Activation of the p38 kinase pathway in immune cells leads to the transcriptional and translational regulation of proinflammatory cytokines. Mitogen-activated protein kinase-activated protein kinase 2 (MK2), a direct downstream substrate of p38 kinase, regulates lipopolysaccharide (LPS)-stimulated tumor necrosis factor α (TNFα) and interleukin-6 (IL-6) production through modulating the stability and translation of these mRNAs. Developing small-molecule inhibitors of MK2 may yield anti-inflammatory efficacy with a different safety profile relative to p38 kinase inhibitors. This article describes the pharmacologic properties of a benzothiophene MK2 inhibitor, PF-3644022 [(10R)-10-methyl-3-(6-methylpyridin-3-yl)-9,10,11,12-tetrahydro-8H-[1,4]diazepino[5′,6′:4,5]thieno[3,2-f]quinolin-8-one]. PF-3644022 is a potent freely reversible ATP-competitive compound that inhibits MK2 activity (Ki = 3 nM) with good selectivity when profiled against 200 human kinases. In the human U937 monocytic cell line or peripheral blood mononuclear cells, PF-3644022 potently inhibits TNFα production with similar activity (IC50 = 160 nM). PF-3644022 blocks TNFα and IL-6 production in LPS-stimulated human whole blood with IC50 values of 1.6 and 10.3 μM, respectively. Inhibition of TNFα in U937 cells and blood correlates closely with inhibition of phospho-heat shock protein 27, a target biomarker of MK2 activity. PF-3644022 displays good pharmacokinetic parameters in rats and is orally efficacious in both the rat acute LPS-induced TNFα model and the chronic streptococcal cell wall-induced arthritis model. Dose-dependent inhibition of TNFα production in the acute model and inhibition of paw swelling in the chronic model is observed with ED50 values of 6.9 and 20 mg/kg, respectively. PF-3644022 efficacy in the chronic inflammation model is strongly correlated with maintaining a Cmin higher than the EC50 measured in the rat LPS-induced TNFα model.
Rheumatoid arthritis (RA) is a chronic inflammatory disease characterized by an imbalance of proinflammatory and anti-inflammatory cytokines, autoimmunity, joint inflammation, and eventual joint destruction (McInnes and Schett, 2007). Evidence supporting the role of the proinflammatory cytokines TNFα, IL-1β, and IL-6 has been demonstrated in both animal models and human clinical trials (Dayer et al., 2001; Scott and Kingsley, 2006; Hennigan and Kavanaugh, 2008). Use of biologic therapeutics that neutralize these cytokines has shown some clinical success in reducing joint pain and inflammation while retarding joint destruction (Smolen and Steiner, 2003). Limits to the use of biologics in RA include the high cost of protein pharmaceuticals, parenteral administration, loss of efficacy over time, risk of infection, and a significant portion of patients who show partial or no response to these agents. Therefore, development of orally active small-molecule inhibitors that target signaling pathways regulating inflammatory cytokine production could add significant value to unmet medical need.
In 1994, p38α kinase was identified as a target that regulates inflammatory cytokine biosynthesis (Lee et al., 1994). Since then, kinases have been hotly pursued as drugable targets that regulate inflammation signaling pathways (Gaestel et al., 2007, 2009). Activation of the p38 kinase pathway in immune cells leads to the transcriptional and translational regulation of proinflammatory cytokine synthesis (Winzen et al., 1999). Many p38 kinase inhibitors have subsequently been developed that demonstrate inhibition of TNFα, IL-1β, and IL-6 production and display anti-inflammatory efficacy in animal models (Pettus and Wurz, 2008). In clinical trials, many p38 kinase inhibitors were discontinued because of unacceptable safety profiles, namely elevated liver enzymes with significant incidence of skin rash (Dominguez et al., 2005). Of the compounds that advanced to testing in RA patients, initial short-term efficacy was observed, but was subsequently lost with further treatment presumably because of feedback control of the p38 kinase network (Hammaker and Firestein, 2010). It is possible, therefore, that other targets upstream or downstream in the p38 kinase pathway may avoid this feedback loop while exhibiting enhanced safety.
Activated p38 kinase directly phosphorylates and activates the mitogen-activated protein kinase-activated protein (MAPKAP) kinases MK2, MK3, and MK5 (also known as PRAK) (Gaestel, 2006). Before 1999, it was unclear which downstream p38 kinase pathway components regulated TNFα production. In MK2(−/−) knockout mice, TNFα levels were reduced approximately 90% through a post-transcriptional mechanism, demonstrating that MK2 is essential for lipopolysachharide (LPS)-induced TNFα biosynthesis (Kotlyarov et al., 1999). IL-6 and interferon-γ were also significantly reduced, whereas IL-1 showed only modest reduction (Kotlyarov et al., 1999). MK2 regulates LPS-stimulated TNFα and IL-6 production through modulating the stability and translation of TNF and IL-6 mRNAs via AU-rich elements in the 3′ untranslated region (Neininger et al., 2002). Interestingly, MK3(−/−) mice showed little reduction in LPS-stimulated TNFα levels, whereas the MK2(−/−)MK3(−/−) double knockout mouse exhibited complete inhibition, indicating that MK2 is the major MAPKAP kinase regulating TNFα production (Ronkina et al., 2007). No effect on TNFα synthesis or other cytokines was observed in PRAK(−/−) mice (Shi et al., 2003). MK2(−/−) mice were also resistant to collagen-induced arthritis in a murine model of RA (Hegen et al., 2006). MK2, therefore, is an attractive target for development of anti-inflammatory kinase inhibitors.
The MK2 knockout mouse has been useful in defining MK2's role in inflammation. Having a potent and selective MK2 kinase inhibitor as an investigative tool, however, would be advantageous for further exploring the biology of MK2 and the p38 kinase pathway. Potent MK2 inhibitors have been recently described (Anderson et al., 2005, 2007, 2009a,b; Trujillo et al., 2007; Wu et al., 2007; Goldberg et al., 2008; Schlapbach et al., 2008; Xiong et al., 2008; Keminer et al., 2009), but few show nanomolar potency in cells (Schlapbach et al., 2008; Anderson et al., 2009b). Developing potent, selective MK2 inhibitors that have optimized pharmacologic properties for activity in blood or in vivo has been extremely difficult, with just one compound from the pyrrolopyridine series reported to have oral efficacy in blocking TNFα production in LPS-challenged rats (Anderson et al., 2007).
PF-3644022 [(10R)-10-methyl-3-(6-methylpyridin-3-yl)-9, 10,11,12-tetrahydro-8H-[1,4]diazepino[5′,6′:4,5]thieno[3,2-f] quinolin-8-one] represents a potent and selective benzothiophene MK2 inhibitor, the first MK2 inhibitor described with oral efficacy in both acute and chronic models of inflammation. This ATP-competitive compound potently inhibits MK2 enzyme activity with good selectivity across 200 human kinases. PF-3644022 potently inhibits LPS-stimulated TNFα production in cells and blood and when dosed orally in LPS-challenged rats. PF-3644022 exhibits good pharmacokinetic properties, demonstrating efficacy in the streptococcal cell wall (SCW)-induced arthritis model.
Materials and Methods
Preparation of PF-3644022.
PF-3644022 was prepared by the Pfizer Discovery Medicinal Chemistry Department (Chesterfield, MO) as described (Anderson et al., 2009b). The molecular weight and formula of the parent compound are 374.47 g/mol and C21H18N4OS. The free base form was used for all dosing studies and calculation of solution concentration. Fresh 10 mM PF-3644022 stock concentrations were made in 100% dimethyl sulfoxide (DMSO) to support enzyme and cell studies and kept at room temperature for no more than 2 weeks.
Generation of Recombinant Protein Kinases.
MAPKAP kinase family recombinant proteins were generated in-house. N-terminally truncated MK2 (amino acids 45–400) was expressed and purified as described (Schindler et al., 2002). To support compound binding studies to MK2, N-terminally biotinylated MK2 (amino acids 45–371) was expressed in Escherichia coli by using the BirA expression system (Smith et al., 1998). MK3 (GenBank accession no. U43784), PRAK (GenBank accession no. AF032437), and MNK1 (GenBank accession no. AB000409) were expressed in E. coli as glutathione S-transferase fusion proteins and affinity-purified over glutathione-Sepharose (GE Healthcare, Little Chalfont, Buckinghamshire, UK). The glutathione S-transferase purification tag was removed by thrombin cleavage, and the proteins were further purified to homogeneity over Mono Q-sepharaose (GE Healthcare). The C-terminal kinase domains of MSK1 (amino acids 367–802) (GenBank accession no. AF074393) and MSK2 (amino acids 351–772) (GenBank accession no. AJ010119) were expressed as N-terminal 6×His-tagged fusion proteins in baculovirus-infected Sf9 insect cells. MSK1 and MSK2 were purified to near homogeneity over Ni-NTA-agarose (QIAGEN, Valencia, CA).
In Vitro Kinase Assays and Inhibition Kinetics.
Activated p38α prepared according to Hope et al. (2009) was used to activate recombinant MAPKAP kinases by incubation at a 1:50 M ratio with 250 μM ATP for 1 h at 30°C. The kinase activity of MK2 was followed by using fluorescently labeled heat shock protein 27 (HSP27) peptide (fluorescein isothiocyanate-KKKALSRQLSVAA) and a Caliper LabChip 3000 (Caliper Life Sciences, Hopkinton, MA). Phosphorylated peptide was separated from substrate peptide electrophoretically and quantified. All kinase reactions were performed at room temperature in 20 mM HEPES, containing 10 mM MgCl2, 1 mM dithiothreitol, 0.01% bovine serum albumin, and 0.0005% Tween 20, pH 7.5. Unless specified otherwise, the reactions were initiated by the addition of enzyme. For endpoint experiments, reactions were terminated during the linear phase by the addition of 30 mM EDTA. The kinase selectivity experiments were performed with the MgATP concentration fixed at the Km(app) determined for each enzyme.
To determine the mechanism of action of PF-3644022 binding, the initial velocities in the presence and absence of PF-3644022 with ATP as the varied substrate while the HSP27 peptide concentration was held constant. The data were fit to the competitive inhibition model (eq. 1), noncompetitive inhibition model (eq. 2), or an uncompetitive inhibition model (eq. 3). In these equations, Vmax is the maximum velocity, Km is the Michaelis–Menton constant for the varied substrate, S is the concentration of the varied substrate, I is the concentration of the inhibitor, and Kis and Kii are the slope and intercept inhibition constants, respectively. The best fit was based on an F test and resulted in the lowest standard errors for the inhibition constants. The apparent inhibition constants (Ki) were determined with GraFit 5.0 (Erithacus Software, Horley, Surrey, UK). Other MAPKAP enzymes were assayed for activity by using an ion exchange separation method for the detection of 33P-labeled product peptide as described (Anderson et al., 2007). PF-3644022 was evaluated for inhibition of 200 human kinases by using an in-house 30 kinase selectivity panel (Card et al., 2009) and 170 kinases from the Upstate Kinase Profiler service (Millipore Corporation, Billerica, MA). The kinase selectivity experiments were performed with the MgATP concentration fixed at the Km(app) determined for each enzyme.
MK2 Inhibitor Binding Studies.
Surface plasmon resonance spectroscopy using standard Biacore methodology on a Biacore 3000 instrument (GE Healthcare) was used to follow real-time binding kinetics of PF-3644022 to immobilized biotinylated MK2. The binding studies were performed at 25°C by using a running buffer of 10 mM HEPES, 150 mM NaCl, 0.005% P20 detergent with a flow rate of 60 μl/min. PF-3644022 (1–100 nM) was injected over immobilized MK2 for 4 min and then dissociation followed for 15 min. For biotin-MK2 capture, a CM5 chip (GE Healthcare) was first preconditioned and then streptavidin was immobilized by using amine coupling with N-hydroxysuccinimide and N-ethyl-N′-(3-dimethylaminopropyl) carbodiimide according to Biacore methods. For kinetic analyses, sensorgrams were double reference-subtracted, and BiaEvaluation (Biacore) was used to determine the association and dissociation constants by using a Langmuir 1:1 model.
The U937 human premonocytic cell line was obtained from the American Type Culture Collection (Manassas, VA) and differentiated to a monocyte/macrophage phenotype with phorbol myristate acetate (Sigma-Aldrich, St. Louis, MO) as described (Burnette et al., 2009). Human peripheral blood mononuclear cells (hPBMCs) were prepared from venous blood of donors collected anonymously with informed consent at an on-site clinic. Venous blood was collected into sodium heparin tubes, and hPBMCs were isolated by density gradient centrifugation using Histopaque 1077 (Sigma-Aldrich) per the manufacturer's directions. U937 cells and hPBMCs were cultured as described (Burnette et al., 2009; Hope et al., 2009).
The ability of PF-3644022 to inhibit LPS-stimulated cytokine production in U937 cells and hPBMCs was evaluated after a 1-h pretreatment of compound in cell culture media containing less than 1% DMSO final concentration. All cell incubations were done at 37°C. Culture media TNFα levels were measured 4 h after LPS stimulation at 100 ng/ml by using an electrochemoluminescence MesoScale Discovery TNFα kit (MesoScale Discovery, Gaithersburg, MD). In hPBMCs, TNFα, IL-1, IL-6, and IL-8 were measured at 16 h after LPS stimulation by using a four-plex human cytokine MSD plate (MesoScale Discovery). For mitogen-activated protein kinase (MAPK) signaling studies in U937 cells, cells were pretreated with PF-3644022 at varying concentrations for 1 h before LPS stimulation at 100 ng/ml for 30 min. Cell lysates were prepared and analyzed for phospho-Ser78-HSP27, phospho-p38, and phospho-c-Jun N-terminal kinase (JNK) levels by Western blot analysis as described (Anderson et al., 2007). Quantitation of Western blots was performed with Alexa-conjugated secondary antibodies (Invitrogen, Carlsbad, CA) and fluorescent LI-COR (LI-COR Biosciences, Lincoln, NE) scanning. MK2 activity in U937 cells, measured by monitoring the phosphorylation of the MK2 substrate HSP27, was quantitated in cell lysates by using a phospho-Ser82 HSP27 and total HSP27 MSD kit. Phospho-HSP27 levels were normalized in cell lysates to total HSP27 protein levels.
LPS-Stimulated Human Whole Blood.
Venous blood from human donors was collected in sodium heparin tubes (Baxter Healthcare, Deerfield, IL) and assayed for LPS-stimulated cytokine production as described (Burnette et al., 2009). In brief, PF-3644022 was added to human whole blood (HWB) ex vivo 1 h before stimulation with 100 ng/ml LPS at 37°C. TNFα was measured 4 h poststimulation by assaying plasma with a human TNFα MSD kit. TNFα, IL-1, IL-6, and IL-8 in plasma was also measured after 16-h LPS stimulation of HWB by using a four-plex human cytokine MSD plate (MesoScale Discovery). MK2 activity in HWB was measured by monitoring phospho-Ser82 HSP27 and total HSP27 levels after a 30-min stimulation of 100 ng/ml LPS by using a dissociation-enhanced lanthanide fluorescent immunoassay (PerkinElmer Life and Analytical Sciences, Waltham, MA) as described (Burnette et al., 2009).
Measurement of PF-3644022 in Plasma.
Plasma was analyzed for total PF-3644022 by high-performance liquid chromatography. In brief, calibration standards ranging from 0.27 nM to 13 μM were prepared by fortifying appropriate amounts of PF-3644022 to blank control plasma by a series of dilutions. Samples and calibration standards were briefly vortex-mixed, and 0.025-ml aliquots were transferred from the vials into corresponding 96-well plates. Internal standard working solution (0.25 μM tolbutamide in 97.5% methanol/2.5% acetonitrile containing 1% formic acid) was then added as a 0.225-ml aliquot to all samples. The plates were centrifuged at approximately 3800 rpm for 5 min. A total of 90 μl of supernatant were transferred to a new 96-well Quadra 96–320 automatic sample handling system (Tomtec, Hamden, CA). The samples (5 μl) were injected onto the liquid chromatography-tandem mass spectrometry (LC-MS/MS) system for analysis. Samples were chromatographed with a Rheos pump (Thermo Fisher Scientific, Waltham, MA) and an Extend C18 (20 × 2.1 mm, 5-μm particle size) column (Agilent Technologies, Santa Clara, CA) connected to a HTS-PAL autosampler from LEAP Technologies (Carrboro, NC). The mobile phases were 10 mM ammonium acetate in 95% water/5% methanol (A) and 10 mM ammonium acetate in methanol (B). The running condition was 70% A for 0.5 min isocratically, ramped to 100% B in 1 min, holding for 0.9 min followed by dropping to 30% B in 0.1 min, and holding at 30% B another 0.5 min before the next injection. The total run time was 3 min. The high-performance liquid chromatography flow rate was maintained at 0.4 ml/min for the entire analysis. An API 4000 triple quadruple mass spectrometer (AB/MDS-Sciex, Concord, Canada) with a turbo-ionspray interface operated in positive ionization mode was used for the multiple reaction monitoring LC-MS/MS analyses. The mass spectrometric conditions were optimized for detection of PF-3644022 and tolbutamide. The following precursor product ion transitions were used for multiple reaction monitoring: PF-3644022: m/z 375 → 291; tolbutamide, m/z 271 → 155.
LPS-Induced TNFα Production in Rats.
All rat in vivo studies were reviewed and approved by the Pfizer Institutional Animal Care and Use Committee according to guidelines sanctioned by the Association for Assessment and Accreditation of Laboratory Animal Care International. In the LPS-induced acute endotoxemia inflammation model, adult male Lewis rats (225–250 g; Harlan, Indianapolis, IN) were fasted 18 h before oral dosing and allowed free access to water. PF-3644022 was prepared as a suspension in a vehicle consisting of 0.5% methylcellulose (Sigma-Aldrich) and 0.025% Tween 20 (Sigma-Aldrich) in water. PF-3644022 or vehicle was orally administered in a volume of 1 ml by using an 18-gauge gavage needle 4 h before LPS challenge. LPS was administered by injection into the penile vein at 1 mg/kg in 0.5 ml of sterile saline, and blood was collected 90 min later by cardiac puncture. Serum TNFα levels were measured by rat TNFα enzyme-linked immunosorbent assay (Burnette et al., 2009), and PF-3644022 levels were determined by LC-MS/MS.
Streptococcal Cell Wall-Induced Arthritis in Rats.
Arthritis was induced in 125- to 140-g female Lewis rats (Harlan) by a single intraperitoneal injection of peptidoglycan–polysaccharide complexes from group A SCW purchased from Lee Laboratories (Grayson, GA) as described (Hope et al., 2009). The disease course was biphasic with an acute inflammatory non-T cell-dependent phase on days 1 to 3 followed by a chronic T cell-dependent inflammatory-erosive arthritis developing over days 14 to 28. Animals developing the acute inflammatory phase were pooled into groups of seven to eight animals per group and dosed with PF-3644022 or methylcellulose-Tween 20 vehicle by oral gavage (1 ml) twice a day for days 10 to 21. Hind paw swelling volumes were measured on day 21 with a displacement plethysmometer. After the last PF-3644022 dose on day 21, plasma was collected at various times up to 11 h to determine compound exposure parameters.
Determination of In Vivo Rat Pharmacokinetic Parameters of PF-3644022.
Male Sprague-Dawley rats weighing 275 to 300 g were purchased from Charles River Laboratories Inc. (Wilmington, MA) and acclimated to their surroundings for approximately 1 week with food and water provided ad libitum. A minimum of 1 day before study, animals were anesthetized with isoflurane (to effect) and then implanted with Culex (BASi, West Lafayette, IN) vascular catheters in the carotid artery. Animals were acclimated in Culex cages overnight before dosing. Patency of the carotid artery catheter was maintained by using the “tend” function of Culex ABS. PF-3644022 was administered dissolved in 70% normal saline/20% polyethylene glycol-400/10% ethanol (intravenously) or suspended in 0.5% hydroxypropylmethylcellulose/0.1% Tween 80 in distilled water (by mouth). Blood collections were obtained from the carotid artery and performed by the Culex at 2 (intravenous only), 5, 15, 30 min and 1, 2, 4, 6, 8, 12, 18, and 24 h. Plasma was separated and frozen for analysis.
Concentrations below the limit of quantitation were reported as zero and used in the evaluation of mean concentrations and the estimation of AUC. The peak plasma concentration (Cmax) and the time to reach peak concentration (Tmax) were recorded directly from individual plasma concentration-time profiles. The terminal log-linear phase of the plasma concentration-time curve was identified by linear regression of data points, which yielded an R2 value. The terminal half-life (t1/2) was calculated as ln(2) divided by absolute value of the slope of the terminal log-linear phase. The area under the plasma concentration-time curve from time 0 to time of the last quantifiable concentration (t) (AUC0-t) was determined by using the linear trapezoidal method. The area under the plasma concentration-time curve from time 0 to infinity (AUC0-∞) was determined as AUC0-t plus the extrapolated area. The extrapolated area was determined by dividing the last observed plasma concentration by the slope of the terminal log-linear phase. The initial plasma concentrations (C0) after intravenous dosing were extrapolated from the apparent distribution phase for individual animals after intravenous administration. Systemic plasma clearance (CLp) was calculated as dose/AUC0-∞, and the volume of distribution at steady state (Vdss) was calculated as CL × mean residence time, which was defined as AUMC0-∞/AUC0-∞. The absolute oral bioavailability (F) was then calculated as a ratio of the mean dose-normalized AUC (0-t or 0-∞) after PO administration to the mean dose-normalized AUC (0-t or 0-∞) after intravenous administration. The extent of binding of PF-3644022 to rat and human plasma was determined in vitro by using an ultracentrifugation method as described (Burnette et al., 2009).
The MK2 Inhibitor PF-3644022 Is Competitive with ATP and Displays Good Selectivity across the Human Kinome.
Compounds from the benzothiophene series were identified in a Pfizer compound library screen showing both potent MK2 inhibitory activity and good cellular potency at blocking LPS-stimulated TNFα production (Anderson et al., 2009a). Modifying the hinge binding element of the scaffold and adding selectivity-promoting substituents led to further increases in MK2 enzyme potency and cellular activity, while enhancing kinase selectivity and oral bioavailability (Anderson et al., 2009b). The structure of our lead compound, PF-3644022, is shown in Fig. 1. PF-3644022 in DMSO stock solutions was shown to lose activity over time. Analysis by LC-MS/MS showed that the compound contained significant oxidation of the thiophene ring by 4 weeks in DMSO at room temperature (data not shown). PF-3644022, however, was stable in aqueous solutions with undetectable levels of oxidation after 4 to 8 weeks. Therefore, PF-3644022 was solubilized in DMSO and used immediately to support analytical, enzyme, and cell studies.
PF-3644022 inhibits recombinant MK2 activity with an IC50 value of 5.2 nM (Table 1). Additional characterization of PF-3644022 showed that it is competitive versus MgATP with a competitive inhibition constant of 3.0 nM (Fig. 2A and Table 1). The competitive inhibition versus MgATP is apparent in the plot (Fig. 2B) that shows a family of lines intersecting on the 1/v axis, indicative of competitive inhibition. Given that a number of studies have demonstrated that the structure of the ATP binding sites of protein kinases can be affected by the phosphorylation state, we also determined the binding kinetics of PF-3644022 against the nonphosphorylated form of MK2 by using surface plasmon resonance. Rapid association and dissociation kinetics were observed for PF-3644022 binding to nonphoposphorlyted MK2 (Table 1). Taken together, these rate constants provided a Kd value of 5.9 nM, which is similar the competitive inhibition constant determined with the phosphorylated form of MK2. Enzyme kinetic studies and crystallographic analyses performed with MK2 support the idea that PF-3644022 binds in the MK2 ATP pocket (Anderson et al., 2009a,b).
The inhibitory activity of PF-3644022 against other MAPKAP kinase family members was evaluated as described under Materials and Methods. At Km levels of ATP for each kinase, PRAK was inhibited with equivalent potency as MK2, whereas close family member MK3 had an IC50 of 53 nM, approximately 10-fold weaker (Table 1). Other than MNK2 with an IC50 of 148 nM, other family members were largely not inhibited, showing at least several hundred-fold selectivity versus MK2. Although PF-3644022 is an ATP-competitive kinase inhibitor, it displayed good selectivity against a diverse panel of 200 human kinases tested at Km levels of ATP (Supplemental Table 1). Of the 200 kinases tested for percentage of inhibition with 1 μM PF-3644022, 16 kinases showed more than 50% inhibition. These kinases were further profiled to generate IC50 values (Table 2). Of the 16 kinases examined for IC50 values, 13 kinases were less than 100-fold selective, and nine kinases showed less than 25-fold selectivity versus MK2. Ten of the 16 kinases that PF-3644022 inhibited were, not surprisingly, members of the CAMK group where MK2 resides. One or two other kinases representing the CMGC, STE, TK, and TKL groups were also inhibited by PF-3644022 (Table 2). More importantly, excluding MK3 and MNK2, which may contribute a minor component to TNFα production, other potential TNFα-regulating kinase targets such as extracellular signal-regulated kinases (ERKs), IκB kinases, JNKs, MAPK/ERK kinases, MAP kinase kinases, and p38α/β were not significantly inhibited by PF-3644022 (Supplemental Table 1).
PF-3644022 Blocks LPS-Stimulated TNFα Production in Cells and Whole Blood.
TNFα production induced by LPS stimulation in the U937 monocytic cell line was measured 4 h after LPS addition, which was previously shown to coincide with peak TNFα levels in cell culture. PF-3644022 blocked LPS-induced TNFα production with an IC50 of 159 nM (Table 3 and Fig. 3). The effect of PF-3644022 on phosphorylation of MAPK pathway members and MK2 activity measured by phosphorylation of HSP27 was quantitated in U937 cell lysates generated after stimulating cells with LPS for 30 min. PF-3644022 had little effect on blocking p38α or JNK1/2 phosphorylation while inhibiting phospho-Ser78-HSP27 levels with a similar concentration response as TNFα production (Fig. 3A). ERK and c-Jun phosphorylation levels were also unchanged by PF-3644022 treatment (data not shown). Inhibition of MK2 activity as measured by inhibition of phospho-Ser82 HSP27 levels indicated an IC50 of 201 nM (Table 3). These data show that TNFα inhibition in LPS-stimulated U937 cells correlates with inhibition of MK2 activity and not inhibition of other MAPK pathway targets implicated in TNFα production.
In addition to the U937 cell line, PF-3644022 inhibits TNFα production in hPBMCs and HWB. Inhibition by PF-3644022 in hPBMCs was similar to U937 cells (Fig. 3B) with an IC50 of 160 nM (Table 3). Of note, the cellular inhibition of TNFα production and MK2 activity was approximately 30-fold weaker than in the MK2 enzyme assay, perhaps because of competition by much higher cellular ATP levels. TNFα inhibition in LPS-stimulated HWB was further right-shifted with an IC50 of 2 μM, which correlated nicely with inhibition of MK2 activity measured in HWB lysates (Table 3). The lower potency of PF-3644022 in HWB can be rationalized by factoring in the measured human plasma protein binding of 93.6% (data not shown), resulting in an unbound concentration or free fraction (ff) IC50 of 126 nM, similar to the PF-3644022 potency in U937 and hPBMCs. Furthermore, the potency of PF-3644022 in cell culture can be right-shifted with increasing amounts of serum (data not shown). The effect of PF-3644022 on other proinflammatory cytokines IL-1β and IL-6 and the chemokine IL-8 was explored in LPS-stimulated hPBMCs and HWB (Fig. 4). Inhibition of IL-1β, IL-6, and IL-8 levels by PF-3644022 in hPBMCs was observed (Fig. 4A), albeit with approximately 10-fold weaker IC50 values of 1 to 2 μM relative to TNFα inhibition (Table 3). Although the concentration-response curves had steep slopes between 1 and 5 μM, no cell toxicity was detected up to 20 μM PF-3644022 (data not shown). The inhibition of LPS-stimulated IL-6 production by PF-3644022 in HWB was similarly 10-fold right-shifted as in hPBMCs, with an IC50 of 10 μM (Fig. 4B and Table 3). Of note, in HWB IL-8 and IL-1β were weakly inhibited by PF-3644022 (Fig. 4B).
PF-3644022 Inhibits TNFα Production in the Acute LPS-Challenged Rat Model.
Although many benzothiophene inhibitors exhibit submicromolar potency in cells, very few have sufficient pharmacokinetic properties to support in vivo evaluation or activity. The pharmacokinetic parameters measured for PF-3644022 in rats are shown in Table 4. PF-3644022 shows good oral bioavailability and is rapidly absorbed in suspension dosing with a Tmax of 0.17 h, a good volume of distribution, and terminal half-life of 9.1 h. Clearance of PF-3644022 from circulating rat blood is also reasonably low. The exposure and pharmacokinetic parameters of PF-3644022 at a suspension dose of 3 mg/kg suggests that PF-3644022 may have sufficient properties to support in vivo evaluation of anti-inflammatory efficacy in rat models.
The oral efficacy of PF-3644022 was evaluated in the acute LPS-challenged rat model. Single doses of PF-3644022 ranging from 0.2 to 60 mg/kg were given to Lewis rats as an oral suspension in methacellulose-Tween vehicle 4 h before LPS challenge. Intravenous administration of LPS to Lewis rats produced a rapid and transient elevation of TNFα levels in plasma that peaked 1 to 2 h after LPS injection. TNFα and compound levels in the rat plasma were measured 90 min after LPS injection. Oral dosing of PF-3644022 yielded a nice dose-dependent response (Fig. 5A) for TNFα inhibition with an ED50 and ED80 of 6.9 and 19 mg/kg, respectively (Table 5), which corresponds to an EC50 and EC80 of 1.4 and 4.2 μM total concentration of PF-3644022, respectively, 5.5 h after an initial oral dose (Fig. 5B and Table 5). The EC50 in the rat acute LPS challenge model was similar to PF-3644022 activity in HWB (Table 3), and when corrected for rat plasma protein binding (92.1%), the unbound free fraction EC50 was 110 nM (Table 5), again correlating well with TNFα inhibition in U937 cells and hPBMCs (Table 3). The pharmacokinetic and pharmacodynamic (PK-PD) response of TNFα inhibition in the endotoxin-stimulated rat model was examined up to 36 h after a single ED80 oral dose of PF-3644022 (Fig. 5C). For each time point when plasma PF-3644022 levels were measured, that group of rats was injected with LPS 90 min before blood collection. A nice mirrored response was seen between PF-3644022 and TNFα levels over time, with minimal TNFα production observed at maximal PF-3644022 plasma levels. Maximal TNFα response returned as compound were cleared from the blood, demonstrating an immediate and reversible pharmacodynamic response to MK2 inhibition in this model.
PF-3644022 Suppresses Chronic Inflammation in the Streptococcal Cell Wall-Induced Arthritis Rat Model.
The rat SCW model is characterized by a biphasic inflammation response with an acute phase on days 1 to 5 followed by a more severe and chronic inflammation phase from days 10 to 21. In the acute phase, hemorrhage and fibrin deposition in the joint synovial space occurs with the accumulation of activated macrophages in the soft tissue. In the more severe chronic phase, intense cell infiltration, joint inflammation, and bone destruction in the rat paw is observed. TNFα and IL-1β play a role in the disease process as neutralizing antibodies to these cytokines show the ability to attenuate the disease (Kuiper et al., 1998). PF-3644022 showed dose-dependent inhibition of chronic paw swelling measured on day 21 after 12 days of oral dosing b.i.d. (Fig. 6A). The observed ED50 for PF-3644022 was 20 mg/kg, whereas the ED80 could be estimated only to be 50 to 100 mg/kg because maximal efficacy for MK2 inhibition appeared to plateau at approximately 80% inhibition in this model (Table 5). At calculated EC50 and EC80 PF-3644022 concentrations in rat plasma, the mean Cmax and Cmin values were 11 and 0.91 μM, respectively (Table 5). Interestingly, the Cmin at EC50 in the chronic rat SCW model was similar to the EC50 for TNFα inhibition in the acute rat LPS model, suggesting that efficacy may be driven by Cmin. This finding was also reinforced when evaluating the exposure-time response generated at various doses in rat SCW experiment 3 on day 21 (Fig. 6B). At efficacious doses generating more than 50% inhibition of paw swelling, the Cmin must exceed the rat LPS EC50 for TNFα inhibition for nearly the entire dosing period. In other words, at least 50% of MK2 activity must be inhibited at all times for an MK2 inhibitor to be efficacious in the rat SCW chronic arthritis inflammation model.
The Pharmacologic Profile of the MK2 Inhibitor PF-3644022 Strongly Links TNFα Inhibition with Efficacy in Acute and Chronic Models of Inflammation.
The potency and efficacy of PF-3644022 was compared across in vitro and in vivo assays, including recombinant human MK2 activity, LPS-induced TNFα production in U937 cells, hPBMCs, HWB, and efficacy in rat acute and chronic models of inflammation. When inhibition was plotted against PF-3644022 concentrations, using unbound free fraction plasma levels in the case of ex vivo or in vivo studies, there was very good correlation between compound concentration and efficacy (Fig. 7). Inhibition of TNFα production in cells through direct inhibition of MK2 enzyme activity led to efficacy in rat acute and chronic models of inflammation. The composite EC50 value for these studies was 139 nM (Fig. 7). This is the minimum unbound plasma concentration needed to inhibit more than 50% cellular activity for TNFα production and may represent a Cmin needed for efficacy in chronic inflammatory disease. The close correlation observed in these assays allows an initial prediction of the exposures that may be required for efficacy in human TNFα-mediated inflammatory diseases.
When it was reported that MK2 is essential for LPS-stimulated TNFα production (Kotlyarov et al., 1999), many pharmaceutical companies who had ongoing p38 kinase programs initiated MK2 projects as alternate approaches to modulating the p38 kinase pathway. Although numerous screening campaigns were run, initial leads were sparse, highly promiscuous, and lacking in cell activity. Although several different crystal forms of MK2 were generated by us and other groups, it is extremely difficult to obtain high diffracting data sets, thus limiting the application of structure-based drug design. Of the MK2 structures that were solved, MK2 was shown to have a narrow and deep ATP-binding cleft (Anderson et al., 2007, 2009a,b; Hillig et al., 2007). Because of this narrow cleft, MK2 inhibitors need to be mostly planar to bind to the ATP pocket. Typical approaches of optimizing selectivity by appending substituents out of the binding plane were not readily applicable to MK2 inhibitors. Most kinase inhibitors bind to the ATP pocket through a nitrogen-containing ring or element that binds to the kinase hinge region and another interaction gained through hydrogen bonding with the conserved catalytic Lys or Asp of the activation loop. The nitrogen atom of the polyaromatic core forms a hydrogen bond to the hinge and the 4-methyl-3-pyridinyl group of PF-3644022 imparts selectivity against other kinases (Anderson et al., 2009b). The configuration of the stereogenic center on the lactam ring is important for potency, and the lactam carbonyl is required for interactions with the conserved catalytic Lys and activation loop Asp.
In this article we refer to PF-3644022 as a selective MK2 inhibitor. In actuality, PF-3644022 is a MK2/MK3/PRAK inhibitor. PRAK is inhibited with equal potency to MK2, whereas MK3 is approximately 10-fold weaker (Table 1). Although no structure of PRAK has been reported, the structure of MK3 recently has (Cheng et al., 2010). The structures of MK2 and MK3 superimpose well, but there are slight differences between MK2 and MK3 noted in the ATP binding pocket. MK2 Leu141 is replaced with slightly larger Met121 in MK3, whose side chain protrudes at the bottom of the adenine pocket. Perhaps this modification or others are responsible for the 10-fold weaker inhibition of MK3 activity by PF-3644022. Based on MK2(−/−), MK3(−/−), and MK2(−/−) MK3(−/−) mouse studies, inhibition of LPS-stimulated TNFα synthesis is predominantly through MK2. Even with a MK2/MK3/PRAK inhibitor, PF-3644022 will inhibit a subset of substrates in the p38 kinase pathway and may offer advantages to the more global effects of a p38 kinase inhibitor. In addition, PF-3644022 inhibits seven other CAMK superfamily members with less than 100-fold selectivity and a scattering of four other kinases across the human kinome (Table 2). None of the other kinases are implicated in TNFα production. PF-3644022 is the most selective MK2 inhibitor described to date.
Of the MK2 inhibitor chemotypes reported, few have submicromolar potency at inhibiting TNFα production in cells, perhaps because of poor physiochemical properties, poor cell permeability, poor biochemical efficiency (BE) (ratio of binding affinity to target versus cellular activity), or inadequate enzyme potency. Of several MK2 chemotypes investigated within Pfizer, only the benzothiophenes have cellular IC50 values less than 500 nM. PF-3644022 is a highly permeable and potent MK2 inhibitor (Ki = 3 nM), yet it exhibits poor BE with at least 30-fold weaker activity at inhibiting TNFα production in cells (Table 3). Given that PF-3644022 is an ATP-competitive inhibitor, the shift in cellular potency may be caused by competition with high cellular concentrations of ATP (approximately 5 mM). We have determined that the binding constant of MgATP for nonactive MK2 is 30 μM (data not shown). The high affinity of nonactive MK2 for ATP is in contrast to the very low affinity of ATP for nonactive p38 (>10 mM). The observation that most ATP-competitive p38 MAPK inhibitors bind with similar affinity to both the activated and nonactivated kinase, whereas MgATP strongly prefers the phosphorylated, active form of p38 kinase, suggests that p38 kinase inhibitors will maintain enzyme potency in cellular systems (Schindler et al., 2007).
Additional increases in MK2 inhibitor BE may be achieved through further increases in Ki, tight binding, slow off-rate kinetics, or through noncompetitive, uncompetitive, or irreversible mechanisms. We believe that the best Ki values achievable with MK2 are low nanomolar, because we were unable to achieve further potency even after gaining additional interactions in the ATP pocket. We also developed several irreversible MK2 inhibitors as tool compounds that did in fact exhibit BEs near one, but had insufficient selectivity to explore as drug leads. In several MK2 screening campaigns, noncompetitive or “allosteric” leads were never identified, and to date, none have been reported by others. In a BE analysis of 50 marketed drugs (Swinney, 2004), 76% had BEs higher than 0.4. For PF-3644022, the calculated BE is 0.03, similar to statin drugs. Swinney concluded that the lower the BE, the more drug would be required for efficacy and the lower the therapeutic index, with higher incidence of toxicity. Although the MK2 knockout mouse validated MK2 as a very attractive target for TNFα inhibition, the very low BE suggests a low probability of success developing MK2 inhibitors as drugs.
Achieving activity in whole blood with MK2 inhibitors has been very challenging, possibly because of high plasma protein binding (e.g., >98%). The measured plasma protein binding for PF-3644022 was 93.6 and 92.1% in human and rat blood, respectively, providing sufficient unbound concentrations to display activity in blood. In LPS-stimulated human monocytes or HWB, p38 kinase inhibitors show nearly equivalent potency in blocking TNFα, IL-1β, and IL-6 through regulating cytokine production at both a transcriptional and post-transcriptional level (Burnette et al., 2009; Hope et al., 2009). In LPS-stimulated hPBMCs, PF-3644022 blocks these cytokines as well, although it is approximately 10-fold more active at inhibiting TNFα (Table 3). In LPS-stimulated HWB, however, PF-3644022 predominantly inhibits TNFα, whereas IL-6 inhibition is 10-fold weaker (Table 3). Little to no inhibition of IL-1β or IL-8 is seen up to 25 μM. These results are consistent with MK2 regulating TNFα and IL-6 through a post-transcriptional mechanism, primarily through modulating the stability and translation of TNFα and IL-6 mRNA (Neininger et al., 2002). The ability of MK2 inhibitors to preferentially block TNFα, IL-6 to a lesser extent, and IL-1 weakly was suggested from LPS-stimulated MK2(−/−) spleenocytes where TNFα, IL-6, and IL-1β were inhibited 92, 72, and 40%, respectively (Kotlyarov et al., 1999).
PF-3644022 is orally efficacious at inhibiting TNFα production in LPS-challenged rats and blocking paw swelling in the chronic SCW-arthritis model. Whereas p38 kinase inhibitors maximally inhibit TNFα production 100% in the rLPS model and 90 to 95% paw swelling in the rSCW model (Burnette et al., 2009; Hope et al., 2009), MK2 inhibitors are somewhat less efficacious. PF-3644022 inhibits up to 90% TNFα levels in rLPS and only 80% paw swelling in the rSCW model. In the rSCW model, a p38 kinase inhibitor was shown to protect against inflammation-mediated joint and bone destruction (Burnette et al., 2009). Unfortunately, bone and cartilage histology were not measured in our rSCW studies with PF-3644022. It would have been interesting to see whether joint preservation was sufficient with MK2 inhibitor-mediated reductions in TNFα and IL-6 alone, especially because IL-1β has been implicated in bone and cartilage destruction in this model (Wilder et al., 1989).
In this article, we show excellent correlation that MK2 inhibition in cells is closely linked to TNFα inhibition in human cells and rats and reduction of paw swelling in a chronic model of arthritis (Fig. 7). To achieve maximal efficacy in the rSCW model, we show that a Cmin equivalent to at least the rat LPS–TNFα EC50 must be maintained throughout dosing of PF-3644022 (Fig. 6). At a half-maximal response level, PF-3644022 total exposure in blood would be 1 to 11 μM (Table 5). However, if an EC80 exposure is needed in humans for RA efficacy, then PF-3644022 total blood levels would be 5 to 50 μM. Projected human doses based on the pharmacology described in this article would be large, more than 300 mg twice a day (unpublished data). Given the biochemical inefficiency of MK2 as an anti-inflammatory target and the constant micromolar blood levels required for MK2 inhibition, sufficient kinase selectivity to establish an acceptable therapeutic index is challenging. That said, we continued developing PF-3644022 and evaluated its safety profile in rats, dogs, and monkeys. Although PF-3644022 is well tolerated in rats, acute hepatotoxicity was observed in dogs and monkeys at insufficient margins to continue developing PF-3644022. Similar toxicity was also observed with other molecules in the benzothiophene series, suggesting that liver toxicity is likely scaffold- related (J. S. Daniels, personal communication).
We thank Thomas L. Fevig, David L. Brown, and Daniel R. Dukesherer for assistance in preparative scale-up of PF-3644022 and Po-Chang Chiang for compound milling to support in vivo animal model testing.
This study was sponsored by Pfizer Inc.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- mitogen-activated protein kinase
- MAPK-activated protein
- MAPKAP kinase 2
- MAPK-interacting kinase
- rheumatoid arthritis
- tumor necrosis factor
- p38-regulated and activated kinase
- streptococcal cell wall
- rat SCW
- mitogen- and stress-activated protein kinase
- heat shock protein 27
- human peripheral blood mononuclear cell
- human whole blood
- liquid chromatography-tandem mass spectrometry
- biochemical efficiency
- dimethyl sulfoxide
- c-Jun N-terminal kinase
- extracellular signal-regulated kinase
- area under the curve
- free fraction.
- Received January 19, 2010.
- Accepted March 16, 2010.
- Copyright © 2010 by The American Society for Pharmacology and Experimental Therapeutics
- Anderson et al., 2005.↵
- Anderson et al., 2009a.↵
- Anderson et al., 2009b.↵
- Anderson et al., 2007.↵
- Burnette et al., 2009.↵
- Card et al., 2009.↵
- Cheng et al., 2010.↵
- Dayer et al., 2001.↵
- Dominguez et al., 2005.↵
- Gaestel, 2006.↵
- Gaestel et al., 2009.↵
- Gaestel et al., 2007.↵
- Goldberg et al., 2008.↵
- Hammaker and Firestein, 2010.↵
- Hegen et al., 2006.↵
- Hennigan and Kavanaugh, 2008.↵
- Hillig et al., 2007.↵
- Hope et al., 2009.↵
- Keminer et al., 2009.↵
- Kotlyarov et al., 1999.↵
- Kuiper et al., 1998.↵
- Lee et al., 1994.↵
- McInnes and Schett, 2007.↵
- Neininger et al., 2002.↵
- Pettus and Wurz, 2008.↵
- Ronkina et al., 2007.↵
- Schindler et al., 2002.↵
- Schindler et al., 2007.↵
- Schlapbach et al., 2008.↵
- Scott and Kingsley, 2006.↵
- Shi et al., 2003.↵
- Smith et al., 1998.↵
- Smolen and Steiner, 2003.↵
- Swinney, 2004.↵
- Trujillo et al., 2007.↵
- Wilder et al., 1989.↵
- Winzen et al., 1999.↵
- Wu et al., 2007.↵
- Xiong et al., 2008.↵