in the PBMC assay containing 5% serum was comparable with the corrected free fraction IC₅₀ of PHA-408 in rat serum (∼0.02 μM).

To study the effect of time on pharmacokinetics (PHA-408 plasma concentrations) and pharmacodynamics (e.g., TNF-α production), rats were pretreated with 50 mg/kg PHA-408 for the indicated times, followed by LPS injection, and 90 min later, blood was collected to measure TNF-α and PHA-408 levels. The data show that the inhibition of TNF-α inhibition inversely correlated with PHA-408 plasma concentrations, with PHA-408 EC₅₀ in the 3.4 μM range, which is consistent with in vitro and ex vivo cell data (Fig. 5C).

PHA-408 Inhibits Joint Swelling and Bone Destruction in Streptococcal Cell Wall-Induced Arthritis in Rats. The effect of PHA-408 was further evaluated in the rat SCW model of chronic inflammation (Kuiper et al., 1998). Initial studies with PHA-408, dosed up to 100 mg/kg, once daily starting at day 10, showed a steep efficacy curve as measured by paw volume and a lack of a correlation between PHA-408 doses and exposures [area under the curve (AUC)] (Fig. 6A). This nonlinear dose versus exposure relationship, suggesting possible PHA-408 accumulation in tissues or saturation of the transporter mechanisms over time, led to modeling efforts for an optimal dosing regimen of this inhibitor. The data show that dosing PHA-408 once every 3 days (Fig. 6B) or t.i.d. with lower PHA-408 doses (Fig. 6C) were more optimal dosing regimens because dose/exposure relationships were linear, and a nice dose-response of efficacy was obtained. Maximal efficacy with the t.i.d. dosing regimen was achieved with the 30 mg/kg dose (3 × 10 mg/kg dose), which corresponds to EC₈₀ of paw swelling inhibition. Further analyses using the t.i.d. experimental protocol show that efficacy was driven by C_min plasma concentrations of PHA-408 (Fig. 6D) and that PHA-408 at a steady-state plasma concentration range of 0.2 to 10 μM inhibited paw swelling and the activity of immunoprecipitated IKK-2 from paw samples with the same potency (Fig. 6E). In addition, micro-CT analysis shows that PHA-408 spared bones from the massive destruction that occurs in this model (Fig. 6F), bone-protecting effects that were dose-dependent (data not shown). Together, the data show that PHA-408 EC₅₀ in the chronic rat SCW model correlates well with EC₅₀ in the acute rat LPS-induced TNF-α production and are consistent with the predominant role of TNF-α in disease development in SCW (Mbalaviele et al., 2006).

To study the effect of time on pharmacokinetic/pharmacodynamic parameters in chronic inflammation settings, rats were treated therapeutically with a single dose of PHA-408 at day 19, then samples were collected over time to measure IKK-2 activity and PHA-408 levels. Consistent with the data above, the inhibition of IKK-2 activity inversely correlated with PHA-408 plasma concentrations (Fig. 6G). Collectively, these data clearly demonstrate for the first time a correlation between IKK-2 target modulation and efficacy in vivo and validate the anti-inflammatory potential of an IKK-2 inhibitor.

PHA-408 Is Well Tolerated at Efficacious Doses in Rats. Female Lewis rats (n = 6/group) were exposed to 0, 9, 30, 45, or 60 mg/kg/day PHA-408 by oral gavage in a 2-week safety study designed to determine whether the IKK-2 target could be fully inhibited in the absence of significant adverse effects. A t.i.d. dosing regimen was used to maintain linear dose to exposure ratio, to limit the peak to trough ratio of PHA-408 in the plasma, and to avoid potential adverse effects related to C_max. Target doses were chosen based on the ED₅₀ (9 mg/kg/day) and ED₈₀ (30 mg/kg/day) exposure values, plus small multiples, which had been determined in rat SCW efficacy studies. Findings included a dose-dependent decrease in body weight (∼10% in the two highest dose groups), which occurred through day 7 (Table 1). However, all dose groups resumed normal weight gain after day 7. Thymus weights decreased approximately 30% at the high dose, but there were no changes in liver, kidney, spleen, or brain weights. Hematology results revealed neutrophilia at the two highest doses, both in absolute numbers (2.5- and 5-fold increase, respectively) and percentage, whereas the percentage of lymphocytes decreased correspondingly. It is interesting that liver enzymes (alanine aminotransferase, aspartate aminotransferase, glutamate dehydrogenase, and sorbitol dehydrogenase) decreased in a dose-dependent manner. For example, alanine aminotransferase values dropped from a control level of approximately 80 to 30 U/l in the high dose group. However, because the inhibitor decreased rather than increased liver enzymes these results were not considered an adverse effect. Microscopic examination of tissues revealed mild chronic-active inflammation in the liver, consistent with neutrophilia, and lymphoid depletion in several tissues (thymus, spleen, lymph nodes). At the two lowest doses (9 and 30 mg/kg/day), no adverse effects were observed. The no observed adverse effect level was determined to be 30 mg/kg/day. Because 30 mg/kg/day is the ED₈₀ for IKK-2 inhibition, these results demonstrate that full inhibition of the target can be achieved in the absence of adverse effects and provide a measure of confidence that, in the absence of chemotype-related safety liabilities, IKK-2 can be successfully targeted by small-molecule inhibitors.

Discussion

IKK-2 has been identified as the primary kinase responsible for NF-κB activation by cytokines, microbial pattern molecules, and other immune activators (Baeuerle and Henkel, 1994). The role of IKK-2 in inflammatory diseases has been demonstrated in multiple disease relevant cells and animal models using selective IKK-2 inhibitors and genetic approaches, including gene knockout or transfer (Miagkov et al., 1998; McIntyre et al., 2003; Podolin et al., 2005; Ziegelbauer et al., 2005; Schopf et al., 2006; Tas et al., 2006; Strnad and Burke, 2007). However, targeting IKK-2 has also been clouded by safety concerns such as liver toxicity, which stems from the phenotype of IKK-2 knockout mice (Beg et al., 1995; Li et al., 1999b; Rudolph et al., 2000). Nevertheless, IKK-2 remains an attractive target for therapeutic intervention, with most of the advanced strategies being based on the development of ATP-competitive inhibitors (Ziegelbauer et al., 2005; Wen et al., 2006). In this study, we describe one of the most potent ATP-competitive inhibitors of IKK-2, PHA-408. In contrast to the previously reported ATP-competitive, -reversible, and -selective IKK-2 inhibitors (Kishore et al., 2003; Ziegelbauer et al., 2005; Wen et al., 2006), PHA-408 is unique with respect to the following points. It binds IKK-2 tightly with slow off-rate kinetics and a dissociation t½ of approximately 2 h. It inhibits IKK-2 with equal potency in living cells and cell-free systems. Furthermore, it is one of the most highly selective IKK-2 inhibitors so far discovered (e.g., >350-fold selectivity against IKK-1). With this profile, PHA-408 should spare IKK-1 in inflammatory conditions where IKK-2 is preferentially activated. However, it is still valuable to evaluate the selectivity of this inhibitor in the alternative NF-κB pathway, which is believed to be IKK-1 dependent. However, this interesting topic is complex (Mills et al., 2007) and, therefore, beyond the scope of this study. Of the several targets that were screened PHA-408 only inhibited PIM1 with a 15-fold selectivity, although the rank order of PHA-408 selectivity (IKK2 > PIM1 > PIM2 > p70S6K > TRKA > IKK-1) is not well understood. Overall, from the standpoint of safety liability, PHA-408 exhibits a desirable selectivity profile for a potential therapeutic pharmacophore.

The slow dissociation kinetics of PHA-408 was demonstrated using multiple approaches, including dissociation rate, gel filtration, dialysis, immunoprecipitation, and cell-based assays. More to the point, in IL-1β- or LPS-stimulated cells, PHA-408 bound tightly to IKK-2 and caused prolonged inhibition of the enzyme activity over 4 h. Another plausible explanation for the prolonged inhibition of IKK-2 activity would be a sustained blockade by PHA-408 of the required upstream kinase(s) that phosphorylate(s) IKK-2 in the activation loop. For example, BAY 11-7082, a putative NF-κB inhibitor, blocked the activity of the IKK complex immunoprecipitated from IL-1β-stimulated cells, and the phosphorylation induced gel supershift of IKK-2 (Kishore et al., 2003). This possibility was ruled out based on the findings that PHA-408 did not block either IL-1β- or LPS-induced phosphorylation of IKK-2 as measured by Western analysis. Alternatively, given the long half-life of the PHA-408-rhIKK-2 complex, the duration of PHA-408-inhibitory effects in cells may be regulated by the turnover of the endogenous IKK-2. Although the mechanism(s) responsible for the prolonged inhibition remain not understood completely, the data clearly show that PHA-408 binds tightly to IKK-2. This property of PHA-408 allowed us to correlate IKK-2 activity and efficacy endpoints in disease relevant cells and in acute and chronic models of inflammation. These findings are novel because in the previous reports utilizing rapidly reversible inhibitors (Burke et al., 2003; McIntyre et al., 2003; Podolin et al., 2005; Ziegelbauer et al., 2005; Schopf et al., 2006; Wen et al., 2006; Izmailova et al., 2007), the assessment of the inhibition of endogenous IKK-2 activity was an indirect readout, relying on the inhibition of efficacy endpoints such as p65 or IκB phosphorylation, IκBα stabilization, gene expression, and/or imaging and histopathological analyses. The ability of PHA-408 to bind tightly to both the inactive and active forms of IKK-2 may explain why, in addition to its excellent pharmaceutical properties, PHA-408 was equally potent against IKK-2 in cell-free and living cell systems. Binding of PHA-408 to the inactive enzyme is a valuable attribute because an ATP-competitive inhibitor that only binds the activated enzyme may face a stronger competition by the increased intracellular ATP concentrations that occur in the activated cell state.

There are multiple points of NF-κB regulation described in the literature, including IκB phosphorylation and degradation and subsequent nuclear translocation of p65 and post-translational modifications of p65 (Ghosh and Karin, 2002). PHA-408 completely blocked IκB and p65 phosphorylation and significantly stabilized IκBα levels after IL-1β or LPS stimulation, similar to observations made using other selective IKK-2 inhibitors (Ziegelbauer et al., 2005; Wen et al., 2006). In the rat LPS model and after extended cell stimulation, PHA-408 appeared to partially stabilize IκB. The interpretation of this discrepancy may be complicated because of the technology used to assess IκB levels, the diverse cellular proteasomal activities, and/or the rapid IκBα nuclear-cytoplasm shuttling (Ding et al., 1998). Despite the incomplete stabilization of IκBα, PHA-408 was efficacious in inhibiting the production of TNF-α with the same efficacy as dexamethasone and the selective p38 MAPK inhibitor, SC-806. These data suggest that IκBα degradation is only one of the mechanisms by which IKK-2 regulates the expression of NFκB targets in these models.

Besides phosphorylating IκBα, IKK-1 and IKK-2 have been shown to phosphorylate p65 on Ser536 in vitro (Kishore et al., 2003), upon cell stimulation by cytokines and LPS (Sakurai et al., 1999; Jiang et al., 2003; Yang et al., 2003), and in response to T cell activation (Mattioli et al., 2004). However, the role of IKK-2 in mediating p65 transcriptional activation is not well understood (Sakurai et al., 1999). The absolute requirement of IKK-2 for LPS-induced phosphorylation of p65 on Ser536 and increased p65 transcriptional activity was demonstrated using knockout cells (Yang et al., 2003). Other investigators have used NF-κB inhibitors such as Bay 11-7082 and 5-aminosalicylic acid to show inhibition of p65 phosphorylation on Ser536 (Sakurai et al., 1999). In agreement with these reports, using high-affinity and selective proprietary p65 antibodies, we found that IL-1β and LPS stimulation of RASFs and PBMCs, respectively, resulted in a time-dependent increase in IKK-2 kinase activity with concomitant p65 Ser536 phosphorylation. It is important that these effects were all dose-dependently inhibited by PHA-408.

From a drug development prospective: 1) we used pharmacokinetics, dynamics, and metabolism modeling tools to define the optimal chronic dosing regimens for PHA-408. We demonstrated that t.i.d dosing of PHA-408 resulted in its linear absorption, while minimizing the peak to trough ratio of PHA-408 in the plasma. Under these conditions, we showed that PHA-408 inhibited joint pathology in a chronic model of arthritis and that the time course inhibition of IKK-2 activity inversely correlated with PHA-408 plasma concentrations. 2) We demonstrated that the efficacy in arthritis was driven by maintaining plasma concentrations of PHA-408 equal or above the IC₅₀ of IKK-2 inhibition at trough (C_min). By using the t.i.d. dosing regimen, we showed that there were no adverse effects at steady-state plasma concentrations that inhibited IKK-2 at least 80% (EC₈₀). These data suggest that in the absence of chemotype-related safety liabilities, IKK-2 is a viable target for therapeutic intervention. 3) We showed that PHA-408 inhibited LPS-stimulated TNF-α production with the same potency in PBMCs, HWB, and in vivo in rats (after correction for PHA-408-free fraction). Furthermore, PHA-408-mediated inhibition of LPS-induced IKK-2 activity, p65 phosphorylation, and TNF-α production was highly correlated in HWB and in the rat model. These data linking inhibitor exposure with potency or efficacy demonstrate the potential translatability of IKK-2 activity and p-p65 readouts in bridging preclinical models to the clinic.

In summary, PHA-408 is a novel, potent, highly selective, and tight binding inhibitor of IKK-2. It is an ideal tool agent to study the relationships between IKK-2 and its substrates and downstream targets in models of inflammation.