Abstract
Reversible janus associated kinase (JAK) inhibitors such as tofacitinib and decernotinib block cytokine signaling and are efficacious in treating autoimmune diseases. However, therapeutic doses are limited due to inhibition of other JAK/signal transducer and activator of transcription pathways associated with hematopoiesis, lipid biogenesis, infection, and immune responses. A selective JAK3 inhibitor may have a better therapeutic index; however, until recently, no compounds have been described that maintain JAK3 selectivity in cells, as well as against the kinome, with good physicochemical properties to test the JAK3 hypothesis in vivo. To quantify the biochemical basis for JAK isozyme selectivity, we determined that the apparent Km value for each JAK isozyme ranged from 31.8 to 2.9 μM for JAK1 and JAK3, respectively. To confirm compound activity in cells, we developed a novel enzyme complementation assay that read activity of single JAK isozymes in a cellular context. Reversible JAK3 inhibitors cannot achieve sufficient selectivity against other isozymes in the cellular context due to inherent differences in enzyme ATP Km values. Therefore, we developed irreversible JAK3 compounds that are potent and highly selective in vitro in cells and against the kinome. Compound 2, a potent inhibitor of JAK3 (0.15 nM) was 4300-fold selective for JAK3 over JAK1 in enzyme assays, 67-fold [interleukin (IL)-2 versus IL-6] or 140-fold [IL-2 versus erythropoietin or granulocyte-macrophage colony-stimulating factor (GMCSF)] selective in cellular reporter assays and >35-fold selective in human peripheral blood mononuclear cell assays (IL-7 versus IL-6 or GMCSF). In vivo, selective JAK3 inhibition was sufficient to block the development of inflammation in a rat model of rheumatoid arthritis, while sparing hematopoiesis.
Introduction
Janus associated kinase (JAK) inhibition has been a breakthrough for the treatment of autoimmune diseases such as rheumatoid arthritis (RA). The pan-JAK inhibitor tofacitinib was efficacious in patients suffering from RA for whom the standard of care therapeutics was ineffective (Burmester et al., 2013; Kremer et al., 2013). However, tofacitinib dosing is also limited by adverse events stemming from the inhibition of certain JAK isozymes, and therefore a selective inhibitor that maintains efficacy but avoids JAK-related adverse events could provide even greater benefit to patients.
The mammalian JAK family of intracellular tyrosine kinases has four members; JAK1, JAK2, JAK3, and TYK2. Cytokines bind to cell surface receptors resulting in receptor dimerization and phosphorylation of associated JAKs. Specific tyrosine residues on the receptor are phosphorylated by activated JAKs and recruit latent cytoplasmic transcription factors known as signal transducers and activators of transcription (STATs). STATs dimerize when phosphorylated by JAKs, translocate to the nucleus where they bind DNA elements, and activate gene transcription. The essential role of JAKs in mediating the biologic effects of cytokines has been confirmed by mutations in humans and targeted disruption in mice (O’Shea et al., 2015).
Selective targeting of JAK3 over other JAK isozymes for autoimmune diseases is attractive because JAK3 is selectively expressed in hematopoietic cells (Brown et al., 1999), and therefore not predicted to be involved in adverse events such as dyslipidemia or anemia. Tofacitinib was initially described as a JAK3 selective compound (Changelian et al., 2003); however, this has been subsequently revised (Flanagan et al., 2010; Thoma et al., 2014). Others have described JAK3 inhibitors, including WYE-151650 (Lin et al., 2010), VX-509 (Clark et al., 2014; Mahajan et al., 2015), and NIBR3049 (Haan et al., 2011; Thoma et al., 2011, 2014; Clark et al., 2014), which were determined to be selective for JAK3 in in vitro enzyme assays; however, this selectivity was not maintained in cellular assays. Several reasons why enzyme selectivity is not translated to the cellular context have been described by Thoma et al. (2011, 2014), including the relative contributions of JAK1 and JAK3 to signaling pathways in intact cells; the use of enzyme fragments including only the kinase domain and no other regulatory domains in in vitro assays; and the difference in ATP Km values for each JAK kinase (Clark et al., 2014; Thorarensen et al., 2014).
A question of therapeutic relevance is whether inhibition of JAK3 alone is sufficient to modify autoimmune phenotypes in vivo. Clinical proof of concept for the selective inhibition of JAK3 cytokines is absent; while the efficacy of anti-interleukin (IL)-6 antibodies such as tocilizumab argues that selective inhibition of JAK1 would be efficacious in RA (Maini et al., 2006), no antibodies against JAK3 cytokines have demonstrated clinical efficacy. To attempt to address the question of whether JAK3 inhibition alone is sufficient to block signaling via the family of receptors that use JAK3 enzymes, Haan et al. (2011) genetically manipulated U4C cells that do not normally express IL-2R or JAKs, to measure the effects of different JAKs on IL-2 signaling. These data suggested that JAK3 phosphorylated JAK1 but did not phosphorylate and activate STAT5, and therefore inhibition of JAK3 alone was not sufficient to block signaling through these receptors. In contrast, data obtained for a set of reversible JAK inhibitors with a wide range of JAK1 / JAK3 selectivity (Thorarensen et al., 2014) strongly suggest that inhibition of either JAK1 or JAK3 is sufficient to completely inhibit IL-15 signaling. Data for unambiguously selective irreversible inhibitors would further cement this conclusion.
Selective, irreversible JAK3 inhibitors have been described but until recently none contain all of the properties required for meaningful in vivo experimentation. These previously discussed compounds demonstrated JAK3 enzymatic selectivity but lacked data describing cellular selectivity in cells (London et al., 2014; Goedken et al., 2015) or appropriate physicochemical properties to be tested in vivo (London et al., 2014; Goedken et al., 2015; Tan et al., 2015). Smith et al. (2016) generated compounds described in a Merck & Co., Inc. (Kenilworth, NJ) patent (Ahearn et al., 2013) and confirmed these were potent and selective JAK3 inhibitors that also inhibited IL-2 signaling in vivo. The compound selected was only 50-fold selective over Bruton’s tyrosine kinase (BTK), a critical kinase for B cell signaling in autoimmune disease, and therefore this compound is not an optimal tool for evaluating JAK3 biology in pathologically relevant models of autoimmune disease. During preparation of this manuscript, Telliez et al. (2016) reported discovery of an irreversible JAK3 inhibitor, PF-06651600, that maintains JAK3 selectivity in cells and possess properties appropriate for in vivo study.
To fully understand the potential of JAK3 inhibition in autoimmune disease, an optimal tool compound is required. This compound needs to be a potent JAK3 inhibitor that is selective against JAK isozymes as well as other members of the kinome, maintains potency and selectivity in cells, and has good pharmacokinetics (PK) properties to enable meaningful in vivo experimentation. Using an evaluation of ATP Km values, we determined that a reversible inhibitor was unlikely to fulfill these requirements due to high-affinity binding of ATP for JAK3. Therefore, we developed a selective, irreversible JAK3 inhibitor (compound 2), demonstrated that JAK3 inhibition alone is sufficient to inhibit JAK3-mediated cytokine signaling, and tested the compound in an in vivo model of RA. We demonstrated that the potent, selective inhibitor 2 could fully inhibit the development of inflammation. Interestingly, the irreversible compound did not demonstrate a striking dissociation between PK and pharmacodynamics (PD) in an ex vivo assay measuring STAT phosphorylation. To understand this result, we determined the half-life and recovery of JAK3 in CD3+ and CD4+ T cells, which revealed a rapid recovery of JAK3 protein, especially under conditions mimicking inflammation.
Materials and Methods
Homogeneous Time-Resolved Fluorescence (HTRF) Assay for Determination of ATP and Peptide Km of JAK Isozymes
For each enzyme [JAK1, JAK2, JAK3, and TYK2 (all four N-terminal GST tagged)], purchased from Thermo Fisher (Waltham, MA), a stock 3X ATP plate was made by 2X serial dilutions of ATP (Sigma-Aldrich, St. Louis, MO) using 1X assay buffer (diluted 5X assay buffer, Thermo Fisher), yielding eight doses. A biotinylated peptide (biotin-EQEDEPEGDYFEWLE-CONH2) was custom made by GenScript (Piscataway, NJ) and subsequently diluted to a 13 μM stock concentration in 1 mM Tris-HCl, pH 8.0. A stock plate of 3X peptide was made by 16 successive 2X serial dilutions, also in assay buffer. Equal amounts of each substrate dilution were then stamped to a reaction plate, and reactions were initiated with the addition of 3X enzyme. At discrete time points, a sample of reaction was collected from the reaction plate into a detection plate containing detection buffer [0.8 nM europium (Eu)-labeled anti-phosphotyrosine antibody PY20 (PerkinElmer, Waltham, MA) and 32 nM streptavidin DyLight (Thermo Fisher Biotechnology) in 250 mM HEPES, 100 mM EDTA, and 1.0% Triton X-100]. After the final time point was taken, the detection plates were incubated for an additional 45 minutes and read on the PerkinElmer EnVision plate reader using excitation of 320 nm and dual emission wavelengths set at 665 and 615 nm.
HTRF Assay for IC50 Determination
Test compound was serially diluted in dimethylsulfoxide (DMSO) and a discrete dose response was performed using an ECHO 555 (Labcyte, Sunnyvale, CA). Next, a 5X enzyme/peptide solution (7.03 nM JAK1, 156.0 pM JAK2, 155 pM JAK3, or 13.31 nM TYK2, and 3.75 μM peptide) was prepared in 1X assay buffer (Thermo Fisher) and subsequently added to the reaction plate already containing titrated compound. The reagents were incubated 30 minutes before the reaction was initiated by addition of 10X ATP stock solution (310.8, 85.0, 29.0, or 69.0 μM for JAK1, JAK2, JAK3, and TYK2, respectively) or 1X buffer solution. The reactions proceeded for 120 minutes before they were quenched with detection buffer (as described above). After 45 minutes at room temperature, the reactions were read on the EnVision as described previously. Percent inhibition was calculated as follows: [(sample signal − background signal)/(uninhibited signal − background signal)] × 100. The dose-response curve data were fit to a four-parameter logistic equation using automated in-house assay analyzer software.
HTRF Assay for JAK3 Potency Determination (kinact/KI)
Test compound was 3X serially diluted in DMSO. For each of the time points tested a separate reaction plate was prepared, where 1000X inhibitor was added to 3.2 nM JAK3. Each reaction was then preincubated for a predetermined amount of time before adding an aliquot of the preincubation complex to wells of a corresponding reaction plate already containing a 1X ATP/peptide solution (290 μM ATP and 750 nM peptide) or assay buffer. The plates were incubated for 20 minutes at room temperature after initiation. Next, an aliquot of reaction was subsequently stamped into a detection plate containing detection solution (as described above). Reactions were quenched for 45 minutes at room temperature before reading on the EnVision plate reader as described previously. The Kobs value was defined as the slope of fractional JAK3 activity remaining for a given preincubation time graphed on a semi-log plot, and kinact and KI were then obtained by plotting Kobs as a nonlinear function of inhibitor concentration using XLFit software (IDBS, Alameda, CA).
HTRF Assay to Determine Inhibitor Reversibility
Three control arms and a single experimental arm were prepared to determine compound reversibility. To the experimental arm, 10X compound 2 (0.8 µM) was added to 1.11X JAK3 (8.4 nM) prepared in 1X kinase buffer. To the control arms, an equivalent amount of DMSO was added. All four arms were incubated for 90 minutes at room temperature. To initiate the reactions, four separate substrate solutions were prepared. To the association-control arm (defined as 50% activity), 100X compound 2 (800 nM) was added to 1.02X substrate solution (300 μM ATP and 750 nM peptide diluted in 1X kinase buffer). Similarly, to the positive-control arm (defined as 100% activity) an equal volume of DMSO was added to 1.02X substrate solution. To the negative-control arm (defined as 0% activity), 100X of compound 2 (100 mM) was added to 1.02X substrate solution. Finally, to the experimental arm an equal amount of DMSO was added to 1.02X substrate solution to keep the DMSO concentration consistent. Each reaction was initiated by simultaneously diluting each preincubation complex 100-fold into the corresponding substrate mixtures.
To determine the τ1/2 value, at discrete times, aliquots from each reaction were simultaneously sampled and quenched in detection buffer (prepared as described above). The signal was read on the EnVision plate reader as described previously. Inhibitor reversibility was measured as the time required for the experimental arm to regain enzymatic velocity (Vs) consistent with the association-control arm. The τ1/2 value was calculated using the following model in GraphPad Prism software: Y = Vs × x + [(Vi-Vs)/k] × [1 − exp(−k × x)] + C, where Vs = reaction velocity measured in the association-control arm at equilibrium; Vi = reaction velocity of the positive-control arm, and C = initial signal.
PathHunter Assays.
JAK1, JAK2, and JAK2/3 chimera PathHunter cell lines (DiscoveRx, Freemont, CA) were plated at 1200 cells per well (1.8 μl of a 2 million cells/ml suspension) in cell plating reagent in 1536-well plates (Corning white polystyrene, square well, nonbinding surface). Compounds were introduced to the wells (30 nl of a DMSO stock solution, 11-point serial dilution yielding final compound test concentrations from 0.5 to 50,000 nM) and incubated at room temperature for 30 minutes, followed by addition of prolactin (1.0 μl of a 43 ng/ml stock solution containing 0.1 mg/ml bovine serum albumin). After 4 hours at room temperature 1.5 μl of PathHunter detection reagent was added to each well, the plate was incubated for a final 2 hours, and read on a ViewLux (PerkinElmer); 30 second read, slow speed, and 2X gain with binning.
CellSensor Assays.
CellSensor assays were performed according to the manufacturer’s specifications (Thermo Fisher).
pSTAT Flow Cytometry Assay in Human Peripheral Blood Mononuclear Cells (PBMCs).
Frozen PBMCs (Seracare, Milford, MA) were thawed in RPMI1640 (Thermo Fisher) with 5% heat-inactivated fetal bovine serum (FBS) (Sigma-Aldrich) and recovered for 40 minutes at 37°C, 5% CO2. Compounds were diluted to give a 0.1% DMSO final concentration and added to assay plates using an ECHO 520 liquid handler (Labcyte). Next, 200,000 cells were added to each well and then incubated for 30 minutes at 37°C, 5% CO2. Cytokines were added to each well and again incubated for 30 minutes at 37°C, 5% CO2. The final concentrations of cytokines added were as follows: human IL-6, 10 ng/ml; human IL-7, 4 ng/ml; and human granulocyte-macrophage colony-stimulating factor (GMCSF), 0.3 ng/ml (all from Thermo Fisher). Cells were fixed with Cytofix reagent (BD Biosciences, San Jose, CA) for 12 minutes at 37°C, spun, and washed twice with Dulbecco’s phosphate-buffered saline (PBS), without calcium or magnesium (Thermo Fisher). To permeablize cells, cold Perm Buffer III (BD Biosciences) was added to the cells and incubated for 30 minutes on ice. Cells were spun again and washed twice with fluorescence-activated cell sorter (FACS) buffer [1X Dulbecco’s PBS, without calcium or magnesium (Thermo Fisher), 1% bovine serum albumin, sulfhydryl-block (Lee Biosolutions, St. Louis, MO), 0.2 mg/ml purified human IgG (Millipore, Billerica, MA), and 0.09% sodium azide (Sigma-Aldrich)]. Cells were resuspended in 50 μl FACS buffer containing antibodies and stained overnight at 4°C with gentle shaking. The antibodies used for staining were the following: anti-CD3-APC, clone UCHT1 (BD Biosciences); anti-CD4-V450, clone RPA-T4 (BD Biosciences); CD14-PC7, clone RM052 (Beckman Coulter, Jersey City, NJ); anti-pSTAT5 (pY694)-PE, clone 47 (BD Biosciences); and anti-pSTAT3 (pY705)-Alexa 488, clone 4/P-STAT3 (BD Biosciences). The next day, cells were washed twice with FACS buffer, and data were acquired using a LSRII or Fortessa flow cytometer (BD Biosciences).
pSTAT Flow Cytometry Assay in Human Whole Blood.
Compounds were diluted and added to assay plates as described previously, and then 100 μl fresh human blood (Merck & Co., Inc., internal blood donor program) was added to each well and incubated for 30 minutes at 37°C, 5% CO2. Cytokines were added and cells were stimulated for an additional 30 minutes at 37°C, 5% CO2. Cytokines were titrated and the pSTAT signal was determined in the major cell subsets. The concentrations that gave 80% maximal stimulation were selected to be the final stimulating concentrations. The final concentrations of cytokines were as follows: human IL-2 500 ng/ml; human IL-3 200 ng/ml; human IL-6 10 ng/ml; human IL-7 10 ng/ml; human IL-10 50 ng/ml; human IL-15 50 ng/ml; human GMCSF 1 ng/ml; human IFNα 200 ng/ml; and human IFNγ 2ng/ml (all from Thermo Fisher). Prewarmed BD Phosflow Lyse/Fix Buffer (BD Biosciences) was added for 12 minutes at 37°C. Cells were washed twice in Dulbecco’s PBS, and incubated for 30 minutes on ice in cold Perm Buffer III (BD Biosciences). Cells were washed twice more with FACS buffer (as previously described). Cells were resuspended in 50 μl FACS buffer with antibodies and stained overnight at 4°C with gentle shaking. The antibodies added were a combination of the following, depending on the stimulating cytokines: anti-CD3-APC, clone UCHT1 (BD Biosciences); anti-CD4-V450, clone RPA-T4 (BD Biosciences); CD14-PC7, clone RM052 (Beckman Coulter); anti-CD20-Percp-Cy5.5, clone 2H7 (BD Biosciences); anti-pSTAT1 (pY701)-Alexa 488, clone 4a (BD Biosciences); anti-pSTAT3 (pY705)-PE, clone 4/P-STAT3 (BD Biosciences); anti-pSTAT5 (pY694)-PE, clone 47 (BD Biosciences); and anti-pSTAT6 (PY641)-Alexa 488, clone 18 (BD Biosciences). The next day, cells were washed twice with FACs buffer and the samples were acquired using the LSRII or Fortessa flow cytometers (BD Biosciences).
Erythropoietin (EPO) Proliferation Assay.
Frozen, purified human CD34+ from mobilized blood (AllCells) were seeded at 100,000 cell/ml in Stempro34 complete medium (Thermo Fisher) supplemented with GMCSF at 20 ng/ml (Thermo Fisher), human stem cell factor at 40 ng/ml (Cell Signaling), and IL-3 at 20 ng/ml (Cell Signaling). Cyclosporin A (Sigma-Aldrich) was added at 1 μg/ml to suppress lymphoid and monocytic lineages. The cells were grown in T225 flasks and placed in a standard cell culture incubator for 5 days (37°C, 5% CO2) with no media change. On day 5, the cell density was determined (ViCellXR, Beckman Coulter) and the total cell harvest was calculated. The cells were pelleted with a 5-minute centrifugation at 216g, followed by resuspension in cryomedium consisting of 10% DMSO (Sigma-Aldrich) and 90% of the 5-day conditioned medium. The cells were frozen in a BioCision CoolCell overnight at −80°C and transferred to a −180°C vapor phase freezer the next day. For EPO stimulation, pre-erythroid cells were thawed and seeded at 100,000 cell/ml in Stempro34 complete medium supplemented with GMCSF at 20 ng/ml, human stem cell factor at 40 ng/ml, IL-3 at 20 ng/ml, EPO at 20 ng/ml (Cell Signaling), cyclosporin A at 1 μg/ml, and 1% antibiotic-antimycotic solution (Sigma-Aldrich). Non-EPO-treated control cells were seeded at the same density. A 3X serial dilution, 10-point dose curve was created using an ECHO acoustic liquid handler (Labcyte) in a 384-well CulturPlate (PerkinElmer). The EPO-treated cells were plated over the compound dose curves at 5000 cells/well. Non-EPO-stimulated cells (negative control) and EPO-stimulated cells (positive control) with DMSO background were included on the same plate. Plates were placed in 37°C, 5% CO2 cell culture incubators for 4 days. On day 4, the cell numbers were quantified by PrestoBlue (Thermo Fisher). After incubation, the plates were read on a fluorometric plate reader at 560 nm excitation and 590 nm emission. In-house data confirmed that the amount of converted PrestoBlue dye was proportional to the amount of viable cells upon EPO stimulation (data not shown). Compound potency was calculated by the amount of PrestoBlue dye conversion per compound dose response. A pan-JAK inhibitor control compound was included to normalize the test compounds against maximal and minimal compound inhibition of EPO-dependent proliferation.
IL-6-Induced MCP-1 Secretion Assay.
Frozen PBMCs (Seracare) were thawed in RPMI1640 (Thermo Fisher) with 10% heat-inactivated FBS (Sigma-Aldrich) and recovered overnight at 37°C, 5% CO2. The next day, cells were diluted to 1 × 106 cells/ml in RPMI1640, 5% FBS. Compounds were serially diluted in DMSO and added to assay plates using an ECHO 520 (Labcyte). Then, 10,000 cells were added per well of a 96-well plate in 100 μl and incubated for 30 minutes at 37°C, 5% CO2. IL-6 (Thermo Fisher) was added to the cells at a final concentration of 400 ng/ml and incubated for 24 hours at 37°C, 5% CO2. MCP-1 levels from sample supernatants were determined using the Human MCP-1 Tissue Culture Kit (Meso Scale Discovery) following the manufacturer’s instructions.
JAK1 and JAK3 Cross-Titration in PBMCs.
Frozen PBMCs (Seracare) were thawed in RPMI1640+10% FBS and recovered overnight at 37°C. JAK3 inhibitor 2 and JAK1 inhibitor 1 were cross-titrated from 10 μM to 4.6 nM and added to the PBMCs. Cells were stimulated with IL-7 (Thermo Fisher) at a final concentration of 2 ng/ml for 30 minutes before being fixed with Cytofix (BD Biosciences) for 30 minutes at 37°C. Cells were washed once in PBS, permeablized with Perm III buffer (BD Biosciences) at 4°C for 30 minutes, and then incubated with antibodies to CD3, CD4, and pSTAT5 (as previously described for antibodies). Cells were washed in FACS buffer (BD Biosciences) and read on an LSRII flow cytometer (BD Biosciences). The mean fluorescence intensity of pSTAT5 in CD3+ and CD4+ T cells was used to plot the dose-response curves for the cross-titrated antibodies. Potential synergy was calculated using the Lowe hybrid method, where the volume of the difference between the expected surface and calculated surface is estimated.
Rat Collagen-Induced Arthritis (CIA) Model.
Female Lewis rats, aged 6 to 7 weeks, were acquired from Harlan Laboratories (Indianapolis, IN) and housed two per cage. Animals had access to drinking water and standard rodent chow ad libitum and were acclimated for 5 days before experiments began. Rats were enrolled into studies between ages 7 and 8 weeks once they reached a body weight between 125 and 150 g. These experiments were conducted in accordance with federal animal care guidelines and all procedures were reviewed by the Institutional Animal Care and Use Committee of Merck & Co., Inc. On day 1 the hind paw thicknesses were measured using a spring-loaded caliper. Following initial measurements the animals were injected at the base of the tail with two 100 μl injections containing 0.3 mg bovine type II collagen emulsified in incomplete Freund’s adjuvant; six animals were similarly injected with 0.9% saline as a negative control (group 1). On day 8 the animals received identical booster injections. Additional hind paw measurements were performed on days 8, 11, 16, 18, 21, 23, 25, 28, and 30 post initial injections. Following day 16 measurements, CIA animals were stratified into groups (groups 2–7, n = 10; group 8, n = 6) according to their respective change in average [(left + right)/2] hind paw thickness. Beginning on day 17 each animal in groups 1–7 was administered either vehicle or compound at 2 ml/kg (vehicle 1: 0.5% methocel; vehicle 2: 10% Tween 80); animals in group 8 received dexamethasone at 5 ml/kg (vehicle: 10:90 polyethylene glycol 400:10% Tween 80). On day 29 blood samples from groups 3–8 were collected at 4 and 8 hours for PK; 0- and 2-hour PK samples were collected on day 30; 0-, 4-, and 8-hour PK samples were collected via jugular vein puncture under isoflurane; and 2-hour PK samples were collected via terminal cardiac stick under isoflurane. At 2 hours on day 30 all animals were euthanized, samples for clinical chemistry and hematology were collected, and hind limbs were removed and placed in 10% formalin for computed tomography (CT) imaging and histologic analysis. Statistical analyses used were one- and two-way analysis of variance (treatment versus time, where applicable) with Bonferroni post hoc tests; comparisons of treatment groups versus arthritic and nonarthritic vehicles are reported. For prophylactic dosing, the rat CIA model was run as previously described. Beginning on day 2 each animal was administered either vehicle or compound 2 at 5 ml/kg (10% Tween 80 was used as a vehicle).
Complete Blood Cell Count in Rats Dosed with Compound 2.
Female Lewis rats were purchased and housed as described previously. A total of 48 rats were divided into six groups (n = 8/group). The rats in group 1 were drug naive, i.e., no compounds were administered throughout the study. On the afternoon of day 1 (4 PM), aminobenzotriazole (ABT) or vehicle (1 ml/kg by mouth) was administered to groups 2–5. On days 2–11 (8 AM), each animal in groups 2–5 was administered ABT 10 mg/kg daily (1 ml/kg by mouth), immediately followed by either vehicle or compound at 5 ml/kg by mouth. Group 6 animals received vehicle only (5 ml/kg by mouth). On days 2–11 (4 PM), groups 2–5 were administered vehicle or compound at 5 ml/kg by mouth. Animals were monitored and weighed throughout the study. On day 10, under isoflurane anesthesia, three animals from groups 2–6 were bled via the jugular vein for PK analysis at 4 and 8 hours post 8 AM dose. On day 11, blood samples were collected, as described previously, at 0 (16-hour post day 10 PM dose) and 2 hours post AM dose for PK, hematology, and clinical chemistry analysis. All remaining animals were euthanized at 2-hour postdosing on day 11 and blood samples were collected for PK, hematology, and clinical chemistry analysis. Data were analyzed using the GraphPad prism software. Statistical analyses were performed using a one-way analysis of variance with Dunnett’s post hoc test for group comparisons to ABT + vehicle treatment.
Cycloheximide (CHX) Chase Assay to Determine JAK3 Half-Life.
Human PBMCs (Seracare) were thawed and recovered overnight at 37°C, 5% CO2. Human T cells were purified using the EasySepHuman CD4+ T Cell Enrichment Kit (STEMCELL, Canada). Cells were incubated with CHX (40 µg/ml), compound 2 (1 μM) or DMSO added at 0 hour, and cells were collected from 0 to 30 hours. The cells were lysed using radioimmunoprecipitation assay buffer with protease and phosphatase inhibitors for immunoprecipitation and western blot assays. For the washout studies, cells were collected after 18-hour incubation with CHX and washed twice with media before being replated in fresh media without CHX for recovery. The lysates were incubated with mouse anti-JAK3 antibody (BD Biosciences) and conjugated to sepharose A beads overnight at 4°C. Beads were washed in PBS, eluted in radioimmunoprecipitation assay lysate buffer, added to 6X lithium dodecyl sulfate sample buffer (NuPage) and run on an SDS-PAGE gel (Criterion Tris-HCl Gel, 4%–15%, NuPage). Proteins were transferred to nitrocellulose membranes and blocked with Odyssey blocking buffer (LI-COR, Lincoln, NE). Immunoprecipitated JAK3 was detected using rabbit anti-JAK3 antibody 1:500 (Santa Cruz Biotech) and IRDye 680CW, donkey anti-rabbit secondary antibody (LI-COR). Fluorescent signal was quantified using an Odyssey Fc dual-mode imaging system (LI-COR).
Results
JAK Enzyme Km ATP Differences Impact Cellular Selectivity of ATP-Competitive Compounds.
Medicinal chemistry compound design to enhance affinity to JAK3 and to minimize binding to JAK1 and JAK2 was guided by enzymatic activity assays run with relatively low [ATP], at the enzyme’s ATP Km value. Under these conditions, the measured IC50 values are proportional to the inhibitor’s binding free (intrinsic) energy in order to guide the structure-activity relationships. To determine the ATP Km value for each JAK enzyme, we titrated ATP for each isozyme and observed that despite the similarities between the ATP binding pockets in the four enzymes, the ATP Km values varied from 31.8 μM for JAK1 to 2.9 μM for JAK3 (Table 1). Since these ATP-pocket ligands bound competitively with ATP and the different JAK isozymes exhibited different intrinsic affinities for ATP, we predicted that JAK inhibitor potency for each JAK isozyme would be differentially reduced under cellular conditions where the ATP concentration was uniformly high (∼5 mM). The predicted potency shift for an ATP-competitive compound for JAK1 was 79-fold, whereas for JAK3 it was 863-fold. Therefore, the shift in potency for a compound in an in vitro biochemical assay to a cellular assay was predicted to be 10.9-fold higher for JAK3 than for JAK1, leading to an in vivo selectivity deficit for JAK3 ATP-competitive compounds that would be hard to overcome.
Experimental ATP Km value for each JAK isozyme, theoretical potency shifts to cellular [ATP] (5 mM), and relative cellular potency compared with a JAK1 inhibitor
The following equation was used to obtain the values: 
.
Validation of Cellular Assays to Measure Individual JAK Activity
To confirm these predictions, we tested ATP-competitive compounds in a series of cellular assays that were dependent only on one JAK isozyme. This was necessary because physiologic cellular assays that depend on cytokines binding to receptors, phosphorylating JAKs, and STATs signal exclusively through heterodimers and heterotrimers of JAKs (excluding JAK2, which can signal as a homodimer). PathHunter β-galactosidase (β-gal) enzyme fragment complementation assays were used to study full-length, single isozyme JAK1, JAK2, and chimeric JAK2/3 target inhibition in transfected U2OS cells. These cell lines contain the prolactin receptor fused to a small peptide fragment of β-gal (termed prolink), and a larger, complementary β-gal fragment fused to the SH2 domain (termed the SH2-enzyme acceptor). Stimulation of the prolactin receptor induces receptor phosphorylation and subsequent binding of the SH2-enzyme acceptor, resulting in complementation and forming functional β-gal, whose enzymatic activity is measured via chemiluminescence. The JAK2/3 chimeric protein contains residues 1–256 from JAK2, which includes the FERM domain, to ensure consistent interaction of all JAK isozymes with the prolactin receptor, while preserving all functional JAK3 domains corresponding to residues 233–1124, including the ATP binding pocket. In particular, the JAK 2/3 chimeric system enabled measurement of intracellular JAK3 activity without interference from JAK1 that complicates all JAK1/JAK3 heterodimeric systems.
To assess the performance of the JAK1, JAK2, and JAK2/3 chimeric PathHunter assays, and in particular to understand their selectivity of response to ATP-competitive inhibitors of JAK isozymes, we generated scatter plots comparing enzymatic and cellular potency. The analysis was restricted to the subset of compounds with IC50 values less than 40 μM in the PathHunter assays (quantitatively reliable range) and sufficiently potent enzymatic IC50 values to permit calculation of the predicted cell shifts, which are at least 10-, 138-, and 1190-fold for JAK1, JAK2, and JAK3 (<4 μM, 290 nM, and 34 nM, respectively). The included compounds were also members of structural clusters containing at least 20 members. Scatter plots were generated separately for each structural cluster since clusters tended to exhibit group behavior, each plot giving rise to a correlation coefficient between the enzyme and cell assay across all three family members in all nine possible combinations (Supplemental Figs. 1–3). The median values of the correlation coefficients across analyzed clusters are shown in the correlation table (Supplemental Fig. 4). While reversible inhibitors, and even the most selective analogs, typically retain significant activity across the JAK family, we found that in general enzymatic assay potency against any particular isozyme best correlated with the PathHunter cellular data for that same isozyme and correlated less well with cell assay data for different isozymes. These results are consistent with the notion that the three PathHunter assays uniquely report activity for each isozyme separately, without crosstalk. Correlation plots, in which each point represents a compound and structural clusters are shaded and shaped by cluster, illustrate that different clusters can behave differently but that the best general enzymatic versus cellular correlation exists when the isozyme is kept constant (e.g., JAK1 enzyme to JAK1 PathHunter).
In addition, we calculated the quantitative cell shift values (the ratio of the IC50 value in the PathHunter cellular assay to the IC50 value in the enzyme assay for the corresponding family member) for all compounds (Supplemental Fig. 5). The median enzymatic-to-cell-shift value was calculated for each structural cluster and the median of those values was used as a representative shift for each pair of associated JAK family member assays. Consistent with the measured differences in the ATP Km values for the different JAKs, the most commonly observed cluster median shift values were lowest for JAK1 and increased in magnitude for JAK2 and JAK3. Not surprisingly, these measured shift values were higher than the shifts predicted based on ATP competition alone, presumably due to other factors such as incomplete cellular permeability, compound loss due to nonspecific binding, or the contribution of other binding motifs of the enzyme (e.g., the pseudokinase domain) to the phosphorylation state, structure, or Km value.
An example of the reduction of selectivity in cells due to ATP shifts is shown by compound 1 (Fig. 1A), which is a compound with 41-fold selectivity for JAK3 over JAK1 in vitro that was near equipotent in single JAK isozyme cellular assays (Fig. 1B; Table 2). These data clearly demonstrated the challenge of developing JAK3 reversible compounds that would maintain true JAK3 selectivity in vivo.
Predicted cellular potency of compound 1 confirmed by potency in single isozyme cellular assay. (A) Compound 1 structure. (B) Compound 1 kinase potency, predicted cellular potency based on the ATP Km value at 5 mM ATP, and actual potency in single isozyme cellular PathHunter assays.
Reversible compound 1 and irreversible compound 2 potencies in enzymatic and cellular assays (nM)
Design of an Irreversible JAK3 Inhibitor.
One approach undertaken to overcome this ATP shift among the JAK isozymes was to capitalize on the amino acid sequence differences among of the JAKs; in particular, differences in the catalytic domains, and more specifically the presence of two cysteine residues in JAK3 (Cys-132 and Cys-909) that were not present in the other JAK isozymes. The goal was to target one of these JAK3 cysteine residues and generate a covalent protein-inhibitor bond with the JAK3 protein, thereby mitigating the ATP Km shift.
Accordingly, an existing JAK3 inhibitor scaffold was modified to produce compound 2, containing an electrophilic acrylamide warhead capable of reacting with the thiol of Cys-909 of JAK3 via a 1,4-conjugate addition (Fig. 2A). Compound synthesis and characterization are described in the Supplemental Material. The covalent modification was confirmed via the X-ray crystal structure, where the distance between the terminal carbon of the olefin and cysteine sulfur is indicative of a covalent bond (Fig. 2B). Compound 2 is selective over JAK1, JAK2, and TYK2 (Table 2) since there is a serine amino acid residue in these kinases, compared with the cysteine in JAK3. Note that the generated JAK3 IC50 value is an apparent IC50 value since true equilibrium cannot be established in the enzymatic assay due to JAK3 being covalently modified. JAK3 selectivity was maintained in the PathHunter cell-based assays with 80-fold selectivity over JAK1. Also, compound 2 was designed to be selective over the other kinases possessing a cysteine in the same region as JAK3, such as BMX, EGFR, ITK, and BTK. Of particular concern was the inhibition of BTK, which if inhibited could confound potential efficacy since BTK is a kinase that plays a crucial role in B-cell signaling (Whang and Chang, 2014). As seen by the waterfall plot in Fig. 2C, compound 2 is exceptionally selective and thus is a powerful tool to explore JAK biology.
Potent, selective irreversible JAK3 inhibitor 2. (A) Compound 2 structure. (B) Confirmation of covalent binding by X-ray crystallography. (C) Waterfall plot demonstrating selectivity of compound 2 against the kinome.
Confirmation of Irreversible Binding of Compound 2.
A comprehensive kinetic measurement of irreversible compound potencies to JAK3 was developed to confirm irreversible compound activity. As demonstrated in Fig. 3, A and B, the time-dependent change in fractional activity for compound 2 at various compound concentrations suggests that the compound undergoes a two-step time-dependent mechanism of inhibition and that compound 2 is extremely effective (kinact/KI = 0.060 ± 0.008 nM−1/min−1), with high affinity (KI = 3.332 ± 0.366 nM) and high specific reactivity (kinact = 0.197 ± 0.003 minute−1) for JAK3.
Fractional activity remaining and relative affinity (kinact/KI) determination. (A) Fractional activity remaining versus preincubation time for compound 2. (B) Re-plot of fractional activity remaining slope versus inhibitor concentration for compound 2; KI = 0.573 ± 00.095 nM, kinact = 0.199 ± 0.011 minute−1, kinact/KI = 0.352 ± 0.046 minute−1·nM−1 (n = 3).
However, since two-step time-dependent inhibition is consistent with either slow-onset reversible or irreversible inhibition, we attempted to further elucidate the mechanism of action with kinetic reversibility studies. Using a traditional jump-dilution technique we were able to demonstrate that no enzymatic activity was observed for 3 hours post reaction initiation (Supplemental Fig. 6). The lack of any product formation indicates that compound 2 did not dissociate from JAK3 during the reaction time. Furthermore, since less than 50% JAK3 activity was regained after this 180-minute reaction, the τ1/2 value for this compound cannot be calculated and is therefore assumed to be >180 minutes. Taken together, these results confirm that compound 2 is a potent and reactive irreversible covalent binder.
Compound 2 Blocks Cytokine Signaling through JAK3, but Not through Other JAK Family Enzymes.
Compounds 1 and 2 were tested in cellular reporter assays to measure inhibition of intact JAK/STAT pathway signaling. CellSensor assays contain a β-lactamase reporter gene under the control of a specific STAT response element stably integrated into relevant cell types. We used IL-2-stimulated CTLL-2 cells that drive β-lactamase gene expression under the control of the STAT5 response element present in the interferon regulatory factor 1 gene promoter to measure JAK1 and JAK3 signaling. GMCSF and EPO were used to stimulate human erythroleukemia TF-1 cells that contained a STAT5-dependent interferon regulatory factor 1 response element to measure JAK2 signaling. Finally, IL-6-stimulated human ME180 cells with an interferon-gamma activated sequence response element were used to measure JAK1 and JAK2 signaling. Both compounds 1 and 2 inhibited IL-2 signaling (30 and 70 nM, respectively). Compound 1 also inhibited IL-6-, GMCSF-, and EPO-stimulated signaling (130, 380, and 570 nM, respectively), whereas compound 2 had very low potency (4700 nM) in the IL-6 assay and was inactive in the GMCSF and EPO assays (Table 2).
Flow cytometry assays were developed that measured the inhibition of cytokine-dependent phosphorylation of intracellular STATs in specific cell subsets in human PBMCs and whole blood. Both JAK3 inhibitors inhibited IL-7-induced pSTAT5 in CD3+ and CD8+ T cells from human PBMCs; compound 1 with an IC50 value of 67 nM and compound 2 with an IC50 value of 280 nM. In contrast, compound 1 inhibited GMCSF-induced pSTAT5 in CD14+ monocytes from human PBMCs with an IC50 value of 480 nM, whereas compound 2 inhibited the same pathway with an IC50 value of 9600 nM. Compound 2 was also inactive against IL-6-stimulated pSTAT3 in CD3+ and CD4+ T cells (Table 2).
To broaden our understanding of compound 2 inhibition of multiple cytokine signaling pathways, we employed a multiparameter flow cytometry approach to look at inhibition of eight cytokines, signaling via four STATs in four different cell types. The IC50 concentration of compound 2 for IL-7-induced pSTAT5 in CD3+ and CD4+ T cells (5100 nM) was added to human whole blood stimulated with GMCSF, IL-7, IL-10, IL-15, IL-2, IL-3, IL-6, IFNα, or IFNγ. The pSTAT 1, 3, 5, and 6 levels were measured in CD3+, CD4+, CD3+, and CD8+ T-cells; CD3− and CD20+ B cells; and CD14+ monocytes. Only pathways where pSTAT levels increased 2-fold over baseline were considered for inhibition by the compound. As shown in Fig. 4, compound 2 inhibited IL-7, IL-15, and IL-2 signaling in T cells, but did not inhibit any other cytokine signaling pathways. Together these data demonstrate that reversible compound 1 is no longer selective in inhibiting physiologic cytokine signaling pathways; however, irreversible compound 2 remains a potent JAK3 inhibitor within the context of intact cytokine signaling in both engineered and primary human cell contexts.
Multiparameter flow cytometry in human whole blood demonstrates pSTAT signature of compound 2 specific to JAK3 signaling. Compound 2 was tested in a multiplexed assay format combining stimulation by nine cytokines in four different cell types and measuring four phospho-STATs. Only pathways that achieved a 2-fold or higher stimulation with the cytokine under control conditions were reported. The fractional inhibition of each pathway at the IC50 concentration of compound 2 in a single-plexed assay measuring IL-7-stimulated pSTAT5 in CD3+, CD4+ was presented (N = 2). Abbreviations: cell types: M: CD14+ monocytes, 4: CD3+, CD4+ T cells, 8: CD3+, CD4- T cells, B: CD20+ B cells. Cytokines: G: GMCSF, 3: IL-3.
JAK3 Inhibition Alone is Sufficient to Inhibit Common Gamma Chain Signaling Pathways.
Using a collection of reversible inhibitors with a full spectrum of JAK1 / JAK3 selectivity, Thorarensen et al. showed that either JAK1 or JAK3 inhibition alone was sufficient to interrupt IL-15 signaling. Furthermore, the selective JAK3 covalent inhibitor, PF-06651600 reported by Telliez and colleagues was shown to block signaling by IL-2, IL-4, IL-7, and IL-15. Using the enzymatic and cellular selective JAK3 inhibitor 2 described here, we demonstrate that JAK3 inhibition alone is sufficient to inhibit IL-2, IL-7, and IL-15 signaling in primary human T cells. Furthermore, we cross-titrated irreversible compound 2 with a JAK1 selective inhibitor 3 (JAK1: 0.96 nM, JAK2: 14 nM, JAK3: >1500 nM, TYK2: 10 nM) (Brubaker et al., 2013) to determine if there was an additive or synergistic effect of co-inhibiting JAK1 and JAK3 enzymes on IL-7 signaling in CD3+ and CD4+ PBMCs. As shown in Fig. 5, the predicted levels of pSTAT5 inhibition based on addition of JAK1 and JAK3 inhibition (rainbow surface) were very close to the measured effects of cross-titrating each compound (black dots) (Supplemental Fig. 7; Supplemental Material for Fig. 5), demonstrating that there was an additive effect but no synergistic effect of inhibiting JAK1 and JAK3 on blocking STAT5 phosphorylation. Furthermore, inhibition of either JAK1 or JAK3 alone was sufficient to fully inhibit pSTAT5.
No synergy observed in pSTAT5 PBMC assay in response to IL-7 with cross-titrated JAK1 and JAK3 inhibitors. pSTAT5 inhibition was observed with both compounds 2 and 3 in a cross-titration in CD3+ and CD4+ T cells in response to IL-7 activation by flow cytometry. The rainbow dotted line shows the inhibition of pSTAT5 by the individual compounds. The rainbow surface shows the expected inhibition achieved assuming no synergy between the compounds. The black dots show the actual inhibition achieved and the red lines show the difference between actual and expected.
JAK3 Inhibition Completely Prevents the Development of an Inflammatory Phenotype in Rats with CIA.
To determine if JAK3 inhibition alone was sufficient to block the development of an inflammatory phenotype in vivo we used the rat CIA model (Trentham et al., 1977, 1978; Brown et al., 1999). Compound 2 was dosed from the beginning of the experiment because it has been reported that the rat CIA model has a biphasic inflammatory response (Stolina et al., 2009): the initial phase of the model (days 1–7) is driven by infiltration and proliferation of immune cells such as T-cells, and this is highly dependent on γc cytokines such as IL-2 and IL-7 that signal through JAK3. Inflammation and joint erosion is coincident with an increase of inflammatory cytokines including IL-6 (JAK1 and JAK2), IL-1β, and CCL2 (which do not signal through JAKs) in the second phase of the model. IL-6 is elevated in the inflamed paw throughout the duration of the model.
Briefly, we injected rats with collagen to initiate the inflammatory response and then treated rats with increasing doses of compound 2 on the subsequent day. Compound or vehicle was dosed daily up to day 30 of the experiment. A second collagen injection was administered on day 8. Paw thickness measurements were taken on days 11 and 16 and then every 2 to 3 days for the duration of the experiment (Fig. 6A). A dose of 0.3 mg/kg dexamethasone was included as a positive control because corticosteroids have been shown to be efficacious in reducing inflammatory arthritis in both rats and humans (Choy et al., 2008; Singh et al., 2016). A PK experiment to determine concentration parameters of each dose in the plasma was also performed. At the end of the study the ankle and knee bone mineral densities were measured by micro-CT (μCT) imaging (Sevilla et al., 2015)
Prophylactic inhibition of JAK3 in a rat CIA model blocks paw swelling and joint degradation. Rats were dosed with compound 2 from day 2 of a CIA model. (A) Schematic of prophylactic dosing. (B) Paw thickness versus time for groups of rats dosed with the JAK3 inhibitor or vehicle controls (n = 10). (C) Calculation of total AUC for paw thickness for each individual rat. (D) Bone mineral density of ankle joints measured by μCT. (E) PK data for each dose of compound. Statistical significance: ***P < 0.001 versus nonarthritic, +P < 0.05 versus vehicle, +++P < 0.001 versus vehicle. Data are representative of two repeated experiments.
The 100 mg/kg dose of compound 2 resulted in partial inhibition (58% reduction compared with vehicle) of the swelling of the rat paw (Fig. 6, B and C). This dose also resulted in significant but incomplete inhibition of bone loss in the ankle measured by μCT (Fig. 6D). The 300 and 600 mg/kg doses resulted in complete inhibition (92% and 96%, respectively) of paw swelling and complete inhibition of bone loss in the ankle. Bone loss in the knee was very similar to the ankle (data not shown). The IC50 value for inhibiting IL-7-driven pSTAT5 in CD3+ and CD4+ T cells in rat whole blood was 1.3 μM. We had previously determined that the time-weighted average (TWA) inhibition of rat IL-7 signaling correlated with inhibition of paw thickness in this model (data not shown). At 100 mg/kg the TWA inhibition of IL-7 signaling was 51%. The 300 and 600 mg/kg doses of compound 2 achieved TWA inhibition of IL-7 signaling of 70% and 80%, respectively. These data clearly demonstrate dose-dependent inhibition of paw swelling and bone loss by a selective, irreversible JAK3 inhibitor in a prophylactic rat CIA model.
JAK3 Inhibition Partially Reversed the Inflammatory Phenotype in Rats with CIA.
As described previously, the primary phase of the rat CIA is driven predominantly by cellular proliferation, whereas the second phase of the model is driven by inflammatory cytokines such as IL-1β and IL-6 that do not signal through JAK3. We hypothesized that a selective JAK3 inhibitor would not be efficacious in reversing inflammation when dosed after the initial paw swelling had occurred. To test this hypothesis, we modified the rat CIA model described previously to start dosing with compound 2 on day 17, after the rats’ paws had already become edematous (Fig. 7A). Again, dexamethasone was included as a positive control. To reduce the metabolism and increase the half-life of compound 2, in this experiment compound 2 was codosed with 1-ABT, a known nonspecific inhibitor of cytochrome P450 (Fig. 7D). Compound 2 dosed at 30 mg/kg with ABT had a minimal effect on reducing paw swelling (11% reduction compared with vehicle) (Fig. 7B) and no effect on inhibiting the reduction in bone mineral density in the ankle (Fig. 7D). This dose had similar area under the curve (AUC) and Cmax values to the 100 mg/kg dose without ABT in the prophylactic study, and a TWA inhibition of IL-7 signaling of 24%. A 300 mg/kg dose of compound 2 dosed with ABT resulted in partial inhibition of paw swelling (31% reduction) (Fig. 7C) and small but significant inhibition of bone loss (Fig. 7D). These results were obtained despite achieving similar plasma AUC and Cmax values to the 600 mg/kg dose in the prophylactic study and TWA inhibition of IL-7 signaling of 82%. The similar dose in the prophylactic study resulted in complete efficacy. These data demonstrate that compound 2 has a partial effect on reversing inflammation and bone loss in the rat CIA model when administered therapeutically during the inflammatory phase of the model at doses that meet or exceed the exposure required for complete inhibition of the phenotype when dosed prophylactically.
Therapeutic inhibition of JAK3 in a rat CIA model partially blocks paw swelling and joint degradation. Rats were dosed with compound 2 from day 17 of a CIA model. (A) Schematic of prophylactic dosing. (B) Paw thickness versus time for groups of rats dosed with the JAK3 inhibitor or vehicle controls. (C) Calculation of total AUC for paw thickness for each individual rat. (D) Bone mineral density of ankle joints measured by microcomputed tomography. (E) PK data for each dose of compound. Statistical significance: ***P < 0.001 versus nonarthritic, ^P < 0.05 versus vehicle, ^^^P < 0.001 versus vehicle. *Calculations based on projected values for twice daily dosing (except dexamethasone); broken bar: below the level of quantification. Data are representative of two repeated experiments.
Compound 2 Impairs Leukocyte Proliferation but Not Erythrocyte Proliferation In Vivo.
To confirm that selectivity of compound 2 was maintained in vivo, we examined hematopoietic cell numbers in blood obtained from rats dosed for 10 days with compound 2. Lymphocytes, specifically T cells, rely on IL-2 (JAK1/JAK3) signaling to promote survival and proliferation, whereas erythrocytes require EPO (JAK2) signaling for maturation. Compound 2 was dosed up to 600 mg/kg twice daily with ABT to maintain high concentrations in the plasma throughout the experiment. At every dose in this study (100, 300, and 600 mg/kg twice daily), the AUC value that was achieved (Fig. 8E) exceeded the AUC required for complete inhibition of paw swelling in the CIA model (300 mg/kg; AUC = 92 μM/h) (Fig. 6, C and E). Compound 2 significantly reduced the total number of white blood cells as a result of the number of lymphocytes (Fig. 8, A and B). There was no significant reduction in the number of monocytes or neutrophils. These data are consistent with compound 2’s inhibition of cytokines such as IL-2, which are essential for T cell survival and proliferation. In contrast, both early (reticulocyte in Fig. 8C) and late (red blood cell, hemoglobin, and hematocrit in Fig. 8, C and D) markers of hematopoiesis were unaffected even at the highest doses of compound 2. Compound 4 is a potent inhibitor of JAK1 JAK2, and TYK2 (JAK1: 0.30 nM; JAK2: 1.5 nM; JAK3: 190 nM; and Tyk2: 0.24 nM) (Lim et al., 2011), and therefore was included as a positive control. These data demonstrate that compound 2 maintains selectivity over JAK2 signaling in vivo.
Inhibition of leukocyte proliferation, but not erythrocyte proliferation, by compound 2 in rats. Number of hematopoietic cells measured in the blood of rats dosed for 10 days with 100, 300, or 600 mg/kg of compound 2. (A) White blood cells and monocytes. (B) Neutrophils and lymphocytes. (C) Reticulocytes and red blood cells. (D) Hemoglobin and hematocrit. (E) PK measurements of compound 2.
Extended Inhibition of IL-7-Induced pSTAT5 In Vivo.
Irreversible compounds can show dissociated PD where inhibition of signaling endures after the compound has been cleared from the plasma (Singh et al., 2011). Compound 2 has reasonable PK for a proof-of-concept molecule but is rapidly cleared from the plasma with a half-life of 0.3 hours. To determine if compound 2 showed dissociated PK/PD, we dosed Lewis rats with 300 mg/kg of compound 2 (n = 4/time point). Animals were sacrificed 6, 15, 24, or 48 hours after dosing and blood was drawn to measure PK and for ex vivo flow cytometry assays to measure pSTAT5 responses to IL-7 and GMCSF, and pSTAT3 responses to IL-6. As expected, compound 2 inhibited IL-7 signaling (Fig. 9A) but not IL-6 signaling (Fig. 9C) or GMCSF signaling (Fig. 9D). At the 6- and 15-hour time points, the mean concentrations of compound 2 were 1.51 and 1.46 μM, respectively, which exceeded the IC50 value of compound 2 for inhibiting IL-7 signaling in rat whole blood (1.3 μM) (Table 2). IL-7 signaling was inhibited by 99% at 6 hours and 104% at 15 hours. By 24 hours the mean compound concentration had dropped to 64 nM and at 48 hours it was just 11 nM. At 24 hours, IL-7 was inhibited by 34% and at 48 hours it was still inhibited by 17%. In contrast, in the corresponding in vitro pSTAT5 CD3+ and CD4+ rat whole blood flow cytometry assay, 64 and 11 nM of compound 2 was not sufficient to inhibit pSTAT5 at all. To determine the IC50 value for inhibition of IL-7 signaling in the IL-7 ex vivo assay, a curve was plotted using data from all time points collected (6, 15, 24, and 48 hours). If JAK3 activity remained inhibited after compound 2 was cleared from the plasma, a left shift in the IC50 value would be expected. The IL-7 ex vivo IC50 value was calculated to be 0.17 μM (Fig. 9B), 7.6-fold lower that the corresponding IC50 value in an in vitro IL-7 assay. These data demonstrate that a modest extended inhibition of IL-7-induced pSTAT5 was observed, even after compound 2 had been cleared from the plasma, resulting in a reduction in the calculated IC50 value.
Ex vivo pSTAT inhibition by compound 2 in rat whole blood. pSTAT signal in whole blood obtained from rats dosed with 300 mg/kg of compound 2 and stimulated with cytokines ex vivo. (A) pSTAT5 in CD3+ and CD4+ T cells from rat whole blood stimulated with IL-7. (B) The IC50 values for inhibition of IL-7 signal calculated from rats dosed for 6–48 hours. (C) pSTAT3 in CD3+ and CD4+ T cells from rat whole blood stimulated with IL-6. (D) pSTAT5 in CD14+ monocytes from rat whole blood stimulated with GMCSF. FI: fluorescence intensity, half-filled symbols represent unstimulated controls (lower-half filled) or stimulated controls taken from animals treated with vehicle (upper-half filled). Data are representative of two repeated experiments.
The Half-Life of JAK3 Protein in Human T Cells.
Chemical, enzymatic, and in vitro data demonstrated that compound 2 irreversibly bound JAK3; however, PK/PD analysis suggested only a modest extension of pSTAT5 inhibition after the compound had been cleared from the plasma. We hypothesized that if JAK3 protein was turned over at a rate faster than compound 2 was cleared from the plasma, then any extended PD effects would be lost. Furthermore, we hypothesized that cell context, activity, and metabolism may impact the turnover of JAK3. Therefore, we used the translational inhibitor CHX to inhibit protein production in human CD3+ T cells, and measured the level of JAK3 protein remaining by western or immunoprecipitation western. After addition of CHX, JAK3 protein levels decreased (Fig. 10A). In the presence of 10% FBS, the τ1/2 value of JAK3 was 14.8 hours. If the cells were serum starved, the τ1/2 value was reduced to 9.0 hours (Fig. 10B), suggesting that JAK3 turnover was increased in the cells under stressed conditions (Fig. 10B). Since production of newly synthesized JAK3, rather than clearance of irreversibly bound JAK3, would essentially be responsible for the lack of an extended PD effect of compound 2, we modified the experiment to determine the rate of recovery of JAK3 protein levels after washout of CHX. When cyclohedamide was washed out, the JAK3 levels slowly increased over 48 hours. However, if the T cells were stimulated with anti-CD3 antibody to mimic T cell activation, the JAK3 levels rapidly rebounded over 12 hours (Fig. 10C). Finally, to determine if the presence of compound 2 impacted turnover of JAK3, the experiment was repeated in the presence of 1 μM compound 2. The τ 1/2 value of JAK3 was not impacted by the presence of compound 2 (τ 1/2 = 14.9 hours), and similar trends were observed for JAK3 recovery in the presence and absence of anti-CD3 antibodies (Fig. 10, C and D). These data suggest that the turnover of JAK3 in human T cells is rapid enough to erase some of the extended PD effects of irreversible compound 2, especially in the context of metabolic challenges and T cell stimulation that are present during disease.
JAK3 turnover in human CD4+ T cells. (A) Western blot demonstrating JAK3 levels in human CD4+ T cells after inhibition with CHX. (B) Quantification of JAK3 levels in (A). (C and D) Quantification of JAK3 in CD4+ T cells after CHX treatment of 18 hours, and subsequent washout with or without anti-CD3 antibodies for 12 hours. Experiment was performed in the presence (D) or absence (C) of 1 μM of compound 2. Data are representative of three repeated experiments (each n = 3).
Discussion
The quest for a selective JAK3 compound to investigate JAK3 biology and as a potential therapeutic has been confounded by the lack of correlation between selectivity in in vitro enzyme assays and cellular cytokine signaling assays. Several factors may contribute to the lack of translation: selectivity for different JAK isozymes that are estimated from individual enzyme assays (but estimated from full-length homo- or heterodimers and homo- or heterotrimers in cellular cytokine signaling assays); differential cellular permeability between compounds; and potentially misleading data generated by screening for enzyme activity on truncated kinases that contain only the JH1 kinase domain (but not other regulatory domains). Thorarensen et al. (2014) presented data determining that different JAK isozymes have different Km values for ATP, and therefore predicted that the potency shift for compounds to cellular assays would vary. These data and results for irreversible JAK3 inhibitor PF-06651600 were supported by inhibition in IL-15 cytokine signaling assays. In this paper, we reported that the different Km values for ATP for each JAK isozyme predict the cellular potency of compounds, and we subsequently measured the potency in cellular assays that have been developed to quantify the activity of only one JAK isozyme at a time. Our data conclusively show that the reduction in selectivity of JAK3 inhibitors in cellular assays can be primarily attributed to the differences in the ATP Km values for different JAK isozymes.
To determine whether JAK3 inhibition alone could block signaling of cytokines through the JAK1/JAK3 heterodimer, we developed an irreversible inhibitor of JAK3. Compound 2 is a highly potent and selective JAK3 inhibitor in kinase assays, as well as in PathHunter assays engineered to signal via a single JAK, CellSensor cytokine reporter assays, and primary cell assays. Compound 1 was 41-fold selective for JAK3 versus JAK1 in enzyme assays, but only 4.5-fold selective for inhibition of IL-2 (JAK1/JAK3) signaling over IL-6 (JAK1/JAK2) signaling or 13-fold selective over GMCSF signaling (JAK2) in CellSensor assays. Compound 1 was 7.1-fold selective for IL-7 (JAK1/JAK3) over GMCSF (JAK2) in PBMCs. On the other hand, compound 2 was 4300-fold selective in enzyme assays, 67-fold or 140-fold selective in CellSensor assays, and >35-fold selective in PBMCs. Some decrease of selectivity observed for compound 2 in cellular assays could be attributed to the two-step process of JAK3 covalent binding, or to differences in the JAK enzyme construct (JH1 domain versus full length). Importantly, in vivo PD assays demonstrated that the concentration of compound 2 achieved in the plasma that resulted in complete inhibition of IL-7 signaling was not sufficient to inhibit IL-6 or GMCSF signaling in rat blood (Fig. 9, A, C, and D). Furthermore, selective JAK3 inhibition by compound 2 was observed when we compared signaling through nine cytokines in four cell types in human whole blood.
In comparison, in these assays we found that tofacitinib was actually slightly more potent in our JAK1 enzyme assays versus JAK3 enzyme assays (JAK1: 1.1 nM; JAK2: 48 nM; JAK3: 6.8 Nm; and TYK2: 20 nM), only 2.2-fold or 4.1-fold selective in cytokine reporter assays (IL-2: 52 nM; IL-6: 110 nM; and GMCSF: 220 nM), and 2.8-fold or 5.8-fold selective in PBMC assays (IL-7: 41 nM; IL-6: 110 nM; and GMCSF: 240 nM). The tofacitinib data are in line with values reported by other groups and was included as a benchmark for individual assays (Flanagan et al., 2010; Thoma et al., 2014).
This study addressed whether inhibition of JAK3 alone is sufficient to inhibit JAK1/JAK3 cytokine signaling. Previous pharmacological studies have been confounded by the lack of sufficient selectivity when compounds move from enzymatic to cellular assays. Engineered cell assays using kinase-dead forms of JAK1 and JAK3 have placed JAK1 as the dominant kinase in the signaling pathway; however, these studies are confounded by the manipulation of the intact cells (Haan et al., 2011; Thoma et al., 2014). Using compound 2, we demonstrate conclusively that JAK3 inhibition alone is sufficient to completely inhibit JAK1/JAK3 cytokine signaling (IL-7). Furthermore, we performed cross-titration with a JAK1 selective inhibitor to demonstrate that JAK1 and JAK3 inhibition is additive in reducing pSTAT5, with no evidence for synergistic activity. Compound 2 is highly selective over all other JAK isozymes, whereas compound 3 is selective over JAK3 (>1500-fold) but only partially selective over JAK2 (15-fold) and TYK2 (10-fold). Since JAK2 and TYK2 do not play a role in IL-7 signaling, compound 3 is a useful tool to dissect JAK1 versus JAK3 signaling in this system. These data place JAK1 and JAK3 equally and nonredundantly in γc receptor signaling.
Selective JAK3 inhibitors have been developed with the aim of maintaining the efficacy demonstrated by compounds such as tofacitininb, while avoiding adverse effects due to inhibition of other JAK isozymes. It was critical to determine whether JAK3 selective inhibitors have efficacy in vivo. While most animal models are insufficiently predictive of human disease, they can be useful to address mechanism-based hypothesis in an in vivo setting. We tested compound 2 in the rat CIA model in two separate dosing paradigms to dissect the therapeutic potential of selective JAK3 inhibition. First, we tested compound 2 in a prophylactic dosing paradigm, where we could inhibit the initial phase of the development of inflammation driven by JAK3 cytokines including IL-2 and IL-7 (Stolina et al., 2009). As expected, compound 2 was completely efficacious in this model. However, the second phase of inflammation in the rat CIA model is driven by JAK3-independent cytokines such as IL-6 and IL-1β. When we dosed compound 2 after inflammation and paw swelling had developed, only partial efficacy was observed, as expected if the inflammation was not entirely JAK3 dependent. Similarly, PF-06651600 was recently shown to be efficacious when dosed prophylactically in two different models of autoimmune disease. In contrast to our experiments evaluating therapeutic dosing after maximal onset of inflammatory symptoms, Telliez and colleagues reported efficacy in an experimental autoimmune encephalomyelitis model where dosing was initiated before symptoms reached maximum levels. In our study, ex vivo PD studies confirmed that the reduced efficacy in the therapeutic dosing paradigm was not due to reduced cytokine inhibition but rather suggested that the pathophysiology of the disease model was underlying these results. Mahajan et al. (2015) reported that JAK3 inhibition was efficacious in the rat CIA model using VX-509 in an intermediate dosing paradigm starting on day 10. VX-509 is a reversible JAK3 inhibitor. The reported data for VX-509 appear to support its selectivity for JAK3; however, due to the absence of any comparative data to a common benchmark in the cellular assays we are unable to confirm that the compound is more selective than other published compounds. Differences in experimental conditions including presence of serum in media, level of cytokine stimulation, and the selection of pSTAT response, can have a significant impact on the apparent compound potency and selectivity (Mahajan et al., 2015). Although VX-509 is clearly efficacious in the rat CIA model, these data do not compellingly demonstrate this is due to inhibition of only JAK3 and not to the combination with other JAK isozymes.
One advantage of irreversible compounds is that they can show dissociated PK/PD, i.e., enzyme inhibition continues after the free compound has been cleared from the animal (Singh et al., 2011). However, this is dependent on the turnover of the protein target. We performed a PK/PD study to identify whether compound 2 had an extended PD effect in vivo. The data revealed a modest effect, with a 7.6-fold decrease in the IC50 value of the compound. To determine if this limited effect was due to rapid JAK3 protein turnover, we determined the half-life of JAK3 in human CD3+ and CD4+ T cells as the most relevant cell type. The half-life of JAK3, under quiescent conditions, was calculated at 14.6 hours, considerably longer than the τ1/2 value of 2 in vivo (0.3 hours) but similar to the time over the IL-7 IC50 value achieved by 300 mg/kg dosing (15 hours). Therefore, an extended PD effect may be anticipated based on these data. However, under conditions that stress the cells, such as serum starvation or stimulation with anti-CD3 antibodies to model T cell activation, the τ1/2 value for JAK3 was reduced to 9 hours, and the rate of recovery of JAK3 protein also increased, suggesting a faster turnover of the enzyme. JAK3 protein recovery, rather than clearance, directly influences the recovery of JAK3 enzyme activity. Extended PD effects with irreversible JAK3 inhibitors may be observed only in normal conditions with reduced effect under inflammatory conditions.
In this study, we have described the discovery and characterization of compound 2, an irreversible JAK3 inhibitor that contains all the properties required for a compound to pharmacologically investigate JAK3 biology in vivo: JAK3 potency; selectivity compared with other JAK isozymes that is maintained in the cellular environment; selectivity across the rest of the kinome, specifically against kinases that are also involved in autoimmune disease pathogenesis; and physicochemical properties sufficient for in vivo experimentation. We demonstrated that compound 2 can fully inhibit JAK3 cytokine signaling in primary cells, and that this translated to partial efficacy in an animal model of RA in vivo. While irreversible JAK3 inhibitors can fully inhibit JAK3 signaling and could avoid adverse events such as anemia that are mediated by JAK2 inhibition and limit the efficacy of less selective JAK inhibitors, other factors such as the short half-life of JAK3 enzyme in activated human T cells may negatively impact the therapeutic efficacy of these compounds in patients. These data support the clinical hypothesis that a selective JAK3 inhibitor would have limited efficacy in RA, and potentially other autoinflammatory diseases, while sparing the hematopoietic adverse events. Patients would be expected to recover from anemia in chronic disease because inflammatory cytokines that inhibit hemapopoiesis would be reduced by the general anti-inflammatory effects of a JAK3 inhibitor. In addition, these patients would not develop additional anemia caused by inhibition of the EPO pathway, which is a concern for pan-JAK inhibitors. A JAK3 inhibitor may be efficacious in blocking the development of flares or periods of increased inflammation; however, based on our observations patients with chronic RA may not achieve full remission by inhibition of JAK3 alone.
Authorship Contributions
Participated in research design: Elwood, Witter, Piesvaux, Kraybill, Alpert, Bays, Goldenblatt, Qu, Ivanovska, Lee, Chiu, Tang, Deshmukh, Zielstorff, Byford, Chakravarthy, Dorosh, Rivkin, Klappenbach, Pan, Kariv, Scott, Dinsmore, Slipetz, Dandliker.
Conducted experiments: Elwood, Witter, Piesvaux, Kraybill, Alpert, Bays, Tang, Scott, Goldenblatt, Qu, Ivanovska, Lee, Chiu, Deshmukh, Zielstorff, Byford, Chakravarthy, Dorosh, Rivkin.
Performed data analysis: Elwood, Piesvaux, Kraybill, Alpert, Bays, Tang, Scott, Goldenblatt, Qu, Ivanovska, Lee, Chiu, Deshmukh, Zielstorff, Byford, Chakravarthy, Dorosh, Klappenbach, Pan, Kariv, Dinsmore, Dandliker.
Wrote or contributed to the writing of the manuscript: Elwood, Witter, Piesvaux, Alpert, Kariv, Dinsmore, Slipetz, Dandliker.
Footnotes
- Received December 22, 2016.
- Accepted February 3, 2017.
↵
This article has supplemental material available at jpet.aspetjournals.org.
Abbreviations
- ABT
- aminobenzotriazole
- AUC
- area under the curve
- BTK
- Bruton’s tyrosine kinase
- CHX
- cycloheximide
- CIA
- collagen-induced arthritis
- CT
- computed tomography
- DMSO
- dimethylsulfoxide
- EPO
- erythropoietin
- Eu
- europium
- FACS
- fluorescence-activated cell sorter
- FBS
- fetal bovine serum
- β-gal
- β-galactosidase
- GMCSF
- granulocyte-macrophage colony-stimulating factor
- HTRF
- homogeneous time-resolved fluorescence
- IL
- interleukin
- JAK
- janus associated kinase
- PBMC
- peripheral blood mononuclear cell
- PBS
- phosphate-buffered saline
- PD
- pharmacodynamics
- PK
- pharmacokinetics
- RA
- rheumatoid arthritis
- STAT
- signal transducer and activator of transcription
- TWA
- time-weighted average
- Copyright © 2017 by The American Society for Pharmacology and Experimental Therapeutics
